On 1/24/25 11:53 AM, Pavel Begunkov wrote:
> [ upstream commit 3181e22fb79910c7071e84a43af93ac89e8a7106 ]
>
> There are reports of mariadb hangs, which is caused by a missing
> barrier in the waking code resulting in waiters losing events.
>
> The problem was introduced in a backport
> 3ab9326f93ec4 ("io_uring: wake up optimisations"),
> and the change restores the barrier present in the original commit
> 3ab9326f93ec4 ("io_uring: wake up optimisations")
>
> Reported by: Xan Charbonnet <xan@xxxxxxxxxxxxxx>
> Fixes: 3ab9326f93ec4 ("io_uring: wake up optimisations")
> Link: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1093243#99
> Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
> ---
> io_uring/io_uring.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
> index 9b58ba4616d40..e5a8ee944ef59 100644
> --- a/io_uring/io_uring.c
> +++ b/io_uring/io_uring.c
> @@ -592,8 +592,10 @@ static inline void __io_cq_unlock_post_flush(struct io_ring_ctx *ctx)
> io_commit_cqring(ctx);
> spin_unlock(&ctx->completion_lock);
> io_commit_cqring_flush(ctx);
> - if (!(ctx->flags & IORING_SETUP_DEFER_TASKRUN))
> + if (!(ctx->flags & IORING_SETUP_DEFER_TASKRUN)) {
> + smp_mb();
> __io_cqring_wake(ctx);
> + }
> }
We could probably just s/__io_cqring_wake/io_cqring_wake here to get
the same effect. Not that it really matters, it's just simpler.
--
Jens Axboe
