2006-06-15 19:48:13 +00:00
|
|
|
/*
|
2012-08-09 10:11:58 +00:00
|
|
|
* include/proto/fd.h
|
|
|
|
* File descriptors states.
|
|
|
|
*
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
* Copyright (C) 2000-2014 Willy Tarreau - w@1wt.eu
|
2012-08-09 10:11:58 +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
|
|
|
|
* exclusively.
|
|
|
|
*
|
|
|
|
* 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
|
|
|
|
*/
|
2006-06-15 19:48:13 +00:00
|
|
|
|
2006-06-26 00:48:02 +00:00
|
|
|
#ifndef _PROTO_FD_H
|
|
|
|
#define _PROTO_FD_H
|
|
|
|
|
2007-04-09 17:29:56 +00:00
|
|
|
#include <stdio.h>
|
2006-06-26 00:48:02 +00:00
|
|
|
#include <sys/time.h>
|
|
|
|
#include <sys/types.h>
|
|
|
|
#include <unistd.h>
|
|
|
|
|
2006-06-29 16:54:54 +00:00
|
|
|
#include <common/config.h>
|
MAJOR: threads/fd: Make fd stuffs thread-safe
Many changes have been made to do so. First, the fd_updt array, where all
pending FDs for polling are stored, is now a thread-local array. Then 3 locks
have been added to protect, respectively, the fdtab array, the fd_cache array
and poll information. In addition, a lock for each entry in the fdtab array has
been added to protect all accesses to a specific FD or its information.
For pollers, according to the poller, the way to manage the concurrency is
different. There is a poller loop on each thread. So the set of monitored FDs
may need to be protected. epoll and kqueue are thread-safe per-se, so there few
things to do to protect these pollers. This is not possible with select and
poll, so there is no sharing between the threads. The poller on each thread is
independant from others.
Finally, per-thread init/deinit functions are used for each pollers and for FD
part for manage thread-local ressources.
Now, you must be carefull when a FD is created during the HAProxy startup. All
update on the FD state must be made in the threads context and never before
their creation. This is mandatory because fd_updt array is thread-local and
initialized only for threads. Because there is no pollers for the main one, this
array remains uninitialized in this context. For this reason, listeners are now
enabled in run_thread_poll_loop function, just like the worker pipe.
2017-05-29 08:40:41 +00:00
|
|
|
|
2006-06-26 00:48:02 +00:00
|
|
|
#include <types/fd.h>
|
|
|
|
|
2012-11-11 14:02:54 +00:00
|
|
|
/* public variables */
|
MAJOR: threads/fd: Make fd stuffs thread-safe
Many changes have been made to do so. First, the fd_updt array, where all
pending FDs for polling are stored, is now a thread-local array. Then 3 locks
have been added to protect, respectively, the fdtab array, the fd_cache array
and poll information. In addition, a lock for each entry in the fdtab array has
been added to protect all accesses to a specific FD or its information.
For pollers, according to the poller, the way to manage the concurrency is
different. There is a poller loop on each thread. So the set of monitored FDs
may need to be protected. epoll and kqueue are thread-safe per-se, so there few
things to do to protect these pollers. This is not possible with select and
poll, so there is no sharing between the threads. The poller on each thread is
independant from others.
Finally, per-thread init/deinit functions are used for each pollers and for FD
part for manage thread-local ressources.
Now, you must be carefull when a FD is created during the HAProxy startup. All
update on the FD state must be made in the threads context and never before
their creation. This is mandatory because fd_updt array is thread-local and
initialized only for threads. Because there is no pollers for the main one, this
array remains uninitialized in this context. For this reason, listeners are now
enabled in run_thread_poll_loop function, just like the worker pipe.
2017-05-29 08:40:41 +00:00
|
|
|
|
2018-01-24 17:17:56 +00:00
|
|
|
extern volatile struct fdlist fd_cache;
|
|
|
|
extern volatile struct fdlist fd_cache_local[MAX_THREADS];
|
|
|
|
|
2018-04-25 14:58:25 +00:00
|
|
|
extern volatile struct fdlist update_list;
|
|
|
|
|
2018-04-26 12:23:07 +00:00
|
|
|
extern unsigned long *polled_mask;
|
|
|
|
|
2018-01-15 10:57:03 +00:00
|
|
|
extern unsigned long fd_cache_mask; // Mask of threads with events in the cache
|
MAJOR: threads/fd: Make fd stuffs thread-safe
Many changes have been made to do so. First, the fd_updt array, where all
pending FDs for polling are stored, is now a thread-local array. Then 3 locks
have been added to protect, respectively, the fdtab array, the fd_cache array
and poll information. In addition, a lock for each entry in the fdtab array has
been added to protect all accesses to a specific FD or its information.
For pollers, according to the poller, the way to manage the concurrency is
different. There is a poller loop on each thread. So the set of monitored FDs
may need to be protected. epoll and kqueue are thread-safe per-se, so there few
things to do to protect these pollers. This is not possible with select and
poll, so there is no sharing between the threads. The poller on each thread is
independant from others.
Finally, per-thread init/deinit functions are used for each pollers and for FD
part for manage thread-local ressources.
Now, you must be carefull when a FD is created during the HAProxy startup. All
update on the FD state must be made in the threads context and never before
their creation. This is mandatory because fd_updt array is thread-local and
initialized only for threads. Because there is no pollers for the main one, this
array remains uninitialized in this context. For this reason, listeners are now
enabled in run_thread_poll_loop function, just like the worker pipe.
