mirror of
https://github.com/torvalds/linux.git
synced 2026-11-20 13:03:48 +00:00
When io_cqring_wait() gets to the point of sleeping and the wait's timeout has already passed, io_cqring_schedule_timeout() still arms an hrtimer for it and calls schedule(). The timer fires right away, but only after the task has switched to idle, so getting -ETIME back costs a round trip through idle plus a timer interrupt. NAPI busy polling hits this on every wait that times out. __io_napi_busy_loop() caps the busy poll at the remaining timeout, so when the loop runs to its end without a completion, the timeout has expired by the time we get here. The task then stops polling for several usec, and packets that arrive meanwhile are left to the device interrupt. This affects one of our production workloads, which currently works around it by having another thread signal an eventfd on the ring before the timeout expires. A wait with a zero timeout, or one that runs out before the task gets to sleep, pays the same cost. io_uring_wait_cqe_timeout() with nothing to reap and a zero timeout takes ~11 usec and a voluntary context switch on a Zen 5 desktop. Return -ETIME without arming the timer once the timeout has passed, as ep_poll() already does with ep_schedule_timeout(). A passed timeout ends the wait with -ETIME whether or not a min_timeout is set, so the check covers both. It costs one io_get_time() per wait that is about to sleep with a timeout. Signed-off-by: Ioannis Coward <[email protected]> Link: https://patch.msgid.link/[email protected] Signed-off-by: Jens Axboe <[email protected]>
359 lines
9.8 KiB
C
359 lines
9.8 KiB
C
// SPDX-License-Identifier: GPL-2.0
|
|
/*
|
|
* Waiting for completion events
|
|
*/
|
|
#include <linux/kernel.h>
|
|
#include <linux/sched/rt.h>
|
|
#include <linux/sched/signal.h>
|
|
#include <linux/io_uring.h>
|
|
#include <linux/time_namespace.h>
|
|
|
|
#include <trace/events/io_uring.h>
|
|
|
|
#include <uapi/linux/io_uring.h>
|
|
|
|
#include "io_uring.h"
|
|
#include "napi.h"
|
|
#include "wait.h"
|
|
|
|
static int io_wake_function(struct wait_queue_entry *curr, unsigned int mode,
|
|
int wake_flags, void *key)
|
|
{
|
|
struct io_wait_queue *iowq = container_of(curr, struct io_wait_queue, wq);
|
|
|
|
/*
|
|
* Cannot safely flush overflowed CQEs from here, ensure we wake up
|
|
* the task, and the next invocation will do it.
|
|
*/
|
|
if (io_should_wake(iowq) || io_has_work(iowq->ctx))
|
|
return autoremove_wake_function(curr, mode, wake_flags, key);
|
|
/* keep going, a waiter further down may be content with less */
|
|
return 0;
|
|
}
|
|
|
|
int io_run_task_work_sig(struct io_ring_ctx *ctx)
|
|
{
|
|
if (io_local_work_pending(ctx)) {
|
|
__set_current_state(TASK_RUNNING);
|
|
if (io_run_local_work(ctx, INT_MAX, IO_LOCAL_TW_DEFAULT_MAX) > 0)
|
|
return 0;
|
|
}
|
|
if (io_run_task_work() > 0)
|
|
return 0;
|
|
if (task_sigpending(current))
|
|
return -EINTR;
|
|
return 0;
|
|
}
|
|
|
|
static bool current_pending_io(void)
|
|
{
|
|
struct io_uring_task *tctx = current->io_uring;
|
|
|
|
if (!tctx)
|
|
return false;
|
|
return percpu_counter_read_positive(&tctx->inflight);
|
|
}
|
|
|
|
static enum hrtimer_restart io_cqring_timer_wakeup(struct hrtimer *timer)
|
|
{
|
|
struct io_wait_queue *iowq = container_of(timer, struct io_wait_queue, t);
|
|
|
|
WRITE_ONCE(iowq->hit_timeout, 1);
|
|
iowq->min_timeout = 0;
|
|
wake_up_process(iowq->wq.private);
|
|
return HRTIMER_NORESTART;
|
|
}
|
|
|
|
/*
|
|
* Doing min_timeout portion. If we saw any timeouts, events, or have work,
|
|
* wake up. If not, and we have a normal timeout, switch to that and keep
|
|
* sleeping.
