Mirrors/libwebsockets
1
0
Fork 0
mirror of https://github.com/warmcat/libwebsockets.git synced 2025-03-16 00:00:07 +01:00
Code Issues Releases Wiki Activity
libwebsockets/lib/core/libwebsockets.c

4030 lines
89 KiB
C
Raw Normal View History

remove bogus bytes on first line of websockets.c Reported-by: Pierre-Paul Lavoie <ppl@idios.org> Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 17:54:06 +00:00
/*
add-COPYING-lgpl-2.1.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-11-08 17:12:19 +00:00
* libwebsockets - small server side websockets and web server implementation
clean-whitespace.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-12-19 22:13:26 +00:00
*
clean
2017-09-23 12:55:21 +08:00
* Copyright (C) 2010-2017 Andy Green <andy@warmcat.com>
add-COPYING-lgpl-2.1.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-11-08 17:12:19 +00:00
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation:
* version 2.1 of the License.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
* MA 02110-1301 USA
add-kernel-doc-script-and-generated-api.txt Signed-off-by: Andy Green <andy@warmcat.com>
2010-10-31 17:51:39 +00:00
*/
refactor: core code in lib/core and private-libwebsockets.h to core/private.h This commit is coverity-clean as tested cmake .. -DLWS_WITH_MINIMAL_EXAMPLES=1 -DLWS_WITHOUT_EXTENSIONS=1 -DLWS_WITH_ACME=1 -DLWS_WITH_LWSWS=1 -DLWS_WITH_LIBUV=1 -DLWS_WITH_HTTP2=1 -DLWS_WITHOUT_CLIENT=0 -DLWS_WITHOUT_SERVER=0 -DLWS_UNIX_SOCK=1 -DLWS_WITH_TLS=0 -DLWS_WITH_MBEDTLS=0 -DLWS_WITH_CGI=1 -DCMAKE_BUILD_TYPE=DEBUG -DLWS_WITH_GENERIC_SESSIONS=1 -DLWS_WITH_RANGES=1 -DLWS_ROLE_WS=1 -DLWS_MAX_SMP=16 -DLWS_ROLE_H1=1 -DLWS_WITH_WOLFSSL=0 -DLWS_WITH_LIBEV=0 -DLWS_WITH_LIBEVENT=1
2018-05-03 10:49:36 +08:00
#include "core/private.h"
wince minor adaptations https://github.com/warmcat/libwebsockets/issues/444
2016-02-29 18:48:55 +08:00
#ifdef LWS_HAVE_SYS_TYPES_H
cgi Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:25:48 +08:00
#include <sys/types.h>
wince minor adaptations https://github.com/warmcat/libwebsockets/issues/444
2016-02-29 18:48:55 +08:00
#endif
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_IPV6
ipv6-allow-binding-to-ipv6-address-in-iface ipv4 and ipv6 binding to a named interface works OK. ipv4 binding to an IP also works, but we need some extra ipv6 magic to identify the ipv6 interface from an ipv6 address. This patch based on code from "user3546716" at http://stackoverflow.com/questions/13504934/binding-sockets-to-ipv6-addresses adds the necessary magic. https://github.com/warmcat/libwebsockets/issues/717
2016-12-15 09:58:20 +08:00
#if defined(WIN32) || defined(_WIN32)
mingw: fixes
2017-12-01 11:36:31 +08:00
#include <wincrypt.h>
Fix Iphlpapi.h -> iphlpapi.h If you cross compile for Windows you will get an error otherwise that the header cannot be found. See here for code example from microsoft: https://docs.microsoft.com/en-us/windows/desktop/iphlp/creating-a-basic-ip-helper-application
2018-08-29 13:29:07 +02:00
#include <iphlpapi.h>
ipv6-allow-binding-to-ipv6-address-in-iface ipv4 and ipv6 binding to a named interface works OK. ipv4 binding to an IP also works, but we need some extra ipv6 magic to identify the ipv6 interface from an ipv6 address. This patch based on code from "user3546716" at http://stackoverflow.com/questions/13504934/binding-sockets-to-ipv6-addresses adds the necessary magic. https://github.com/warmcat/libwebsockets/issues/717
2016-12-15 09:58:20 +08:00
#else
#include <net/if.h>
#endif
#endif
refactor context related functions into context.c Signed-off-by: Andy Green <andy.green@linaro.org>
2014-04-03 07:29:50 +08:00
int log_level = LLL_ERR | LLL_WARN | LLL_NOTICE;
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
static void (*lwsl_emit)(int level, const char *line)
#ifndef LWS_PLAT_OPTEE
= lwsl_emit_stderr
#endif
;
#ifndef LWS_PLAT_OPTEE
style cleanup This brings the library sources into compliance with checkpatch style except for three or four exceptions like WIN32 related stuff and one long string constant I don't want to break into multiple sprintf calls. There should be no functional or compilability change from all this (hopefully). Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-11 17:13:32 +08:00
static const char * const log_level_names[] = {
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
"ERR",
"WARN",
add lwsl_notice Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-19 12:18:07 +08:00
"NOTICE",
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
"INFO",
"DEBUG",
"PARSER",
"HEADER",
"EXTENSION",
"CLIENT",
introduce lws_latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-29 12:36:17 +08:00
"LATENCY",
coverity 175437: forgot to add name to array for new USER log level
2017-02-06 10:16:45 +08:00
"USER",
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
"THREAD",
coverity 175437: forgot to add name to array for new USER log level
2017-02-06 10:16:45 +08:00
"?",
"?"
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
};
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#endif
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
#if defined (_DEBUG)
void lwsi_set_role(struct lws *wsi, lws_wsi_state_t role)
{
wsi->wsistate = (wsi->wsistate & (~LWSI_ROLE_MASK)) | role;
lwsl_debug("lwsi_set_role(%p, 0x%x)\n", wsi, wsi->wsistate);
}
void lwsi_set_state(struct lws *wsi, lws_wsi_state_t lrs)
{
wsi->wsistate = (wsi->wsistate & (~LRS_MASK)) | lrs;
lwsl_debug("lwsi_set_state(%p, 0x%x)\n", wsi, wsi->wsistate);
}
#endif
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
signed char char_to_hex(const char c)
{
if (c >= '0' && c <= '9')
return c - '0';
if (c >= 'a' && c <= 'f')
return c - 'a' + 10;
if (c >= 'A' && c <= 'F')
return c - 'A' + 10;
return -1;
}
fd_cloexec: add and use lws_open wrapper and lws_plat_apply_FD_CLOEXEC() on cgi
2018-06-23 12:56:21 +08:00
int lws_open(const char *__file, int __oflag, ...)
{
va_list ap;
int n;
va_start(ap, __oflag);
if (((__oflag & O_CREAT) == O_CREAT)
#if defined(O_TMPFILE)
|| ((__oflag & O_TMPFILE) == O_TMPFILE)
#endif
)
/* last arg is really a mode_t. But windows... */
n = open(__file, __oflag, va_arg(ap, uint32_t));
else
n = open(__file, __oflag);
va_end(ap);
coverity: cleanup
2018-08-01 06:52:03 +08:00
if (n != -1 && lws_plat_apply_FD_CLOEXEC(n)) {
close(n);
return -1;
}
fd_cloexec: add and use lws_open wrapper and lws_plat_apply_FD_CLOEXEC() on cgi
2018-06-23 12:56:21 +08:00
return n;
}
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
void
lws_vhost_bind_wsi(struct lws_vhost *vh, struct lws *wsi)
{
lws_vhost_bind_wsi: avoid if already bound to same wsi if wsi->vhost == vhost then it is already bound to that vhost. doubling binding causes a problem during shutdown by trashing the reference counting.
2018-07-19 09:36:20 -05:00
if (wsi->vhost == vh)
return;
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
lws_context_lock(vh->context, __func__); /* ---------- context { */
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
wsi->vhost = vh;
vh->count_bound_wsi++;
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
lws_context_unlock(vh->context); /* } context ---------- */
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
lwsl_info("%s: vh %s: count_bound_wsi %d\n",
__func__, vh->name, vh->count_bound_wsi);
assert(wsi->vhost->count_bound_wsi > 0);
}
void
lws_vhost_unbind_wsi(struct lws *wsi)
{
smp: context lock around vhost destroy
2018-06-27 07:49:04 +08:00
if (!wsi->vhost)
return;
lws_context_lock(wsi->context, __func__); /* ---------- context { */
assert(wsi->vhost->count_bound_wsi > 0);
wsi->vhost->count_bound_wsi--;
lwsl_info("%s: vh %s: count_bound_wsi %d\n", __func__,
wsi->vhost->name, wsi->vhost->count_bound_wsi);
if (!wsi->vhost->count_bound_wsi &&
wsi->vhost->being_destroyed) {
/*
* We have closed all wsi that were bound to this vhost
* by any pt: nothing can be servicing any wsi belonging
* to it any more.
*
* Finalize the vh destruction
*/
__lws_vhost_destroy2(wsi->vhost);
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
}
smp: context lock around vhost destroy
2018-06-27 07:49:04 +08:00
wsi->vhost = NULL;
lws_context_unlock(wsi->context); /* } context ---------- */
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
}
Fix potential memory leak - Got rid of ifdef _WIN32 stuff adn moved to plat_ files instead. - Also, check all calls to lws_zalloc, was potential failure on WIN32 - Made context destory enable to destroy a half inited context as well. This way I got get rid of some of the error handling complexity in libwebsocket_create_context - Added TODOs for some potential problems I see where things might be leaking and such
2015-06-25 17:51:07 +02:00
void
pthreads: more locking
2018-03-05 16:49:28 +08:00
__lws_free_wsi(struct lws *wsi)
Fix potential memory leak - Got rid of ifdef _WIN32 stuff adn moved to plat_ files instead. - Also, check all calls to lws_zalloc, was potential failure on WIN32 - Made context destory enable to destroy a half inited context as well. This way I got get rid of some of the error handling complexity in libwebsocket_create_context - Added TODOs for some potential problems I see where things might be leaking and such
2015-06-25 17:51:07 +02:00
{
if (!wsi)
return;
pmd: autobahn fixes
2017-07-28 13:12:03 +08:00
clean
2017-09-23 12:55:21 +08:00
/*
* Protocol user data may be allocated either internally by lws
* or by specified the user. We should only free what we allocated.
*/
cleanups after api changes and mbed update Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 16:54:12 +08:00
if (wsi->protocol && wsi->protocol->per_session_data_size &&
wsi->user_space && !wsi->user_space_externally_allocated)
Fix potential memory leak - Got rid of ifdef _WIN32 stuff adn moved to plat_ files instead. - Also, check all calls to lws_zalloc, was potential failure on WIN32 - Made context destory enable to destroy a half inited context as well. This way I got get rid of some of the error handling complexity in libwebsocket_create_context - Added TODOs for some potential problems I see where things might be leaking and such
2015-06-25 17:51:07 +02:00
lws_free(wsi->user_space);
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
lws_buflist_destroy_all_segments(&wsi->buflist);
partial: replace ad-hoc code with a wsi buflist_out Various kinds of input stashing were replaced with a single buflist before v3.0... this patch replaces the partial send arrangements with its own buflist in the same way. Buflists as the name says are growable lists of allocations in a linked-list that take care of book-keeping what's added and removed (even if what is removed is less than the current buffer on the list). The immediate result is that we no longer have to freak out if we had a partial buffered and new output is coming... we can just pile it on the end of the buflist and keep draining the front of it. Likewise we no longer need to be rabid about reporting multiple attempts to send stuff without going back to the event loop, although not doing that will introduce inefficiencies we don't have to term it "illegal" any more. Since buflists have proven reliable on the input side and the logic for dealing with truncated "non-network events" was already there this internal-only change should be relatively self-contained.
2018-08-20 12:02:26 +08:00
lws_buflist_destroy_all_segments(&wsi->buflist_out);
udp
2018-03-24 08:07:00 +08:00
lws_free_set_NULL(wsi->udp);
ah owns rxbuf This is intended to solve a longstanding problem with the relationship between http/1.1 keep-alive and the service loop. Ah now contain an rx buffer which is used during header processing, and the ah may not be detached from the wsi until the rx buffer is exhausted. Having the rx buffer in the ah means we can delay using the rx until a later service loop. Ah which have pending rx force POLLIN service on the wsi they are attached to automatically, so we can interleave general service / connections with draining each ah rx buffer. The possible http/1.1 situations and their dispositions are: 1) exactly one set of http headers come. After processing, the ah is detached since no pending rx left. If more headers come later, a fresh ah is aqcuired when available and the rx flow control blocks the read until then. 2) more that one whole set of headers come and we remain in http mode (no upgrade). The ah is left attached and returns to the service loop after the first set of headers. We will get forced service due to the ah having pending content (respecting flowcontrol) and process the pending rx in the ah. If we use it all up, we will detach the ah. 3) one set of http headers come with ws traffic appended. We service the headers, do the upgrade, and keep the ah until the remaining ws content is used. When we exhausted the ws traffix in the ah rx buffer, we detach the ah. Since there can be any amount of http/1.1 pipelining on a connection, and each may be expensive to service, it's now enforced there is a return to the service loop after each header set is serviced on a connection. When I added the forced service for ah with pending buffering, I added support for it to the windows plat code. However this is untested. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-15 12:37:04 +08:00
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
if (wsi->vhost && wsi->vhost->lserv_wsi == wsi)
wsi->vhost->lserv_wsi = NULL;
client: oom4: unify with __lws_free_wsi
2018-10-06 08:00:32 +08:00
#if !defined(LWS_NO_CLIENT)
lws_dll_lws_remove(&wsi->dll_active_client_conns);
#endif
wsi->context->count_wsi_allocated--;
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
client: oom4: unify with __lws_free_wsi
2018-10-06 08:00:32 +08:00
#if defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
__lws_header_table_detach(wsi, 0);
#endif
__lws_same_vh_protocol_remove(wsi);
#if !defined(LWS_NO_CLIENT)
lws_client_stash_destroy(wsi);
lws_free_set_NULL(wsi->client_hostname_copy);
#endif
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
minimal-raw-netcat Adapt attack.sh to use it instead of OS netcat and fox various bugs that exposed.
2018-05-02 08:46:16 +08:00
if (wsi->role_ops->destroy_role)
wsi->role_ops->destroy_role(wsi);
LWS_WITH_PEER_LIMITS Introduces an optional, settable restriction on the number of simultaneous wsi connections and ah that a single peer IP can obtain. The default is disabled for build, and if enabled, unlimited. However when enabled at CMake, setting info.ip_limit_ah and / or info.ip_limit_wsi at context creation time will enforce the limits. Connections past the info.ip_limit_wsi limit are dropped, and connections needing an ah are forced to stay in the ah waiting list even when ah are available, while the peer is at the number of ah in info.ip_limit_ah.
2017-09-14 13:14:11 +08:00
#if defined(LWS_WITH_PEER_LIMITS)
lws_peer_track_wsi_close(wsi->context, wsi->peer);
wsi->peer = NULL;
#endif
close path: make sure a second time timeout and ssl buffered lists are scrubbed lws_meta children can have a different close path
2017-08-04 13:27:34 +08:00
/* since we will destroy the wsi, make absolutely sure now */
smp: fixes
2018-03-02 14:22:49 +08:00
#if defined(LWS_WITH_OPENSSL)
__lws_ssl_remove_wsi_from_buffered_list(wsi);
#endif
__lws_remove_from_timeout_list(wsi);
refactor: apply ops structs to event loop handlers
2018-04-29 10:44:36 +08:00
if (wsi->context->event_loop_ops->destroy_wsi)
wsi->context->event_loop_ops->destroy_wsi(wsi);
libevent: fix memory leak - Added event free function for accept.
2018-01-17 02:05:45 +01:00
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
lws_vhost_unbind_wsi(wsi);
libuv integration This gets the libuv stuff plumbed in and working. Currently it's only workable for some service thread, and there is an isolated valgrind problem left ==28425== 128 bytes in 1 blocks are definitely lost in loss record 3 of 3 ==28425== at 0x4C28C50: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x4C2AB1E: realloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x58BBB27: maybe_resize (core.c:748) ==28425== by 0x58BBB27: uv__io_start (core.c:787) ==28425== by 0x58C1B80: uv__signal_loop_once_init (signal.c:225) ==28425== by 0x58C1B80: uv_signal_init (signal.c:260) ==28425== by 0x58BF7A6: uv_loop_init (loop.c:66) ==28425== by 0x4157F5: lws_uv_initloop (libuv.c:89) ==28425== by 0x405536: main (test-server-libuv.c:284) libuv wants to sign off on all libuv 'handles' that will close, and callback to do the close confirmation asynchronously. The wsi close function is adapted when libuv is in use to work with libuv accordingly and exit the uv loop the number of remaining wsi is zero. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-14 09:27:41 +08:00
lwsl_debug("%s: %p, remaining wsi %d\n", __func__, wsi,
wsi->context->count_wsi_allocated);
Fix potential memory leak - Got rid of ifdef _WIN32 stuff adn moved to plat_ files instead. - Also, check all calls to lws_zalloc, was potential failure on WIN32 - Made context destory enable to destroy a half inited context as well. This way I got get rid of some of the error handling complexity in libwebsocket_create_context - Added TODOs for some potential problems I see where things might be leaking and such
2015-06-25 17:51:07 +02:00
lws_free(wsi);
}
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
void
lws_dll_add_front(struct lws_dll *d, struct lws_dll *phead)
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
{
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
if (d->prev)
return;
/* our next guy is current first guy */
d->next = phead->next;
/* if there is a next guy, set his prev ptr to our next ptr */
if (d->next)
d->next->prev = d;
/* our prev ptr is first ptr */
d->prev = phead;
/* set the first guy to be us */
phead->next = d;
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
}
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
/* situation is:
*
* HEAD: struct lws_dll * = &entry1
*
* Entry 1: struct lws_dll .pprev = &HEAD , .next = Entry 2
* Entry 2: struct lws_dll .pprev = &entry1 , .next = &entry2
* Entry 3: struct lws_dll .pprev = &entry2 , .next = NULL
*
* Delete Entry1:
*
* - HEAD = &entry2
* - Entry2: .pprev = &HEAD, .next = &entry3
* - Entry3: .pprev = &entry2, .next = NULL
*
* Delete Entry2:
*
* - HEAD = &entry1
* - Entry1: .pprev = &HEAD, .next = &entry3
* - Entry3: .pprev = &entry1, .next = NULL
*
* Delete Entry3:
*
* - HEAD = &entry1
* - Entry1: .pprev = &HEAD, .next = &entry2
* - Entry2: .pprev = &entry1, .next = NULL
*
*/
client connect: oom4 clean up timeout list https://github.com/warmcat/libwebsockets/issues/810
2017-02-21 23:38:40 +08:00
void
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
lws_dll_remove(struct lws_dll *d)
timeout move to doubly linked list In the case we have a lot of connections, checking them all for timeout state once a second becomes burdensome. At the moment if you have 100K connections, once a second they all get checked for timeout in a loop. This patch adds a doubly-linked list based in the context to each wsi, and only wsi with pending timeouts appear on it. At checking time, we traverse the list, which costs nothing if empty because nobody has a pending timeout. Similarly adding and removing from the list costs almost nothing since no iteration is required no matter how big the list. The extra 8 or 16 bytes in the wsi are offset a little bit by demoting .pps from int to char (save 3 bytes). And trim max act exts to 2, since we only provide one, saving 8 /16 bytes by itself if exts enabled. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 04:32:14 +08:00
{
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
if (!d->prev) /* ie, not part of the list */
timeout move to doubly linked list In the case we have a lot of connections, checking them all for timeout state once a second becomes burdensome. At the moment if you have 100K connections, once a second they all get checked for timeout in a loop. This patch adds a doubly-linked list based in the context to each wsi, and only wsi with pending timeouts appear on it. At checking time, we traverse the list, which costs nothing if empty because nobody has a pending timeout. Similarly adding and removing from the list costs almost nothing since no iteration is required no matter how big the list. The extra 8 or 16 bytes in the wsi are offset a little bit by demoting .pps from int to char (save 3 bytes). And trim max act exts to 2, since we only provide one, saving 8 /16 bytes by itself if exts enabled. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 04:32:14 +08:00
return;
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
/*
* remove us
*
* USp <-> us <-> USn --> USp <-> USn
*/
improve timeout and ah list comments Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 21:39:01 +08:00
/* if we have a next guy, set his prev to our prev */
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
if (d->next)
d->next->prev = d->prev;
improve timeout and ah list comments Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 21:39:01 +08:00
/* set our prev guy to our next guy instead of us */
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
if (d->prev)
d->prev->next = d->next;
timeout also fixup for0middle guy deletion case Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 23:11:39 +08:00
improve timeout and ah list comments Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 21:39:01 +08:00
/* we're out of the list, we should not point anywhere any more */
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
d->prev = NULL;
d->next = NULL;
}
void
__lws_remove_from_timeout_list(struct lws *wsi)
{
lws_dll_lws_remove(&wsi->dll_timeout);
smp: fixes
2018-03-02 14:22:49 +08:00
}
void
lws_remove_from_timeout_list(struct lws *wsi)
{
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
lws_pt_lock(pt, __func__);
__lws_remove_from_timeout_list(wsi);
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
lws_pt_unlock(pt);
timeout move to doubly linked list In the case we have a lot of connections, checking them all for timeout state once a second becomes burdensome. At the moment if you have 100K connections, once a second they all get checked for timeout in a loop. This patch adds a doubly-linked list based in the context to each wsi, and only wsi with pending timeouts appear on it. At checking time, we traverse the list, which costs nothing if empty because nobody has a pending timeout. Similarly adding and removing from the list costs almost nothing since no iteration is required no matter how big the list. The extra 8 or 16 bytes in the wsi are offset a little bit by demoting .pps from int to char (save 3 bytes). And trim max act exts to 2, since we only provide one, saving 8 /16 bytes by itself if exts enabled. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 04:32:14 +08:00
}
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
void
lws_dll_dump(struct lws_dll_lws *head, const char *title)
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
{
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
int n = 0;
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
(void)n;
lwsl_notice("%s: %s (head.next %p)\n", __func__, title, head->next);
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
lws_start_foreach_dll_safe(struct lws_dll_lws *, d, d1, head->next) {
struct lws *wsi = lws_container_of(d, struct lws, dll_hrtimer);
(void)wsi;
lwsl_notice(" %d: wsi %p: %llu\n", n++, wsi,
(unsigned long long)wsi->pending_timer);
} lws_end_foreach_dll_safe(d, d1);
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
}
pthreads: more locking
2018-03-05 16:49:28 +08:00
void
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
__lws_set_timer_usecs(struct lws *wsi, lws_usec_t usecs)
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
{
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
struct lws_dll_lws *dd = &pt->dll_head_hrtimer;
struct timeval now;
struct lws *wsi1;
int bef = 0;
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
lws_dll_lws_remove(&wsi->dll_hrtimer);
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
if (usecs == LWS_SET_TIMER_USEC_CANCEL)
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
return;
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
gettimeofday(&now, NULL);
wsi->pending_timer = ((now.tv_sec * 1000000ll) + now.tv_usec) + usecs;
/*
* we sort the hrtimer list with the earliest timeout first
*/
lws_start_foreach_dll_safe(struct lws_dll_lws *, d, d1,
pt->dll_head_hrtimer.next) {
dd = d;
wsi1 = lws_container_of(d, struct lws, dll_hrtimer);
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
if (wsi1->pending_timer >= wsi->pending_timer) {
/* d, dprev's next, is >= our time */
bef = 1;
break;
}
} lws_end_foreach_dll_safe(d, d1);
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
if (bef) {
/*
* we go before dd
* DDp <-> DD <-> DDn --> DDp <-> us <-> DD <-> DDn
*/
/* we point forward to dd */
wsi->dll_hrtimer.next = dd;
/* we point back to what dd used to point back to */
wsi->dll_hrtimer.prev = dd->prev;
/* DDp points forward to us now */
dd->prev->next = &wsi->dll_hrtimer;
/* DD points back to us now */
dd->prev = &wsi->dll_hrtimer;
} else {
/*
* we go after dd
* DDp <-> DD <-> DDn --> DDp <-> DD <-> us <-> DDn
*/
/* we point forward to what dd used to point forward to */
wsi->dll_hrtimer.next = dd->next;
/* we point back to dd */
wsi->dll_hrtimer.prev = dd;
/* DDn points back to us */
if (dd->next)
dd->next->prev = &wsi->dll_hrtimer;
/* DD points forward to us */
dd->next = &wsi->dll_hrtimer;
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
}
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
// lws_dll_dump(&pt->dll_head_hrtimer, "after set_timer_usec");
lws_set_timer This adds a new api lws_set_timer(wsi, secs), which schedules a callback LWS_CALLBACK_TIMER secs seconds into the future. The timer can be continuously deferred by calling lws_set_timer() again before it expires. Calling lws_set_timer(wsi, -1) cancels any pending timer.
2017-12-05 20:02:29 +08:00
}
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
LWS_VISIBLE lws_usec_t
lws_now_usecs(void)
{
struct timeval now;
gettimeofday(&now, NULL);
return (now.tv_sec * 1000000ll) + now.tv_usec;
}
pthreads: more locking
2018-03-05 16:49:28 +08:00
LWS_VISIBLE void
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
lws_set_timer_usecs(struct lws *wsi, lws_usec_t usecs)
{
__lws_set_timer_usecs(wsi, usecs);
}
lws_usec_t
__lws_hrtimer_service(struct lws_context_per_thread *pt)
pthreads: more locking
2018-03-05 16:49:28 +08:00
{
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
struct timeval now;
struct lws *wsi;
lws_usec_t t;
pthreads: more locking
2018-03-05 16:49:28 +08:00
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
gettimeofday(&now, NULL);
t = (now.tv_sec * 1000000ll) + now.tv_usec;
lws_start_foreach_dll_safe(struct lws_dll_lws *, d, d1,
pt->dll_head_hrtimer.next) {
wsi = lws_container_of(d, struct lws, dll_hrtimer);
/*
* if we met one in the future, we are done, because the list
* is sorted by time in the future.
*/
if (wsi->pending_timer > t)
break;
lws_set_timer_usecs(wsi, LWS_SET_TIMER_USEC_CANCEL);
/* it's time for the timer to be serviced */
if (wsi->protocol &&
wsi->protocol->callback(wsi, LWS_CALLBACK_TIMER,
wsi->user_space, NULL, 0))
__lws_close_free_wsi(wsi, LWS_CLOSE_STATUS_NOSTATUS,
"timer cb errored");
} lws_end_foreach_dll_safe(d, d1);
/* return an estimate how many us until next timer hit */
if (!pt->dll_head_hrtimer.next)
return LWS_HRTIMER_NOWAIT;
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
wsi = lws_container_of(pt->dll_head_hrtimer.next, struct lws,
dll_hrtimer);
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
gettimeofday(&now, NULL);
t = (now.tv_sec * 1000000ll) + now.tv_usec;
if (wsi->pending_timer < t)
return 0;
return wsi->pending_timer - t;
pthreads: more locking
2018-03-05 16:49:28 +08:00
}
smp: fixes
2018-03-02 14:22:49 +08:00
void
__lws_set_timeout(struct lws *wsi, enum pending_timeout reason, int secs)
{
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
smp: fixes
2018-03-02 14:22:49 +08:00
time_t now;
time(&now);
http: enlarge headers buffers since they may meet large headers from vhost config
2018-08-25 12:27:00 +08:00
lwsl_debug("%s: %p: %d secs (reason %d)\n", __func__, wsi, secs, reason);
smp: fixes
2018-03-02 14:22:49 +08:00
wsi->pending_timeout_limit = secs;
wsi->pending_timeout_set = now;
wsi->pending_timeout = reason;
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
if (!reason)
lws_dll_lws_remove(&wsi->dll_timeout);
else
lws_dll_lws_add_front(&wsi->dll_timeout, &pt->dll_head_timeout);
smp: fixes
2018-03-02 14:22:49 +08:00
}
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
LWS_VISIBLE void
lws_set_timeout(struct lws *wsi, enum pending_timeout reason, int secs)
{
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
lws-meta
2017-07-19 04:39:14 +08:00
if (secs == LWS_TO_KILL_SYNC) {
lws_remove_from_timeout_list(wsi);
lwsl_debug("synchronously killing %p\n", wsi);
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lws_close_free_wsi(wsi, LWS_CLOSE_STATUS_NOSTATUS,
"to sync kill");
lws-meta
2017-07-19 04:39:14 +08:00
return;
}
lws_ring: implement lagging client culling
2018-05-05 10:16:54 +08:00
if (secs == LWS_TO_KILL_ASYNC)
secs = 0;
smp: fixes
2018-03-02 14:22:49 +08:00
lws_pt_lock(pt, __func__);
__lws_set_timeout(wsi, reason, secs);
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
lws_pt_unlock(pt);
}
timeout move to doubly linked list In the case we have a lot of connections, checking them all for timeout state once a second becomes burdensome. At the moment if you have 100K connections, once a second they all get checked for timeout in a loop. This patch adds a doubly-linked list based in the context to each wsi, and only wsi with pending timeouts appear on it. At checking time, we traverse the list, which costs nothing if empty because nobody has a pending timeout. Similarly adding and removing from the list costs almost nothing since no iteration is required no matter how big the list. The extra 8 or 16 bytes in the wsi are offset a little bit by demoting .pps from int to char (save 3 bytes). And trim max act exts to 2, since we only provide one, saving 8 /16 bytes by itself if exts enabled. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 04:32:14 +08:00
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
/* requires context + vh lock */
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
int
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
__lws_timed_callback_remove(struct lws_vhost *vh, struct lws_timed_vh_protocol *p)
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
{
lws_start_foreach_llp(struct lws_timed_vh_protocol **, pt,
vh->timed_vh_protocol_list) {
if (*pt == p) {
*pt = p->next;
lws_free(p);
return 0;
}
} lws_end_foreach_llp(pt, next);
return 1;
}
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
int
lws_pthread_self_to_tsi(struct lws_context *context)
{
#if LWS_MAX_SMP > 1
pthread_t ps = pthread_self();
struct lws_context_per_thread *pt = &context->pt[0];
int n;
for (n = 0; n < context->count_threads; n++) {
if (pthread_equal(ps, pt->self))
return n;
pt++;
}
return -1;
#else
return 0;
#endif
}
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
LWS_VISIBLE LWS_EXTERN int
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
lws_timed_callback_vh_protocol(struct lws_vhost *vh,
const struct lws_protocols *prot, int reason,
int secs)
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
{
struct lws_timed_vh_protocol *p = (struct lws_timed_vh_protocol *)
lws_malloc(sizeof(*p), "timed_vh");
if (!p)
return 1;
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
p->tsi_req = lws_pthread_self_to_tsi(vh->context);
if (p->tsi_req < 0) /* not called from a service thread --> tsi 0 */
p->tsi_req = 0;
lws_context_lock(vh->context, __func__); /* context ----------------- */
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
p->protocol = prot;
p->reason = reason;
p->time = lws_now_secs() + secs;
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
lws_vhost_lock(vh); /* vhost ---------------------------------------- */
p->next = vh->timed_vh_protocol_list;
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
vh->timed_vh_protocol_list = p;
smp: lws_timed_callback_vh_protocol: request sets handler tsi and other fixes
2018-11-15 16:33:54 +08:00
lws_vhost_unlock(vh); /* -------------------------------------- vhost */
lws_context_unlock(vh->context); /* ------------------------- context */
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
return 0;
}
cgi: fix for https git server
2018-06-16 09:35:07 +08:00
void
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
lws_remove_child_from_any_parent(struct lws *wsi)
{
struct lws **pwsi;
lws_remove_child_from_any_parent: clear parent pointer and fix failure message
2017-03-30 08:31:36 +08:00
int seen = 0;
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
lws_remove_child_from_any_parent: clear parent pointer and fix failure message
2017-03-30 08:31:36 +08:00
if (!wsi->parent)
return;
/* detach ourselves from parent's child list */
pwsi = &wsi->parent->child_list;
while (*pwsi) {
if (*pwsi == wsi) {
clean
2017-09-23 12:55:21 +08:00
lwsl_info("%s: detach %p from parent %p\n", __func__,
wsi, wsi->parent);
lws-meta
2017-07-19 04:39:14 +08:00
if (wsi->parent->protocol)
wsi->parent->protocol->callback(wsi,
LWS_CALLBACK_CHILD_CLOSING,
wsi->parent->user_space, wsi, 0);
lws_remove_child_from_any_parent: clear parent pointer and fix failure message
2017-03-30 08:31:36 +08:00
*pwsi = wsi->sibling_list;
seen = 1;
break;
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
}
lws_remove_child_from_any_parent: clear parent pointer and fix failure message
2017-03-30 08:31:36 +08:00
pwsi = &(*pwsi)->sibling_list;
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
}
lws_remove_child_from_any_parent: clear parent pointer and fix failure message
2017-03-30 08:31:36 +08:00
if (!seen)
lwsl_err("%s: failed to detach from parent\n", __func__);
wsi->parent = NULL;
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
}
raw: enable server and client raw sockets
2017-03-07 16:06:05 +08:00
int
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lws_bind_protocol(struct lws *wsi, const struct lws_protocols *p,
const char *reason)
raw: enable server and client raw sockets
2017-03-07 16:06:05 +08:00
{
// if (wsi->protocol == p)
// return 0;
raw: take care about same vh protocol linked list
2017-04-06 13:49:17 +08:00
const struct lws_protocols *vp = wsi->vhost->protocols, *vpo;
raw: enable server and client raw sockets
2017-03-07 16:06:05 +08:00
DROP_PROTOCOL: guarantee only one DROP per BIND
2018-04-03 10:37:14 +08:00
if (wsi->protocol && wsi->protocol_bind_balance) {
client: bind and drop protocol like server HTTP server protocols have had for a while LWS_CALLBACK_HTTP_DROP/BIND_PROTOCOL callbacks that mark when a wsi is attched to a protocol and detached. It turns out this is generally useful for everything to know when a wsi is joining a protocol and definitively completely finished with a protocol. Particularly with client wsi where you provided the userdata externally, this makes a clear point to free() it on the protocol binding being dropped. This patch adds protocol bind / unbind callbacks to the role definition and lets them operate on all roles. For the various roles HTTP server: LWS_CALLBACK_HTTP_BIND/DROP_PROTOCOL as before HTTP client: LWS_CALLBACK_CLIENT_HTTP_BIND/DROP_PROTOCOL ws server: LWS_CALLBACK_WS_SERVER_BIND/DROP_PROTOCOL ws client: LWS_CALLBACK_WS_CLIENT_BIND/DROP_PROTOCOL raw file: LWS_CALLBACK_RAW_FILE_BIND/DROP_PROTOCOL raw skt: LWS_CALLBACK_RAW_SKT_BIND/DROP_PROTOCOL
2018-08-18 14:11:29 +08:00
wsi->protocol->callback(wsi,
wsi->role_ops->protocol_unbind_cb[!!lwsi_role_server(wsi)],
threadpool
2018-09-02 14:35:37 +08:00
wsi->user_space, (void *)reason, 0);
DROP_PROTOCOL: guarantee only one DROP per BIND
2018-04-03 10:37:14 +08:00
wsi->protocol_bind_balance = 0;
}
raw: enable server and client raw sockets
2017-03-07 16:06:05 +08:00
if (!wsi->user_space_externally_allocated)
lws_free_set_NULL(wsi->user_space);
raw: take care about same vh protocol linked list
2017-04-06 13:49:17 +08:00
lws_same_vh_protocol_remove(wsi);
raw: enable server and client raw sockets
2017-03-07 16:06:05 +08:00
wsi->protocol = p;
if (!p)
return 0;
if (lws_ensure_user_space(wsi))
return 1;
raw: take care about same vh protocol linked list
2017-04-06 13:49:17 +08:00
if (p > vp && p < &vp[wsi->vhost->count_protocols])
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
lws_same_vh_protocol_insert(wsi, (int)(p - vp));
raw: take care about same vh protocol linked list
2017-04-06 13:49:17 +08:00
else {
int n = wsi->vhost->count_protocols;
int hit = 0;
vpo = vp;
while (n--) {
bind protocol: fix for NULL names
2017-04-07 18:22:54 +08:00
if (p->name && vp->name && !strcmp(p->name, vp->name)) {
raw: take care about same vh protocol linked list
2017-04-06 13:49:17 +08:00
hit = 1;
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
lws_same_vh_protocol_insert(wsi, (int)(vp - vpo));
raw: take care about same vh protocol linked list
2017-04-06 13:49:17 +08:00
break;
}
vp++;
}
if (!hit)
clean
2017-09-23 12:55:21 +08:00
lwsl_err("%s: %p is not in vhost '%s' protocols list\n",
__func__, p, wsi->vhost->name);
raw: take care about same vh protocol linked list
2017-04-06 13:49:17 +08:00
}
client: bind and drop protocol like server HTTP server protocols have had for a while LWS_CALLBACK_HTTP_DROP/BIND_PROTOCOL callbacks that mark when a wsi is attched to a protocol and detached. It turns out this is generally useful for everything to know when a wsi is joining a protocol and definitively completely finished with a protocol. Particularly with client wsi where you provided the userdata externally, this makes a clear point to free() it on the protocol binding being dropped. This patch adds protocol bind / unbind callbacks to the role definition and lets them operate on all roles. For the various roles HTTP server: LWS_CALLBACK_HTTP_BIND/DROP_PROTOCOL as before HTTP client: LWS_CALLBACK_CLIENT_HTTP_BIND/DROP_PROTOCOL ws server: LWS_CALLBACK_WS_SERVER_BIND/DROP_PROTOCOL ws client: LWS_CALLBACK_WS_CLIENT_BIND/DROP_PROTOCOL raw file: LWS_CALLBACK_RAW_FILE_BIND/DROP_PROTOCOL raw skt: LWS_CALLBACK_RAW_SKT_BIND/DROP_PROTOCOL
2018-08-18 14:11:29 +08:00
if (wsi->protocol->callback(wsi, wsi->role_ops->protocol_bind_cb[
!!lwsi_role_server(wsi)],
raw: enable server and client raw sockets
2017-03-07 16:06:05 +08:00
wsi->user_space, NULL, 0))
return 1;
DROP_PROTOCOL: guarantee only one DROP per BIND
2018-04-03 10:37:14 +08:00
wsi->protocol_bind_balance = 1;
raw: enable server and client raw sockets
2017-03-07 16:06:05 +08:00
return 0;
}
clean-whitespace.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-12-19 22:13:26 +00:00
void
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
__lws_close_free_wsi(struct lws *wsi, enum lws_close_status reason,
const char *caller)
add-per-session-user-data.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-11-03 11:13:06 +00:00
{
coverity 156860 wsi dereference before NULL check ... bit in reality, wsi can never be NULL Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 21:32:08 +08:00
struct lws_context_per_thread *pt;
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
struct lws *wsi1, *wsi2;
add explicit parent child wsi relationships wsi can have a full tree relationship with each other using linked lists. closing the parent ensures the children are closed first. Convert cgi to use this instead of his cgi-specific sub-wsi management. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-02 09:17:22 +08:00
struct lws_context *context;
autobahn fixes This replaces the old test-app for echo with separate client and server minimal versions. The autobahn test script is made more autonomous and tests both client and server.