2017-05-29 08:40:41 +00:00
|
|
|
|
|
|
|
extern THREAD_LOCAL int *fd_updt; // FD updates list
|
|
|
|
extern THREAD_LOCAL int fd_nbupdt; // number of updates in the list
|
|
|
|
|
2018-07-26 15:55:11 +00:00
|
|
|
extern int poller_wr_pipe[MAX_THREADS];
|
|
|
|
|
2017-11-26 10:07:34 +00:00
|
|
|
__decl_hathreads(extern HA_RWLOCK_T __attribute__((aligned(64))) fdcache_lock); /* global lock to protect fd_cache array */
|
2012-11-11 14:02:54 +00:00
|
|
|
|
2018-01-26 20:48:23 +00:00
|
|
|
/* Deletes an FD from the fdsets.
|
2006-06-26 00:48:02 +00:00
|
|
|
* The file descriptor is also closed.
|
|
|
|
*/
|
|
|
|
void fd_delete(int fd);
|
2006-06-15 19:48:13 +00:00
|
|
|
|
2018-01-26 20:48:23 +00:00
|
|
|
/* Deletes an FD from the fdsets.
|
2017-04-05 23:05:05 +00:00
|
|
|
* The file descriptor is kept open.
|
|
|
|
*/
|
|
|
|
void fd_remove(int fd);
|
|
|
|
|
2007-04-08 14:39:58 +00:00
|
|
|
/* disable the specified poller */
|
|
|
|
void disable_poller(const char *poller_name);
|
2006-06-15 19:48:13 +00:00
|
|
|
|
2018-07-26 15:55:11 +00:00
|
|
|
void poller_pipe_io_handler(int fd);
|
|
|
|
|
2006-10-15 12:52:29 +00:00
|
|
|
/*
|
2007-04-08 14:39:58 +00:00
|
|
|
* Initialize the pollers till the best one is found.
|
|
|
|
* If none works, returns 0, otherwise 1.
|
2007-04-15 22:25:25 +00:00
|
|
|
* The pollers register themselves just before main() is called.
|
2006-10-15 12:52:29 +00:00
|
|
|
*/
|
2007-04-08 14:39:58 +00:00
|
|
|
int init_pollers();
|
2006-10-15 12:52:29 +00:00
|
|
|
|
[MEDIUM] Fix memory freeing at exit
New functions implemented:
- deinit_pollers: called at the end of deinit())
- prune_acl: called via list_for_each_entry_safe
Add missing pool_destroy2 calls:
- p->hdr_idx_pool
- pool2_tree64
Implement all task stopping:
- health-check: needs new "struct task" in the struct server
- queue processing: queue_mgt
- appsess_refresh: appsession_refresh
before (idle system):
==6079== LEAK SUMMARY:
==6079== definitely lost: 1,112 bytes in 75 blocks.
==6079== indirectly lost: 53,356 bytes in 2,090 blocks.
==6079== possibly lost: 52 bytes in 1 blocks.
==6079== still reachable: 150,996 bytes in 504 blocks.
==6079== suppressed: 0 bytes in 0 blocks.
after (idle system):
==6945== LEAK SUMMARY:
==6945== definitely lost: 7,644 bytes in 137 blocks.
==6945== indirectly lost: 9,913 bytes in 587 blocks.
==6945== possibly lost: 0 bytes in 0 blocks.
==6945== still reachable: 0 bytes in 0 blocks.
==6945== suppressed: 0 bytes in 0 blocks.
before (running system for ~2m):
==9343== LEAK SUMMARY:
==9343== definitely lost: 1,112 bytes in 75 blocks.
==9343== indirectly lost: 54,199 bytes in 2,122 blocks.
==9343== possibly lost: 52 bytes in 1 blocks.
==9343== still reachable: 151,128 bytes in 509 blocks.
==9343== suppressed: 0 bytes in 0 blocks.
after (running system for ~2m):
==11616== LEAK SUMMARY:
==11616== definitely lost: 7,644 bytes in 137 blocks.
==11616== indirectly lost: 9,981 bytes in 591 blocks.
==11616== possibly lost: 0 bytes in 0 blocks.
==11616== still reachable: 4 bytes in 1 blocks.
==11616== suppressed: 0 bytes in 0 blocks.
Still not perfect but significant improvement.
2008-05-29 21:53:44 +00:00
|
|
|
/*
|
|
|
|
* Deinitialize the pollers.
|
|
|
|
*/
|
|
|
|
void deinit_pollers();
|
|
|
|
|
2007-04-09 17:29:56 +00:00
|
|
|
/*
|
|
|
|
* Some pollers may lose their connection after a fork(). It may be necessary
|
|
|
|
* to create initialize part of them again. Returns 0 in case of failure,
|
|
|
|
* otherwise 1. The fork() function may be NULL if unused. In case of error,
|
|
|
|
* the the current poller is destroyed and the caller is responsible for trying
|
|
|
|
* another one by calling init_pollers() again.
|
|
|
|
*/
|
|
|
|
int fork_poller();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lists the known pollers on <out>.
|
|
|
|
* Should be performed only before initialization.
|
|
|
|
*/
|
|
|
|
int list_pollers(FILE *out);
|
|
|
|
|
2007-04-08 14:39:58 +00:00
|
|
|
/*
|
|
|
|
* Runs the polling loop
|
|
|
|
*/
|
|
|
|
void run_poller();
|
|
|
|
|
2014-01-25 18:24:15 +00:00
|
|
|
/* Scan and process the cached events. This should be called right after
|
2012-11-11 15:43:45 +00:00
|
|
|
* the poller.
|
|
|
|
*/
|
2014-01-25 18:24:15 +00:00
|
|
|
void fd_process_cached_events();
|
2012-11-11 15:43:45 +00:00
|
|
|
|
2018-04-25 13:10:30 +00:00
|
|
|
void fd_add_to_fd_list(volatile struct fdlist *list, int fd, int off);
|
|
|
|
void fd_rm_from_fd_list(volatile struct fdlist *list, int fd, int off);
|
2018-02-05 16:14:55 +00:00
|
|
|
|
MAJOR: polling: centralize calls to I/O callbacks
In order for HTTP/2 not to eat too much memory, we'll have to support
on-the-fly buffer allocation, since most streams will have an empty
request buffer at some point. Supporting allocation on the fly means
being able to sleep inside I/O callbacks if a buffer is not available.
Till now, the I/O callbacks were called from two locations :
- when processing the cached events
- when processing the polled events from the poller
This change cleans up the design a bit further than what was started in
1.5. It now ensures that we never call any iocb from the poller itself
and that instead, events learned by the poller are put into the cache.
