1
0
Fork 0
mirror of https://github.com/warmcat/libwebsockets.git synced 2025-03-23 00:00:06 +01:00
libwebsockets/lib/service.c

903 lines
23 KiB
C
Raw Normal View History

/*
* libwebsockets - small server side websockets and web server implementation
*
* Copyright (C) 2010-2018 Andy Green <andy@warmcat.com>
*
* 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
*/
#include "private-libwebsockets.h"
int
lws_callback_as_writeable(struct lws *wsi)
{
2017-05-07 10:02:03 +08:00
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
int n, m;
2017-05-07 10:02:03 +08:00
lws_stats_atomic_bump(wsi->context, pt, LWSSTATS_C_WRITEABLE_CB, 1);
#if defined(LWS_WITH_STATS)
if (wsi->active_writable_req_us) {
2017-09-23 12:55:21 +08:00
uint64_t ul = time_in_microseconds() -
wsi->active_writable_req_us;
2017-05-07 10:02:03 +08:00
2017-09-23 12:55:21 +08:00
lws_stats_atomic_bump(wsi->context, pt,
LWSSTATS_MS_WRITABLE_DELAY, ul);
lws_stats_atomic_max(wsi->context, pt,
LWSSTATS_MS_WORST_WRITABLE_DELAY, ul);
2017-05-07 10:02:03 +08:00
wsi->active_writable_req_us = 0;
}
#endif
n = wsi->role_ops->writeable_cb[lwsi_role_server(wsi)];
m = user_callback_handle_rxflow(wsi->protocol->callback,
wsi, (enum lws_callback_reasons) n,
wsi->user_space, NULL, 0);
return m;
}
LWS_VISIBLE int
lws_handle_POLLOUT_event(struct lws *wsi, struct lws_pollfd *pollfd)
{
volatile struct lws *vwsi = (volatile struct lws *)wsi;
int n;
//lwsl_notice("%s: %p\n", __func__, wsi);
vwsi->leave_pollout_active = 0;
vwsi->handling_pollout = 1;
/*
* if another thread wants POLLOUT on us, from here on while
* handling_pollout is set, he will only set leave_pollout_active.
* If we are going to disable POLLOUT, we will check that first.
*/
wsi->could_have_pending = 0; /* clear back-to-back write detection */
/*
* user callback is lowest priority to get these notifications
* actually, since other pending things cannot be disordered
*
* Priority 1: pending truncated sends are incomplete ws fragments
* If anything else sent first the protocol would be
* corrupted.
*/
if (wsi->trunc_len) {
//lwsl_notice("%s: completing partial\n", __func__);
if (lws_issue_raw(wsi, wsi->trunc_alloc + wsi->trunc_offset,
wsi->trunc_len) < 0) {
lwsl_info("%s signalling to close\n", __func__);
goto bail_die;
}
/* leave POLLOUT active either way */
goto bail_ok;
} else
if (lwsi_state(wsi) == LRS_FLUSHING_BEFORE_CLOSE) {
wsi->socket_is_permanently_unusable = 1;
goto bail_die; /* retry closing now */
}
#ifdef LWS_WITH_CGI
2017-09-23 12:55:21 +08:00
/*
* A cgi master's wire protocol remains h1 or h2. He is just getting
* his data from his child cgis.
