Mirrors/libwebsockets
1
0
Fork 0
mirror of https://github.com/warmcat/libwebsockets.git synced 2025-03-09 00:00:04 +01:00
Code Issues Releases Wiki Activity
libwebsockets/lib/roles/h2/http2.c

2906 lines
76 KiB
C
Raw Permalink Normal View History

http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
/*
* libwebsockets - small server side websockets and web server implementation
*
license: switch LGPLv2.1+SLE parts to MIT
2019-08-14 10:44:14 +01:00
* Copyright (C) 2010 - 2019 Andy Green <andy@warmcat.com>
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
*
license: switch LGPLv2.1+SLE parts to MIT
2019-08-14 10:44:14 +01:00
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
*
license: switch LGPLv2.1+SLE parts to MIT
2019-08-14 10:44:14 +01:00
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
*
license: switch LGPLv2.1+SLE parts to MIT
2019-08-14 10:44:14 +01:00
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
*/
private.h: rename to contain dir Having unique private header names is a requirement of a particular platform build system it's desirable to work with
2019-08-15 10:49:52 +01:00
#include "private-lib-core.h"
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/*
* bitmap of control messages that are valid to receive for each http2 state
*/
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
static const uint16_t http2_rx_validity[] = {
/* LWS_H2S_IDLE */
(1 << LWS_H2_FRAME_TYPE_SETTINGS) |
(1 << LWS_H2_FRAME_TYPE_PRIORITY) |
// (1 << LWS_H2_FRAME_TYPE_WINDOW_UPDATE)| /* ignore */
(1 << LWS_H2_FRAME_TYPE_HEADERS) |
(1 << LWS_H2_FRAME_TYPE_CONTINUATION),
/* LWS_H2S_RESERVED_LOCAL */
(1 << LWS_H2_FRAME_TYPE_SETTINGS) |
(1 << LWS_H2_FRAME_TYPE_RST_STREAM) |
(1 << LWS_H2_FRAME_TYPE_PRIORITY) |
(1 << LWS_H2_FRAME_TYPE_WINDOW_UPDATE),
/* LWS_H2S_RESERVED_REMOTE */
(1 << LWS_H2_FRAME_TYPE_SETTINGS) |
(1 << LWS_H2_FRAME_TYPE_HEADERS) |
(1 << LWS_H2_FRAME_TYPE_CONTINUATION) |
(1 << LWS_H2_FRAME_TYPE_RST_STREAM) |
(1 << LWS_H2_FRAME_TYPE_PRIORITY),
/* LWS_H2S_OPEN */
(1 << LWS_H2_FRAME_TYPE_DATA) |
(1 << LWS_H2_FRAME_TYPE_HEADERS) |
(1 << LWS_H2_FRAME_TYPE_PRIORITY) |
(1 << LWS_H2_FRAME_TYPE_RST_STREAM) |
(1 << LWS_H2_FRAME_TYPE_SETTINGS) |
(1 << LWS_H2_FRAME_TYPE_PUSH_PROMISE) |
(1 << LWS_H2_FRAME_TYPE_PING) |
(1 << LWS_H2_FRAME_TYPE_GOAWAY) |
(1 << LWS_H2_FRAME_TYPE_WINDOW_UPDATE) |
(1 << LWS_H2_FRAME_TYPE_CONTINUATION),
/* LWS_H2S_HALF_CLOSED_REMOTE */
(1 << LWS_H2_FRAME_TYPE_SETTINGS) |
(1 << LWS_H2_FRAME_TYPE_WINDOW_UPDATE) |
(1 << LWS_H2_FRAME_TYPE_PRIORITY) |
(1 << LWS_H2_FRAME_TYPE_RST_STREAM),
/* LWS_H2S_HALF_CLOSED_LOCAL */
(1 << LWS_H2_FRAME_TYPE_DATA) |
(1 << LWS_H2_FRAME_TYPE_HEADERS) |
(1 << LWS_H2_FRAME_TYPE_PRIORITY) |
(1 << LWS_H2_FRAME_TYPE_RST_STREAM) |
(1 << LWS_H2_FRAME_TYPE_SETTINGS) |
(1 << LWS_H2_FRAME_TYPE_PUSH_PROMISE) |
(1 << LWS_H2_FRAME_TYPE_PING) |
(1 << LWS_H2_FRAME_TYPE_GOAWAY) |
(1 << LWS_H2_FRAME_TYPE_WINDOW_UPDATE) |
(1 << LWS_H2_FRAME_TYPE_CONTINUATION),
/* LWS_H2S_CLOSED */
(1 << LWS_H2_FRAME_TYPE_SETTINGS) |
(1 << LWS_H2_FRAME_TYPE_PRIORITY) |
(1 << LWS_H2_FRAME_TYPE_WINDOW_UPDATE) |
(1 << LWS_H2_FRAME_TYPE_RST_STREAM),
};
static const char *preface = "PRI * HTTP/2.0\x0d\x0a\x0d\x0aSM\x0d\x0a\x0d\x0a";
static const char * const h2_state_names[] = {
"LWS_H2S_IDLE",
"LWS_H2S_RESERVED_LOCAL",
"LWS_H2S_RESERVED_REMOTE",
"LWS_H2S_OPEN",
"LWS_H2S_HALF_CLOSED_REMOTE",
"LWS_H2S_HALF_CLOSED_LOCAL",
"LWS_H2S_CLOSED",
};
#if 0
static const char * const h2_setting_names[] = {
"",
"H2SET_HEADER_TABLE_SIZE",
"H2SET_ENABLE_PUSH",
"H2SET_MAX_CONCURRENT_STREAMS",
"H2SET_INITIAL_WINDOW_SIZE",
"H2SET_MAX_FRAME_SIZE",
"H2SET_MAX_HEADER_LIST_SIZE",
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
"reserved",
"H2SET_ENABLE_CONNECT_PROTOCOL"
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
};
void
lws_h2_dump_settings(struct http2_settings *set)
{
int n;
http2 add hpack decode support Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-09 16:57:47 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
for (n = 1; n < H2SET_COUNT; n++)
lwsl_notice(" %30s: %10d\n", h2_setting_names[n], set->s[n]);
}
#else
void
lws_h2_dump_settings(struct http2_settings *set)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
{
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
#endif
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
struct lws_h2_protocol_send *
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lws_h2_new_pps(enum lws_h2_protocol_send_type type)
{
struct lws_h2_protocol_send *pps = lws_malloc(sizeof(*pps), "pps");
if (pps)
pps->type = type;
return pps;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
void lws_h2_init(struct lws *wsi)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
{
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
wsi->h2.h2n->our_set = wsi->a.vhost->h2.set;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->h2.h2n->peer_set = lws_h2_defaults;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
h2: tx END_STREAM should only set HALF_CLOSED_LOCAL
2018-04-03 12:00:46 +08:00
void
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_state(struct lws *wsi, enum lws_h2_states s)
{
if (!wsi)
return;
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: state %s -> %s\n", __func__, lws_wsi_tag(wsi),
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2_state_names[wsi->h2.h2_state],
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2_state_names[s]);
http2: missing return when openssl too old
2017-10-22 10:38:43 +08:00
(void)h2_state_names;
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
wsi->h2.h2_state = (uint8_t)s;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
int
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
lws_h2_update_peer_txcredit(struct lws *wsi, unsigned int sid, int bump)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
{
struct lws *nwsi = lws_get_network_wsi(wsi);
struct lws_h2_protocol_send *pps;
assert(wsi);
cleaning
2020-01-02 08:32:23 +00:00
if (!bump)
return 0;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
if (sid == (unsigned int)-1)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
sid = wsi->mux.my_sid;
lwsl_info("%s: sid %d: bump %d -> %d\n", __func__, sid, bump,
linkit: support build using public sdk This provides support to build lws using the linkit 7697 public SDK from here https://docs.labs.mediatek.com/resource/mt7687-mt7697/en/downloads This toolchain has some challenges, its int32_t / uint32_t are long, so assumptions about format strings for those being %u / %d / %x all break. This fixes all the cases for the features enabled by the default cmake settings.
2020-01-16 07:16:01 +00:00
(int)wsi->txc.peer_tx_cr_est + bump);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
pps = lws_h2_new_pps(LWS_H2_PPS_UPDATE_WINDOW);
if (!pps)
return 1;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
pps->u.update_window.sid = (unsigned int)sid;
pps->u.update_window.credit = (unsigned int)bump;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->txc.peer_tx_cr_est += bump;
lws_wsi_txc_describe(&wsi->txc, __func__, wsi->mux.my_sid);
lws_pps_schedule(wsi, pps);
pps = lws_h2_new_pps(LWS_H2_PPS_UPDATE_WINDOW);
if (!pps)
return 1;
pps->u.update_window.sid = 0;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
pps->u.update_window.credit = (unsigned int)bump;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
nwsi->txc.peer_tx_cr_est += bump;
lws_wsi_txc_describe(&nwsi->txc, __func__, nwsi->mux.my_sid);
lws_pps_schedule(nwsi, pps);
return 0;
}
int
lws_h2_get_peer_txcredit_estimate(struct lws *wsi)
{
lws_wsi_txc_describe(&wsi->txc, __func__, wsi->mux.my_sid);
return (int)wsi->txc.peer_tx_cr_est;
}
static int
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
lws_h2_update_peer_txcredit_thresh(struct lws *wsi, unsigned int sid, int threshold, int bump)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
{
if (wsi->txc.peer_tx_cr_est > threshold)
return 0;
return lws_h2_update_peer_txcredit(wsi, sid, bump);
}
smp: more lock assertions
2021-04-04 04:06:24 +01:00
/* cx + vh lock */
static struct lws *
__lws_wsi_server_new(struct lws_vhost *vh, struct lws *parent_wsi,
unsigned int sid)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
{
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws *nwsi = lws_get_network_wsi(parent_wsi);
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
struct lws_h2_netconn *h2n = nwsi->h2.h2n;
route: on change also check source address still extant
2021-02-05 13:08:41 +00:00
char tmp[50], tmp1[50];
unsigned int n, b = 0;
evlib: count extant evlib handles instead of wsi
2021-01-19 14:19:12 +00:00
struct lws *wsi;
route: on change also check source address still extant
2021-02-05 13:08:41 +00:00
const char *p;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
smp: more lock assertions
2021-04-04 04:06:24 +01:00
lws_context_assert_lock_held(vh->context);
lws_vhost_assert_lock_held(vh);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/*
* The identifier of a newly established stream MUST be numerically
* greater than all streams that the initiating endpoint has opened or
* reserved. This governs streams that are opened using a HEADERS frame
* and streams that are reserved using PUSH_PROMISE. An endpoint that
* receives an unexpected stream identifier MUST respond with a
* connection error (Section 5.4.1) of type PROTOCOL_ERROR.
*/
if (sid <= h2n->highest_sid_opened) {
h2: tolerate unexpected server behaviour Make sure we only do one SETTINGS ack handling Fix corner case of ignored header with 0 data being the last in the HEADERS
2019-10-10 16:37:38 +01:00
lwsl_info("%s: tried to open lower sid %d (%d)\n", __func__,
format strings: fixes for when toolchain has uint32_t as unsigned long
2019-11-05 21:00:45 +00:00
sid, (int)h2n->highest_sid_opened);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(nwsi, H2_ERR_PROTOCOL_ERROR, "Bad sid");
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
return NULL;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
/* no more children allowed by parent */
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (parent_wsi->mux.child_count + 1 >
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
parent_wsi->h2.h2n->our_set.s[H2SET_MAX_CONCURRENT_STREAMS]) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lwsl_notice("reached concurrent stream limit\n");
return NULL;
}
route: on change also check source address still extant
2021-02-05 13:08:41 +00:00
n = 0;
p = &parent_wsi->lc.gutag[1];
do {
if (*p == '|') {
b++;
if (b == 3)
continue;
}
tmp1[n++] = *p++;
} while (b < 3 && n < sizeof(tmp1) - 2);
tmp1[n] = '\0';
lws_snprintf(tmp, sizeof(tmp), "h2_sid%u_(%s)", sid, tmp1);
wsi = lws_create_new_server_wsi(vh, parent_wsi->tsi, tmp);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!wsi) {
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_notice("new server wsi failed (%s)\n", lws_vh_tag(vh));
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return NULL;
}
lws_metrics There are a few build options that are trying to keep and report various statistics - DETAILED_LATENCY - SERVER_STATUS - WITH_STATS remove all those and establish a generic rplacement, lws_metrics. lws_metrics makes its stats available via an lws_system ops function pointer that the user code can set. Openmetrics export is supported, for, eg, prometheus scraping.
2021-01-06 15:08:22 +00:00
#if defined(LWS_WITH_SERVER)
Wextra Add -Wextra (with -Wno-unused-parameter) to unix builds in addition to -Wall -Werror. This can successfully build everything in Sai without warnings / errors.
2021-06-28 11:51:44 +01:00
if (lwsi_role_server(parent_wsi)) {
lws_metrics There are a few build options that are trying to keep and report various statistics - DETAILED_LATENCY - SERVER_STATUS - WITH_STATS remove all those and establish a generic rplacement, lws_metrics. lws_metrics makes its stats available via an lws_system ops function pointer that the user code can set. Openmetrics export is supported, for, eg, prometheus scraping.
2021-01-06 15:08:22 +00:00
lws_metrics_caliper_bind(wsi->cal_conn, wsi->a.context->mth_srv);
Wextra Add -Wextra (with -Wno-unused-parameter) to unix builds in addition to -Wall -Werror. This can successfully build everything in Sai without warnings / errors.
2021-06-28 11:51:44 +01:00
}
lws_metrics There are a few build options that are trying to keep and report various statistics - DETAILED_LATENCY - SERVER_STATUS - WITH_STATS remove all those and establish a generic rplacement, lws_metrics. lws_metrics makes its stats available via an lws_system ops function pointer that the user code can set. Openmetrics export is supported, for, eg, prometheus scraping.
2021-01-06 15:08:22 +00:00
#endif
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2n->highest_sid_opened = sid;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
lws_wsi_mux_insert(wsi, parent_wsi, sid);
h2: take care to retain initial sid
2020-04-06 13:22:01 +01:00
if (sid >= h2n->highest_sid)
h2n->highest_sid = sid + 2;
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
wsi->mux_substream = 1;
wsi->seen_nonpseudoheader = 0;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
wsi->txc.tx_cr = (int32_t)nwsi->h2.h2n->peer_set.s[H2SET_INITIAL_WINDOW_SIZE];
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->txc.peer_tx_cr_est =
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
(int32_t)nwsi->h2.h2n->our_set.s[H2SET_INITIAL_WINDOW_SIZE];
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_set_state(wsi, LRS_ESTABLISHED);
lwsi_set_role(wsi, lwsi_role(parent_wsi));
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
wsi->a.protocol = &vh->protocols[0];
http2 handle error path on ensure_user_space Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-23 09:36:18 +08:00
if (lws_ensure_user_space(wsi))
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
goto bail1;
ss: server: h1, h2, ws basic support Add initial support for defining servers using Secure Streams policy and api semantics. Serving h1, h2 and ws should be functional, the new minimal example shows a combined http + SS server with an incrementing ws message shown in the browser over tls, in around 200 lines of user code. NOP out anything to do with plugins, they're not currently used. Update the docs correspondingly.
2020-07-27 10:03:12 +01:00
#if defined(LWS_WITH_SERVER) && defined(LWS_WITH_SECURE_STREAMS)
if (lws_adopt_ss_server_accept(wsi))
goto bail1;
#endif
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
/* get the ball rolling */
lws_validity_confirmed(wsi);
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s new ch %s, sid %d, usersp=%p\n", __func__,
lws_wsi_tag(parent_wsi), lws_wsi_tag(wsi), sid, wsi->user_space);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
lws_wsi_txc_describe(&wsi->txc, __func__, wsi->mux.my_sid);
lws_wsi_txc_describe(&nwsi->txc, __func__, 0);
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
return wsi;
http2 handle error path on ensure_user_space Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-23 09:36:18 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
bail1:
/* undo the insert */
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
parent_wsi->mux.child_list = wsi->mux.sibling_list;
parent_wsi->mux.child_count--;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (wsi->user_space)
lws_free_set_NULL(wsi->user_space);
vh->protocols[0].callback(wsi, LWS_CALLBACK_WSI_DESTROY, NULL, NULL, 0);
smp: more lock assertions
2021-04-04 04:06:24 +01:00
__lws_vhost_unbind_wsi(wsi);
http2 handle error path on ensure_user_space Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-23 09:36:18 +08:00
lws_free(wsi);
return NULL;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
struct lws *
lws_wsi_h2_adopt(struct lws *parent_wsi, struct lws *wsi)
{
struct lws *nwsi = lws_get_network_wsi(parent_wsi);
/* no more children allowed by parent */
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (parent_wsi->mux.child_count + 1 >
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
parent_wsi->h2.h2n->our_set.s[H2SET_MAX_CONCURRENT_STREAMS]) {
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lwsl_notice("reached concurrent stream limit\n");
return NULL;
}
/* sid is set just before issuing the headers, ensuring monoticity */
wsi->seen_nonpseudoheader = 0;
clean: internally use LWS_WITH_CLIENT and _SERVER Remove some more things in LWS_WITH_SERVER=0 case
2019-08-18 05:04:15 +01:00
#if defined(LWS_WITH_CLIENT)
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
wsi->client_mux_substream = 1;
lws_diskcache: split generic parts from gitohashi into lws
2018-11-12 15:24:42 +08:00
#endif
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
wsi->h2.initialized = 1;
h2: client sid: must be allocated at header send There is no way to allocate a client wsi mux sid before the headers are send, because we don't know the order in which new wsi headers will be sent and so seen by the peer. The peer inisists that sid indexes only increase... we cannot allocat sids monotonically at the client and then send them disordered...
2020-05-04 08:26:48 +01:00
#if 0
/* only assign sid at header send time when we know it */
minimal-http-client-multi: add POST This adds support for POST in both h1 and h2 queues / stream binding. The previous queueing tried to keep the "leader" wsi who made the actual connection around and have it act on the transaction queue tail if it had done its own thing. This refactors it so instead, who is the "leader" moves down the queue and the queued guys inherit the fd, SSL * and queue from the old leader as they take over. This lets them operate in their own wsi identity directly and gets rid of all the "effective wsi" checks, which was applied incompletely and getting out of hand considering the separate lws_mux checks for h2 and other muxed protocols alongside it. This change also allows one wsi at a time to own the transaction for POST. --post is added as an option to lws-minimal-http-client-multi and 6 extra selftests with POST on h1/h2, pipelined or not and staggered or not are added to the CI.
2020-01-30 13:19:11 +00:00
if (!wsi->mux.my_sid) {
wsi->mux.my_sid = nwsi->h2.h2n->highest_sid;
nwsi->h2.h2n->highest_sid += 2;
}
h2: client sid: must be allocated at header send There is no way to allocate a client wsi mux sid before the headers are send, because we don't know the order in which new wsi headers will be sent and so seen by the peer. The peer inisists that sid indexes only increase... we cannot allocat sids monotonically at the client and then send them disordered...
2020-05-04 08:26:48 +01:00
#endif
minimal-http-client-multi: add POST This adds support for POST in both h1 and h2 queues / stream binding. The previous queueing tried to keep the "leader" wsi who made the actual connection around and have it act on the transaction queue tail if it had done its own thing. This refactors it so instead, who is the "leader" moves down the queue and the queued guys inherit the fd, SSL * and queue from the old leader as they take over. This lets them operate in their own wsi identity directly and gets rid of all the "effective wsi" checks, which was applied incompletely and getting out of hand considering the separate lws_mux checks for h2 and other muxed protocols alongside it. This change also allows one wsi at a time to own the transaction for POST. --post is added as an option to lws-minimal-http-client-multi and 6 extra selftests with POST on h1/h2, pipelined or not and staggered or not are added to the CI.