The benefit is important in terms of stability : we don't have to care
anymore about the risk that new events are added into the poller while
processing its events, and we're certain that updates are processed at
a single location.
To achieve this, we now modify all the fd_* functions so that instead of
creating updates, they add/remove the fd to/from the cache depending on
its state, and only create an update when the polling status reaches a
state where it will have to change. Since the pollers make use of these
functions to notify readiness (using fd_may_recv/fd_may_send), the cache
is always up to date with the poller.
Creating updates only when the polling status needs to change saves a
significant amount of work for the pollers : a benchmark showed that on
a typical TCP proxy test, the amount of updates per connection dropped
from 11 to 1 on average. This also means that the update list is smaller
and has more chances of not thrashing too many CPU cache lines. The first
observed benefit is a net 2% performance gain on the connection rate.
A second benefit is that when a connection is accepted, it's only when
we're processing the cache, and the recv event is automatically added
into the cache *after* the current one, resulting in this event to be
processed immediately during the same loop. Previously we used to have
a second run over the updates to detect if new events were added to
catch them before waking up tasks.
The next gain will be offered by the next steps on this subject consisting
in implementing an I/O queue containing all cached events ordered by priority
just like the run queue, and to be able to leave some events pending there
as long as needed. That will allow us *not* to perform some FD processing
if it's not the proper time for this (typically keep waiting for a buffer
to be allocated if none is available for an recv()). And by only processing
a small bunch of them, we'll allow priorities to take place even at the I/O
level.
As a result of this change, functions fd_alloc_or_release_cache_entry()
and fd_process_polled_events() have disappeared, and the code dedicated
to checking for new fd events after the callback during the poll() loop
was removed as well. Despite the patch looking large, it's mostly a
change of what function is falled upon fd_*() and almost nothing was
added.
2014-11-19 18:43:05 +00:00
|
|
|
/* Mark fd <fd> as updated for polling and allocate an entry in the update list
|
|
|
|
* for this if it was not already there. This can be done at any time.
|
2014-01-25 08:58:06 +00:00
|
|
|
*/
|
MAJOR: polling: centralize calls to I/O callbacks
In order for HTTP/2 not to eat too much memory, we'll have to support
on-the-fly buffer allocation, since most streams will have an empty
request buffer at some point. Supporting allocation on the fly means
being able to sleep inside I/O callbacks if a buffer is not available.
Till now, the I/O callbacks were called from two locations :
- when processing the cached events
- when processing the polled events from the poller
This change cleans up the design a bit further than what was started in
1.5. It now ensures that we never call any iocb from the poller itself
and that instead, events learned by the poller are put into the cache.
The benefit is important in terms of stability : we don't have to care
anymore about the risk that new events are added into the poller while
processing its events, and we're certain that updates are processed at
a single location.
To achieve this, we now modify all the fd_* functions so that instead of
creating updates, they add/remove the fd to/from the cache depending on
its state, and only create an update when the polling status reaches a
state where it will have to change. Since the pollers make use of these
functions to notify readiness (using fd_may_recv/fd_may_send), the cache
is always up to date with the poller.
Creating updates only when the polling status needs to change saves a
significant amount of work for the pollers : a benchmark showed that on
a typical TCP proxy test, the amount of updates per connection dropped
from 11 to 1 on average. This also means that the update list is smaller
and has more chances of not thrashing too many CPU cache lines. The first
observed benefit is a net 2% performance gain on the connection rate.
A second benefit is that when a connection is accepted, it's only when
we're processing the cache, and the recv event is automatically added
into the cache *after* the current one, resulting in this event to be
processed immediately during the same loop. Previously we used to have
a second run over the updates to detect if new events were added to
catch them before waking up tasks.
The next gain will be offered by the next steps on this subject consisting
in implementing an I/O queue containing all cached events ordered by priority
just like the run queue, and to be able to leave some events pending there
as long as needed. That will allow us *not* to perform some FD processing
if it's not the proper time for this (typically keep waiting for a buffer
to be allocated if none is available for an recv()). And by only processing
a small bunch of them, we'll allow priorities to take place even at the I/O
level.
As a result of this change, functions fd_alloc_or_release_cache_entry()
and fd_process_polled_events() have disappeared, and the code dedicated
to checking for new fd events after the callback during the poll() loop
was removed as well. Despite the patch looking large, it's mostly a
change of what function is falled upon fd_*() and almost nothing was
added.
2014-11-19 18:43:05 +00:00
|
|
|
static inline void updt_fd_polling(const int fd)
|
2012-11-11 14:02:54 +00:00
|
|
|
{
|
2018-08-17 11:37:59 +00:00
|
|
|
if ((fdtab[fd].thread_mask & all_threads_mask) == tid_bit) {
|
2018-04-25 14:58:25 +00:00
|
|
|
unsigned int oldupdt;
|
|
|
|
|
|
|
|
/* note: we don't have a test-and-set yet in hathreads */
|
|
|
|
|
|
|
|
if (HA_ATOMIC_BTS(&fdtab[fd].update_mask, tid))
|
|
|
|
return;
|
|
|
|
|
|
|
|
oldupdt = HA_ATOMIC_ADD(&fd_nbupdt, 1) - 1;
|
|
|
|
fd_updt[oldupdt] = fd;
|
|
|
|
} else {
|
|
|
|
unsigned long update_mask = fdtab[fd].update_mask;
|
|
|
|
do {
|
|
|
|
if (update_mask == fdtab[fd].thread_mask)
|
|
|
|
return;
|
|
|
|
} while (!HA_ATOMIC_CAS(&fdtab[fd].update_mask, &update_mask,
|
|
|
|
fdtab[fd].thread_mask));
|
|
|
|
fd_add_to_fd_list(&update_list, fd, offsetof(struct fdtab, update));
|
|
|
|
}
|
2018-01-17 21:57:54 +00:00
|
|
|
|
2018-04-25 14:58:25 +00:00
|
|
|
}
|
2018-01-17 21:57:54 +00:00
|
|
|
|
2018-04-25 14:58:25 +00:00
|
|
|
/* Called from the poller to acknoledge we read an entry from the global
|
|
|
|
* update list, to remove our bit from the update_mask, and remove it from
|
|
|
|
* the list if we were the last one.
