qemu/include/block/aio.h

640 lines
20 KiB
C
Raw Normal View History

/*
* QEMU aio implementation
*
* Copyright IBM, Corp. 2008
*
* Authors:
* Anthony Liguori <aliguori@us.ibm.com>
*
* This work is licensed under the terms of the GNU GPL, version 2. See
* the COPYING file in the top-level directory.
*
*/
#ifndef QEMU_AIO_H
#define QEMU_AIO_H
#include "qemu/queue.h"
#include "qemu/event_notifier.h"
#include "qemu/thread.h"
#include "qemu/timer.h"
typedef struct BlockAIOCB BlockAIOCB;
typedef void BlockCompletionFunc(void *opaque, int ret);
typedef struct AIOCBInfo {
void (*cancel_async)(BlockAIOCB *acb);
AioContext *(*get_aio_context)(BlockAIOCB *acb);
size_t aiocb_size;
} AIOCBInfo;
struct BlockAIOCB {
const AIOCBInfo *aiocb_info;
BlockDriverState *bs;
BlockCompletionFunc *cb;
void *opaque;
int refcnt;
};
void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs,
BlockCompletionFunc *cb, void *opaque);
void qemu_aio_unref(void *p);
void qemu_aio_ref(void *p);
typedef struct AioHandler AioHandler;
typedef void QEMUBHFunc(void *opaque);
typedef bool AioPollFn(void *opaque);
typedef void IOHandler(void *opaque);
struct Coroutine;
struct ThreadPool;
struct LinuxAioState;
struct LuringState;
struct AioContext {
GSource source;
/* Used by AioContext users to protect from multi-threaded access. */
QemuRecMutex lock;
/* The list of registered AIO handlers. Protected by ctx->list_lock. */
QLIST_HEAD(, AioHandler) aio_handlers;
AioContext: fix broken ctx->dispatching optimization This patch rewrites the ctx->dispatching optimization, which was the cause of some mysterious hangs that could be reproduced on aarch64 KVM only. The hangs were indirectly caused by aio_poll() and in particular by flash memory updates's call to blk_write(), which invokes aio_poll(). Fun stuff: they had an extremely short race window, so much that adding all kind of tracing to either the kernel or QEMU made it go away (a single printf made it half as reproducible). On the plus side, the failure mode (a hang until the next keypress) made it very easy to examine the state of the process with a debugger. And there was a very nice reproducer from Laszlo, which failed pretty often (more than half of the time) on any version of QEMU with a non-debug kernel; it also failed fast, while still in the firmware. So, it could have been worse. For some unknown reason they happened only with virtio-scsi, but that's not important. It's more interesting that they disappeared with io=native, making thread-pool.c a likely suspect for where the bug arose. thread-pool.c is also one of the few places which use bottom halves across threads, by the way. I hope that no other similar bugs exist, but just in case :) I am going to describe how the successful debugging went... Since the likely culprit was the ctx->dispatching optimization, which mostly affects bottom halves, the first observation was that there are two qemu_bh_schedule() invocations in the thread pool: the one in the aio worker and the one in thread_pool_completion_bh. The latter always causes the optimization to trigger, the former may or may not. In order to restrict the possibilities, I introduced new functions qemu_bh_schedule_slow() and qemu_bh_schedule_fast(): /* qemu_bh_schedule_slow: */ ctx = bh->ctx; bh->idle = 0; if (atomic_xchg(&bh->scheduled, 1) == 0) { event_notifier_set(&ctx->notifier); } /* qemu_bh_schedule_fast: */ ctx = bh->ctx; bh->idle = 0; assert(ctx->dispatching); atomic_xchg(&bh->scheduled, 1); Notice how the atomic_xchg is still in qemu_bh_schedule_slow(). This was already debated a few months ago, so I assumed it to be correct. In retrospect this was a very good idea, as you'll see later. Changing thread_pool_completion_bh() to qemu_bh_schedule_fast() didn't trigger the assertion (as expected). Changing the worker's invocation to qemu_bh_schedule_slow() didn't hide the bug (another assumption which luckily held). This already limited heavily the amount of interaction between the threads, hinting that the problematic events must have triggered around thread_pool_completion_bh(). As mentioned early, invoking a debugger to examine the state of a hung process was pretty easy; the iothread was always waiting on a poll(..., -1) system call. Infinite timeouts are much rarer on x86, and this could be the reason why the bug was never observed there. With the buggy sequence more or less resolved to an interaction between thread_pool_completion_bh() and poll(..., -1), my "tracing" strategy was to just add a few qemu_clock_get_ns(QEMU_CLOCK_REALTIME) calls, hoping that the ordering of aio_ctx_prepare(), aio_ctx_dispatch, poll() and qemu_bh_schedule_fast() would provide some hint. The output was: (gdb) p last_prepare $3 = 103885451 (gdb) p last_dispatch $4 = 103876492 (gdb) p last_poll $5 = 115909333 (gdb) p last_schedule $6 = 115925212 Notice how the last call to qemu_poll_ns() came after aio_ctx_dispatch(). This makes little sense unless there is an aio_poll() call involved, and indeed with a slightly different instrumentation you can see that there is one: (gdb) p last_prepare $3 = 107569679 (gdb) p last_dispatch $4 = 107561600 (gdb) p last_aio_poll $5 = 110671400 (gdb) p last_schedule $6 = 110698917 So the scenario becomes clearer: iothread VCPU thread -------------------------------------------------------------------------- aio_ctx_prepare aio_ctx_check qemu_poll_ns(timeout=-1) aio_poll aio_dispatch thread_pool_completion_bh qemu_bh_schedule() At this point bh->scheduled = 1 and the iothread has not been woken up. The solution must be close, but this alone should not be a problem, because the bottom half is only rescheduled to account for rare situations (see commit 3c80ca1, thread-pool: avoid deadlock in nested aio_poll() calls, 2014-07-15). Introducing a third thread---a thread pool worker thread, which also does qemu_bh_schedule()---does bring out the problematic case. The third thread must be awakened *after* the callback is complete and thread_pool_completion_bh has redone the whole loop, explaining the short race window. And then this is what happens: thread pool worker -------------------------------------------------------------------------- <I/O completes> qemu_bh_schedule() Tada, bh->scheduled is already 1, so qemu_bh_schedule() does nothing and the iothread is never woken up. This is where the bh->scheduled optimization comes into play---it is correct, but removing it would have masked the bug. So, what is the bug? Well, the question asked by the ctx->dispatching optimization ("is any active aio_poll dispatching?") was wrong. The right question to ask instead is "is any active aio_poll *not* dispatching", i.e. in the prepare or poll phases? In that case, the aio_poll is sleeping or might go to sleep anytime soon, and the EventNotifier must be invoked to wake it up. In any other case (including if there is *no* active aio_poll at all!) we can just wait for the next prepare phase to pick up the event (e.g. a bottom half); the prepare phase will avoid the blocking and service the bottom half. Expressing the invariant with a logic formula, the broken one looked like: !