2020-01-30 13:19:11 +00:00
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: binding wsi %s to sid %d (next %d)\n", __func__,
lws_wsi_tag(wsi), (int)wsi->mux.my_sid, (int)nwsi->h2.h2n->highest_sid);
h2: take care to retain initial sid
2020-04-06 13:22:01 +01:00
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
lws_wsi_mux_insert(wsi, parent_wsi, wsi->mux.my_sid);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
wsi->txc.tx_cr = (int32_t)nwsi->h2.h2n->peer_set.s[H2SET_INITIAL_WINDOW_SIZE];
wsi->txc.peer_tx_cr_est = (int32_t)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
nwsi->h2.h2n->our_set.s[H2SET_INITIAL_WINDOW_SIZE];
lws_wsi_txc_describe(&wsi->txc, __func__, wsi->mux.my_sid);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (lws_ensure_user_space(wsi))
goto bail1;
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
lws_role_transition(wsi, LWSIFR_CLIENT, LRS_H2_WAITING_TO_SEND_HEADERS,
&role_ops_h2);
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lws_callback_on_writable(wsi);
return wsi;
bail1:
/* undo the insert */
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
parent_wsi->mux.child_list = wsi->mux.sibling_list;
parent_wsi->mux.child_count--;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (wsi->user_space)
lws_free_set_NULL(wsi->user_space);
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
wsi->a.protocol->callback(wsi, LWS_CALLBACK_WSI_DESTROY, NULL, NULL, 0);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lws_free(wsi);
return NULL;
}
h2: post: add states to wait for body
2020-12-19 09:13:24 +00:00
int
lws_h2_issue_preface(struct lws *wsi)
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
{
struct lws_h2_netconn *h2n = wsi->h2.h2n;
struct lws_h2_protocol_send *pps;
http: redirect: dont try from h2 at the moment h1->h1, h1->h2 are OK, but h2->anything needs more work. Just fail it early at the moment.
2021-04-10 06:45:21 +01:00
if (!h2n) {
lwsl_warn("%s: no valid h2n\n", __func__);
return 1;
}
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
if (h2n->sent_preface)
return 1;
smd: add more scenarios to tests Let's have 4 x forked clients all intercommunicate via the SS proxy, and add it to ctest.
2021-01-01 09:01:09 +00:00
lwsl_debug("%s: %s: fd %d\n", __func__, lws_wsi_tag(wsi), (int)wsi->desc.sockfd);
h2: post: add states to wait for body
2020-12-19 09:13:24 +00:00
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (lws_issue_raw(wsi, (uint8_t *)preface, strlen(preface)) !=
(int)strlen(preface))
return 1;
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
h2n->sent_preface = 1;
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
lws_role_transition(wsi, LWSIFR_CLIENT, LRS_H2_WAITING_TO_SEND_HEADERS,
&role_ops_h2);
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
h2n->count = 0;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->txc.tx_cr = 65535;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
/*
* we must send a settings frame
*/
pps = lws_h2_new_pps(LWS_H2_PPS_MY_SETTINGS);
if (!pps)
return 1;
lws_pps_schedule(wsi, pps);
lwsl_info("%s: h2 client sending settings\n", __func__);
return 0;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
void
lws_pps_schedule(struct lws *wsi, struct lws_h2_protocol_send *pps)
{
struct lws *nwsi = lws_get_network_wsi(wsi);
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
struct lws_h2_netconn *h2n = nwsi->h2.h2n;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
mbedtls: improve api detection mbedtls cmake api detection was not able to work on esp-idf well. Improve diagnostics and reaction if we ever see that again.
2022-02-01 07:59:41 +00:00
if (!h2n) {
lwsl_warn("%s: null h2n\n", __func__);
lws_free(pps);
return;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
pps->next = h2n->pps;
h2n->pps = pps;
lws_rx_flow_control(wsi, LWS_RXFLOW_REASON_APPLIES_DISABLE |
LWS_RXFLOW_REASON_H2_PPS_PENDING);
lws_callback_on_writable(wsi);
}
int
lws_h2_goaway(struct lws *wsi, uint32_t err, const char *reason)
{
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
struct lws_h2_netconn *h2n = wsi->h2.h2n;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws_h2_protocol_send *pps;
if (h2n->type == LWS_H2_FRAME_TYPE_COUNT)
return 0;
pps = lws_h2_new_pps(LWS_H2_PPS_GOAWAY);
if (!pps)
return 1;
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: ERR 0x%x, '%s'\n", __func__, lws_wsi_tag(wsi), (int)err, reason);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
pps->u.ga.err = err;
refactor: subdirs for source in lib Split out some optional code into own sources to shrink down libwebsockets.c and server.c a bit
2017-10-16 12:52:32 +08:00
pps->u.ga.highest_sid = h2n->highest_sid;
lws_snprintf
2018-03-12 09:28:26 +08:00
lws_strncpy(pps->u.ga.str, reason, sizeof(pps->u.ga.str));
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_pps_schedule(wsi, pps);
h2n->type = LWS_H2_FRAME_TYPE_COUNT; /* ie, IGNORE */
return 0;
}
int
lws_h2_rst_stream(struct lws *wsi, uint32_t err, const char *reason)
{
struct lws *nwsi = lws_get_network_wsi(wsi);
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
struct lws_h2_netconn *h2n = nwsi->h2.h2n;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws_h2_protocol_send *pps;
http proxy: client: unix socket support This allows the client stuff to understand that addresses beginning with '+' represent unix sockets. If the first character after the '+' is '@', it understands that the '@' should be read as '\0', in order to use Linux "abstract namespace" sockets. Further the lws_parse_uri() helper is extended to understand the convention that an address starting with + is a unix socket, and treats the socket path as delimited by ':', eg http://+/var/run/mysocket:/my/path HTTP Proxy is updated to allow mounts to these unix socket paths. Proxy connections go out on h1, but are dynamically translated to h1 or h2 on the incoming side. Proxy usage of libhubbub is separated out... LWS_WITH_HTTP_PROXY is on by default, and LWS_WITH_HUBBUB is off by default.
2018-09-04 08:06:46 +08:00
if (!h2n)
return 0;
ws-over-h2: wrong sid on rst_stream rst_stream is almost always called from the h2 parser... in the case a ws-over-h2 is closing though, it can't reference the h2n parser sid context to get the sid to send... it's unrelated at that time. https://bugzilla.mozilla.org/show_bug.cgi?id=1590299 https://github.com/warmcat/libwebsockets/issues/1752
2019-11-04 21:19:53 +00:00
if (!wsi->h2_stream_carries_ws && h2n->type == LWS_H2_FRAME_TYPE_COUNT)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 0;
pps = lws_h2_new_pps(LWS_H2_PPS_RST_STREAM);
if (!pps)
return 1;
cleaning
2020-01-02 08:32:23 +00:00
lwsl_info("%s: RST_STREAM 0x%x, sid %d, REASON '%s'\n", __func__,
(int)err, wsi->mux.my_sid, reason);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
pps->u.rs.sid = wsi->mux.my_sid;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
pps->u.rs.err = err;
ws-over-h2: wrong sid on rst_stream rst_stream is almost always called from the h2 parser... in the case a ws-over-h2 is closing though, it can't reference the h2n parser sid context to get the sid to send... it's unrelated at that time. https://bugzilla.mozilla.org/show_bug.cgi?id=1590299 https://github.com/warmcat/libwebsockets/issues/1752
2019-11-04 21:19:53 +00:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_pps_schedule(wsi, pps);
h2n->type = LWS_H2_FRAME_TYPE_COUNT; /* ie, IGNORE */
lws_h2_state(wsi, LWS_H2_STATE_CLOSED);
return 0;
}
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
int
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_settings(struct lws *wsi, struct http2_settings *settings,
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
unsigned char *buf, int len)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
{
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws *nwsi = lws_get_network_wsi(wsi);
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
unsigned int a, b;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
if (!len)
return 0;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (len < LWS_H2_SETTINGS_LEN)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
return 1;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
while (len >= LWS_H2_SETTINGS_LEN) {
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
a = (unsigned int)((buf[0] << 8) | buf[1]);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!a || a >= H2SET_COUNT)
goto skip;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
b = (unsigned int)(buf[2] << 24 | buf[3] << 16 | buf[4] << 8 | buf[5]);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
switch (a) {
case H2SET_HEADER_TABLE_SIZE:
break;
case H2SET_ENABLE_PUSH:
if (b > 1) {
lws_h2_goaway(nwsi, H2_ERR_PROTOCOL_ERROR,
refactor: subdirs for source in lib Split out some optional code into own sources to shrink down libwebsockets.c and server.c a bit
2017-10-16 12:52:32 +08:00
"ENABLE_PUSH invalid arg");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 1;
}
break;
case H2SET_MAX_CONCURRENT_STREAMS:
break;
case H2SET_INITIAL_WINDOW_SIZE:
if (b > 0x7fffffff) {
lws_h2_goaway(nwsi, H2_ERR_FLOW_CONTROL_ERROR,
refactor: subdirs for source in lib Split out some optional code into own sources to shrink down libwebsockets.c and server.c a bit
2017-10-16 12:52:32 +08:00
"Inital Window beyond max");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 1;
}
rtos: add support for AMAZON_RTOS
2019-06-05 05:04:17 +01:00
h2: overflow quirk Some servers set the tx credit to the absolute max and then add to it... this is illegal (and checked for in h2spec). Add a quirk flag that works around it by reducing the initial tx credit size by a factor of 16.
2019-10-03 07:58:24 -07:00
#if defined(LWS_WITH_CLIENT)
build: change AMAZON_NOART to AMAZON_LINUX
2019-07-13 14:50:58 -07:00
#if defined(LWS_AMAZON_RTOS) || defined(LWS_AMAZON_LINUX)
h2: overflow quirk Some servers set the tx credit to the absolute max and then add to it... this is illegal (and checked for in h2spec). Add a quirk flag that works around it by reducing the initial tx credit size by a factor of 16.
2019-10-03 07:58:24 -07:00
if (
#else
if (wsi->flags & LCCSCF_H2_QUIRK_OVERFLOWS_TXCR &&
#endif
b == 0x7fffffff) {
b >>= 4;
rtos: add support for AMAZON_RTOS
2019-06-05 05:04:17 +01:00
break;
}
#endif
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/*
* In addition to changing the flow-control window for
* streams that are not yet active, a SETTINGS frame
* can alter the initial flow-control window size for
* streams with active flow-control windows (that is,
* streams in the "open" or "half-closed (remote)"
* state). When the value of
* SETTINGS_INITIAL_WINDOW_SIZE changes, a receiver
* MUST adjust the size of all stream flow-control
* windows that it maintains by the difference between
* the new value and the old value.
*/
lws_start_foreach_ll(struct lws *, w,
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
nwsi->mux.child_list) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lwsl_info("%s: adi child tc cr %d +%d -> %d",
linkit: support build using public sdk This provides support to build lws using the linkit 7697 public SDK from here https://docs.labs.mediatek.com/resource/mt7687-mt7697/en/downloads This toolchain has some challenges, its int32_t / uint32_t are long, so assumptions about format strings for those being %u / %d / %x all break. This fixes all the cases for the features enabled by the default cmake settings.
2020-01-16 07:16:01 +00:00
__func__, (int)w->txc.tx_cr,
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
b - (unsigned int)settings->s[a],
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
(int)(w->txc.tx_cr + (int)b -
(int)settings->s[a]));
w->txc.tx_cr += (int)b - (int)settings->s[a];
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (w->txc.tx_cr > 0 &&
w->txc.tx_cr <=
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
(int32_t)(b - settings->s[a]))
cleaning
2020-01-02 08:32:23 +00:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_callback_on_writable(w);
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
} lws_end_foreach_ll(w, mux.sibling_list);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
case H2SET_MAX_FRAME_SIZE:
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
if (b < wsi->a.vhost->h2.set.s[H2SET_MAX_FRAME_SIZE]) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(nwsi, H2_ERR_PROTOCOL_ERROR,
"Frame size < initial");
return 1;
}
h2: SETTINGS: allow 16M-1 peer max frame size We only allow 8M-1 when we should allow 16M-1 without blowing a PROTOCOL ERROR.
2019-07-07 12:19:54 +01:00
if (b > 0x00ffffff) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(nwsi, H2_ERR_PROTOCOL_ERROR,
"Settings Frame size above max");
return 1;
}
break;
case H2SET_MAX_HEADER_LIST_SIZE:
break;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
settings->s[a] = b;
lwsl_info("http2 settings %d <- 0x%x\n", a, b);
skip:
len -= LWS_H2_SETTINGS_LEN;
buf += LWS_H2_SETTINGS_LEN;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
if (len)
return 1;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_dump_settings(settings);
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
return 0;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/* RFC7640 Sect 6.9
*
* The WINDOW_UPDATE frame can be specific to a stream or to the entire
* connection. In the former case, the frame's stream identifier
* indicates the affected stream; in the latter, the value "0" indicates
* that the entire connection is the subject of the frame.
*
* ...
*
* Two flow-control windows are applicable: the stream flow-control
* window and the connection flow-control window. The sender MUST NOT
* send a flow-controlled frame with a length that exceeds the space
* available in either of the flow-control windows advertised by the
* receiver. Frames with zero length with the END_STREAM flag set (that
* is, an empty DATA frame) MAY be sent if there is no available space
* in either flow-control window.
*/
int
lws_h2_tx_cr_get(struct lws *wsi)
http2 track content length add END_STREAM Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-18 12:23:05 +08:00
{
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
int c = wsi->txc.tx_cr;
h2 client: deal correctly with zero length DATA with END_STREAM
2019-10-16 15:08:25 +01:00
struct lws *nwsi = lws_get_network_wsi(wsi);
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (!wsi->mux_substream && !nwsi->upgraded_to_http2)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return ~0x80000000;
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info ("%s: %s: own tx credit %d: nwsi credit %d\n",
__func__, lws_wsi_tag(wsi), c, (int)nwsi->txc.tx_cr);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (nwsi->txc.tx_cr < c)
c = nwsi->txc.tx_cr;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (c < 0)
return 0;
return c;
http2 track content length add END_STREAM Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-18 12:23:05 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
void
lws_h2_tx_cr_consume(struct lws *wsi, int consumed)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
{
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws *nwsi = lws_get_network_wsi(wsi);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->txc.tx_cr -= consumed;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (nwsi != wsi)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
nwsi->txc.tx_cr -= consumed;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
int lws_h2_frame_write(struct lws *wsi, int type, int flags,
unsigned int sid, unsigned int len, unsigned char *buf)
{
struct lws *nwsi = lws_get_network_wsi(wsi);
unsigned char *p = &buf[-LWS_H2_FRAME_HEADER_LENGTH];
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
int n;
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
//if (wsi->h2_stream_carries_ws)
// lwsl_hexdump_level(LLL_NOTICE, buf, len);
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*p++ = (uint8_t)(len >> 16);
*p++ = (uint8_t)(len >> 8);
*p++ = (uint8_t)len;
*p++ = (uint8_t)type;
*p++ = (uint8_t)flags;
*p++ = (uint8_t)(sid >> 24);
*p++ = (uint8_t)(sid >> 16);
*p++ = (uint8_t)(sid >> 8);
*p++ = (uint8_t)sid;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_debug("%s: %s (eff %s). typ %d, fl 0x%x, sid=%d, len=%d, "
"txcr=%d, nwsi->txcr=%d\n", __func__, lws_wsi_tag(wsi),
lws_wsi_tag(nwsi), type, flags,
linkit: support build using public sdk This provides support to build lws using the linkit 7697 public SDK from here https://docs.labs.mediatek.com/resource/mt7687-mt7697/en/downloads This toolchain has some challenges, its int32_t / uint32_t are long, so assumptions about format strings for those being %u / %d / %x all break. This fixes all the cases for the features enabled by the default cmake settings.
2020-01-16 07:16:01 +00:00
sid, len, (int)wsi->txc.tx_cr, (int)nwsi->txc.tx_cr);
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (type == LWS_H2_FRAME_TYPE_DATA) {
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (wsi->txc.tx_cr < (int)len)
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: sending payload len %d"
" but tx_cr only %d!\n", __func__,
lws_wsi_tag(wsi), len, (int)wsi->txc.tx_cr);
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
lws_h2_tx_cr_consume(wsi, (int)len);
http2 window update Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-29 09:39:08 +08:00
}
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
n = lws_issue_raw(nwsi, &buf[-LWS_H2_FRAME_HEADER_LENGTH],
len + LWS_H2_FRAME_HEADER_LENGTH);
if (n < 0)
return n;
if (n >= LWS_H2_FRAME_HEADER_LENGTH)
return n - LWS_H2_FRAME_HEADER_LENGTH;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
return n;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
static void lws_h2_set_bin(struct lws *wsi, int n, unsigned char *buf)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
{
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*buf++ = (uint8_t)(n >> 8);
*buf++ = (uint8_t)n;
*buf++ = (uint8_t)(wsi->h2.h2n->our_set.s[n] >> 24);
*buf++ = (uint8_t)(wsi->h2.h2n->our_set.s[n] >> 16);
*buf++ = (uint8_t)(wsi->h2.h2n->our_set.s[n] >> 8);
*buf = (uint8_t)wsi->h2.h2n->our_set.s[n];
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
rxflow buflist: handle forced service
2018-04-17 11:43:20 +08:00
/* we get called on the network connection */
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
int lws_h2_do_pps_send(struct lws *wsi)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
{
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
struct lws_h2_netconn *h2n = wsi->h2.h2n;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws_h2_protocol_send *pps = NULL;
struct lws *cwsi;
uint8_t set[LWS_PRE + 64], *p = &set[LWS_PRE], *q;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
int n, m = 0, flags = 0;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!h2n)
return 1;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/* get the oldest pps */
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_start_foreach_llp(struct lws_h2_protocol_send **, pps1, h2n->pps) {
if ((*pps1)->next == NULL) { /* we are the oldest in the list */
pps = *pps1; /* remove us from the list */
*pps1 = NULL;
continue;
}
} lws_end_foreach_llp(pps1, next);
if (!pps)
return 1;
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: %d\n", __func__, lws_wsi_tag(wsi), pps->type);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
switch (pps->type) {
case LWS_H2_PPS_MY_SETTINGS:
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/*
* if any of our settings varies from h2 "default defaults"
esp32: align with esp-idf changes LWIP_SOCKET_OFFSET is now nonzero, which I handled a while back. But the C api support for it is broken in esp-idf. select() takes unmodified fds, but FD_SET / FD_ISSET etc must have the offset subtracted on their args. With this we are working on current HEAD esp-idf.