*/
2017-03-22 20:15:01 +08:00
if (wsi->cgi) {
/* also one shot */
if (pollfd)
if (lws_change_pollfd(wsi, LWS_POLLOUT, 0)) {
lwsl_info("failed at set pollfd\n");
return 1;
}
goto user_service_go_again;
2017-03-22 20:15:01 +08:00
}
#endif
/* if we got here, we should have wire protocol ops set on the wsi */
assert(wsi->role_ops);
if (!wsi->role_ops->handle_POLLOUT)
goto bail_ok;
switch ((wsi->role_ops->handle_POLLOUT)(wsi)) {
case LWS_HP_RET_BAIL_OK:
goto bail_ok;
case LWS_HP_RET_BAIL_DIE:
goto bail_die;
case LWS_HP_RET_USER_SERVICE:
break;
default:
assert(0);
}
/* one shot */
2017-07-19 04:39:14 +08:00
if (wsi->parent_carries_io) {
vwsi->handling_pollout = 0;
vwsi->leave_pollout_active = 0;
2017-07-19 04:39:14 +08:00
return lws_callback_as_writeable(wsi);
2017-07-19 04:39:14 +08:00
}
if (pollfd) {
int eff = vwsi->leave_pollout_active;
2018-04-17 11:43:20 +08:00
if (!eff) {
if (lws_change_pollfd(wsi, LWS_POLLOUT, 0)) {
lwsl_info("failed at set pollfd\n");
goto bail_die;
}
2018-04-17 11:43:20 +08:00
}
vwsi->handling_pollout = 0;
/* cannot get leave_pollout_active set after the above */
if (!eff && wsi->leave_pollout_active) {
/*
* got set inbetween sampling eff and clearing
* handling_pollout, force POLLOUT on
*/
lwsl_debug("leave_pollout_active\n");
if (lws_change_pollfd(wsi, 0, LWS_POLLOUT)) {
lwsl_info("failed at set pollfd\n");
goto bail_die;
}
}
vwsi->leave_pollout_active = 0;
}
if (lwsi_role_client(wsi) &&
!wsi->hdr_parsing_completed &&
2018-04-17 11:43:20 +08:00
lwsi_state(wsi) != LRS_H2_WAITING_TO_SEND_HEADERS &&
lwsi_state(wsi) != LRS_ISSUE_HTTP_BODY
)
goto bail_ok;
#ifdef LWS_WITH_CGI
user_service_go_again:
#endif
if (wsi->role_ops->perform_user_POLLOUT) {
if (wsi->role_ops->perform_user_POLLOUT(wsi) == -1)
goto bail_die;
else
goto bail_ok;
}
lwsl_debug("%s: %p: non mux: wsistate 0x%x, ops %s\n", __func__, wsi,
wsi->wsistate, wsi->role_ops->name);
2018-03-11 11:26:06 +08:00
vwsi = (volatile struct lws *)wsi;
vwsi->leave_pollout_active = 0;
n = lws_callback_as_writeable(wsi);
vwsi->handling_pollout = 0;
if (vwsi->leave_pollout_active)
lws_change_pollfd(wsi, 0, LWS_POLLOUT);
return n;
/*
* since these don't disable the POLLOUT, they are always doing the
* right thing for leave_pollout_active whether it was set or not.
*/
bail_ok:
vwsi->handling_pollout = 0;
vwsi->leave_pollout_active = 0;
return 0;
bail_die:
vwsi->handling_pollout = 0;
vwsi->leave_pollout_active = 0;
return -1;
}
2018-03-05 16:49:28 +08:00
static int
__lws_service_timeout_check(struct lws *wsi, time_t sec)
{
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
int n = 0;
2016-07-23 14:18:25 +08:00
(void)n;
/*
* if we went beyond the allowed time, kill the
* connection
*/
if (wsi->dll_timeout.prev &&
lws_compare_time_t(wsi->context, sec, wsi->pending_timeout_set) >
2018-03-05 16:49:28 +08:00
wsi->pending_timeout_limit) {
2017-09-23 12:55:21 +08:00
if (wsi->desc.sockfd != LWS_SOCK_INVALID &&
wsi->position_in_fds_table >= 0)
2016-07-23 14:18:25 +08:00
n = pt->fds[wsi->position_in_fds_table].events;
2017-05-07 10:02:03 +08:00
lws_stats_atomic_bump(wsi->context, pt, LWSSTATS_C_TIMEOUTS, 1);
/* no need to log normal idle keepalive timeout */
if (wsi->pending_timeout != PENDING_TIMEOUT_HTTP_KEEPALIVE_IDLE)
2017-09-23 12:55:21 +08:00
lwsl_info("wsi %p: TIMEDOUT WAITING on %d "
"(did hdr %d, ah %p, wl %d, pfd "
"events %d) %llu vs %llu\n",
(void *)wsi, wsi->pending_timeout,
wsi->hdr_parsing_completed, wsi->ah,
2017-09-23 12:55:21 +08:00
pt->ah_wait_list_length, n,
(unsigned long long)sec,
(unsigned long long)wsi->pending_timeout_limit);
#if defined(LWS_WITH_CGI)
if (wsi->cgi)
lwsl_notice("CGI timeout: %s\n", wsi->cgi->summary);
#endif
2017-09-23 12:55:21 +08:00
/*
* Since he failed a timeout, he already had a chance to do
* something and was unable to... that includes situations like
* half closed connections. So process this "failed timeout"
* close as a violent death and don't try to do protocol
* cleanup like flush partials.
