haproxy/include/proto/fd.h
Willy Tarreau 5be2f35231 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.
2014-11-21 20:37:32 +01:00

351 lines
9.5 KiB
C

/*
* include/proto/fd.h
* File descriptors states.
*
* Copyright (C) 2000-2014 Willy Tarreau - w@1wt.eu
*
* 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
*/
#ifndef _PROTO_FD_H
#define _PROTO_FD_H
#include <stdio.h>
#include <sys/time.h>
#include <sys/types.h>
#include <unistd.h>
#include <common/config.h>
#include <types/fd.h>
/* public variables */
extern unsigned int *fd_cache; // FD events cache
extern unsigned int *fd_updt; // FD updates list
extern int fd_cache_num; // number of events in the cache
extern int fd_nbupdt; // number of updates in the list
/* Deletes an FD from the fdsets, and recomputes the maxfd limit.
* The file descriptor is also closed.
*/
void fd_delete(int fd);
/* disable the specified poller */
void disable_poller(const char *poller_name);
/*
* Initialize the pollers till the best one is found.
* If none works, returns 0, otherwise 1.
* The pollers register themselves just before main() is called.
*/
int init_pollers();
/*
* Deinitialize the pollers.
*/
void deinit_pollers();
/*
* 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);
/*
* Runs the polling loop
*/
void run_poller();
/* Scan and process the cached events. This should be called right after
* the poller.
*/
void fd_process_cached_events();
/* 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.
*/
static inline void updt_fd_polling(const int fd)
{
if (fdtab[fd].updated)
/* already scheduled for update */
return;
fdtab[fd].updated = 1;
fd_updt[fd_nbupdt++] = fd;
}
/* 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)
{
if (fdtab[fd].cache)
return;
fd_cache_num++;
fdtab[fd].cache = fd_cache_num;
fd_cache[fd_cache_num-1] = fd;
}
/* Removes entry used by fd <fd> from the FD cache and replaces it with the
* last one. The fdtab.cache is adjusted to match the back reference if needed.
* If the fd has no entry assigned, return immediately.
*/
static inline void fd_release_cache_entry(int fd)
{
unsigned int pos;
pos = fdtab[fd].cache;
if (!pos)
return;
fdtab[fd].cache = 0;
fd_cache_num--;
if (likely(pos <= fd_cache_num)) {
/* was not the last entry */
fd = fd_cache[fd_cache_num];
fd_cache[pos - 1] = fd;
fdtab[fd].cache = pos;
}
}
/* Computes the new polled status based on the active and ready statuses, for
* each direction. This is meant to be used by pollers while processing updates.
*/
static inline int fd_compute_new_polled_status(int state)
{
if (state & FD_EV_ACTIVE_R) {
if (!(state & FD_EV_READY_R))
state |= FD_EV_POLLED_R;
}
else
state &= ~FD_EV_POLLED_R;
if (state & FD_EV_ACTIVE_W) {
if (!(state & FD_EV_READY_W))
state |= FD_EV_POLLED_W;
}
else
state &= ~FD_EV_POLLED_W;
return state;
}
/* This function automatically enables/disables caching for an entry depending
* on its state, and also possibly creates an update entry so that the poller
* does its job as well. It is only called on state changes.