2017-11-24 10:54:59 +08:00
* then we must inform the peer
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
*/
for (n = 1; n < H2SET_COUNT; n++)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (h2n->our_set.s[n] != lws_h2_defaults.s[n]) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lwsl_debug("sending SETTING %d 0x%x\n", n,
cleaning
2020-01-02 08:32:23 +00:00
(unsigned int)
wsi->h2.h2n->our_set.s[n]);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_set_bin(wsi, n, &set[LWS_PRE + m]);
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
m += (int)sizeof(h2n->one_setting);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
n = lws_h2_frame_write(wsi, LWS_H2_FRAME_TYPE_SETTINGS,
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
flags, LWS_H2_STREAM_ID_MASTER, (unsigned int)m,
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
&set[LWS_PRE]);
if (n != m) {
lwsl_info("send %d %d\n", n, m);
goto bail;
}
break;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
case LWS_H2_PPS_SETTINGS_INITIAL_UPDATE_WINDOW:
q = &set[LWS_PRE];
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*q++ = (uint8_t)(H2SET_INITIAL_WINDOW_SIZE >> 8);
*q++ = (uint8_t)(H2SET_INITIAL_WINDOW_SIZE);
*q++ = (uint8_t)(pps->u.update_window.credit >> 24);
*q++ = (uint8_t)(pps->u.update_window.credit >> 16);
*q++ = (uint8_t)(pps->u.update_window.credit >> 8);
*q = (uint8_t)(pps->u.update_window.credit);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
lwsl_debug("%s: resetting initial window to %d\n", __func__,
(int)pps->u.update_window.credit);
n = lws_h2_frame_write(wsi, LWS_H2_FRAME_TYPE_SETTINGS,
flags, LWS_H2_STREAM_ID_MASTER, 6,
&set[LWS_PRE]);
if (n != 6) {
lwsl_info("send %d %d\n", n, m);
goto bail;
}
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_PPS_ACK_SETTINGS:
/* send ack ... always empty */
n = lws_h2_frame_write(wsi, LWS_H2_FRAME_TYPE_SETTINGS, 1,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
LWS_H2_STREAM_ID_MASTER, 0,
&set[LWS_PRE]);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (n) {
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
lwsl_err("%s: writing settings ack frame failed %d\n", __func__, n);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
goto bail;
}
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->h2_acked_settings = 0;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/* this is the end of the preface dance then? */
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
if (lwsi_state(wsi) == LRS_H2_AWAIT_SETTINGS) {
lwsi_set_state(wsi, LRS_ESTABLISHED);
freertos: rename esp32 plat to freertos
2019-08-18 10:35:43 +01:00
#if defined(LWS_WITH_FILE_OPS)
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
wsi->http.fop_fd = NULL;
freertos: rename esp32 plat to freertos
2019-08-18 10:35:43 +01:00
#endif
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (lws_is_ssl(lws_get_network_wsi(wsi)))
break;
h2: add prior knowledge support
2020-05-07 15:21:48 -07:00
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
if (wsi->a.vhost->options &
h2: add prior knowledge support
2020-05-07 15:21:48 -07:00
LWS_SERVER_OPTION_H2_PRIOR_KNOWLEDGE)
break;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
/*
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
* we need to treat the headers from the upgrade as the
* first job. So these need to get shifted to sid 1.
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
*/
smp: more lock assertions
2021-04-04 04:06:24 +01:00
lws_context_lock(wsi->a.context, "h2 mig");
lws_vhost_lock(wsi->a.vhost);
h2n->swsi = __lws_wsi_server_new(wsi->a.vhost, wsi, 1);
lws_vhost_unlock(wsi->a.vhost);
lws_context_unlock(wsi->a.context);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!h2n->swsi)
goto bail;
/* pass on the initial headers to SID 1 */
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
h2n->swsi->http.ah = wsi->http.ah;
wsi->http.ah = NULL;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lwsl_info("%s: inherited headers %p\n", __func__,
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
h2n->swsi->http.ah);
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
h2n->swsi->txc.tx_cr = (int32_t)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
h2n->our_set.s[H2SET_INITIAL_WINDOW_SIZE];
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("initial tx credit on %s: %d\n",
lws_wsi_tag(h2n->swsi),
(int)h2n->swsi->txc.tx_cr);
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2n->swsi->h2.initialized = 1;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/* demanded by HTTP2 */
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2n->swsi->h2.END_STREAM = 1;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lwsl_info("servicing initial http request\n");
freertos: rename esp32 plat to freertos
2019-08-18 10:35:43 +01:00
#if defined(LWS_WITH_SERVER)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (lws_http_action(h2n->swsi))
goto bail;
freertos: rename esp32 plat to freertos
2019-08-18 10:35:43 +01:00
#endif
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
}
break;
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
/*
* h2 only has PING... ACK = 0 = ping, ACK = 1 = pong
*/
case LWS_H2_PPS_PING:
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_PPS_PONG:
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
if (pps->type == LWS_H2_PPS_PING)
lwsl_info("sending PING\n");
else {
lwsl_info("sending PONG\n");
flags = LWS_H2_FLAG_SETTINGS_ACK;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
memcpy(&set[LWS_PRE], pps->u.ping.ping_payload, 8);
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
n = lws_h2_frame_write(wsi, LWS_H2_FRAME_TYPE_PING, flags,
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
LWS_H2_STREAM_ID_MASTER, 8,
&set[LWS_PRE]);
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
if (n != 8)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
goto bail;
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_PPS_GOAWAY:
lwsl_info("LWS_H2_PPS_GOAWAY\n");
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*p++ = (uint8_t)(pps->u.ga.highest_sid >> 24);
*p++ = (uint8_t)(pps->u.ga.highest_sid >> 16);
*p++ = (uint8_t)(pps->u.ga.highest_sid >> 8);
*p++ = (uint8_t)(pps->u.ga.highest_sid);
*p++ = (uint8_t)(pps->u.ga.err >> 24);
*p++ = (uint8_t)(pps->u.ga.err >> 16);
*p++ = (uint8_t)(pps->u.ga.err >> 8);
*p++ = (uint8_t)(pps->u.ga.err);
refactor: subdirs for source in lib Split out some optional code into own sources to shrink down libwebsockets.c and server.c a bit
2017-10-16 12:52:32 +08:00
q = (unsigned char *)pps->u.ga.str;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
n = 0;
build: enable signed vs unsigned warnings on gcc This enables selected things from -Wextra, can't use -Wextra because it is fussy enough to complain about unused params on functions... they are there for a reason. -Wsign-compare -Wignored-qualifiers not -Wimplicit-fallthrough=3 ... only on gcc 7 -Wtype-limits -Wuninitialized not -Wclobbered ... only on gcc 7ish fix the warnings everywhere they were found.
2017-10-20 17:45:02 +08:00
while (*q && n++ < (int)sizeof(pps->u.ga.str))
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
*p++ = *q++;
h2n->we_told_goaway = 1;
n = lws_h2_frame_write(wsi, LWS_H2_FRAME_TYPE_GOAWAY, 0,
LWS_H2_STREAM_ID_MASTER,
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
(unsigned int)lws_ptr_diff(p, &set[LWS_PRE]),
appveyor: Enable 64-bit build with OpenSSL and HTTP2 Fix warnings found from that Introduce lws_ptr_diff(head, tail) helper to normalize pointers to char *, do the subtraction and cast the result to int.
2017-10-25 08:00:23 +08:00
&set[LWS_PRE]);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (n != 4) {
lwsl_info("send %d %d\n", n, m);
goto bail;
}
goto bail;
case LWS_H2_PPS_RST_STREAM:
lwsl_info("LWS_H2_PPS_RST_STREAM\n");
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*p++ = (uint8_t)(pps->u.rs.err >> 24);
*p++ = (uint8_t)(pps->u.rs.err >> 16);
*p++ = (uint8_t)(pps->u.rs.err >> 8);
*p++ = (uint8_t)(pps->u.rs.err);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
n = lws_h2_frame_write(wsi, LWS_H2_FRAME_TYPE_RST_STREAM,
0, pps->u.rs.sid, 4, &set[LWS_PRE]);
if (n != 4) {
lwsl_info("send %d %d\n", n, m);
goto bail;
}
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
cwsi = lws_wsi_mux_from_id(wsi, pps->u.rs.sid);
rxflow buflist: handle forced service
2018-04-17 11:43:20 +08:00
if (cwsi) {
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_debug("%s: closing cwsi %s %s %s (wsi %s)\n",
__func__, lws_wsi_tag(cwsi),
cwsi->role_ops->name,
cwsi->a.protocol->name, lws_wsi_tag(wsi));
internal close api: add caller id
2018-02-03 13:48:18 +08:00
lws_close_free_wsi(cwsi, 0, "reset stream");
rxflow buflist: handle forced service
2018-04-17 11:43:20 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
case LWS_H2_PPS_UPDATE_WINDOW:
cleaning
2020-01-02 08:32:23 +00:00
lwsl_info("Issuing LWS_H2_PPS_UPDATE_WINDOW: sid %d: add %d\n",
format strings: fixes for when toolchain has uint32_t as unsigned long
2019-11-05 21:00:45 +00:00
(int)pps->u.update_window.sid,
(int)pps->u.update_window.credit);
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*p++ = (uint8_t)((pps->u.update_window.credit >> 24) & 0x7f); /* 31b */
*p++ = (uint8_t)(pps->u.update_window.credit >> 16);
*p++ = (uint8_t)(pps->u.update_window.credit >> 8);
*p++ = (uint8_t)(pps->u.update_window.credit);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
n = lws_h2_frame_write(wsi, LWS_H2_FRAME_TYPE_WINDOW_UPDATE,
refactor: subdirs for source in lib Split out some optional code into own sources to shrink down libwebsockets.c and server.c a bit
2017-10-16 12:52:32 +08:00
0, pps->u.update_window.sid, 4,
&set[LWS_PRE]);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (n != 4) {
lwsl_info("send %d %d\n", n, m);
goto bail;
}
break;
default:
break;
}
lws_free(pps);
return 0;
bail:
lws_free(pps);
return 1;
}
h2 client: deal correctly with zero length DATA with END_STREAM
2019-10-16 15:08:25 +01:00
static int
lws_h2_parse_end_of_frame(struct lws *wsi);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/*
* The frame header part has just completely arrived.
esp32: align with esp-idf changes LWIP_SOCKET_OFFSET is now nonzero, which I handled a while back. But the C api support for it is broken in esp-idf. select() takes unmodified fds, but FD_SET / FD_ISSET etc must have the offset subtracted on their args. With this we are working on current HEAD esp-idf.
2017-11-24 10:54:59 +08:00
* Perform actions for header completion.
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
*/
static int
lws_h2_parse_frame_header(struct lws *wsi)
{
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
struct lws_h2_netconn *h2n = wsi->h2.h2n;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws_h2_protocol_send *pps;
int n;
/*
* We just got the frame header
*/
h2n->count = 0;
h2n->swsi = wsi;
/* b31 is a reserved bit */
h2n->sid = h2n->sid & 0x7fffffff;
if (h2n->sid && !(h2n->sid & 1)) {
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
char pes[32];
lws_snprintf(pes, sizeof(pes), "Even Stream ID 0x%x", (unsigned int)h2n->sid);
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR, pes);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 0;
}
/* let the network wsi live a bit longer if subs are active */
rtos: add support for AMAZON_RTOS
2019-06-05 05:04:17 +01:00
lws_http_mark_sse https://github.com/warmcat/libwebsockets/issues/1486
2019-01-30 20:59:56 +08:00
if (!wsi->immortal_substream_count)
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
lws_set_timeout(wsi, PENDING_TIMEOUT_HTTP_KEEPALIVE_IDLE,
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
wsi->a.vhost->keepalive_timeout ?
wsi->a.vhost->keepalive_timeout : 31);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (h2n->sid)
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
h2n->swsi = lws_wsi_mux_from_id(wsi, h2n->sid);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_debug("%s (%s): fr hdr: typ 0x%x, fla 0x%x, sid 0x%x, len 0x%x\n",
lws_wsi_tag(wsi), lws_wsi_tag(h2n->swsi), h2n->type,
h2n->flags, (unsigned int)h2n->sid, (unsigned int)h2n->length);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (h2n->we_told_goaway && h2n->sid > h2n->highest_sid)
h2n->type = LWS_H2_FRAME_TYPE_COUNT; /* ie, IGNORE */
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
if (h2n->type >= LWS_H2_FRAME_TYPE_COUNT) {
lwsl_info("%s: ignoring unknown frame type %d (len %d)\n", __func__, h2n->type, (unsigned int)h2n->length);
/* we MUST ignore frames we don't understand */
h2n->type = LWS_H2_FRAME_TYPE_COUNT;
}
/*
* Even if we have decided to logically ignore this frame, we must
* consume the correct "frame length" amount of data to retain sync
*/
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (h2n->length > h2n->our_set.s[H2SET_MAX_FRAME_SIZE]) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/*
* peer sent us something bigger than we told
* it we would allow
*/
cleaning
2020-01-02 08:32:23 +00:00
lwsl_info("%s: received oversize frame %d\n", __func__,
(unsigned int)h2n->length);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_FRAME_SIZE_ERROR,
"Peer ignored our frame size setting");
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
return 1;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
if (h2n->swsi)
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s, State: %s, received cmd %d\n",
__func__, lws_wsi_tag(h2n->swsi),
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2_state_names[h2n->swsi->h2.h2_state], h2n->type);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
else {
/* if it's data, either way no swsi means CLOSED state */
if (h2n->type == LWS_H2_FRAME_TYPE_DATA) {
lws_diskcache: split generic parts from gitohashi into lws
2018-11-12 15:24:42 +08:00
if (h2n->sid <= h2n->highest_sid_opened
clean: internally use LWS_WITH_CLIENT and _SERVER Remove some more things in LWS_WITH_SERVER=0 case
2019-08-18 05:04:15 +01:00
#if defined(LWS_WITH_CLIENT)
lws_diskcache: split generic parts from gitohashi into lws
2018-11-12 15:24:42 +08:00
&& wsi->client_h2_alpn
#endif
) {
h2 client: deal correctly with zero length DATA with END_STREAM
2019-10-16 15:08:25 +01:00
lwsl_notice("ignoring straggling data fl 0x%x\n",
h2n->flags);
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
/* ie, IGNORE */
h2n->type = LWS_H2_FRAME_TYPE_COUNT;
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
} else {
h2: post: add states to wait for body
2020-12-19 09:13:24 +00:00
lwsl_info("%s: received %d bytes data for unknown sid %d, highest known %d\n",
__func__, (int)h2n->length, (int)h2n->sid, (int)h2n->highest_sid_opened);
// if (h2n->sid > h2n->highest_sid_opened) {
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
lws_h2_goaway(wsi, H2_ERR_STREAM_CLOSED,
lib: Spelling fixes based on debian packaging
2017-11-28 07:46:22 +08:00
"Data for nonexistent sid");
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
return 0;
h2: post: add states to wait for body
2020-12-19 09:13:24 +00:00
// }
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
/* if the sid is credible, treat as wsi for it closed */
if (h2n->sid > h2n->highest_sid_opened &&
h2n->type != LWS_H2_FRAME_TYPE_HEADERS &&
h2n->type != LWS_H2_FRAME_TYPE_PRIORITY) {
/* if not credible, reject it */
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s, No child for sid %d, rxcmd %d\n",
__func__, lws_wsi_tag(h2n->swsi), (unsigned int)h2n->sid, h2n->type);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_STREAM_CLOSED,
lib: Spelling fixes based on debian packaging
2017-11-28 07:46:22 +08:00
"Data for nonexistent sid");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 0;
}
}
h2: ignore unknown frame types a bit earlier Chrome has started being able to issue frame type 0x42, we drop the connection before we realize we wanted to ignore it. This explicitly ignores it a bit earlier.
2020-11-24 17:02:36 +00:00
if (h2n->swsi && h2n->sid && h2n->type != LWS_H2_FRAME_TYPE_COUNT &&
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
!(http2_rx_validity[h2n->swsi->h2.h2_state] & (1 << h2n->type))) {
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s, State: %s, ILLEGAL cmdrx %d (OK 0x%x)\n",
__func__, lws_wsi_tag(h2n->swsi),
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2_state_names[h2n->swsi->h2.h2_state], h2n->type,
http2_rx_validity[h2n->swsi->h2.h2_state]);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (h2n->swsi->h2.h2_state == LWS_H2_STATE_CLOSED ||
h2n->swsi->h2.h2_state == LWS_H2_STATE_HALF_CLOSED_REMOTE)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
n = H2_ERR_STREAM_CLOSED;
else
n = H2_ERR_PROTOCOL_ERROR;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
lws_h2_goaway(wsi, (unsigned int)n, "invalid rx for state");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 0;
}
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
if (h2n->cont_exp && h2n->type != LWS_H2_FRAME_TYPE_COUNT &&
(h2n->cont_exp_sid != h2n->sid ||
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2n->type != LWS_H2_FRAME_TYPE_CONTINUATION)) {
format strings: fixes for when toolchain has uint32_t as unsigned long
2019-11-05 21:00:45 +00:00
lwsl_info("%s: expected cont on sid %u (got %d on sid %u)\n",
__func__, (unsigned int)h2n->cont_exp_sid, h2n->type,
(unsigned int)h2n->sid);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2n->cont_exp = 0;
if (h2n->cont_exp_headers)
n = H2_ERR_COMPRESSION_ERROR;
else
n = H2_ERR_PROTOCOL_ERROR;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
lws_h2_goaway(wsi, (unsigned int)n, "Continuation hdrs State");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 0;
}
switch (h2n->type) {
case LWS_H2_FRAME_TYPE_DATA:
lwsl_info("seen incoming LWS_H2_FRAME_TYPE_DATA start\n");
if (!h2n->sid) {
travis: h2spec integration
2018-04-26 15:27:02 +08:00
lwsl_info("DATA: 0 sid\n");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR, "DATA 0 sid");
break;
}
format strings: fixes for when toolchain has uint32_t as unsigned long
2019-11-05 21:00:45 +00:00
lwsl_info("Frame header DATA: sid %u, flags 0x%x, len %u\n",
(unsigned int)h2n->sid, h2n->flags,
(unsigned int)h2n->length);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: move window mangament into DATA body handling
2018-02-06 09:58:48 +08:00
if (!h2n->swsi) {
lwsl_notice("DATA: NULL swsi\n");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
}
travis: h2spec integration
2018-04-26 15:27:02 +08:00
lwsl_info("DATA rx on state %d\n", h2n->swsi->h2.h2_state);
refactor: subdirs for source in lib Split out some optional code into own sources to shrink down libwebsockets.c and server.c a bit
2017-10-16 12:52:32 +08:00
if (
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2n->swsi->h2.h2_state == LWS_H2_STATE_HALF_CLOSED_REMOTE ||
h2n->swsi->h2.h2_state == LWS_H2_STATE_CLOSED) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_STREAM_CLOSED, "conn closed");
break;
}
h2 client: deal correctly with zero length DATA with END_STREAM
2019-10-16 15:08:25 +01:00
if (h2n->length == 0)
lws_h2_parse_end_of_frame(wsi);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
h2 client: deal correctly with zero length DATA with END_STREAM
2019-10-16 15:08:25 +01:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_FRAME_TYPE_PRIORITY:
lwsl_info("LWS_H2_FRAME_TYPE_PRIORITY complete frame\n");
if (!h2n->sid) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"Priority has 0 sid");
break;
}
if (h2n->length != 5) {
lws_h2_goaway(wsi, H2_ERR_FRAME_SIZE_ERROR,
"Priority has length other than 5");
break;
}
break;
case LWS_H2_FRAME_TYPE_PUSH_PROMISE:
lwsl_info("LWS_H2_FRAME_TYPE_PUSH_PROMISE complete frame\n");
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR, "Server only");
break;
case LWS_H2_FRAME_TYPE_GOAWAY:
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
lwsl_debug("LWS_H2_FRAME_TYPE_GOAWAY received\n");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
case LWS_H2_FRAME_TYPE_RST_STREAM:
if (!h2n->sid)
return 1;
if (!h2n->swsi) {
if (h2n->sid <= h2n->highest_sid_opened)
http2 padding and ping to autopong Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-27 16:46:44 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"crazy sid on RST_STREAM");
return 1;
}
if (h2n->length != 4) {
lws_h2_goaway(wsi, H2_ERR_FRAME_SIZE_ERROR,
"RST_STREAM can only be length 4");
break;
}
lws_h2_state(h2n->swsi, LWS_H2_STATE_CLOSED);
break;
case LWS_H2_FRAME_TYPE_SETTINGS:
lwsl_info("LWS_H2_FRAME_TYPE_SETTINGS complete frame\n");
/* nonzero sid on settings is illegal */
if (h2n->sid) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"Settings has nonzero sid");
break;
}
if (!(h2n->flags & LWS_H2_FLAG_SETTINGS_ACK)) {
h2: allow empty SETTINGS https://libwebsockets.org/pipermail/libwebsockets/2020-August/008676.html
2020-08-26 11:06:31 +01:00
if (h2n->length % 6) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_FRAME_SIZE_ERROR,
"Settings length error");
http2 padding and ping to autopong Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-27 16:46:44 +08:00
break;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (h2n->type == LWS_H2_FRAME_TYPE_COUNT)
return 0;
travis: h2spec integration
2018-04-26 15:27:02 +08:00
h2: ack SETTINGS just the once
2019-10-02 08:55:12 -07:00
if (wsi->upgraded_to_http2 &&
#if defined(LWS_WITH_CLIENT)
(!(wsi->flags & LCCSCF_H2_QUIRK_NGHTTP2_END_STREAM) ||
#else
(
#endif
!wsi->h2_acked_settings)) {
travis: h2spec integration
2018-04-26 15:27:02 +08:00
pps = lws_h2_new_pps(LWS_H2_PPS_ACK_SETTINGS);
if (!pps)
return 1;
lws_pps_schedule(wsi, pps);
h2: ack SETTINGS just the once
2019-10-02 08:55:12 -07:00
wsi->h2_acked_settings = 1;
travis: h2spec integration
2018-04-26 15:27:02 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
}
/* came to us with ACK set... not allowed to have payload */
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (h2n->length) {
lws_h2_goaway(wsi, H2_ERR_FRAME_SIZE_ERROR,
"Settings with ACK not allowed payload");
break;
}
break;
case LWS_H2_FRAME_TYPE_PING:
if (h2n->sid) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"Ping has nonzero sid");
break;
}
if (h2n->length != 8) {
lws_h2_goaway(wsi, H2_ERR_FRAME_SIZE_ERROR,
"Ping payload can only be 8");
break;
}
break;
case LWS_H2_FRAME_TYPE_CONTINUATION:
http: chunked client
2019-11-12 10:17:36 +00:00
lwsl_info("LWS_H2_FRAME_TYPE_CONTINUATION: sid = %u %d %d\n",
(unsigned int)h2n->sid, (int)h2n->cont_exp,
(int)h2n->cont_exp_sid);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!h2n->cont_exp ||
h2n->cont_exp_sid != h2n->sid ||
!h2n->sid ||
!h2n->swsi) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"unexpected CONTINUATION");
break;
}
http: chunked client
2019-11-12 10:17:36 +00:00
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (h2n->swsi->h2.END_HEADERS) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"END_HEADERS already seen");
break;
}
/* END_STREAM is in HEADERS, skip resetting it */
goto update_end_headers;
case LWS_H2_FRAME_TYPE_HEADERS:
format strings: fixes for when toolchain has uint32_t as unsigned long
2019-11-05 21:00:45 +00:00
lwsl_info("HEADERS: frame header: sid = %u\n",
(unsigned int)h2n->sid);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!h2n->sid) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR, "sid 0");
return 1;
}
http2 hpack basic decode ok including huff Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-12 08:38:16 +08:00
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
if (h2n->swsi && !h2n->swsi->h2.END_STREAM &&
h2n->swsi->h2.END_HEADERS &&
!(h2n->flags & LWS_H2_FLAG_END_STREAM)) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"extra HEADERS together");
travis: h2spec integration
2018-04-26 15:27:02 +08:00
return 1;
}
clean: internally use LWS_WITH_CLIENT and _SERVER Remove some more things in LWS_WITH_SERVER=0 case
2019-08-18 05:04:15 +01:00
#if defined(LWS_WITH_CLIENT)
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (wsi->client_h2_alpn) {
if (h2n->sid) {
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
h2n->swsi = lws_wsi_mux_from_id(wsi, h2n->sid);
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("HEADERS: nwsi %s: sid %u mapped "
"to wsi %s\n", lws_wsi_tag(wsi),
(unsigned int)h2n->sid,
lws_wsi_tag(h2n->swsi));
h2: tx END_STREAM should only set HALF_CLOSED_LOCAL
2018-04-03 12:00:46 +08:00
if (!h2n->swsi)
break;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
}
goto update_end_headers;
}
#endif
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!h2n->swsi) {
/* no more children allowed by parent */
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (wsi->mux.child_count + 1 >
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->h2.h2n->our_set.s[H2SET_MAX_CONCURRENT_STREAMS]) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"Another stream not allowed");
return 1;
http2 hpack basic decode ok including huff Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-12 08:38:16 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
/*
* The peer has sent us a HEADERS implying the creation
* of a new stream
*/
smp: more lock assertions
2021-04-04 04:06:24 +01:00
lws_context_lock(wsi->a.context, "h2 new str");
lws_vhost_lock(wsi->a.vhost);
h2n->swsi = __lws_wsi_server_new(wsi->a.vhost, wsi,
h2n->sid);
lws_vhost_unlock(wsi->a.vhost);
lws_context_unlock(wsi->a.context);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!h2n->swsi) {
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"OOM");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 1;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
h2: explicit credit on connect
2018-02-05 10:54:18 +08:00
h2: take care to retain initial sid
2020-04-06 13:22:01 +01:00
if (h2n->sid >= h2n->highest_sid)
h2n->highest_sid = h2n->sid + 2;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
h2n->swsi->h2.initialized = 1;
h2: explicit credit on connect
2018-02-05 10:54:18 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (lws_h2_update_peer_txcredit(h2n->swsi,
h2n->swsi->mux.my_sid, 4 * 65536))
vhost_destroy: use vhost wsi reference counting to trigger destroy This changes the vhost destroy flow to only hand off the listen socket if another vhost sharing it, and mark the vhost as being_destroyed. Each tsi calls lws_check_deferred_free() once a second, if it sees any vhost being_destroyed there, it closes all wsi on its tsi on the same vhost, one time. As the wsi on the vhost complete close (ie, after libuv async close if on libuv event loop), they decrement a reference count for all wsi open on the vhost. The tsi who closes the last one then completes the destroy flow for the vhost itself... it's random which tsi completes the vhost destroy but since there are no wsi left on the vhost, and it holds the context lock, nothing can conflict. The advantage of this is that owning tsi do the close for wsi that are bound to the vhost under destruction, at a time when they are guaranteed to be idle for service, and they do it with both vhost and context locks owned, so no other service thread can conflict for stuff protected by those either. For the situation the user code may have allocations attached to the vhost, this adds args to lws_vhost_destroy() to allow destroying the user allocations just before the vhost is freed.