|
|
|
|
*/
|
|
|
|
static inline void done_update_polling(int fd)
|
|
|
|
{
|
|
|
|
unsigned long update_mask;
|
|
|
|
|
|
|
|
update_mask = HA_ATOMIC_AND(&fdtab[fd].update_mask, ~tid_bit);
|
|
|
|
while ((update_mask & all_threads_mask)== 0) {
|
|
|
|
/* If we were the last one that had to update that entry, remove it from the list */
|
|
|
|
fd_rm_from_fd_list(&update_list, fd, offsetof(struct fdtab, update));
|
|
|
|
if (update_list.first == fd)
|
|
|
|
abort();
|
|
|
|
update_mask = (volatile unsigned long)fdtab[fd].update_mask;
|
|
|
|
if ((update_mask & all_threads_mask) != 0) {
|
|
|
|
/* Maybe it's been re-updated in the meanwhile, and we
|
|
|
|
* wrongly removed it from the list, if so, re-add it
|
|
|
|
*/
|
|
|
|
fd_add_to_fd_list(&update_list, fd, offsetof(struct fdtab, update));
|
|
|
|
update_mask = (volatile unsigned long)(fdtab[fd].update_mask);
|
|
|
|
/* And then check again, just in case after all it
|
|
|
|
* should be removed, even if it's very unlikely, given
|
|
|
|
* the current thread wouldn't have been able to take
|
|
|
|
* care of it yet */
|
|
|
|
} else
|
|
|
|
break;
|
|
|
|
|
|
|
|
}
|
2012-11-11 14:02:54 +00:00
|
|
|
}
|
|
|
|
|
2014-01-25 18:20:35 +00:00
|
|
|
/* Allocates a cache entry for a file descriptor if it does not yet have one.
|
|
|
|
* This can be done at any time.
|
|
|
|
*/
|
|
|
|
static inline void fd_alloc_cache_entry(const int fd)
|
2012-11-11 14:02:54 +00:00
|
|
|
{
|
2018-03-20 18:06:52 +00:00
|
|
|
HA_ATOMIC_OR(&fd_cache_mask, fdtab[fd].thread_mask);
|
2018-01-24 17:17:56 +00:00
|
|
|
if (!(fdtab[fd].thread_mask & (fdtab[fd].thread_mask - 1)))
|
2018-04-25 13:10:30 +00:00
|
|
|
fd_add_to_fd_list(&fd_cache_local[my_ffsl(fdtab[fd].thread_mask) - 1], fd, offsetof(struct fdtab, cache));
|
2018-01-24 17:17:56 +00:00
|
|
|
else
|
2018-04-25 13:10:30 +00:00
|
|
|
fd_add_to_fd_list(&fd_cache, fd, offsetof(struct fdtab, cache));
|
2012-11-11 14:02:54 +00:00
|
|
|
}
|
|
|
|
|
2014-01-25 18:20:35 +00:00
|
|
|
/* Removes entry used by fd <fd> from the FD cache and replaces it with the
|
2018-01-24 17:17:56 +00:00
|
|
|
* last one.
|
2012-11-11 14:02:54 +00:00
|
|
|
* If the fd has no entry assigned, return immediately.
|
|
|
|
*/
|
2018-02-05 16:14:55 +00:00
|
|
|
static inline void fd_release_cache_entry(const int fd)
|
2012-11-11 14:02:54 +00:00
|
|
|
{
|
2018-01-24 17:17:56 +00:00
|
|
|
if (!(fdtab[fd].thread_mask & (fdtab[fd].thread_mask - 1)))
|
2018-04-25 13:10:30 +00:00
|
|
|
fd_rm_from_fd_list(&fd_cache_local[my_ffsl(fdtab[fd].thread_mask) - 1], fd, offsetof(struct fdtab, cache));
|
2018-01-24 17:17:56 +00:00
|
|
|
else
|
2018-04-25 13:10:30 +00:00
|
|
|
fd_rm_from_fd_list(&fd_cache, fd, offsetof(struct fdtab, cache));
|
2012-11-11 14:02:54 +00:00
|
|
|
}
|
2006-10-15 12:52:29 +00:00
|
|
|
|
MAJOR: polling: centralize calls to I/O callbacks
In order for HTTP/2 not to eat too much memory, we'll have to support
on-the-fly buffer allocation, since most streams will have an empty
request buffer at some point. Supporting allocation on the fly means
being able to sleep inside I/O callbacks if a buffer is not available.
Till now, the I/O callbacks were called from two locations :
- when processing the cached events
- when processing the polled events from the poller
This change cleans up the design a bit further than what was started in
1.5. It now ensures that we never call any iocb from the poller itself
and that instead, events learned by the poller are put into the cache.
The benefit is important in terms of stability : we don't have to care
anymore about the risk that new events are added into the poller while
processing its events, and we're certain that updates are processed at
a single location.
To achieve this, we now modify all the fd_* functions so that instead of
creating updates, they add/remove the fd to/from the cache depending on
its state, and only create an update when the polling status reaches a
state where it will have to change. Since the pollers make use of these
functions to notify readiness (using fd_may_recv/fd_may_send), the cache
is always up to date with the poller.
Creating updates only when the polling status needs to change saves a
significant amount of work for the pollers : a benchmark showed that on
a typical TCP proxy test, the amount of updates per connection dropped
from 11 to 1 on average. This also means that the update list is smaller
and has more chances of not thrashing too many CPU cache lines. The first
observed benefit is a net 2% performance gain on the connection rate.
A second benefit is that when a connection is accepted, it's only when
we're processing the cache, and the recv event is automatically added
into the cache *after* the current one, resulting in this event to be
processed immediately during the same loop. Previously we used to have
a second run over the updates to detect if new events were added to
catch them before waking up tasks.
The next gain will be offered by the next steps on this subject consisting
in implementing an I/O queue containing all cached events ordered by priority
just like the run queue, and to be able to leave some events pending there
as long as needed. That will allow us *not* to perform some FD processing
if it's not the proper time for this (typically keep waiting for a buffer
to be allocated if none is available for an recv()). And by only processing
a small bunch of them, we'll allow priorities to take place even at the I/O
level.