*/
static inline void fd_update_cache(int fd)
{
/* 3 states for each direction require a polling update */
if ((fdtab[fd].state & (FD_EV_POLLED_R | FD_EV_ACTIVE_R)) == FD_EV_POLLED_R ||
(fdtab[fd].state & (FD_EV_POLLED_R | FD_EV_READY_R | FD_EV_ACTIVE_R)) == FD_EV_ACTIVE_R ||
(fdtab[fd].state & (FD_EV_POLLED_W | FD_EV_ACTIVE_W)) == FD_EV_POLLED_W ||
(fdtab[fd].state & (FD_EV_POLLED_W | FD_EV_READY_W | FD_EV_ACTIVE_W)) == FD_EV_ACTIVE_W)
updt_fd_polling(fd);
/* 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))) {
fd_alloc_cache_entry(fd);
}
else {
fd_release_cache_entry(fd);
}
}
/*
* returns the FD's recv state (FD_EV_*)
*/
static inline int fd_recv_state(const int fd)
{
return ((unsigned)fdtab[fd].state >> (4 * DIR_RD)) & FD_EV_STATUS;
}
/*
* returns true if the FD is active for recv
*/
static inline int fd_recv_active(const int fd)
{
return (unsigned)fdtab[fd].state & FD_EV_ACTIVE_R;
}
/*
* returns true if the FD is ready for recv
*/
static inline int fd_recv_ready(const int fd)
{
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;
}
/*
* returns true if the FD is ready for send
*/
static inline int fd_send_ready(const int fd)
{
return (unsigned)fdtab[fd].state & FD_EV_READY_W;
}
/*
* 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;
}
/* Disable processing recv events on fd <fd> */
static inline void fd_stop_recv(int fd)
{
if (!((unsigned int)fdtab[fd].state & FD_EV_ACTIVE_R))
return; /* already disabled */
fdtab[fd].state &= ~FD_EV_ACTIVE_R;
fd_update_cache(fd); /* need an update entry to change the state */
}
/* Disable processing send events on fd <fd> */
static inline void fd_stop_send(int fd)
{
if (!((unsigned int)fdtab[fd].state & FD_EV_ACTIVE_W))
return; /* already disabled */
fdtab[fd].state &= ~FD_EV_ACTIVE_W;
fd_update_cache(fd); /* need an update entry to change the state */
}
/* Disable processing of events on fd <fd> for both directions. */
static inline void fd_stop_both(int fd)
{
if (!((unsigned int)fdtab[fd].state & FD_EV_ACTIVE_RW))
return; /* already disabled */
fdtab[fd].state &= ~FD_EV_ACTIVE_RW;
fd_update_cache(fd); /* need an update entry to change the state */
}
/* Report that FD <fd> cannot receive anymore without polling (EAGAIN detected). */
static inline void fd_cant_recv(const int fd)
{
if (!(((unsigned int)fdtab[fd].state) & FD_EV_READY_R))
return; /* already marked as blocked */
fdtab[fd].state &= ~FD_EV_READY_R;
fd_update_cache(fd);
}
/* Report that FD <fd> can receive anymore without polling. */
static inline void fd_may_recv(const int fd)
{
if (((unsigned int)fdtab[fd].state) & FD_EV_READY_R)
return; /* already marked as blocked */
fdtab[fd].state |= FD_EV_READY_R;
fd_update_cache(fd);
}
/* 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)
{
if (fd_recv_polled(fd))
fd_cant_recv(fd);
}
/* Report that FD <fd> cannot send anymore without polling (EAGAIN detected). */
static inline void fd_cant_send(const int fd)
{
if (!(((unsigned int)fdtab[fd].state) & FD_EV_READY_W))
return; /* already marked as blocked */
fdtab[fd].state &= ~FD_EV_READY_W;
fd_update_cache(fd);
}
/* Report that FD <fd> can send anymore without polling (EAGAIN detected). */
static inline void fd_may_send(const int fd)
{
if (((unsigned int)fdtab[fd].state) & FD_EV_READY_W)
return; /* already marked as blocked */
fdtab[fd].state |= FD_EV_READY_W;
fd_update_cache(fd);
}
/* Prepare FD <fd> to try to receive */
static inline void fd_want_recv(int fd)
{
if (((unsigned int)fdtab[fd].state & FD_EV_ACTIVE_R))
return; /* already enabled */
fdtab[fd].state |= FD_EV_ACTIVE_R;
fd_update_cache(fd); /* need an update entry to change the state */
}
/* Prepare FD <fd> to try to send */
static inline void fd_want_send(int fd)
{
if (((unsigned int)fdtab[fd].state & FD_EV_ACTIVE_W))
return; /* already enabled */
fdtab[fd].state |= FD_EV_ACTIVE_W;
fd_update_cache(fd); /* need an update entry to change the state */
}
/* Prepares <fd> for being polled */
static inline void fd_insert(int fd)
{
fdtab[fd].ev = 0;
fdtab[fd].new = 1;
fdtab[fd].linger_risk = 0;
fdtab[fd].cloned = 0;
if (fd + 1 > maxfd)
maxfd = fd + 1;
}
#endif /* _PROTO_FD_H */
/*
* Local variables:
* c-indent-level: 8
* c-basic-offset: 8
* End:
*/