io_uring: fix racy overflow count reporting

All ->cq_overflow modifications should be under completion_lock,
otherwise it can report a wrong number to the userspace. Fix it in
io_uring_cancel_files().

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 18:43:47 +03:00 committed by Jens Axboe
parent 81b68a5ca0
commit b2bd1cf99f

View file

@ -7847,10 +7847,9 @@ static void io_uring_cancel_files(struct io_ring_ctx *ctx,
clear_bit(0, &ctx->cq_check_overflow);
ctx->rings->sq_flags &= ~IORING_SQ_CQ_OVERFLOW;
}
spin_unlock_irq(&ctx->completion_lock);
WRITE_ONCE(ctx->rings->cq_overflow,
atomic_inc_return(&ctx->cached_cq_overflow));
spin_unlock_irq(&ctx->completion_lock);
/*
* Put inflight ref and overflow ref. If that's