|
|
*/
|
|
static enum hrtimer_restart io_cqring_min_timer_wakeup(struct hrtimer *timer)
|
|
{
|
|
struct io_wait_queue *iowq = container_of(timer, struct io_wait_queue, t);
|
|
struct io_ring_ctx *ctx = iowq->ctx;
|
|
|
|
/* no general timeout, or shorter (or equal), we are done */
|
|
if (iowq->timeout == KTIME_MAX ||
|
|
ktime_compare(iowq->min_timeout, iowq->timeout) >= 0)
|
|
goto out_wake;
|
|
/*
|
|
* Work we need to run. It isn't known if that is going to post a CQE,
|
|
* so this isn't a timeout. Carry on with the normal one if it doesn't.
|
|
*/
|
|
if (io_has_work(ctx)) {
|
|
iowq->min_timeout = 0;
|
|
iowq->cq_tail = iowq->cq_min_tail + 1;
|
|
wake_up_process(iowq->wq.private);
|
|
return HRTIMER_NORESTART;
|
|
}
|
|
/* got events since we started waiting, min timeout is done */
|
|
scoped_guard(rcu) {
|
|
struct io_rings *rings = io_get_rings(ctx);
|
|
|
|
if (iowq->cq_min_tail != READ_ONCE(rings->cq.tail))
|
|
goto out_wake;
|
|
/* if we have any events and min timeout expired, we're done */
|
|
if (io_cqring_events(ctx))
|
|
goto out_wake;
|
|
}
|
|
/*
|
|
* If using deferred task_work running and application is waiting on
|
|
* more than one request, ensure we reset it now where we are switching
|
|
* to normal sleeps. Any request completion post min_wait should wake
|
|
* the task and return.
|
|
*/
|
|
if (ctx->flags & IORING_SETUP_DEFER_TASKRUN) {
|
|
atomic_set(&ctx->cq_wait_nr, 1);
|
|
smp_mb();
|
|
if (io_local_work_pending(ctx))
|
|
goto out_wake;
|
|
}
|
|
|
|
/* any generated CQE posted past this time should wake us up */
|
|
iowq->cq_tail = iowq->cq_min_tail + 1;
|
|
|
|
hrtimer_update_function(&iowq->t, io_cqring_timer_wakeup);
|
|
hrtimer_set_expires(timer, iowq->timeout);
|
|
return HRTIMER_RESTART;
|
|
out_wake:
|
|
return io_cqring_timer_wakeup(timer);
|
|
}
|
|
|
|
/* Same policy as for hrtimer sleepers, see __hrtimer_setup_sleeper(). */
|
|
static enum hrtimer_mode io_cqring_timer_mode(void)
|
|
{
|
|
if (IS_ENABLED(CONFIG_PREEMPT_RT) && rt_or_dl_task_policy(current))
|
|
return HRTIMER_MODE_ABS_HARD;
|
|
return HRTIMER_MODE_ABS;
|
|
}
|
|
|
|
static int io_cqring_schedule_timeout(struct io_wait_queue *iowq,
|
|
clockid_t clock_id, ktime_t start_time)
|
|
{
|
|
enum hrtimer_mode mode = io_cqring_timer_mode();
|
|
ktime_t timeout;
|
|
|
|
if (iowq->min_timeout) {
|
|
timeout = ktime_add_ns(iowq->min_timeout, start_time);
|
|
hrtimer_setup_on_stack(&iowq->t, io_cqring_min_timer_wakeup, clock_id,
|
|
mode);
|
|
} else {
|
|
timeout = iowq->timeout;
|
|
hrtimer_setup_on_stack(&iowq->t, io_cqring_timer_wakeup, clock_id,
|
|
mode);
|
|
}
|
|
|
|
hrtimer_set_expires_range_ns(&iowq->t, timeout, 0);
|
|
hrtimer_start_expires(&iowq->t, mode);
|
|
|
|
if (!READ_ONCE(iowq->hit_timeout))
|
|
schedule();
|
|
|
|
hrtimer_cancel(&iowq->t);
|
|
destroy_hrtimer_on_stack(&iowq->t);
|
|
__set_current_state(TASK_RUNNING);
|
|
|
|
return READ_ONCE(iowq->hit_timeout) ? -ETIME : 0;
|
|
}
|
|
|
|
static int __io_cqring_wait_schedule(struct io_ring_ctx *ctx,
|
|
struct io_wait_queue *iowq,
|
|
struct ext_arg *ext_arg,
|
|
ktime_t start_time)
|
|
{
|
|
int ret = 0;
|
|
|
|
/*
|
|
* Don't arm a timer for a timeout that has already passed, it would
|
|
* only fire to return -ETIME. NAPI busy polling gets here like that.