As a result of this change, functions fd_alloc_or_release_cache_entry()
and fd_process_polled_events() have disappeared, and the code dedicated
to checking for new fd events after the callback during the poll() loop
was removed as well. Despite the patch looking large, it's mostly a
change of what function is falled upon fd_*() and almost nothing was
added.
2014-11-19 18:43:05 +00:00
|
|
|
/* This function automatically enables/disables caching for an entry depending
|
2018-01-17 20:25:57 +00:00
|
|
|
* on its state. It is only called on state changes.
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
*/
|
MAJOR: polling: centralize calls to I/O callbacks
In order for HTTP/2 not to eat too much memory, we'll have to support
on-the-fly buffer allocation, since most streams will have an empty
request buffer at some point. Supporting allocation on the fly means
being able to sleep inside I/O callbacks if a buffer is not available.
Till now, the I/O callbacks were called from two locations :
- when processing the cached events
- when processing the polled events from the poller
This change cleans up the design a bit further than what was started in
1.5. It now ensures that we never call any iocb from the poller itself
and that instead, events learned by the poller are put into the cache.
The benefit is important in terms of stability : we don't have to care
anymore about the risk that new events are added into the poller while
processing its events, and we're certain that updates are processed at
a single location.
To achieve this, we now modify all the fd_* functions so that instead of
creating updates, they add/remove the fd to/from the cache depending on
its state, and only create an update when the polling status reaches a
state where it will have to change. Since the pollers make use of these
functions to notify readiness (using fd_may_recv/fd_may_send), the cache
is always up to date with the poller.
Creating updates only when the polling status needs to change saves a
significant amount of work for the pollers : a benchmark showed that on
a typical TCP proxy test, the amount of updates per connection dropped
from 11 to 1 on average. This also means that the update list is smaller
and has more chances of not thrashing too many CPU cache lines. The first
observed benefit is a net 2% performance gain on the connection rate.
A second benefit is that when a connection is accepted, it's only when
we're processing the cache, and the recv event is automatically added
into the cache *after* the current one, resulting in this event to be
processed immediately during the same loop. Previously we used to have
a second run over the updates to detect if new events were added to
catch them before waking up tasks.
The next gain will be offered by the next steps on this subject consisting
in implementing an I/O queue containing all cached events ordered by priority
just like the run queue, and to be able to leave some events pending there
as long as needed. That will allow us *not* to perform some FD processing
if it's not the proper time for this (typically keep waiting for a buffer
to be allocated if none is available for an recv()). And by only processing
a small bunch of them, we'll allow priorities to take place even at the I/O
level.
As a result of this change, functions fd_alloc_or_release_cache_entry()
and fd_process_polled_events() have disappeared, and the code dedicated
to checking for new fd events after the callback during the poll() loop
was removed as well. Despite the patch looking large, it's mostly a
change of what function is falled upon fd_*() and almost nothing was
added.
2014-11-19 18:43:05 +00:00
|
|
|
static inline void fd_update_cache(int fd)
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
{
|
MAJOR: polling: centralize calls to I/O callbacks
In order for HTTP/2 not to eat too much memory, we'll have to support
on-the-fly buffer allocation, since most streams will have an empty
request buffer at some point. Supporting allocation on the fly means
being able to sleep inside I/O callbacks if a buffer is not available.
Till now, the I/O callbacks were called from two locations :
- when processing the cached events
- when processing the polled events from the poller
This change cleans up the design a bit further than what was started in
1.5. It now ensures that we never call any iocb from the poller itself
and that instead, events learned by the poller are put into the cache.
The benefit is important in terms of stability : we don't have to care
anymore about the risk that new events are added into the poller while
processing its events, and we're certain that updates are processed at
a single location.
To achieve this, we now modify all the fd_* functions so that instead of
creating updates, they add/remove the fd to/from the cache depending on
its state, and only create an update when the polling status reaches a
state where it will have to change. Since the pollers make use of these
functions to notify readiness (using fd_may_recv/fd_may_send), the cache
is always up to date with the poller.
Creating updates only when the polling status needs to change saves a
significant amount of work for the pollers : a benchmark showed that on
a typical TCP proxy test, the amount of updates per connection dropped
from 11 to 1 on average. This also means that the update list is smaller
and has more chances of not thrashing too many CPU cache lines. The first
observed benefit is a net 2% performance gain on the connection rate.
A second benefit is that when a connection is accepted, it's only when
we're processing the cache, and the recv event is automatically added
into the cache *after* the current one, resulting in this event to be
processed immediately during the same loop. Previously we used to have
a second run over the updates to detect if new events were added to
catch them before waking up tasks.
The next gain will be offered by the next steps on this subject consisting
in implementing an I/O queue containing all cached events ordered by priority
just like the run queue, and to be able to leave some events pending there
as long as needed. That will allow us *not* to perform some FD processing
if it's not the proper time for this (typically keep waiting for a buffer
to be allocated if none is available for an recv()). And by only processing
a small bunch of them, we'll allow priorities to take place even at the I/O
level.
As a result of this change, functions fd_alloc_or_release_cache_entry()
and fd_process_polled_events() have disappeared, and the code dedicated
to checking for new fd events after the callback during the poll() loop
was removed as well. Despite the patch looking large, it's mostly a
change of what function is falled upon fd_*() and almost nothing was
added.