*/
wsi->socket_is_permanently_unusable = 1;
if (lwsi_state(wsi) == LRS_WAITING_SSL && wsi->protocol)
wsi->protocol->callback(wsi,
LWS_CALLBACK_CLIENT_CONNECTION_ERROR,
2017-09-23 12:55:21 +08:00
wsi->user_space,
(void *)"Timed out waiting SSL", 21);
2018-03-05 16:49:28 +08:00
__lws_close_free_wsi(wsi, LWS_CLOSE_STATUS_NOSTATUS, "timeout");
return 1;
}
return 0;
}
int lws_rxflow_cache(struct lws *wsi, unsigned char *buf, int n, int len)
{
struct lws_context_per_thread *pt = &wsi->context->pt[(int)wsi->tsi];
uint8_t *buffered;
size_t blen;
int ret = 0, m;
/* his RX is flowcontrolled, don't send remaining now */
blen = lws_buflist_next_segment_len(&wsi->buflist, &buffered);
if (blen) {
if (buf >= buffered && buf + len <= buffered + blen) {
/* rxflow while we were spilling prev rxflow */
lwsl_info("%s: staying in rxflow buf\n", __func__);
return 1;
}
ret = 1;
}
/* a new rxflow, buffer it and warn caller */
2017-09-23 12:55:21 +08:00
m = lws_buflist_append_segment(&wsi->buflist, buf + n, len - n);
2018-04-17 11:43:20 +08:00
if (m < 0)
return -1;
2018-04-17 11:43:20 +08:00
if (m) {
lwsl_debug("%s: added %p to rxflow list\n", __func__, wsi);
lws_dll_lws_add_front(&wsi->dll_buflist, &pt->dll_head_buflist);
2018-04-17 11:43:20 +08:00
}
return ret;
}
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
/* this is used by the platform service code to stop us waiting for network
* activity in poll() when we have something that already needs service
*/
2016-10-07 03:19:17 +08:00
LWS_VISIBLE LWS_EXTERN int
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
lws_service_adjust_timeout(struct lws_context *context, int timeout_ms, int tsi)
{
struct lws_context_per_thread *pt = &context->pt[tsi];
/* Figure out if we really want to wait in poll()
* We only need to wait if really nothing already to do and we have
* to wait for something from network
*/
#if defined(LWS_ROLE_WS) && !defined(LWS_WITHOUT_EXTENSIONS)
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
/* 1) if we know we are draining rx ext, do not wait in poll */
if (pt->ws.rx_draining_ext_list)
return 0;
#endif
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
#if defined(LWS_WITH_TLS)
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
/* 2) if we know we have non-network pending data, do not wait in poll */
if (lws_ssl_anybody_has_buffered_read_tsi(context, tsi)) {
lwsl_info("ssl buffered read\n");
return 0;
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
}
#endif
/* 3) If there is any wsi with rxflow buffered and in a state to process
* it, we should not wait in poll
*/
lws_start_foreach_dll(struct lws_dll_lws *, d, pt->dll_head_buflist.next) {
struct lws *wsi = lws_container_of(d, struct lws, dll_buflist);
if (lwsi_state(wsi) != LRS_DEFERRING_ACTION)
return 0;
} lws_end_foreach_dll(d);
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
return timeout_ms;
}
int
lws_buflist_aware_read(struct lws_context_per_thread *pt, struct lws *wsi,
struct lws_tokens *ebuf)
{
ebuf->len = (int)lws_buflist_next_segment_len(&wsi->buflist,
(uint8_t **)&ebuf->token);
if (!ebuf->len) {
ebuf->token = (char *)pt->serv_buf;
ebuf->len = lws_ssl_capable_read(wsi, pt->serv_buf,
wsi->context->pt_serv_buf_size);
// if (ebuf->len > 0)
// lwsl_hexdump_notice(ebuf->token, ebuf->len);
return 0; /* fresh */
}
return 1; /* buffered */
}
int
lws_buflist_aware_consume(struct lws *wsi, struct lws_tokens *ebuf, int used,