2018-04-20 10:33:23 +08:00
int n;
fork-sever-process-and-introduce-broadcast-api.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-12-18 15:13:50 +00:00
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
lwsl_info("%s: %p: caller: %s\n", __func__, wsi, caller);
lws-meta
2017-07-19 04:39:14 +08:00
unify server and client close Signed-off-by: Andy Green <andy@warmcat.com>
2011-02-14 08:03:48 +00:00
if (!wsi)
fork-sever-process-and-introduce-broadcast-api.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-12-18 15:13:50 +00:00
return;
lws access log option and lwsws conf This adds the ability to store apache-compatible logs to a file given at vhost-creation time. lwsws conf can set it per-vhost using "access-log": "<filepath>" The feature defaults to disabled at cmake, it can be set independently but LWS_WITH_LWSWS set it on. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 12:00:23 +08:00
lws_access_log(wsi);
coverity 156860 wsi dereference before NULL check ... bit in reality, wsi can never be NULL Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 21:32:08 +08:00
context = wsi->context;
pt = &context->pt[(int)wsi->tsi];
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
lws_stats_atomic_bump(wsi->context, pt, LWSSTATS_C_API_CLOSE, 1);
add-per-session-user-data.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-11-03 11:13:06 +00:00
client: http1.1 pipelining
2018-03-26 12:05:04 +08:00
#if !defined(LWS_NO_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
lws_free_set_NULL(wsi->client_hostname_copy);
client: http1.1 pipelining
2018-03-26 12:05:04 +08:00
/* we are no longer an active client connection that can piggyback */
lws_dll_lws_remove(&wsi->dll_active_client_conns);
/*
* if we have wsi in our transaction queue, if we are closing we
* must go through and close all those first
*/
fix minimal-server-tls
2018-04-27 09:13:23 +08:00
if (wsi->vhost) {
refactor: change event lib minimal examples to serve as the test apps
2018-04-30 19:17:32 +08:00
if ((int)reason != -1)
lws_vhost_lock(wsi->vhost);
fix minimal-server-tls
2018-04-27 09:13:23 +08:00
lws_start_foreach_dll_safe(struct lws_dll_lws *, d, d1,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
wsi->dll_client_transaction_queue_head.next) {
fix minimal-server-tls
2018-04-27 09:13:23 +08:00
struct lws *w = lws_container_of(d, struct lws,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
dll_client_transaction_queue);
client: http1.1 pipelining
2018-03-26 12:05:04 +08:00
refactor: change event lib minimal examples to serve as the test apps
2018-04-30 19:17:32 +08:00
__lws_close_free_wsi(w, -1, "trans q leader closing");
fix minimal-server-tls
2018-04-27 09:13:23 +08:00
} lws_end_foreach_dll_safe(d, d1);
client: http1.1 pipelining
2018-03-26 12:05:04 +08:00
fix minimal-server-tls
2018-04-27 09:13:23 +08:00
/*
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
* !!! If we are closing, but we have pending pipelined
* transaction results we already sent headers for, that's going
* to destroy sync for HTTP/1 and leave H2 stream with no live
* swsi.
fix minimal-server-tls
2018-04-27 09:13:23 +08:00
*
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
* However this is normal if we are being closed because the
* transaction queue leader is closing.
fix minimal-server-tls
2018-04-27 09:13:23 +08:00
*/
lws_dll_lws_remove(&wsi->dll_client_transaction_queue);
refactor: change event lib minimal examples to serve as the test apps
2018-04-30 19:17:32 +08:00
if ((int)reason !=-1)
lws_vhost_unlock(wsi->vhost);
fix minimal-server-tls
2018-04-27 09:13:23 +08:00
}
client: http1.1 pipelining
2018-03-26 12:05:04 +08:00
#endif
add explicit parent child wsi relationships wsi can have a full tree relationship with each other using linked lists. closing the parent ensures the children are closed first. Convert cgi to use this instead of his cgi-specific sub-wsi management. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-02 09:17:22 +08:00
/* if we have children, close them first */
if (wsi->child_list) {
wsi2 = wsi->child_list;
while (wsi2) {
wsi1 = wsi2->sibling_list;
proxy rewrite If you enable -DLWS_WITH_HTTP_PROXY=1 at cmake, the test server has a new URI path http://localhost:7681/proxytest If you visit here, a client connection to http://example.com:80 is spawned, and the results piped on to your original connection. Also with LWS_WITH_HTTP_PROXY enabled at cmake, lws wants to link to an additional library, "libhubbub". This allows lws to do html rewriting on the fly, adjusting proxied urls in a lightweight and fast way.
2016-03-20 11:59:53 +08:00
wsi2->parent = NULL;
/* stop it doing shutdown processing */
wsi2->socket_is_permanently_unusable = 1;
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
__lws_close_free_wsi(wsi2, reason,
"general child recurse");
add explicit parent child wsi relationships wsi can have a full tree relationship with each other using linked lists. closing the parent ensures the children are closed first. Convert cgi to use this instead of his cgi-specific sub-wsi management. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-02 09:17:22 +08:00
wsi2 = wsi1;
}
proxy rewrite If you enable -DLWS_WITH_HTTP_PROXY=1 at cmake, the test server has a new URI path http://localhost:7681/proxytest If you visit here, a client connection to http://example.com:80 is spawned, and the results piped on to your original connection. Also with LWS_WITH_HTTP_PROXY enabled at cmake, lws wants to link to an additional library, "libhubbub". This allows lws to do html rewriting on the fly, adjusting proxied urls in a lightweight and fast way.
2016-03-20 11:59:53 +08:00
wsi->child_list = NULL;
add explicit parent child wsi relationships wsi can have a full tree relationship with each other using linked lists. closing the parent ensures the children are closed first. Convert cgi to use this instead of his cgi-specific sub-wsi management. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-02 09:17:22 +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
if (wsi->role_ops == &role_ops_raw_file) {
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_remove_child_from_any_parent(wsi);
pthreads: more locking
2018-03-05 16:49:28 +08:00
__remove_wsi_socket_from_fds(wsi);
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
wsi->protocol->callback(wsi, wsi->role_ops->close_cb[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
wsi->user_space, NULL, 0);
goto async_close;
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
}
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
wsi->wsistate_pre_close = wsi->wsistate;
close: make sure all paths capture original state
2017-11-24 12:18:49 +08:00
cgi Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:25:48 +08:00
#ifdef LWS_WITH_CGI
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
if (wsi->role_ops == &role_ops_cgi) {
cgi Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:25:48 +08:00
/* we are not a network connection, but a handler for CGI io */
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
if (wsi->parent && wsi->parent->http.cgi) {
cgi: remove and kill CGI after closing its STDOUT handler Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>
2017-08-26 12:00:16 +08:00
if (wsi->cgi_channel == LWS_STDOUT)
lws_cgi_remove_and_kill(wsi->parent);
lwsws cgi integration Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-13 11:49:07 +08:00
/* end the binding between us and master */
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
wsi->parent->http.cgi->stdwsi[(int)wsi->cgi_channel] =
NULL;
cgi: remove and kill CGI after closing its STDOUT handler Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>
2017-08-26 12:00:16 +08:00
}
cgi Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:25:48 +08:00
wsi->socket_is_permanently_unusable = 1;
goto just_kill_connection;
}
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
if (wsi->http.cgi)
cgi: remove and kill CGI after closing its STDOUT handler Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>
2017-08-26 12:00:16 +08:00
lws_cgi_remove_and_kill(wsi);
cgi Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:25:48 +08:00
#endif
client-fix-header-stash-leak-on-close-before-success
2017-07-21 21:49:24 +08:00
#if !defined(LWS_NO_CLIENT)
client: convert stash to exactly sized mallocs
2017-12-07 10:09:56 +08:00
lws_client_stash_destroy(wsi);
client-fix-header-stash-leak-on-close-before-success
2017-07-21 21:49:24 +08:00
#endif
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
if (wsi->role_ops == &role_ops_raw_skt) {
raw: adoption and processing
2017-02-12 20:32:49 +08:00
wsi->socket_is_permanently_unusable = 1;
goto just_kill_connection;
}
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
#if defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
if (lwsi_role_http(wsi) && lwsi_role_server(wsi) &&
wsi->http.fop_fd != NULL)
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
lws_vfs_file_close(&wsi->http.fop_fd);
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
#endif
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
debug: with debug build and uv, repeat some closes The closes have to complete after libuv returned to the event loop. So make sure to catch any problem that gets introduced where another close entry in the meanwhile is unsafe. https://github.com/warmcat/libwebsockets/issues/1245
2018-04-19 10:08:48 +08:00
if (lwsi_state(wsi) == LRS_DEAD_SOCKET)
return;
timeout counts as pernmanently unusable close Signed-off-by: Andy Green <andy.green@linaro.org>
2015-04-12 08:17:26 +08:00
if (wsi->socket_is_permanently_unusable ||
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
reason == LWS_CLOSE_STATUS_NOSTATUS_CONTEXT_DESTROY ||
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_state(wsi) == LRS_SHUTDOWN)
trac 89 add flag on fatal send fail to use simple close Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-16 08:23:46 +08:00
goto just_kill_connection;
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
switch (lwsi_state_PRE_CLOSE(wsi)) {
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
case LRS_DEAD_SOCKET:
introduce-new-05-socket-closure-rules.patch Signed-off-by: Andy Green <andy.green@linaro.org>
2011-02-10 09:07:05 +00:00
return;
close if we tried to close politely just close next time Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-15 22:48:58 +08:00
/* we tried the polite way... */
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
case LRS_WAITING_TO_SEND_CLOSE:
case LRS_AWAITING_CLOSE_ACK:
autobahn fixes This replaces the old test-app for echo with separate client and server minimal versions. The autobahn test script is made more autonomous and tests both client and server.
2018-04-20 10:33:23 +08:00
case LRS_RETURNED_CLOSE:
close if we tried to close politely just close next time Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-15 22:48:58 +08:00
goto just_kill_connection;
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
case LRS_FLUSHING_BEFORE_CLOSE:
http: compression methods Add generic http compression layer eanbled at cmake with LWS_WITH_HTTP_STREAM_COMPRESSION. This is wholly a feature of the HTTP role (used by h1 and h2 roles) and doesn't exist outside that context. Currently provides 'deflate' and 'br' compression methods for server side only. 'br' requires also -DLWS_WITH_HTTP_BROTLI=1 at cmake and the brotli libraries (available in your distro already) and dev package. Other compression methods can be added nicely using an ops struct. The built-in file serving stuff will use this is the client says he can handle it, and the mimetype of the file either starts with "text/" (html and css etc) or is the mimetype of Javascript. zlib allocates quite a bit while in use, it seems to be around 256KiB per stream. So this is only useful on relatively strong servers with lots of memory. However for some usecases where you are serving a lot of css and js assets, it's a nice help. The patch performs special treatment for http/1.1 pipelining, since the compression is performed on the fly the compressed content-length is not known until the end. So for h1 only, chunked transfer-encoding is automatically added so pipelining can continue of the connection. For h2 the chunking is neither supported nor required, so it "just works". User code can also request to add a compression transform before the reply headers were sent using the new api LWS_VISIBLE int lws_http_compression_apply(struct lws *wsi, const char *name, unsigned char **p, unsigned char *end, char decomp); ... this allows transparent compression of dynamically generated HTTP. The requested compression (eg, "deflate") is only applied if the client headers indicated it was supported, otherwise it's a NOP. Name may be NULL in which case the first compression method in the internal table at stream.c that is mentioned as acceptable by the client will be used. NOTE: the compression translation, same as h2 support, relies on the user code using LWS_WRITE_HTTP and then LWS_WRITE_HTTP_FINAL on the last part written. The internal lws fileserving code already does this.
2018-09-02 14:43:05 +08:00
if (lws_has_buffered_out(wsi)
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
|| wsi->http.comp_ctx.buflist_comp ||
wsi->http.comp_ctx.may_have_more
#endif
) {
public api remove superfluous context params API BREAK Extend the cleanout caused by wsi having a context pointer into the public api. There's no point keeping the 1.5 compatibility work, we have changed the api in several places and rebuilt wasn't going to be enough a while ago. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-16 18:19:08 +08:00
lws_callback_on_writable(wsi);
close should flush any truncated send before really closing Signed-off-by: Andy Green <andy.green@linaro.org>
2014-04-10 14:25:24 +08:00
return;
}
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsl_info("%p: end LRS_FLUSHING_BEFORE_CLOSE\n", wsi);
close should flush any truncated send before really closing Signed-off-by: Andy Green <andy.green@linaro.org>
2014-04-10 14:25:24 +08:00
goto just_kill_connection;
default:
http: compression methods Add generic http compression layer eanbled at cmake with LWS_WITH_HTTP_STREAM_COMPRESSION. This is wholly a feature of the HTTP role (used by h1 and h2 roles) and doesn't exist outside that context. Currently provides 'deflate' and 'br' compression methods for server side only. 'br' requires also -DLWS_WITH_HTTP_BROTLI=1 at cmake and the brotli libraries (available in your distro already) and dev package. Other compression methods can be added nicely using an ops struct. The built-in file serving stuff will use this is the client says he can handle it, and the mimetype of the file either starts with "text/" (html and css etc) or is the mimetype of Javascript. zlib allocates quite a bit while in use, it seems to be around 256KiB per stream. So this is only useful on relatively strong servers with lots of memory. However for some usecases where you are serving a lot of css and js assets, it's a nice help. The patch performs special treatment for http/1.1 pipelining, since the compression is performed on the fly the compressed content-length is not known until the end. So for h1 only, chunked transfer-encoding is automatically added so pipelining can continue of the connection. For h2 the chunking is neither supported nor required, so it "just works". User code can also request to add a compression transform before the reply headers were sent using the new api LWS_VISIBLE int lws_http_compression_apply(struct lws *wsi, const char *name, unsigned char **p, unsigned char *end, char decomp); ... this allows transparent compression of dynamically generated HTTP. The requested compression (eg, "deflate") is only applied if the client headers indicated it was supported, otherwise it's a NOP. Name may be NULL in which case the first compression method in the internal table at stream.c that is mentioned as acceptable by the client will be used. NOTE: the compression translation, same as h2 support, relies on the user code using LWS_WRITE_HTTP and then LWS_WRITE_HTTP_FINAL on the last part written. The internal lws fileserving code already does this.
2018-09-02 14:43:05 +08:00
if (lws_has_buffered_out(wsi)
#if defined(LWS_WITH_HTTP_STREAM_COMPRESSION)
|| wsi->http.comp_ctx.buflist_comp ||
wsi->http.comp_ctx.may_have_more
#endif
) {
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsl_info("%p: LRS_FLUSHING_BEFORE_CLOSE\n", wsi);
lwsi_set_state(wsi, LRS_FLUSHING_BEFORE_CLOSE);
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
__lws_set_timeout(wsi,
cleanups
2017-10-28 07:42:44 +08:00
PENDING_FLUSH_STORED_SEND_BEFORE_CLOSE, 5);
close should flush any truncated send before really closing Signed-off-by: Andy Green <andy.green@linaro.org>
2014-04-10 14:25:24 +08:00
return;
}
break;
}
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
if (lwsi_state(wsi) == LRS_WAITING_CONNECT ||
lwsi_state(wsi) == LRS_H1C_ISSUE_HANDSHAKE)
trac 40 client connection properly nonblocking As spotted by JM on Trac#40 http://libwebsockets.org/trac/libwebsockets/ticket/40 client connect didn't do anything about being truly nonblocking. This patch should hopefully solve that. Signed-off-by: Andy Green <andy.green@linaro.org>
2013-09-20 20:26:12 +08:00
goto just_kill_connection;
client: bind and drop protocol like server HTTP server protocols have had for a while LWS_CALLBACK_HTTP_DROP/BIND_PROTOCOL callbacks that mark when a wsi is attched to a protocol and detached. It turns out this is generally useful for everything to know when a wsi is joining a protocol and definitively completely finished with a protocol. Particularly with client wsi where you provided the userdata externally, this makes a clear point to free() it on the protocol binding being dropped. This patch adds protocol bind / unbind callbacks to the role definition and lets them operate on all roles. For the various roles HTTP server: LWS_CALLBACK_HTTP_BIND/DROP_PROTOCOL as before HTTP client: LWS_CALLBACK_CLIENT_HTTP_BIND/DROP_PROTOCOL ws server: LWS_CALLBACK_WS_SERVER_BIND/DROP_PROTOCOL ws client: LWS_CALLBACK_WS_CLIENT_BIND/DROP_PROTOCOL raw file: LWS_CALLBACK_RAW_FILE_BIND/DROP_PROTOCOL raw skt: LWS_CALLBACK_RAW_SKT_BIND/DROP_PROTOCOL
2018-08-18 14:11:29 +08:00
if (!wsi->told_user_closed && wsi->user_space && wsi->protocol &&
wsi->protocol_bind_balance) {
wsi->protocol->callback(wsi,
wsi->role_ops->protocol_unbind_cb[
!!lwsi_role_server(wsi)],
threadpool
2018-09-02 14:35:37 +08:00
wsi->user_space, (void *)__func__, 0);
client: bind and drop protocol like server HTTP server protocols have had for a while LWS_CALLBACK_HTTP_DROP/BIND_PROTOCOL callbacks that mark when a wsi is attched to a protocol and detached. It turns out this is generally useful for everything to know when a wsi is joining a protocol and definitively completely finished with a protocol. Particularly with client wsi where you provided the userdata externally, this makes a clear point to free() it on the protocol binding being dropped. This patch adds protocol bind / unbind callbacks to the role definition and lets them operate on all roles. For the various roles HTTP server: LWS_CALLBACK_HTTP_BIND/DROP_PROTOCOL as before HTTP client: LWS_CALLBACK_CLIENT_HTTP_BIND/DROP_PROTOCOL ws server: LWS_CALLBACK_WS_SERVER_BIND/DROP_PROTOCOL ws client: LWS_CALLBACK_WS_CLIENT_BIND/DROP_PROTOCOL raw file: LWS_CALLBACK_RAW_FILE_BIND/DROP_PROTOCOL raw skt: LWS_CALLBACK_RAW_SKT_BIND/DROP_PROTOCOL
2018-08-18 14:11:29 +08:00
wsi->protocol_bind_balance = 0;
introduce urlencode decode and sql escape public apis This adds - simple lws_urlencode() - simple lws_urldecode() - simple lws_sql_purify Those expect the data to all be there and process it up until the first '\0'. There is also a larger opaque apis for handling POST_BODY urldecode. To enable these, you need to give cmake -DLWS_WITH_STATEFUL_URLDECODE=1 (or arrange any larger feature that relies on it sets that in CMakeLists.txt) - stateful urldecode with parameter array These have create / process / destroy semantics on a struct that maintains decode state. Stateful urldecode is capable of dealing with large POST data in multiple POST_BODY callbacks cleanly, eg, file transfer by POST. Stateful urldecode with parameter array wraps the above with a canned callback that stores the urldecoded data and indexes them in a pointer array matching an array of parameter names. You may also pass it an optional callback when creating it, that will recieve uploaded file content. The test html is updated to support both urlencoded and multipart forms, with some javascript to do clientside validation of an arbitrary 100KB file size limit (there is no file size limit in the apis). Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-08 10:07:02 +08:00
}
Callback with reason LWS_CALLBACK_CLOSED_HTTP when HTTP closed before serve. If the remote HTTP client closes the connection before serving commences, the only notification client code receives is LWS_CALLBACK_WSI_DESTROY. This commit gives LWS_CALLBACK_CLOSED_HTTP as well, as would happen if HTTP serving had already commenced.
2015-01-28 04:15:13 +08:00
add extra state for waiting on close ack with timeout Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 07:08:12 +00:00
/*
api rationalization use new names internally Change all internal uses of rationalized public apis to reflect the new names. Theer are a few things that got changed as side effect of search/replace matches, but these are almost all internal. I added a compatibility define for the public enum that got renamed. Theoretically existing code should not notice the difference from these two patches. And new code will find the new names. https://github.com/warmcat/libwebsockets/issues/357 Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 08:43:54 +08:00
* signal we are closing, lws_write will
add extra state for waiting on close ack with timeout Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 07:08:12 +00:00
* add any necessary version-specific stuff. If the write fails,
* no worries we are closing anyway. If we didn't initiate this
* close, then our state has been changed to
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
* LRS_RETURNED_CLOSE and we will skip this.
add extra state for waiting on close ack with timeout Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 07:08:12 +00:00
*
* Likewise if it's a second call to close this connection after we
* sent the close indication to the peer already, we are in state
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
* LRS_AWAITING_CLOSE_ACK and will skip doing this a second time.
add extra state for waiting on close ack with timeout Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 07:08:12 +00:00
*/
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
if (wsi->role_ops->close_via_role_protocol &&
wsi->role_ops->close_via_role_protocol(wsi, reason))
close: make close notification go through writable Until now we took the approach if just writing the close notification broke something, we didn't care because we were closing the connection anyway. But with lws_meta, breaking stuff in the parent connection would be a sticky problem outliving the closing child connection. So this adds a new wsi state LWSS_WAITING_TO_SEND_CLOSE_NOTIFICATION and makes the send go via the writable callback mechanism.
2017-07-17 10:11:17 +08:00
return;
add extra state for waiting on close ack with timeout Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 07:08:12 +00:00
flush extensions when closing Extensions might be caching stuff that we should spill before a controlled close. It's not allowed to send anything on the wire after the close request, so we need to make the extensions spill just before. Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 07:08:18 +00:00
just_kill_connection:
esp8266 initial support
2016-07-23 14:18:25 +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
if (wsi->role_ops->close_kill_connection)
wsi->role_ops->close_kill_connection(wsi, reason);
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
lws_remove_child_from_any_parent(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
n = 0;
adopt: allow binding to parent at same time - if protocol set, allocate own user_space If the child wsi wants the parent wsi user_space, it can use lws_wsi_user(lws_get_parent(child_wsi)) - raw file close processing handles parent-child relationship
2017-03-03 07:36:08 +08:00
DROP_PROTOCOL: guarantee only one DROP per BIND
2018-04-03 10:37:14 +08:00
if (!wsi->told_user_closed && wsi->user_space &&
wsi->protocol_bind_balance) {
lwsl_debug("%s: %p: DROP_PROTOCOL %s\n", __func__, wsi,
client: reorder close callbacks This will make sure you get - DROP_PROTOCOL - CONNECTION_ERROR - CLOSED_CLIENT_HTTP as expected for failed client connections.
2017-10-15 12:15:28 +08:00
wsi->protocol->name);
client: bind and drop protocol like server HTTP server protocols have had for a while LWS_CALLBACK_HTTP_DROP/BIND_PROTOCOL callbacks that mark when a wsi is attched to a protocol and detached. It turns out this is generally useful for everything to know when a wsi is joining a protocol and definitively completely finished with a protocol. Particularly with client wsi where you provided the userdata externally, this makes a clear point to free() it on the protocol binding being dropped. This patch adds protocol bind / unbind callbacks to the role definition and lets them operate on all roles. For the various roles HTTP server: LWS_CALLBACK_HTTP_BIND/DROP_PROTOCOL as before HTTP client: LWS_CALLBACK_CLIENT_HTTP_BIND/DROP_PROTOCOL ws server: LWS_CALLBACK_WS_SERVER_BIND/DROP_PROTOCOL ws client: LWS_CALLBACK_WS_CLIENT_BIND/DROP_PROTOCOL raw file: LWS_CALLBACK_RAW_FILE_BIND/DROP_PROTOCOL raw skt: LWS_CALLBACK_RAW_SKT_BIND/DROP_PROTOCOL
2018-08-18 14:11:29 +08:00
wsi->protocol->callback(wsi,
wsi->role_ops->protocol_unbind_cb[
!!lwsi_role_server(wsi)],
threadpool
2018-09-02 14:35:37 +08:00
wsi->user_space, (void *)__func__, 0);
DROP_PROTOCOL: guarantee only one DROP per BIND
2018-04-03 10:37:14 +08:00
wsi->protocol_bind_balance = 0;
client: reorder close callbacks This will make sure you get - DROP_PROTOCOL - CONNECTION_ERROR - CLOSED_CLIENT_HTTP as expected for failed client connections.
2017-10-15 12:15:28 +08:00
}
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
if ((lwsi_state(wsi) == LRS_WAITING_SERVER_REPLY ||
lwsi_state(wsi) == LRS_WAITING_CONNECT) && !wsi->already_did_cce)
client: http1.1 pipelining
2018-03-26 12:05:04 +08:00
wsi->protocol->callback(wsi,
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
LWS_CALLBACK_CLIENT_CONNECTION_ERROR,
client: reorder close callbacks This will make sure you get - DROP_PROTOCOL - CONNECTION_ERROR - CLOSED_CLIENT_HTTP as expected for failed client connections.
2017-10-15 12:15:28 +08:00
wsi->user_space, NULL, 0);
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
/*
* Testing with ab shows that we have to stage the socket close when
* the system is under stress... shutdown any further TX, change the
* state to one that won't emit anything more, and wait with a timeout
* for the POLLIN to show a zero-size rx before coming back and doing
* the actual close.
*/
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
if (wsi->role_ops != &role_ops_raw_skt && !lwsi_role_client(wsi) &&
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_state(wsi) != LRS_SHUTDOWN &&
lwsi_state(wsi) != LRS_UNCONNECTED &&
libuv integration This gets the libuv stuff plumbed in and working. Currently it's only workable for some service thread, and there is an isolated valgrind problem left ==28425== 128 bytes in 1 blocks are definitely lost in loss record 3 of 3 ==28425== at 0x4C28C50: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x4C2AB1E: realloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x58BBB27: maybe_resize (core.c:748) ==28425== by 0x58BBB27: uv__io_start (core.c:787) ==28425== by 0x58C1B80: uv__signal_loop_once_init (signal.c:225) ==28425== by 0x58C1B80: uv_signal_init (signal.c:260) ==28425== by 0x58BF7A6: uv_loop_init (loop.c:66) ==28425== by 0x4157F5: lws_uv_initloop (libuv.c:89) ==28425== by 0x405536: main (test-server-libuv.c:284) libuv wants to sign off on all libuv 'handles' that will close, and callback to do the close confirmation asynchronously. The wsi close function is adapted when libuv is in use to work with libuv accordingly and exit the uv loop the number of remaining wsi is zero. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-14 09:27:41 +08:00
reason != LWS_CLOSE_STATUS_NOSTATUS_CONTEXT_DESTROY &&
!wsi->socket_is_permanently_unusable) {
tls: split out common, openssl and mbedtls code - introduce lib/tls/mbedtls lib/tls/openssl - move wrapper into lib/tls/mbedtls/wrapper - introduce private helpers to hide backend This patch doesn't replace or remove the wrapper, it moves it to lib/tls/mbedtls/wrapper. But it should be now that the ONLY functions directly consuming wrapper apis are isolated in - lib/tls/mbedtls/client.c (180 lines) - lib/tls/mbedtls/server.c (317 lines) - lib/tls/mbedtls/ssl.c (325 lines) In particular there are no uses of openssl or mbedtls-related constants outside of ./lib/tls any more.
2017-10-18 09:41:44 +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
#if defined(LWS_WITH_TLS)
refactor: also migrate tls to the ops struct and private.h pattern Several new ops are planned for tls... so better to bite the bullet and clean it out to the same level as roles + event-libs first. Also adds a new travis target "mbedtls" and all the tests except autobahn against mbedtls build.
2018-05-01 12:41:42 +08:00
if (lws_is_ssl(wsi) && wsi->tls.ssl) {
tls: split out common, openssl and mbedtls code - introduce lib/tls/mbedtls lib/tls/openssl - move wrapper into lib/tls/mbedtls/wrapper - introduce private helpers to hide backend This patch doesn't replace or remove the wrapper, it moves it to lib/tls/mbedtls/wrapper. But it should be now that the ONLY functions directly consuming wrapper apis are isolated in - lib/tls/mbedtls/client.c (180 lines) - lib/tls/mbedtls/server.c (317 lines) - lib/tls/mbedtls/ssl.c (325 lines) In particular there are no uses of openssl or mbedtls-related constants outside of ./lib/tls any more.
