Re: [PATCHv2] io_uring: consistently use rcu semantics with sqpoll thread

[Jens Axboe <axboe@kernel.dk>]

On 6/11/25 3:01 PM, Jens Axboe wrote:
> 
> On Wed, 11 Jun 2025 13:53:43 -0700, Keith Busch wrote:
>> The sqpoll thread is dereferenced with rcu read protection in one place,
>> so it needs to be annotated as an __rcu type, and should consistently
>> use rcu helpers for access and assignment to make sparse happy.
>>
>> Since most of the accesses occur under the sqd->lock, we can use
>> rcu_dereference_protected() without declaring an rcu read section.
>> Provide a simple helper to get the thread from a locked context.
>>
>> [...]
> 
> Applied, thanks!
> 
> [1/1] io_uring: consistently use rcu semantics with sqpoll thread
>       commit: 89f607e88c9f885187a0144d0b540fb257b912ea

Folded in the below, looks like that one was missed. We should probably
have renamed ->thread to avoid this kind of stuff, but oh well.

diff --git a/io_uring/register.c b/io_uring/register.c
index cc23a4c205cd..a59589249fce 100644
--- a/io_uring/register.c
+++ b/io_uring/register.c
@@ -273,6 +273,8 @@ static __cold int io_register_iowq_max_workers(struct io_ring_ctx *ctx,
 	if (ctx->flags & IORING_SETUP_SQPOLL) {
 		sqd = ctx->sq_data;
 		if (sqd) {
+			struct task_struct *tsk;
+
 			/*
 			 * Observe the correct sqd->lock -> ctx->uring_lock
 			 * ordering. Fine to drop uring_lock here, we hold
@@ -282,8 +284,9 @@ static __cold int io_register_iowq_max_workers(struct io_ring_ctx *ctx,
 			mutex_unlock(&ctx->uring_lock);
 			mutex_lock(&sqd->lock);
 			mutex_lock(&ctx->uring_lock);
-			if (sqd->thread)
-				tctx = sqd->thread->io_uring;
+			tsk = sqpoll_task_locked(sqd);
+			if (tsk)
+				tctx = tsk->io_uring;
 		}
 	} else {
 		tctx = current->io_uring;

-- 
Jens Axboe