int buffered)
{
int m;
if (used && buffered) {
m = lws_buflist_use_segment(&wsi->buflist, used);
lwsl_info("%s: draining rxflow: used %d, next %d\n",
__func__, used, m);
if (m)
return 0;
lwsl_info("%s: removed %p from dll_buflist\n", __func__, wsi);
lws_dll_lws_remove(&wsi->dll_buflist);
return 0;
}
/* any remainder goes on the buflist */
if (used != ebuf->len &&
lws_buflist_append_segment(&wsi->buflist, (uint8_t *)ebuf->token +
used, ebuf->len - used) < 0)
return 1; /* OOM */
return 0;
}
2018-04-17 11:43:20 +08:00
void
lws_service_do_ripe_rxflow(struct lws_context_per_thread *pt)
{
struct lws_pollfd pfd;
if (!pt->dll_head_buflist.next)
2018-04-17 11:43:20 +08:00
return;
/*
* service all guys with pending rxflow that reached a state they can
* accept the pending data
*/
lws_pt_lock(pt, __func__);
lws_start_foreach_dll_safe(struct lws_dll_lws *, d, d1,
pt->dll_head_buflist.next) {
struct lws *wsi = lws_container_of(d, struct lws, dll_buflist);
2018-04-17 11:43:20 +08:00
pfd.events = LWS_POLLIN;
pfd.revents = LWS_POLLIN;
2018-04-17 11:43:20 +08:00
pfd.fd = -1;
lwsl_debug("%s: rxflow processing: %p 0x%x\n", __func__, wsi,
wsi->wsistate);
if (!lws_is_flowcontrolled(wsi) &&
lwsi_state(wsi) != LRS_DEFERRING_ACTION &&
(wsi->role_ops->handle_POLLIN)(pt, wsi, &pfd) ==
LWS_HPI_RET_PLEASE_CLOSE_ME)
2018-04-17 11:43:20 +08:00
lws_close_free_wsi(wsi, LWS_CLOSE_STATUS_NOSTATUS,
"close_and_handled");
} lws_end_foreach_dll_safe(d, d1);
lws_pt_unlock(pt);
}
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
/*
* guys that need POLLIN service again without waiting for network action
* can force POLLIN here if not flowcontrolled, so they will get service.
*
* Return nonzero if anybody got their POLLIN faked
*/
int
lws_service_flag_pending(struct lws_context *context, int tsi)
{
struct lws_context_per_thread *pt = &context->pt[tsi];
#if defined(LWS_WITH_TLS)
struct lws *wsi, *wsi_next;
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
#endif
int forced = 0;
2018-03-02 14:22:49 +08:00
lws_pt_lock(pt, __func__);
2018-04-17 11:43:20 +08:00
/*
* 1) If there is any wsi with a buflist and in a state to process
2018-04-17 11:43:20 +08:00
* it, we should not wait in poll
*/
lws_start_foreach_dll(struct lws_dll_lws *, d, pt->dll_head_buflist.next) {
struct lws *wsi = lws_container_of(d, struct lws, dll_buflist);
2018-04-17 11:43:20 +08:00
if (lwsi_state(wsi) != LRS_DEFERRING_ACTION) {
forced = 1;
break;
}
} lws_end_foreach_dll(d);
#if defined(LWS_ROLE_WS)
forced |= role_ops_ws.service_flag_pending(context, tsi);
#endif
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
#if defined(LWS_WITH_TLS)
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
/*
* 2) For all guys with buffered SSL read data already saved up, if they
* are not flowcontrolled, fake their POLLIN status so they'll get
* service to use up the buffered incoming data, even though their
* network socket may have nothing
*/
wsi = pt->pending_read_list;
while (wsi) {
wsi_next = wsi->pending_read_list_next;
pt->fds[wsi->position_in_fds_table].revents |=
pt->fds[wsi->position_in_fds_table].events & LWS_POLLIN;
if (pt->fds[wsi->position_in_fds_table].revents & LWS_POLLIN) {
forced = 1;
/*
* he's going to get serviced now, take him off the
* list of guys with buffered SSL. If he still has some
* at the end of the service, he'll get put back on the
* list then.