2017-10-18 09:41:44 +08:00
n = 0;
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
switch (__lws_tls_shutdown(wsi)) {
tls: split out common, openssl and mbedtls code - introduce lib/tls/mbedtls lib/tls/openssl - move wrapper into lib/tls/mbedtls/wrapper - introduce private helpers to hide backend This patch doesn't replace or remove the wrapper, it moves it to lib/tls/mbedtls/wrapper. But it should be now that the ONLY functions directly consuming wrapper apis are isolated in - lib/tls/mbedtls/client.c (180 lines) - lib/tls/mbedtls/server.c (317 lines) - lib/tls/mbedtls/ssl.c (325 lines) In particular there are no uses of openssl or mbedtls-related constants outside of ./lib/tls any more.
2017-10-18 09:41:44 +08:00
case LWS_SSL_CAPABLE_DONE:
case LWS_SSL_CAPABLE_ERROR:
case LWS_SSL_CAPABLE_MORE_SERVICE_READ:
case LWS_SSL_CAPABLE_MORE_SERVICE_WRITE:
case LWS_SSL_CAPABLE_MORE_SERVICE:
break;
}
} else
ssl close: do explicit ssl shutdown instead of socket shutdown if ssl mode
2017-02-20 05:44:56 +08:00
#endif
{
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
lwsl_info("%s: shutdown conn: %p (sk %d, state 0x%x)\n",
clean
2017-09-23 12:55:21 +08:00
__func__, wsi, (int)(long)wsi->desc.sockfd,
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_state(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
if (!wsi->socket_is_permanently_unusable &&
windows: cleanup wrong and duplicated socket validity helpers https://github.com/warmcat/libwebsockets/issues/1259
2018-05-06 07:19:21 +08:00
lws_socket_is_valid(wsi->desc.sockfd)) {
close: only tell user closed once
2018-01-03 09:22:32 +08:00
wsi->socket_is_permanently_unusable = 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 = shutdown(wsi->desc.sockfd, SHUT_WR);
close: only tell user closed once
2018-01-03 09:22:32 +08:00
}
ssl close: do explicit ssl shutdown instead of socket shutdown if ssl mode
2017-02-20 05:44:56 +08:00
}
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
if (n)
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsl_debug("closing: shutdown (state 0x%x) ret %d\n",
lwsi_state(wsi), LWS_ERRNO);
defeat POLLOUT if socket in shutdown wait After andrejs.hanins@ubnt.com Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-24 21:27:46 +08:00
clean
2017-09-23 12:55:21 +08:00
/*
* This causes problems on WINCE / ESP32 with disconnection
* when the events are half closing connection
*/
esp32: separate factory setup
2017-03-16 10:46:31 +08:00
#if !defined(_WIN32_WCE) && !defined(LWS_WITH_ESP32)
libuv when in use skip shutdown close phase Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-29 10:34:29 +08:00
/* libuv: no event available to guarantee 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
if (!wsi->socket_is_permanently_unusable &&
windows: cleanup wrong and duplicated socket validity helpers https://github.com/warmcat/libwebsockets/issues/1259
2018-05-06 07:19:21 +08:00
lws_socket_is_valid(wsi->desc.sockfd) &&
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_state(wsi) != LRS_SHUTDOWN &&
refactor: apply ops structs to event loop handlers
2018-04-29 10:44:36 +08:00
context->event_loop_ops->periodic_events_available) {
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
__lws_change_pollfd(wsi, LWS_POLLOUT, LWS_POLLIN);
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_set_state(wsi, LRS_SHUTDOWN);
h2 ws: fixes against chrome
2018-03-11 11:26:06 +08:00
__lws_set_timeout(wsi, PENDING_TIMEOUT_SHUTDOWN_FLUSH,
refactor: apply ops structs to event loop handlers
2018-04-29 10:44:36 +08:00
context->timeout_secs);
esp32: separate factory setup
2017-03-16 10:46:31 +08:00
libuv when in use skip shutdown close phase Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-29 10:34:29 +08:00
return;
}
wince minor adaptations https://github.com/warmcat/libwebsockets/issues/444
2016-02-29 18:48:55 +08:00
#endif
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
}
clean
2017-09-23 12:55:21 +08:00
lwsl_debug("%s: real just_kill_connection: %p (sockfd %d)\n", __func__,
wsi, wsi->desc.sockfd);
esp8266 initial support
2016-07-23 14:18:25 +08:00
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
#ifdef LWS_WITH_HUBBUB
refactor: move everything related to HTTP_PROXY out of struct lws and private-libwebsockets.h into http role private
2018-04-27 14:36:10 +08:00
if (wsi->http.rw) {
lws_rewrite_destroy(wsi->http.rw);
wsi->http.rw = NULL;
proxy rewrite If you enable -DLWS_WITH_HTTP_PROXY=1 at cmake, the test server has a new URI path http://localhost:7681/proxytest If you visit here, a client connection to http://example.com:80 is spawned, and the results piped on to your original connection. Also with LWS_WITH_HTTP_PROXY enabled at cmake, lws wants to link to an additional library, "libhubbub". This allows lws to do html rewriting on the fly, adjusting proxied urls in a lightweight and fast way.
2016-03-20 11:59:53 +08:00
}
cgi env Improve cgi support so it's capable of running cgit
2016-03-20 11:55:25 +08:00
#endif
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 (wsi->http.pending_return_headers)
lws_free_set_NULL(wsi->http.pending_return_headers);
add extra state for waiting on close ack with timeout Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 07:08:12 +00:00
/*
* we won't be servicing or receiving anything further from this guy
replace hashtable polltable management This rips out the connection hashtable implementation along with MAX_CLIENTS and replaces it with a dynamically allocated fds array and lookup table along the same lines as the new extpoll implementation from Edwin van den Oetelaar. It detects the max number of file descriptors possible at context init time and allocates accordingly; this can be externally controlled by ulimit and the server run as a specific user to facilitate targeting specific ulimit rules at it. Many operations that translated between socket descriptors and struct websocket or pollfd objects have had iteration removed by this patch and under load will be a lot faster. Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-17 12:26:48 +08:00
* delete socket from the internal poll list if still present
add extra state for waiting on close ack with timeout Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-07 07:08:12 +00:00
*/
pthreads: more locking
2018-03-05 16:49:28 +08:00
__lws_ssl_remove_wsi_from_buffered_list(wsi);
__lws_remove_from_timeout_list(wsi);
lws_set_timer_usecs: change to usec resolution This replaces the existing, unreleased lws_set_timer(wsi, secs) with lws_set_timer_usecs(wsi, usecs). wsi with a timer waiting are added to a linked-list sorted by the timer trigger time. 1) poll() timeout (ie, poll wait) is trimmed to the nearest ms of the first waiting timer if the default poll wait is longer than the interval until the first waiting timer. The linked-list of waiting timers is checked every entry and exit from poll()... if no timers waiting or none reached their time this costs almost nothing. 2) libuv: the earliest hrtimer is checked after every IO, again this is costing nothing if the list head is NULL. If the case there are hrtimers on the list, it costs a getimeofday (a VDSO in linux) and more only if any of the timers have fired. In addition on entry to libuv idle, if there are any waiting hrtimers on the list, a libuv timer is used to force a wake in case we stay idle (the libuv timer has ms resolution). 3) libev: not implemented 4) libevent: not implemented Warnings are logged in the api is used on an event backend without support. Patches welcome to add support similarly to libuv.
2018-03-19 16:37:37 +08:00
lws_dll_lws_remove(&wsi->dll_hrtimer);
ssl pending buffered reads use linked list Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-29 08:36:18 +08:00
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
/* don't repeat event loop stuff */
if (wsi->told_event_loop_closed)
return;
cleanups after api changes and mbed update Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 16:54:12 +08:00
/* checking return redundant since we anyway close */
adoption: make union for socket and file fds This lets lws support adopting raw file FDs and raw socket fds. A test plugin creates a FIFO and prints data sent on it, using the lws event loop.
2017-02-27 12:55:56 +08:00
if (wsi->desc.sockfd != LWS_SOCK_INVALID)
pthreads: more locking
2018-03-05 16:49:28 +08:00
__remove_wsi_socket_from_fds(wsi);
lws-meta
2017-07-19 04:39:14 +08:00
else
client: oom4: unify with __lws_free_wsi
2018-10-06 08:00:32 +08:00
__lws_same_vh_protocol_remove(wsi);
unify server and client close Signed-off-by: Andy Green <andy@warmcat.com>
2011-02-14 08:03:48 +00:00
refactor mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsi_set_state(wsi, LRS_DEAD_SOCKET);
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
lws_buflist_destroy_all_segments(&wsi->buflist);
lws_dll_lws_remove(&wsi->dll_buflist);
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
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
if (wsi->role_ops->close_role)
wsi->role_ops->close_role(pt, wsi);
extension permessage deflate Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-11 11:34:01 +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
/* tell the user it's all over for this guy */
http2 can keep upgraded connection up Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-08 12:00:53 +08:00
client: allow pipelined http to call LWS_CALLBACK_CLOSED_CLIENT_HTTP
2018-08-15 12:49:32 +08:00
if ((lwsi_state_est_PRE_CLOSE(wsi) ||
lwsi_state_PRE_CLOSE(wsi) == LRS_WAITING_SERVER_REPLY) &&
!wsi->told_user_closed &&
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
wsi->role_ops->close_cb[lwsi_role_server(wsi)]) {
const struct lws_protocols *pro = wsi->protocol;
cleanups after api changes and mbed update Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 16:54:12 +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
if (!wsi->protocol)
pro = &wsi->vhost->protocols[0];
minimal-raw-netcat Adapt attack.sh to use it instead of OS netcat and fox various bugs that exposed.
2018-05-02 08:46:16 +08:00
h2 client: no close callback for network stream
2018-05-24 11:26:09 +08:00
if (!wsi->upgraded_to_http2 || !lwsi_role_client(wsi))
/*
* The network wsi for a client h2 connection shouldn't
* call back for its role: the child stream connections
* own the role. Otherwise h2 will call back closed
* one too many times as the children do it and then
* the closing network stream.
*/
pro->callback(wsi,
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
wsi->role_ops->close_cb[lwsi_role_server(wsi)],
wsi->user_space, NULL, 0);
wsi->told_user_closed = 1;
remove fixed rx buffer allow definition per protocol A new protocol member is defined that controls the size of rx buffer allocation per connection. For compatibility 0 size allocates 4096, but you should adapt your protocol definition array in the user code to declare an appropriate value. See the changelog for more detail. The advantage is the rx frame buffer size is now tailored to what is expected from the protocol, rather than being fixed to a default of 4096. If your protocol only sends frames of a dozen bytes this allows you to only allocate an rx frame buffer of the same size. For example the per-connection allocation (excluding headers) for the test server fell from ~4500 to < 750 bytes with this. Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-06 21:10:16 +09:00
}
fops-zip
2017-03-03 12:38:10 +08:00
async_close:
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
wsi->socket_is_permanently_unusable = 1;
refactor: apply ops structs to event loop handlers
2018-04-29 10:44:36 +08:00
if (wsi->context->event_loop_ops->wsi_logical_close)
if (wsi->context->event_loop_ops->wsi_logical_close(wsi))
lws-meta
2017-07-19 04:39:14 +08:00
return;
libuv integration This gets the libuv stuff plumbed in and working. Currently it's only workable for some service thread, and there is an isolated valgrind problem left ==28425== 128 bytes in 1 blocks are definitely lost in loss record 3 of 3 ==28425== at 0x4C28C50: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x4C2AB1E: realloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x58BBB27: maybe_resize (core.c:748) ==28425== by 0x58BBB27: uv__io_start (core.c:787) ==28425== by 0x58C1B80: uv__signal_loop_once_init (signal.c:225) ==28425== by 0x58C1B80: uv_signal_init (signal.c:260) ==28425== by 0x58BF7A6: uv_loop_init (loop.c:66) ==28425== by 0x4157F5: lws_uv_initloop (libuv.c:89) ==28425== by 0x405536: main (test-server-libuv.c:284) libuv wants to sign off on all libuv 'handles' that will close, and callback to do the close confirmation asynchronously. The wsi close function is adapted when libuv is in use to work with libuv accordingly and exit the uv loop the number of remaining wsi is zero. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-14 09:27:41 +08:00
pthreads: more locking
2018-03-05 16:49:28 +08:00
__lws_close_free_wsi_final(wsi);
libuv integration This gets the libuv stuff plumbed in and working. Currently it's only workable for some service thread, and there is an isolated valgrind problem left ==28425== 128 bytes in 1 blocks are definitely lost in loss record 3 of 3 ==28425== at 0x4C28C50: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x4C2AB1E: realloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x58BBB27: maybe_resize (core.c:748) ==28425== by 0x58BBB27: uv__io_start (core.c:787) ==28425== by 0x58C1B80: uv__signal_loop_once_init (signal.c:225) ==28425== by 0x58C1B80: uv_signal_init (signal.c:260) ==28425== by 0x58BF7A6: uv_loop_init (loop.c:66) ==28425== by 0x4157F5: lws_uv_initloop (libuv.c:89) ==28425== by 0x405536: main (test-server-libuv.c:284) libuv wants to sign off on all libuv 'handles' that will close, and callback to do the close confirmation asynchronously. The wsi close function is adapted when libuv is in use to work with libuv accordingly and exit the uv loop the number of remaining wsi is zero. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-14 09:27:41 +08:00
}
void
pthreads: more locking
2018-03-05 16:49:28 +08:00
__lws_close_free_wsi_final(struct lws *wsi)
libuv integration This gets the libuv stuff plumbed in and working. Currently it's only workable for some service thread, and there is an isolated valgrind problem left ==28425== 128 bytes in 1 blocks are definitely lost in loss record 3 of 3 ==28425== at 0x4C28C50: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x4C2AB1E: realloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==28425== by 0x58BBB27: maybe_resize (core.c:748) ==28425== by 0x58BBB27: uv__io_start (core.c:787) ==28425== by 0x58C1B80: uv__signal_loop_once_init (signal.c:225) ==28425== by 0x58C1B80: uv_signal_init (signal.c:260) ==28425== by 0x58BF7A6: uv_loop_init (loop.c:66) ==28425== by 0x4157F5: lws_uv_initloop (libuv.c:89) ==28425== by 0x405536: main (test-server-libuv.c:284) libuv wants to sign off on all libuv 'handles' that will close, and callback to do the close confirmation asynchronously. The wsi close function is adapted when libuv is in use to work with libuv accordingly and exit the uv loop the number of remaining wsi is zero. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-14 09:27:41 +08:00
{
int n;
role: dbus This adds support for the integrating libdbus into the lws event loop. Unlike the other roles, lws doesn't completely adopt the fd and libdbus insists to retain control over the fd lifecycle. However libdbus provides apis for foreign code (lws) to provide event loop services to libdbus for the fd. Accordingly, unlike the other roles rx and writeable are not subsumed into lws callback messages and the events remain the property of libdbus. A context struct wrapper is provided that is available in the libdbus callbacks to bridge between the lws and dbus worlds, along with a minimal example dbus client and server.
2018-10-02 10:50:24 +08:00
if (!wsi->shadow &&
lws_socket_is_valid(wsi->desc.sockfd) && !lws_ssl_close(wsi)) {
cgi: fix for https git server
2018-06-16 09:35:07 +08:00
lwsl_debug("%s: wsi %p: fd %d\n", __func__, wsi, wsi->desc.sockfd);
adoption: make union for socket and file fds This lets lws support adopting raw file FDs and raw socket fds. A test plugin creates a FIFO and prints data sent on it, using the lws event loop.
2017-02-27 12:55:56 +08:00
n = compatible_close(wsi->desc.sockfd);
refactor move ssl server service to ssl.c Signed-off-by: Andy Green <andy.green@linaro.org>
2014-04-12 10:07:02 +08:00
if (n)
lwsl_debug("closing: close ret %d\n", LWS_ERRNO);
mbed3 plat Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-02 20:34:12 +08:00
adoption: make union for socket and file fds This lets lws support adopting raw file FDs and raw socket fds. A test plugin creates a FIFO and prints data sent on it, using the lws event loop.
2017-02-27 12:55:56 +08:00
wsi->desc.sockfd = LWS_SOCK_INVALID;
add-wss-ssl-openssl-support.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-11-08 17:03:03 +00:00
}
add outermost wsi lifetime callbacks Signed-off-by: Andy Green <andy.green@linaro.org>
2014-02-15 19:25:50 +08:00
/* outermost destroy notification for wsi (user_space still intact) */
cancel_service: change to event-loop agnostic implementation - Add platform helpers for pipe creation. - Change the direct-to-fds implementation to create a wsi for each pt and use the normal apis to bind it to the event loop. - Modifiy context creation and destroy to create and remove the event pipe wsis. - Create the event pipe wsis during context create if using the default poll() event loop, or when the other event loops start otherwise. - Add handler that calls back user code with LWS_CALLBACK_EVENT_WAIT_CANCELLED This patch allows you to call `lws_cancel_service(struct lws_context *context)` from another thread. It's very cheap for the other thread to call and is safe without locking. Every use protocol receives a LWS_CALLBACK_EVENT_WAIT_CANCELLED from the main thread serialized normally in the event loop.
2017-10-24 11:59:44 +08:00
if (wsi->vhost)
wsi->vhost->protocols[0].callback(wsi, LWS_CALLBACK_WSI_DESTROY,
wsi->user_space, NULL, 0);
add outermost wsi lifetime callbacks Signed-off-by: Andy Green <andy.green@linaro.org>
2014-02-15 19:25:50 +08:00
cgi post Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-13 11:42:53 +08:00
#ifdef LWS_WITH_CGI
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
if (wsi->http.cgi) {
cgi: fix not getting POLLHUP on fd associated to CGI STDOUT Signed-off-by: Petar Paradzik <petar.paradzik@sartura.hr>
2017-07-28 18:01:14 +02:00
for (n = 0; n < 3; n++) {
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
if (wsi->http.cgi->pipe_fds[n][!!(n == 0)] == 0)
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
lwsl_err("ZERO FD IN CGI CLOSE");
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
if (wsi->http.cgi->pipe_fds[n][!!(n == 0)] >= 0)
close(wsi->http.cgi->pipe_fds[n][!!(n == 0)]);
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
}
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
lws_free(wsi->http.cgi);
cgi post Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-13 11:42:53 +08:00
}
#endif
pthreads: more locking
2018-03-05 16:49:28 +08:00
__lws_free_wsi(wsi);
}
void
lws_close_free_wsi(struct lws *wsi, enum lws_close_status reason, const char *caller)
{
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
lws_pt_lock(pt, __func__);
__lws_close_free_wsi(wsi, reason, caller);
lws_pt_unlock(pt);
add-per-session-user-data.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-11-03 11:13:06 +00:00
}
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
/* lws_buflist */
int
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
lws_buflist_append_segment(struct lws_buflist **head, const uint8_t *buf,
size_t len)
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
{
buflist: debug
2018-04-28 07:49:31 +08:00
struct lws_buflist *nbuf;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
int first = !*head;
rxflow buflist: handle forced service
2018-04-17 11:43:20 +08:00
void *p = *head;
buflist: debug
2018-04-28 07:49:31 +08:00
int sanity = 1024;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
assert(buf);
assert(len);
/* append at the tail */
buflist: debug
2018-04-28 07:49:31 +08:00
while (*head) {
if (!--sanity || head == &((*head)->next)) {
lwsl_err("%s: corrupt list points to self\n", __func__);
return -1;
}
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
head = &((*head)->next);
buflist: debug
2018-04-28 07:49:31 +08:00
}
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
rxflow buflist: handle forced service
2018-04-17 11:43:20 +08:00
lwsl_info("%s: len %u first %d %p\n", __func__, (uint32_t)len, first, p);
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
nbuf = (struct lws_buflist *)lws_malloc(sizeof(**head) + len, __func__);
buflist: debug
2018-04-28 07:49:31 +08:00
if (!nbuf) {
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
lwsl_err("%s: OOM\n", __func__);
return -1;
}
buflist: debug
2018-04-28 07:49:31 +08:00
nbuf->len = len;
nbuf->pos = 0;
nbuf->next = NULL;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
buflist: debug
2018-04-28 07:49:31 +08:00
p = (void *)nbuf->buf;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
memcpy(p, buf, len);
buflist: debug
2018-04-28 07:49:31 +08:00
*head = nbuf;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
return first; /* returns 1 if first segment just created */
}
static int
lws_buflist_destroy_segment(struct lws_buflist **head)
{
struct lws_buflist *old = *head;
assert(*head);
*head = (*head)->next;
buflist: debug
2018-04-28 07:49:31 +08:00
old->next = NULL;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
lws_free(old);
return !*head; /* returns 1 if last segment just destroyed */
}
void
lws_buflist_destroy_all_segments(struct lws_buflist **head)
{
struct lws_buflist *p = *head, *p1;
while (p) {
p1 = p->next;
buflist: debug
2018-04-28 07:49:31 +08:00
p->next = NULL;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
lws_free(p);
p = p1;
}
*head = NULL;
}
size_t
lws_buflist_next_segment_len(struct lws_buflist **head, uint8_t **buf)
{
if (!*head) {
if (buf)
*buf = NULL;
return 0;
}
if (!(*head)->len && (*head)->next)
lws_buflist_destroy_segment(head);
if (!*head) {
if (buf)
*buf = NULL;
return 0;
}
assert((*head)->pos < (*head)->len);
if (buf)
*buf = (*head)->buf + (*head)->pos;
return (*head)->len - (*head)->pos;
}
int
lws_buflist_use_segment(struct lws_buflist **head, size_t len)
{
assert(*head);
assert(len);
assert((*head)->pos + len <= (*head)->len);
(*head)->pos += len;
if ((*head)->pos == (*head)->len)
lws_buflist_destroy_segment(head);
if (!*head)
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
return (int)((*head)->len - (*head)->pos);
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
}
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
void
lws_buflist_describe(struct lws_buflist **head, void *id)
{
refactor: change event lib minimal examples to serve as the test apps
2018-04-30 19:17:32 +08:00
struct lws_buflist *old;
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
int n = 0;
if (*head == NULL)
lwsl_notice("%p: buflist empty\n", id);
while (*head) {
lwsl_notice("%p: %d: %llu / %llu (%llu left)\n", id, n,
(unsigned long long)(*head)->pos,
(unsigned long long)(*head)->len,
(unsigned long long)(*head)->len - (*head)->pos);
refactor: change event lib minimal examples to serve as the test apps
2018-04-30 19:17:32 +08:00
old = *head;
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
head = &((*head)->next);
refactor: change event lib minimal examples to serve as the test apps
2018-04-30 19:17:32 +08:00
if (*head == old) {
lwsl_err("%s: next points to self\n", __func__);
break;
}
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
n++;
}
}
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
/* ... */
lws_get_urlarg_by_name Adds a convenient way to directly get the value of a URL argument like ...?x=y&v=1, regardless of position in the parameter list. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 09:04:15 +08:00
LWS_VISIBLE LWS_EXTERN const char *
lws_get_urlarg_by_name(struct lws *wsi, const char *name, char *buf, int len)
{
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
int n = 0, sl = (int)strlen(name);
lws_get_urlarg_by_name Adds a convenient way to directly get the value of a URL argument like ...?x=y&v=1, regardless of position in the parameter list. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 09:04:15 +08:00
while (lws_hdr_copy_fragment(wsi, buf, len,
WSI_TOKEN_HTTP_URI_ARGS, n) >= 0) {
if (!strncmp(buf, name, sl))
return buf + sl;
n++;
}
return NULL;
}
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
#if !defined(LWS_WITH_ESP32)
lws_interface_to_sa Expose interface handling code as an api
2016-01-16 12:09:38 +08:00
LWS_VISIBLE int
clean
2017-09-23 12:55:21 +08:00
interface_to_sa(struct lws_vhost *vh, const char *ifname,
struct sockaddr_in *addr, size_t addrlen)
lws_interface_to_sa Expose interface handling code as an api
2016-01-16 12:09:38 +08:00
{
int ipv6 = 0;
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_IPV6
ipv6 move disable to vhost option Server ipv6 support disable is now controlled by vhost->options rather than context->options, allowing it to be set per-vhost. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 21:19:40 +08:00
ipv6 = LWS_IPV6_ENABLED(vh);
lws_interface_to_sa Expose interface handling code as an api
2016-01-16 12:09:38 +08:00
#endif
ipv6 move disable to vhost option Server ipv6 support disable is now controlled by vhost->options rather than context->options, allowing it to be set per-vhost. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 21:19:40 +08:00
(void)vh;
lws_interface_to_sa Expose interface handling code as an api
2016-01-16 12:09:38 +08:00
return lws_interface_to_sa(ipv6, ifname, addr, addrlen);
}
#endif
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#ifndef LWS_PLAT_OPTEE
ipv6 move disable to vhost option Server ipv6 support disable is now controlled by vhost->options rather than context->options, allowing it to be set per-vhost. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 21:19:40 +08:00
static int
lws_get_addresses(struct lws_vhost *vh, void *ads, char *name,
clean tidy the worst whitespace alignment probs due to mass token name length changes Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:30:53 +08:00
int name_len, char *rip, int rip_len)
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
{
struct addrinfo ai, *res;
Fix getaddrinfo memory leak
2015-10-15 13:02:03 +03:00
struct sockaddr_in addr4;
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
coverity 182068: 155650: unnecessary check against illegal NULL
2017-07-28 14:19:24 +08:00
rip[0] = '\0';
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
name[0] = '\0';
Fix getaddrinfo memory leak
2015-10-15 13:02:03 +03:00
addr4.sin_family = AF_UNSPEC;
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_IPV6
ipv6 move disable to vhost option Server ipv6 support disable is now controlled by vhost->options rather than context->options, allowing it to be set per-vhost. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 21:19:40 +08:00
if (LWS_IPV6_ENABLED(vh)) {
clean
2017-09-23 12:55:21 +08:00
if (!lws_plat_inet_ntop(AF_INET6,
&((struct sockaddr_in6 *)ads)->sin6_addr,
rip, rip_len)) {
gcc-format-strings: ipv6
2017-02-14 17:55:13 +08:00
lwsl_err("inet_ntop: %s", strerror(LWS_ERRNO));
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
return -1;
}
// Strip off the IPv4 to IPv6 header if one exists
if (strncmp(rip, "::ffff:", 7) == 0)
memmove(rip, rip + 7, strlen(rip) - 6);
cleanups
2017-10-28 07:42:44 +08:00
getnameinfo((struct sockaddr *)ads, sizeof(struct sockaddr_in6),
name, name_len, NULL, 0, 0);
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
return 0;
} else
#endif
{
Fix getaddrinfo memory leak
2015-10-15 13:02:03 +03:00
struct addrinfo *result;
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
memset(&ai, 0, sizeof ai);
ai.ai_family = PF_UNSPEC;
ai.ai_socktype = SOCK_STREAM;
ESP32 platform This is enough for all the test app features to work on ESP32 without SSL.
2017-02-18 17:26:40 +08:00
#if !defined(LWS_WITH_ESP32)
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
if (getnameinfo((struct sockaddr *)ads,
sizeof(struct sockaddr_in),
name, name_len, NULL, 0, 0))
return -1;
ESP32 platform This is enough for all the test app features to work on ESP32 without SSL.
2017-02-18 17:26:40 +08:00
#endif
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
Fix getaddrinfo memory leak
2015-10-15 13:02:03 +03:00
if (getaddrinfo(name, NULL, &ai, &result))
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
return -1;
Fix getaddrinfo memory leak
2015-10-15 13:02:03 +03:00
res = result;
while (addr4.sin_family == AF_UNSPEC && res) {
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
switch (res->ai_family) {
case AF_INET:
clean
2017-09-23 12:55:21 +08:00
addr4.sin_addr =
((struct sockaddr_in *)res->ai_addr)->sin_addr;
Fix getaddrinfo memory leak
2015-10-15 13:02:03 +03:00
addr4.sin_family = AF_INET;
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
break;
}
res = res->ai_next;
}
Fix getaddrinfo memory leak
2015-10-15 13:02:03 +03:00
freeaddrinfo(result);
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
}
Fix getaddrinfo memory leak
2015-10-15 13:02:03 +03:00
if (addr4.sin_family == AF_UNSPEC)
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
return -1;
introduce LWS_WARN_UNUSED_RESULT Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-20 16:56:06 +08:00
if (lws_plat_inet_ntop(AF_INET, &addr4.sin_addr, rip, rip_len) == NULL)
return -1;
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
return 0;
}
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
lws_get_peer_simple move to user api https://github.com/warmcat/libwebsockets/issues/537 Signed-off-by: Andy Green <andy@warmcat.com>
2016-05-25 08:04:52 +08:00
LWS_VISIBLE const char *
lws access log option and lwsws conf This adds the ability to store apache-compatible logs to a file given at vhost-creation time. lwsws conf can set it per-vhost using "access-log": "<filepath>" The feature defaults to disabled at cmake, it can be set independently but LWS_WITH_LWSWS set it on. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 12:00:23 +08:00
lws_get_peer_simple(struct lws *wsi, char *name, int namelen)
{
socklen_t len, olen;
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_IPV6
lws access log option and lwsws conf This adds the ability to store apache-compatible logs to a file given at vhost-creation time. lwsws conf can set it per-vhost using "access-log": "<filepath>" The feature defaults to disabled at cmake, it can be set independently but LWS_WITH_LWSWS set it on. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 12:00:23 +08:00
struct sockaddr_in6 sin6;
#endif
struct sockaddr_in sin4;
int af = AF_INET;
void *p, *q;
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
wsi = lws_get_network_wsi(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
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_IPV6
ipv6 move disable to vhost option Server ipv6 support disable is now controlled by vhost->options rather than context->options, allowing it to be set per-vhost. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 21:19:40 +08:00
if (LWS_IPV6_ENABLED(wsi->vhost)) {
lws access log option and lwsws conf This adds the ability to store apache-compatible logs to a file given at vhost-creation time. lwsws conf can set it per-vhost using "access-log": "<filepath>" The feature defaults to disabled at cmake, it can be set independently but LWS_WITH_LWSWS set it on. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 12:00:23 +08:00
len = sizeof(sin6);
p = &sin6;
af = AF_INET6;
q = &sin6.sin6_addr;
} else
#endif
{
len = sizeof(sin4);
p = &sin4;
q = &sin4.sin_addr;
}
olen = len;
adoption: make union for socket and file fds This lets lws support adopting raw file FDs and raw socket fds. A test plugin creates a FIFO and prints data sent on it, using the lws event loop.