(exists(thread): in_dispatching(thread)) => !optimize or equivalently: !(exists(thread): in_aio_poll(thread) && in_dispatching(thread)) => !optimize In the correct one, the negation is in a slightly different place: (exists(thread): in_aio_poll(thread) && !in_dispatching(thread)) => !optimize or equivalently: (exists(thread): in_prepare_or_poll(thread)) => !optimize Even if the difference boils down to moving an exclamation mark :) the implementation is quite different. However, I think the new one is simpler to understand. In the old implementation, the "exists" was implemented with a boolean value. This didn't really support well the case of multiple concurrent event loops, but I thought that this was okay: aio_poll holds the AioContext lock so there cannot be concurrent aio_poll invocations, and I was just considering nested event loops. However, aio_poll _could_ indeed be concurrent with the GSource. This is why I came up with the wrong invariant. In the new implementation, "exists" is computed simply by counting how many threads are in the prepare or poll phases. There are some interesting points to consider, but the gist of the idea remains: 1) AioContext can be used through GSource as well; as mentioned in the patch, bit 0 of the counter is reserved for the GSource. 2) the counter need not be updated for a non-blocking aio_poll, because it won't sleep forever anyway. This is just a matter of checking the "blocking" variable. This requires some changes to the win32 implementation, but is otherwise not too complicated. 3) as mentioned above, the new implementation will not call aio_notify when there is *no* active aio_poll at all. The tests have to be adjusted for this change. The calls to aio_notify in async.c are fine; they only want to kick aio_poll out of a blocking wait, but need not do anything if aio_poll is not running. 4) nested aio_poll: these just work with the new implementation; when a nested event loop is invoked, the outer event loop is never in the prepare or poll phases. The outer event loop thus has already decremented the counter. Reported-by: Richard W. M. Jones <rjones@redhat.com> Reported-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Fam Zheng <famz@redhat.com> Tested-by: Richard W.M. Jones <rjones@redhat.com> Message-id: 1437487673-23740-5-git-send-email-pbonzini@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2015-07-21 17:07:51 +03:00
/* Used to avoid unnecessary event_notifier_set calls in aio_notify;
* accessed with atomic primitives. If this field is 0, everything
* (file descriptors, bottom halves, timers) will be re-evaluated
* before the next blocking poll(), thus the event_notifier_set call
* can be skipped. If it is non-zero, you may need to wake up a
* concurrent aio_poll or the glib main event loop, making
* event_notifier_set necessary.
*
* Bit 0 is reserved for GSource usage of the AioContext, and is 1
* between a call to aio_ctx_prepare and the next call to aio_ctx_check.
AioContext: fix broken ctx->dispatching optimization This patch rewrites the ctx->dispatching optimization, which was the cause of some mysterious hangs that could be reproduced on aarch64 KVM only. The hangs were indirectly caused by aio_poll() and in particular by flash memory updates's call to blk_write(), which invokes aio_poll(). Fun stuff: they had an extremely short race window, so much that adding all kind of tracing to either the kernel or QEMU made it go away (a single printf made it half as reproducible). On the plus side, the failure mode (a hang until the next keypress) made it very easy to examine the state of the process with a debugger. And there was a very nice reproducer from Laszlo, which failed pretty often (more than half of the time) on any version of QEMU with a non-debug kernel; it also failed fast, while still in the firmware. So, it could have been worse. For some unknown reason they happened only with virtio-scsi, but that's not important. It's more interesting that they disappeared with io=native, making thread-pool.c a likely suspect for where the bug arose. thread-pool.c is also one of the few places which use bottom halves across threads, by the way. I hope that no other similar bugs exist, but just in case :) I am going to describe how the successful debugging went... Since the likely culprit was the ctx->dispatching optimization, which mostly affects bottom halves, the first observation was that there are two qemu_bh_schedule() invocations in the thread pool: the one in the aio worker and the one in thread_pool_completion_bh. The latter always causes the optimization to trigger, the former may or may not. In order to restrict the possibilities, I introduced new functions qemu_bh_schedule_slow() and qemu_bh_schedule_fast(): /* qemu_bh_schedule_slow: */ ctx = bh->ctx; bh->idle = 0; if (atomic_xchg(&bh->scheduled, 1) == 0) { event_notifier_set(&ctx->notifier); } /* qemu_bh_schedule_fast: */ ctx = bh->ctx; bh->idle = 0; assert(ctx->dispatching); atomic_xchg(&bh->scheduled, 1); Notice how the atomic_xchg is still in qemu_bh_schedule_slow(). This was already debated a few months ago, so I assumed it to be correct. In retrospect this was a very good idea, as you'll see later. Changing thread_pool_completion_bh() to qemu_bh_schedule_fast() didn't trigger the assertion (as expected). Changing the worker's invocation to qemu_bh_schedule_slow() didn't hide the bug (another