[PATCH] io_uring: fix skipping disabling sqo on exec

[PATCH] io_uring: fix skipping disabling sqo on exec

[Pavel Begunkov]

If there are no requests at the time __io_uring_task_cancel() is called,
tctx_inflight() returns zero and and it terminates not getting a chance
to go through __io_uring_files_cancel() and do
io_disable_sqo_submit(). And we absolutely want them disabled by the
time cancellation ends.

Also a fix potential false positive warning because of ctx->sq_data
check before io_disable_sqo_submit().

Reported-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---

To not grow diffstat now will be cleaned for-next

 fs/io_uring.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index d494c4269fc5..0d50845f1f3f 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -8937,10 +8937,12 @@ static void io_uring_cancel_task_requests(struct io_ring_ctx *ctx,
 {
 	struct task_struct *task = current;
 
-	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
+	if (ctx->flags & IORING_SETUP_SQPOLL) {
 		/* for SQPOLL only sqo_task has task notes */
 		WARN_ON_ONCE(ctx->sqo_task != current);
 		io_disable_sqo_submit(ctx);
+	}
+	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
 		task = ctx->sq_data->thread;
 		atomic_inc(&task->io_uring->in_idle);
 		io_sq_thread_park(ctx->sq_data);
@@ -9085,6 +9087,10 @@ void __io_uring_task_cancel(void)
 	/* make sure overflow events are dropped */
 	atomic_inc(&tctx->in_idle);
 
+	/* trigger io_disable_sqo_submit() */
+	if (tctx->sqpoll)
+		__io_uring_files_cancel(NULL);
+
 	do {
 		/* read completions before cancelations */
 		inflight = tctx_inflight(tctx);
-- 
2.24.0

Re: [PATCH] io_uring: fix skipping disabling sqo on exec

[Jens Axboe]

On 1/16/21 4:37 PM, Pavel Begunkov wrote:
> If there are no requests at the time __io_uring_task_cancel() is called,
> tctx_inflight() returns zero and and it terminates not getting a chance
> to go through __io_uring_files_cancel() and do
> io_disable_sqo_submit(). And we absolutely want them disabled by the
> time cancellation ends.
> 
> Also a fix potential false positive warning because of ctx->sq_data
> check before io_disable_sqo_submit().
> 
> Reported-by: Jens Axboe <axboe@kernel.dk>
> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
> ---
> 
> To not grow diffstat now will be cleaned for-next
> 
>  fs/io_uring.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index d494c4269fc5..0d50845f1f3f 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -8937,10 +8937,12 @@ static void io_uring_cancel_task_requests(struct io_ring_ctx *ctx,
>  {
>  	struct task_struct *task = current;
>  
> -	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
> +	if (ctx->flags & IORING_SETUP_SQPOLL) {
>  		/* for SQPOLL only sqo_task has task notes */
>  		WARN_ON_ONCE(ctx->sqo_task != current);
>  		io_disable_sqo_submit(ctx);
> +	}
> +	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
>  		task = ctx->sq_data->thread;
>  		atomic_inc(&task->io_uring->in_idle);
>  		io_sq_thread_park(ctx->sq_data);

Maybe just nest that inside?

	if (ctx->flags & IORING_SETUP_SQPOLL) {
  		/* for SQPOLL only sqo_task has task notes */
  		WARN_ON_ONCE(ctx->sqo_task != current);
  		io_disable_sqo_submit(ctx);
		if (ctx->sq_data) {
			...
		}
	}

That'd look a bit cleaner imho.
 
-- 
Jens Axboe

Re: [PATCH] io_uring: fix skipping disabling sqo on exec

[Pavel Begunkov]

On 17/01/2021 02:13, Jens Axboe wrote:
> On 1/16/21 4:37 PM, Pavel Begunkov wrote:
>> If there are no requests at the time __io_uring_task_cancel() is called,
>> tctx_inflight() returns zero and and it terminates not getting a chance
>> to go through __io_uring_files_cancel() and do
>> io_disable_sqo_submit(). And we absolutely want them disabled by the
>> time cancellation ends.
>>
>> Also a fix potential false positive warning because of ctx->sq_data
>> check before io_disable_sqo_submit().
>>
>> Reported-by: Jens Axboe <axboe@kernel.dk>
>> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
>> ---
>>
>> To not grow diffstat now will be cleaned for-next
>>
>>  fs/io_uring.c | 8 +++++++-
>>  1 file changed, 7 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index d494c4269fc5..0d50845f1f3f 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -8937,10 +8937,12 @@ static void io_uring_cancel_task_requests(struct io_ring_ctx *ctx,
>>  {
>>  	struct task_struct *task = current;
>>  
>> -	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
>> +	if (ctx->flags & IORING_SETUP_SQPOLL) {
>>  		/* for SQPOLL only sqo_task has task notes */
>>  		WARN_ON_ONCE(ctx->sqo_task != current);
>>  		io_disable_sqo_submit(ctx);
>> +	}
>> +	if ((ctx->flags & IORING_SETUP_SQPOLL) && ctx->sq_data) {
>>  		task = ctx->sq_data->thread;
>>  		atomic_inc(&task->io_uring->in_idle);
>>  		io_sq_thread_park(ctx->sq_data);
> 
> Maybe just nest that inside?
> 
> 	if (ctx->flags & IORING_SETUP_SQPOLL) {
>   		/* for SQPOLL only sqo_task has task notes */
>   		WARN_ON_ONCE(ctx->sqo_task != current);
>   		io_disable_sqo_submit(ctx);
> 		if (ctx->sq_data) {
> 			...
> 		}
> 	}
> 
> That'd look a bit cleaner imho.

second thought, it can't even happen because if not set we failed
during creation, and it has its own disable hooks.
sent v2 without that part

-- 
Pavel Begunkov