2017-02-27 12:55:56 +08:00
if (getpeername(wsi->desc.sockfd, p, &len) < 0 || len > olen) {
lws access log option and lwsws conf This adds the ability to store apache-compatible logs to a file given at vhost-creation time. lwsws conf can set it per-vhost using "access-log": "<filepath>" The feature defaults to disabled at cmake, it can be set independently but LWS_WITH_LWSWS set it on. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 12:00:23 +08:00
lwsl_warn("getpeername: %s\n", strerror(LWS_ERRNO));
return NULL;
}
Fix inet_ntop linking error on Windows
2016-04-17 19:26:04 -03:00
return lws_plat_inet_ntop(af, q, name, namelen);
lws access log option and lwsws conf This adds the ability to store apache-compatible logs to a file given at vhost-creation time. lwsws conf can set it per-vhost using "access-log": "<filepath>" The feature defaults to disabled at cmake, it can be set independently but LWS_WITH_LWSWS set it on. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 12:00:23 +08:00
}
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#endif
lws access log option and lwsws conf This adds the ability to store apache-compatible logs to a file given at vhost-creation time. lwsws conf can set it per-vhost using "access-log": "<filepath>" The feature defaults to disabled at cmake, it can be set independently but LWS_WITH_LWSWS set it on. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 12:00:23 +08:00
Hide private library symbols Signed-off-by: Peter Pentchev <roam@ringlet.net>
2013-03-30 09:52:21 +08:00
LWS_VISIBLE void
public api remove superfluous context params API BREAK Extend the cleanout caused by wsi having a context pointer into the public api. There's no point keeping the 1.5 compatibility work, we have changed the api in several places and rebuilt wasn't going to be enough a while ago. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-16 18:19:08 +08:00
lws_get_peer_addresses(struct lws *wsi, lws_sockfd_type fd, char *name,
int name_len, char *rip, int rip_len)
introduce-network-connect-filter-callback.patch Signed-off-by: Andy Green <andy.green@linaro.org>
2011-02-13 08:37:12 +00:00
{
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#ifndef LWS_PLAT_OPTEE
cygwin compatibility Signed-off-by: Bob Roberts <bobroberts177@gmail.com>
2013-04-25 09:16:30 +08:00
socklen_t len;
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_IPV6
add ipv6 support (changed to support runtime disable + integration by andy@warmcat.com)
2014-03-24 16:09:25 +08:00
struct sockaddr_in6 sin6;
#endif
struct sockaddr_in sin4;
public api remove superfluous context params API BREAK Extend the cleanout caused by wsi having a context pointer into the public api. There's no point keeping the 1.5 compatibility work, we have changed the api in several places and rebuilt wasn't going to be enough a while ago. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-16 18:19:08 +08:00
struct lws_context *context = wsi->context;
add ipv6 support (changed to support runtime disable + integration by andy@warmcat.com)
2014-03-24 16:09:25 +08:00
int ret = -1;
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
void *p;
introduce-network-connect-filter-callback.patch Signed-off-by: Andy Green <andy.green@linaro.org>
2011-02-13 08:37:12 +00:00
rip[0] = '\0';
name[0] = '\0';
change get_peer_addresses to use context wsi latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-30 08:12:20 +08:00
lws_latency_pre(context, wsi);
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_IPV6
ipv6 move disable to vhost option Server ipv6 support disable is now controlled by vhost->options rather than context->options, allowing it to be set per-vhost. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 21:19:40 +08:00
if (LWS_IPV6_ENABLED(wsi->vhost)) {
add ipv6 support (changed to support runtime disable + integration by andy@warmcat.com)
2014-03-24 16:09:25 +08:00
len = sizeof(sin6);
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
p = &sin6;
add ipv6 support (changed to support runtime disable + integration by andy@warmcat.com)
2014-03-24 16:09:25 +08:00
} else
Re: Windows port of your libwebsocket The latest code changed need some win32 fixes: Signed-off-by: Peter Hinz <cerebusrc@gmail.com>
2011-03-10 18:14:01 +00:00
#endif
add ipv6 support (changed to support runtime disable + integration by andy@warmcat.com)
2014-03-24 16:09:25 +08:00
{
len = sizeof(sin4);
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
p = &sin4;
}
introduce-network-connect-filter-callback.patch Signed-off-by: Andy Green <andy.green@linaro.org>
2011-02-13 08:37:12 +00:00
gethostbyname removal from lws get peer addresses This replaces gethostbyname in libwebsockets_get_peer_addresses and in the client handshake path. There's one left in lws-plat-win but that can be done another time. Let's see how much damage that did to the cross-platform and option builds... Signed-off-by: Andy Green <andy.green@linaro.org>
2015-01-28 21:03:49 +08:00
if (getpeername(fd, p, &len) < 0) {
lwsl_warn("getpeername: %s\n", strerror(LWS_ERRNO));
goto bail;
introduce-network-connect-filter-callback.patch Signed-off-by: Andy Green <andy.green@linaro.org>
2011-02-13 08:37:12 +00:00
}
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
ipv6 move disable to vhost option Server ipv6 support disable is now controlled by vhost->options rather than context->options, allowing it to be set per-vhost. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 21:19:40 +08:00
ret = lws_get_addresses(wsi->vhost, p, name, name_len, rip, rip_len);
change get_peer_addresses to use context wsi latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-30 08:12:20 +08:00
bail:
api rationalization use new names internally Change all internal uses of rationalized public apis to reflect the new names. Theer are a few things that got changed as side effect of search/replace matches, but these are almost all internal. I added a compatibility define for the public enum that got renamed. Theoretically existing code should not notice the difference from these two patches. And new code will find the new names. https://github.com/warmcat/libwebsockets/issues/357 Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 08:43:54 +08:00
lws_latency(context, wsi, "lws_get_peer_addresses", ret, 1);
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#endif
mbed3 plat Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-02 20:34:12 +08:00
(void)wsi;
(void)fd;
(void)name;
(void)name_len;
(void)rip;
(void)rip_len;
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
introduce-network-connect-filter-callback.patch Signed-off-by: Andy Green <andy.green@linaro.org>
2011-02-13 08:37:12 +00:00
}
export-libwebsocket_service_fd-api.patch Signed-off-by: Andy Green <andy.green@linaro.org>
2011-02-12 11:57:45 +00:00
vhost: allow user pointer to be attached at creation time https://github.com/warmcat/libwebsockets/issues/1008
2017-08-29 15:37:16 +08:00
LWS_EXTERN void *
lws_vhost_user(struct lws_vhost *vhost)
{
return vhost->user;
}
libwebsocket_context: add userspace pointer for use before wsi creation Signed-off-by: Alon Levy <alevy@redhat.com>
2012-10-19 11:21:56 +02:00
LWS_EXTERN void *
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
lws_context_user(struct lws_context *context)
libwebsocket_context: add userspace pointer for use before wsi creation Signed-off-by: Alon Levy <alevy@redhat.com>
2012-10-19 11:21:56 +02:00
{
style cleanup This brings the library sources into compliance with checkpatch style except for three or four exceptions like WIN32 related stuff and one long string constant I don't want to break into multiple sprintf calls. There should be no functional or compilability change from all this (hopefully). Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-11 17:13:32 +08:00
return context->user_space;
libwebsocket_context: add userspace pointer for use before wsi creation Signed-off-by: Alon Levy <alevy@redhat.com>
2012-10-19 11:21:56 +02:00
}
plugins This adds support for dynamically loaded plugins at runtime, which can expose their own protocols or extensions transparently. With these changes lwsws defaults to OFF in cmake, and if enabled it automatically enables plugins and libuv support. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-06 16:15:40 +08:00
LWS_VISIBLE struct lws_vhost *
lws_vhost_get(struct lws *wsi)
{
return wsi->vhost;
}
deprecate misnamed lws_vhost_get add lws_get_vhost All the other simple accessors begin lws_get_... Signed-off-by: Andy Green <andy@warmcat.com>
2016-05-17 13:47:44 +08:00
LWS_VISIBLE struct lws_vhost *
lws_get_vhost(struct lws *wsi)
{
return wsi->vhost;
}
plugins This adds support for dynamically loaded plugins at runtime, which can expose their own protocols or extensions transparently. With these changes lwsws defaults to OFF in cmake, and if enabled it automatically enables plugins and libuv support. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-06 16:15:40 +08:00
LWS_VISIBLE const struct lws_protocols *
lws_protocol_get(struct lws *wsi)
{
return wsi->protocol;
}
udp
2018-03-24 08:07:00 +08:00
LWS_VISIBLE const struct lws_udp *
lws_get_udp(const struct lws *wsi)
{
return wsi->udp;
}
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_VISIBLE struct lws *
lws_get_network_wsi(struct lws *wsi)
{
if (!wsi)
return NULL;
#if defined(LWS_WITH_HTTP2)
lws_diskcache: split generic parts from gitohashi into lws
2018-11-12 15:24:42 +08:00
if (!wsi->http2_substream
#if !defined(LWS_NO_CLIENT)
&& !wsi->client_h2_substream
#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
return 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
while (wsi->h2.parent_wsi)
wsi = wsi->h2.parent_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
#endif
return wsi;
}
genaes: generic AES layer independent of tls library Although RSA can be used directly for signing / JWS on large chunks of data since it's only operating on the hash, when JWE support arrives, which allows bulk encryption, it's going to be mandatory to support secondary AES ciphers to use on the bulk data. This adds generic support for all AES modes that OpenSSL and mbedTLS have in common, works on both mbedTLS and OpenSSL the same, and adds unit tests for each mode in api-test-gencrypto, to run in CI.
2018-12-04 08:03:58 +08:00
LWS_VISIBLE void
lws_explicit_bzero(void *p, size_t len)
{
volatile uint8_t *vp = p;
while (len--)
*vp++ = 0;
}
lws_raw_transaction_completed This provides a way to defer closing if the output buflist has unsent content for the wsi, until the buflist is drained. It doesn't make any assumption about the content being related to http, so you can use it on raw. It follows the semantics of the http transaction completed, ie if (lws_raw_transaction_completed(wsi)) return -1 return 0;
2018-12-01 06:45:23 +08:00
LWS_VISIBLE int LWS_WARN_UNUSED_RESULT
lws_raw_transaction_completed(struct lws *wsi)
{
if (lws_has_buffered_out(wsi)) {
/*
* ...so he tried to send something large, but it went out
* as a partial, but he immediately called us to say he wants
* to close the connection.
*
* Defer the close until the last part of the partial is sent.
*
*/
lwsl_debug("%s: %p: deferring due to partial\n", __func__, wsi);
wsi->close_when_buffered_out_drained = 1;
lws_callback_on_writable(wsi);
return 0;
}
return -1;
}
client: allow http[s] to select targeted protocol name in vhost
2017-02-15 09:12:39 +08:00
LWS_VISIBLE LWS_EXTERN const struct lws_protocols *
lws_vhost_name_to_protocol(struct lws_vhost *vh, const char *name)
{
int n;
for (n = 0; n < vh->count_protocols; n++)
if (!strcmp(name, vh->protocols[n].name))
return &vh->protocols[n];
return NULL;
}
libwebsocket_callback_all_protocol Signed-off-by: Andy Green <andy.green@linaro.org>
2014-02-15 16:36:38 +08:00
LWS_VISIBLE int
make protocols const require explicit context API BREAK The user protocols struct has not been const until now. This has been painful for a while because the semantics of the protocols struct look like it's going to be treated as const. At context creation, the protocols struct has been getting marked with the context, and three apis exploited that to only need to be passed a pointer to a protocol to get access to the context. This patch removes the two writeable members in the context (these were never directly used by user code), changes all pointers to protocols to be const, and adds an explicit first argument to the three affected apis so they can have access to context. The three affected apis are these LWS_VISIBLE LWS_EXTERN int -lws_callback_on_writable_all_protocol(const struct lws_protocols *protocol); +lws_callback_on_writable_all_protocol(const struct lws_context *context, + const struct lws_protocols *protocol); LWS_VISIBLE LWS_EXTERN int -lws_callback_all_protocol(const struct lws_protocols *protocol, int reason); +lws_callback_all_protocol(struct lws_context *context, + const struct lws_protocols *protocol, int reason); LWS_VISIBLE LWS_EXTERN void -lws_rx_flow_allow_all_protocol(const struct lws_protocols *protocol); +lws_rx_flow_allow_all_protocol(const struct lws_context *context, + const struct lws_protocols *protocol); unfortunately the original apis can no longer be emulated and users of them must update. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-11 10:45:35 +08:00
lws_callback_all_protocol(struct lws_context *context,
const struct lws_protocols *protocol, int reason)
libwebsocket_callback_all_protocol Signed-off-by: Andy Green <andy.green@linaro.org>
2014-02-15 16:36:38 +08:00
{
multithreaded service This adds support for multithreaded service to lws without adding any threading or locking code in the library. At context creation time you can request split the service part of the context into n service domains, which are load-balanced so that the most idle one gets the next listen socket accept. There's a single listen socket on one port still. User code may then spawn n threads doing n service loops / poll()s simultaneously. Locking is only required (I think) in the existing FD lock callbacks already handled by the pthreads server example, and that locking takes place in user code. So the library remains completely agnostic about the threading / locking scheme. And by default, it's completely compatible with one service thread so no changes are required by people uninterested in multithreaded service. However for people interested in extremely lightweight mass http[s]/ ws[s] service with minimum provisioning, the library can now do everything out of the box. To test it, just try $ libwebsockets-test-server-pthreads -j 8 where -j controls the number of service threads Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 03:34:24 +08:00
struct lws_context_per_thread *pt = &context->pt[0];
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
unsigned int n, m = context->count_threads;
clean pre 1.7 Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-29 21:18:54 +08:00
struct lws *wsi;
multithreaded service This adds support for multithreaded service to lws without adding any threading or locking code in the library. At context creation time you can request split the service part of the context into n service domains, which are load-balanced so that the most idle one gets the next listen socket accept. There's a single listen socket on one port still. User code may then spawn n threads doing n service loops / poll()s simultaneously. Locking is only required (I think) in the existing FD lock callbacks already handled by the pthreads server example, and that locking takes place in user code. So the library remains completely agnostic about the threading / locking scheme. And by default, it's completely compatible with one service thread so no changes are required by people uninterested in multithreaded service. However for people interested in extremely lightweight mass http[s]/ ws[s] service with minimum provisioning, the library can now do everything out of the box. To test it, just try $ libwebsockets-test-server-pthreads -j 8 where -j controls the number of service threads Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 03:34:24 +08:00
while (m--) {
for (n = 0; n < pt->fds_count; n++) {
wsi = wsi_from_fd(context, pt->fds[n].fd);
if (!wsi)
continue;
if (wsi->protocol == protocol)
clean pre 1.7 Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-29 21:18:54 +08:00
protocol->callback(wsi, reason, wsi->user_space,
NULL, 0);
multithreaded service This adds support for multithreaded service to lws without adding any threading or locking code in the library. At context creation time you can request split the service part of the context into n service domains, which are load-balanced so that the most idle one gets the next listen socket accept. There's a single listen socket on one port still. User code may then spawn n threads doing n service loops / poll()s simultaneously. Locking is only required (I think) in the existing FD lock callbacks already handled by the pthreads server example, and that locking takes place in user code. So the library remains completely agnostic about the threading / locking scheme. And by default, it's completely compatible with one service thread so no changes are required by people uninterested in multithreaded service. However for people interested in extremely lightweight mass http[s]/ ws[s] service with minimum provisioning, the library can now do everything out of the box. To test it, just try $ libwebsockets-test-server-pthreads -j 8 where -j controls the number of service threads Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 03:34:24 +08:00
}
pt++;
libwebsocket_callback_all_protocol Signed-off-by: Andy Green <andy.green@linaro.org>
2014-02-15 16:36:38 +08:00
}
return 0;
}
plugins This adds support for dynamically loaded plugins at runtime, which can expose their own protocols or extensions transparently. With these changes lwsws defaults to OFF in cmake, and if enabled it automatically enables plugins and libuv support. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-06 16:15:40 +08:00
LWS_VISIBLE int
lws_callback_all_protocol_vhost_args
2017-08-05 10:38:59 +08:00
lws_callback_all_protocol_vhost_args(struct lws_vhost *vh,
const struct lws_protocols *protocol, int reason,
void *argp, size_t len)
plugins This adds support for dynamically loaded plugins at runtime, which can expose their own protocols or extensions transparently. With these changes lwsws defaults to OFF in cmake, and if enabled it automatically enables plugins and libuv support. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-06 16:15:40 +08:00
{
struct lws_context *context = vh->context;
struct lws_context_per_thread *pt = &context->pt[0];
unsigned int n, m = context->count_threads;
struct lws *wsi;
while (m--) {
for (n = 0; n < pt->fds_count; n++) {
wsi = wsi_from_fd(context, pt->fds[n].fd);
if (!wsi)
continue;
lws_callback_all_protocol_vhost_args
2017-08-05 10:38:59 +08:00
if (wsi->vhost == vh && (wsi->protocol == protocol ||
!protocol))
wsi->protocol->callback(wsi, reason,
wsi->user_space, argp, len);
plugins This adds support for dynamically loaded plugins at runtime, which can expose their own protocols or extensions transparently. With these changes lwsws defaults to OFF in cmake, and if enabled it automatically enables plugins and libuv support. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-06 16:15:40 +08:00
}
pt++;
}
return 0;
}
lws_callback_all_protocol_vhost_args
2017-08-05 10:38:59 +08:00
LWS_VISIBLE int
lws_callback_all_protocol_vhost(struct lws_vhost *vh,
const struct lws_protocols *protocol, int reason)
{
return lws_callback_all_protocol_vhost_args(vh, protocol, reason, NULL, 0);
}
lws_callback_vhost_protocols This gives protocols a way to talk to each other via per-vhost callbacks, one per protocol (including the sender). Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-15 10:46:58 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_callback_vhost_protocols(struct lws *wsi, int reason, void *in, int len)
{
int n;
for (n = 0; n < wsi->vhost->count_protocols; n++)
if (wsi->vhost->protocols[n].callback(wsi, reason, NULL, in, len))
return 1;
return 0;
}
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_callback_vhost_protocols_vhost(struct lws_vhost *vh, int reason, void *in,
size_t len)
{
int n;
struct lws *wsi = lws_zalloc(sizeof(*wsi), "fake wsi");
wsi->context = vh->context;
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
lws_vhost_bind_wsi(vh, wsi);
lws_timed_callback_vh_protocol
2017-12-07 10:16:17 +08:00
for (n = 0; n < wsi->vhost->count_protocols; n++) {
wsi->protocol = &vh->protocols[n];
if (wsi->protocol->callback(wsi, reason, NULL, in, len)) {
lws_free(wsi);
return 1;
}
}
lws_free(wsi);
return 0;
}
ESP32 platform This is enough for all the test app features to work on ESP32 without SSL.
2017-02-18 17:26:40 +08:00
LWS_VISIBLE LWS_EXTERN void
esp32: align fops member name defines
2017-03-08 11:11:41 +08:00
lws_set_fops(struct lws_context *context, const struct lws_plat_file_ops *fops)
ESP32 platform This is enough for all the test app features to work on ESP32 without SSL.
2017-02-18 17:26:40 +08:00
{
esp32: align fops member name defines
2017-03-08 11:11:41 +08:00
context->fops = fops;
fops-zip
2017-03-03 12:38:10 +08:00
}
LWS_VISIBLE LWS_EXTERN lws_filepos_t
lws_vfs_tell(lws_fop_fd_t fop_fd)
{
return fop_fd->pos;
}
LWS_VISIBLE LWS_EXTERN lws_filepos_t
lws_vfs_get_length(lws_fop_fd_t fop_fd)
{
return fop_fd->len;
}
LWS_VISIBLE LWS_EXTERN uint32_t
lws_vfs_get_mod_time(lws_fop_fd_t fop_fd)
{
return fop_fd->mod_time;
}
LWS_VISIBLE lws_fileofs_t
lws_vfs_file_seek_set(lws_fop_fd_t fop_fd, lws_fileofs_t offset)
{
lws_fileofs_t ofs;
clean
2017-09-23 12:55:21 +08:00
fops-zip
2017-03-03 12:38:10 +08:00
ofs = fop_fd->fops->LWS_FOP_SEEK_CUR(fop_fd, offset - fop_fd->pos);
clean
2017-09-23 12:55:21 +08:00
fops-zip
2017-03-03 12:38:10 +08:00
return ofs;
}
LWS_VISIBLE lws_fileofs_t
lws_vfs_file_seek_end(lws_fop_fd_t fop_fd, lws_fileofs_t offset)
{
clean
2017-09-23 12:55:21 +08:00
return fop_fd->fops->LWS_FOP_SEEK_CUR(fop_fd, fop_fd->len +
fop_fd->pos + offset);
fops: allow setting from context creation and introduce lws_select_fops_by_vfs_path 1) There's now a .fops pointer that can be set in the context creation info. If set, the array of fops it points to (terminated by an entry with .open = NULL) is walked to find out the best vfs filesystem path match (comparing the vfs path to fops.path_prefix) for which fops to use. If none given (.fops is NULL in info) then behaviour is as before, fops are the platform-provided one only. 2) The built in fileserving now walks any array of fops looking for the best fops match automatically. 3) lws_plat_file_... apis are renamed to lws_vfs_file_...
2017-03-01 14:28:56 +08:00
}
fops-zip
2017-03-03 12:38:10 +08:00
const struct lws_plat_file_ops *
lws_vfs_select_fops(const struct lws_plat_file_ops *fops, const char *vfs_path,
const char **vpath)
fops: allow setting from context creation and introduce lws_select_fops_by_vfs_path 1) There's now a .fops pointer that can be set in the context creation info. If set, the array of fops it points to (terminated by an entry with .open = NULL) is walked to find out the best vfs filesystem path match (comparing the vfs path to fops.path_prefix) for which fops to use. If none given (.fops is NULL in info) then behaviour is as before, fops are the platform-provided one only. 2) The built in fileserving now walks any array of fops looking for the best fops match automatically. 3) lws_plat_file_... apis are renamed to lws_vfs_file_...
2017-03-01 14:28:56 +08:00
{
fops-zip
2017-03-03 12:38:10 +08:00
const struct lws_plat_file_ops *pf;
const char *p = vfs_path;
int n;
fops: allow setting from context creation and introduce lws_select_fops_by_vfs_path 1) There's now a .fops pointer that can be set in the context creation info. If set, the array of fops it points to (terminated by an entry with .open = NULL) is walked to find out the best vfs filesystem path match (comparing the vfs path to fops.path_prefix) for which fops to use. If none given (.fops is NULL in info) then behaviour is as before, fops are the platform-provided one only. 2) The built in fileserving now walks any array of fops looking for the best fops match automatically. 3) lws_plat_file_... apis are renamed to lws_vfs_file_...
2017-03-01 14:28:56 +08:00
fops-zip
2017-03-03 12:38:10 +08:00
*vpath = NULL;
/* no non-platform fops, just use that */
if (!fops->next)
return fops;
/*
* scan the vfs path looking for indications we are to be
* handled by a specific fops
*/
fops: allow setting from context creation and introduce lws_select_fops_by_vfs_path 1) There's now a .fops pointer that can be set in the context creation info. If set, the array of fops it points to (terminated by an entry with .open = NULL) is walked to find out the best vfs filesystem path match (comparing the vfs path to fops.path_prefix) for which fops to use. If none given (.fops is NULL in info) then behaviour is as before, fops are the platform-provided one only. 2) The built in fileserving now walks any array of fops looking for the best fops match automatically. 3) lws_plat_file_... apis are renamed to lws_vfs_file_...
2017-03-01 14:28:56 +08:00
esp32: align fops member name defines
2017-03-08 11:11:41 +08:00
while (p && *p) {
fops-zip
2017-03-03 12:38:10 +08:00
if (*p != '/') {
p++;
continue;
}
/* the first one is always platform fops, so skip */
pf = fops->next;
while (pf) {
n = 0;
libwebsockets.h: clean out some boilerplate better put in core/private.h https://github.com/warmcat/libwebsockets/issues/1370
2018-08-16 19:10:32 +08:00
while (n < (int)LWS_ARRAY_SIZE(pf->fi) && pf->fi[n].sig) {
fops-zip
2017-03-03 12:38:10 +08:00
if (p >= vfs_path + pf->fi[n].len)
if (!strncmp(p - (pf->fi[n].len - 1),
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
pf->fi[n].sig,
pf->fi[n].len - 1)) {
fops-zip
2017-03-03 12:38:10 +08:00
*vpath = p + 1;
return pf;
}
n++;
fops: allow setting from context creation and introduce lws_select_fops_by_vfs_path 1) There's now a .fops pointer that can be set in the context creation info. If set, the array of fops it points to (terminated by an entry with .open = NULL) is walked to find out the best vfs filesystem path match (comparing the vfs path to fops.path_prefix) for which fops to use. If none given (.fops is NULL in info) then behaviour is as before, fops are the platform-provided one only. 2) The built in fileserving now walks any array of fops looking for the best fops match automatically. 3) lws_plat_file_... apis are renamed to lws_vfs_file_...
2017-03-01 14:28:56 +08:00
}
fops-zip
2017-03-03 12:38:10 +08:00
pf = pf->next;
}
p++;
fops: allow setting from context creation and introduce lws_select_fops_by_vfs_path 1) There's now a .fops pointer that can be set in the context creation info. If set, the array of fops it points to (terminated by an entry with .open = NULL) is walked to find out the best vfs filesystem path match (comparing the vfs path to fops.path_prefix) for which fops to use. If none given (.fops is NULL in info) then behaviour is as before, fops are the platform-provided one only. 2) The built in fileserving now walks any array of fops looking for the best fops match automatically. 3) lws_plat_file_... apis are renamed to lws_vfs_file_...
2017-03-01 14:28:56 +08:00
}
fops-zip
2017-03-03 12:38:10 +08:00
return fops;
}
LWS_VISIBLE LWS_EXTERN lws_fop_fd_t LWS_WARN_UNUSED_RESULT
lws_vfs_file_open(const struct lws_plat_file_ops *fops, const char *vfs_path,
lws_fop_flags_t *flags)
{
esp32: align fops member name defines
2017-03-08 11:11:41 +08:00
const char *vpath = "";
clean
2017-09-23 12:55:21 +08:00
const struct lws_plat_file_ops *selected;
selected = lws_vfs_select_fops(fops, vfs_path, &vpath);
fops: allow setting from context creation and introduce lws_select_fops_by_vfs_path 1) There's now a .fops pointer that can be set in the context creation info. If set, the array of fops it points to (terminated by an entry with .open = NULL) is walked to find out the best vfs filesystem path match (comparing the vfs path to fops.path_prefix) for which fops to use. If none given (.fops is NULL in info) then behaviour is as before, fops are the platform-provided one only. 2) The built in fileserving now walks any array of fops looking for the best fops match automatically. 3) lws_plat_file_... apis are renamed to lws_vfs_file_...
2017-03-01 14:28:56 +08:00
fops-zip
2017-03-03 12:38:10 +08:00
return selected->LWS_FOP_OPEN(fops, vfs_path, vpath, flags);
ESP32 platform This is enough for all the test app features to work on ESP32 without SSL.
2017-02-18 17:26:40 +08:00
}
fops-zip
2017-03-03 12:38:10 +08:00
lws_now_secs Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-17 09:41:22 +08:00
/**
* lws_now_secs() - seconds since 1970-1-1
*
*/
LWS_VISIBLE LWS_EXTERN unsigned long
lws_now_secs(void)
{
struct timeval tv;
gettimeofday(&tv, NULL);
return tv.tv_sec;
}
lws_compare_time_t: conceal time discontiguities on all platforms This provides platform-independent support for time discontiguities. On embedded without battery RTC, commonly we only get time after NTP completes. This makes the cert checking happen when we have a reasonable time and introduces lws_compare_time_t() to correctly compare time_t s that may sit on either side of a time discontiguity.
2017-11-24 11:00:59 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_compare_time_t(struct lws_context *context, time_t t1, time_t t2)
{
if (t1 < context->time_discontiguity)
t1 += context->time_fixup;
if (t2 < context->time_discontiguity)
t2 += context->time_fixup;
return (int)(t1 - t2);
}
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
LWS_VISIBLE lws_sockfd_type
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
lws_get_socket_fd(struct lws *wsi)
introduce-test-ping-app.patch Signed-off-by: Andy Green <andy@warmcat.com>
2011-01-27 20:06:03 +00:00
{
cgi: also allow for generic spawn
2017-09-19 07:58:04 +08:00
if (!wsi)
return -1;
adoption: make union for socket and file fds This lets lws support adopting raw file FDs and raw socket fds. A test plugin creates a FIFO and prints data sent on it, using the lws event loop.
2017-02-27 12:55:56 +08:00
return wsi->desc.sockfd;
introduce-test-ping-app.patch Signed-off-by: Andy Green <andy@warmcat.com>
2011-01-27 20:06:03 +00:00
}
introduce lws_latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-29 12:36:17 +08:00
#ifdef LWS_LATENCY
void
clean tidy the worst whitespace alignment probs due to mass token name length changes Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:30:53 +08:00
lws_latency(struct lws_context *context, struct lws *wsi, const char *action,
int ret, int completed)
introduce lws_latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-29 12:36:17 +08:00
{
Avoid calls to gettimeofday() Add a new function to get the current time in microseconds, since gettimeofday() does not exist on Windows. Keep the current implementation for the test applications.
2014-02-26 21:37:31 +01:00
unsigned long long u;
introduce lws_latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-29 12:36:17 +08:00
char buf[256];
lws_time_in_microseconds: export Rename time_in_microseconds() and export the internal api as lws_time_in_microseconds()
2018-10-14 06:15:36 +08:00
u = lws_time_in_microseconds();
introduce lws_latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-29 12:36:17 +08:00
clean libwebsockets.c Signed-off-by: None <None>
2014-03-31 11:01:32 +08:00
if (!action) {
introduce lws_latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-29 12:36:17 +08:00
wsi->latency_start = u;
if (!wsi->action_start)
wsi->action_start = u;
clean libwebsockets.c Signed-off-by: None <None>
2014-03-31 11:01:32 +08:00
return;
introduce lws_latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-29 12:36:17 +08:00
}
clean libwebsockets.c Signed-off-by: None <None>
2014-03-31 11:01:32 +08:00
if (completed) {
if (wsi->action_start == wsi->latency_start)
sprintf(buf,
fix confusion ull in latency reporting
2014-07-05 10:50:47 +08:00
"Completion first try lat %lluus: %p: ret %d: %s\n",
clean libwebsockets.c Signed-off-by: None <None>
2014-03-31 11:01:32 +08:00
u - wsi->latency_start,
(void *)wsi, ret, action);
else
sprintf(buf,
fix confusion ull in latency reporting
2014-07-05 10:50:47 +08:00
"Completion %lluus: lat %lluus: %p: ret %d: %s\n",
clean libwebsockets.c Signed-off-by: None <None>
2014-03-31 11:01:32 +08:00
u - wsi->action_start,
u - wsi->latency_start,
(void *)wsi, ret, action);
wsi->action_start = 0;
} else
fix confusion ull in latency reporting
2014-07-05 10:50:47 +08:00
sprintf(buf, "lat %lluus: %p: ret %d: %s\n",
clean libwebsockets.c Signed-off-by: None <None>
2014-03-31 11:01:32 +08:00
u - wsi->latency_start, (void *)wsi, ret, action);
if (u - wsi->latency_start > context->worst_latency) {
context->worst_latency = u - wsi->latency_start;
strcpy(context->worst_latency_info, buf);
}
lwsl_latency("%s", buf);
introduce lws_latency Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-29 12:36:17 +08:00
}
#endif
Hide private library symbols Signed-off-by: Peter Pentchev <roam@ringlet.net>
2013-03-30 09:52:21 +08:00
LWS_VISIBLE 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_rx_flow_control(struct lws *wsi, int _enable)
solve flowcontrol problems Problems with rx flow control implementation were the underlying cause of the connection stalling issue that was covered up with the udelay() patch that was removed recently. This get rx flow control working properly and corrects problems with fifo management in the test server mirror protocol code too. The rxfow control api has been changed to just set a flag, so it's very cheap to call from user code. After the callbacks that might use the rxflow control api the flag is checked and any pending actions done. rx flow control now stops any rx packet coming immediately, with compessed connections "just what was left in the pipe" might be hundreds of KBytes. To implement that the current packet being decoded is copied into a malloc'd buffer by the rx processing code now. When rxflow is allows to come again, the buffer is drained and freed before any new packet content is accepted. Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-17 16:50:35 +08:00
{
smp: fixes
2018-03-02 14:22:49 +08:00
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
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 en = _enable;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
// h2 ignores rx flow control atm
if (lwsi_role_h2(wsi) || wsi->http2_substream ||
lwsi_role_h2_ENCAPSULATION(wsi))
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
lwsl_info("%s: %p 0x%x\n", __func__, wsi, _enable);
if (!(_enable & LWS_RXFLOW_REASON_APPLIES)) {
/*
* convert user bool style to bitmap style... in user simple
* bool style _enable = 0 = flow control it, = 1 = allow rx
*/
en = LWS_RXFLOW_REASON_APPLIES | LWS_RXFLOW_REASON_USER_BOOL;
if (_enable & 1)
en |= LWS_RXFLOW_REASON_APPLIES_ENABLE_BIT;
}
smp: fixes
2018-03-02 14:22:49 +08:00
lws_pt_lock(pt, __func__);
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
/* any bit set in rxflow_bitmap DISABLEs rxflow control */
if (en & LWS_RXFLOW_REASON_APPLIES_ENABLE_BIT)
wsi->rxflow_bitmap &= ~(en & 0xff);
else
wsi->rxflow_bitmap |= en & 0xff;
if ((LWS_RXFLOW_PENDING_CHANGE | (!wsi->rxflow_bitmap)) ==
wsi->rxflow_change_to)
smp: fixes
2018-03-02 14:22:49 +08:00
goto skip;
rxflow remove recursion and simplify Signed-off-by: Andy Green <andy.green@linaro.org>
2013-03-16 11:24:23 +08:00
codacy: minor fixes
2018-10-10 13:54:43 +08:00
wsi->rxflow_change_to = LWS_RXFLOW_PENDING_CHANGE |
(!wsi->rxflow_bitmap);
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 mode and states into roles and states
2018-04-02 11:55:17 +08:00
lwsl_info("%s: %p: bitmap 0x%x: en 0x%x, ch 0x%x\n", __func__, 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
wsi->rxflow_bitmap, en, wsi->rxflow_change_to);
if (_enable & LWS_RXFLOW_REASON_FLAG_PROCESS_NOW ||
smp: fixes
2018-03-02 14:22:49 +08:00
!wsi->rxflow_will_be_applied) {
en = __lws_rx_flow_control(wsi);
lws_pt_unlock(pt);
return en;
}
skip:
lws_pt_unlock(pt);
solve flowcontrol problems Problems with rx flow control implementation were the underlying cause of the connection stalling issue that was covered up with the udelay() patch that was removed recently. This get rx flow control working properly and corrects problems with fifo management in the test server mirror protocol code too. The rxfow control api has been changed to just set a flag, so it's very cheap to call from user code. After the callbacks that might use the rxflow control api the flag is checked and any pending actions done. rx flow control now stops any rx packet coming immediately, with compessed connections "just what was left in the pipe" might be hundreds of KBytes. To implement that the current packet being decoded is copied into a malloc'd buffer by the rx processing code now. When rxflow is allows to come again, the buffer is drained and freed before any new packet content is accepted. Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-17 16:50:35 +08:00
return 0;
}
Hide private library symbols Signed-off-by: Peter Pentchev <roam@ringlet.net>
2013-03-30 09:52:21 +08:00
LWS_VISIBLE void
make protocols const require explicit context API BREAK The user protocols struct has not been const until now. This has been painful for a while because the semantics of the protocols struct look like it's going to be treated as const. At context creation, the protocols struct has been getting marked with the context, and three apis exploited that to only need to be passed a pointer to a protocol to get access to the context. This patch removes the two writeable members in the context (these were never directly used by user code), changes all pointers to protocols to be const, and adds an explicit first argument to the three affected apis so they can have access to context. The three affected apis are these LWS_VISIBLE LWS_EXTERN int -lws_callback_on_writable_all_protocol(const struct lws_protocols *protocol); +lws_callback_on_writable_all_protocol(const struct lws_context *context, + const struct lws_protocols *protocol); LWS_VISIBLE LWS_EXTERN int -lws_callback_all_protocol(const struct lws_protocols *protocol, int reason); +lws_callback_all_protocol(struct lws_context *context, + const struct lws_protocols *protocol, int reason); LWS_VISIBLE LWS_EXTERN void -lws_rx_flow_allow_all_protocol(const struct lws_protocols *protocol); +lws_rx_flow_allow_all_protocol(const struct lws_context *context, + const struct lws_protocols *protocol); unfortunately the original apis can no longer be emulated and users of them must update. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-11 10:45:35 +08:00
lws_rx_flow_allow_all_protocol(const struct lws_context *context,
const struct lws_protocols *protocol)
introduce api for unthrottle all connections of protocol Signed-off-by: Andy Green <andy.green@linaro.org>
2013-03-16 12:32:27 +08:00
{
multithreaded service This adds support for multithreaded service to lws without adding any threading or locking code in the library. At context creation time you can request split the service part of the context into n service domains, which are load-balanced so that the most idle one gets the next listen socket accept. There's a single listen socket on one port still. User code may then spawn n threads doing n service loops / poll()s simultaneously. Locking is only required (I think) in the existing FD lock callbacks already handled by the pthreads server example, and that locking takes place in user code. So the library remains completely agnostic about the threading / locking scheme. And by default, it's completely compatible with one service thread so no changes are required by people uninterested in multithreaded service. However for people interested in extremely lightweight mass http[s]/ ws[s] service with minimum provisioning, the library can now do everything out of the box. To test it, just try $ libwebsockets-test-server-pthreads -j 8 where -j controls the number of service threads Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 03:34:24 +08:00
const struct lws_context_per_thread *pt = &context->pt[0];
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
struct lws *wsi;
multithread stability Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-26 20:56:56 +08:00
unsigned int n, m = context->count_threads;
multithreaded service This adds support for multithreaded service to lws without adding any threading or locking code in the library. At context creation time you can request split the service part of the context into n service domains, which are load-balanced so that the most idle one gets the next listen socket accept. There's a single listen socket on one port still. User code may then spawn n threads doing n service loops / poll()s simultaneously. Locking is only required (I think) in the existing FD lock callbacks already handled by the pthreads server example, and that locking takes place in user code. So the library remains completely agnostic about the threading / locking scheme. And by default, it's completely compatible with one service thread so no changes are required by people uninterested in multithreaded service. However for people interested in extremely lightweight mass http[s]/ ws[s] service with minimum provisioning, the library can now do everything out of the box. To test it, just try $ libwebsockets-test-server-pthreads -j 8 where -j controls the number of service threads Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 03:34:24 +08:00
while (m--) {
for (n = 0; n < pt->fds_count; n++) {
wsi = wsi_from_fd(context, pt->fds[n].fd);
if (!wsi)
continue;
if (wsi->protocol == protocol)
lws_rx_flow_control(wsi, LWS_RXFLOW_ALLOW);
}
pt++;
introduce api for unthrottle all connections of protocol Signed-off-by: Andy Green <andy.green@linaro.org>
2013-03-16 12:32:27 +08:00
}
}
vhost: check cert validity dates After startup, and once per day, check the validity dates on our ssl certs, and broadcast callbacks with the information so interested plugins can know. If our clock is < May 2016, we don't try to judge the certs, because clearly we don't know what time it is.