assumption which luckily held). This already limited heavily the amount of interaction between the threads, hinting that the problematic events must have triggered around thread_pool_completion_bh(). As mentioned early, invoking a debugger to examine the state of a hung process was pretty easy; the iothread was always waiting on a poll(..., -1) system call. Infinite timeouts are much rarer on x86, and this could be the reason why the bug was never observed there. With the buggy sequence more or less resolved to an interaction between thread_pool_completion_bh() and poll(..., -1), my "tracing" strategy was to just add a few qemu_clock_get_ns(QEMU_CLOCK_REALTIME) calls, hoping that the ordering of aio_ctx_prepare(), aio_ctx_dispatch, poll() and qemu_bh_schedule_fast() would provide some hint. The output was: (gdb) p last_prepare $3 = 103885451 (gdb) p last_dispatch $4 = 103876492 (gdb) p last_poll $5 = 115909333 (gdb) p last_schedule $6 = 115925212 Notice how the last call to qemu_poll_ns() came after aio_ctx_dispatch(). This makes little sense unless there is an aio_poll() call involved, and indeed with a slightly different instrumentation you can see that there is one: (gdb) p last_prepare $3 = 107569679 (gdb) p last_dispatch $4 = 107561600 (gdb) p last_aio_poll $5 = 110671400 (gdb) p last_schedule $6 = 110698917 So the scenario becomes clearer: iothread VCPU thread -------------------------------------------------------------------------- aio_ctx_prepare aio_ctx_check qemu_poll_ns(timeout=-1) aio_poll aio_dispatch thread_pool_completion_bh qemu_bh_schedule() At this point bh->scheduled = 1 and the iothread has not been woken up. The solution must be close, but this alone should not be a problem, because the bottom half is only rescheduled to account for rare situations (see commit 3c80ca1, thread-pool: avoid deadlock in nested aio_poll() calls, 2014-07-15). Introducing a third thread---a thread pool worker thread, which also does qemu_bh_schedule()---does bring out the problematic case. The third thread must be awakened *after* the callback is complete and thread_pool_completion_bh has redone the whole loop, explaining the short race window. And then this is what happens: thread pool worker -------------------------------------------------------------------------- <I/O completes> qemu_bh_schedule() Tada, bh->scheduled is already 1, so qemu_bh_schedule() does nothing and the iothread is never woken up. This is where the bh->scheduled optimization comes into play---it is correct, but removing it would have masked the bug. So, what is the bug? Well, the question asked by the ctx->dispatching optimization ("is any active aio_poll dispatching?") was wrong. The right question to ask instead is "is any active aio_poll *not* dispatching", i.e. in the prepare or poll phases? In that case, the aio_poll is sleeping or might go to sleep anytime soon, and the EventNotifier must be invoked to wake it up. In any other case (including if there is *no* active aio_poll at all!) we can just wait for the next prepare phase to pick up the event (e.g. a bottom half); the prepare phase will avoid the blocking and service the bottom half. Expressing the invariant with a logic formula, the broken one looked like: !(exists(thread): in_dispatching(thread)) => !optimize or equivalently: !(exists(thread): in_aio_poll(thread) && in_dispatching(thread)) => !optimize In the correct one, the negation is in a slightly different place: (exists(thread): in_aio_poll(thread) && !in_dispatching(thread)) => !optimize or equivalently: (exists(thread): in_prepare_or_poll(thread)) => !optimize Even if the difference boils down to moving an exclamation mark :) the implementation is quite different. However, I think the new one is simpler to understand. In the old implementation, the "exists" was implemented with a boolean value. This didn't really support well the case of multiple concurrent event loops, but I thought that this was okay: aio_poll holds the AioContext lock so there cannot be concurrent aio_poll invocations, and I was just considering nested event loops. However, aio_poll _could_ indeed be concurrent with the GSource. This is why I came up with the wrong invariant. In the new implementation, "exists" is computed simply by counting how many threads are in the prepare or poll phases. There are some interesting points to consider, but the gist of the idea remains: 1) AioContext can be used through GSource as well; as mentioned in the patch, bit 0 of the counter is reserved for the GSource. 2) the counter need not be updated for a non-blocking aio_poll, because it won't sleep forever anyway. This is just a matter of checking the "blocking" variable. This requires some changes to the win32 implementation, but is otherwise not too complicated. 3) as mentioned above, the new implementation will not call aio_notify when there is *no* active aio_poll at all. The tests have to be adjusted for this change. The calls to aio_notify in async.c are fine; they only want to kick aio_poll out of a blocking wait, but need not do anything if aio_poll is not running. 4) nested aio_poll: these just work with the new implementation; when a nested event loop is invoked, the outer event loop is never in the prepare or poll phases. The outer event loop thus has already decremented the counter. Reported-by: Richard W. M. Jones <rjones@redhat.com> Reported-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Fam Zheng <famz@redhat.com> Tested-by: Richard W.M. Jones <rjones@redhat.com> Message-id: 1437487673-23740-5-git-send-email-pbonzini@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2015-07-21 17:07:51 +03:00
* Bits 1-31 simply count the number of active calls to aio_poll
* that are in the prepare or poll phase.
*
* The GSource and aio_poll must use a different mechanism because
* there is no certainty that a call to GSource's prepare callback
* (via g_main_context_prepare) is indeed followed by check and
* dispatch. It's not clear whether this would be a bug, but let's
* play safe and allow it---it will just cause extra calls to
* event_notifier_set until the next call to dispatch.
*
* Instead, the aio_poll calls include both the prepare and the
* dispatch phase, hence a simple counter is enough for them.