*/
2018-03-05 16:49:28 +08:00
__lws_ssl_remove_wsi_from_buffered_list(wsi);
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
}
wsi = wsi_next;
}
#endif
2018-03-02 14:22:49 +08:00
lws_pt_unlock(pt);
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
return forced;
}
static int
lws_service_periodic_checks(struct lws_context *context,
struct lws_pollfd *pollfd, int tsi)
{
struct lws_context_per_thread *pt = &context->pt[tsi];
lws_sockfd_type our_fd = 0, tmp_fd;
struct allocated_headers *ah;
struct lws *wsi;
int timed_out = 0;
time_t now;
#if defined(LWS_WITH_TLS)
int n = 0;
#endif
int m;
if (!context->protocol_init_done)
2017-11-08 14:21:03 +08:00
if (lws_protocol_init(context))
return -1;
time(&now);
/*
* handle case that system time was uninitialized when lws started
* at boot, and got initialized a little later
*/
if (context->time_up < 1464083026 && now > 1464083026)
context->time_up = now;
if (context->last_timeout_check_s &&
now - context->last_timeout_check_s > 100) {
/*
* There has been a discontiguity. Any stored time that is
* less than context->time_discontiguity should have context->
* time_fixup added to it.
*
* Some platforms with no RTC will experience this as a normal
* event when ntp sets their clock, but we can have started
* long before that with a 0-based unix time.
*/
context->time_discontiguity = now;
context->time_fixup = now - context->last_timeout_check_s;
lwsl_notice("time discontiguity: at old time %llus, "
"new time %llus: +%llus\n",
(unsigned long long)context->last_timeout_check_s,
(unsigned long long)context->time_discontiguity,
(unsigned long long)context->time_fixup);
context->last_timeout_check_s = now - 1;
}
if (!lws_compare_time_t(context, context->last_timeout_check_s, now))
return 0;
context->last_timeout_check_s = now;
2017-05-07 10:02:03 +08:00
#if defined(LWS_WITH_STATS)
if (!tsi && now - context->last_dump > 10) {
lws_stats_log_dump(context);
context->last_dump = now;
}
2017-05-07 10:02:03 +08:00
#endif
lws_plat_service_periodic(context);
lws_check_deferred_free(context, 0);
#if defined(LWS_WITH_PEER_LIMITS)
lws_peer_cull_peer_wait_list(context);
#endif
/* retire unused deprecated context */
#if !defined(LWS_PLAT_OPTEE) && !defined(LWS_WITH_ESP32)
#if !defined(_WIN32)
if (context->deprecated && !context->count_wsi_allocated) {
lwsl_notice("%s: ending deprecated context\n", __func__);
kill(getpid(), SIGINT);
return 0;
}
2017-01-17 07:01:02 +08:00
#endif
#endif
/* global timeout check once per second */
if (pollfd)
our_fd = pollfd->fd;
/*
* Phase 1: check every wsi on the timeout check list
*/
lws_pt_lock(pt, __func__);
lws_start_foreach_dll_safe(struct lws_dll_lws *, d, d1,
context->pt[tsi].dll_head_timeout.next) {
wsi = lws_container_of(d, struct lws, dll_timeout);
tmp_fd = wsi->desc.sockfd;
if (__lws_service_timeout_check(wsi, now)) {
/* he did time out... */
if (tmp_fd == our_fd)
/* it was the guy we came to service! */
timed_out = 1;
/* he's gone, no need to mark as handled */
}
} lws_end_foreach_dll_safe(d, d1);
/*
* Phase 2: double-check active ah timeouts independent of wsi
* timeout status
*/
ah = pt->ah_list;
while (ah) {
int len;
char buf[256];
const unsigned char *c;
if (!ah->in_use || !ah->wsi || !ah->assigned ||
(ah->wsi->vhost &&
lws_compare_time_t(context, now, ah->assigned) <
ah->wsi->vhost->timeout_secs_ah_idle + 360)) {
ah = ah->next;
continue;
}
/*
* a single ah session somehow got held for
* an unreasonable amount of time.
*
* Dump info on the connection...