2018-06-16 09:31:07 +08:00
goto cleanup_wsi;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
/*
* ah needs attaching to child wsi, even though
* we only fill it from network wsi
*/
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
if (!h2n->swsi->http.ah)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (lws_header_table_attach(h2n->swsi, 0)) {
lwsl_err("%s: Failed to get ah\n", __func__);
return 1;
}
/*
* The first use of a new stream identifier implicitly closes
* all streams in the "idle" state that might have been
* initiated by that peer with a lower-valued stream identifier.
*
* For example, if a client sends a HEADERS frame on stream 7
* without ever sending a frame on stream 5, then stream 5
* transitions to the "closed" state when the first frame for
* stream 7 is sent or received.
*/
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
lws_start_foreach_ll(struct lws *, w, wsi->mux.child_list) {
if (w->mux.my_sid < h2n->sid &&
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
w->h2.h2_state == LWS_H2_STATE_IDLE)
internal close api: add caller id
2018-02-03 13:48:18 +08:00
lws_close_free_wsi(w, 0, "h2 sid close");
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
assert(w->mux.sibling_list != w);
} lws_end_foreach_ll(w, mux.sibling_list);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: make sure we see END_STREAM with END_HEADERS on client
2021-01-15 19:53:55 +00:00
h2n->cont_exp = !(h2n->flags & LWS_H2_FLAG_END_HEADERS);
h2n->cont_exp_sid = h2n->sid;
h2n->cont_exp_headers = 1;
// lws_header_table_reset(h2n->swsi, 0);
update_end_headers:
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
if (lws_check_opt(h2n->swsi->a.vhost->options,
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL)) {
/*
* We don't directly timeout streams that enter the
* half-closed remote state, allowing immortal long
* poll
*/
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
lws_mux_mark_immortal(h2n->swsi);
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: h2 stream entering long poll\n",
__func__, lws_wsi_tag(h2n->swsi));
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
} else {
h2n->swsi->h2.END_STREAM =
refactor: subdirs for source in lib Split out some optional code into own sources to shrink down libwebsockets.c and server.c a bit
2017-10-16 12:52:32 +08:00
!!(h2n->flags & LWS_H2_FLAG_END_STREAM);
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
lwsl_debug("%s: hdr END_STREAM = %d\n",__func__,
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2n->swsi->h2.END_STREAM);
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/* no END_HEADERS means CONTINUATION must come */
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2n->swsi->h2.END_HEADERS =
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
!!(h2n->flags & LWS_H2_FLAG_END_HEADERS);
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: END_HEADERS %d\n", __func__, lws_wsi_tag(h2n->swsi),
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
h2n->swsi->h2.END_HEADERS);
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (h2n->swsi->h2.END_HEADERS)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2n->cont_exp = 0;
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
lwsl_debug("END_HEADERS %d\n", h2n->swsi->h2.END_HEADERS);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
vhost_destroy: use vhost wsi reference counting to trigger destroy This changes the vhost destroy flow to only hand off the listen socket if another vhost sharing it, and mark the vhost as being_destroyed. Each tsi calls lws_check_deferred_free() once a second, if it sees any vhost being_destroyed there, it closes all wsi on its tsi on the same vhost, one time. As the wsi on the vhost complete close (ie, after libuv async close if on libuv event loop), they decrement a reference count for all wsi open on the vhost. The tsi who closes the last one then completes the destroy flow for the vhost itself... it's random which tsi completes the vhost destroy but since there are no wsi left on the vhost, and it holds the context lock, nothing can conflict. The advantage of this is that owning tsi do the close for wsi that are bound to the vhost under destruction, at a time when they are guaranteed to be idle for service, and they do it with both vhost and context locks owned, so no other service thread can conflict for stuff protected by those either. For the situation the user code may have allocations attached to the vhost, this adds args to lws_vhost_destroy() to allow destroying the user allocations just before the vhost is freed.
2018-06-16 09:31:07 +08:00
cleanup_wsi:
return 1;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_FRAME_TYPE_WINDOW_UPDATE:
if (h2n->length != 4) {
lws_h2_goaway(wsi, H2_ERR_FRAME_SIZE_ERROR,
"window update frame not 4");
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
break;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lwsl_info("LWS_H2_FRAME_TYPE_WINDOW_UPDATE\n");
break;
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
case LWS_H2_FRAME_TYPE_COUNT:
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
if (h2n->length == 0)
lws_h2_parse_end_of_frame(wsi);
else
lwsl_debug("%s: going on to deal with unknown frame remaining len %d\n", __func__, (unsigned int)h2n->length);
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
default:
lwsl_info("%s: ILLEGAL FRAME TYPE %d\n", __func__, h2n->type);
h2n->type = LWS_H2_FRAME_TYPE_COUNT; /* ie, IGNORE */
break;
}
if (h2n->length == 0)
h2n->frame_state = 0;
return 0;
}
h2: duplicate :path into ah method-specific header slot for all methods
2019-07-29 00:31:41 +01:00
static const char * const method_names[] = {
rtos diet: http: remove headers at buildtime according to config Headers related to ws or h2 are now elided if the ws or h2 role is not enabled for build. In addition, a new build-time option LWS_WITH_HTTP_UNCOMMON_HEADERS on by default allows removal of less-common http headers to shrink the parser footprint. Minilex is adapted to produce 8 different versions of the lex table, chosen at build-time according to which headers are included in the build. If you don't need the unusual headers, or aren't using h2 or ws, this chops down the size of the ah and the rodata needed to hold the parsing table from 87 strings / pointers to 49, and the parsing table from 1177 to 696 bytes.
2020-02-28 10:31:04 +00:00
"GET", "POST",
#if defined(LWS_WITH_HTTP_UNCOMMON_HEADERS)
"OPTIONS", "PUT", "PATCH", "DELETE",
#endif
"CONNECT", "HEAD"
h2: duplicate :path into ah method-specific header slot for all methods
2019-07-29 00:31:41 +01:00
};
static unsigned char method_index[] = {
WSI_TOKEN_GET_URI,
WSI_TOKEN_POST_URI,
rtos diet: http: remove headers at buildtime according to config Headers related to ws or h2 are now elided if the ws or h2 role is not enabled for build. In addition, a new build-time option LWS_WITH_HTTP_UNCOMMON_HEADERS on by default allows removal of less-common http headers to shrink the parser footprint. Minilex is adapted to produce 8 different versions of the lex table, chosen at build-time according to which headers are included in the build. If you don't need the unusual headers, or aren't using h2 or ws, this chops down the size of the ah and the rodata needed to hold the parsing table from 87 strings / pointers to 49, and the parsing table from 1177 to 696 bytes.
2020-02-28 10:31:04 +00:00
#if defined(LWS_WITH_HTTP_UNCOMMON_HEADERS)
h2: duplicate :path into ah method-specific header slot for all methods
2019-07-29 00:31:41 +01:00
WSI_TOKEN_OPTIONS_URI,
WSI_TOKEN_PUT_URI,
WSI_TOKEN_PATCH_URI,
WSI_TOKEN_DELETE_URI,
rtos diet: http: remove headers at buildtime according to config Headers related to ws or h2 are now elided if the ws or h2 role is not enabled for build. In addition, a new build-time option LWS_WITH_HTTP_UNCOMMON_HEADERS on by default allows removal of less-common http headers to shrink the parser footprint. Minilex is adapted to produce 8 different versions of the lex table, chosen at build-time according to which headers are included in the build. If you don't need the unusual headers, or aren't using h2 or ws, this chops down the size of the ah and the rodata needed to hold the parsing table from 87 strings / pointers to 49, and the parsing table from 1177 to 696 bytes.
2020-02-28 10:31:04 +00:00
#endif
h2: duplicate :path into ah method-specific header slot for all methods
2019-07-29 00:31:41 +01:00
WSI_TOKEN_CONNECT,
WSI_TOKEN_HEAD_URI,
};
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/*
* The last byte of the whole frame has been handled.
* Perform actions for frame completion.
truncated: detect and flag any write on a wsi that could have pending truncated
2017-12-07 07:20:47 +08:00
*
* This is the crunch time for parsing that may have occured on a network
* wsi with a pending partial send... we may call lws_http_action() to send
* a response, conflicting with the partial.
*
* So in that case we change the wsi state and do the lws_http_action() in the
* WRITABLE handler as a priority.
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
*/
static int
lws_h2_parse_end_of_frame(struct lws *wsi)
{
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
struct lws_h2_netconn *h2n = wsi->h2.h2n;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws *eff_wsi = wsi;
const char *p;
int n;
h2n->frame_state = 0;
h2n->count = 0;
if (h2n->sid)
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
h2n->swsi = lws_wsi_mux_from_id(wsi, h2n->sid);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (h2n->sid > h2n->highest_sid)
h2n->highest_sid = h2n->sid;
LWS_WITH_ASAN This lets you build using the runtime Address Sanitizer in gcc. LWS is heavily tested with valgrind routinely during development. But ASAN did find some theoretical-only issues with shifting, strictly ~(1 << 31) is a signed int, it should be ~(1u << 31). Gcc does the same for both, but it's good to have the ability to find these.
2018-11-03 14:47:48 +08:00
if (h2n->collected_priority && (h2n->dep & ~(1u << 31)) == h2n->sid) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR, "depends on own sid");
return 0;
}
switch (h2n->type) {
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
case LWS_H2_FRAME_TYPE_SETTINGS:
clean: internally use LWS_WITH_CLIENT and _SERVER Remove some more things in LWS_WITH_SERVER=0 case
2019-08-18 05:04:15 +01:00
#if defined(LWS_WITH_CLIENT)
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (wsi->client_h2_alpn && !wsi->client_mux_migrated &&
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
!(h2n->flags & LWS_H2_FLAG_SETTINGS_ACK)) {
codacy: minor fixes
2018-10-10 13:54:43 +08:00
struct lws_h2_protocol_send *pps;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
/* migrate original client ask on to substream 1 */
freertos: rename esp32 plat to freertos
2019-08-18 10:35:43 +01:00
#if defined(LWS_WITH_FILE_OPS)
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
wsi->http.fop_fd = NULL;
freertos: rename esp32 plat to freertos
2019-08-18 10:35:43 +01:00
#endif
h2: tolerate unexpected server behaviour Make sure we only do one SETTINGS ack handling Fix corner case of ignored header with 0 data being the last in the HEADERS
2019-10-10 16:37:38 +01:00
lwsl_info("%s: migrating\n", __func__);
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
wsi->client_mux_migrated = 1;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
/*
* we need to treat the headers from the upgrade as the
* first job. So these need to get shifted to sid 1.
*/
smp: more lock assertions
2021-04-04 04:06:24 +01:00
lws_context_lock(wsi->a.context, "h2 mig");
lws_vhost_lock(wsi->a.vhost);
h2n->swsi = __lws_wsi_server_new(wsi->a.vhost, wsi, 1);
lws_vhost_unlock(wsi->a.vhost);
lws_context_unlock(wsi->a.context);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (!h2n->swsi)
return 1;
h2n->sid = 1;
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
assert(lws_wsi_mux_from_id(wsi, 1) == h2n->swsi);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
// lws_role_transition(wsi, LWSIFR_CLIENT,
// LRS_H2_WAITING_TO_SEND_HEADERS,
// &role_ops_h2);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
lws_role_transition(h2n->swsi, LWSIFR_CLIENT,
LRS_H2_WAITING_TO_SEND_HEADERS,
&role_ops_h2);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
/* pass on the initial headers to SID 1 */
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
h2n->swsi->http.ah = wsi->http.ah;
Fault injection add lws_xos: xoshiro256 PRNG
2021-03-16 13:32:05 +00:00
#if defined(LWS_WITH_SYS_FAULT_INJECTION)
lws_fi_import(&h2n->swsi->fic, &wsi->fic);
#endif
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
h2n->swsi->client_mux_substream = 1;
h2: client: do not apply host header if h2 negotiated by alpn If client negotiated over alpn, it will have done SNI. Some peers like youtube.com reject with 400 if you sent a host: as well.
2020-05-05 12:17:27 +01:00
h2n->swsi->client_h2_alpn = 1;
h2: end stream with end headers This shouldn't be necessary; just END_HEADERS flag should be enough. But nghttp2 will not talk to us unless we end the stream from our side. Unfortunately ending the stream at the time we sent the headers means we cannot support the long poll half-close scheme. So add a quirk flag to optionally support this behaviour of nghttp2 when the client is creating the connection.
2019-10-02 10:49:18 -07:00
#if defined(LWS_WITH_CLIENT)
h2n->swsi->flags = wsi->flags;
lws_conmon: connection monitoring and stats generation This provides a build option LWS_WITH_CONMON that lets user code recover detailed connection stats on client connections with the LCCSCF_CONMON flag. In addition to latencies for dns, socket connection, tls and first protocol response where possible, it also provides the user code an unfiltered list of DNS responses that the client received, and the peer it actually succeded to connect to.
2021-03-15 05:20:15 +00:00
#if defined(LWS_WITH_CONMON)
/* sid1 needs to represent the connection experience
* ... we take over responsibility for the DNS list
* copy as well
*/
h2n->swsi->conmon = wsi->conmon;
h2n->swsi->conmon_datum = wsi->conmon_datum;
h2n->swsi->sa46_peer = wsi->sa46_peer;
wsi->conmon.dns_results_copy = NULL;
h2: end stream with end headers This shouldn't be necessary; just END_HEADERS flag should be enough. But nghttp2 will not talk to us unless we end the stream from our side. Unfortunately ending the stream at the time we sent the headers means we cannot support the long poll half-close scheme. So add a quirk flag to optionally support this behaviour of nghttp2 when the client is creating the connection.
2019-10-02 10:49:18 -07:00
#endif
lws_conmon: connection monitoring and stats generation This provides a build option LWS_WITH_CONMON that lets user code recover detailed connection stats on client connections with the LCCSCF_CONMON flag. In addition to latencies for dns, socket connection, tls and first protocol response where possible, it also provides the user code an unfiltered list of DNS responses that the client received, and the peer it actually succeded to connect to.
2021-03-15 05:20:15 +00:00
#endif /* CLIENT */
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
h2: migration: migrate for_ss The attribute indicating that a wsi belongs to an SS object also must be migrated when we split out the original transaction wsi into a new nwsi, if we're not going to lose track of its affiliation. Likewise if the affiliated SS object points to the original wsi, we have to migrate his pointer when we migrate the wsi. Taking care of this book-keeping is necessary to get correct behaviours at close-time.