*/
AioContext: fix broken ctx->dispatching optimization This patch rewrites the ctx->dispatching optimization, which was the cause of some mysterious hangs that could be reproduced on aarch64 KVM only. The hangs were indirectly caused by aio_poll() and in particular by flash memory updates's call to blk_write(), which invokes aio_poll(). Fun stuff: they had an extremely short race window, so much that adding all kind of tracing to either the kernel or QEMU made it go away (a single printf made it half as reproducible). On the plus side, the failure mode (a hang until the next keypress) made it very easy to examine the state of the process with a debugger. And there was a very nice reproducer from Laszlo, which failed pretty often (more than half of the time) on any version of QEMU with a non-debug kernel; it also failed fast, while still in the firmware. So, it could have been worse. For some unknown reason they happened only with virtio-scsi, but that's not important. It's more interesting that they disappeared with io=native, making thread-pool.c a likely suspect for where the bug arose. thread-pool.c is also one of the few places which use bottom halves across threads, by the way. I hope that no other similar bugs exist, but just in case :) I am going to describe how the successful debugging went... Since the likely culprit was the ctx->dispatching optimization, which mostly affects bottom halves, the first observation was that there are two qemu_bh_schedule() invocations in the thread pool: the one in the aio worker and the one in thread_pool_completion_bh. The latter always causes the optimization to trigger, the former may or may not. In order to restrict the possibilities, I introduced new functions qemu_bh_schedule_slow() and qemu_bh_schedule_fast(): /* qemu_bh_schedule_slow: */ ctx = bh->ctx; bh->idle = 0; if (atomic_xchg(&bh->scheduled, 1) == 0) { event_notifier_set(&ctx->notifier); } /* qemu_bh_schedule_fast: */ ctx = bh->ctx; bh->idle = 0; assert(ctx->dispatching); atomic_xchg(&bh->scheduled, 1); Notice how the atomic_xchg is still in qemu_bh_schedule_slow(). This was already debated a few months ago, so I assumed it to be correct. In retrospect this was a very good idea, as you'll see later. Changing thread_pool_completion_bh() to qemu_bh_schedule_fast() didn't trigger the assertion (as expected). Changing the worker's invocation to qemu_bh_schedule_slow() didn't hide the bug (another assumption which luckily held). This already limited heavily the amount of interaction between the threads, hinting that the problematic events must have triggered around thread_pool_completion_bh(). As mentioned early, invoking a debugger to examine the state of a hung process was pretty easy; the iothread was always waiting on a poll(..., -1) system call. Infinite timeouts are much rarer on x86, and this could be the reason why the bug was never observed there. With the buggy sequence more or less resolved to an interaction between thread_pool_completion_bh() and poll(..., -1), my "tracing" strategy was to just add a few qemu_clock_get_ns(QEMU_CLOCK_REALTIME) calls, hoping that the ordering of aio_ctx_prepare(), aio_ctx_dispatch, poll() and qemu_bh_schedule_fast() would provide some hint. The output was: (gdb) p last_prepare $3 = 103885451 (gdb) p last_dispatch $4 = 103876492 (gdb) p last_poll $5 = 115909333 (gdb) p last_schedule $6 = 115925212 Notice how the last call to qemu_poll_ns() came after aio_ctx_dispatch(). This makes little sense unless there is an aio_poll() call involved, and indeed with a slightly different instrumentation you can see that there is one: (gdb) p last_prepare $3 = 107569679 (gdb) p last_dispatch $4 = 107561600 (gdb) p last_aio_poll $5 = 110671400 (gdb) p last_schedule $6 = 110698917 So the scenario becomes clearer: iothread VCPU thread -------------------------------------------------------------------------- aio_ctx_prepare aio_ctx_check qemu_poll_ns(timeout=-1) aio_poll aio_dispatch thread_pool_completion_bh qemu_bh_schedule() At this point bh->scheduled = 1 and the iothread has not been woken up. The solution must be close, but this alone should not be a problem, because the bottom half is only rescheduled to account for rare situations (see commit 3c80ca1, thread-pool: avoid deadlock in nested aio_poll() calls, 2014-07-15). Introducing a third thread---a thread pool worker thread, which also does qemu_bh_schedule()---does bring out the problematic case. The third thread must be awakened *after* the callback is complete and thread_pool_completion_bh has redone the whole loop, explaining the short race window. And then this is what happens: thread pool worker -------------------------------------------------------------------------- <I/O completes> qemu_bh_schedule() Tada, bh->scheduled is already 1, so qemu_bh_schedule() does nothing and the iothread is never woken up. This is where the bh->scheduled optimization comes into play---it is correct, but removing it would have masked the bug. So, what is the bug? Well, the question asked by the ctx->dispatching optimization ("is any active aio_poll dispatching?") was wrong. The right question to ask instead is "is any active aio_poll *not* dispatching", i.e. in the prepare or poll phases? In that case, the aio_poll is sleeping or might go to sleep anytime soon, and the EventNotifier must be invoked to wake it up. In any other case (including if there is *no* active aio_poll at all!) we can just wait for the next prepare phase to pick up the event (e.g. a bottom half); the prepare phase will avoid the blocking and service the bottom half. Expressing the invariant with a logic formula, the broken one looked like: !(exists(thread): in_dispatching(thread)) => !optimize or equivalently: !(exists(thread): in_aio_poll(thread) && in_dispatching(thread)) => !optimize In the correct one, the negation is in a slightly different place: (exists(thread): in_aio_poll(thread) && !in_dispatching(thread)) => !optimize or equivalently: (exists(thread): in_prepare_or_poll(thread)) => !optimize Even if the difference boils down to moving an exclamation mark :) the implementation is quite different. However, I think the new one is simpler to understand. In the old implementation, the "exists" was implemented with a boolean value. This didn't really support well the case of multiple concurrent event loops, but I thought that this was okay: aio_poll holds the AioContext lock so there cannot be concurrent aio_poll invocations, and I was just considering nested event loops. However, aio_poll _could_ indeed be concurrent with the GSource. This is why I came up with the wrong invariant. In the new implementation, "exists" is computed simply by counting how many threads are in the prepare or poll phases. There are some interesting points to consider, but the gist of the idea remains: 1) AioContext can be used through GSource as well; as mentioned in the patch, bit 0 of the counter is reserved for the GSource. 2) the counter need not be updated for a non-blocking aio_poll, because it won't sleep forever anyway. This is just a matter of checking the "blocking" variable. This requires some changes to the win32 implementation, but is otherwise not too complicated. 3) as mentioned above, the new implementation will not call aio_notify when there is *no* active aio_poll at all. The tests have to be adjusted for this change. The calls to aio_notify in async.c are fine; they only want to kick aio_poll out of a blocking wait, but need not do anything if aio_poll is not running. 4) nested aio_poll: these just work with the new implementation; when a nested event loop is invoked, the outer event loop is never in the prepare or poll phases. The outer event loop thus has already decremented the counter. Reported-by: Richard W. M. Jones <rjones@redhat.com> Reported-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Fam Zheng <famz@redhat.com> Tested-by: Richard W.M. Jones <rjones@redhat.com> Message-id: 1437487673-23740-5-git-send-email-pbonzini@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2015-07-21 17:07:51 +03:00
uint32_t notify_me;
/* A lock to protect between QEMUBH and AioHandler adders and deleter,
* and to ensure that no callbacks are removed while we're walking and
* dispatching them.
*/
QemuLockCnt list_lock;
/* Anchor of the list of Bottom Halves belonging to the context */
struct QEMUBH *first_bh;
/* Used by aio_notify.
*
* "notified" is used to avoid expensive event_notifier_test_and_clear
* calls. When it is clear, the EventNotifier is clear, or one thread
* is going to clear "notified" before processing more events. False
* positives are possible, i.e. "notified" could be set even though the
* EventNotifier is clear.
*
* Note that event_notifier_set *cannot* be optimized the same way. For
* more information on the problem that would result, see "#ifdef BUG2"
* in the docs/aio_notify_accept.promela formal model.
*/
bool notified;
EventNotifier notifier;
QSLIST_HEAD(, Coroutine) scheduled_coroutines;
QEMUBH *co_schedule_bh;
/* Thread pool for performing work and receiving completion callbacks.
