[PATCH] io_uring: add missing sigmask restore in io_cqring_wait()

[PATCH] io_uring: add missing sigmask restore in io_cqring_wait()

[Xiaoguang Wang]

Found this by learning codes.

Signed-off-by: Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
---
 fs/io_uring.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 30d959416eba..d7b0aeda1d84 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -7326,8 +7326,10 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 	if (uts) {
 		struct timespec64 ts;
 
-		if (get_timespec64(&ts, uts))
+		if (get_timespec64(&ts, uts)) {
+			restore_saved_sigmask_unless(false);
 			return -EFAULT;
+		}
 		timeout = timespec64_to_jiffies(&ts);
 	}
 
-- 
2.14.4.44.g2045bb6

Re: [PATCH] io_uring: add missing sigmask restore in io_cqring_wait()

[Jens Axboe]

On 9/14/21 2:41 AM, Xiaoguang Wang wrote:
> Found this by learning codes.

Does look like a real bug. But how about we move the get_timespec() section
before the sigmask saving?

-- 
Jens Axboe

Re: [PATCH] io_uring: add missing sigmask restore in io_cqring_wait()

[Xiaoguang Wang]

hi,


> On 9/14/21 2:41 AM, Xiaoguang Wang wrote:
>> Found this by learning codes.
> Does look like a real bug. But how about we move the get_timespec() section
> before the sigmask saving?

Ok,  I thought about this method before :)


Regards,

Xiaoguang Wang

>

Re: [PATCH] io_uring: add missing sigmask restore in io_cqring_wait()

[Jens Axboe]

On 9/14/21 8:28 AM, Xiaoguang Wang wrote:
> hi,
> 
> 
>> On 9/14/21 2:41 AM, Xiaoguang Wang wrote:
>>> Found this by learning codes.
>> Does look like a real bug. But how about we move the get_timespec() section
>> before the sigmask saving?
> 
> Ok,  I thought about this method before :)

When you send a v2 with that, please do also write a proper commit message.
It should explain why the change is needed. Thanks!

-- 
Jens Axboe