io_uring: get rid of atomic FAA for cq_timeouts

If ->cq_timeouts modifications are done under ->completion_lock, we
don't really nee any fetch-and-add and other complex atomics. Replace it
with non-atomic FAA, that saves an implicit full memory barrier.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This commit is contained in:
Pavel Begunkov 2020-07-30 18:43:50 +03:00 committed by Jens Axboe
parent 4693014340
commit 01cec8c18f

View File

@ -1205,7 +1205,8 @@ static void io_kill_timeout(struct io_kiocb *req)
ret = hrtimer_try_to_cancel(&req->io->timeout.timer);
if (ret != -1) {
atomic_inc(&req->ctx->cq_timeouts);
atomic_set(&req->ctx->cq_timeouts,
atomic_read(&req->ctx->cq_timeouts) + 1);
list_del_init(&req->timeout.list);
req->flags |= REQ_F_COMP_LOCKED;
io_cqring_fill_event(req, 0);
@ -4972,9 +4973,10 @@ static enum hrtimer_restart io_timeout_fn(struct hrtimer *timer)
struct io_ring_ctx *ctx = req->ctx;
unsigned long flags;
atomic_inc(&ctx->cq_timeouts);
spin_lock_irqsave(&ctx->completion_lock, flags);
atomic_set(&req->ctx->cq_timeouts,
atomic_read(&req->ctx->cq_timeouts) + 1);
/*
* We could be racing with timeout deletion. If the list is empty,
* then timeout lookup already found it and will be handling it.