* Has its own locking.
*/
struct ThreadPool *thread_pool;
#ifdef CONFIG_LINUX_AIO
/*
* State for native Linux AIO. Uses aio_context_acquire/release for
* locking.
*/
struct LinuxAioState *linux_aio;
#endif
#ifdef CONFIG_LINUX_IO_URING
/*
* State for Linux io_uring. Uses aio_context_acquire/release for
* locking.
*/
struct LuringState *linux_io_uring;
#endif
/* TimerLists for calling timers - one per clock type. Has its own
* locking.
*/
QEMUTimerListGroup tlg;
int external_disable_cnt;
aio: add polling mode to AioContext The AioContext event loop uses ppoll(2) or epoll_wait(2) to monitor file descriptors or until a timer expires. In cases like virtqueues, Linux AIO, and ThreadPool it is technically possible to wait for events via polling (i.e. continuously checking for events without blocking). Polling can be faster than blocking syscalls because file descriptors, the process scheduler, and system calls are bypassed. The main disadvantage to polling is that it increases CPU utilization. In classic polling configuration a full host CPU thread might run at 100% to respond to events as quickly as possible. This patch implements a timeout so we fall back to blocking syscalls if polling detects no activity. After the timeout no CPU cycles are wasted on polling until the next event loop iteration. The run_poll_handlers_begin() and run_poll_handlers_end() trace events are added to aid performance analysis and troubleshooting. If you need to know whether polling mode is being used, trace these events to find out. Note that the AioContext is now re-acquired before disabling notify_me in the non-polling case. This makes the code cleaner since notify_me was enabled outside the non-polling AioContext release region. This change is correct since it's safe to keep notify_me enabled longer (disabling is an optimization) but potentially causes unnecessary event_notifer_set() calls. I think the chance of performance regression is small here. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Message-id: 20161201192652.9509-4-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2016-12-01 22:26:42 +03:00
/* Number of AioHandlers without .io_poll() */
int poll_disable_cnt;
/* Polling mode parameters */
int64_t poll_ns; /* current polling time in nanoseconds */
int64_t poll_max_ns; /* maximum polling time in nanoseconds */
int64_t poll_grow; /* polling time growth factor */
int64_t poll_shrink; /* polling time shrink factor */
aio: add polling mode to AioContext The AioContext event loop uses ppoll(2) or epoll_wait(2) to monitor file descriptors or until a timer expires. In cases like virtqueues, Linux AIO, and ThreadPool it is technically possible to wait for events via polling (i.e. continuously checking for events without blocking). Polling can be faster than blocking syscalls because file descriptors, the process scheduler, and system calls are bypassed. The main disadvantage to polling is that it increases CPU utilization. In classic polling configuration a full host CPU thread might run at 100% to respond to events as quickly as possible. This patch implements a timeout so we fall back to blocking syscalls if polling detects no activity. After the timeout no CPU cycles are wasted on polling until the next event loop iteration. The run_poll_handlers_begin() and run_poll_handlers_end() trace events are added to aid performance analysis and troubleshooting. If you need to know whether polling mode is being used, trace these events to find out. Note that the AioContext is now re-acquired before disabling notify_me in the non-polling case. This makes the code cleaner since notify_me was enabled outside the non-polling AioContext release region. This change is correct since it's safe to keep notify_me enabled longer (disabling is an optimization) but potentially causes unnecessary event_notifer_set() calls. I think the chance of performance regression is small here. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Message-id: 20161201192652.9509-4-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2016-12-01 22:26:42 +03:00
/* Are we in polling mode or monitoring file descriptors? */
bool poll_started;
/* epoll(7) state used when built with CONFIG_EPOLL */
int epollfd;
bool epoll_enabled;
bool epoll_available;
};
/**
* aio_context_new: Allocate a new AioContext.
*
* AioContext provide a mini event-loop that can be waited on synchronously.
* They also provide bottom halves, a service to execute a piece of code
* as soon as possible.
*/
AioContext *aio_context_new(Error **errp);
/**
* aio_context_ref:
* @ctx: The AioContext to operate on.
*
* Add a reference to an AioContext.
*/
void aio_context_ref(AioContext *ctx);
/**
* aio_context_unref:
* @ctx: The AioContext to operate on.
*
* Drop a reference to an AioContext.
*/
void aio_context_unref(AioContext *ctx);
/* Take ownership of the AioContext. If the AioContext will be shared between
* threads, and a thread does not want to be interrupted, it will have to
* take ownership around calls to aio_poll(). Otherwise, aio_poll()
* automatically takes care of calling aio_context_acquire and
* aio_context_release.
*
* Note that this is separate from bdrv_drained_begin/bdrv_drained_end. A
* thread still has to call those to avoid being interrupted by the guest.
*
* Bottom halves, timers and callbacks can be created or removed without
* acquiring the AioContext.
*/
void aio_context_acquire(AioContext *ctx);
/* Relinquish ownership of the AioContext. */
void aio_context_release(AioContext *ctx);
/**
* aio_bh_schedule_oneshot: Allocate a new bottom half structure that will run
* only once and as soon as possible.
*/
void aio_bh_schedule_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque);
/**
* aio_bh_new: Allocate a new bottom half structure.
*
* Bottom halves are lightweight callbacks whose invocation is guaranteed
* to be wait-free, thread-safe and signal-safe. The #QEMUBH structure
* is opaque and must be allocated prior to its use.
*/
QEMUBH *aio_bh_new(AioContext *ctx, QEMUBHFunc *cb, void *opaque);
/**
* aio_notify: Force processing of pending events.
*
* Similar to signaling a condition variable, aio_notify forces
* aio_poll to exit, so that the next call will re-examine pending events.
* The caller of aio_notify will usually call aio_poll again very soon,
* or go through another iteration of the GLib main loop. Hence, aio_notify
* also has the side effect of recalculating the sets of file descriptors
* that the main loop waits for.
*
* Calling aio_notify is rarely necessary, because for example scheduling
* a bottom half calls it already.
*/
void aio_notify(AioContext *ctx);
/**
* aio_notify_accept: Acknowledge receiving an aio_notify.
*
* aio_notify() uses an EventNotifier in order to wake up a sleeping
* aio_poll() or g_main_context_iteration(). Calls to aio_notify() are
* usually rare, but the AioContext has to clear the EventNotifier on
* every aio_poll() or g_main_context_iteration() in order to avoid
* busy waiting. This event_notifier_test_and_clear() cannot be done
* using the usual aio_context_set_event_notifier(), because it must
* be done before processing all events (file descriptors, bottom halves,
* timers).