|
|
*/
|
|
if (iowq->timeout != KTIME_MAX &&
|
|
ktime_compare(io_get_time(ctx), iowq->timeout) >= 0)
|
|
return -ETIME;
|
|
|
|
/*
|
|
* Mark us as being in io_wait if we have pending requests, so cpufreq
|
|
* can take into account that the task is waiting for IO - turns out
|
|
* to be important for low QD IO.
|
|
*/
|
|
if (ext_arg->iowait && current_pending_io())
|
|
current->in_iowait = 1;
|
|
if (iowq->timeout != KTIME_MAX || iowq->min_timeout)
|
|
ret = io_cqring_schedule_timeout(iowq, ctx->clockid, start_time);
|
|
else
|
|
schedule();
|
|
current->in_iowait = 0;
|
|
return ret;
|
|
}
|
|
|
|
/* If this returns > 0, the caller should retry */
|
|
static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,
|
|
struct io_wait_queue *iowq,
|
|
struct ext_arg *ext_arg,
|
|
ktime_t start_time)
|
|
{
|
|
if (unlikely(READ_ONCE(ctx->check_cq)))
|
|
return 1;
|
|
if (unlikely(io_local_work_pending(ctx)))
|
|
return 1;
|
|
if (unlikely(task_work_pending(current)))
|
|
return 1;
|
|
if (unlikely(task_sigpending(current)))
|
|
return -EINTR;
|
|
if (unlikely(io_should_wake(iowq)))
|
|
return 0;
|
|
|
|
return __io_cqring_wait_schedule(ctx, iowq, ext_arg, start_time);
|
|
}
|
|
|
|
/*
|
|
* Wait until events become available, if we don't already have some. The
|
|
* application must reap them itself, as they reside on the shared cq ring.
|
|
*/
|
|
int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, u32 flags,
|
|
struct ext_arg *ext_arg)
|
|
{
|
|
struct io_wait_queue iowq;
|
|
struct io_rings *rings;
|
|
ktime_t start_time;
|
|
int ret, nr_wait;
|
|
|
|
min_events = min_t(int, min_events, ctx->cq_entries);
|
|
|
|
if (!io_allowed_run_tw(ctx))
|
|
return -EEXIST;
|
|
if (io_local_work_pending(ctx))
|
|
io_run_local_work(ctx, min_events,
|
|
max(IO_LOCAL_TW_DEFAULT_MAX, min_events));
|
|
io_run_task_work();
|
|
|
|
if (unlikely(test_bit(IO_CHECK_CQ_OVERFLOW_BIT, &ctx->check_cq)))
|
|
io_cqring_do_overflow_flush(ctx);
|
|
|
|
rcu_read_lock();
|
|
rings = io_get_rings(ctx);
|
|
if (__io_cqring_events_user(ctx) >= min_events) {
|
|
rcu_read_unlock();
|
|
return 0;
|
|
}
|
|
|
|
init_waitqueue_func_entry(&iowq.wq, io_wake_function);
|
|
iowq.wq.private = current;
|
|
INIT_LIST_HEAD(&iowq.wq.entry);
|
|
iowq.ctx = ctx;
|
|
iowq.cq_tail = READ_ONCE(rings->cq.head) + min_events;
|
|
iowq.cq_min_tail = READ_ONCE(rings->cq.tail);
|
|
nr_wait = (int) iowq.cq_tail - READ_ONCE(rings->cq.tail);
|
|
rcu_read_unlock();
|
|
rings = NULL;
|
|
iowq.nr_timeouts = atomic_read(&ctx->cq_timeouts);
|
|
iowq.hit_timeout = 0;
|
|
iowq.min_timeout = ext_arg->min_time;
|
|
iowq.timeout = KTIME_MAX;
|
|
start_time = io_get_time(ctx);
|
|
|
|
if (ext_arg->ts_set) {
|
|
iowq.timeout = timespec64_to_ktime(ext_arg->ts);
|
|
if (flags & IORING_ENTER_ABS_TIMER)