2014-11-19 18:43:05 +00:00
|
|
|
/* only READY and ACTIVE states (the two with both flags set) require a cache entry */
|
|
|
|
if (((fdtab[fd].state & (FD_EV_READY_R | FD_EV_ACTIVE_R)) == (FD_EV_READY_R | FD_EV_ACTIVE_R)) ||
|
|
|
|
((fdtab[fd].state & (FD_EV_READY_W | FD_EV_ACTIVE_W)) == (FD_EV_READY_W | FD_EV_ACTIVE_W))) {
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
fd_alloc_cache_entry(fd);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
fd_release_cache_entry(fd);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-11-11 15:05:19 +00:00
|
|
|
/*
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
* returns the FD's recv state (FD_EV_*)
|
2012-11-11 15:05:19 +00:00
|
|
|
*/
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
static inline int fd_recv_state(const int fd)
|
2012-11-11 15:05:19 +00:00
|
|
|
{
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
return ((unsigned)fdtab[fd].state >> (4 * DIR_RD)) & FD_EV_STATUS;
|
2012-11-11 15:05:19 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/*
|
|
|
|
* returns true if the FD is active for recv
|
2012-11-11 15:05:19 +00:00
|
|
|
*/
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
static inline int fd_recv_active(const int fd)
|
2012-11-11 15:05:19 +00:00
|
|
|
{
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
return (unsigned)fdtab[fd].state & FD_EV_ACTIVE_R;
|
2012-11-11 15:05:19 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/*
|
|
|
|
* returns true if the FD is ready for recv
|
2012-11-11 15:05:19 +00:00
|
|
|
*/
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
static inline int fd_recv_ready(const int fd)
|
2012-11-11 15:05:19 +00:00
|
|
|
{
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
return (unsigned)fdtab[fd].state & FD_EV_READY_R;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* returns true if the FD is polled for recv
|
|
|
|
*/
|
|
|
|
static inline int fd_recv_polled(const int fd)
|
|
|
|
{
|
|
|
|
return (unsigned)fdtab[fd].state & FD_EV_POLLED_R;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* returns the FD's send state (FD_EV_*)
|
|
|
|
*/
|
|
|
|
static inline int fd_send_state(const int fd)
|
|
|
|
{
|
|
|
|
return ((unsigned)fdtab[fd].state >> (4 * DIR_WR)) & FD_EV_STATUS;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* returns true if the FD is active for send
|
|
|
|
*/
|
|
|
|
static inline int fd_send_active(const int fd)
|
|
|
|
{
|
|
|
|
return (unsigned)fdtab[fd].state & FD_EV_ACTIVE_W;
|
2012-11-11 15:05:19 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/*
|
|
|
|
* returns true if the FD is ready for send
|
2012-11-11 15:05:19 +00:00
|
|
|
*/
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
static inline int fd_send_ready(const int fd)
|
2012-11-11 15:05:19 +00:00
|
|
|
{
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
return (unsigned)fdtab[fd].state & FD_EV_READY_W;
|
|
|
|
}
|
2012-11-11 15:05:19 +00:00
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/*
|
|
|
|
* returns true if the FD is polled for send
|
|
|
|
*/
|
|
|
|
static inline int fd_send_polled(const int fd)
|
|
|
|
{
|
|
|
|
return (unsigned)fdtab[fd].state & FD_EV_POLLED_W;
|
|
|
|
}
|
|
|
|
|
2017-08-30 07:59:38 +00:00
|
|
|
/*
|
|
|
|
* returns true if the FD is active for recv or send
|
|
|
|
*/
|
|
|
|
static inline int fd_active(const int fd)
|
|
|
|
{
|
|
|
|
return (unsigned)fdtab[fd].state & FD_EV_ACTIVE_RW;
|
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Disable processing recv events on fd <fd> */
|
|
|
|
static inline void fd_stop_recv(int fd)
|
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
unsigned char old, new;
|
|
|
|
|
|
|
|
old = fdtab[fd].state;
|
|
|
|
do {
|
|
|
|
if (!(old & FD_EV_ACTIVE_R))
|
|
|
|
return;
|
|
|
|
new = old & ~FD_EV_ACTIVE_R;
|
|
|
|
new &= ~FD_EV_POLLED_R;
|
|
|
|
} while (unlikely(!HA_ATOMIC_CAS(&fdtab[fd].state, &old, new)));
|
|
|
|
|
|
|
|
if ((old ^ new) & FD_EV_POLLED_R)
|
|
|
|
updt_fd_polling(fd);
|
|
|
|
|
2018-01-17 21:57:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-11-11 15:05:19 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Disable processing send events on fd <fd> */
|
|
|
|
static inline void fd_stop_send(int fd)
|
2012-11-11 15:05:19 +00:00
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
unsigned char old, new;
|
|
|
|
|
|
|
|
old = fdtab[fd].state;
|
|
|
|
do {
|
|
|
|
if (!(old & FD_EV_ACTIVE_W))
|
|
|
|
return;
|
|
|
|
new = old & ~FD_EV_ACTIVE_W;
|
|
|
|
new &= ~FD_EV_POLLED_W;
|
|
|
|
} while (unlikely(!HA_ATOMIC_CAS(&fdtab[fd].state, &old, new)));
|
|
|
|
|
|
|
|
if ((old ^ new) & FD_EV_POLLED_W)
|
|
|
|
updt_fd_polling(fd);
|
|
|
|
|
2018-01-17 21:57:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-11-11 15:05:19 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Disable processing of events on fd <fd> for both directions. */
|
|
|
|
static inline void fd_stop_both(int fd)
|
2012-08-09 10:11:58 +00:00
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
unsigned char old, new;
|
|
|
|
|
|
|
|
old = fdtab[fd].state;
|
|
|
|
do {
|
|
|
|
if (!(old & FD_EV_ACTIVE_RW))
|
|
|
|
return;
|
|
|
|
new = old & ~FD_EV_ACTIVE_RW;
|
|
|
|
new &= ~FD_EV_POLLED_RW;
|
|
|
|
} while (unlikely(!HA_ATOMIC_CAS(&fdtab[fd].state, &old, new)));
|
|
|
|
|
|
|
|
if ((old ^ new) & FD_EV_POLLED_RW)
|
|
|
|
updt_fd_polling(fd);
|
|
|
|
|
2018-01-17 21:57:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-08-09 10:11:58 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Report that FD <fd> cannot receive anymore without polling (EAGAIN detected). */
|
|
|
|
static inline void fd_cant_recv(const int fd)
|
2012-08-09 10:11:58 +00:00
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
unsigned char old, new;
|
|
|
|
|
|
|
|
old = fdtab[fd].state;
|
|
|
|
do {
|
|
|
|
if (!(old & FD_EV_READY_R))
|
|
|
|
return;
|
|
|
|
new = old & ~FD_EV_READY_R;
|
|
|
|
if (new & FD_EV_ACTIVE_R)
|
|
|
|
new |= FD_EV_POLLED_R;
|
|
|
|
} while (unlikely(!HA_ATOMIC_CAS(&fdtab[fd].state, &old, new)));
|
|
|
|
|
|
|
|
if ((old ^ new) & FD_EV_POLLED_R)
|
|
|
|
updt_fd_polling(fd);
|
|
|
|
|
2018-01-17 21:57:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-08-09 10:11:58 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Report that FD <fd> can receive anymore without polling. */
|
|
|
|
static inline void fd_may_recv(const int fd)
|
2012-08-09 10:14:03 +00:00
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
/* marking ready never changes polled status */
|
|
|
|
HA_ATOMIC_OR(&fdtab[fd].state, FD_EV_READY_R);
|
|
|
|
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-08-09 10:14:03 +00:00
|
|
|
}
|
|
|
|
|
2014-01-23 23:54:27 +00:00
|
|
|
/* Disable readiness when polled. This is useful to interrupt reading when it
|
|
|
|
* is suspected that the end of data might have been reached (eg: short read).