*
* aio_notify_accept() is an optimized event_notifier_test_and_clear()
* that is specific to an AioContext's notifier; it is used internally
* to clear the EventNotifier only if aio_notify() had been called.
*/
void aio_notify_accept(AioContext *ctx);
/**
* aio_bh_call: Executes callback function of the specified BH.
*/
void aio_bh_call(QEMUBH *bh);
/**
* aio_bh_poll: Poll bottom halves for an AioContext.
*
* These are internal functions used by the QEMU main loop.
* And notice that multiple occurrences of aio_bh_poll cannot
* be called concurrently
*/
int aio_bh_poll(AioContext *ctx);
/**
* qemu_bh_schedule: Schedule a bottom half.
*
* Scheduling a bottom half interrupts the main loop and causes the
* execution of the callback that was passed to qemu_bh_new.
*
* Bottom halves that are scheduled from a bottom half handler are instantly
* invoked. This can create an infinite loop if a bottom half handler
* schedules itself.
*
* @bh: The bottom half to be scheduled.
*/
void qemu_bh_schedule(QEMUBH *bh);
/**
* qemu_bh_cancel: Cancel execution of a bottom half.
*
* Canceling execution of a bottom half undoes the effect of calls to
* qemu_bh_schedule without freeing its resources yet. While cancellation
* itself is also wait-free and thread-safe, it can of course race with the
* loop that executes bottom halves unless you are holding the iothread
* mutex. This makes it mostly useless if you are not holding the mutex.
*
* @bh: The bottom half to be canceled.
*/
void qemu_bh_cancel(QEMUBH *bh);
/**
*qemu_bh_delete: Cancel execution of a bottom half and free its resources.
*
* Deleting a bottom half frees the memory that was allocated for it by
* qemu_bh_new. It also implies canceling the bottom half if it was
* scheduled.
* This func is async. The bottom half will do the delete action at the finial
* end.
*
* @bh: The bottom half to be deleted.
*/
void qemu_bh_delete(QEMUBH *bh);
/* Return whether there are any pending callbacks from the GSource
* attached to the AioContext, before g_poll is invoked.
*
* This is used internally in the implementation of the GSource.
*/
bool aio_prepare(AioContext *ctx);
/* Return whether there are any pending callbacks from the GSource
* attached to the AioContext, after g_poll is invoked.
*
* This is used internally in the implementation of the GSource.
*/
bool aio_pending(AioContext *ctx);
/* Dispatch any pending callbacks from the GSource attached to the AioContext.
*
* This is used internally in the implementation of the GSource.
*/
void aio_dispatch(AioContext *ctx);
/* Progress in completing AIO work to occur. This can issue new pending
* aio as a result of executing I/O completion or bh callbacks.
*
AioContext: do not rely on aio_poll(ctx, true) result to end a loop Currently, whenever aio_poll(ctx, true) has completed all pending work it returns true *and* the next call to aio_poll(ctx, true) will not block. This invariant has its roots in qemu_aio_flush()'s implementation as "while (qemu_aio_wait()) {}". However, qemu_aio_flush() does not exist anymore and bdrv_drain_all() is implemented differently; and this invariant is complicated to maintain and subtly different from the return value of GMainLoop's g_main_context_iteration. All calls to aio_poll(ctx, true) except one are guarded by a while() loop checking for a request to be incomplete, or a BlockDriverState to be idle. The one remaining call (in iothread.c) uses this to delay the aio_context_release/acquire pair until the AioContext is quiescent, however: - we can do the same just by using non-blocking aio_poll, similar to how vl.c invokes main_loop_wait - it is buggy, because it does not ensure that the AioContext is released between an aio_notify and the next time the iothread goes to sleep. This leads to hangs when stopping the dataplane thread. In the end, these semantics are a bad match for the current users of AioContext. So modify that one exception in iothread.c, which also fixes the hangs, as well as the testcase so that it use the same idiom as the actual QEMU code. Reported-by: Christian Borntraeger <borntraeger@de.ibm.com> Tested-by: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2014-07-09 12:49:46 +04:00
* Return whether any progress was made by executing AIO or bottom half
* handlers. If @blocking == true, this should always be true except
* if someone called aio_notify.
*
* If there are no pending bottom halves, but there are pending AIO
* operations, it may not be possible to make any progress without
* blocking. If @blocking is true, this function will wait until one
* or more AIO events have completed, to ensure something has moved
* before returning.
*/
bool aio_poll(AioContext *ctx, bool blocking);
/* Register a file descriptor and associated callbacks. Behaves very similarly
* to qemu_set_fd_handler. Unlike qemu_set_fd_handler, these callbacks will
* be invoked when using aio_poll().
*
* Code that invokes AIO completion functions should rely on this function
* instead of qemu_set_fd_handler[2].
*/
void aio_set_fd_handler(AioContext *ctx,
int fd,
bool is_external,
IOHandler *io_read,
IOHandler *io_write,
AioPollFn *io_poll,
void *opaque);
/* Set polling begin/end callbacks for a file descriptor that has already been
* registered with aio_set_fd_handler. Do nothing if the file descriptor is
* not registered.
*/
void aio_set_fd_poll(AioContext *ctx, int fd,
IOHandler *io_poll_begin,
IOHandler *io_poll_end);
/* Register an event notifier and associated callbacks. Behaves very similarly
* to event_notifier_set_handler. Unlike event_notifier_set_handler, these callbacks
* will be invoked when using aio_poll().
*
* Code that invokes AIO completion functions should rely on this function
* instead of event_notifier_set_handler.
*/
void aio_set_event_notifier(AioContext *ctx,
EventNotifier *notifier,
bool is_external,
EventNotifierHandler *io_read,
AioPollFn *io_poll);
/* Set polling begin/end callbacks for an event notifier that has already been
* registered with aio_set_event_notifier. Do nothing if the event notifier is
* not registered.
*/
void aio_set_event_notifier_poll(AioContext *ctx,
EventNotifier *notifier,
EventNotifierHandler *io_poll_begin,
EventNotifierHandler *io_poll_end);
/* Return a GSource that lets the main loop poll the file descriptors attached
* to this AioContext.