*/
wsi = ah->wsi;
buf[0] = '\0';
2017-12-20 10:44:21 +08:00
#if !defined(LWS_PLAT_OPTEE)
lws_get_peer_simple(wsi, buf, sizeof(buf));
2017-12-20 10:44:21 +08:00
#else
buf[0] = '\0';
#endif
lwsl_notice("ah excessive hold: wsi %p\n"
" peer address: %s\n"
" ah pos %u\n",
wsi, buf, ah->pos);
buf[0] = '\0';
m = 0;
do {
c = lws_token_to_string(m);
if (!c)
break;
if (!(*c))
break;
len = lws_hdr_total_length(wsi, m);
if (!len || len > (int)sizeof(buf) - 1) {
m++;
continue;
}
if (lws_hdr_copy(wsi, buf,
sizeof buf, m) > 0) {
buf[sizeof(buf) - 1] = '\0';
lwsl_notice(" %s = %s\n",
(const char *)c, buf);
}
m++;
} while (1);
/* explicitly detach the ah */
lws_header_table_detach(wsi, 0);
/* ... and then drop the connection */
m = 0;
if (wsi->desc.sockfd == our_fd) {
m = timed_out;
/* it was the guy we came to service! */
timed_out = 1;
}
2018-03-05 16:49:28 +08:00
if (!m) /* if he didn't already timeout */
__lws_close_free_wsi(wsi, LWS_CLOSE_STATUS_NOSTATUS,
"excessive ah");
ah = pt->ah_list;
}
lws_pt_unlock(pt);
2017-12-07 10:16:17 +08:00
#if 0
{
char s[300], *p = s;
for (n = 0; n < context->count_threads; n++)
p += sprintf(p, " %7lu (%5d), ",
context->pt[n].count_conns,
context->pt[n].fds_count);
lwsl_notice("load: %s\n", s);
}
#endif
/*
* Phase 3: vhost / protocol timer callbacks
*/
wsi = NULL;
lws_start_foreach_ll(struct lws_vhost *, v, context->vhost_list) {
struct lws_timed_vh_protocol *nx;
if (v->timed_vh_protocol_list) {
lws_start_foreach_ll(struct lws_timed_vh_protocol *,
q, v->timed_vh_protocol_list) {
if (now >= q->time) {
if (!wsi)
wsi = lws_zalloc(sizeof(*wsi), "cbwsi");
wsi->context = context;
wsi->vhost = v;
wsi->protocol = q->protocol;
lwsl_debug("timed cb: vh %s, protocol %s, reason %d\n", v->name, q->protocol->name, q->reason);
q->protocol->callback(wsi, q->reason, NULL, NULL, 0);
nx = q->next;
lws_timed_callback_remove(v, q);
q = nx;
continue; /* we pointed ourselves to the next from the now-deleted guy */
}
} lws_end_foreach_ll(q, next);
}
} lws_end_foreach_ll(v, vhost_next);
if (wsi)
lws_free(wsi);
2018-03-02 15:31:35 +08:00
/*
* Phase 4: check for unconfigured vhosts due to required
* interface missing before
*/
2018-03-02 15:31:35 +08:00
lws_context_lock(context);
lws_start_foreach_llp(struct lws_vhost **, pv,
context->no_listener_vhost_list) {
struct lws_vhost *v = *pv;
lwsl_debug("deferred iface: checking if on vh %s\n", (*pv)->name);
if (lws_context_init_server(NULL, *pv) == 0) {
/* became happy */
lwsl_notice("vh %s: became connected\n", v->name);
*pv = v->no_listener_vhost_list;
v->no_listener_vhost_list = NULL;
break;
}
} lws_end_foreach_llp(pv, no_listener_vhost_list);
lws_context_unlock(context);
/*
* Phase 5: role periodic checks
*/
#if defined(LWS_ROLE_WS)
role_ops_ws.periodic_checks(context, tsi, now);
#endif
#if defined(LWS_ROLE_CGI)
role_ops_cgi.periodic_checks(context, tsi, now);
#endif
/*
* Phase 6: check the remaining cert lifetime daily
*/
#if defined(LWS_WITH_TLS)
n = lws_compare_time_t(context, now, context->last_cert_check_s);
if ((!context->last_cert_check_s || n > (24 * 60 * 60)) &&
!lws_tls_check_all_cert_lifetimes(context))
context->last_cert_check_s = now;
#endif
return timed_out;
