[PATCH] io_uring: set -EINTR directly when a signal wakes up in io_cqring_wait

[PATCH] io_uring: set -EINTR directly when a signal wakes up in io_cqring_wait

[Jackie Liu]

We didn't use -ERESTARTSYS to tell the application layer to restart the
system call, but instead return -EINTR. we can set -EINTR directly when
wakeup by the signal, which can help us save an assignment operation and
comparison operation.

Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
---
 fs/io_uring.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index a30c4f6..63eee7e 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2878,7 +2878,7 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 		.to_wait	= min_events,
 	};
 	struct io_rings *rings = ctx->rings;
-	int ret;
+	int ret = 0;
 
 	if (io_cqring_events(rings) >= min_events)
 		return 0;
@@ -2896,7 +2896,6 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 			return ret;
 	}
 
-	ret = 0;
 	iowq.nr_timeouts = atomic_read(&ctx->cq_timeouts);
 	do {
 		prepare_to_wait_exclusive(&ctx->wait, &iowq.wq,
@@ -2905,15 +2904,13 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 			break;
 		schedule();
 		if (signal_pending(current)) {
-			ret = -ERESTARTSYS;
+			ret = -EINTR;
 			break;
 		}
 	} while (1);
 	finish_wait(&ctx->wait, &iowq.wq);
 
-	restore_saved_sigmask_unless(ret == -ERESTARTSYS);
-	if (ret == -ERESTARTSYS)
-		ret = -EINTR;
+	restore_saved_sigmask_unless(ret == -EINTR);
 
 	return READ_ONCE(rings->cq.head) == READ_ONCE(rings->cq.tail) ? ret : 0;
 }
-- 
2.7.4

Re: [PATCH] io_uring: set -EINTR directly when a signal wakes up in io_cqring_wait

[Bob Liu]

On 10/29/19 11:16 AM, Jackie Liu wrote:
> We didn't use -ERESTARTSYS to tell the application layer to restart the
> system call, but instead return -EINTR. we can set -EINTR directly when
> wakeup by the signal, which can help us save an assignment operation and
> comparison operation.
> 
> Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>

Looks good to me.
Reviewed-by: Bob Liu <bob.liu@oracle.com>

> ---
>  fs/io_uring.c | 9 +++------
>  1 file changed, 3 insertions(+), 6 deletions(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index a30c4f6..63eee7e 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -2878,7 +2878,7 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
>  		.to_wait	= min_events,
>  	};
>  	struct io_rings *rings = ctx->rings;
> -	int ret;
> +	int ret = 0;
>  
>  	if (io_cqring_events(rings) >= min_events)
>  		return 0;
> @@ -2896,7 +2896,6 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
>  			return ret;
>  	}
>  
> -	ret = 0;
>  	iowq.nr_timeouts = atomic_read(&ctx->cq_timeouts);
>  	do {
>  		prepare_to_wait_exclusive(&ctx->wait, &iowq.wq,
> @@ -2905,15 +2904,13 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
>  			break;
>  		schedule();
>  		if (signal_pending(current)) {
> -			ret = -ERESTARTSYS;
> +			ret = -EINTR;
>  			break;
>  		}
>  	} while (1);
>  	finish_wait(&ctx->wait, &iowq.wq);
>  
> -	restore_saved_sigmask_unless(ret == -ERESTARTSYS);
> -	if (ret == -ERESTARTSYS)
> -		ret = -EINTR;
> +	restore_saved_sigmask_unless(ret == -EINTR);
>  
>  	return READ_ONCE(rings->cq.head) == READ_ONCE(rings->cq.tail) ? ret : 0;
>  }
> 

Re: [PATCH] io_uring: set -EINTR directly when a signal wakes up in io_cqring_wait

[Jens Axboe]

On 10/28/19 9:16 PM, Jackie Liu wrote:
> We didn't use -ERESTARTSYS to tell the application layer to restart the
> system call, but instead return -EINTR. we can set -EINTR directly when
> wakeup by the signal, which can help us save an assignment operation and
> comparison operation.

Looks good - patch looks like it's against 5.3 though, I've queued it up
for 5.4 with a bit of mangling.

-- 
Jens Axboe