2021-05-07 10:01:59 +01:00
#if defined(LWS_WITH_SECURE_STREAMS)
if (wsi->for_ss) {
lws_ss_handle_t *h = (lws_ss_handle_t *)lws_get_opaque_user_data(wsi);
h2n->swsi->for_ss = 1;
wsi->for_ss = 0;
if (h->wsi == wsi)
h->wsi = h2n->swsi;
}
#endif
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
h2n->swsi->a.protocol = wsi->a.protocol;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (h2n->swsi->user_space &&
!h2n->swsi->user_space_externally_allocated)
h2 client: dont leak original user_space for network wsi When creating the stream from the nwsi, the stream was created with its own user_space that gets overwritten with the nwsi one as it is demoted to be the stream. Stop that leaking.
2019-08-08 21:15:57 +01:00
lws_free(h2n->swsi->user_space);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
h2n->swsi->user_space = wsi->user_space;
h2n->swsi->user_space_externally_allocated =
wsi->user_space_externally_allocated;
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
h2n->swsi->a.opaque_user_data = wsi->a.opaque_user_data;
wsi->a.opaque_user_data = NULL;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
h2n->swsi->txc.manual_initial_tx_credit =
wsi->txc.manual_initial_tx_credit;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
#if defined(LWS_WITH_TLS)
lws_strncpy(h2n->swsi->alpn, wsi->alpn,
sizeof(wsi->alpn));
#endif
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
wsi->user_space = NULL;
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
if (h2n->swsi->http.ah)
h2n->swsi->http.ah->wsi = h2n->swsi;
wsi->http.ah = NULL;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: MIGRATING nwsi %s -> swsi %s\n", __func__,
lws_wsi_tag(wsi), lws_wsi_tag(h2n->swsi));
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
h2n->swsi->txc.tx_cr = (int32_t)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
h2n->peer_set.s[H2SET_INITIAL_WINDOW_SIZE];
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: initial tx credit on %s: %d\n",
__func__, lws_wsi_tag(h2n->swsi),
(int)h2n->swsi->txc.tx_cr);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
h2n->swsi->h2.initialized = 1;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
/* set our initial window size */
if (!wsi->h2.initialized) {
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
wsi->txc.tx_cr = (int32_t)
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
h2n->peer_set.s[H2SET_INITIAL_WINDOW_SIZE];
cleaning
2020-01-02 08:32:23 +00:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
lwsl_info("%s: initial tx credit for us to "
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
"write on nwsi %s: %d\n", __func__,
lws_wsi_tag(wsi), (int)wsi->txc.tx_cr);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->h2.initialized = 1;
}
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lws_callback_on_writable(h2n->swsi);
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (!wsi->h2_acked_settings ||
!(wsi->flags & LCCSCF_H2_QUIRK_NGHTTP2_END_STREAM)
h2: ack SETTINGS just the once
2019-10-02 08:55:12 -07:00
) {
pps = lws_h2_new_pps(LWS_H2_PPS_ACK_SETTINGS);
if (!pps)
return 1;
lws_pps_schedule(wsi, pps);
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
lwsl_info("%s: SETTINGS ack PPS\n", __func__);
h2: ack SETTINGS just the once
2019-10-02 08:55:12 -07:00
wsi->h2_acked_settings = 1;
}
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
/* also attach any queued guys */
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
lws_wsi_mux_apply_queue(wsi);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
}
#endif
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_FRAME_TYPE_CONTINUATION:
case LWS_H2_FRAME_TYPE_HEADERS:
h2: tx END_STREAM should only set HALF_CLOSED_LOCAL
2018-04-03 12:00:46 +08:00
if (!h2n->swsi)
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/* service the http request itself */
if (h2n->last_action_dyntable_resize) {
lws_h2_goaway(wsi, H2_ERR_COMPRESSION_ERROR,
"dyntable resize last in headers");
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (!h2n->swsi->h2.END_HEADERS) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/* we are not finished yet */
lwsl_info("witholding http action for continuation\n");
http: chunked client
2019-11-12 10:17:36 +00:00
h2n->cont_exp_sid = h2n->sid;
h2n->cont_exp = 1;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
/* confirm the hpack stream state is reasonable for finishing */
if (h2n->hpack != HPKS_TYPE) {
/* hpack incomplete */
format strings: fixes for when toolchain has uint32_t as unsigned long
2019-11-05 21:00:45 +00:00
lwsl_info("hpack incomplete %d (type %d, len %u)\n",
h2n->hpack, h2n->type,
(unsigned int)h2n->hpack_len);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_COMPRESSION_ERROR,
"hpack incomplete");
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
/* this is the last part of HEADERS */
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
switch (h2n->swsi->h2.h2_state) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_STATE_IDLE:
lws_h2_state(h2n->swsi, LWS_H2_STATE_OPEN);
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_STATE_RESERVED_REMOTE:
lws_h2_state(h2n->swsi, LWS_H2_STATE_HALF_CLOSED_LOCAL);
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
break;
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("http req, %s, h2n->swsi=%s\n", lws_wsi_tag(wsi),
lws_wsi_tag(h2n->swsi));
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2n->swsi->hdr_parsing_completed = 1;
clean: internally use LWS_WITH_CLIENT and _SERVER Remove some more things in LWS_WITH_SERVER=0 case
2019-08-18 05:04:15 +01:00
#if defined(LWS_WITH_CLIENT)
cleaning
2020-01-02 08:32:23 +00:00
if (h2n->swsi->client_mux_substream &&
lws_client_interpret_server_handshake(h2n->swsi)) {
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
/*
* This is more complicated than it looks, one exit from
* interpret_server_handshake() is to do a close that
* turns into a redirect.
*
* In that case, the wsi survives having being reset
* and detached from any h2 identity. We need to get
* our parents out from touching it any more
*/
lwsl_info("%s: cli int serv hs closed, or redir\n", __func__);
return 2;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
}
lws_diskcache: split generic parts from gitohashi into lws
2018-11-12 15:24:42 +08:00
#endif
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (lws_hdr_extant(h2n->swsi, WSI_TOKEN_HTTP_CONTENT_LENGTH)) {
coverity: 62535: another simple_ptr NULL check after confirmed by other means
2020-08-18 12:57:58 +01:00
const char *simp = lws_hdr_simple_ptr(h2n->swsi,
WSI_TOKEN_HTTP_CONTENT_LENGTH);
if (!simp) /* coverity */
return 1;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
h2n->swsi->http.rx_content_length = (unsigned long long)atoll(simp);
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2n->swsi->http.rx_content_remain =
h2n->swsi->http.rx_content_length;
h2: post buflist: track rx_content_length On h2 server POST, there's a race to see if the POST body is going to be received coalesced with the headers. The problem is on h2, we can't action the stream http request or body until the stream is writeable, since we may start issuing the response right away; there's already DEFERRING_ACTION state to manage this. And indeed, the coalesced, not-immediately-actionable POST body is buflisted properly. However when we come to action the POST using buflisted data, we don't follow the same pattern as dealing with the incoming data immediately. This patch aligns the pattern dumping the buflist content to track expected rx_content_length and handle BODY_COMPLETION if we got to the end of it, along with removal from the pt list of wsi with pending buflists if we used it up.
2021-01-18 14:20:37 +00:00
h2n->swsi->http.content_length_given = 1;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lwsl_info("setting rx_content_length %lld\n",
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
(long long)h2n->swsi->http.rx_content_length);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
{
int n = 0, len;
char buf[256];
const unsigned char *c;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
do {
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
c = lws_token_to_string((enum lws_token_indexes)n);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!c) {
n++;
continue;
}
http2 handle flags Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-17 08:38:44 +08:00
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
len = lws_hdr_total_length(h2n->swsi, (enum lws_token_indexes)n);
build: enable signed vs unsigned warnings on gcc This enables selected things from -Wextra, can't use -Wextra because it is fussy enough to complain about unused params on functions... they are there for a reason. -Wsign-compare -Wignored-qualifiers not -Wimplicit-fallthrough=3 ... only on gcc 7 -Wtype-limits -Wuninitialized not -Wclobbered ... only on gcc 7ish fix the warnings everywhere they were found.
2017-10-20 17:45:02 +08:00
if (!len || len > (int)sizeof(buf) - 1) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
n++;
continue;
http2 update integration https://github.com/warmcat/libwebsockets/issues/489 This 1) fixes the vhost changes on master 2) works around the ah pool changes 3) fixes some other build problems that appeared 4) hacks out physical flow control for internal streams 5) updates the advertised protocol to h2 needed by, eg, chrome 51 That gets it able to serve small (<4K, ie, one packet) files over http2 Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-10 09:33:54 +08:00
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
if (lws_hdr_copy(h2n->swsi, buf, sizeof buf,
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
(enum lws_token_indexes)n) < 0) {
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lwsl_info(" %s !oversize!\n",
(char *)c);
coverity 192910: check length in h2 header dump Since the NUL was always taken care of at the end of buf, the worst it could do was log some junk. This makes it explicitly log it as oversize separately.
2018-11-13 16:53:41 +08:00
} else {
buf[sizeof(buf) - 1] = '\0';
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lwsl_info(" %s = %s\n",
(char *)c, buf);
coverity 192910: check length in h2 header dump Since the NUL was always taken care of at the end of buf, the worst it could do was log some junk. This makes it explicitly log it as oversize separately.
2018-11-13 16:53:41 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
n++;
} while (c);
}
http2 track content length add END_STREAM Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-18 12:23:05 +08:00
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (h2n->swsi->h2.h2_state == LWS_H2_STATE_HALF_CLOSED_REMOTE ||
h2n->swsi->h2.h2_state == LWS_H2_STATE_CLOSED) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_STREAM_CLOSED,
"Banning service on CLOSED_REMOTE");
break;
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
switch (h2n->swsi->h2.h2_state) {
h2: take care to retain initial sid
2020-04-06 13:22:01 +01:00
case LWS_H2_STATE_IDLE:
lws_h2_state(h2n->swsi, LWS_H2_STATE_OPEN);
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_STATE_OPEN:
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (h2n->swsi->h2.END_STREAM)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_state(h2n->swsi,
LWS_H2_STATE_HALF_CLOSED_REMOTE);
break;
case LWS_H2_STATE_HALF_CLOSED_LOCAL:
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (h2n->swsi->h2.END_STREAM)
h2: client: action END_STREAM and END_HEADERS immediately
2021-01-23 05:46:56 +00:00
/*
* action the END_STREAM
*/
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_state(h2n->swsi, LWS_H2_STATE_CLOSED);
break;
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
clean: internally use LWS_WITH_CLIENT and _SERVER Remove some more things in LWS_WITH_SERVER=0 case
2019-08-18 05:04:15 +01:00
#if defined(LWS_WITH_CLIENT)
h2: client: handle END_STREAM and END_HEADERS togther on all paths
2021-01-28 19:32:11 +00:00
/*
* If we already had the END_STREAM along with the END_HEADERS,
* we have already transitioned to STATE_CLOSED and we are not
* going to be doing anything further on this stream.
*
* In that case handle the transaction completion and
* finalize the stream for the peer
*/
if (h2n->swsi->h2.h2_state == LWS_H2_STATE_CLOSED &&
h2n->swsi->client_mux_substream) {
lws_h2_rst_stream(h2n->swsi, H2_ERR_NO_ERROR,
"client done");
if (lws_http_transaction_completed_client(h2n->swsi))
lwsl_debug("tx completed returned close\n");
break;
}
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (h2n->swsi->client_mux_substream) {
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: headers: client path (h2 state %s)\n",
__func__, lws_wsi_tag(wsi),
h2_state_names[h2n->swsi->h2.h2_state]);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
break;
}
lws_diskcache: split generic parts from gitohashi into lws
2018-11-12 15:24:42 +08:00
#endif
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!lws_hdr_total_length(h2n->swsi, WSI_TOKEN_HTTP_COLON_PATH) ||
!lws_hdr_total_length(h2n->swsi, WSI_TOKEN_HTTP_COLON_METHOD) ||
!lws_hdr_total_length(h2n->swsi, WSI_TOKEN_HTTP_COLON_SCHEME) ||
lws_hdr_total_length(h2n->swsi, WSI_TOKEN_HTTP_COLON_STATUS) ||
lws_hdr_extant(h2n->swsi, WSI_TOKEN_CONNECTION)) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"Pseudoheader checks");
break;
}
http2 track content length add END_STREAM Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-18 12:23:05 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (lws_hdr_extant(h2n->swsi, WSI_TOKEN_TE)) {
n = lws_hdr_total_length(h2n->swsi, WSI_TOKEN_TE);
if (n != 8 ||
coverity: 62535: another simple_ptr NULL check after confirmed by other means
2020-08-18 12:57:58 +01:00
!lws_hdr_simple_ptr(h2n->swsi, WSI_TOKEN_TE) ||
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
strncmp(lws_hdr_simple_ptr(h2n->swsi, WSI_TOKEN_TE),
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
"trailers", (unsigned int)n)) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"Illegal transfer-encoding");
http2 window update Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-29 09:39:08 +08:00
break;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
threadpool
2018-09-02 14:35:37 +08:00
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
lws_http_compression_validate(h2n->swsi);
#endif
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
p = lws_hdr_simple_ptr(h2n->swsi, WSI_TOKEN_HTTP_COLON_METHOD);
h2: duplicate :path into ah method-specific header slot for all methods
2019-07-29 00:31:41 +01:00
/*
* duplicate :path into the individual method uri header
* index, so that it looks the same as h1 in the ah
*/
for (n = 0; n < (int)LWS_ARRAY_SIZE(method_names); n++)
coverity: 62535: another simple_ptr NULL check after confirmed by other means
2020-08-18 12:57:58 +01:00
if (p && !strcasecmp(p, method_names[n])) {
h2: duplicate :path into ah method-specific header slot for all methods
2019-07-29 00:31:41 +01:00
h2n->swsi->http.ah->frag_index[method_index[n]] =
h2n->swsi->http.ah->frag_index[
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
WSI_TOKEN_HTTP_COLON_PATH];
h2: duplicate :path into ah method-specific header slot for all methods
2019-07-29 00:31:41 +01:00
break;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: client: action END_STREAM and END_HEADERS immediately
2021-01-23 05:46:56 +00:00
{
lwsl_debug("%s: setting DEF_ACT from 0x%x\n", __func__,
(unsigned int)h2n->swsi->wsistate);
lwsi_set_state(h2n->swsi, LRS_DEFERRING_ACTION);
lws_callback_on_writable(h2n->swsi);
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
case LWS_H2_FRAME_TYPE_DATA:
cleaning
2020-01-02 08:32:23 +00:00
lwsl_info("%s: DATA flags 0x%x\n", __func__, h2n->flags);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!h2n->swsi)
break;
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
if (lws_hdr_total_length(h2n->swsi,
WSI_TOKEN_HTTP_CONTENT_LENGTH) &&
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
h2n->swsi->h2.END_STREAM &&
h2n->swsi->http.rx_content_length &&
h2n->swsi->http.rx_content_remain) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_rst_stream(h2n->swsi, H2_ERR_PROTOCOL_ERROR,
"Not enough rx content");
break;
}
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (h2n->swsi->h2.END_STREAM &&
h2n->swsi->h2.h2_state == LWS_H2_STATE_OPEN)
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lws_h2_state(h2n->swsi,
LWS_H2_STATE_HALF_CLOSED_REMOTE);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
if (h2n->swsi->h2.END_STREAM &&
h2n->swsi->h2.h2_state == LWS_H2_STATE_HALF_CLOSED_LOCAL)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_h2_state(h2n->swsi, LWS_H2_STATE_CLOSED);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
clean: internally use LWS_WITH_CLIENT and _SERVER Remove some more things in LWS_WITH_SERVER=0 case
2019-08-18 05:04:15 +01:00
#if defined(LWS_WITH_CLIENT)
LWS_TLS_CERT_INFO_OPAQUE_PUBLIC_KEY
2018-04-05 20:48:08 +08:00
/*
* client... remote END_STREAM implies we weren't going to
* send anything else anyway.
*/
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (h2n->swsi->client_mux_substream &&
h2 client: deal correctly with zero length DATA with END_STREAM
2019-10-16 15:08:25 +01:00
(h2n->flags & LWS_H2_FLAG_END_STREAM)) {
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: DATA: end stream\n",
__func__, lws_wsi_tag(h2n->swsi));
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
if (h2n->swsi->h2.h2_state == LWS_H2_STATE_OPEN) {
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lws_h2_state(h2n->swsi,
LWS_H2_STATE_HALF_CLOSED_REMOTE);
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
// lws_h2_rst_stream(h2n->swsi, H2_ERR_NO_ERROR,
// "client done");
// if (lws_http_transaction_completed_client(h2n->swsi))
// lwsl_debug("tx completed returned close\n");
}
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
LWS_TLS_CERT_INFO_OPAQUE_PUBLIC_KEY
2018-04-05 20:48:08 +08:00
//if (h2n->swsi->h2.h2_state == LWS_H2_STATE_HALF_CLOSED_LOCAL)
{
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lws_h2_state(h2n->swsi, LWS_H2_STATE_CLOSED);
lws_h2_rst_stream(h2n->swsi, H2_ERR_NO_ERROR,
"client done");
if (lws_http_transaction_completed_client(h2n->swsi))
lwsl_debug("tx completed returned close\n");
}
}
lws_diskcache: split generic parts from gitohashi into lws
2018-11-12 15:24:42 +08:00
#endif
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
case LWS_H2_FRAME_TYPE_PING:
cleaning
2020-01-02 08:32:23 +00:00
if (h2n->flags & LWS_H2_FLAG_SETTINGS_ACK)
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
lws_validity_confirmed(wsi);
cleaning
2020-01-02 08:32:23 +00:00
else {
/* they're sending us a ping request */
codacy: minor fixes
2018-10-10 13:54:43 +08:00
struct lws_h2_protocol_send *pps =
lws_h2_new_pps(LWS_H2_PPS_PONG);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!pps)
return 1;
codacy: minor fixes
2018-10-10 13:54:43 +08:00
lwsl_info("rx ping, preparing pong\n");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
memcpy(pps->u.ping.ping_payload, h2n->ping_payload, 8);
lws_pps_schedule(wsi, pps);
}
break;
case LWS_H2_FRAME_TYPE_WINDOW_UPDATE:
cleaning
2020-01-02 08:32:23 +00:00
/*
* We only have an unsigned 31-bit (positive) increment possible
*/
LWS_WITH_ASAN This lets you build using the runtime Address Sanitizer in gcc. LWS is heavily tested with valgrind routinely during development. But ASAN did find some theoretical-only issues with shifting, strictly ~(1 << 31) is a signed int, it should be ~(1u << 31). Gcc does the same for both, but it's good to have the ability to find these.