2017-10-26 18:53:34 +08:00
int
lws_broadcast(struct lws_context *context, int reason, void *in, size_t len)
{
struct lws_vhost *v = context->vhost_list;
struct lws wsi;
int n, ret = 0;
memset(&wsi, 0, sizeof(wsi));
wsi.context = context;
while (v) {
const struct lws_protocols *p = v->protocols;
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
wsi.vhost = v; /* not a real bound wsi */
vhost: check cert validity dates After startup, and once per day, check the validity dates on our ssl certs, and broadcast callbacks with the information so interested plugins can know. If our clock is < May 2016, we don't try to judge the certs, because clearly we don't know what time it is.
2017-10-26 18:53:34 +08:00
for (n = 0; n < v->count_protocols; n++) {
wsi.protocol = p;
if (p->callback &&
p->callback(&wsi, reason, NULL, in, len))
ret |= 1;
p++;
}
v = v->vhost_next;
}
return ret;
}
Hide private library symbols Signed-off-by: Peter Pentchev <roam@ringlet.net>
2013-03-30 09:52:21 +08:00
LWS_VISIBLE extern const char *
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
lws_canonical_hostname(struct lws_context *context)
fix-host-header-contents--introduce-canonical-hostname-api.patch I?aki pointed out the dummy host field used in client test and ping is not valid http. This patch changes it to use the actual host name and adds an api to collect that from the context cheaply. Reported-by: I?aki Baz Castillo <ibc@aliax.net> Signed-off-by: Andy Green <andy@warmcat.com>
2011-01-28 10:00:18 +00:00
{
introduce win32 build capability This adds win32 build compatability to libwebsockets. The patch is from Peter Hinz, Andy Green has cleaned it up a bit and possibly broken win32 compatability since I can't test it, so there may be followup patches. It compiles fine under Linux after this patch anyway. Much of the patch is changing a reserved keyword for Visual C compiler "this" to "context", but there is no real C99 support in the MSFT compiler even though it is 2011 so C99 style array declarations have been mangled back into "ancient C" style. Some windows-isms are also added like closesocket() but these are quite localized. Win32 random is just using C library random() call at the moment vs Linux /dev/urandom. canonical hostname detection is broken in win32 at the moment. Signed-off-by: Peter Hinz <cerebusrc@gmail.com> Signed-off-by: Andy Green <andy@warmcat.com>
2011-03-02 22:03:47 +00:00
return (const char *)context->canonical_hostname;
fix-host-header-contents--introduce-canonical-hostname-api.patch I?aki pointed out the dummy host field used in client test and ping is not valid http. This patch changes it to use the actual host name and adds an api to collect that from the context cheaply. Reported-by: I?aki Baz Castillo <ibc@aliax.net> Signed-off-by: Andy Green <andy@warmcat.com>
2011-01-28 10:00:18 +00:00
}
vhost: check cert validity dates After startup, and once per day, check the validity dates on our ssl certs, and broadcast callbacks with the information so interested plugins can know. If our clock is < May 2016, we don't try to judge the certs, because clearly we don't know what time it is.
2017-10-26 18:53:34 +08:00
LWS_VISIBLE LWS_EXTERN const char *
lws_get_vhost_name(struct lws_vhost *vhost)
{
return vhost->name;
}
additional vhost member accessors
2017-10-29 16:27:13 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_get_vhost_port(struct lws_vhost *vhost)
{
return vhost->listen_port;
}
LWS_VISIBLE LWS_EXTERN void *
lws_get_vhost_user(struct lws_vhost *vhost)
{
return vhost->user;
}
LWS_VISIBLE LWS_EXTERN const char *
lws_get_vhost_iface(struct lws_vhost *vhost)
{
return vhost->iface;
}
fix callback typedefs and declarations Remove declarations of callback and extension_callback as these are functions declared in header but not defined anywhere. Also rename typedefs callback_function and extension_callback_function to lws_callback_function and lws_extension_callback_function so all symbolx exported by header have lws prefix; Signed-off-by: Denis Osvald <denis.osvald@sartura.hr>
2015-12-21 18:06:38 +01:00
int user_callback_handle_rxflow(lws_callback_function callback_function,
public api remove context from user callback API BREAK Since struct lws (wsi) now has his own context pointer, we were able to remove the need for passing context almost everywhere in the apis. In turn, that means there's no real use for context being passed to every callback; in the rare cases context is needed user code can get it with lws_get_ctx(wsi) Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-17 07:54:44 +08:00
struct lws *wsi,
clean tidy the worst whitespace alignment probs due to mass token name length changes Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:30:53 +08:00
enum lws_callback_reasons reason, void *user,
void *in, size_t len)
solve flowcontrol problems Problems with rx flow control implementation were the underlying cause of the connection stalling issue that was covered up with the udelay() patch that was removed recently. This get rx flow control working properly and corrects problems with fifo management in the test server mirror protocol code too. The rxfow control api has been changed to just set a flag, so it's very cheap to call from user code. After the callbacks that might use the rxflow control api the flag is checked and any pending actions done. rx flow control now stops any rx packet coming immediately, with compessed connections "just what was left in the pipe" might be hundreds of KBytes. To implement that the current packet being decoded is copied into a malloc'd buffer by the rx processing code now. When rxflow is allows to come again, the buffer is drained and freed before any new packet content is accepted. Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-17 16:50:35 +08:00
{
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
wsi->rxflow_will_be_applied = 1;
public api remove context from user callback API BREAK Since struct lws (wsi) now has his own context pointer, we were able to remove the need for passing context almost everywhere in the apis. In turn, that means there's no real use for context being passed to every callback; in the rare cases context is needed user code can get it with lws_get_ctx(wsi) Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-17 07:54:44 +08:00
n = callback_function(wsi, reason, user, in, 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
wsi->rxflow_will_be_applied = 0;
dont close in user callback wrapper let ancestor do it Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-10 21:21:24 +08:00
if (!n)
smp: fixes
2018-03-02 14:22:49 +08:00
n = __lws_rx_flow_control(wsi);
solve flowcontrol problems Problems with rx flow control implementation were the underlying cause of the connection stalling issue that was covered up with the udelay() patch that was removed recently. This get rx flow control working properly and corrects problems with fifo management in the test server mirror protocol code too. The rxfow control api has been changed to just set a flag, so it's very cheap to call from user code. After the callbacks that might use the rxflow control api the flag is checked and any pending actions done. rx flow control now stops any rx packet coming immediately, with compessed connections "just what was left in the pipe" might be hundreds of KBytes. To implement that the current packet being decoded is copied into a malloc'd buffer by the rx processing code now. When rxflow is allows to come again, the buffer is drained and freed before any new packet content is accepted. Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-17 16:50:35 +08:00
dont close in user callback wrapper let ancestor do it Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-10 21:21:24 +08:00
return n;
solve flowcontrol problems Problems with rx flow control implementation were the underlying cause of the connection stalling issue that was covered up with the udelay() patch that was removed recently. This get rx flow control working properly and corrects problems with fifo management in the test server mirror protocol code too. The rxfow control api has been changed to just set a flag, so it's very cheap to call from user code. After the callbacks that might use the rxflow control api the flag is checked and any pending actions done. rx flow control now stops any rx packet coming immediately, with compessed connections "just what was left in the pipe" might be hundreds of KBytes. To implement that the current packet being decoded is copied into a malloc'd buffer by the rx processing code now. When rxflow is allows to come again, the buffer is drained and freed before any new packet content is accepted. Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-17 16:50:35 +08:00
}
refactor: also migrate tls to the ops struct and private.h pattern Several new ops are planned for tls... so better to bite the bullet and clean it out to the same level as roles + event-libs first. Also adds a new travis target "mbedtls" and all the tests except autobahn against mbedtls build.
2018-05-01 12:41:42 +08:00
#if !defined(LWS_WITHOUT_CLIENT)
manual proxy forcing api add function to manually setup proxy. Useful on iOS where getenv doesn't return proxy settings Simplified by AG Signed-off-by: shys <shyswork@zoho.com> Signed-off-by: Andy Green <andy@warmcat.com>
2013-10-24 22:12:03 +08:00
LWS_VISIBLE int
introduce vhosts This patch splits out some lws_context members into a new lws_vhost struct. - ssl state and options per vhost - SSL_CTX for serving and client per vhost - protocols[] per vhost - extensions[] per vhost lws_context maintains a linked list of lws_vhosts. The same lws_context_creation_info struct is used to regulate both the context creation and to create vhosts: for backward compatibility if you didn't provide the new LWS_SERVER_OPTION_EXPLICIT_VHOSTS option, then a default vhost is created at context creation time using the same info data as the context itself. If you will have multiple vhosts though, you should give the LWS_SERVER_OPTION_EXPLICIT_VHOSTS option at context creation time, create the context first and then the vhosts afterwards using lws_create_vhost(contest, &info); Although there is a lot of housekeeping to implement this change, there is almost no additional overhead if you don't use multiple vhosts and very little api impact (no changes to test apps). Signed-off-by: Andy Green <andy@warmcat.com>
2016-03-28 10:10:43 +08:00
lws_set_proxy(struct lws_vhost *vhost, const char *proxy)
manual proxy forcing api add function to manually setup proxy. Useful on iOS where getenv doesn't return proxy settings Simplified by AG Signed-off-by: shys <shyswork@zoho.com> Signed-off-by: Andy Green <andy@warmcat.com>
2013-10-24 22:12:03 +08:00
{
char *p;
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
char authstring[96];
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
manual proxy forcing api add function to manually setup proxy. Useful on iOS where getenv doesn't return proxy settings Simplified by AG Signed-off-by: shys <shyswork@zoho.com> Signed-off-by: Andy Green <andy@warmcat.com>
2013-10-24 22:12:03 +08:00
if (!proxy)
return -1;
http_proxy: deal with redundant protocol leader https://github.com/warmcat/libwebsockets/issues/764
2017-01-26 07:27:11 +08:00
/* we have to deal with a possible redundant leading http:// */
if (!strncmp(proxy, "http://", 7))
proxy += 7;
proxy: allow use of at char in passwords Applies to both http proxy and socks proxy https://github.com/warmcat/libwebsockets/issues/1247
2018-04-20 10:49:29 +08:00
p = strrchr(proxy, '@');
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
if (p) { /* auth is around */
mbed3 workable plus or minus mbed3 net stack bug https://github.com/ARMmbed/sockets/issues/35 Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 12:10:26 +08:00
if ((unsigned int)(p - proxy) > sizeof(authstring) - 1)
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
goto auth_too_long;
lws_snprintf
2018-03-12 09:28:26 +08:00
lws_strncpy(authstring, proxy, p - proxy + 1);
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
// null termination not needed on input
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
if (lws_b64_encode_string(authstring, lws_ptr_diff(p, proxy),
introduce vhosts This patch splits out some lws_context members into a new lws_vhost struct. - ssl state and options per vhost - SSL_CTX for serving and client per vhost - protocols[] per vhost - extensions[] per vhost lws_context maintains a linked list of lws_vhosts. The same lws_context_creation_info struct is used to regulate both the context creation and to create vhosts: for backward compatibility if you didn't provide the new LWS_SERVER_OPTION_EXPLICIT_VHOSTS option, then a default vhost is created at context creation time using the same info data as the context itself. If you will have multiple vhosts though, you should give the LWS_SERVER_OPTION_EXPLICIT_VHOSTS option at context creation time, create the context first and then the vhosts afterwards using lws_create_vhost(contest, &info); Although there is a lot of housekeeping to implement this change, there is almost no additional overhead if you don't use multiple vhosts and very little api impact (no changes to test apps). Signed-off-by: Andy Green <andy@warmcat.com>
2016-03-28 10:10:43 +08:00
vhost->proxy_basic_auth_token,
sizeof vhost->proxy_basic_auth_token) < 0)
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
goto auth_too_long;
esp8266 initial support
2016-07-23 14:18:25 +08:00
lwsl_info(" Proxy auth in use\n");
whitespace trailing mass cleanout Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 08:52:03 +08:00
codacy: minor fixes
2018-10-10 13:54:43 +08:00
#if defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
proxy = p + 1;
codacy: minor fixes
2018-10-10 13:54:43 +08:00
#endif
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
} else
introduce vhosts This patch splits out some lws_context members into a new lws_vhost struct. - ssl state and options per vhost - SSL_CTX for serving and client per vhost - protocols[] per vhost - extensions[] per vhost lws_context maintains a linked list of lws_vhosts. The same lws_context_creation_info struct is used to regulate both the context creation and to create vhosts: for backward compatibility if you didn't provide the new LWS_SERVER_OPTION_EXPLICIT_VHOSTS option, then a default vhost is created at context creation time using the same info data as the context itself. If you will have multiple vhosts though, you should give the LWS_SERVER_OPTION_EXPLICIT_VHOSTS option at context creation time, create the context first and then the vhosts afterwards using lws_create_vhost(contest, &info); Although there is a lot of housekeeping to implement this change, there is almost no additional overhead if you don't use multiple vhosts and very little api impact (no changes to test apps). Signed-off-by: Andy Green <andy@warmcat.com>
2016-03-28 10:10:43 +08:00
vhost->proxy_basic_auth_token[0] = '\0';
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
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 defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
lws_strncpy(vhost->http.http_proxy_address, proxy,
sizeof(vhost->http.http_proxy_address));
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
p = strchr(vhost->http.http_proxy_address, ':');
if (!p && !vhost->http.http_proxy_port) {
manual proxy forcing api add function to manually setup proxy. Useful on iOS where getenv doesn't return proxy settings Simplified by AG Signed-off-by: shys <shyswork@zoho.com> Signed-off-by: Andy Green <andy@warmcat.com>
2013-10-24 22:12:03 +08:00
lwsl_err("http_proxy needs to be ads:port\n");
return -1;
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
} else {
coverity 155649 medium possible write to null pointer Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-06 11:04:05 +08:00
if (p) {
*p = '\0';
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
vhost->http.http_proxy_port = atoi(p + 1);
coverity 155649 medium possible write to null pointer Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-06 11:04:05 +08:00
}
manual proxy forcing api add function to manually setup proxy. Useful on iOS where getenv doesn't return proxy settings Simplified by AG Signed-off-by: shys <shyswork@zoho.com> Signed-off-by: Andy Green <andy@warmcat.com>
2013-10-24 22:12:03 +08:00
}
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
lwsl_info(" Proxy %s:%u\n", vhost->http.http_proxy_address,
vhost->http.http_proxy_port);
#endif
manual proxy forcing api add function to manually setup proxy. Useful on iOS where getenv doesn't return proxy settings Simplified by AG Signed-off-by: shys <shyswork@zoho.com> Signed-off-by: Andy Green <andy@warmcat.com>
2013-10-24 22:12:03 +08:00
return 0;
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
auth_too_long:
lwsl_err("proxy auth too long\n");
plat: remove esp8266 ESP32 module price is now within range of 8266 price. ESP32 capability and OS support is hugely better than 8266, we can support mbedtls tls, http/2 etc with ESP32. I'm no longer testing on ESP8266... there's no more user traffic... it's time to go.
2017-11-30 12:40:46 +08:00
proxy auth Simplifies proxy code to use the existing libwebsocket_set_proxy. Enables libwebsocket_set_proxy() to parse username:password@ at front of servername in both http_proxy and info->http_proxy_address. If given the base64 version of the credentials are sent in the CONNECT header to the proxy. Port is now taken from info->http_proxy_address server:port syntax, but if a port is given in the now deprecated info->http_proxy_port (ie, is nonzero) then it is allowed to be missed out and the info port used instead for backwards compatibility. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-08 10:15:01 +08:00
return -1;
manual proxy forcing api add function to manually setup proxy. Useful on iOS where getenv doesn't return proxy settings Simplified by AG Signed-off-by: shys <shyswork@zoho.com> Signed-off-by: Andy Green <andy@warmcat.com>
2013-10-24 22:12:03 +08:00
}
refactor: also migrate tls to the ops struct and private.h pattern Several new ops are planned for tls... so better to bite the bullet and clean it out to the same level as roles + event-libs first. Also adds a new travis target "mbedtls" and all the tests except autobahn against mbedtls build.
2018-05-01 12:41:42 +08:00
#endif
manual proxy forcing api add function to manually setup proxy. Useful on iOS where getenv doesn't return proxy settings Simplified by AG Signed-off-by: shys <shyswork@zoho.com> Signed-off-by: Andy Green <andy@warmcat.com>
2013-10-24 22:12:03 +08:00
client: added socks5 proxy support AG: - move creation info members to end of struct - add LWS_WITH_SOCKS5 CMake var, defaults to OFF - cast away some warnings about signed / unsigned in strncpy Signed-off-by: Andy Ning <andy.ning@windriver.com>
2017-05-05 11:38:34 -04:00
#if defined(LWS_WITH_SOCKS5)
LWS_VISIBLE int
lws_set_socks(struct lws_vhost *vhost, const char *socks)
{
char *p_at, *p_colon;
char user[96];
char password[96];
if (!socks)
return -1;
vhost->socks_user[0] = '\0';
vhost->socks_password[0] = '\0';
proxy: allow use of at char in passwords Applies to both http proxy and socks proxy https://github.com/warmcat/libwebsockets/issues/1247
2018-04-20 10:49:29 +08:00
p_at = strrchr(socks, '@');
client: added socks5 proxy support AG: - move creation info members to end of struct - add LWS_WITH_SOCKS5 CMake var, defaults to OFF - cast away some warnings about signed / unsigned in strncpy Signed-off-by: Andy Ning <andy.ning@windriver.com>
2017-05-05 11:38:34 -04:00
if (p_at) { /* auth is around */
if ((unsigned int)(p_at - socks) > (sizeof(user)
+ sizeof(password) - 2)) {
lwsl_err("Socks auth too long\n");
goto bail;
}
p_colon = strchr(socks, ':');
if (p_colon) {
if ((unsigned int)(p_colon - socks) > (sizeof(user)
- 1) ) {
lwsl_err("Socks user too long\n");
goto bail;
}
if ((unsigned int)(p_at - p_colon) > (sizeof(password)
- 1) ) {
lwsl_err("Socks password too long\n");
goto bail;
}
coverity fixes
2017-09-23 13:27:11 +08:00
lws_snprintf
2018-03-12 09:28:26 +08:00
lws_strncpy(vhost->socks_user, socks, p_colon - socks + 1);
lws_strncpy(vhost->socks_password, p_colon + 1,
p_at - (p_colon + 1) + 1);
client: added socks5 proxy support AG: - move creation info members to end of struct - add LWS_WITH_SOCKS5 CMake var, defaults to OFF - cast away some warnings about signed / unsigned in strncpy Signed-off-by: Andy Ning <andy.ning@windriver.com>
2017-05-05 11:38:34 -04:00
}
lwsl_info(" Socks auth, user: %s, password: %s\n",
vhost->socks_user, vhost->socks_password );
socks = p_at + 1;
}
lws_snprintf
2018-03-12 09:28:26 +08:00
lws_strncpy(vhost->socks_proxy_address, socks,
sizeof(vhost->socks_proxy_address));
client: added socks5 proxy support AG: - move creation info members to end of struct - add LWS_WITH_SOCKS5 CMake var, defaults to OFF - cast away some warnings about signed / unsigned in strncpy Signed-off-by: Andy Ning <andy.ning@windriver.com>
2017-05-05 11:38:34 -04:00
p_colon = strchr(vhost->socks_proxy_address, ':');
if (!p_colon && !vhost->socks_proxy_port) {
lwsl_err("socks_proxy needs to be address:port\n");
return -1;
} else {
if (p_colon) {
*p_colon = '\0';
vhost->socks_proxy_port = atoi(p_colon + 1);
}
}
lwsl_info(" Socks %s:%u\n", vhost->socks_proxy_address,
vhost->socks_proxy_port);
return 0;
bail:
return -1;
}
#endif
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
LWS_VISIBLE const struct lws_protocols *
lws_get_protocol(struct lws *wsi)
fork-sever-process-and-introduce-broadcast-api.patch Signed-off-by: Andy Green <andy@warmcat.com>
2010-12-18 15:13:50 +00:00
{
return wsi->protocol;
}
Added support for extensions that only manipulate application data.
2013-01-09 18:06:55 +08:00
remove lws_ensure_user_space from public api change return The function has a logical problem when the size of the requested allocation is 0, it will return NULL which is overloaded as failure. Actually the whole function is evil as an api, this patch moves it out of the public API space and fixes it to return 0 for success or 1 for fail. Private code does not need to to return wsi->user_space and public code should only get that from the callback as discussed on trac recently. Thanks to Edwin for debugging the problem. Reported-by: Edwin van den Oetelaar <oetelaar.automatisering@gmail.com> Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-18 16:30:10 +08:00
int
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
lws_ensure_user_space(struct lws *wsi)
Add libwebsockets_ensure_user_space This patch unifies the code for per-connection user allocation, and allows it to be allocated earlier, duiring HTTP service time. It's also OK to attempt allocation multiple times, it will just return any existing allocation. Signed-off-by: Alex Bligh <alex@alex.org.uk>
2011-11-07 17:19:25 +08:00
{
robustness protect and document ensure_user_space Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-15 22:31:55 +08:00
if (!wsi->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
return 0;
robustness protect and document ensure_user_space Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-15 22:31:55 +08:00
Add libwebsockets_ensure_user_space This patch unifies the code for per-connection user allocation, and allows it to be allocated earlier, duiring HTTP service time. It's also OK to attempt allocation multiple times, it will just return any existing allocation. Signed-off-by: Alex Bligh <alex@alex.org.uk>
2011-11-07 17:19:25 +08:00
/* allocate the per-connection user memory (if any) */
if (wsi->protocol->per_session_data_size && !wsi->user_space) {
cleanups
2017-10-28 07:42:44 +08:00
wsi->user_space = lws_zalloc(
wsi->protocol->per_session_data_size, "user space");
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 == NULL) {
clean
2017-09-23 12:55:21 +08:00
lwsl_err("%s: OOM\n", __func__);
remove lws_ensure_user_space from public api change return The function has a logical problem when the size of the requested allocation is 0, it will return NULL which is overloaded as failure. Actually the whole function is evil as an api, this patch moves it out of the public API space and fixes it to return 0 for success or 1 for fail. Private code does not need to to return wsi->user_space and public code should only get that from the callback as discussed on trac recently. Thanks to Edwin for debugging the problem. Reported-by: Edwin van den Oetelaar <oetelaar.automatisering@gmail.com> Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-18 16:30:10 +08:00
return 1;
Add libwebsockets_ensure_user_space This patch unifies the code for per-connection user allocation, and allows it to be allocated earlier, duiring HTTP service time. It's also OK to attempt allocation multiple times, it will just return any existing allocation. Signed-off-by: Alex Bligh <alex@alex.org.uk>
2011-11-07 17:19:25 +08:00
}
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
} else
clean
2017-09-23 12:55:21 +08:00
lwsl_debug("%s: %p protocol pss %lu, user_space=%p\n", __func__,
wsi, (long)wsi->protocol->per_session_data_size,
wsi->user_space);
remove lws_ensure_user_space from public api change return The function has a logical problem when the size of the requested allocation is 0, it will return NULL which is overloaded as failure. Actually the whole function is evil as an api, this patch moves it out of the public API space and fixes it to return 0 for success or 1 for fail. Private code does not need to to return wsi->user_space and public code should only get that from the callback as discussed on trac recently. Thanks to Edwin for debugging the problem. Reported-by: Edwin van den Oetelaar <oetelaar.automatisering@gmail.com> Signed-off-by: Andy Green <andy.green@linaro.org>
2013-02-18 16:30:10 +08:00
return 0;
Add libwebsockets_ensure_user_space This patch unifies the code for per-connection user allocation, and allows it to be allocated earlier, duiring HTTP service time. It's also OK to attempt allocation multiple times, it will just return any existing allocation. Signed-off-by: Alex Bligh <alex@alex.org.uk>
2011-11-07 17:19:25 +08:00
}
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
Plugins: add ssh-base ssh server plugin
2017-10-16 16:59:57 +08:00
LWS_VISIBLE void *
lws_adjust_protocol_psds(struct lws *wsi, size_t new_size)
{
((struct lws_protocols *)lws_get_protocol(wsi))->per_session_data_size =
new_size;
if (lws_ensure_user_space(wsi))
return NULL;
return wsi->user_space;
}
logging timestamp creation expose as api Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 15:01:55 +08:00
LWS_VISIBLE int
lwsl_timestamp(int level, char *p, int len)
logging extend level set api to allow setting emission function Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-12 09:17:42 +08:00
{
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#ifndef LWS_PLAT_OPTEE
codacy: minor fixes
2018-10-10 13:54:43 +08:00
#ifndef _WIN32_WCE
lwsl stderr sink use formatted time AG change to use localtime_r for threadsafety and leave old method as fallback
2016-02-21 07:42:49 +08:00
time_t o_now = time(NULL);
codacy: minor fixes
2018-10-10 13:54:43 +08:00
#endif
Avoid calls to gettimeofday() Add a new function to get the current time in microseconds, since gettimeofday() does not exist on Windows. Keep the current implementation for the test applications.
2014-02-26 21:37:31 +01:00
unsigned long long now;
windows doesnt have localtime_r Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:36:48 +08:00
struct tm *ptm = NULL;
#ifndef WIN32
struct tm tm;
#endif
int n;
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
wince minor adaptations https://github.com/warmcat/libwebsockets/issues/444
2016-02-29 18:48:55 +08:00
#ifndef _WIN32_WCE
windows doesnt have localtime_r Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:36:48 +08:00
#ifdef WIN32
ptm = localtime(&o_now);
#else
if (localtime_r(&o_now, &tm))
ptm = &tm;
wince minor adaptations https://github.com/warmcat/libwebsockets/issues/444
2016-02-29 18:48:55 +08:00
#endif
windows doesnt have localtime_r Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:36:48 +08:00
#endif
logging timestamp creation expose as api Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 15:01:55 +08:00
p[0] = '\0';
cleanups after api changes and mbed update Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 16:54:12 +08:00
for (n = 0; n < LLL_COUNT; n++) {
if (level != (1 << n))
continue;
lws_time_in_microseconds: export Rename time_in_microseconds() and export the internal api as lws_time_in_microseconds()
2018-10-14 06:15:36 +08:00
now = lws_time_in_microseconds() / 100;
windows doesnt have localtime_r Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:36:48 +08:00
if (ptm)
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
n = lws_snprintf(p, len,
logging timestamp creation expose as api Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 15:01:55 +08:00
"[%04d/%02d/%02d %02d:%02d:%02d:%04d] %s: ",
windows doesnt have localtime_r Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:36:48 +08:00
ptm->tm_year + 1900,
lwsl_timestamp month off by one
2016-06-01 08:32:18 +08:00
ptm->tm_mon + 1,
windows doesnt have localtime_r Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:36:48 +08:00
ptm->tm_mday,
ptm->tm_hour,
ptm->tm_min,
ptm->tm_sec,
lwsl stderr sink use formatted time AG change to use localtime_r for threadsafety and leave old method as fallback
2016-02-21 07:42:49 +08:00
(int)(now % 10000), log_level_names[n]);
else
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
n = lws_snprintf(p, len, "[%llu:%04d] %s: ",
lwsl stderr sink use formatted time AG change to use localtime_r for threadsafety and leave old method as fallback
2016-02-21 07:42:49 +08:00
(unsigned long long) now / 10000,
(int)(now % 10000), log_level_names[n]);
logging timestamp creation expose as api Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 15:01:55 +08:00
return n;
cleanups after api changes and mbed update Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 16:54:12 +08:00
}
plat optee: updates
2017-12-20 10:44:21 +08:00
#else
p[0] = '\0';
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#endif
plat optee: updates
2017-12-20 10:44:21 +08:00
logging timestamp creation expose as api Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 15:01:55 +08:00
return 0;
}
plat optee: updates
2017-12-20 10:44:21 +08:00
#ifndef LWS_PLAT_OPTEE
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 char * const colours[] = {
"[31;1m", /* LLL_ERR */
"[36;1m", /* LLL_WARN */
"[35;1m", /* LLL_NOTICE */
"[32;1m", /* LLL_INFO */
"[34;1m", /* LLL_DEBUG */
"[33;1m", /* LLL_PARSER */
logging: stderr: reduce brightness of some ANSI loglevels
2018-06-30 09:40:26 +08:00
"[33m", /* LLL_HEADER */
"[33m", /* LLL_EXT */
"[33m", /* LLL_CLIENT */
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
"[33;1m", /* LLL_LATENCY */
"[30;1m", /* LLL_USER */
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
"[31m", /* LLL_THREAD */
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
};
logs: go back to ansi only on tty and add lwsl_emit_stderr_notimestamp
2018-11-13 17:03:33 +08:00
static char tty;
LWS_VISIBLE void
lwsl_emit_stderr(int level, const char *line)
logging timestamp creation expose as api Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 15:01:55 +08:00
{
char buf[50];
libwebsockets.h: clean out some boilerplate better put in core/private.h https://github.com/warmcat/libwebsockets/issues/1370
2018-08-16 19:10:32 +08:00
int n, m = LWS_ARRAY_SIZE(colours) - 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 (!tty)
tty = isatty(2) | 2;
logging timestamp creation expose as api Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-25 15:01:55 +08:00
lwsl_timestamp(level, buf, sizeof(buf));
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 (tty == 3) {
libwebsockets.h: clean out some boilerplate better put in core/private.h https://github.com/warmcat/libwebsockets/issues/1370
2018-08-16 19:10:32 +08:00
n = 1 << (LWS_ARRAY_SIZE(colours) - 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
while (n) {
if (level & n)
break;
m--;
n >>= 1;
}
fprintf(stderr, "%c%s%s%s%c[0m", 27, colours[m], buf, line, 27);
} else
fprintf(stderr, "%s%s", buf, line);
expose log level in emit Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-19 11:17:56 +08:00
}
logs: go back to ansi only on tty and add lwsl_emit_stderr_notimestamp
2018-11-13 17:03:33 +08:00
LWS_VISIBLE void
lwsl_emit_stderr_notimestamp(int level, const char *line)
{
int n, m = LWS_ARRAY_SIZE(colours) - 1;
if (!tty)
tty = isatty(2) | 2;
if (tty == 3) {
n = 1 << (LWS_ARRAY_SIZE(colours) - 1);
while (n) {
if (level & n)
break;
m--;
n >>= 1;
}
fprintf(stderr, "%c%s%s%c[0m", 27, colours[m], line, 27);
} else
fprintf(stderr, "%s", line);
}
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#endif
helper api: log through syslog Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-19 11:12:16 +08:00
Subject: [PATCH] add _lws_logv to make loging accessible to other wrappers. use _lws_logv in _lws_log
2014-12-10 18:50:28 -06:00
LWS_VISIBLE void _lws_logv(int filter, const char *format, va_list vl)
expose log level in emit Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-19 11:17:56 +08:00
{
char buf[256];
esp8266 initial support
2016-07-23 14:18:25 +08:00
int n;
expose log level in emit Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-19 11:17:56 +08:00
if (!(log_level & filter))
return;
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
esp8266 initial support
2016-07-23 14:18:25 +08:00
n = vsnprintf(buf, sizeof(buf) - 1, format, vl);
(void)n;
/* vnsprintf returns what it would have written, even if truncated */
logging: ellipsisize logs longer than our line buffer Currently the line buffer for vsnprintf() is 256, lines longer than that end abruptly without a CRLF. Change it to end with "...\n\0" when it truncates the line. CSP header additions, logged on vhost init, longer than this are going to become normal...