|
|
|
|
* This can only be done using level-triggered pollers, so if any edge-triggered
|
|
|
|
* is ever implemented, a test will have to be added here.
|
|
|
|
*/
|
|
|
|
static inline void fd_done_recv(const int fd)
|
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
unsigned char old, new;
|
|
|
|
|
|
|
|
old = fdtab[fd].state;
|
|
|
|
do {
|
|
|
|
if ((old & (FD_EV_POLLED_R|FD_EV_READY_R)) != (FD_EV_POLLED_R|FD_EV_READY_R))
|
|
|
|
return;
|
|
|
|
new = old & ~FD_EV_READY_R;
|
|
|
|
if (new & FD_EV_ACTIVE_R)
|
|
|
|
new |= FD_EV_POLLED_R;
|
|
|
|
} while (unlikely(!HA_ATOMIC_CAS(&fdtab[fd].state, &old, new)));
|
|
|
|
|
|
|
|
if ((old ^ new) & FD_EV_POLLED_R)
|
|
|
|
updt_fd_polling(fd);
|
|
|
|
|
2018-01-17 21:57:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2014-01-23 23:54:27 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Report that FD <fd> cannot send anymore without polling (EAGAIN detected). */
|
|
|
|
static inline void fd_cant_send(const int fd)
|
2012-08-09 10:11:58 +00:00
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
unsigned char old, new;
|
|
|
|
|
|
|
|
old = fdtab[fd].state;
|
|
|
|
do {
|
|
|
|
if (!(old & FD_EV_READY_W))
|
|
|
|
return;
|
|
|
|
new = old & ~FD_EV_READY_W;
|
|
|
|
if (new & FD_EV_ACTIVE_W)
|
|
|
|
new |= FD_EV_POLLED_W;
|
|
|
|
} while (unlikely(!HA_ATOMIC_CAS(&fdtab[fd].state, &old, new)));
|
|
|
|
|
|
|
|
if ((old ^ new) & FD_EV_POLLED_W)
|
|
|
|
updt_fd_polling(fd);
|
|
|
|
|
2018-01-17 21:57:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-08-09 10:11:58 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Report that FD <fd> can send anymore without polling (EAGAIN detected). */
|
|
|
|
static inline void fd_may_send(const int fd)
|
2012-08-09 10:11:58 +00:00
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
/* marking ready never changes polled status */
|
|
|
|
HA_ATOMIC_OR(&fdtab[fd].state, FD_EV_READY_W);
|
|
|
|
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-08-09 10:11:58 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Prepare FD <fd> to try to receive */
|
|
|
|
static inline void fd_want_recv(int fd)
|
2012-08-09 10:14:03 +00:00
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
unsigned char old, new;
|
|
|
|
|
|
|
|
old = fdtab[fd].state;
|
|
|
|
do {
|
|
|
|
if (old & FD_EV_ACTIVE_R)
|
|
|
|
return;
|
|
|
|
new = old | FD_EV_ACTIVE_R;
|
|
|
|
if (!(new & FD_EV_READY_R))
|
|
|
|
new |= FD_EV_POLLED_R;
|
|
|
|
} while (unlikely(!HA_ATOMIC_CAS(&fdtab[fd].state, &old, new)));
|
|
|
|
|
|
|
|
if ((old ^ new) & FD_EV_POLLED_R)
|
|
|
|
updt_fd_polling(fd);
|
|
|
|
|
2018-01-17 21:57:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-08-09 10:14:03 +00:00
|
|
|
}
|
|
|
|
|
MAJOR: polling: rework the whole polling system
This commit heavily changes the polling system in order to definitely
fix the frequent breakage of SSL which needs to remember the last
EAGAIN before deciding whether to poll or not. Now we have a state per
direction for each FD, as opposed to a previous and current state
previously. An FD can have up to 8 different states for each direction,
each of which being the result of a 3-bit combination. These 3 bits
indicate a wish to access the FD, the readiness of the FD and the
subscription of the FD to the polling system.
This means that it will now be possible to remember the state of a
file descriptor across disable/enable sequences that generally happen
during forwarding, where enabling reading on a previously disabled FD
would result in forgetting the EAGAIN flag it met last time.
Several new state manipulation functions have been introduced or
adapted :
- fd_want_{recv,send} : enable receiving/sending on the FD regardless
of its state (sets the ACTIVE flag) ;
- fd_stop_{recv,send} : stop receiving/sending on the FD regardless
of its state (clears the ACTIVE flag) ;
- fd_cant_{recv,send} : report a failure to receive/send on the FD
corresponding to EAGAIN (clears the READY flag) ;
- fd_may_{recv,send} : report the ability to receive/send on the FD
as reported by poll() (sets the READY flag) ;
Some functions are used to report the current FD status :
- fd_{recv,send}_active
- fd_{recv,send}_ready
- fd_{recv,send}_polled
Some functions were removed :
- fd_ev_clr(), fd_ev_set(), fd_ev_rem(), fd_ev_wai()
The POLLHUP/POLLERR flags are now reported as ready so that the I/O layers
knows it can try to access the file descriptor to get this information.