*/
GSource *aio_get_g_source(AioContext *ctx);
/* Return the ThreadPool bound to this AioContext */
struct ThreadPool *aio_get_thread_pool(AioContext *ctx);
/* Setup the LinuxAioState bound to this AioContext */
struct LinuxAioState *aio_setup_linux_aio(AioContext *ctx, Error **errp);
/* Return the LinuxAioState bound to this AioContext */
struct LinuxAioState *aio_get_linux_aio(AioContext *ctx);
/* Setup the LuringState bound to this AioContext */
struct LuringState *aio_setup_linux_io_uring(AioContext *ctx, Error **errp);
/* Return the LuringState bound to this AioContext */
struct LuringState *aio_get_linux_io_uring(AioContext *ctx);
/**
* aio_timer_new_with_attrs:
* @ctx: the aio context
* @type: the clock type
* @scale: the scale
* @attributes: 0, or one to multiple OR'ed QEMU_TIMER_ATTR_<id> values
* to assign
* @cb: the callback to call on timer expiry
* @opaque: the opaque pointer to pass to the callback
*
* Allocate a new timer (with attributes) attached to the context @ctx.
* The function is responsible for memory allocation.
*
* The preferred interface is aio_timer_init or aio_timer_init_with_attrs.
* Use that unless you really need dynamic memory allocation.
*
* Returns: a pointer to the new timer
*/
static inline QEMUTimer *aio_timer_new_with_attrs(AioContext *ctx,
QEMUClockType type,
int scale, int attributes,
QEMUTimerCB *cb, void *opaque)
{
return timer_new_full(&ctx->tlg, type, scale, attributes, cb, opaque);
}
/**
* aio_timer_new:
* @ctx: the aio context
* @type: the clock type
* @scale: the scale
* @cb: the callback to call on timer expiry
* @opaque: the opaque pointer to pass to the callback
*
* Allocate a new timer attached to the context @ctx.
* See aio_timer_new_with_attrs for details.
*
* Returns: a pointer to the new timer
*/
static inline QEMUTimer *aio_timer_new(AioContext *ctx, QEMUClockType type,
int scale,
QEMUTimerCB *cb, void *opaque)
{
return timer_new_full(&ctx->tlg, type, scale, 0, cb, opaque);
}
/**
* aio_timer_init_with_attrs:
* @ctx: the aio context
* @ts: the timer
* @type: the clock type
* @scale: the scale
* @attributes: 0, or one to multiple OR'ed QEMU_TIMER_ATTR_<id> values
* to assign
* @cb: the callback to call on timer expiry
* @opaque: the opaque pointer to pass to the callback
*
* Initialise a new timer (with attributes) attached to the context @ctx.
* The caller is responsible for memory allocation.
*/
static inline void aio_timer_init_with_attrs(AioContext *ctx,
QEMUTimer *ts, QEMUClockType type,
int scale, int attributes,
QEMUTimerCB *cb, void *opaque)
{
timer_init_full(ts, &ctx->tlg, type, scale, attributes, cb, opaque);
}
/**
* aio_timer_init:
* @ctx: the aio context
* @ts: the timer
* @type: the clock type
* @scale: the scale
* @cb: the callback to call on timer expiry
* @opaque: the opaque pointer to pass to the callback
*
* Initialise a new timer attached to the context @ctx.
* See aio_timer_init_with_attrs for details.
*/
static inline void aio_timer_init(AioContext *ctx,
QEMUTimer *ts, QEMUClockType type,
int scale,
QEMUTimerCB *cb, void *opaque)
{
timer_init_full(ts, &ctx->tlg, type, scale, 0, cb, opaque);
}
/**
* aio_compute_timeout:
* @ctx: the aio context
*
* Compute the timeout that a blocking aio_poll should use.
*/
int64_t aio_compute_timeout(AioContext *ctx);
/**
* aio_disable_external:
* @ctx: the aio context
*
* Disable the further processing of external clients.
*/
static inline void aio_disable_external(AioContext *ctx)
{
atomic_inc(&ctx->external_disable_cnt);
}
/**
* aio_enable_external:
* @ctx: the aio context
*
* Enable the processing of external clients.
*/
static inline void aio_enable_external(AioContext *ctx)
{
aio: add missing aio_notify() to aio_enable_external() The main loop uses aio_disable_external()/aio_enable_external() to temporarily disable processing of external AioContext clients like device emulation. This allows monitor commands to quiesce I/O and prevent the guest from submitting new requests while a monitor command is in progress. The aio_enable_external() API is currently broken when an IOThread is in aio_poll() waiting for fd activity when the main loop re-enables external clients. Incrementing ctx->external_disable_cnt does not wake the IOThread from ppoll(2) so fd processing remains suspended and leads to unresponsive emulated devices. This patch adds an aio_notify() call to aio_enable_external() so the IOThread is kicked out of ppoll(2) and will re-arm the file descriptors. The bug can be reproduced as follows: $ qemu -M accel=kvm -m 1024 \ -object iothread,id=iothread0 \ -device virtio-scsi-pci,iothread=iothread0,id=virtio-scsi-pci0 \ -drive if=none,id=drive0,aio=native,cache=none,format=raw,file=test.img \ -device scsi-hd,id=scsi-hd0,drive=drive0 \ -qmp tcp::5555,server,nowait $ scripts/qmp/qmp-shell localhost:5555 (qemu) blockdev-snapshot-sync device=drive0 snapshot-file=sn1.qcow2 mode=absolute-paths format=qcow2 After blockdev-snapshot-sync completes the SCSI disk will be unresponsive. This leads to request timeouts inside the guest. Reported-by: Qianqian Zhu <qizhu@redhat.com> Reviewed-by: Fam Zheng <famz@redhat.com> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Message-id: 20170508180705.20609-1-stefanha@redhat.com Suggested-by: Fam Zheng <famz@redhat.com> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2017-05-08 21:07:05 +03:00
int old;
old = atomic_fetch_dec(&ctx->external_disable_cnt);
assert(old > 0);
if (old == 1) {
/* Kick event loop so it re-arms file descriptors */
aio_notify(ctx);
}
}
/**
* aio_external_disabled:
* @ctx: the aio context
*
* Return true if the external clients are disabled.
*/
static inline bool aio_external_disabled(AioContext *ctx)
{
return atomic_read(&ctx->external_disable_cnt);
}
/**
* aio_node_check:
* @ctx: the aio context
* @is_external: Whether or not the checked node is an external event source.
*
* Check if the node's is_external flag is okay to be polled by the ctx at this
* moment. True means green light.