2018-11-03 14:47:48 +08:00
h2n->hpack_e_dep &= ~(1u << 31);
format strings: fixes for when toolchain has uint32_t as unsigned long
2019-11-05 21:00:45 +00:00
lwsl_info("WINDOW_UPDATE: sid %u %u (0x%x)\n",
(unsigned int)h2n->sid,
(unsigned int)h2n->hpack_e_dep,
(unsigned int)h2n->hpack_e_dep);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (h2n->sid)
eff_wsi = h2n->swsi;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!eff_wsi) {
if (h2n->sid > h2n->highest_sid_opened)
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"alien sid");
break; /* ignore */
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
if (eff_wsi->a.vhost->options &
h2-tx-credit-disallow-overflow
2019-07-08 08:48:58 +01:00
LWS_SERVER_OPTION_H2_JUST_FIX_WINDOW_UPDATE_OVERFLOW &&
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
(uint64_t)eff_wsi->txc.tx_cr + (uint64_t)h2n->hpack_e_dep >
h2-tx-credit-disallow-overflow
2019-07-08 08:48:58 +01:00
(uint64_t)0x7fffffff)
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
h2n->hpack_e_dep = (uint32_t)(0x7fffffff - eff_wsi->txc.tx_cr);
h2-tx-credit-disallow-overflow
2019-07-08 08:48:58 +01:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if ((uint64_t)eff_wsi->txc.tx_cr + (uint64_t)h2n->hpack_e_dep >
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
(uint64_t)0x7fffffff) {
minimal-http-client: set option to fix broken server window update
2021-02-26 11:38:45 +00:00
lwsl_warn("%s: WINDOW_UPDATE 0x%llx + 0x%llx = 0x%llx, too high\n",
__func__, (unsigned long long)eff_wsi->txc.tx_cr,
(unsigned long long)h2n->hpack_e_dep,
(unsigned long long)eff_wsi->txc.tx_cr + (unsigned long long)h2n->hpack_e_dep);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (h2n->sid)
lws_h2_rst_stream(h2n->swsi,
H2_ERR_FLOW_CONTROL_ERROR,
"Flow control exceeded max");
else
lws_h2_goaway(wsi, H2_ERR_FLOW_CONTROL_ERROR,
"Flow control exceeded max");
break;
}
if (!h2n->hpack_e_dep) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"Zero length window update");
break;
}
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
n = eff_wsi->txc.tx_cr;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
eff_wsi->txc.tx_cr += (int32_t)h2n->hpack_e_dep;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
lws_wsi_txc_report_manual_txcr_in(eff_wsi,
(int32_t)h2n->hpack_e_dep);
lws_wsi_txc_describe(&eff_wsi->txc, "WINDOW_UPDATE in",
eff_wsi->mux.my_sid);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (n <= 0 && eff_wsi->txc.tx_cr <= 0)
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
/* it helps, but won't change sendability for anyone */
break;
/*
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
* It may have changed sendability (depends on SID 0 tx credit
* too)... for us and any children waiting on us... reassess
* blockage for all children first
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
*/
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
lws_start_foreach_ll(struct lws *, w, wsi->mux.child_list) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_callback_on_writable(w);
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
} lws_end_foreach_ll(w, mux.sibling_list);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (eff_wsi->txc.skint &&
!lws_wsi_txc_check_skint(&eff_wsi->txc,
lws_h2_tx_cr_get(eff_wsi)))
/*
* This one became un-skint, schedule a writeable
* callback
*/
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
lws_callback_on_writable(eff_wsi);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
case LWS_H2_FRAME_TYPE_GOAWAY:
cleaning
2020-01-02 08:32:23 +00:00
lwsl_notice("GOAWAY: last sid %u, error 0x%08X, string '%s'\n",
rtos: h2 in toolchain with long uint32_t
2019-10-17 11:08:47 +01:00
(unsigned int)h2n->goaway_last_sid,
(unsigned int)h2n->goaway_err, h2n->goaway_str);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
return 1;
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
case LWS_H2_FRAME_TYPE_RST_STREAM:
format strings: fixes for when toolchain has uint32_t as unsigned long
2019-11-05 21:00:45 +00:00
lwsl_info("LWS_H2_FRAME_TYPE_RST_STREAM: sid %u: reason 0x%x\n",
cleaning
2020-01-02 08:32:23 +00:00
(unsigned int)h2n->sid,
(unsigned int)h2n->hpack_e_dep);
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
case LWS_H2_FRAME_TYPE_COUNT: /* IGNORING FRAME */
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
break;
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
return 0;
}
truncated: detect and flag any write on a wsi that could have pending truncated
2017-12-07 07:20:47 +08:00
/*
* This may want to send something on the network wsi, which may be in the
* middle of a partial send. PPS sends are OK because they are queued to
* go through the WRITABLE handler already.
*
* The read parser for the network wsi has no choice but to parse its stream
* anyway, because otherwise it will not be able to get tx credit window
* messages.
*
* Therefore if we will send non-PPS, ie, lws_http_action() for a stream
* wsi, we must change its state and handle it as a priority in the
* POLLOUT handler instead of writing it here.
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
*
* About closing... for the main network wsi, it should return nonzero to
* close it all. If it needs to close an swsi, it can do it here.
truncated: detect and flag any write on a wsi that could have pending truncated
2017-12-07 07:20:47 +08:00
*/
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
int
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
lws_h2_parser(struct lws *wsi, unsigned char *in, lws_filepos_t _inlen,
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
lws_filepos_t *inused)
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
{
refactor-eliminate-lws-union The union used to make a lot of sense to save space between mutually exclusive modes. But the fact the http2 struct contains the http1 struct as well as it appearing in the union means the http1 struct belongs outside the union. This patch - eliminates the union - puts the http_related struct directly in struct lws - removes http_related from h2 - puts h2 directly in struct lws if enabled for build - changes ws to be a pointer, allocated if we upgrade to ws (the ws part contains a 135 byte char array for ping / close) Again all of this is entirely private / internal and doesn't affect any apis.
2017-12-01 11:09:32 +08:00
struct lws_h2_netconn *h2n = wsi->h2.h2n;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
struct lws_h2_protocol_send *pps;
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
unsigned char c, *oldin = in, *iend = in + (size_t)_inlen;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
int n, m;
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
if (!h2n)
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
goto fail;
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
while (in < iend) {
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
c = *in++;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
switch (lwsi_state(wsi)) {
case LRS_H2_AWAIT_PREFACE:
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (preface[h2n->count++] != c)
goto fail;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (preface[h2n->count])
break;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("http2: %s: established\n", lws_wsi_tag(wsi));
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_set_state(wsi, LRS_H2_AWAIT_SETTINGS);
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
lws_validity_confirmed(wsi);
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
h2n->count = 0;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
wsi->txc.tx_cr = 65535;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
/*
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
* we must send a settings frame -- empty one is OK...
* that must be the first thing sent by server
* and the peer must send a SETTINGS with ACK flag...
http2 working nghttp fetch multi files completely Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-18 18:54:04 +08:00
*/
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
pps = lws_h2_new_pps(LWS_H2_PPS_MY_SETTINGS);
if (!pps)
goto fail;
lws_pps_schedule(wsi, pps);
break;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
case LRS_H2_WAITING_TO_SEND_HEADERS:
case LRS_ESTABLISHED:
case LRS_H2_AWAIT_SETTINGS:
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (h2n->frame_state != LWS_H2_FRAME_HEADER_LENGTH)
goto try_frame_start;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
/*
* post-header, preamble / payload / padding part
*/
h2n->count++;
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
if (h2n->type == LWS_H2_FRAME_TYPE_COUNT) { /* IGNORING FRAME */
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
//lwsl_debug("%s: consuming for ignored %u %u\n", __func__, (unsigned int)h2n->count, (unsigned int)h2n->length);
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
goto frame_end;
}
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
if (h2n->flags & LWS_H2_FLAG_PADDED &&
!h2n->pad_length) {
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
/*
* Get the padding count... actual padding is
* at the end of the frame.
*/
h2n->padding = c;
h2n->pad_length = 1;
h2n->preamble++;
if (h2n->padding > h2n->length - 1)
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lws_h2_goaway(wsi,
H2_ERR_PROTOCOL_ERROR,
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
"execssive padding");
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break; /* we consumed this */
}
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (h2n->flags & LWS_H2_FLAG_PRIORITY &&
!h2n->collected_priority) {
/* going to be 5 preamble bytes */
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
lwsl_debug("PRIORITY FLAG: 0x%x\n", c);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (h2n->preamble++ - h2n->pad_length < 4) {
h2n->dep = ((h2n->dep) << 8) | c;
break; /* we consumed this */
}
h2n->weight_temp = c;
h2n->collected_priority = 1;
lwsl_debug("PRI FL: dep 0x%x, weight 0x%02X\n",
cleaning
2020-01-02 08:32:23 +00:00
(unsigned int)h2n->dep,
h2n->weight_temp);
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
break; /* we consumed this */
}
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
if (h2n->padding && h2n->count >
(h2n->length - h2n->padding)) {
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (c) {
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
"nonzero padding");
break;
}
goto frame_end;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
/* applies to wsi->h2.swsi which may be wsi */
switch(h2n->type) {
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_SETTINGS:
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
n = (int)(h2n->count - 1u - h2n->preamble) %
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
LWS_H2_SETTINGS_LEN;
h2n->one_setting[n] = c;
if (n != LWS_H2_SETTINGS_LEN - 1)
break;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
lws_h2_settings(wsi, &h2n->peer_set,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
h2n->one_setting,
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
LWS_H2_SETTINGS_LEN);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_CONTINUATION:
case LWS_H2_FRAME_TYPE_HEADERS:
if (!h2n->swsi)
break;
if (lws_hpack_interpret(h2n->swsi, c)) {
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lwsl_info("%s: hpack failed\n",
__func__);
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
goto fail;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_GOAWAY:
switch (h2n->inside++) {
case 0:
case 1:
case 2:
case 3:
h2n->goaway_last_sid <<= 8;
h2n->goaway_last_sid |= c;
h2n->goaway_str[0] = '\0';
break;
h2: extend network stream timeout during body processing
2018-02-03 13:53:45 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case 4:
case 5:
case 6:
case 7:
h2n->goaway_err <<= 8;
h2n->goaway_err |= c;
break;
h2: extend network stream timeout during body processing
2018-02-03 13:53:45 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
default:
if (h2n->inside - 9 <
sizeof(h2n->goaway_str) - 1)
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
h2n->goaway_str[
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
h2n->inside - 9] = (char)c;
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
h2n->goaway_str[
sizeof(h2n->goaway_str) - 1] = '\0';
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
break;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_DATA:
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
h2: stay a client if a client
2020-12-17 22:16:57 +00:00
// lwsl_info("%s: LWS_H2_FRAME_TYPE_DATA: fl 0x%x\n",
// __func__, h2n->flags);
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
/*
* let the network wsi live a bit longer if
* subs are active... our frame may take a long
* time to chew through
*/
lws_http_mark_sse https://github.com/warmcat/libwebsockets/issues/1486
2019-01-30 20:59:56 +08:00
if (!wsi->immortal_substream_count)
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lws_set_timeout(wsi,
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
PENDING_TIMEOUT_HTTP_KEEPALIVE_IDLE,
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
wsi->a.vhost->keepalive_timeout ?
wsi->a.vhost->keepalive_timeout : 31);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (!h2n->swsi)
break;
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
if (lws_buflist_next_segment_len(
&h2n->swsi->buflist, NULL))
lwsl_info("%s: substream has pending\n",
__func__);
refactor: absorb other caches into buflist 1) Remove the whole ah rxbuf and put things on to the wsi buflist This eliminates the whole detachability thing based on ah rxbuf state... ah can always be detached. 2) Remove h2 scratch and put it on the wsi buflist 3) Remove preamble_rx and use the wsi buflist This was used in the case adopted sockets had already been read. Basically there are now only three forced service scenarios - something in buflist (and not in state LRS_DEFERRING_ACTION) - tls layer has buffered rx - extension has buffered rx This is a net removal of around 400 lines of special-casing.
2018-04-17 15:35:15 +08:00
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
if (lwsi_role_http(h2n->swsi) &&
lwsi_state(h2n->swsi) == LRS_ESTABLISHED) {
lwsi_set_state(h2n->swsi, LRS_BODY);
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s to LRS_BODY\n",
__func__, lws_wsi_tag(h2n->swsi));
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
}
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
h2: data: frame header already parsed out
2021-08-09 14:14:06 +01:00
/*
* in + length may cover multiple frames, we
* can only consider the length of the DATA
* in front of us
*/
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (lws_hdr_total_length(h2n->swsi,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
WSI_TOKEN_HTTP_CONTENT_LENGTH) &&
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
h2n->swsi->http.rx_content_length &&
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
h2n->swsi->http.rx_content_remain <
h2: data: frame header already parsed out
2021-08-09 14:14:06 +01:00
h2n->length && /* last */
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
h2n->inside < h2n->length) {
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
ss: support huge urls
2021-03-02 16:34:33 +00:00
lwsl_warn("%s: %lu %lu %lu %lu\n", __func__,
(unsigned long)h2n->swsi->http.rx_content_remain,
h2: data: frame header already parsed out
2021-08-09 14:14:06 +01:00
(unsigned long)(lws_ptr_diff_size_t(iend, in) + 1),
ss: support huge urls
2021-03-02 16:34:33 +00:00
(unsigned long)h2n->inside, (unsigned long)h2n->length);
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
/* unread data in frame */
lws_h2_goaway(wsi,
H2_ERR_PROTOCOL_ERROR,
"More rx than content_length told");
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
break;
}
h2: move window mangament into DATA body handling
2018-02-06 09:58:48 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
/*
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
* We operate on a frame. The RX we have at
* hand may exceed the current frame.
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
*/
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
n = (int)lws_ptr_diff_size_t(iend, in) + 1;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (n > (int)(h2n->length - h2n->count + 1)) {
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
if (h2n->count > h2n->length)
goto close_swsi_and_return;
n = (int)(h2n->length - h2n->count) + 1;
cleaning
2020-01-02 08:32:23 +00:00
lwsl_debug("---- restricting len to %d "
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
"\n", n);
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
}
clean: internally use LWS_WITH_CLIENT and _SERVER Remove some more things in LWS_WITH_SERVER=0 case
2019-08-18 05:04:15 +01:00
#if defined(LWS_WITH_CLIENT)
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (h2n->swsi->client_mux_substream) {
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
if (!h2n->swsi->a.protocol) {
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_err("%s: %p doesn't have protocol\n",
__func__, lws_wsi_tag(h2n->swsi));
client: improve redirect
2019-08-26 18:41:40 +01:00
m = 1;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
} else {
h2n->swsi->txc.peer_tx_cr_est -= n;
wsi->txc.peer_tx_cr_est -= n;
lws_wsi_txc_describe(&h2n->swsi->txc,
__func__,
h2n->swsi->mux.my_sid);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
m = user_callback_handle_rxflow(
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
h2n->swsi->a.protocol->callback,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
h2n->swsi,
LWS_CALLBACK_RECEIVE_CLIENT_HTTP_READ,
h2n->swsi->user_space,
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
in - 1, (unsigned int)n);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
}
client: improve redirect
2019-08-26 18:41:40 +01:00
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
in += n - 1;
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
h2n->inside += (unsigned int)n;
h2n->count += (unsigned int)n - 1;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (m) {
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lwsl_info("RECEIVE_CLIENT_HTTP "
"closed it\n");
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
goto close_swsi_and_return;
}
h2: bulk rx: deal with window We don't always keep updating the rxflow window on bulk rx, and just run down the initial. This keeps us going with 64K in flight
2021-08-30 14:18:14 +01:00
goto do_windows;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
}
lws_diskcache: split generic parts from gitohashi into lws
2018-11-12 15:24:42 +08:00
#endif
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (lwsi_state(h2n->swsi) == LRS_DEFERRING_ACTION) {
m = lws_buflist_append_segment(
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
&h2n->swsi->buflist, in - 1, (unsigned int)n);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
if (m < 0)
return -1;
h2: post buflist: track rx_content_length On h2 server POST, there's a race to see if the POST body is going to be received coalesced with the headers. The problem is on h2, we can't action the stream http request or body until the stream is writeable, since we may start issuing the response right away; there's already DEFERRING_ACTION state to manage this. And indeed, the coalesced, not-immediately-actionable POST body is buflisted properly. However when we come to action the POST using buflisted data, we don't follow the same pattern as dealing with the incoming data immediately. This patch aligns the pattern dumping the buflist content to track expected rx_content_length and handle BODY_COMPLETION if we got to the end of it, along with removal from the pt list of wsi with pending buflists if we used it up.
2021-01-18 14:20:37 +00:00
/*
* Since we're in an open-ended
* DEFERRING_ACTION, don't add this swsi
* to the pt list of wsi holding buflist
* content yet, we are not in a position
* to consume it until we get out of
* DEFERRING_ACTION.
*/
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
in += n - 1;
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
h2n->inside += (unsigned int)n;
h2n->count += (unsigned int)n - 1;
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
lwsl_debug("%s: deferred %d\n", __func__, n);
goto do_windows;
}
h2n->swsi->outer_will_close = 1;
/*
* choose the length for this go so that we end at
* the frame boundary, in the case there is already
* more waiting leave it for next time around
*/
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
n = lws_read_h1(h2n->swsi, in - 1, (unsigned int)n);
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
// lwsl_notice("%s: lws_read_h1 %d\n", __func__, n);
h2n->swsi->outer_will_close = 0;
/*
* can return 0 in POST body with
* content len exhausted somehow.
*/
if (n < 0 ||
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
(!n && h2n->swsi->http.content_length_given && !lws_buflist_next_segment_len(
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
&wsi->buflist, NULL))) {
lwsl_info("%s: lws_read_h1 told %d %u / %u\n",
__func__, n,
(unsigned int)h2n->count,
(unsigned int)h2n->length);
in += h2n->length - h2n->count;
h2n->inside = h2n->length;
h2n->count = h2n->length - 1;
//if (n < 0)
// goto already_closed_swsi;
goto close_swsi_and_return;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
}
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
lwsl_info("%s: lws_read_h1 telling %d %u / %u\n",
__func__, n,
(unsigned int)h2n->count,
(unsigned int)h2n->length);
coverity 472861: http2: deal with zero length data without overflow
2025-01-16 11:40:26 +00:00
if (n) {
in += (unsigned int)n - 1;
h2n->inside += (unsigned int)n;
h2n->count += (unsigned int)n - 1;
h2: remove length tracking just use pointer arithmetic
2020-12-20 08:34:23 +00:00
coverity 472861: http2: deal with zero length data without overflow
2025-01-16 11:40:26 +00:00
h2n->swsi->txc.peer_tx_cr_est -= n;
wsi->txc.peer_tx_cr_est -= n;
}
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
rxflow buflist: handle forced service
2018-04-17 11:43:20 +08:00
do_windows:
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
#if defined(LWS_WITH_CLIENT)
if (!(h2n->swsi->flags & LCCSCF_H2_MANUAL_RXFLOW))
#endif
{
/*
* The default behaviour is we just keep
* cranking the other side's tx credit
* back up, for simple bulk transfer as
* fast as we can take it
*/
h2: bulk rx: deal with window We don't always keep updating the rxflow window on bulk rx, and just run down the initial. This keeps us going with 64K in flight
2021-08-30 14:18:14 +01:00
m = n + 65536;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
/* update both the stream and nwsi */
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
lws_h2_update_peer_txcredit_thresh(h2n->swsi,
h2n->sid, m, m);
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
}
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
#if defined(LWS_WITH_CLIENT)
else {
/*
* If he's handling it himself, only
* repair the nwsi credit but allow the
* stream credit to run down until the
* user code deals with it
*/
lws_h2_update_peer_txcredit(wsi, 0, n);
h2n->swsi->txc.manual = 1;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
}
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
#endif
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
break;
h2: move window mangament into DATA body handling
2018-02-06 09:58:48 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_PRIORITY:
if (h2n->count <= 4) {
h2n->dep <<= 8;
h2n->dep |= c;
cleaning
2020-01-02 08:32:23 +00:00
break;
}
h2n->weight_temp = c;
lwsl_info("PRIORITY: dep 0x%x, weight 0x%02X\n",
(unsigned int)h2n->dep, h2n->weight_temp);
if ((h2n->dep & ~(1u << 31)) == h2n->sid) {
lws_h2_goaway(wsi, H2_ERR_PROTOCOL_ERROR,
"cant depend on own sid");
break;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
}
break;
esp32: align with esp-idf changes LWIP_SOCKET_OFFSET is now nonzero, which I handled a while back. But the C api support for it is broken in esp-idf. select() takes unmodified fds, but FD_SET / FD_ISSET etc must have the offset subtracted on their args. With this we are working on current HEAD esp-idf.