2018-09-02 07:13:56 +08:00
if (n > (int)sizeof(buf) - 1) {
n = sizeof(buf) - 5;
buf[n++] = '.';
buf[n++] = '.';
buf[n++] = '.';
buf[n++] = '\n';
buf[n] = '\0';
}
esp8266 initial support
2016-07-23 14:18:25 +08:00
if (n > 0)
buf[n] = '\0';
logging extend level set api to allow setting emission function Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-12 09:17:42 +08:00
expose log level in emit Signed-off-by: Andy Green <andy.green@linaro.org>
2013-01-19 11:17:56 +08:00
lwsl_emit(filter, buf);
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
}
Subject: [PATCH] add _lws_logv to make loging accessible to other wrappers. use _lws_logv in _lws_log
2014-12-10 18:50:28 -06:00
LWS_VISIBLE void _lws_log(int filter, const char *format, ...)
{
va_list ap;
va_start(ap, format);
_lws_logv(filter, format, ap);
va_end(ap);
}
clean tidy the worst whitespace alignment probs due to mass token name length changes Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:30:53 +08:00
LWS_VISIBLE void lws_set_log_level(int level,
cleanups after api changes and mbed update Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 16:54:12 +08:00
void (*func)(int level, const char *line))
introduce logging api and convert all library output to use it - multiple debug context calls lwsl_ err, warn, debug, parser, ext, client - api added to set which contexts output to stderr using a bitfield log_level - --disable-debug on configure removes all code that is not err or warn severity - err and warn contexts always output to stderr unless disabled by log_level - err and warn enabled by default in log_level Signed-off-by: Andy Green <andy@warmcat.com>
2013-01-10 19:50:35 +08:00
{
log_level = level;
cleanups after api changes and mbed update Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 16:54:12 +08:00
if (func)
lwsl_emit = func;
Disable compile warning: no newline at end of file
2014-04-27 13:28:22 +02:00
}
add lws_is_ssl api Suggested by Michael Harbler on the mailing list Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-16 09:54:27 +08:00
lwsl_visible
2017-02-09 09:11:17 +08:00
LWS_VISIBLE int lwsl_visible(int level)
{
return log_level & level;
}
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
LWS_VISIBLE void
lwsl_hexdump_level(int hexdump_level, const void *vbuf, size_t len)
{
unsigned char *buf = (unsigned char *)vbuf;
codacy: minor fixes
2018-10-10 13:54:43 +08:00
unsigned int n;
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 (!lwsl_visible(hexdump_level))
return;
autobahn fixes This replaces the old test-app for echo with separate client and server minimal versions. The autobahn test script is made more autonomous and tests both client and server.
2018-04-20 10:33:23 +08:00
if (!len)
return;
if (!vbuf)
return;
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
_lws_log(hexdump_level, "\n");
for (n = 0; n < len;) {
codacy: minor fixes
2018-10-10 13:54:43 +08:00
unsigned int start = n, m;
char line[80], *p = line;
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
p += sprintf(p, "%04X: ", start);
for (m = 0; m < 16 && n < len; m++)
p += sprintf(p, "%02X ", buf[n++]);
while (m++ < 16)
p += sprintf(p, " ");
p += sprintf(p, " ");
for (m = 0; m < 16 && (start + m) < len; m++) {
if (buf[start + m] >= ' ' && buf[start + m] < 127)
*p++ = buf[start + m];
else
*p++ = '.';
}
while (m++ < 16)
*p++ = ' ';
*p++ = '\n';
*p = '\0';
_lws_log(hexdump_level, "%s", line);
(void)line;
}
_lws_log(hexdump_level, "\n");
}
LWS_VISIBLE void
lwsl_hexdump(const void *vbuf, size_t len)
{
lwsl_debug: NOP if no _DEBUG
2018-05-26 09:07:31 +08:00
#if defined(_DEBUG)
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
lwsl_hexdump_level(LLL_DEBUG, vbuf, len);
lwsl_debug: NOP if no _DEBUG
2018-05-26 09:07:31 +08:00
#endif
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
}
add lws_is_ssl api Suggested by Michael Harbler on the mailing list Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-16 09:54:27 +08:00
LWS_VISIBLE int
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
lws_is_ssl(struct lws *wsi)
add lws_is_ssl api Suggested by Michael Harbler on the mailing list Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-16 09:54:27 +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
#if defined(LWS_WITH_TLS)
refactor: also migrate tls to the ops struct and private.h pattern Several new ops are planned for tls... so better to bite the bullet and clean it out to the same level as roles + event-libs first. Also adds a new travis target "mbedtls" and all the tests except autobahn against mbedtls build.
2018-05-01 12:41:42 +08:00
return wsi->tls.use_ssl & LCCSCF_USE_SSL;
fix build without ssl support Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-19 08:41:26 +08:00
#else
mbed3 warning cleaning Signed-off-by: Andy Green <andy.green@linaro.org>
2015-11-02 13:10:33 +08:00
(void)wsi;
fix build without ssl support Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-19 08:41:26 +08:00
return 0;
#endif
add lws_is_ssl api Suggested by Michael Harbler on the mailing list Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-16 09:54:27 +08:00
}
introduce lws_partial_buffered Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-18 22:49:39 +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
#if defined(LWS_WITH_TLS) && !defined(LWS_WITH_MBEDTLS)
tls: split out common, openssl and mbedtls code - introduce lib/tls/mbedtls lib/tls/openssl - move wrapper into lib/tls/mbedtls/wrapper - introduce private helpers to hide backend This patch doesn't replace or remove the wrapper, it moves it to lib/tls/mbedtls/wrapper. But it should be now that the ONLY functions directly consuming wrapper apis are isolated in - lib/tls/mbedtls/client.c (180 lines) - lib/tls/mbedtls/server.c (317 lines) - lib/tls/mbedtls/ssl.c (325 lines) In particular there are no uses of openssl or mbedtls-related constants outside of ./lib/tls any more.
2017-10-18 09:41:44 +08:00
LWS_VISIBLE lws_tls_conn*
ssl expose public wsi->ssl getter Signed-off-by: Denis Osvald <denis.osvald@sartura.hr>
2017-01-23 19:34:46 +08:00
lws_get_ssl(struct lws *wsi)
{
refactor: also migrate tls to the ops struct and private.h pattern Several new ops are planned for tls... so better to bite the bullet and clean it out to the same level as roles + event-libs first. Also adds a new travis target "mbedtls" and all the tests except autobahn against mbedtls build.
2018-05-01 12:41:42 +08:00
return wsi->tls.ssl;
ssl expose public wsi->ssl getter Signed-off-by: Denis Osvald <denis.osvald@sartura.hr>
2017-01-23 19:34:46 +08:00
}
#endif
introduce lws_partial_buffered Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-18 22:49:39 +08:00
LWS_VISIBLE int
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
lws_partial_buffered(struct lws *wsi)
introduce lws_partial_buffered Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-18 22:49:39 +08:00
{
partial: replace ad-hoc code with a wsi buflist_out Various kinds of input stashing were replaced with a single buflist before v3.0... this patch replaces the partial send arrangements with its own buflist in the same way. Buflists as the name says are growable lists of allocations in a linked-list that take care of book-keeping what's added and removed (even if what is removed is less than the current buffer on the list). The immediate result is that we no longer have to freak out if we had a partial buffered and new output is coming... we can just pile it on the end of the buflist and keep draining the front of it. Likewise we no longer need to be rabid about reporting multiple attempts to send stuff without going back to the event loop, although not doing that will introduce inefficiencies we don't have to term it "illegal" any more. Since buflists have proven reliable on the input side and the logic for dealing with truncated "non-network events" was already there this internal-only change should be relatively self-contained.
2018-08-20 12:02:26 +08:00
return lws_has_buffered_out(wsi);
introduce lws_partial_buffered Signed-off-by: Andy Green <andy.green@linaro.org>
2014-08-18 22:49:39 +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
lws_get_peer_write_allowance: return lws_fileofs_t
2018-04-26 08:30:12 +08:00
LWS_VISIBLE lws_fileofs_t
api rationalization: eliminate all libwebsocket[s]_ prefixes This nukes all the oldstyle prefixes except in the compatibility code. struct libwebsockets becomes struct lws too. The api docs are updated accordingly as are the READMEs that mention those apis. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 11:08:32 +08:00
lws_get_peer_write_allowance(struct lws *wsi)
http2 window update Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-29 09:39:08 +08:00
{
lws_get_peer_write_allowance: cope with no role callback for it https://github.com/warmcat/libwebsockets/issues/1339
2018-07-17 18:28:23 +08:00
if (!wsi->role_ops->tx_credit)
return -1;
lws_get_peer_write_allowance: return lws_fileofs_t
2018-04-26 08:30:12 +08:00
return wsi->role_ops->tx_credit(wsi);
http2 window update Signed-off-by: Andy Green <andy.green@linaro.org>
2014-10-29 09:39:08 +08:00
}
lws_union_transition Signed-off-by: Andy Green <andy.green@linaro.org>
2014-11-08 11:18:47 +08:00
LWS_VISIBLE void
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(struct lws *wsi, enum lwsi_role role, enum lwsi_state state,
adopt: allow associated accepted vhost connections to specific role Normalize the vhost options around optionally handling noncompliant traffic at the listening socket for both non-tls and tls cases. By default everything is as before. However it's now possible to tell the vhost to allow noncompliant connects to fall back to a specific role and protocol, both set by name in the vhost creation info struct. The original vhost flags allowing http redirect to https and direct http serving from https server (which is a security downgrade if enabled) are cleaned up and tested. A minimal example minimal-raw-fallback-http-server is added with switches to confirm operation of all the valid possibilities (see the readme on that).
2018-11-29 08:47:49 +08:00
const struct lws_role_ops *ops)
lws_union_transition Signed-off-by: Andy Green <andy.green@linaro.org>
2014-11-08 11:18:47 +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
#if defined(_DEBUG)
const char *name = "(unset)";
#endif
wsi->wsistate = role | state;
if (ops)
wsi->role_ops = ops;
#if defined(_DEBUG)
if (wsi->role_ops)
name = wsi->role_ops->name;
lwsl_debug("%s: %p: wsistate 0x%x, ops %s\n", __func__, wsi,
wsi->wsistate, name);
#endif
lws_union_transition Signed-off-by: Andy Green <andy.green@linaro.org>
2014-11-08 11:18:47 +08:00
}
api rationalization add cmake switch to export old api wrappers This is off by default, use -D LWS_WITH_OLD_API_WRAPPERS=1 on cmake to get the old api names exported from the library as wrappers around the new api names. This allows the library to continue to be compatible with apps that are not rebuilt with the new libwebsockets.h api compatibility defines. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 10:39:23 +08:00
lws_plat_fd implement platform default handlers This is a rewrite of the patch from Soapyman here https://github.com/warmcat/libwebsockets/pull/363 The main changes compared to Soapyman's original patch are - There's no new stuff in the info struct user code does any overrides it may want to do explicitly after lws_context_create returns - User overrides for file ops can call through (subclass) to the original platform implementation using lws_get_fops_plat() - A typedef is provided for plat-specific fd type - Public helpers are provided to allow user code to be platform-independent about file access, using the lws platform file operations underneath: static inline lws_filefd_type lws_plat_file_open(struct lws_plat_file_ops *fops, const char *filename, unsigned long *filelen, int flags) static inline int lws_plat_file_close(struct lws_plat_file_ops *fops, lws_filefd_type fd) static inline unsigned long lws_plat_file_seek_cur(struct lws_plat_file_ops *fops, lws_filefd_type fd, long offset_from_cur_pos) static inline int lws_plat_file_read(struct lws_plat_file_ops *fops, lws_filefd_type fd, unsigned long *amount, unsigned char *buf, unsigned long len) static inline int lws_plat_file_write(struct lws_plat_file_ops *fops, lws_filefd_type fd, unsigned long *amount, unsigned char *buf, unsigned long len) There's example documentation and implementation in the test server. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-10 07:58:58 +08:00
LWS_VISIBLE struct lws_plat_file_ops *
lws_get_fops(struct lws_context *context)
{
fops-zip
2017-03-03 12:38:10 +08:00
return (struct lws_plat_file_ops *)context->fops;
lws_plat_fd implement platform default handlers This is a rewrite of the patch from Soapyman here https://github.com/warmcat/libwebsockets/pull/363 The main changes compared to Soapyman's original patch are - There's no new stuff in the info struct user code does any overrides it may want to do explicitly after lws_context_create returns - User overrides for file ops can call through (subclass) to the original platform implementation using lws_get_fops_plat() - A typedef is provided for plat-specific fd type - Public helpers are provided to allow user code to be platform-independent about file access, using the lws platform file operations underneath: static inline lws_filefd_type lws_plat_file_open(struct lws_plat_file_ops *fops, const char *filename, unsigned long *filelen, int flags) static inline int lws_plat_file_close(struct lws_plat_file_ops *fops, lws_filefd_type fd) static inline unsigned long lws_plat_file_seek_cur(struct lws_plat_file_ops *fops, lws_filefd_type fd, long offset_from_cur_pos) static inline int lws_plat_file_read(struct lws_plat_file_ops *fops, lws_filefd_type fd, unsigned long *amount, unsigned char *buf, unsigned long len) static inline int lws_plat_file_write(struct lws_plat_file_ops *fops, lws_filefd_type fd, unsigned long *amount, unsigned char *buf, unsigned long len) There's example documentation and implementation in the test server. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-10 07:58:58 +08:00
}
api rationalization add cmake switch to export old api wrappers This is off by default, use -D LWS_WITH_OLD_API_WRAPPERS=1 on cmake to get the old api names exported from the library as wrappers around the new api names. This allows the library to continue to be compatible with apps that are not rebuilt with the new libwebsockets.h api compatibility defines. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-04 10:39:23 +08:00
lws_get_ctx conversion Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-11 09:36:14 +08:00
LWS_VISIBLE LWS_EXTERN struct lws_context *
lws_get_context not _ctx Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-17 18:25:25 +08:00
lws_get_context(const struct lws *wsi)
lws_get_ctx conversion Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-11 09:36:14 +08:00
{
return wsi->context;
}
multithreaded service This adds support for multithreaded service to lws without adding any threading or locking code in the library. At context creation time you can request split the service part of the context into n service domains, which are load-balanced so that the most idle one gets the next listen socket accept. There's a single listen socket on one port still. User code may then spawn n threads doing n service loops / poll()s simultaneously. Locking is only required (I think) in the existing FD lock callbacks already handled by the pthreads server example, and that locking takes place in user code. So the library remains completely agnostic about the threading / locking scheme. And by default, it's completely compatible with one service thread so no changes are required by people uninterested in multithreaded service. However for people interested in extremely lightweight mass http[s]/ ws[s] service with minimum provisioning, the library can now do everything out of the box. To test it, just try $ libwebsockets-test-server-pthreads -j 8 where -j controls the number of service threads Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 03:34:24 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_get_count_threads(struct lws_context *context)
{
return context->count_threads;
}
introduce lws_wsi_user Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-14 07:16:32 +08:00
LWS_VISIBLE LWS_EXTERN void *
lws_wsi_user(struct lws *wsi)
{
return wsi->user_space;
}
close add api to control sent close frame contents This adds an api lws_close_reason() which lets you control what will be sent in the close frame when the connection is closed by returning nonzero from the user callback. The test server demo is extended to prove it works in both directions. With this, we should have nice close support. https://github.com/warmcat/libwebsockets/issues/196 Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-26 17:20:34 +08:00
client: allow change externally-defined wsi user_data https://github.com/warmcat/libwebsockets/issues/819 AG: add explanation in doxygen comment and check user_space was externally set
2017-02-28 21:17:25 +08:00
LWS_VISIBLE LWS_EXTERN void
lws_set_wsi_user(struct lws *wsi, void *data)
{
if (wsi->user_space_externally_allocated)
wsi->user_space = data;
else
lwsl_err("%s: Cannot set internally-allocated user_space\n",
__func__);
}
add explicit parent child wsi relationships wsi can have a full tree relationship with each other using linked lists. closing the parent ensures the children are closed first. Convert cgi to use this instead of his cgi-specific sub-wsi management. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-02 09:17:22 +08:00
LWS_VISIBLE LWS_EXTERN struct lws *
lws_get_parent(const struct lws *wsi)
{
return wsi->parent;
}
LWS_VISIBLE LWS_EXTERN struct lws *
lws_get_child(const struct lws *wsi)
{
return wsi->child_list;
}
lws-meta
2017-07-19 04:39:14 +08:00
LWS_VISIBLE LWS_EXTERN void *
lws_get_opaque_parent_data(const struct lws *wsi)
{
return wsi->opaque_parent_data;
}
LWS_VISIBLE LWS_EXTERN void
lws_set_opaque_parent_data(struct lws *wsi, void *data)
{
wsi->opaque_parent_data = data;
}
wsi: opaque_user_data and accessors Under some circumstances it's useful to tag a wsi with user data, while still having an lws-allocated and destroyed pss.
2018-12-01 06:35:20 +08:00
LWS_VISIBLE LWS_EXTERN void *
lws_get_opaque_user_data(const struct lws *wsi)
{
return wsi->opaque_user_data;
}
LWS_VISIBLE LWS_EXTERN void
lws_set_opaque_user_data(struct lws *wsi, void *data)
{
wsi->opaque_user_data = data;
}
lws-meta
2017-07-19 04:39:14 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_get_child_pending_on_writable(const struct lws *wsi)
{
return wsi->parent_pending_cb_on_writable;
}
LWS_VISIBLE LWS_EXTERN void
lws_clear_child_pending_on_writable(struct lws *wsi)
{
wsi->parent_pending_cb_on_writable = 0;
}
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
LWS_EXTERN int
smp: fixes
2018-03-02 14:22:49 +08:00
__lws_rx_flow_control(struct lws *wsi)
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
{
rx flow: handle child state change during parent cb
2017-03-05 15:32:47 +08:00
struct lws *wsic = wsi->child_list;
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
// h2 ignores rx flow control atm
if (lwsi_role_h2(wsi) || wsi->http2_substream ||
lwsi_role_h2_ENCAPSULATION(wsi))
return 0; // !!!
rx flow: handle child state change during parent cb
2017-03-05 15:32:47 +08:00
/* if he has children, do those if they were changed */
while (wsic) {
if (wsic->rxflow_change_to & LWS_RXFLOW_PENDING_CHANGE)
smp: fixes
2018-03-02 14:22:49 +08:00
__lws_rx_flow_control(wsic);
rx flow: handle child state change during parent cb
2017-03-05 15:32:47 +08:00
wsic = wsic->sibling_list;
}
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
/* there is no pending change */
clean
2017-09-23 12:55:21 +08:00
if (!(wsi->rxflow_change_to & LWS_RXFLOW_PENDING_CHANGE))
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
return 0;
/* stuff is still buffered, not ready to really accept new input */
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
if (lws_buflist_next_segment_len(&wsi->buflist, NULL)) {
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
/* get ourselves called back to deal with stashed buffer */
lws_callback_on_writable(wsi);
return 0;
}
lws_buflist For h1 / ws, a combination of removing POLLIN wait and stashing any unused rx lets us immediately respond to rx flow control requests in a simple and effective way, because the tcp socket is the stream. But for muxed protocols like h2, that technique cannot be used because we cannot silence the whole bundle of streams because one can't handle any more rx dynamically. There are control frames and content for other streams serialized inbetween the flow controlled stream content. We have no choice but to read to so we can see the other things. Therefore for muxed protocols like h2 and spdy, rx flow control boils down to tx credit manipulation on individual streams to staunch the flow at the peer. However that requires a round trip to take effect, any transmitted packets that were in flight before the tx credit reduction arrives at the remote peer are still going to come and have to be dealt with by adding them to the stash. This patch introduces lws_buflist scatter-gather type buffer management for rxflow handling, so we can append buffer segments in a linked-list to handle whatever rx is unavoidably in flight on a stream that is trying to assert rx flow control.
2018-04-13 16:01:38 +08:00
/* now the pending is cleared, we can change rxflow state */
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
wsi->rxflow_change_to &= ~LWS_RXFLOW_PENDING_CHANGE;
lwsl_info("rxflow: wsi %p change_to %d\n", wsi,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
wsi->rxflow_change_to & LWS_RXFLOW_ALLOW);
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
/* adjust the pollfd for this wsi */
if (wsi->rxflow_change_to & LWS_RXFLOW_ALLOW) {
smp: fixes
2018-03-02 14:22:49 +08:00
if (__lws_change_pollfd(wsi, 0, LWS_POLLIN)) {
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
lwsl_info("%s: fail\n", __func__);
return -1;
}
} else
smp: fixes
2018-03-02 14:22:49 +08:00
if (__lws_change_pollfd(wsi, LWS_POLLIN, 0))
autobahn add same serverside rxflow cache to client Server side has had immediate RX flow control for quite a while. But client side made do with RX continuing until what had been received was exhausted. For what Autobahn tests, that's not enough. This patch gives clientside RX flow control the same immediate effect as the server side enjoys, re-using the same code. Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-28 14:24:49 +08:00
return -1;
return 0;
}
utf8 check compatible with extensions Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-30 11:43:36 +08:00
lws_tokenize
2018-10-09 10:29:42 +08:00
static const unsigned char e0f4[] = {
0xa0 | ((2 - 1) << 2) | 1, /* e0 */
0x80 | ((4 - 1) << 2) | 1, /* e1 */
0x80 | ((4 - 1) << 2) | 1, /* e2 */
0x80 | ((4 - 1) << 2) | 1, /* e3 */
0x80 | ((4 - 1) << 2) | 1, /* e4 */
0x80 | ((4 - 1) << 2) | 1, /* e5 */
0x80 | ((4 - 1) << 2) | 1, /* e6 */
0x80 | ((4 - 1) << 2) | 1, /* e7 */
0x80 | ((4 - 1) << 2) | 1, /* e8 */
0x80 | ((4 - 1) << 2) | 1, /* e9 */
0x80 | ((4 - 1) << 2) | 1, /* ea */
0x80 | ((4 - 1) << 2) | 1, /* eb */
0x80 | ((4 - 1) << 2) | 1, /* ec */
0x80 | ((2 - 1) << 2) | 1, /* ed */
0x80 | ((4 - 1) << 2) | 1, /* ee */
0x80 | ((4 - 1) << 2) | 1, /* ef */
0x90 | ((3 - 1) << 2) | 2, /* f0 */
0x80 | ((4 - 1) << 2) | 2, /* f1 */
0x80 | ((4 - 1) << 2) | 2, /* f2 */
0x80 | ((4 - 1) << 2) | 2, /* f3 */
0x80 | ((1 - 1) << 2) | 2, /* f4 */
0, /* s0 */
0x80 | ((4 - 1) << 2) | 0, /* s2 */
0x80 | ((4 - 1) << 2) | 1, /* s3 */
};
LWS_EXTERN int
lws_check_byte_utf8(unsigned char state, unsigned char c)
{
unsigned char s = state;
if (!s) {
if (c >= 0x80) {
if (c < 0xc2 || c > 0xf4)
return -1;
if (c < 0xe0)
return 0x80 | ((4 - 1) << 2);
else
return e0f4[c - 0xe0];
}
return s;
}
if (c < (s & 0xf0) || c >= (s & 0xf0) + 0x10 + ((s << 2) & 0x30))
return -1;
return e0f4[21 + (s & 3)];
}
utf8 check compatible with extensions Signed-off-by: Andy Green <andy.green@linaro.org>
2015-12-30 11:43:36 +08:00
LWS_EXTERN int
lws_check_utf8(unsigned char *state, unsigned char *buf, size_t len)
{
unsigned char s = *state;
while (len--) {
unsigned char c = *buf++;
if (!s) {
if (c >= 0x80) {
if (c < 0xc2 || c > 0xf4)
return 1;
if (c < 0xe0)
s = 0x80 | ((4 - 1) << 2);
else
s = e0f4[c - 0xe0];
}
} else {
if (c < (s & 0xf0) ||
c >= (s & 0xf0) + 0x10 + ((s << 2) & 0x30))
return 1;
s = e0f4[21 + (s & 3)];
}
}
*state = s;
return 0;
}
extension permessage deflate Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-11 11:34:01 +08:00
parse_uri Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-14 13:39:02 +08:00
LWS_VISIBLE LWS_EXTERN int
multithreaded service This adds support for multithreaded service to lws without adding any threading or locking code in the library. At context creation time you can request split the service part of the context into n service domains, which are load-balanced so that the most idle one gets the next listen socket accept. There's a single listen socket on one port still. User code may then spawn n threads doing n service loops / poll()s simultaneously. Locking is only required (I think) in the existing FD lock callbacks already handled by the pthreads server example, and that locking takes place in user code. So the library remains completely agnostic about the threading / locking scheme. And by default, it's completely compatible with one service thread so no changes are required by people uninterested in multithreaded service. However for people interested in extremely lightweight mass http[s]/ ws[s] service with minimum provisioning, the library can now do everything out of the box. To test it, just try $ libwebsockets-test-server-pthreads -j 8 where -j controls the number of service threads Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 03:34:24 +08:00
lws_parse_uri(char *p, const char **prot, const char **ads, int *port,
const char **path)
parse_uri Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-14 13:39:02 +08:00
{
const char *end;
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
char unix_skt = 0;
parse_uri Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-14 13:39:02 +08:00
/* cut up the location into address, port and path */
*prot = p;
while (*p && (*p != ':' || p[1] != '/' || p[2] != '/'))
p++;
if (!*p) {
end = p;
p = (char *)*prot;
*prot = end;
} else {
*p = '\0';
p += 3;
}
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 (*p == '+') /* unix skt */
unix_skt = 1;
parse_uri Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-14 13:39:02 +08:00
*ads = p;
if (!strcmp(*prot, "http") || !strcmp(*prot, "ws"))
*port = 80;
else if (!strcmp(*prot, "https") || !strcmp(*prot, "wss"))
*port = 443;
timeout move to doubly linked list In the case we have a lot of connections, checking them all for timeout state once a second becomes burdensome. At the moment if you have 100K connections, once a second they all get checked for timeout in a loop. This patch adds a doubly-linked list based in the context to each wsi, and only wsi with pending timeouts appear on it. At checking time, we traverse the list, which costs nothing if empty because nobody has a pending timeout. Similarly adding and removing from the list costs almost nothing since no iteration is required no matter how big the list. The extra 8 or 16 bytes in the wsi are offset a little bit by demoting .pps from int to char (save 3 bytes). And trim max act exts to 2, since we only provide one, saving 8 /16 bytes by itself if exts enabled. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-19 04:32:14 +08:00
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 (*p == '[') {
++(*ads);
while (*p && *p != ']')
p++;
if (*p)
*p++ = '\0';
} else
while (*p && *p != ':' && (unix_skt || *p != '/'))
p++;
parse_uri Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-14 13:39:02 +08:00
if (*p == ':') {
*p++ = '\0';
*port = atoi(p);
while (*p && *p != '/')
p++;
}
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
*path = "/";
parse_uri Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-14 13:39:02 +08:00
if (*p) {
*p++ = '\0';
if (*p)
*path = p;
}
return 0;
}
refactor: split out adoption and client apis to core - split raw role into separate skt and file - remove all special knowledge from the adoption apis and migrate to core - remove all special knowledge from client_connect stuff, and have it discovered by iterating the role callbacks to let those choose how to bind; migrate to core - retire the old deprecated client apis pre- client_connect_info
2018-05-10 16:13:26 +08:00
char *
lws_strdup(const char *s)
{
char *d = lws_malloc(strlen(s) + 1, "strdup");
if (d)
strcpy(d, s);
return d;
}
align private LWS_NO_EXTENSIONS to be same as public LWS_WITHOUT_EXTENSIONS This is just an internal mass change of LWS_NO_EXTENSIONS to LWS_WITHOUT_EXTENSIONS to match the public name and eliminate all instances of LWS_NO_EXTENSIONS.
2018-03-19 09:33:55 +08:00
#if defined(LWS_WITHOUT_EXTENSIONS)
extension permessage deflate Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-11 11:34:01 +08:00
/* we need to provide dummy callbacks for internal exts
* so user code runs when faced with a lib compiled with
* extensions disabled.
*/
qnx: add support https://libwebsockets.org/pipermail/libwebsockets/2018-January/003600.html
2018-01-04 10:37:25 +08:00
LWS_VISIBLE int
extension permessage deflate Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-11 11:34:01 +08:00
lws_extension_callback_pm_deflate(struct lws_context *context,
const struct lws_extension *ext,
struct lws *wsi,
enum lws_extension_callback_reasons reason,
void *user, void *in, size_t len)
{
mbed align with pt changes Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-20 17:35:18 +08:00
(void)context;
(void)ext;
(void)wsi;
(void)reason;
(void)user;
(void)in;
(void)len;
extension permessage deflate Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-11 11:34:01 +08:00
return 0;
}
refactor: move out extension lws members into ws and establish role structs for vhost content
2018-04-25 06:53:30 +08:00
LWS_EXTERN int
lws_set_extension_option(struct lws *wsi, const char *ext_name,
const char *opt_name, const char *opt_val)
{
return -1;
}
extension permessage deflate Signed-off-by: Andy Green <andy.green@linaro.org>
2016-01-11 11:34:01 +08:00
#endif
bind: treat EADDRINUSE as fatal The retry stuff for bind failures is actually aimed at the scenarios the interface either doesn't exist yet, or is not configured enough (having an IP) to be bindable yet. This patch treats EADDRINUSE as fatal at vhost init.
2018-11-22 17:58:30 +08:00
/* note: this returns a random port, or one of these <= 0 return codes:
*
* LWS_ITOSA_USABLE: the interface is usable, returned if so and sockfd invalid
* LWS_ITOSA_NOT_EXIST: the requested iface does not even exist
* LWS_ITOSA_NOT_USABLE: the requested iface exists but is not usable (eg, no IP)
* LWS_ITOSA_BUSY: the port at the requested iface + port is already in use
*/
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
LWS_EXTERN int
lws_socket_bind: use lws_sockfd_type
2016-11-16 08:59:47 +08:00
lws_socket_bind(struct lws_vhost *vhost, lws_sockfd_type sockfd, int port,
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
const char *iface)
{
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_UNIX_SOCK
Add unix domain socket
2016-03-30 22:47:02 -07:00
struct sockaddr_un serv_unix;
#endif
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_IPV6
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
struct sockaddr_in6 serv_addr6;
#endif
struct sockaddr_in serv_addr4;
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#ifndef LWS_PLAT_OPTEE
Subject: Support to IPv6 on Windows
2017-05-30 18:31:06 -03:00
socklen_t len = sizeof(struct sockaddr_storage);
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#endif
alpn: assemble defaults from roles and allow override Since new roles may be incompatible with http, add support for alpn names at the role struct, automatic generation of the default list of alpn names that servers advertise, and the ability to override the used alpn names per-vhost and per- client connection. This not only lets you modulate visibility or use of h2, but also enables vhosts that only offer non-http roles, as well as restricting http role vhosts to only alpn identifiers related to http roles.
2018-04-12 15:56:38 +08:00
int n;
#if !defined(LWS_WITH_ESP32)
int m;
#endif
Subject: Support to IPv6 on Windows
2017-05-30 18:31:06 -03:00
struct sockaddr_storage sin;
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
struct sockaddr *v;
codacy: minor fixes
2018-10-10 13:54:43 +08:00
memset(&sin, 0, sizeof(sin));
unix socket: fixes and improvements Auto-remove any unix socket file already there. Correctly identify if it's in use per-vhost. Make the peer-limits stuff ignore it.
2018-08-02 19:13:53 +08:00
#if defined(LWS_WITH_UNIX_SOCK)
unix sockets make vhost option Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 12:11:51 +08:00
if (LWS_UNIX_SOCK_ENABLED(vhost)) {
Add unix domain socket
2016-03-30 22:47:02 -07:00
v = (struct sockaddr *)&serv_unix;
n = sizeof(struct sockaddr_un);
bzero((char *) &serv_unix, sizeof(serv_unix));
serv_unix.sun_family = AF_UNIX;
coverity 188416: possibility of NULL iface
2018-03-19 19:20:29 +08:00
if (!iface)
bind: treat EADDRINUSE as fatal The retry stuff for bind failures is actually aimed at the scenarios the interface either doesn't exist yet, or is not configured enough (having an IP) to be bindable yet. This patch treats EADDRINUSE as fatal at vhost init.
2018-11-22 17:58:30 +08:00
return LWS_ITOSA_NOT_EXIST;
Add unix domain socket
2016-03-30 22:47:02 -07:00
if (sizeof(serv_unix.sun_path) <= strlen(iface)) {
lwsl_err("\"%s\" too long for UNIX domain socket\n",
iface);
bind: treat EADDRINUSE as fatal The retry stuff for bind failures is actually aimed at the scenarios the interface either doesn't exist yet, or is not configured enough (having an IP) to be bindable yet. This patch treats EADDRINUSE as fatal at vhost init.
