Re: [PATCH] io_uring: use io_poll_get_double in io_poll_double_wake

From: Pavel Begunkov <asml.silence@gmail.com>
To: Hao Xu <haoxu@linux.alibaba.com>, Jens Axboe <axboe@kernel.dk>
Cc: io-uring@vger.kernel.org, Joseph Qi <joseph.qi@linux.alibaba.com>
Subject: Re: [PATCH] io_uring: use io_poll_get_double in io_poll_double_wake
Date: Thu, 10 Jun 2021 09:59:43 +0100	[thread overview]
Message-ID: <3e3894ee-398d-02ba-c72c-f40ef121a724@gmail.com> (raw)
In-Reply-To: <1623314716-55024-1-git-send-email-haoxu@linux.alibaba.com>

On 6/10/21 9:45 AM, Hao Xu wrote:
> Correct wrong use of io_poll_get_single in io_poll_double_wake, which
> I think is a slip-up.

Not really. It removes both poll entries

> [   55.204528] WARNING: CPU: 0 PID: 2660 at fs/io_uring.c:1512 io_poll_double_wake+0x1d6/0x1f0
> [   55.204546] Modules linked in:
> [   55.204560] CPU: 0 PID: 2660 Comm: a.out Not tainted 5.13.0-rc3+ #1
> [   55.204575] RIP: 0010:io_poll_double_wake+0x1d6/0x1f0
> [   55.204584] Code: ff 48 89 eb e9 8b fe ff ff e8 c6 68 d3 ff 49 c7 44 24 08 00 00 00 00 48 8b 7b 08 e8 e4 7b 9d 00 e9 5d ff ff ff e8 aa 68 d3 ff <0f> 0b e9 76 ff ff ff e8 9e 68 d3 ff 0f 0b e9 59 ff ff ff 0f 1f 80
> [   55.204592] RSP: 0018:ffffc90003c73cc8 EFLAGS: 00010093
> [   55.204599] RAX: 0000000000000000 RBX: ffff88810fcf6500 RCX: 0000000000000000
> [   55.204604] RDX: ffff88810d38b500 RSI: ffffffff814d3dd6 RDI: ffff88810ff5fd60
> [   55.204610] RBP: ffff88810fcf6500 R08: ffff88810d38bf38 R09: 00000000fffffffe
> [   55.204615] R10: 00000000e2886200 R11: 000000005dbbc615 R12: ffff88810d068658
> [   55.204620] R13: 0000000000000000 R14: 0000000000000001 R15: 0000000040000000
> [   55.204625] FS:  00007f84d6170700(0000) GS:ffff88813bc00000(0000) knlGS:0000000000000000
> [   55.204635] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [   55.204640] CR2: 00007f84d616fef8 CR3: 000000010ffd2006 CR4: 00000000001706f0
> [   55.204646] Call Trace:
> [   55.204656]  __wake_up_common+0x9f/0x1b0
> [   55.204674]  __wake_up_common_lock+0x7a/0xc0
> [   55.204694]  tty_ldisc_lock+0x44/0x80
> [   55.204705]  tty_ldisc_hangup+0xe3/0x240
> [   55.204719]  __tty_hangup+0x26b/0x360
> [   55.204736]  tty_ioctl+0x6be/0xb20
> [   55.204747]  ? do_vfs_ioctl+0x1af/0xaa0
> [   55.204757]  ? __fget_files+0x15a/0x260
> [   55.204774]  ? tty_vhangup+0x20/0x20
> [   55.204787]  __x64_sys_ioctl+0xbb/0x100
> [   55.204801]  do_syscall_64+0x36/0x70
> [   55.204813]  entry_SYSCALL_64_after_hwframe+0x44/0xae
> 
> Reported-by: Abaci <abaci@linux.alibaba.com>
> Fixes: d4e7cd36a90e ("io_uring: sanitize double poll handling")
> Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
> ---
>  fs/io_uring.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index b2cc1e76d660..2be2db156094 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -4950,7 +4950,7 @@ static int io_poll_double_wake(struct wait_queue_entry *wait, unsigned mode,
>  			       int sync, void *key)
>  {
>  	struct io_kiocb *req = wait->private;
> -	struct io_poll_iocb *poll = io_poll_get_single(req);
> +	struct io_poll_iocb *poll = io_poll_get_double(req);
>  	__poll_t mask = key_to_poll(key);
>  
>  	/* for instances that support it check for an event match first: */
> 

-- 
Pavel Begunkov