|
|
iowq.timeout = timens_ktime_to_host(ctx->clockid,
|
|
iowq.timeout);
|
|
else
|
|
iowq.timeout = ktime_add(iowq.timeout, start_time);
|
|
}
|
|
|
|
if (ext_arg->sig) {
|
|
#ifdef CONFIG_COMPAT
|
|
if (in_compat_syscall())
|
|
ret = set_compat_user_sigmask((const compat_sigset_t __user *)ext_arg->sig,
|
|
ext_arg->argsz);
|
|
else
|
|
#endif
|
|
ret = set_user_sigmask(ext_arg->sig, ext_arg->argsz);
|
|
|
|
if (ret)
|
|
return ret;
|
|
}
|
|
|
|
io_napi_busy_loop(ctx, &iowq);
|
|
|
|
trace_io_uring_cqring_wait(ctx, min_events);
|
|
do {
|
|
unsigned long check_cq;
|
|
|
|
if (ctx->flags & IORING_SETUP_DEFER_TASKRUN) {
|
|
io_cq_wait_arm(ctx, nr_wait);
|
|
} else {
|
|
prepare_to_wait_exclusive(&ctx->cq_wait, &iowq.wq,
|
|
TASK_INTERRUPTIBLE);
|
|
}
|
|
|
|
ret = io_cqring_wait_schedule(ctx, &iowq, ext_arg, start_time);
|
|
__set_current_state(TASK_RUNNING);
|
|
atomic_set(&ctx->cq_wait_nr, IO_CQ_WAKE_INIT);
|
|
|
|
/*
|
|
* Run task_work after scheduling and before io_should_wake().
|
|
* If we got woken because of task_work being processed, run it
|
|
* now rather than let the caller do another wait loop.
|
|
*/
|
|
if (io_local_work_pending(ctx))
|
|
io_run_local_work(ctx, nr_wait, nr_wait);
|
|
io_run_task_work();
|
|
|
|
/*
|
|
* Non-local task_work will be run on exit to userspace, but
|
|
* if we're using DEFER_TASKRUN, then we could have waited
|
|
* with a timeout for a number of requests. If the timeout
|
|
* hits, we could have some requests ready to process. Ensure
|
|
* this break is _after_ we have run task_work, to avoid
|
|
* deferring running potentially pending requests until the
|
|
* next time we wait for events.
|
|
*/
|
|
if (ret < 0)
|
|
break;
|
|
|
|
check_cq = READ_ONCE(ctx->check_cq);
|
|
if (unlikely(check_cq)) {
|
|
if (check_cq & BIT(IO_CHECK_CQ_OVERFLOW_BIT)) {
|
|
io_cqring_do_overflow_flush(ctx);
|
|
/*
|
|
* No room for what has overflown. It takes the
|
|
* application to change that, so return to it.
|
|
*/
|
|
if (test_bit(IO_CHECK_CQ_OVERFLOW_BIT,
|
|
&ctx->check_cq)) {
|
|
ret = 0;
|
|
break;
|
|
}
|
|
}
|
|
if (check_cq & BIT(IO_CHECK_CQ_DROPPED_BIT)) {
|
|
ret = -EBADR;
|
|
break;
|
|
}
|
|
}
|
|
|
|
if (io_should_wake(&iowq)) {
|
|
ret = 0;
|
|
break;
|
|
}
|
|
cond_resched();
|
|
|
|
/* if min timeout has been hit, don't reset wait count */
|
|
if (!iowq.hit_timeout)
|
|
scoped_guard(rcu)
|
|
nr_wait = (int) iowq.cq_tail -
|
|
READ_ONCE(io_get_rings(ctx)->cq.tail);
|
|
else
|
|
nr_wait = 1;
|
|
} while (1);
|
|
|
|
if (!(ctx->flags & IORING_SETUP_DEFER_TASKRUN))
|
|
finish_wait(&ctx->cq_wait, &iowq.wq);
|
|
restore_saved_sigmask_unless(ret == -EINTR);
|
|
|
|
guard(rcu)();
|
|
return READ_ONCE(io_get_rings(ctx)->cq.head) == READ_ONCE(io_get_rings(ctx)->cq.tail) ? ret : 0;
|
|
}
|