2018-11-22 17:58:30 +08:00
return LWS_ITOSA_NOT_EXIST;
Add unix domain socket
2016-03-30 22:47:02 -07:00
}
strcpy(serv_unix.sun_path, iface);
Support for abstract socket
2016-04-10 13:19:16 +09:00
if (serv_unix.sun_path[0] == '@')
serv_unix.sun_path[0] = '\0';
unix socket: fixes and improvements Auto-remove any unix socket file already there. Correctly identify if it's in use per-vhost. Make the peer-limits stuff ignore it.
2018-08-02 19:13:53 +08:00
else
unlink(serv_unix.sun_path);
Support for abstract socket
2016-04-10 13:19:16 +09:00
Add unix domain socket
2016-03-30 22:47:02 -07:00
} else
#endif
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#if defined(LWS_WITH_IPV6) && !defined(LWS_WITH_ESP32)
ipv6 move disable to vhost option Server ipv6 support disable is now controlled by vhost->options rather than context->options, allowing it to be set per-vhost. Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-03 21:19:40 +08:00
if (LWS_IPV6_ENABLED(vhost)) {
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
v = (struct sockaddr *)&serv_addr6;
n = sizeof(struct sockaddr_in6);
bzero((char *) &serv_addr6, sizeof(serv_addr6));
ipv6-allow-binding-to-ipv6-address-in-iface ipv4 and ipv6 binding to a named interface works OK. ipv4 binding to an IP also works, but we need some extra ipv6 magic to identify the ipv6 interface from an ipv6 address. This patch based on code from "user3546716" at http://stackoverflow.com/questions/13504934/binding-sockets-to-ipv6-addresses adds the necessary magic. https://github.com/warmcat/libwebsockets/issues/717
2016-12-15 09:58:20 +08:00
if (iface) {
network interface: defer bindings to absent network interfaces Previously down network interfaces without an IPv4 address are removed from the posix api that lists network interfaces. That means if you bound a vhost listen socket to a particular interface, it will fail at startup time. This patch adds these vhosts to a list, starts the vhost without a listen socket, and checks to see if the vhost's network interface has appeared while the rest of lws is running. If it appears, the listen socket is opened on the network interface and the vhost becomes reachable.
2018-03-29 13:32:33 +08:00
m = interface_to_sa(vhost, iface,
(struct sockaddr_in *)v, n);
if (m == LWS_ITOSA_NOT_USABLE) {
lwsl_info("%s: netif %s: Not usable\n",
__func__, iface);
return m;
}
if (m == LWS_ITOSA_NOT_EXIST) {
lwsl_info("%s: netif %s: Does not exist\n",
__func__, iface);
return m;
ipv6-allow-binding-to-ipv6-address-in-iface ipv4 and ipv6 binding to a named interface works OK. ipv4 binding to an IP also works, but we need some extra ipv6 magic to identify the ipv6 interface from an ipv6 address. This patch based on code from "user3546716" at http://stackoverflow.com/questions/13504934/binding-sockets-to-ipv6-addresses adds the necessary magic. https://github.com/warmcat/libwebsockets/issues/717
2016-12-15 09:58:20 +08:00
}
Subject: windows: support to bind to a specific IPv6 address
2017-06-05 13:59:22 -03:00
serv_addr6.sin6_scope_id = lws_get_addr_scope(iface);
ipv6-allow-binding-to-ipv6-address-in-iface ipv4 and ipv6 binding to a named interface works OK. ipv4 binding to an IP also works, but we need some extra ipv6 magic to identify the ipv6 interface from an ipv6 address. This patch based on code from "user3546716" at http://stackoverflow.com/questions/13504934/binding-sockets-to-ipv6-addresses adds the necessary magic. https://github.com/warmcat/libwebsockets/issues/717
2016-12-15 09:58:20 +08:00
}
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
serv_addr6.sin6_family = AF_INET6;
serv_addr6.sin6_port = htons(port);
} else
#endif
{
v = (struct sockaddr *)&serv_addr4;
n = sizeof(serv_addr4);
bzero((char *) &serv_addr4, sizeof(serv_addr4));
serv_addr4.sin_addr.s_addr = INADDR_ANY;
serv_addr4.sin_family = AF_INET;
bind: treat EADDRINUSE as fatal The retry stuff for bind failures is actually aimed at the scenarios the interface either doesn't exist yet, or is not configured enough (having an IP) to be bindable yet. This patch treats EADDRINUSE as fatal at vhost init.
2018-11-22 17:58:30 +08:00
#if !defined(LWS_WITH_ESP32)
network interface: defer bindings to absent network interfaces Previously down network interfaces without an IPv4 address are removed from the posix api that lists network interfaces. That means if you bound a vhost listen socket to a particular interface, it will fail at startup time. This patch adds these vhosts to a list, starts the vhost without a listen socket, and checks to see if the vhost's network interface has appeared while the rest of lws is running. If it appears, the listen socket is opened on the network interface and the vhost becomes reachable.
2018-03-29 13:32:33 +08:00
if (iface) {
m = interface_to_sa(vhost, iface,
(struct sockaddr_in *)v, n);
if (m == LWS_ITOSA_NOT_USABLE) {
lwsl_info("%s: netif %s: Not usable\n",
__func__, iface);
return m;
}
if (m == LWS_ITOSA_NOT_EXIST) {
lwsl_info("%s: netif %s: Does not exist\n",
__func__, iface);
return m;
}
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
}
ESP32 platform This is enough for all the test app features to work on ESP32 without SSL.
2017-02-18 17:26:40 +08:00
#endif
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
serv_addr4.sin_port = htons(port);
} /* ipv4 */
network interface: defer bindings to absent network interfaces Previously down network interfaces without an IPv4 address are removed from the posix api that lists network interfaces. That means if you bound a vhost listen socket to a particular interface, it will fail at startup time. This patch adds these vhosts to a list, starts the vhost without a listen socket, and checks to see if the vhost's network interface has appeared while the rest of lws is running. If it appears, the listen socket is opened on the network interface and the vhost becomes reachable.
2018-03-29 13:32:33 +08:00
/* just checking for the interface extant */
if (sockfd == LWS_SOCK_INVALID)
bind: treat EADDRINUSE as fatal The retry stuff for bind failures is actually aimed at the scenarios the interface either doesn't exist yet, or is not configured enough (having an IP) to be bindable yet. This patch treats EADDRINUSE as fatal at vhost init.
2018-11-22 17:58:30 +08:00
return LWS_ITOSA_USABLE;
network interface: defer bindings to absent network interfaces Previously down network interfaces without an IPv4 address are removed from the posix api that lists network interfaces. That means if you bound a vhost listen socket to a particular interface, it will fail at startup time. This patch adds these vhosts to a list, starts the vhost without a listen socket, and checks to see if the vhost's network interface has appeared while the rest of lws is running. If it appears, the listen socket is opened on the network interface and the vhost becomes reachable.
2018-03-29 13:32:33 +08:00
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
n = bind(sockfd, v, n);
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_UNIX_SOCK
unix sockets make vhost option Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 12:11:51 +08:00
if (n < 0 && LWS_UNIX_SOCK_ENABLED(vhost)) {
Add unix domain socket
2016-03-30 22:47:02 -07:00
lwsl_err("ERROR on binding fd %d to \"%s\" (%d %d)\n",
unix socket: fixes and improvements Auto-remove any unix socket file already there. Correctly identify if it's in use per-vhost. Make the peer-limits stuff ignore it.
2018-08-02 19:13:53 +08:00
sockfd, iface, n, LWS_ERRNO);
bind: treat EADDRINUSE as fatal The retry stuff for bind failures is actually aimed at the scenarios the interface either doesn't exist yet, or is not configured enough (having an IP) to be bindable yet. This patch treats EADDRINUSE as fatal at vhost init.
2018-11-22 17:58:30 +08:00
return LWS_ITOSA_NOT_EXIST;
Add unix domain socket
2016-03-30 22:47:02 -07:00
} else
#endif
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
if (n < 0) {
introduce vhosts This patch splits out some lws_context members into a new lws_vhost struct. - ssl state and options per vhost - SSL_CTX for serving and client per vhost - protocols[] per vhost - extensions[] per vhost lws_context maintains a linked list of lws_vhosts. The same lws_context_creation_info struct is used to regulate both the context creation and to create vhosts: for backward compatibility if you didn't provide the new LWS_SERVER_OPTION_EXPLICIT_VHOSTS option, then a default vhost is created at context creation time using the same info data as the context itself. If you will have multiple vhosts though, you should give the LWS_SERVER_OPTION_EXPLICIT_VHOSTS option at context creation time, create the context first and then the vhosts afterwards using lws_create_vhost(contest, &info); Although there is a lot of housekeeping to implement this change, there is almost no additional overhead if you don't use multiple vhosts and very little api impact (no changes to test apps). Signed-off-by: Andy Green <andy@warmcat.com>
2016-03-28 10:10:43 +08:00
lwsl_err("ERROR on binding fd %d to port %d (%d %d)\n",
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
sockfd, port, n, LWS_ERRNO);
bind: treat EADDRINUSE as fatal The retry stuff for bind failures is actually aimed at the scenarios the interface either doesn't exist yet, or is not configured enough (having an IP) to be bindable yet. This patch treats EADDRINUSE as fatal at vhost init.
2018-11-22 17:58:30 +08:00
/* if something already listening, tell caller to fail permanently */
if (LWS_ERRNO == LWS_EADDRINUSE)
return LWS_ITOSA_BUSY;
/* otherwise ask caller to retry later */
return LWS_ITOSA_NOT_EXIST;
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
}
unix socket: fixes and improvements Auto-remove any unix socket file already there. Correctly identify if it's in use per-vhost. Make the peer-limits stuff ignore it.
2018-08-02 19:13:53 +08:00
#if defined(LWS_WITH_UNIX_SOCK)
unix skt: default on
2018-08-24 12:20:38 +08:00
if (LWS_UNIX_SOCK_ENABLED(vhost) && vhost->context->uid)
if (chown(serv_unix.sun_path, vhost->context->uid,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
vhost->context->gid))
unix skt: default on
2018-08-24 12:20:38 +08:00
lwsl_notice("%s: chown for unix skt %s failed\n",
__func__, serv_unix.sun_path);
unix socket: fixes and improvements Auto-remove any unix socket file already there. Correctly identify if it's in use per-vhost. Make the peer-limits stuff ignore it.
2018-08-02 19:13:53 +08:00
#endif
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#ifndef LWS_PLAT_OPTEE
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
if (getsockname(sockfd, (struct sockaddr *)&sin, &len) == -1)
lwsl_warn("getsockname: %s\n", strerror(LWS_ERRNO));
else
plat-optee and boringssl adaptations
2017-01-17 07:01:02 +08:00
#endif
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#if defined(LWS_WITH_IPV6)
Subject: Support to IPv6 on Windows
2017-05-30 18:31:06 -03:00
port = (sin.ss_family == AF_INET6) ?
cleanups
2017-10-28 07:42:44 +08:00
ntohs(((struct sockaddr_in6 *) &sin)->sin6_port) :
ntohs(((struct sockaddr_in *) &sin)->sin_port);
Subject: Support to IPv6 on Windows
2017-05-30 18:31:06 -03:00
#else
build: strict-aliasing on old compiler workaround https://github.com/warmcat/libwebsockets/issues/1067
2017-10-24 19:25:46 +08:00
{
struct sockaddr_in sain;
memcpy(&sain, &sin, sizeof(sain));
port = ntohs(sain.sin_port);
}
socket interface bind generalize to lws_socket_bind Move the socket bind to interface code out of server into libwebsockets.c and make a private api for it. Signed-off-by: Andy Green <andy.green@linaro.org>
2016-03-12 08:18:58 +08:00
#endif
return port;
}
lws_get_vhost_listen_port
2018-04-27 09:01:41 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_get_vhost_listen_port(struct lws_vhost *vhost)
{
return vhost->listen_port;
}
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#if defined(LWS_WITH_IPV6)
Subject: windows: support to bind to a specific IPv6 address
2017-06-05 13:59:22 -03:00
LWS_EXTERN unsigned long
lws_get_addr_scope(const char *ipaddr)
{
unsigned long scope = 0;
#ifndef WIN32
struct ifaddrs *addrs, *addr;
char ip[NI_MAXHOST];
unsigned int i;
getifaddrs(&addrs);
for (addr = addrs; addr; addr = addr->ifa_next) {
if (!addr->ifa_addr ||
addr->ifa_addr->sa_family != AF_INET6)
continue;
getnameinfo(addr->ifa_addr,
sizeof(struct sockaddr_in6),
ip, sizeof(ip),
NULL, 0, NI_NUMERICHOST);
i = 0;
while (ip[i])
if (ip[i++] == '%') {
ip[i - 1] = '\0';
break;
}
if (!strcmp(ip, ipaddr)) {
scope = if_nametoindex(addr->ifa_name);
break;
}
}
freeifaddrs(addrs);
#else
PIP_ADAPTER_ADDRESSES adapter, addrs = NULL;
PIP_ADAPTER_UNICAST_ADDRESS addr;
ULONG size = 0;
DWORD ret;
struct sockaddr_in6 *sockaddr;
char ip[NI_MAXHOST];
unsigned int i;
int found = 0;
for (i = 0; i < 5; i++)
{
ret = GetAdaptersAddresses(AF_INET6, GAA_FLAG_INCLUDE_PREFIX,
clean: general whitespace cleanup
2018-11-23 08:47:56 +08:00
NULL, addrs, &size);
Subject: windows: support to bind to a specific IPv6 address
2017-06-05 13:59:22 -03:00
if ((ret == NO_ERROR) || (ret == ERROR_NO_DATA)) {
break;
} else if (ret == ERROR_BUFFER_OVERFLOW)
{
if (addrs)
free(addrs);
heap instrumentation
2017-10-04 07:10:39 +08:00
addrs = (IP_ADAPTER_ADDRESSES *)malloc(size);
Subject: windows: support to bind to a specific IPv6 address
2017-06-05 13:59:22 -03:00
} else
{
if (addrs)
{
free(addrs);
addrs = NULL;
}
lwsl_err("Failed to get IPv6 address table (%d)", ret);
break;
}
}
clean
2017-09-23 12:55:21 +08:00
if ((ret == NO_ERROR) && (addrs)) {
Subject: windows: support to bind to a specific IPv6 address
2017-06-05 13:59:22 -03:00
adapter = addrs;
clean
2017-09-23 12:55:21 +08:00
while (adapter && !found) {
Subject: windows: support to bind to a specific IPv6 address
2017-06-05 13:59:22 -03:00
addr = adapter->FirstUnicastAddress;
clean
2017-09-23 12:55:21 +08:00
while (addr && !found) {
cleanups
2017-10-28 07:42:44 +08:00
if (addr->Address.lpSockaddr->sa_family ==
AF_INET6) {
clean
2017-09-23 12:55:21 +08:00
sockaddr = (struct sockaddr_in6 *)
(addr->Address.lpSockaddr);
Subject: windows: support to bind to a specific IPv6 address
2017-06-05 13:59:22 -03:00
clean
2017-09-23 12:55:21 +08:00
lws_plat_inet_ntop(sockaddr->sin6_family,
&sockaddr->sin6_addr,
Subject: windows: support to bind to a specific IPv6 address
2017-06-05 13:59:22 -03:00
ip, sizeof(ip));
if (!strcmp(ip, ipaddr)) {
scope = sockaddr->sin6_scope_id;
found = 1;
break;
}
}
addr = addr->Next;
}
adapter = adapter->Next;
}
}
if (addrs)
free(addrs);
#endif
return scope;
}
#endif
introduce urlencode decode and sql escape public apis This adds - simple lws_urlencode() - simple lws_urldecode() - simple lws_sql_purify Those expect the data to all be there and process it up until the first '\0'. There is also a larger opaque apis for handling POST_BODY urldecode. To enable these, you need to give cmake -DLWS_WITH_STATEFUL_URLDECODE=1 (or arrange any larger feature that relies on it sets that in CMakeLists.txt) - stateful urldecode with parameter array These have create / process / destroy semantics on a struct that maintains decode state. Stateful urldecode is capable of dealing with large POST data in multiple POST_BODY callbacks cleanly, eg, file transfer by POST. Stateful urldecode with parameter array wraps the above with a canned callback that stores the urldecoded data and indexes them in a pointer array matching an array of parameter names. You may also pass it an optional callback when creating it, that will recieve uploaded file content. The test html is updated to support both urlencoded and multipart forms, with some javascript to do clientside validation of an arbitrary 100KB file size limit (there is no file size limit in the apis). Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-08 10:07:02 +08:00
static const char *hex = "0123456789ABCDEF";
LWS_VISIBLE LWS_EXTERN const char *
lws_sql_purify(char *escaped, const char *string, int len)
{
const char *p = string;
char *q = escaped;
while (*p && len-- > 2) {
if (*p == '\'') {
protocol generic sessions Signed-off-by: Andy Green <andy@warmcat.com>
2016-05-19 15:28:31 +08:00
*q++ = '\'';
introduce urlencode decode and sql escape public apis This adds - simple lws_urlencode() - simple lws_urldecode() - simple lws_sql_purify Those expect the data to all be there and process it up until the first '\0'. There is also a larger opaque apis for handling POST_BODY urldecode. To enable these, you need to give cmake -DLWS_WITH_STATEFUL_URLDECODE=1 (or arrange any larger feature that relies on it sets that in CMakeLists.txt) - stateful urldecode with parameter array These have create / process / destroy semantics on a struct that maintains decode state. Stateful urldecode is capable of dealing with large POST data in multiple POST_BODY callbacks cleanly, eg, file transfer by POST. Stateful urldecode with parameter array wraps the above with a canned callback that stores the urldecoded data and indexes them in a pointer array matching an array of parameter names. You may also pass it an optional callback when creating it, that will recieve uploaded file content. The test html is updated to support both urlencoded and multipart forms, with some javascript to do clientside validation of an arbitrary 100KB file size limit (there is no file size limit in the apis). Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-08 10:07:02 +08:00
*q++ = '\'';
len --;
p++;
} else
*q++ = *p++;
}
*q = '\0';
return escaped;
}
lws_json_escape Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-17 10:05:23 +08:00
LWS_VISIBLE LWS_EXTERN const char *
lws_json_purify(char *escaped, const char *string, int len)
{
const char *p = string;
char *q = escaped;
plugin table dirlisting
2016-06-26 06:29:20 +08:00
if (!p) {
escaped[0] = '\0';
return escaped;
}
lws_json_escape Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-17 10:05:23 +08:00
while (*p && len-- > 6) {
plugins: gitws libjsongit2 support This adds a plugin that interfaces to libjsongit2 https://warmcat.com/git/libjsongit2 to provide a per-vhost service for presenting bare git repos in a web interface.
2018-08-14 08:00:25 +08:00
if (*p == '\t') {
p++;
*q++ = '\\';
*q++ = 't';
continue;
}
if (*p == '\n') {
p++;
*q++ = '\\';
*q++ = 'n';
continue;
}
if (*p == '\r') {
p++;
*q++ = '\\';
*q++ = 'r';
continue;
}
lws_json_escape Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-17 10:05:23 +08:00
if (*p == '\"' || *p == '\\' || *p < 0x20) {
*q++ = '\\';
*q++ = 'u';
*q++ = '0';
*q++ = '0';
*q++ = hex[((*p) >> 4) & 15];
*q++ = hex[(*p) & 15];
len -= 5;
p++;
} else
*q++ = *p++;
}
*q = '\0';
return escaped;
}
lws_filename_purify_inplace
2018-03-29 09:28:41 +08:00
LWS_VISIBLE LWS_EXTERN void
lws_filename_purify_inplace(char *filename)
{
while (*filename) {
if (*filename == '.' && filename[1] == '.') {
*filename = '_';
filename[1] = '_';
}
if (*filename == ':' ||
*filename == '\\' ||
*filename == '$' ||
*filename == '%')
*filename = '_';
filename++;
}
}
introduce urlencode decode and sql escape public apis This adds - simple lws_urlencode() - simple lws_urldecode() - simple lws_sql_purify Those expect the data to all be there and process it up until the first '\0'. There is also a larger opaque apis for handling POST_BODY urldecode. To enable these, you need to give cmake -DLWS_WITH_STATEFUL_URLDECODE=1 (or arrange any larger feature that relies on it sets that in CMakeLists.txt) - stateful urldecode with parameter array These have create / process / destroy semantics on a struct that maintains decode state. Stateful urldecode is capable of dealing with large POST data in multiple POST_BODY callbacks cleanly, eg, file transfer by POST. Stateful urldecode with parameter array wraps the above with a canned callback that stores the urldecoded data and indexes them in a pointer array matching an array of parameter names. You may also pass it an optional callback when creating it, that will recieve uploaded file content. The test html is updated to support both urlencoded and multipart forms, with some javascript to do clientside validation of an arbitrary 100KB file size limit (there is no file size limit in the apis). Signed-off-by: Andy Green <andy@warmcat.com>
2016-06-08 10:07:02 +08:00
LWS_VISIBLE LWS_EXTERN const char *
lws_urlencode(char *escaped, const char *string, int len)
{
const char *p = string;
char *q = escaped;
while (*p && len-- > 3) {
if (*p == ' ') {
*q++ = '+';
p++;
continue;
}
if ((*p >= '0' && *p <= '9') ||
(*p >= 'A' && *p <= 'Z') ||
(*p >= 'a' && *p <= 'z')) {
*q++ = *p++;
continue;
}
*q++ = '%';
*q++ = hex[(*p >> 4) & 0xf];
*q++ = hex[*p & 0xf];
len -= 2;
p++;
}
*q = '\0';
return escaped;
}
LWS_VISIBLE LWS_EXTERN int
lws_urldecode(char *string, const char *escaped, int len)
{
int state = 0, n;
char sum = 0;
while (*escaped && len) {
switch (state) {
case 0:
if (*escaped == '%') {
state++;
escaped++;
continue;
}
if (*escaped == '+') {
escaped++;
*string++ = ' ';
len--;
continue;
}
*string++ = *escaped++;
len--;
break;
case 1:
n = char_to_hex(*escaped);
if (n < 0)
return -1;
escaped++;
sum = n << 4;
state++;
break;
case 2:
n = char_to_hex(*escaped);
if (n < 0)
return -1;
escaped++;
*string++ = sum | n;
len--;
state = 0;
break;
}
}
*string = '\0';
return 0;
}
introduce vhosts This patch splits out some lws_context members into a new lws_vhost struct. - ssl state and options per vhost - SSL_CTX for serving and client per vhost - protocols[] per vhost - extensions[] per vhost lws_context maintains a linked list of lws_vhosts. The same lws_context_creation_info struct is used to regulate both the context creation and to create vhosts: for backward compatibility if you didn't provide the new LWS_SERVER_OPTION_EXPLICIT_VHOSTS option, then a default vhost is created at context creation time using the same info data as the context itself. If you will have multiple vhosts though, you should give the LWS_SERVER_OPTION_EXPLICIT_VHOSTS option at context creation time, create the context first and then the vhosts afterwards using lws_create_vhost(contest, &info); Although there is a lot of housekeeping to implement this change, there is almost no additional overhead if you don't use multiple vhosts and very little api impact (no changes to test apps). Signed-off-by: Andy Green <andy@warmcat.com>
2016-03-28 10:10:43 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_finalize_startup(struct lws_context *context)
{
struct lws_context_creation_info info;
info.uid = context->uid;
info.gid = context->gid;
capabilities support
2017-06-07 06:10:02 +08:00
#if defined(LWS_HAVE_SYS_CAPABILITY_H) && defined(LWS_HAVE_LIBCAP)
memcpy(info.caps, context->caps, sizeof(info.caps));
info.count_caps = context->count_caps;
#endif
introduce vhosts This patch splits out some lws_context members into a new lws_vhost struct. - ssl state and options per vhost - SSL_CTX for serving and client per vhost - protocols[] per vhost - extensions[] per vhost lws_context maintains a linked list of lws_vhosts. The same lws_context_creation_info struct is used to regulate both the context creation and to create vhosts: for backward compatibility if you didn't provide the new LWS_SERVER_OPTION_EXPLICIT_VHOSTS option, then a default vhost is created at context creation time using the same info data as the context itself. If you will have multiple vhosts though, you should give the LWS_SERVER_OPTION_EXPLICIT_VHOSTS option at context creation time, create the context first and then the vhosts afterwards using lws_create_vhost(contest, &info); Although there is a lot of housekeeping to implement this change, there is almost no additional overhead if you don't use multiple vhosts and very little api impact (no changes to test apps). Signed-off-by: Andy Green <andy@warmcat.com>
2016-03-28 10:10:43 +08:00
if (lws_check_opt(context->options, LWS_SERVER_OPTION_EXPLICIT_VHOSTS))
lws_plat_drop_app_privileges(&info);
return 0;
}
lws_get_effective_uid_gid This helper lets you find out the eventual uid / gid lws will run under before the root privileges are actually dropped
2018-08-02 19:15:19 +08:00
LWS_VISIBLE LWS_EXTERN void
lws_get_effective_uid_gid(struct lws_context *context, int *uid, int *gid)
{
*uid = context->uid;
*gid = context->gid;
}
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
int
lws_snprintf(char *str, size_t size, const char *format, ...)
{
va_list ap;
int n;
if (!size)
return 0;
va_start(ap, format);
n = vsnprintf(str, size, format, ap);
va_end(ap);
win32: enable 64-bit file lengths https://github.com/warmcat/libwebsockets/issues/916 AG added more appveyor warning cleaning and stitched in cmake tests
2017-06-09 20:20:42 +08:00
if (n >= (int)size)
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
return (int)size;
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
return n;
}
lws_snprintf
2018-03-12 09:28:26 +08:00
char *
lws_strncpy(char *dest, const char *src, size_t size)
{
strncpy(dest, src, size - 1);
dest[size - 1] = '\0';
return dest;
}
lws_tokenize
2018-10-09 10:29:42 +08:00
typedef enum {
LWS_TOKZS_LEADING_WHITESPACE,
LWS_TOKZS_QUOTED_STRING,
LWS_TOKZS_TOKEN,
LWS_TOKZS_TOKEN_POST_TERMINAL
} lws_tokenize_state;
int
lws_tokenize(struct lws_tokenize *ts)
{
const char *rfc7230_delims = "(),/:;<=>?@[\\]{}";
lws_tokenize_state state = LWS_TOKZS_LEADING_WHITESPACE;
tokenizer: flags for dot not delim and floats as strings
2018-11-13 09:34:10 +08:00
char c, flo = 0, d_minus = '-', d_dot = '.', s_minus = '\0',
s_dot = '\0';
tokenizer: use signed char where needed
2018-10-26 13:48:31 +08:00
signed char num = -1;
lws_tokenize
2018-10-09 10:29:42 +08:00
int utf8 = 0;
tokenizer: flags for dot not delim and floats as strings
2018-11-13 09:34:10 +08:00
/* for speed, compute the effect of the flags outside the loop */
if (ts->flags & LWS_TOKENIZE_F_MINUS_NONTERM) {
d_minus = '\0';
s_minus = '-';
}
if (ts->flags & LWS_TOKENIZE_F_DOT_NONTERM) {
d_dot = '\0';
s_dot = '.';
}
lws_tokenize
2018-10-09 10:29:42 +08:00
ts->token = NULL;
ts->token_len = 0;
while (ts->len) {
c = *ts->start++;
ts->len--;
utf8 = lws_check_byte_utf8((unsigned char)utf8, c);
if (utf8 < 0)
return LWS_TOKZE_ERR_BROKEN_UTF8;
if (!c)
break;
/* whitespace */
if (c == ' ' || c == '\t' || c == '\n' || c == '\r' ||
c == '\f') {
switch (state) {
case LWS_TOKZS_LEADING_WHITESPACE:
case LWS_TOKZS_TOKEN_POST_TERMINAL:
continue;
case LWS_TOKZS_QUOTED_STRING:
ts->token_len++;
continue;
case LWS_TOKZS_TOKEN:
/* we want to scan forward to look for = */
state = LWS_TOKZS_TOKEN_POST_TERMINAL;
continue;
}
}
/* quoted string */
if (c == '\"') {
if (state == LWS_TOKZS_QUOTED_STRING)
return LWS_TOKZE_QUOTED_STRING;
/* starting a quoted string */
if (ts->flags & LWS_TOKENIZE_F_COMMA_SEP_LIST) {
if (ts->delim == LWSTZ_DT_NEED_DELIM)
return LWS_TOKZE_ERR_COMMA_LIST;
ts->delim = LWSTZ_DT_NEED_DELIM;
}
state = LWS_TOKZS_QUOTED_STRING;
ts->token = ts->start;
ts->token_len = 0;
continue;
}
/* token= aggregation */
if (c == '=' && (state == LWS_TOKZS_TOKEN_POST_TERMINAL ||
state == LWS_TOKZS_TOKEN)) {
if (num == 1)
return LWS_TOKZE_ERR_NUM_ON_LHS;
/* swallow the = */
return LWS_TOKZE_TOKEN_NAME_EQUALS;
}
/* optional token: aggregation */
if ((ts->flags & LWS_TOKENIZE_F_AGG_COLON) && c == ':' &&
(state == LWS_TOKZS_TOKEN_POST_TERMINAL ||
state == LWS_TOKZS_TOKEN))
/* swallow the : */
return LWS_TOKZE_TOKEN_NAME_COLON;
/* aggregate . in a number as a float */
tokenizer: flags for dot not delim and floats as strings
2018-11-13 09:34:10 +08:00
if (c == '.' && !(ts->flags & LWS_TOKENIZE_F_NO_FLOATS) &&
state == LWS_TOKZS_TOKEN && num == 1) {
lws_tokenize
2018-10-09 10:29:42 +08:00
if (flo)
return LWS_TOKZE_ERR_MALFORMED_FLOAT;
flo = 1;
ts->token_len++;
continue;
}
/*
* Delimiter... by default anything that:
*
* - isn't matched earlier, or
* - is [A-Z, a-z, 0-9, _], and
* - is not a partial utf8 char
*
* is a "delimiter", it marks the end of a token and is itself
* reported as a single LWS_TOKZE_DELIMITER each time.
*
* However with LWS_TOKENIZE_F_RFC7230_DELIMS flag, tokens may
* contain any noncontrol character that isn't defined in
* rfc7230_delims, and only characters listed there are treated
* as delimiters.
*/
if (!utf8 &&
((ts->flags & LWS_TOKENIZE_F_RFC7230_DELIMS &&
strchr(rfc7230_delims, c) && c > 32) ||
((!(ts->flags & LWS_TOKENIZE_F_RFC7230_DELIMS) &&
(c < '0' || c > '9') && (c < 'A' || c > 'Z') &&
tokenizer: flags for dot not delim and floats as strings
2018-11-13 09:34:10 +08:00
(c < 'a' || c > 'z') && c != '_') &&
c != s_minus && c != s_dot) ||
c == d_minus || c == d_dot
lws_tokenize
2018-10-09 10:29:42 +08:00
)) {
switch (state) {
case LWS_TOKZS_LEADING_WHITESPACE:
if (ts->flags & LWS_TOKENIZE_F_COMMA_SEP_LIST) {
if (c != ',' ||
ts->delim != LWSTZ_DT_NEED_DELIM)
return LWS_TOKZE_ERR_COMMA_LIST;
ts->delim = LWSTZ_DT_NEED_NEXT_CONTENT;
}
ts->token = ts->start - 1;
ts->token_len = 1;
return LWS_TOKZE_DELIMITER;
case LWS_TOKZS_QUOTED_STRING:
ts->token_len++;
continue;
case LWS_TOKZS_TOKEN_POST_TERMINAL:
case LWS_TOKZS_TOKEN:
/* report the delimiter next time */
ts->start--;
ts->len++;
goto token_or_numeric;
}
}
/* anything that's not whitespace or delimiter is payload */
switch (state) {
case LWS_TOKZS_LEADING_WHITESPACE:
if (ts->flags & LWS_TOKENIZE_F_COMMA_SEP_LIST) {
if (ts->delim == LWSTZ_DT_NEED_DELIM)
return LWS_TOKZE_ERR_COMMA_LIST;
ts->delim = LWSTZ_DT_NEED_DELIM;
}
state = LWS_TOKZS_TOKEN;
ts->token = ts->start - 1;
ts->token_len = 1;
if (c < '0' || c > '9')
num = 0;
else
if (num < 0)
num = 1;
continue;
case LWS_TOKZS_QUOTED_STRING:
case LWS_TOKZS_TOKEN:
if (c < '0' || c > '9')
num = 0;
else
if (num < 0)
num = 1;
ts->token_len++;
continue;
case LWS_TOKZS_TOKEN_POST_TERMINAL:
/* report the new token next time */
ts->start--;
ts->len++;
goto token_or_numeric;
}
}
/* we ran out of content */
if (utf8) /* ended partway through a multibyte char */
return LWS_TOKZE_ERR_BROKEN_UTF8;
if (state == LWS_TOKZS_QUOTED_STRING)
return LWS_TOKZE_ERR_UNTERM_STRING;
if (state != LWS_TOKZS_TOKEN_POST_TERMINAL &&
state != LWS_TOKZS_TOKEN) {
if ((ts->flags & LWS_TOKENIZE_F_COMMA_SEP_LIST) &&
ts->delim == LWSTZ_DT_NEED_NEXT_CONTENT)
return LWS_TOKZE_ERR_COMMA_LIST;
return LWS_TOKZE_ENDED;
}
/* report the pending token */
token_or_numeric:
if (num != 1)
return LWS_TOKZE_TOKEN;
if (flo)
return LWS_TOKZE_FLOAT;
return LWS_TOKZE_INTEGER;
}
LWS_VISIBLE LWS_EXTERN int
lws_tokenize_cstr(struct lws_tokenize *ts, char *str, int max)
{
if (ts->token_len + 1 >= max)
return 1;
memcpy(str, ts->token, ts->token_len);
str[ts->token_len] = '\0';
return 0;
}
LWS_VISIBLE LWS_EXTERN void
lws_tokenize_init(struct lws_tokenize *ts, const char *start, int flags)
{
ts->start = start;
ts->len = 0x7fffffff;
ts->flags = flags;
ts->delim = LWSTZ_DT_NEED_FIRST_CONTENT;
}
lws_mutex_refcount This creates a "pthread mutex with a reference count" using gcc / clang atomic intrinsics + pthreads. Both pt and context locks are moved to use this, pt already had reference counting but it's new for context.