}
LWS_VISIBLE int
lws_service_fd_tsi(struct lws_context *context, struct lws_pollfd *pollfd,
int tsi)
{
struct lws_context_per_thread *pt = &context->pt[tsi];
struct lws *wsi;
/* the socket we came to service timed out, nothing to do */
if (lws_service_periodic_checks(context, pollfd, tsi) || !pollfd)
return 0;
/* no, here to service a socket descriptor */
wsi = wsi_from_fd(context, pollfd->fd);
if (!wsi)
/* not lws connection ... leave revents alone and return */
return 0;
/*
* so that caller can tell we handled, past here we need to
* zero down pollfd->revents after handling
*/
/* handle session socket closed */
if ((!(pollfd->revents & pollfd->events & LWS_POLLIN)) &&
(pollfd->revents & LWS_POLLHUP)) {
wsi->socket_is_permanently_unusable = 1;
lwsl_debug("Session Socket %p (fd=%d) dead\n",
(void *)wsi, pollfd->fd);
goto close_and_handled;
}
#ifdef _WIN32
if (pollfd->revents & LWS_POLLOUT)
wsi->sock_send_blocking = FALSE;
#endif
if ((!(pollfd->revents & pollfd->events & LWS_POLLIN)) &&
(pollfd->revents & LWS_POLLHUP)) {
lwsl_debug("pollhup\n");
wsi->socket_is_permanently_unusable = 1;
goto close_and_handled;
}
#if defined(LWS_WITH_TLS)
if (lwsi_state(wsi) == LRS_SHUTDOWN && lws_is_ssl(wsi) && wsi->ssl) {
2018-03-11 11:26:06 +08:00
switch (__lws_tls_shutdown(wsi)) {
case LWS_SSL_CAPABLE_DONE:
case LWS_SSL_CAPABLE_ERROR:
2017-04-06 19:37:14 +02:00
goto close_and_handled;
2017-09-29 11:27:58 +08:00
case LWS_SSL_CAPABLE_MORE_SERVICE_READ:
case LWS_SSL_CAPABLE_MORE_SERVICE_WRITE:
case LWS_SSL_CAPABLE_MORE_SERVICE:
2017-04-06 19:37:14 +02:00
goto handled;
}
}
#endif
wsi->could_have_pending = 0; /* clear back-to-back write detection */
/* okay, what we came here to do... */
/* if we got here, we should have wire protocol ops set on the wsi */
assert(wsi->role_ops);
// lwsl_notice("%s: %s: wsistate 0x%x\n", __func__, wsi->role_ops->name,
// wsi->wsistate);
switch ((wsi->role_ops->handle_POLLIN)(pt, wsi, pollfd)) {
case LWS_HPI_RET_WSI_ALREADY_DIED:
return 1;
case LWS_HPI_RET_HANDLED:
break;
case LWS_HPI_RET_PLEASE_CLOSE_ME:
close_and_handled:
lwsl_debug("%p: Close and handled\n", wsi);
lws_close_free_wsi(wsi, LWS_CLOSE_STATUS_NOSTATUS,
"close_and_handled");
#if defined(_DEBUG)
/*
* confirm close has no problem being called again while
* it waits for libuv service to complete the first async
* close
*/
if (LWS_LIBUV_ENABLED(context))
lws_close_free_wsi(wsi, LWS_CLOSE_STATUS_NOSTATUS,
"close_and_handled uv repeat test");
#endif
/*
* pollfd may point to something else after the close
* due to pollfd swapping scheme on delete on some platforms
* we can't clear revents now because it'd be the wrong guy's
* revents
*/
return 1;
default:
assert(0);
}
#if defined(LWS_WITH_TLS)
handled:
#endif
pollfd->revents = 0;
return 0;
}
LWS_VISIBLE int
lws_service_fd(struct lws_context *context, struct lws_pollfd *pollfd)
{
return lws_service_fd_tsi(context, pollfd, 0);
}
LWS_VISIBLE int
lws_service(struct lws_context *context, int timeout_ms)
{
return lws_plat_service(context, timeout_ms);
}
LWS_VISIBLE int
lws_service_tsi(struct lws_context *context, int timeout_ms, int tsi)
{
return _lws_plat_service_tsi(context, timeout_ms, tsi);
}