*/
static inline bool aio_node_check(AioContext *ctx, bool is_external)
{
return !is_external || !atomic_read(&ctx->external_disable_cnt);
}
/**
* aio_co_schedule:
* @ctx: the aio context
* @co: the coroutine
*
* Start a coroutine on a remote AioContext.
*
* The coroutine must not be entered by anyone else while aio_co_schedule()
* is active. In addition the coroutine must have yielded unless ctx
* is the context in which the coroutine is running (i.e. the value of
* qemu_get_current_aio_context() from the coroutine itself).
*/
void aio_co_schedule(AioContext *ctx, struct Coroutine *co);
/**
* aio_co_wake:
* @co: the coroutine
*
* Restart a coroutine on the AioContext where it was running last, thus
* preventing coroutines from jumping from one context to another when they
* go to sleep.
*
* aio_co_wake may be executed either in coroutine or non-coroutine
* context. The coroutine must not be entered by anyone else while
* aio_co_wake() is active.
*/
void aio_co_wake(struct Coroutine *co);
/**
* aio_co_enter:
* @ctx: the context to run the coroutine
* @co: the coroutine to run
*
* Enter a coroutine in the specified AioContext.
*/
void aio_co_enter(AioContext *ctx, struct Coroutine *co);
/**
* Return the AioContext whose event loop runs in the current thread.
*
* If called from an IOThread this will be the IOThread's AioContext. If
* called from another thread it will be the main loop AioContext.
*/
AioContext *qemu_get_current_aio_context(void);
/**
* in_aio_context_home_thread:
* @ctx: the aio context
*
* Return whether we are running in the thread that normally runs @ctx. Note
* that acquiring/releasing ctx does not affect the outcome, each AioContext
* still only has one home thread that is responsible for running it.
*/
static inline bool in_aio_context_home_thread(AioContext *ctx)
{
return ctx == qemu_get_current_aio_context();
}
/**
* aio_context_setup:
* @ctx: the aio context
*
* Initialize the aio context.
*/
void aio_context_setup(AioContext *ctx);
/**
* aio_context_destroy:
* @ctx: the aio context
*
* Destroy the aio context.
*/
void aio_context_destroy(AioContext *ctx);
aio: add polling mode to AioContext The AioContext event loop uses ppoll(2) or epoll_wait(2) to monitor file descriptors or until a timer expires. In cases like virtqueues, Linux AIO, and ThreadPool it is technically possible to wait for events via polling (i.e. continuously checking for events without blocking). Polling can be faster than blocking syscalls because file descriptors, the process scheduler, and system calls are bypassed. The main disadvantage to polling is that it increases CPU utilization. In classic polling configuration a full host CPU thread might run at 100% to respond to events as quickly as possible. This patch implements a timeout so we fall back to blocking syscalls if polling detects no activity. After the timeout no CPU cycles are wasted on polling until the next event loop iteration. The run_poll_handlers_begin() and run_poll_handlers_end() trace events are added to aid performance analysis and troubleshooting. If you need to know whether polling mode is being used, trace these events to find out. Note that the AioContext is now re-acquired before disabling notify_me in the non-polling case. This makes the code cleaner since notify_me was enabled outside the non-polling AioContext release region. This change is correct since it's safe to keep notify_me enabled longer (disabling is an optimization) but potentially causes unnecessary event_notifer_set() calls. I think the chance of performance regression is small here. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Message-id: 20161201192652.9509-4-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2016-12-01 22:26:42 +03:00
/**
* aio_context_set_poll_params:
* @ctx: the aio context
* @max_ns: how long to busy poll for, in nanoseconds
* @grow: polling time growth factor
* @shrink: polling time shrink factor
aio: add polling mode to AioContext The AioContext event loop uses ppoll(2) or epoll_wait(2) to monitor file descriptors or until a timer expires. In cases like virtqueues, Linux AIO, and ThreadPool it is technically possible to wait for events via polling (i.e. continuously checking for events without blocking). Polling can be faster than blocking syscalls because file descriptors, the process scheduler, and system calls are bypassed. The main disadvantage to polling is that it increases CPU utilization. In classic polling configuration a full host CPU thread might run at 100% to respond to events as quickly as possible. This patch implements a timeout so we fall back to blocking syscalls if polling detects no activity. After the timeout no CPU cycles are wasted on polling until the next event loop iteration. The run_poll_handlers_begin() and run_poll_handlers_end() trace events are added to aid performance analysis and troubleshooting. If you need to know whether polling mode is being used, trace these events to find out. Note that the AioContext is now re-acquired before disabling notify_me in the non-polling case. This makes the code cleaner since notify_me was enabled outside the non-polling AioContext release region. This change is correct since it's safe to keep notify_me enabled longer (disabling is an optimization) but potentially causes unnecessary event_notifer_set() calls. I think the chance of performance regression is small here. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Message-id: 20161201192652.9509-4-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2016-12-01 22:26:42 +03:00
*
* Poll mode can be disabled by setting poll_max_ns to 0.
*/
void aio_context_set_poll_params(AioContext *ctx, int64_t max_ns,
int64_t grow, int64_t shrink,
aio: add polling mode to AioContext The AioContext event loop uses ppoll(2) or epoll_wait(2) to monitor file descriptors or until a timer expires. In cases like virtqueues, Linux AIO, and ThreadPool it is technically possible to wait for events via polling (i.e. continuously checking for events without blocking). Polling can be faster than blocking syscalls because file descriptors, the process scheduler, and system calls are bypassed. The main disadvantage to polling is that it increases CPU utilization. In classic polling configuration a full host CPU thread might run at 100% to respond to events as quickly as possible. This patch implements a timeout so we fall back to blocking syscalls if polling detects no activity. After the timeout no CPU cycles are wasted on polling until the next event loop iteration. The run_poll_handlers_begin() and run_poll_handlers_end() trace events are added to aid performance analysis and troubleshooting. If you need to know whether polling mode is being used, trace these events to find out. Note that the AioContext is now re-acquired before disabling notify_me in the non-polling case. This makes the code cleaner since notify_me was enabled outside the non-polling AioContext release region. This change is correct since it's safe to keep notify_me enabled longer (disabling is an optimization) but potentially causes unnecessary event_notifer_set() calls. I think the chance of performance regression is small here. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Message-id: 20161201192652.9509-4-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2016-12-01 22:26:42 +03:00
Error **errp);
#endif