In order to simplify the conditions to add/remove cache entries, a new
function fd_alloc_or_release_cache_entry() was created to be used from
pollers while scanning for updates.
The following pollers have been updated :
ev_select() : done, built, tested on Linux 3.10
ev_poll() : done, built, tested on Linux 3.10
ev_epoll() : done, built, tested on Linux 3.10 & 3.13
ev_kqueue() : done, built, tested on OpenBSD 5.2
2014-01-10 15:58:45 +00:00
|
|
|
/* Prepare FD <fd> to try to send */
|
|
|
|
static inline void fd_want_send(int fd)
|
2012-08-09 10:11:58 +00:00
|
|
|
{
|
2018-01-17 20:25:57 +00:00
|
|
|
unsigned char old, new;
|
|
|
|
|
|
|
|
old = fdtab[fd].state;
|
|
|
|
do {
|
|
|
|
if (old & FD_EV_ACTIVE_W)
|
|
|
|
return;
|
|
|
|
new = old | FD_EV_ACTIVE_W;
|
|
|
|
if (!(new & FD_EV_READY_W))
|
|
|
|
new |= FD_EV_POLLED_W;
|
|
|
|
} while (unlikely(!HA_ATOMIC_CAS(&fdtab[fd].state, &old, new)));
|
|
|
|
|
|
|
|
if ((old ^ new) & FD_EV_POLLED_W)
|
|
|
|
updt_fd_polling(fd);
|
|
|
|
|
2018-01-17 21:57:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-17 20:25:57 +00:00
|
|
|
fd_update_cache(fd); /* need an update entry to change the state */
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2012-08-09 10:11:58 +00:00
|
|
|
}
|
2006-10-15 12:52:29 +00:00
|
|
|
|
2017-08-30 08:30:04 +00:00
|
|
|
/* Update events seen for FD <fd> and its state if needed. This should be called
|
|
|
|
* by the poller to set FD_POLL_* flags. */
|
|
|
|
static inline void fd_update_events(int fd, int evts)
|
|
|
|
{
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2017-08-30 08:30:04 +00:00
|
|
|
fdtab[fd].ev &= FD_POLL_STICKY;
|
|
|
|
fdtab[fd].ev |= evts;
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2017-08-30 08:30:04 +00:00
|
|
|
|
|
|
|
if (fdtab[fd].ev & (FD_POLL_IN | FD_POLL_HUP | FD_POLL_ERR))
|
|
|
|
fd_may_recv(fd);
|
|
|
|
|
|
|
|
if (fdtab[fd].ev & (FD_POLL_OUT | FD_POLL_ERR))
|
|
|
|
fd_may_send(fd);
|
|
|
|
}
|
|
|
|
|
2008-01-18 16:20:13 +00:00
|
|
|
/* Prepares <fd> for being polled */
|
2018-01-25 06:22:13 +00:00
|
|
|
static inline void fd_insert(int fd, void *owner, void (*iocb)(int fd), unsigned long thread_mask)
|
2006-06-26 00:48:02 +00:00
|
|
|
{
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_LOCK(FD_LOCK, &fdtab[fd].lock);
|
2018-01-25 06:22:13 +00:00
|
|
|
fdtab[fd].owner = owner;
|
|
|
|
fdtab[fd].iocb = iocb;
|
2008-01-18 16:20:13 +00:00
|
|
|
fdtab[fd].ev = 0;
|
2013-12-15 13:19:38 +00:00
|
|
|
fdtab[fd].linger_risk = 0;
|
2014-05-20 12:28:24 +00:00
|
|
|
fdtab[fd].cloned = 0;
|
2017-10-31 15:06:06 +00:00
|
|
|
fdtab[fd].thread_mask = thread_mask;
|
2018-01-17 17:44:46 +00:00
|
|
|
/* note: do not reset polled_mask here as it indicates which poller
|
|
|
|
* still knows this FD from a possible previous round.
|
|
|
|
*/
|
2017-11-07 09:42:54 +00:00
|
|
|
HA_SPIN_UNLOCK(FD_LOCK, &fdtab[fd].lock);
|
2006-06-26 00:48:02 +00:00
|
|
|
}
|
|
|
|
|
2018-01-25 15:37:04 +00:00
|
|
|
/* These are replacements for FD_SET, FD_CLR, FD_ISSET, working on uints */
|
|
|
|
static inline void hap_fd_set(int fd, unsigned int *evts)
|
|
|
|
{
|
2018-01-25 15:59:09 +00:00
|
|
|
HA_ATOMIC_OR(&evts[fd / (8*sizeof(*evts))], 1U << (fd & (8*sizeof(*evts) - 1)));
|
2018-01-25 15:37:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void hap_fd_clr(int fd, unsigned int *evts)
|
|
|
|
{
|
2018-01-25 15:59:09 +00:00
|
|
|
HA_ATOMIC_AND(&evts[fd / (8*sizeof(*evts))], ~(1U << (fd & (8*sizeof(*evts) - 1))));
|
2018-01-25 15:37:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline unsigned int hap_fd_isset(int fd, unsigned int *evts)
|
|
|
|
{
|
|
|
|
return evts[fd / (8*sizeof(*evts))] & (1U << (fd & (8*sizeof(*evts) - 1)));
|
|
|
|
}
|
|
|
|
|
2018-07-26 15:55:11 +00:00
|
|
|
static inline void wake_thread(int tid)
|
|
|
|
{
|
|
|
|
char c = 'c';
|
|
|
|
|
|
|
|
shut_your_big_mouth_gcc(write(poller_wr_pipe[tid], &c, 1));
|
|
|
|
}
|
|
|
|
|
2006-06-26 00:48:02 +00:00
|
|
|
|
|
|
|
#endif /* _PROTO_FD_H */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Local variables:
|
|
|
|
* c-indent-level: 8
|
|
|
|
* c-basic-offset: 8
|
|
|
|
* End:
|
|
|
|
*/
|