2017-11-24 10:54:59 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_RST_STREAM:
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
h2n->hpack_e_dep <<= 8;
h2n->hpack_e_dep |= c;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_PUSH_PROMISE:
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_PING:
if (h2n->flags & LWS_H2_FLAG_SETTINGS_ACK) { // ack
} else { /* they're sending us a ping request */
if (h2n->count > 8)
return 1;
h2n->ping_payload[h2n->count - 1] = c;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_WINDOW_UPDATE:
h2n->hpack_e_dep <<= 8;
h2n->hpack_e_dep |= c;
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case LWS_H2_FRAME_TYPE_COUNT: /* IGNORING FRAME */
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
//lwsl_debug("%s: consuming for ignored %u %u\n", __func__, (unsigned int)h2n->count, (unsigned int)h2n->length);
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
h2n->count++;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
default:
lwsl_notice("%s: unhandled frame type %d\n",
__func__, h2n->type);
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
goto fail;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
frame_end:
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
if (h2n->count > h2n->length) {
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
lwsl_notice("%s: count > length %u %u (type %d)\n",
rtos: h2 in toolchain with long uint32_t
2019-10-17 11:08:47 +01:00
__func__, (unsigned int)h2n->count,
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
(unsigned int)h2n->length, h2n->type);
} else
if (h2n->count != h2n->length)
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
/*
* end of frame just happened
*/
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
n = lws_h2_parse_end_of_frame(wsi);
if (n == 2) {
*inused = (lws_filepos_t)lws_ptr_diff_size_t(in, oldin);
return 2;
}
if (n)
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
goto fail;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
break;
try_frame_start:
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
if (h2n->frame_state <= 8) {
switch (h2n->frame_state++) {
case 0:
h2n->pad_length = 0;
h2n->collected_priority = 0;
h2n->padding = 0;
h2n->preamble = 0;
h2n->length = c;
h2n->inside = 0;
break;
case 1:
case 2:
h2n->length <<= 8;
h2n->length |= c;
break;
case 3:
h2n->type = c;
break;
case 4:
h2n->flags = c;
break;
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
case 5:
case 6:
case 7:
case 8:
h2n->sid <<= 8;
h2n->sid |= c;
break;
}
http2: make usable HTTP/2 support is now able to serve the test server, complete with websockets, from a single vhost. - This works the same with both OpenSSL and mbedTLS. - POST is now wired up and works (also for file upload). - CGI is wired up and works. - Redirect is adapted and works - lwsws works. - URI urldecode, sanitation and argument parsing wired up for :path valgrind clean (aside from openssl-style false uninit data usage in mbedtls send occasionally) h2spec reports: $ h2spec -h 127.0.0.1 -p 7681 -t -k -o 1 ... 145 tests, 145 passed, 0 skipped, 0 failed" Incorporates: - "https://github.com/warmcat/libwebsockets/pull/1039 Fixes issue with -Werror=unused-variable flag - https://github.com/sartura/libwebsockets/commit/2c843a1395bdca5c23b40d989f508fc96aa5ba3b ssl: fix infinite loop on client cert verification failure Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>" Caused and fixes Coverity 184887 - 184892
2017-10-13 10:33:02 +08:00
}
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
if (h2n->frame_state == LWS_H2_FRAME_HEADER_LENGTH &&
lws_h2_parse_frame_header(wsi))
goto fail;
h2: POST: restrict block handling to end of frame
2018-02-19 05:31:14 +08:00
break;
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
default:
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
if (h2n->type == LWS_H2_FRAME_TYPE_COUNT) { /* IGNORING FRAME */
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
//lwsl_debug("%s: consuming for ignored %u %u\n", __func__, (unsigned int)h2n->count, (unsigned int)h2n->length);
h2: fix ignoring unknown frame payloads
2020-12-01 13:11:28 +00:00
h2n->count++;
}
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
break;
}
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
}
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*inused = (lws_filepos_t)lws_ptr_diff_size_t(in, oldin);
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
return 0;
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
close_swsi_and_return:
lws_close_free_wsi(h2n->swsi, 0, "close_swsi_and_return");
h2n->swsi = NULL;
h2n->frame_state = 0;
h2n->count = 0;
cmake: allow setting LWS_ROLE_WS You can build lws without support for ws, with -DLWS_ROLE_WS=0. This is thanks to the role ops isolating all the ws-specific business in the ws role. Also retire more test apps replaced by minmal-examples.
2018-04-25 08:42:18 +08:00
// already_closed_swsi:
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*inused = (lws_filepos_t)lws_ptr_diff_size_t(in, oldin);
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
return 2;
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
fail:
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
*inused = (lws_filepos_t)lws_ptr_diff_size_t(in, oldin);
h2: provide parser buffer at a time
2017-11-25 10:10:33 +08:00
return 1;
http2 alpn npn pollout This adds npn / alpn support if your openssl can handle it. Then, browsers that understand alpn will by default negotiate http/1.1 and work as normal. Clients that understand http2.0 can negotiate h2-14 and use the basic but working http2.0 support automatically Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-22 15:37:28 +08:00
}
client h2: auth bearer option flag This affects max header size since we use the latter half of the pt_serv_buf to prepare the (possibly huge) auth token. Adapt the pt_serv_buf_size in the hugeurl example.
2019-10-03 08:31:34 -07:00
#if defined(LWS_WITH_CLIENT)
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
int
lws_h2_client_handshake(struct lws *wsi)
{
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
struct lws_context_per_thread *pt = &wsi->a.context->pt[(int)wsi->tsi];
h2 client: deal correctly with zero length DATA with END_STREAM
2019-10-16 15:08:25 +01:00
uint8_t *buf, *start, *p, *p1, *end;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
char *meth = lws_hdr_simple_ptr(wsi, _WSI_TOKEN_CLIENT_METHOD),
coverity: 62123: explicitly check things to guide coverity Coverity sees we sometimes check if header length is 0 and extrapolates from that not checking header length return is a violation. But often we are OK if the header length is 0 and there is no error return to check from that. It also doesn't understand that if we saw a nonzero length for a header, then we are going to get a non-null simple_ptr() return for sure. Just give up and explicitly, unneccessarily check everything so coverity can stop telling us about it.
2020-08-19 07:11:59 +01:00
*uri = lws_hdr_simple_ptr(wsi, _WSI_TOKEN_CLIENT_URI), *simp;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
struct lws *nwsi = lws_get_network_wsi(wsi);
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
const char *path = "/";
h2: end stream with end headers This shouldn't be necessary; just END_HEADERS flag should be enough. But nghttp2 will not talk to us unless we end the stream from our side. Unfortunately ending the stream at the time we sent the headers means we cannot support the long poll half-close scheme. So add a quirk flag to optionally support this behaviour of nghttp2 when the client is creating the connection.
2019-10-02 10:49:18 -07:00
int n, m;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
/*
* The identifier of a newly established stream MUST be numerically
* greater than all streams that the initiating endpoint has opened or
* reserved. This governs streams that are opened using a HEADERS frame
* and streams that are reserved using PUSH_PROMISE. An endpoint that
* receives an unexpected stream identifier MUST respond with a
* connection error (Section 5.4.1) of type PROTOCOL_ERROR.
*/
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
unsigned int sid = nwsi->h2.h2n->highest_sid_opened + 2;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
cleaning
2020-01-02 08:32:23 +00:00
lwsl_debug("%s\n", __func__);
h2: client sid: must be allocated at header send There is no way to allocate a client wsi mux sid before the headers are send, because we don't know the order in which new wsi headers will be sent and so seen by the peer. The peer inisists that sid indexes only increase... we cannot allocat sids monotonically at the client and then send them disordered...
2020-05-04 08:26:48 +01:00
/*
* We MUST allocate our sid here at the point we're about to send the
* stream open. It's because we don't know the order in which multiple
* open streams will send their headers... in h2, sending the headers
* is the point the stream is opened. The peer requires that we only
* open streams in ascending sid order
*/
wsi->mux.my_sid = nwsi->h2.h2n->highest_sid_opened = sid;
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_info("%s: %s: assigning SID %d at header send\n", __func__,
lws_wsi_tag(wsi), sid);
h2: client sid: must be allocated at header send There is no way to allocate a client wsi mux sid before the headers are send, because we don't know the order in which new wsi headers will be sent and so seen by the peer. The peer inisists that sid indexes only increase... we cannot allocat sids monotonically at the client and then send them disordered...
2020-05-04 08:26:48 +01:00
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lwsl_info("%s: CLIENT_WAITING_TO_SEND_HEADERS: pollout (sid %d)\n",
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
__func__, wsi->mux.my_sid);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
p = start = buf = pt->serv_buf + LWS_PRE;
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
end = start + (wsi->a.context->pt_serv_buf_size / 2) - LWS_PRE - 1;
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
/* it's time for us to send our client stream headers */
if (!meth)
meth = "GET";
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
/* h2 pseudoheaders must be in a bunch at the start */
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_COLON_METHOD,
(unsigned char *)meth,
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
(int)strlen(meth), &p, end))
goto fail_length;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (lws_add_http_header_by_token(wsi,
WSI_TOKEN_HTTP_COLON_SCHEME,
h2: end stream with end headers This shouldn't be necessary; just END_HEADERS flag should be enough. But nghttp2 will not talk to us unless we end the stream from our side. Unfortunately ending the stream at the time we sent the headers means we cannot support the long poll half-close scheme. So add a quirk flag to optionally support this behaviour of nghttp2 when the client is creating the connection.
2019-10-02 10:49:18 -07:00
(unsigned char *)"https", 5,
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
&p, end))
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
goto fail_length;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
coverity: 62123: explicitly check things to guide coverity Coverity sees we sometimes check if header length is 0 and extrapolates from that not checking header length return is a violation. But often we are OK if the header length is 0 and there is no error return to check from that. It also doesn't understand that if we saw a nonzero length for a header, then we are going to get a non-null simple_ptr() return for sure. Just give up and explicitly, unneccessarily check everything so coverity can stop telling us about it.
2020-08-19 07:11:59 +01:00
n = lws_hdr_total_length(wsi, _WSI_TOKEN_CLIENT_URI);
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
if (n)
path = uri;
else
if (wsi->stash && wsi->stash->cis[CIS_PATH]) {
path = wsi->stash->cis[CIS_PATH];
n = (int)strlen(path);
} else
n = 1;
if (n > 1 && path[0] == '/' && path[1] == '/') {
path++;
n--;
}
coverity: 62123: explicitly check things to guide coverity Coverity sees we sometimes check if header length is 0 and extrapolates from that not checking header length return is a violation. But often we are OK if the header length is 0 and there is no error return to check from that. It also doesn't understand that if we saw a nonzero length for a header, then we are going to get a non-null simple_ptr() return for sure. Just give up and explicitly, unneccessarily check everything so coverity can stop telling us about it.
2020-08-19 07:11:59 +01:00
if (n && lws_add_http_header_by_token(wsi,
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
WSI_TOKEN_HTTP_COLON_PATH,
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
(unsigned char *)path, n, &p, end))
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
goto fail_length;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
n = lws_hdr_total_length(wsi, _WSI_TOKEN_CLIENT_HOST);
simp = lws_hdr_simple_ptr(wsi, _WSI_TOKEN_CLIENT_HOST);
if (!n && wsi->stash && wsi->stash->cis[CIS_ADDRESS]) {
n = (int)strlen(wsi->stash->cis[CIS_ADDRESS]);
simp = wsi->stash->cis[CIS_ADDRESS];
}
// n = lws_hdr_total_length(wsi, _WSI_TOKEN_CLIENT_ORIGIN);
// simp = lws_hdr_simple_ptr(wsi, _WSI_TOKEN_CLIENT_ORIGIN);
#if 0
coverity: 62123: explicitly check things to guide coverity Coverity sees we sometimes check if header length is 0 and extrapolates from that not checking header length return is a violation. But often we are OK if the header length is 0 and there is no error return to check from that. It also doesn't understand that if we saw a nonzero length for a header, then we are going to get a non-null simple_ptr() return for sure. Just give up and explicitly, unneccessarily check everything so coverity can stop telling us about it.
2020-08-19 07:11:59 +01:00
if (n && simp && lws_add_http_header_by_token(wsi,
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
WSI_TOKEN_HTTP_COLON_AUTHORITY,
coverity: 62123: explicitly check things to guide coverity Coverity sees we sometimes check if header length is 0 and extrapolates from that not checking header length return is a violation. But often we are OK if the header length is 0 and there is no error return to check from that. It also doesn't understand that if we saw a nonzero length for a header, then we are going to get a non-null simple_ptr() return for sure. Just give up and explicitly, unneccessarily check everything so coverity can stop telling us about it.
2020-08-19 07:11:59 +01:00
(unsigned char *)simp, n, &p, end))
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
goto fail_length;
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
#endif
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
coverity: 62123: explicitly check things to guide coverity Coverity sees we sometimes check if header length is 0 and extrapolates from that not checking header length return is a violation. But often we are OK if the header length is 0 and there is no error return to check from that. It also doesn't understand that if we saw a nonzero length for a header, then we are going to get a non-null simple_ptr() return for sure. Just give up and explicitly, unneccessarily check everything so coverity can stop telling us about it.
2020-08-19 07:11:59 +01:00
http: redo redirects to follow close flow Until now although we can follow redirects, and they can promote the protocol from h1->h2, we couldn't handle h2 wsi reuse since there are many states in the wsi affected by being h2. This wipes the related states in lws_wsi_reset() and follows the generic wsi close flow before deviating into the redirect really close to the end, ensuring we cleaned out evidence of our previous life properly. h2->h2 redirects work properly after this. The max number of redirects is increased from 3 -> 4 since this was seen in the wild with www and then geographic-based redirects.
2021-06-17 10:07:04 +01:00
if (/*!wsi->client_h2_alpn && */n && simp &&
h2: client: do not apply host header if h2 negotiated by alpn If client negotiated over alpn, it will have done SNI. Some peers like youtube.com reject with 400 if you sent a host: as well.
2020-05-05 12:17:27 +01:00
lws_add_http_header_by_token(wsi, WSI_TOKEN_HOST,
coverity: 62123: explicitly check things to guide coverity Coverity sees we sometimes check if header length is 0 and extrapolates from that not checking header length return is a violation. But often we are OK if the header length is 0 and there is no error return to check from that. It also doesn't understand that if we saw a nonzero length for a header, then we are going to get a non-null simple_ptr() return for sure. Just give up and explicitly, unneccessarily check everything so coverity can stop telling us about it.
2020-08-19 07:11:59 +01:00
(unsigned char *)simp, n, &p, end))
h2: end stream with end headers This shouldn't be necessary; just END_HEADERS flag should be enough. But nghttp2 will not talk to us unless we end the stream from our side. Unfortunately ending the stream at the time we sent the headers means we cannot support the long poll half-close scheme. So add a quirk flag to optionally support this behaviour of nghttp2 when the client is creating the connection.
2019-10-02 10:49:18 -07:00
goto fail_length;
client: multipart mime generation helpers lws has been able to generate client multipart mime as shown in minimal-http-client-post, but it requires a lot of user boilerplate to handle the boundary, related transaction header, and multipart headers. This patch adds a client creation flag to indicate it will carry multipart mime, which autocreates the boundary string and applies the transaction header with it, and an api to form the boundary headers between the different mime parts and the terminating boundary.
2019-10-07 14:28:18 +01:00
if (wsi->flags & LCCSCF_HTTP_MULTIPART_MIME) {
p1 = lws_http_multipart_headers(wsi, p);
if (!p1)
goto fail_length;
p = p1;
}
wsi flags: extra encodings
2019-10-09 19:11:20 +01:00
if (wsi->flags & LCCSCF_HTTP_X_WWW_FORM_URLENCODED) {
if (lws_add_http_header_by_token(wsi, WSI_TOKEN_HTTP_CONTENT_TYPE,
cleaning
2020-01-02 08:32:23 +00:00
(unsigned char *)"application/x-www-form-urlencoded",
33, &p, end))
wsi flags: extra encodings
2019-10-09 19:11:20 +01:00
goto fail_length;
lws_client_http_body_pending(wsi, 1);
}
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
/* give userland a chance to append, eg, cookies */
http: cookies: support cookie jar in and out
2021-07-05 16:41:41 +08:00
#if defined(LWS_WITH_CACHE_NSCOOKIEJAR) && defined(LWS_WITH_CLIENT)
if (wsi->flags & LCCSCF_CACHE_COOKIES)
lws_cookie_send_cookies(wsi, (char **)&p, (char *)end);
#endif
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
if (wsi->a.protocol->callback(wsi,
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
LWS_CALLBACK_CLIENT_APPEND_HANDSHAKE_HEADER,
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
wsi->user_space, &p, lws_ptr_diff_size_t(end, p) - 12))
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
goto fail_length;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
if (lws_finalize_http_header(wsi, &p, end))
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
goto fail_length;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
h2: end stream with end headers This shouldn't be necessary; just END_HEADERS flag should be enough. But nghttp2 will not talk to us unless we end the stream from our side. Unfortunately ending the stream at the time we sent the headers means we cannot support the long poll half-close scheme. So add a quirk flag to optionally support this behaviour of nghttp2 when the client is creating the connection.
2019-10-02 10:49:18 -07:00
m = LWS_WRITE_HTTP_HEADERS;
#if defined(LWS_WITH_CLIENT)
/* below is not needed in spec, indeed it destroys the long poll
* feature, but required by nghttp2 */
minimal-http-client-multi: add POST This adds support for POST in both h1 and h2 queues / stream binding. The previous queueing tried to keep the "leader" wsi who made the actual connection around and have it act on the transaction queue tail if it had done its own thing. This refactors it so instead, who is the "leader" moves down the queue and the queued guys inherit the fd, SSL * and queue from the old leader as they take over. This lets them operate in their own wsi identity directly and gets rid of all the "effective wsi" checks, which was applied incompletely and getting out of hand considering the separate lws_mux checks for h2 and other muxed protocols alongside it. This change also allows one wsi at a time to own the transaction for POST. --post is added as an option to lws-minimal-http-client-multi and 6 extra selftests with POST on h1/h2, pipelined or not and staggered or not are added to the CI.
2020-01-30 13:19:11 +00:00
if ((wsi->flags & LCCSCF_H2_QUIRK_NGHTTP2_END_STREAM) &&
http: post: defer body pending end until no more buffered out
2021-03-23 07:14:26 +00:00
!(wsi->client_http_body_pending || lws_has_buffered_out(wsi)))
h2: end stream with end headers This shouldn't be necessary; just END_HEADERS flag should be enough. But nghttp2 will not talk to us unless we end the stream from our side. Unfortunately ending the stream at the time we sent the headers means we cannot support the long poll half-close scheme. So add a quirk flag to optionally support this behaviour of nghttp2 when the client is creating the connection.
2019-10-02 10:49:18 -07:00
m |= LWS_WRITE_H2_STREAM_END;
#endif
cleaning
2020-01-02 08:32:23 +00:00
// lwsl_hexdump_notice(start, p - start);
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
n = lws_write(wsi, start, lws_ptr_diff_size_t(p, start), (enum lws_write_protocol)m);
h2: end stream with end headers This shouldn't be necessary; just END_HEADERS flag should be enough. But nghttp2 will not talk to us unless we end the stream from our side. Unfortunately ending the stream at the time we sent the headers means we cannot support the long poll half-close scheme. So add a quirk flag to optionally support this behaviour of nghttp2 when the client is creating the connection.
2019-10-02 10:49:18 -07:00
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
if (n != lws_ptr_diff(p, start)) {
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lwsl_err("_write returned %d from %ld\n", n,
(long)(p - start));
return -1;
}
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
/*
* Normally let's charge up the peer tx credit a bit. But if
* MANUAL_REFLOW is set, just set it to the initial credit given in
* the client create info
*/
n = 4 * 65536;
if (wsi->flags & LCCSCF_H2_MANUAL_RXFLOW) {
n = wsi->txc.manual_initial_tx_credit;
wsi->txc.manual = 1;
}
h2: take care to retain initial sid
2020-04-06 13:22:01 +01:00
if (lws_h2_update_peer_txcredit(wsi, wsi->mux.my_sid, n))
h2: LCCSCF_H2_MANUAL_RXFLOW and refactor txcr This changes the approach of tx credit management to set the initial stream tx credit window to zero. This is the only way with RFC7540 to gain the ability to selectively precisely rx flow control incoming streams. At the time the headers are sent, a WINDOW_UPDATE is sent with the initial tx credit towards us for that specific stream. By default, this acts as before with a 256KB window added for both the stream and the nwsi, and additional window management sent as stuff is received. It's now also possible to set a member in the client info struct and a new option LCCSCF_H2_MANUAL_RXFLOW to precisely manage both the initial tx credit for a specific stream and the ongoing rate limit by meting out further tx credit manually. Add another minimal example http-client-h2-rxflow demonstrating how to force a connection's peer's initial budget to transmit to us and control it during the connection lifetime to restrict the amount of incoming data we have to buffer.