2018-06-27 07:15:39 +08:00
#if LWS_MAX_SMP > 1
void
lws_mutex_refcount_init(struct lws_mutex_refcount *mr)
{
pthread_mutex_init(&mr->lock, NULL);
mr->last_lock_reason = NULL;
mr->lock_depth = 0;
mr->metadata = 0;
mr->lock_owner = 0;
}
void
lws_mutex_refcount_destroy(struct lws_mutex_refcount *mr)
{
pthread_mutex_destroy(&mr->lock);
}
void
lws_mutex_refcount_lock(struct lws_mutex_refcount *mr, const char *reason)
{
/* if true, this sequence is atomic because our thread has the lock
*
* - if true, only guy who can race to make it untrue is our thread,
* and we are here.
*
* - if false, only guy who could race to make it true is our thread,
* and we are here
*
* - it can be false and change to a different tid that is also false
*/
if (mr->lock_owner == pthread_self()) {
/* atomic because we only change it if we own the lock */
mr->lock_depth++;
return;
}
pthread_mutex_lock(&mr->lock);
/* atomic because only we can have the lock */
mr->last_lock_reason = reason;
mr->lock_owner = pthread_self();
mr->lock_depth = 1;
//lwsl_notice("tid %d: lock %s\n", mr->tid, reason);
}
void
lws_mutex_refcount_unlock(struct lws_mutex_refcount *mr)
{
if (--mr->lock_depth)
/* atomic because only thread that has the lock can unlock */
return;
mr->last_lock_reason = "free";
mr->lock_owner = 0;
//lwsl_notice("tid %d: unlock %s\n", mr->tid, mr->last_lock_reason);
pthread_mutex_unlock(&mr->lock);
}
#endif /* SMP */
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
cgi Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:25:48 +08:00
LWS_VISIBLE LWS_EXTERN int
lws_is_cgi(struct lws *wsi) {
#ifdef LWS_WITH_CGI
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
return !!wsi->http.cgi;
cgi Signed-off-by: Andy Green <andy.green@linaro.org>
2016-02-21 21:25:48 +08:00
#else
return 0;
#endif
}
minimal: updates
2018-03-12 10:45:32 +08:00
const struct lws_protocol_vhost_options *
lws_pvo_search(const struct lws_protocol_vhost_options *pvo, const char *name)
{
while (pvo) {
if (!strcmp(pvo->name, name))
break;
cgi: also allow for generic spawn
2017-09-19 07:58:04 +08:00
minimal: updates
2018-03-12 10:45:32 +08:00
pvo = pvo->next;
}
return pvo;
}
extensions add api for user code option manipulation Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-08 09:45:49 +08:00
lws_pvo_get_str
2018-08-15 09:23:35 +08:00
int
lws_pvo_get_str(void *in, const char *name, const char **result)
{
const struct lws_protocol_vhost_options *pv =
lws_pvo_search((const struct lws_protocol_vhost_options *)in,
name);
if (!pv)
return 1;
*result = (const char *)pv->value;
return 0;
}
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
void
lws_sum_stats(const struct lws_context *ctx, struct lws_conn_stats *cs)
{
const struct lws_vhost *vh = ctx->vhost_list;
while (vh) {
cs->rx += vh->conn_stats.rx;
cs->tx += vh->conn_stats.tx;
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
cs->h1_conn += vh->conn_stats.h1_conn;
cs->h1_trans += vh->conn_stats.h1_trans;
cs->h2_trans += vh->conn_stats.h2_trans;
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
cs->ws_upg += vh->conn_stats.ws_upg;
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
cs->h2_upg += vh->conn_stats.h2_upg;
cs->h2_alpn += vh->conn_stats.h2_alpn;
cs->h2_subs += vh->conn_stats.h2_subs;
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
cs->rejected += vh->conn_stats.rejected;
vh = vh->vhost_next;
}
}
alpn: assemble defaults from roles and allow override Since new roles may be incompatible with http, add support for alpn names at the role struct, automatic generation of the default list of alpn names that servers advertise, and the ability to override the used alpn names per-vhost and per- client connection. This not only lets you modulate visibility or use of h2, but also enables vhosts that only offer non-http roles, as well as restricting http role vhosts to only alpn identifiers related to http roles.
2018-04-12 15:56:38 +08:00
const char *
lws_cmdline_option(int argc, const char **argv, const char *val)
{
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 n = (int)strlen(val), c = argc;
alpn: assemble defaults from roles and allow override Since new roles may be incompatible with http, add support for alpn names at the role struct, automatic generation of the default list of alpn names that servers advertise, and the ability to override the used alpn names per-vhost and per- client connection. This not only lets you modulate visibility or use of h2, but also enables vhosts that only offer non-http roles, as well as restricting http role vhosts to only alpn identifiers related to http roles.
2018-04-12 15:56:38 +08:00
coverity: check argv
2018-05-05 06:58:52 +08:00
while (--c > 0) {
lws_ring: implement lagging client culling
2018-05-05 10:16:54 +08:00
minimal: add switches and exit results
2018-04-16 07:32:02 +08:00
if (!strncmp(argv[c], val, n)) {
lws_ring: implement lagging client culling
2018-05-05 10:16:54 +08:00
if (!*(argv[c] + n) && c < argc - 1) {
/* coverity treats unchecked argv as "tainted" */
if (!argv[c + 1] || strlen(argv[c + 1]) > 1024)
return NULL;
alpn: assemble defaults from roles and allow override Since new roles may be incompatible with http, add support for alpn names at the role struct, automatic generation of the default list of alpn names that servers advertise, and the ability to override the used alpn names per-vhost and per- client connection. This not only lets you modulate visibility or use of h2, but also enables vhosts that only offer non-http roles, as well as restricting http role vhosts to only alpn identifiers related to http roles.
2018-04-12 15:56:38 +08:00
return argv[c + 1];
lws_ring: implement lagging client culling
2018-05-05 10:16:54 +08:00
}
alpn: assemble defaults from roles and allow override Since new roles may be incompatible with http, add support for alpn names at the role struct, automatic generation of the default list of alpn names that servers advertise, and the ability to override the used alpn names per-vhost and per- client connection. This not only lets you modulate visibility or use of h2, but also enables vhosts that only offer non-http roles, as well as restricting http role vhosts to only alpn identifiers related to http roles.
2018-04-12 15:56:38 +08:00
return argv[c] + n;
}
coverity: check argv
2018-05-05 06:58:52 +08:00
}
alpn: assemble defaults from roles and allow override Since new roles may be incompatible with http, add support for alpn names at the role struct, automatic generation of the default list of alpn names that servers advertise, and the ability to override the used alpn names per-vhost and per- client connection. This not only lets you modulate visibility or use of h2, but also enables vhosts that only offer non-http roles, as well as restricting http role vhosts to only alpn identifiers related to http roles.
2018-04-12 15:56:38 +08:00
return NULL;
}
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
#ifdef LWS_WITH_SERVER_STATUS
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
LWS_EXTERN int
lws_json_dump_vhost(const struct lws_vhost *vh, char *buf, int len)
{
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
#if defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
static const char * const prots[] = {
"http://",
"https://",
"file://",
"cgi://",
">http://",
">https://",
add protocol plugin for post demo Signed-off-by: Andy Green <andy@warmcat.com>
2016-05-09 09:37:01 +08:00
"callback://"
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
};
refactor: most preparation for -DLWS_ROLE_H1=0
2018-04-27 19:16:50 +08:00
#endif
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
char *orig = buf, *end = buf + len - 1, first = 1;
int n = 0;
if (len < 100)
return 0;
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf,
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
"{\n \"name\":\"%s\",\n"
" \"port\":\"%d\",\n"
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
" \"use_ssl\":\"%d\",\n"
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
" \"sts\":\"%d\",\n"
stats upgrade rx tx to long long Signed-off-by: Andy Green <andy@warmcat.com>
2016-05-04 15:59:27 +08:00
" \"rx\":\"%llu\",\n"
" \"tx\":\"%llu\",\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
" \"h1_conn\":\"%lu\",\n"
" \"h1_trans\":\"%lu\",\n"
" \"h2_trans\":\"%lu\",\n"
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
" \"ws_upg\":\"%lu\",\n"
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
" \"rejected\":\"%lu\",\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
" \"h2_upg\":\"%lu\",\n"
" \"h2_alpn\":\"%lu\",\n"
" \"h2_subs\":\"%lu\""
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
,
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
vh->name, vh->listen_port,
refactor role ops This only refactors internal architecture and representations, the user api is unaffected.
2018-04-11 13:39:42 +08:00
#if defined(LWS_WITH_TLS)
refactor: also migrate tls to the ops struct and private.h pattern Several new ops are planned for tls... so better to bite the bullet and clean it out to the same level as roles + event-libs first. Also adds a new travis target "mbedtls" and all the tests except autobahn against mbedtls build.
2018-05-01 12:41:42 +08:00
vh->tls.use_ssl & LCCSCF_USE_SSL,
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
#else
0,
#endif
!!(vh->options & LWS_SERVER_OPTION_STS),
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
vh->conn_stats.rx, vh->conn_stats.tx,
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
vh->conn_stats.h1_conn,
vh->conn_stats.h1_trans,
vh->conn_stats.h2_trans,
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
vh->conn_stats.ws_upg,
vh->conn_stats.rejected,
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
vh->conn_stats.h2_upg,
vh->conn_stats.h2_alpn,
vh->conn_stats.h2_subs
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
);
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 defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
if (vh->http.mount_list) {
const struct lws_http_mount *m = vh->http.mount_list;
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, ",\n \"mounts\":[");
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
while (m) {
if (!first)
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, ",");
buf += lws_snprintf(buf, end - buf,
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
"\n {\n \"mountpoint\":\"%s\",\n"
http cache policy This allows mounts to define the caching policy of the files inside them. Support is added in lwsws for controlling it from the config files. The api for serializing a mount struct opaquely is removed and lws_http_mount struct made public... it was getting out of control trying to hide the options. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-22 08:53:49 +08:00
" \"origin\":\"%s%s\",\n"
" \"cache_max_age\":\"%d\",\n"
" \"cache_reuse\":\"%d\",\n"
" \"cache_revalidate\":\"%d\",\n"
" \"cache_intermediaries\":\"%d\"\n"
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
,
m->mountpoint,
prots[m->origin_protocol],
http cache policy This allows mounts to define the caching policy of the files inside them. Support is added in lwsws for controlling it from the config files. The api for serializing a mount struct opaquely is removed and lws_http_mount struct made public... it was getting out of control trying to hide the options. Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-22 08:53:49 +08:00
m->origin,
m->cache_max_age,
m->cache_reusable,
m->cache_revalidate,
m->cache_intermediaries);
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
if (m->def)
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf,
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
",\n \"default\":\"%s\"",
m->def);
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, "\n }");
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
first = 0;
m = m->mount_next;
}
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, "\n ]");
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
}
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
#endif
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
if (vh->protocols) {
n = 0;
first = 1;
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, ",\n \"ws-protocols\":[");
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
while (n < vh->count_protocols) {
if (!first)
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, ",");
buf += lws_snprintf(buf, end - buf,
win fix warnings from appveyor Signed-off-by: Andy Green <andy@warmcat.com>
2016-05-15 08:29:37 +08:00
"\n {\n \"%s\":{\n"
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
" \"status\":\"ok\"\n }\n }"
,
vh->protocols[n].name);
first = 0;
n++;
}
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, "\n ]");
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
}
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, "\n}");
json dump vhost Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-14 15:07:44 +08:00
return buf - orig;
}
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
LWS_EXTERN LWS_VISIBLE int
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
lws_json_dump_context(const struct lws_context *context, char *buf, int len,
int hide_vhosts)
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
{
char *orig = buf, *end = buf + len - 1, first = 1;
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
const struct lws_vhost *vh = context->vhost_list;
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
const struct lws_context_per_thread *pt;
time_t t = time(NULL);
coverity 175436: dead cruft
2017-02-06 10:04:04 +08:00
int n, listening = 0, cgi_count = 0;
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
struct lws_conn_stats cs;
double d = 0;
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
#ifdef LWS_WITH_CGI
struct lws_cgi * const *pcgi;
#endif
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
CMake: convert all LWS_USE_... to LWS_WITH_... Almost all the CMake options begin with LWS_WITH_..., but many of the symbols passed to lws are LWS_USE_... , this causes neededless confusion, compounded by the fact that a few CMake options also begin with LWS_USE_. This patch globally converts all LWS_USE_... to LWS_WITH_..., so there is only one prefix to remember in both CMake and the code. The affected public CMake options are LWS_USE_BORINGSSL -> LWS_WITH_BORINGSSL LWS_USE_CYASSL -> LWS_WITH_CYASSL LWS_USE_WOLFSSL -> LWS_WITH_WOLFSSL LWS_USE_MBEDTLS -> LWS_WITH_MBEDTLS LWS_USE_BUNDLED_ZLIB -> LWS_WITH_BUNDLED_ZLIB
2017-09-28 11:29:03 +08:00
#ifdef LWS_WITH_LIBUV
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
uv_uptime(&d);
#endif
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, "{ "
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
"\"version\":\"%s\",\n"
"\"uptime\":\"%ld\",\n",
lws_get_library_version(),
(long)d);
server status add getloadavg and lib version and more css Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-26 07:45:45 +08:00
#ifdef LWS_HAVE_GETLOADAVG
{
double d[3];
int m;
m = getloadavg(d, 3);
for (n = 0; n < m; n++) {
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf,
server status add getloadavg and lib version and more css Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-26 07:45:45 +08:00
"\"l%d\":\"%.2f\",\n",
n + 1, d[n]);
}
}
#endif
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
buf += lws_snprintf(buf, end - buf, "\"contexts\":[\n");
buf += lws_snprintf(buf, end - buf, "{ "
"\"context_uptime\":\"%ld\",\n"
"\"cgi_spawned\":\"%d\",\n"
"\"pt_fd_max\":\"%d\",\n"
"\"ah_pool_max\":\"%d\",\n"
"\"deprecated\":\"%d\",\n"
"\"wsi_alive\":\"%d\",\n",
(unsigned long)(t - context->time_up),
context->count_cgi_spawned,
context->fd_limit_per_thread,
context->max_http_header_pool,
context->deprecated,
context->count_wsi_allocated);
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, "\"pt\":[\n ");
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
for (n = 0; n < context->count_threads; n++) {
pt = &context->pt[n];
if (n)
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, ",");
buf += lws_snprintf(buf, end - buf,
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
"\n {\n"
" \"fds_count\":\"%d\",\n"
" \"ah_pool_inuse\":\"%d\",\n"
" \"ah_wait_list\":\"%d\"\n"
" }",
pt->fds_count,
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
pt->http.ah_count_in_use,
pt->http.ah_wait_list_length);
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
}
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
buf += lws_snprintf(buf, end - buf, "]");
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
buf += lws_snprintf(buf, end - buf, ", \"vhosts\":[\n ");
first = 1;
vh = context->vhost_list;
listening = 0;
cs = context->conn_stats;
lws_sum_stats(context, &cs);
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
while (vh) {
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
if (!hide_vhosts) {
if (!first)
if(buf != end)
*buf++ = ',';
buf += lws_json_dump_vhost(vh, buf, end - buf);
first = 0;
}
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
if (vh->lserv_wsi)
listening++;
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
vh = vh->vhost_next;
}
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
buf += lws_snprintf(buf, end - buf,
"],\n\"listen_wsi\":\"%d\",\n"
" \"rx\":\"%llu\",\n"
" \"tx\":\"%llu\",\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
" \"h1_conn\":\"%lu\",\n"
" \"h1_trans\":\"%lu\",\n"
" \"h2_trans\":\"%lu\",\n"
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
" \"ws_upg\":\"%lu\",\n"
" \"rejected\":\"%lu\",\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
" \"h2_alpn\":\"%lu\",\n"
" \"h2_subs\":\"%lu\",\n"
" \"h2_upg\":\"%lu\"",
listening, cs.rx, cs.tx,
cs.h1_conn,
cs.h1_trans,
cs.h2_trans,
cs.ws_upg,
cs.rejected,
cs.h2_alpn,
cs.h2_subs,
cs.h2_upg);
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
#ifdef LWS_WITH_CGI
for (n = 0; n < context->count_threads; n++) {
pt = &context->pt[n];
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
pcgi = &pt->http.cgi_list;
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
while (*pcgi) {
pcgi = &(*pcgi)->cgi_list;
cgi_count++;
}
}
#endif
lws_snprintf Thanks to Fabrice Gilot for reporting the problem that led to uncovering this. Due to a misunderstanding of the return value of snprintf (it is not truncated according to the max size passed in) in several places relying on snprintf to truncate the length overflows are possible. This patch wraps snprintf with a new lws_snprintf() which does truncate its length to allow the buffer limiting scheme to work properly. All users should update with these fixes.
2016-09-15 02:22:57 +08:00
buf += lws_snprintf(buf, end - buf, ",\n \"cgi_alive\":\"%d\"\n ",
cgi lifecycle logging Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-20 06:10:56 +08:00
cgi_count);
context deprecation 1) This makes lwsws run a parent process with the original permissions. But this process is only able to respond to SIGHUP, it doesn't do anything else. 2) You can send this parent process a SIGHUP now to cause it to - close listening sockets in existing lwsws processes - mark those processes as to exit when the number of active connections on the falls to zero - spawn a fresh child process from scratch, using latest configuration file content, latest plugins, etc. It can now reopen listening sockets if it chooses to, or open different listen ports or whatever. Notes: 1) lws_context_destroy() has been split into two pieces... the reason for the split is the first part closes the per-vhost protocols, but since they may have created libuv objects in the per-vhost protocol storage, these cannot be freed until after the loop has been run. That's the purpose of the second part of the context destruction, lws_context_destroy2(). For compatibility, if you are not using libuv, the first part calls the second part. However if you are using libuv, you must now call the second part from your own main.c after the first part.
2016-12-16 07:37:43 +08:00
buf += lws_snprintf(buf, end - buf, "}");
buf += lws_snprintf(buf, end - buf, "]}\n ");
plugin lws server status Signed-off-by: Andy Green <andy@warmcat.com>
2016-04-15 14:01:29 +08:00
return buf - orig;
}
#endif
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
#if defined(LWS_WITH_STATS)
LWS_VISIBLE LWS_EXTERN uint64_t
lws_stats_get(struct lws_context *context, int index)
{
if (index >= LWSSTATS_SIZE)
return 0;
return context->lws_stats[index];
}
LWS_VISIBLE LWS_EXTERN void
lws_stats_log_dump(struct lws_context *context)
{
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
struct lws_vhost *v = context->vhost_list;
codacy: minor fixes
2018-10-10 13:54:43 +08:00
int n;
#if defined(LWS_WITH_PEER_LIMITS)
int m;
#endif
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
if (!context->updated)
return;
context->updated = 0;
lwsl_notice("\n");
lwsl_notice("LWS internal statistics dump ----->\n");
cleanups
2017-10-28 07:42:44 +08:00
lwsl_notice("LWSSTATS_C_CONNECTIONS: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_CONNECTIONS));
lwsl_notice("LWSSTATS_C_API_CLOSE: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_API_CLOSE));
lwsl_notice("LWSSTATS_C_API_READ: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_API_READ));
lwsl_notice("LWSSTATS_C_API_LWS_WRITE: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_API_LWS_WRITE));
lwsl_notice("LWSSTATS_C_API_WRITE: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_API_WRITE));
lwsl_notice("LWSSTATS_C_WRITE_PARTIALS: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_WRITE_PARTIALS));
lwsl_notice("LWSSTATS_C_WRITEABLE_CB_REQ: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_WRITEABLE_CB_REQ));
lwsl_notice("LWSSTATS_C_WRITEABLE_CB_EFF_REQ: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_WRITEABLE_CB_EFF_REQ));
lwsl_notice("LWSSTATS_C_WRITEABLE_CB: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_WRITEABLE_CB));
lwsl_notice("LWSSTATS_C_SSL_CONNECTIONS_ACCEPT_SPIN: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_SSL_CONNECTIONS_ACCEPT_SPIN));
lwsl_notice("LWSSTATS_C_SSL_CONNECTIONS_FAILED: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_SSL_CONNECTIONS_FAILED));
lwsl_notice("LWSSTATS_C_SSL_CONNECTIONS_ACCEPTED: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_SSL_CONNECTIONS_ACCEPTED));
lwsl_notice("LWSSTATS_C_SSL_CONNS_HAD_RX: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_SSL_CONNS_HAD_RX));
lwsl_notice("LWSSTATS_C_PEER_LIMIT_AH_DENIED: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_PEER_LIMIT_AH_DENIED));
lwsl_notice("LWSSTATS_C_PEER_LIMIT_WSI_DENIED: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_PEER_LIMIT_WSI_DENIED));
lwsl_notice("LWSSTATS_C_TIMEOUTS: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_TIMEOUTS));
lwsl_notice("LWSSTATS_C_SERVICE_ENTRY: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_C_SERVICE_ENTRY));
lwsl_notice("LWSSTATS_B_READ: %8llu\n",
(unsigned long long)lws_stats_get(context, LWSSTATS_B_READ));
lwsl_notice("LWSSTATS_B_WRITE: %8llu\n",
(unsigned long long)lws_stats_get(context, LWSSTATS_B_WRITE));
lwsl_notice("LWSSTATS_B_PARTIALS_ACCEPTED_PARTS: %8llu\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_B_PARTIALS_ACCEPTED_PARTS));
lwsl_notice("LWSSTATS_MS_SSL_CONNECTIONS_ACCEPTED_DELAY: %8llums\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_MS_SSL_CONNECTIONS_ACCEPTED_DELAY) / 1000);
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
if (lws_stats_get(context, LWSSTATS_C_SSL_CONNECTIONS_ACCEPTED))
lwsl_notice(" Avg accept delay: %8llums\n",
cleanups
2017-10-28 07:42:44 +08:00
(unsigned long long)(lws_stats_get(context,
LWSSTATS_MS_SSL_CONNECTIONS_ACCEPTED_DELAY) /
lws_stats_get(context,
LWSSTATS_C_SSL_CONNECTIONS_ACCEPTED)) / 1000);
lwsl_notice("LWSSTATS_MS_SSL_RX_DELAY: %8llums\n",
(unsigned long long)lws_stats_get(context,
LWSSTATS_MS_SSL_RX_DELAY) / 1000);
esp32: multi ap slots
2017-05-13 10:26:59 +08:00
if (lws_stats_get(context, LWSSTATS_C_SSL_CONNS_HAD_RX))
lwsl_notice(" Avg accept-rx delay: %8llums\n",
cleanups
2017-10-28 07:42:44 +08:00
(unsigned long long)(lws_stats_get(context,
LWSSTATS_MS_SSL_RX_DELAY) /
lws_stats_get(context,
LWSSTATS_C_SSL_CONNS_HAD_RX)) / 1000);
esp32: multi ap slots
2017-05-13 10:26:59 +08:00
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
lwsl_notice("LWSSTATS_MS_WRITABLE_DELAY: %8lluus\n",
cleanups
2017-10-28 07:42:44 +08:00
(unsigned long long)lws_stats_get(context,
LWSSTATS_MS_WRITABLE_DELAY));
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
lwsl_notice("LWSSTATS_MS_WORST_WRITABLE_DELAY: %8lluus\n",
cleanups
2017-10-28 07:42:44 +08:00
(unsigned long long)lws_stats_get(context,
LWSSTATS_MS_WORST_WRITABLE_DELAY));
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
if (lws_stats_get(context, LWSSTATS_C_WRITEABLE_CB))
lwsl_notice(" Avg writable delay: %8lluus\n",
cleanups
2017-10-28 07:42:44 +08:00
(unsigned long long)(lws_stats_get(context,
LWSSTATS_MS_WRITABLE_DELAY) /
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
lws_stats_get(context, LWSSTATS_C_WRITEABLE_CB)));
refactor: also migrate tls to the ops struct and private.h pattern Several new ops are planned for tls... so better to bite the bullet and clean it out to the same level as roles + event-libs first. Also adds a new travis target "mbedtls" and all the tests except autobahn against mbedtls build.
2018-05-01 12:41:42 +08:00
lwsl_notice("Simultaneous SSL restriction: %8d/%d\n",
cleanups
2017-10-28 07:42:44 +08:00
context->simultaneous_ssl,
refactor: also migrate tls to the ops struct and private.h pattern Several new ops are planned for tls... so better to bite the bullet and clean it out to the same level as roles + event-libs first. Also adds a new travis target "mbedtls" and all the tests except autobahn against mbedtls build.
2018-05-01 12:41:42 +08:00
context->simultaneous_ssl_restriction);
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
cleanups
2017-10-28 07:42:44 +08:00
lwsl_notice("Live wsi: %8d\n",
context->count_wsi_allocated);
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
context->updated = 1;
while (v) {
windows: cleanup wrong and duplicated socket validity helpers https://github.com/warmcat/libwebsockets/issues/1259
2018-05-06 07:19:21 +08:00
if (v->lserv_wsi &&
v->lserv_wsi->position_in_fds_table != LWS_NO_FDS_POS) {
WITH_STATS: remove accidental dependency on ssl
2017-06-28 09:52:22 +08:00
cleanups
2017-10-28 07:42:44 +08:00
struct lws_context_per_thread *pt =
&context->pt[(int)v->lserv_wsi->tsi];
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
struct lws_pollfd *pfd;
pfd = &pt->fds[v->lserv_wsi->position_in_fds_table];
lwsl_notice(" Listen port %d actual POLLIN: %d\n",
cleanups
2017-10-28 07:42:44 +08:00
v->listen_port,
(int)pfd->events & LWS_POLLIN);
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
}
v = v->vhost_next;
}
for (n = 0; n < context->count_threads; n++) {
struct lws_context_per_thread *pt = &context->pt[n];
struct lws *wl;
int m = 0;
lwsl_notice("PT %d\n", n + 1);
smp: fixes
2018-03-02 14:22:49 +08:00
lws_pt_lock(pt, __func__);
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
lwsl_notice(" AH in use / max: %d / %d\n",
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
pt->http.ah_count_in_use,
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
context->max_http_header_pool);
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
wl = pt->http.ah_wait_list;
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
while (wl) {
m++;
LWS_WITH_STATS: clean refactor dust https://github.com/warmcat/libwebsockets/issues/1383
2018-08-25 05:43:31 +08:00
wl = wl->http.ah_wait_list;
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
}
lwsl_notice(" AH wait list count / actual: %d / %d\n",
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
pt->http.ah_wait_list_length, m);
stats: ah and ssl restriction stats
2017-05-15 07:30:06 +08:00
lws_pt_unlock(pt);
}
LWS_WITH_PEER_LIMITS Introduces an optional, settable restriction on the number of simultaneous wsi connections and ah that a single peer IP can obtain. The default is disabled for build, and if enabled, unlimited. However when enabled at CMake, setting info.ip_limit_ah and / or info.ip_limit_wsi at context creation time will enforce the limits. Connections past the info.ip_limit_wsi limit are dropped, and connections needing an ah are forced to stay in the ah waiting list even when ah are available, while the peer is at the number of ah in info.ip_limit_ah.
2017-09-14 13:14:11 +08:00
#if defined(LWS_WITH_PEER_LIMITS)
m = 0;
WITH_STATS: fix signed build warnings
2017-11-05 06:54:31 +08:00
for (n = 0; n < (int)context->pl_hash_elements; n++) {
lws_start_foreach_llp(struct lws_peer **, peer,
context->pl_hash_table[n]) {
LWS_WITH_PEER_LIMITS Introduces an optional, settable restriction on the number of simultaneous wsi connections and ah that a single peer IP can obtain. The default is disabled for build, and if enabled, unlimited. However when enabled at CMake, setting info.ip_limit_ah and / or info.ip_limit_wsi at context creation time will enforce the limits. Connections past the info.ip_limit_wsi limit are dropped, and connections needing an ah are forced to stay in the ah waiting list even when ah are available, while the peer is at the number of ah in info.ip_limit_ah.
2017-09-14 13:14:11 +08:00
m++;
} lws_end_foreach_llp(peer, next);
}
lwsl_notice(" Peers: total active %d\n", m);
if (m > 10) {
m = 10;
lwsl_notice(" (showing 10 peers only)\n");
}
if (m) {
WITH_STATS: fix signed build warnings
2017-11-05 06:54:31 +08:00
for (n = 0; n < (int)context->pl_hash_elements; n++) {
LWS_WITH_PEER_LIMITS Introduces an optional, settable restriction on the number of simultaneous wsi connections and ah that a single peer IP can obtain. The default is disabled for build, and if enabled, unlimited. However when enabled at CMake, setting info.ip_limit_ah and / or info.ip_limit_wsi at context creation time will enforce the limits. Connections past the info.ip_limit_wsi limit are dropped, and connections needing an ah are forced to stay in the ah waiting list even when ah are available, while the peer is at the number of ah in info.ip_limit_ah.
2017-09-14 13:14:11 +08:00
char buf[72];
cleanups
2017-10-28 07:42:44 +08:00
lws_start_foreach_llp(struct lws_peer **, peer,
context->pl_hash_table[n]) {
LWS_WITH_PEER_LIMITS Introduces an optional, settable restriction on the number of simultaneous wsi connections and ah that a single peer IP can obtain. The default is disabled for build, and if enabled, unlimited. However when enabled at CMake, setting info.ip_limit_ah and / or info.ip_limit_wsi at context creation time will enforce the limits. Connections past the info.ip_limit_wsi limit are dropped, and connections needing an ah are forced to stay in the ah waiting list even when ah are available, while the peer is at the number of ah in info.ip_limit_ah.
2017-09-14 13:14:11 +08:00
struct lws_peer *df = *peer;
if (!lws_plat_inet_ntop(df->af, df->addr, buf,
sizeof(buf) - 1))
strcpy(buf, "unknown");
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 defined(LWS_ROLE_H1) || defined(LWS_ROLE_H2)
LWS_WITH_PEER_LIMITS Introduces an optional, settable restriction on the number of simultaneous wsi connections and ah that a single peer IP can obtain. The default is disabled for build, and if enabled, unlimited. However when enabled at CMake, setting info.ip_limit_ah and / or info.ip_limit_wsi at context creation time will enforce the limits. Connections past the info.ip_limit_wsi limit are dropped, and connections needing an ah are forced to stay in the ah waiting list even when ah are available, while the peer is at the number of ah in info.ip_limit_ah.
2017-09-14 13:14:11 +08:00
lwsl_notice(" peer %s: count wsi: %d, count ah: %d\n",
LWS_WITH_STATS: clean refactor dust https://github.com/warmcat/libwebsockets/issues/1383
2018-08-25 05:43:31 +08:00
buf, df->count_wsi,
df->http.count_ah);
refactor: move all ah and http specific vars to http private and conditional inclusion from there
2018-04-27 15:20:56 +08:00
#else
lwsl_notice(" peer %s: count wsi: %d\n",
buf, df->count_wsi);
#endif
LWS_WITH_PEER_LIMITS Introduces an optional, settable restriction on the number of simultaneous wsi connections and ah that a single peer IP can obtain. The default is disabled for build, and if enabled, unlimited. However when enabled at CMake, setting info.ip_limit_ah and / or info.ip_limit_wsi at context creation time will enforce the limits. Connections past the info.ip_limit_wsi limit are dropped, and connections needing an ah are forced to stay in the ah waiting list even when ah are available, while the peer is at the number of ah in info.ip_limit_ah.
2017-09-14 13:14:11 +08:00
if (!--m)
break;
} lws_end_foreach_llp(peer, next);
}
}
#endif
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
lwsl_notice("\n");
}
void
lws_stats_atomic_bump(struct lws_context * context,
struct lws_context_per_thread *pt, int index, uint64_t bump)
{
smp: fixes
2018-03-02 14:22:49 +08:00
lws_pt_stats_lock(pt);
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
context->lws_stats[index] += bump;
if (index != LWSSTATS_C_SERVICE_ENTRY)
context->updated = 1;
smp: fixes
2018-03-02 14:22:49 +08:00
lws_pt_stats_unlock(pt);
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
}
void
lws_stats_atomic_max(struct lws_context * context,
struct lws_context_per_thread *pt, int index, uint64_t val)
{
smp: fixes
2018-03-02 14:22:49 +08:00
lws_pt_stats_lock(pt);
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
if (val > context->lws_stats[index]) {
context->lws_stats[index] = val;
context->updated = 1;
}
smp: fixes
2018-03-02 14:22:49 +08:00
lws_pt_stats_unlock(pt);
LWS_WITH_STATS
2017-05-07 10:02:03 +08:00
}
#endif
lws_ring: introduce generic flexible ringbuffer abstraction
2017-09-20 10:37:59 +08:00
Reference in a new issue Copy permalink