2019-12-26 02:35:41 +00:00
return 1;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
lws_h2_state(wsi, LWS_H2_STATE_OPEN);
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_set_state(wsi, LRS_ESTABLISHED);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
cleaning
2020-01-02 08:32:23 +00:00
if (wsi->flags & LCCSCF_HTTP_MULTIPART_MIME)
lws_callback_on_writable(wsi);
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
return 0;
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
fail_length:
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lwsl_err("Client hdrs too long: incr context info.pt_serv_buf_size\n");
selftests This converts several of the selftests to return a status in their exit code about whether they 'worked'. A small bash script framework is added, with a selftest.sh in the mininmal example dirs that support it, and a ./minimal-examples/selftests.sh script that can be run from the build dir with no args that discovers and runs all the selftest.sh scripts underneath. That is also integrated into travis and the enabled tests must pass now for travis to pass. Travis does not have a modern libuv so it can't run a couple of tests which are nulled out if it sees it's running in travis env.
2018-04-16 19:52:28 +08:00
return -1;
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
}
client h2: auth bearer option flag This affects max header size since we use the latter half of the pt_serv_buf to prepare the (possibly huge) auth token. Adapt the pt_serv_buf_size in the hugeurl example.
2019-10-03 08:31:34 -07:00
#endif
client: h2 This adds h2 http support for the client api. The public client api requires no changes, it will detect by ALPN if the server can handle http/2, if so, it will use it. Multiple client connections using the lws api will be mapped on to the same single http/2 + tls socket using http/2 streams that are serviced simultaneously where possible.
2018-03-27 09:17:19 +08:00
h2: cmake: build with NO_SERVER
2020-04-16 06:53:10 +01:00
#if defined(LWS_ROLE_WS) && defined(LWS_WITH_SERVER)
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
int
lws_h2_ws_handshake(struct lws *wsi)
{
ws-over-h2: unbreak
2018-10-02 05:42:56 +08:00
uint8_t buf[LWS_PRE + 2048], *p = buf + LWS_PRE, *start = p,
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
*end = &buf[sizeof(buf) - 1];
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
const struct lws_http_mount *hit;
const char * uri_ptr;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
size_t m;
int n;
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
if (lws_add_http_header_status(wsi, HTTP_STATUS_OK, &p, end))
return -1;
if (lws_hdr_total_length(wsi, WSI_TOKEN_PROTOCOL) > 64)
return -1;
ws proxy: also proxy h1 ws to h1 and h2 lws has been able to proxy h2 or h1 inbound connections to an h1 onward connection for a while now. It's simple to use just build with LWS_WITH_HTTP_PROXY and make a mount where the origin is the onward connection details. Unix sockets can also be used as the onward connection. This patch extends the support to be able to also do the same for inbound h2 or h1 ws upgrades to an h1 ws onward connection as well. This allows you to offer completely different services in a common URL space, including ones that connect back by ws / wss.
2019-03-19 11:53:57 +08:00
if (wsi->proxied_ws_parent && wsi->child_list) {
if (lws_hdr_simple_ptr(wsi, WSI_TOKEN_PROTOCOL)) {
if (lws_add_http_header_by_token(wsi, WSI_TOKEN_PROTOCOL,
(uint8_t *)lws_hdr_simple_ptr(wsi,
WSI_TOKEN_PROTOCOL),
fix integer conversions from 64 bit
2019-10-27 12:55:13 +00:00
(int)strlen(lws_hdr_simple_ptr(wsi,
ws proxy: also proxy h1 ws to h1 and h2 lws has been able to proxy h2 or h1 inbound connections to an h1 onward connection for a while now. It's simple to use just build with LWS_WITH_HTTP_PROXY and make a mount where the origin is the onward connection details. Unix sockets can also be used as the onward connection. This patch extends the support to be able to also do the same for inbound h2 or h1 ws upgrades to an h1 ws onward connection as well. This allows you to offer completely different services in a common URL space, including ones that connect back by ws / wss.
2019-03-19 11:53:57 +08:00
WSI_TOKEN_PROTOCOL)),
&p, end))
return -1;
}
} else {
/* we can only return the protocol header if:
* - one came in, and ... */
if (lws_hdr_total_length(wsi, WSI_TOKEN_PROTOCOL) &&
/* - it is not an empty string */
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
wsi->a.protocol->name && wsi->a.protocol->name[0]) {
ss: server: h1, h2, ws basic support Add initial support for defining servers using Secure Streams policy and api semantics. Serving h1, h2 and ws should be functional, the new minimal example shows a combined http + SS server with an incrementing ws message shown in the browser over tls, in around 200 lines of user code. NOP out anything to do with plugins, they're not currently used. Update the docs correspondingly.
2020-07-27 10:03:12 +01:00
#if defined(LWS_WITH_SECURE_STREAMS) && defined(LWS_WITH_SERVER)
/*
* This is the h2 version of server-ws.c understanding that it
* did the ws upgrade on a ss server object, therefore it needs
* to pass back to the peer the policy ws-protocol name, not
* the generic ss-ws.c protocol name
*/
ss: server: check more carefully if built but not used on a wsi
2020-08-08 07:24:13 +01:00
if (wsi->a.vhost && wsi->a.vhost->ss_handle &&
wsi->a.vhost->ss_handle->policy->u.http.u.ws.subprotocol) {
ss: server: h1, h2, ws basic support Add initial support for defining servers using Secure Streams policy and api semantics. Serving h1, h2 and ws should be functional, the new minimal example shows a combined http + SS server with an incrementing ws message shown in the browser over tls, in around 200 lines of user code. NOP out anything to do with plugins, they're not currently used. Update the docs correspondingly.
2020-07-27 10:03:12 +01:00
lws_ss_handle_t *h =
(lws_ss_handle_t *)wsi->a.opaque_user_data;
lws_lifecycle This adds some new objects and helpers for keeping and logging info on grouped allocations, a group is, eg, SS handles or client wsis. Allocated objects get a context-unique "tag" string intended to replace %p / wsi pointers etc. Pointers quickly become confusing when allocations are freed and reused, the tag string won't repeat until you produce 2^64 objects in a context. In addition the tag string documents the object group, with prefixes like "wsi-" or "vh-" and contain object-specific additional information like the vhost name, address / port or the role of the wsi. At creation time the lws code can use a format string and args to add whatever group-specific info makes sense, eg, a wsi bound to a secure stream can also append the guid of the secure stream, it's copied into the new object tag and so is still available cleanly after the stream is destroyed if the wsi outlives it.
2020-12-25 05:54:19 +00:00
lwsl_notice("%s: Server SS %s .wsi %s switching to ws protocol\n",
__func__, lws_ss_tag(h), lws_wsi_tag(h->wsi));
ss: server: h1, h2, ws basic support Add initial support for defining servers using Secure Streams policy and api semantics. Serving h1, h2 and ws should be functional, the new minimal example shows a combined http + SS server with an incrementing ws message shown in the browser over tls, in around 200 lines of user code. NOP out anything to do with plugins, they're not currently used. Update the docs correspondingly.
2020-07-27 10:03:12 +01:00
wsi->a.protocol = &protocol_secstream_ws;
/*
* inform the SS user code that this has done a one-way
* upgrade to some other protocol... it will likely
* want to treat subsequent payloads differently
*/
lws_ss_event_helper(h, LWSSSCS_SERVER_UPGRADE);
lws_mux_mark_immortal(wsi);
if (lws_add_http_header_by_token(wsi, WSI_TOKEN_PROTOCOL,
(unsigned char *)wsi->a.vhost->ss_handle->policy->
u.http.u.ws.subprotocol,
(int)strlen(wsi->a.vhost->ss_handle->policy->
u.http.u.ws.subprotocol), &p, end))
return -1;
} else
#endif
ws proxy: also proxy h1 ws to h1 and h2 lws has been able to proxy h2 or h1 inbound connections to an h1 onward connection for a while now. It's simple to use just build with LWS_WITH_HTTP_PROXY and make a mount where the origin is the onward connection details. Unix sockets can also be used as the onward connection. This patch extends the support to be able to also do the same for inbound h2 or h1 ws upgrades to an h1 ws onward connection as well. This allows you to offer completely different services in a common URL space, including ones that connect back by ws / wss.
2019-03-19 11:53:57 +08:00
if (lws_add_http_header_by_token(wsi, WSI_TOKEN_PROTOCOL,
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
(unsigned char *)wsi->a.protocol->name,
(int)strlen(wsi->a.protocol->name), &p, end))
ss: server: h1, h2, ws basic support Add initial support for defining servers using Secure Streams policy and api semantics. Serving h1, h2 and ws should be functional, the new minimal example shows a combined http + SS server with an incrementing ws message shown in the browser over tls, in around 200 lines of user code. NOP out anything to do with plugins, they're not currently used. Update the docs correspondingly.
2020-07-27 10:03:12 +01:00
return -1;
ws proxy: also proxy h1 ws to h1 and h2 lws has been able to proxy h2 or h1 inbound connections to an h1 onward connection for a while now. It's simple to use just build with LWS_WITH_HTTP_PROXY and make a mount where the origin is the onward connection details. Unix sockets can also be used as the onward connection. This patch extends the support to be able to also do the same for inbound h2 or h1 ws upgrades to an h1 ws onward connection as well. This allows you to offer completely different services in a common URL space, including ones that connect back by ws / wss.
2019-03-19 11:53:57 +08:00
}
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
}
if (lws_finalize_http_header(wsi, &p, end))
return -1;
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
m = lws_ptr_diff_size_t(p, start);
ws proxy: also proxy h1 ws to h1 and h2 lws has been able to proxy h2 or h1 inbound connections to an h1 onward connection for a while now. It's simple to use just build with LWS_WITH_HTTP_PROXY and make a mount where the origin is the onward connection details. Unix sockets can also be used as the onward connection. This patch extends the support to be able to also do the same for inbound h2 or h1 ws upgrades to an h1 ws onward connection as well. This allows you to offer completely different services in a common URL space, including ones that connect back by ws / wss.
2019-03-19 11:53:57 +08:00
// lwsl_hexdump_notice(start, m);
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
n = lws_write(wsi, start, m, LWS_WRITE_HTTP_HEADERS);
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
if (n != (int)m) {
lwsl_err("_write returned %d from %d\n", n, (int)m);
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
return -1;
}
/*
* alright clean up, set our state to generic ws established, the
* mode / state of the nwsi will get the h2 processing done.
*/
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_set_state(wsi, LRS_ESTABLISHED);
cmake: allow setting LWS_ROLE_WS You can build lws without support for ws, with -DLWS_ROLE_WS=0. This is thanks to the role ops isolating all the ws-specific business in the ws role. Also retire more test apps replaced by minmal-examples.
2018-04-25 08:42:18 +08:00
wsi->lws_rx_parse_state = 0; // ==LWS_RXPS_NEW;
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
uri_ptr = lws_hdr_simple_ptr(wsi, WSI_TOKEN_HTTP_COLON_PATH);
n = lws_hdr_total_length(wsi, WSI_TOKEN_HTTP_COLON_PATH);
hit = lws_find_mount(wsi, uri_ptr, n);
if (hit && hit->cgienv &&
fakewsi: replace with smaller substructure Currently we always reserve a fakewsi per pt so events that don't have a related actual wsi, like vhost-protocol-init or vhost cert init via protocol callback can make callbacks that look reasonable to user protocol handler code expecting a valid wsi every time. This patch splits out stuff that user callbacks often unconditionally expect to be in a wsi, like context pointer, vhost pointer etc into a substructure, which is composed into struct lws at the top of it. Internal references (struct lws is opaque, so there are only internal references) are all updated to go via the substructre, the compiler should make that a NOP. Helpers are added when fakewsi is used and referenced. If not PLAT_FREERTOS, we continue to provide a full fakewsi in the pt as before, although the helpers improve consistency by zeroing down the substructure. There is a huge amount of user code out there over the last 10 years that did not always have the minimal examples to follow, some of it does some unexpected things. If it is PLAT_FREERTOS, that is a newer thing in lws and users have the benefit of being able to follow the minimal examples' approach. For PLAT_FREERTOS we don't reserve the fakewsi in the pt any more, saving around 800 bytes. The helpers then create a struct lws_a (the substructure) on the stack, zero it down (but it is only like 4 pointers) and prepare it with whatever we know like the context. Then we cast it to a struct lws * and use it in the user protocol handler call. In this case, the remainder of the struct lws is undefined. However the amount of old protocol handlers that might touch things outside of the substructure in PLAT_FREERTOS is very limited compared to legacy lws user code and the saving is significant on constrained devices. User handlers should not be touching everything in a wsi every time anyway, there are several cases where there is no valid wsi to do the call with. Dereference of things outside the substructure should only happen when the callback reason shows there is a valid wsi bound to the activity (as in all the minimal examples).
2020-07-19 08:33:46 +01:00
wsi->a.protocol->callback(wsi, LWS_CALLBACK_HTTP_PMO, wsi->user_space,
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
(void *)hit->cgienv, 0))
return 1;
lws_validity: unified connection validity tracking Refactor everything around ping / pong handling in ws and h2, so there is instead a protocol-independent validity lws_sul tracking how long it has been since the last exchange that confirms the operation of the network connection in both directions. Clean out periodic role callback and replace the last two role users with discrete lws_sul for each pt.
2019-09-18 13:09:32 +01:00
lws_validity_confirmed(wsi);
h2-ws-draft-02 This implements: https://www.ietf.org/id/draft-mcmanus-httpbis-h2-websockets-02.txt it's untested since no compatible clients/browsers available yet.
2017-11-14 07:35:05 +08:00
return 0;
}
rtos diet: http: remove headers at buildtime according to config Headers related to ws or h2 are now elided if the ws or h2 role is not enabled for build. In addition, a new build-time option LWS_WITH_HTTP_UNCOMMON_HEADERS on by default allows removal of less-common http headers to shrink the parser footprint. Minilex is adapted to produce 8 different versions of the lex table, chosen at build-time according to which headers are included in the build. If you don't need the unusual headers, or aren't using h2 or ws, this chops down the size of the ah and the rodata needed to hold the parsing table from 87 strings / pointers to 49, and the parsing table from 1177 to 696 bytes.
2020-02-28 10:31:04 +00:00
#endif
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
int
lws_read_h2(struct lws *wsi, unsigned char *buf, lws_filepos_t len)
{
unsigned char *oldbuf = buf;
// lwsl_notice("%s: h2 path: wsistate 0x%x len %d\n", __func__,
// wsi->wsistate, (int)len);
/*
* wsi here is always the network connection wsi, not a stream
* wsi. Once we unpicked the framing we will find the right
* swsi and make it the target of the frame.
*
* If it's ws over h2, the nwsi will get us here to do the h2
* processing, and that will call us back with the swsi +
* ESTABLISHED state for the inner payload, handled in a later
* case.
*/
while (len) {
clean: gcc13 false positive uninitialized https://github.com/warmcat/libwebsockets/issues/3004
2023-11-10 11:13:24 +00:00
lws_filepos_t body_chunk_len = 0;
codacy: minor fixes
2018-10-10 13:54:43 +08:00
int m;
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
/*
* we were accepting input but now we stopped doing so
*/
if (lws_is_flowcontrolled(wsi)) {
type comparisons: fixes This is a huge patch that should be a global NOP. For unix type platforms it enables -Wconversion to issue warnings (-> error) for all automatic casts that seem less than ideal but are normally concealed by the toolchain. This is things like passing an int to a size_t argument. Once enabled, I went through all args on my default build (which build most things) and tried to make the removed default cast explicit. With that approach it neither change nor bloat the code, since it compiles to whatever it was doing before, just with the casts made explicit... in a few cases I changed some length args from int to size_t but largely left the causes alone. From now on, new code that is relying on less than ideal casting will complain and nudge me to improve it by warnings.
2020-12-12 06:21:40 +00:00
lws_rxflow_cache(wsi, buf, 0, (size_t)len);
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
buf += len;
break;
}
/*
* lws_h2_parser() may send something; when it gets the
* whole frame, it will want to perform some action
* involving a reply. But we may be in a partial send
* situation on the network wsi...
*
* Even though we may be in a partial send and unable to
* send anything new, we still have to parse the network
* wsi in order to gain tx credit to send, which is
* potentially necessary to clear the old partial send.
*
* ALL network wsi-specific frames are sent by PPS
* already, these are sent as a priority on the writable
* handler, and so respect partial sends. The only
* problem is when a stream wsi wants to send an, eg,
* reply headers frame in response to the parsing
* we will do now... the *stream wsi* must stall in a
* different state until it is able to do so from a
* priority on the WRITABLE callback, same way that
* file transfers operate.
*/
m = lws_h2_parser(wsi, buf, len, &body_chunk_len);
if (m && m != 2) {
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lwsl_debug("%s: http2_parser bail: %d\n", __func__, m);
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
lws_close_free_wsi(wsi, LWS_CLOSE_STATUS_NOSTATUS,
"lws_read_h2 bail");
return -1;
}
if (m == 2) {
/* swsi has been closed */
buf += body_chunk_len;
break;
}
buf += body_chunk_len;
len -= body_chunk_len;
}
return lws_ptr_diff(buf, oldbuf);
}
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
int
lws_h2_client_stream_long_poll_rxonly(struct lws *wsi)
{
mux children: generalize helpers out of h2 implementation This should be a NOP for h2 support and only affects internal apis. But it lets us reuse the working and reliable h2 mux arrangements directly in other protocols later, and share code so building for h2 + new protocols can take advantage of common mux child handling struct and code. Break out common mux handling struct into its own type. Convert all uses of members that used to be in wsi->h2 to wsi->mux Audit all references to the members and break out generic helpers for anything that is useful for other mux-capable protocols to reuse wsi->mux related features.
2019-12-23 11:31:57 +00:00
if (!wsi->mux_substream)
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
return 1;
/*
* Elect to send an empty DATA with END_STREAM, to force the stream
* into HALF_CLOSED LOCAL
*/
wsi->h2.long_poll = 1;
wsi->h2.send_END_STREAM = 1;
client: secure streams Secure Streams is an optional layer on top of lws that separates policy like endpoint selection and tls cert validation into a device JSON policy document. Code that wants to open a client connection just specifies a streamtype name, and no longer deals with details like the endpoint, the protocol (!) or anything else other than payloads and optionally generic metadata; the JSON policy contains all the details for each streamtype. h1, h2, ws and mqtt client connections are supported. Logical secure streams outlive any particular connection and supports "nailed-up" connectivity regardless of underlying connection stability.
2020-02-29 12:37:24 +00:00
// lws_header_table_detach(wsi, 0);
ah: drop on LONG_POLL and allow dropping client ah early
2019-11-06 21:34:14 +00:00
h2: unify immortal stream tracking across SSE and ws substreams It was already correct but add helpers to isolate and deduplicate processing adding and closing a generically immortal stream. Change the default 31s h2 network connection timeout to be settable by .keepalive_timeout if nonzero. Add a public api allowing a client h2 stream to transition to half-closed LOCAL (by sending a 0-byte DATA with END_STREAM) and mark itself as immortal to create a read-only long-poll stream if the server allows it. Add a vhost server option flag LWS_SERVER_OPTION_VH_H2_HALF_CLOSED_LONG_POLL which allows the vhost to treat half-closed remotes as immortal long poll streams.
2019-09-13 10:33:24 +01:00
lws_callback_on_writable(wsi);
return 0;
}
Reference in a new issue Copy permalink