io-uring.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Pavel Begunkov <asml.silence@gmail.com>
To: Joseph Qi <joseph.qi@linux.alibaba.com>,
	Jens Axboe <axboe@kernel.dk>, io-uring <io-uring@vger.kernel.org>
Cc: Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Subject: Re: [PATCH] io_uring: fix SQPOLL IORING_OP_CLOSE cancelation state
Date: Wed, 20 Jan 2021 02:25:23 +0000	[thread overview]
Message-ID: <6faaeee6-c8f1-b0c4-ef07-fc1bd8b9d87f@gmail.com> (raw)
In-Reply-To: <7c176c50-0f62-6753-eeef-bbb7a803febf@linux.alibaba.com>

On 20/01/2021 02:07, Joseph Qi wrote:
[...]
>> Fix this by moving the IO_WQ_WORK_NO_CANCEL until _after_ we've modified
>> the fdtable. Canceling before this point is totally fine, and running
>> it in the io-wq context _after_ that point is also fine.
>>
>> For 5.12, we'll handle this internally and get rid of the no-cancel
>> flag, as IORING_OP_CLOSE is the only user of it.
>>
>> Fixes: 14587a46646d ("io_uring: enable file table usage for SQPOLL rings")
> 
> As discussed with Pavel, this can not only happen in case sqpoll, but
> also in case async cancel is from io-wq.

Right, and it's handled because execution-during-cancellation can't
anymore get into close_fd_get_file(), either it was already done or
we didn't yet set IO_WQ_WORK_NO_CANCEL. And blkcg is not a problem.

> 
>> Reported-by: Joseph Qi <joseph.qi@linux.alibaba.com>
> 
> In fact, it is reported by "Abaci <abaci@linux.alibaba.com>"
> 
>> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> 
> Reviewed-and-tested-by: Joseph Qi <joseph.qi@linux.alibaba.com>

Thanks!

>>
>> ---
>>
>> Joseph, can you test this patch and see if this fixes it for you?
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index b76bb50f18c7..5f6f1e48954e 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -4472,7 +4472,6 @@ static int io_close_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
>>  	 * io_wq_work.flags, so initialize io_wq_work firstly.
>>  	 */
>>  	io_req_init_async(req);
>> -	req->work.flags |= IO_WQ_WORK_NO_CANCEL;
>>  
>>  	if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
>>  		return -EINVAL;
>> @@ -4505,6 +4504,8 @@ static int io_close(struct io_kiocb *req, bool force_nonblock,
>>  
>>  	/* if the file has a flush method, be safe and punt to async */
>>  	if (close->put_file->f_op->flush && force_nonblock) {
>> +		/* not safe to cancel at this point */
>> +		req->work.flags |= IO_WQ_WORK_NO_CANCEL;
>>  		/* was never set, but play safe */
>>  		req->flags &= ~REQ_F_NOWAIT;
>>  		/* avoid grabbing files - we don't need the files */
>>

-- 
Pavel Begunkov

      parent reply	other threads:[~2021-01-20  3:36 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-19 17:18 [PATCH] io_uring: fix SQPOLL IORING_OP_CLOSE cancelation state Jens Axboe
2021-01-20  2:07 ` Joseph Qi
2021-01-20  2:20   ` Jens Axboe
2021-01-20  2:25   ` Pavel Begunkov [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=6faaeee6-c8f1-b0c4-ef07-fc1bd8b9d87f@gmail.com \
    --to=asml.silence@gmail.com \
    --cc=axboe@kernel.dk \
    --cc=io-uring@vger.kernel.org \
    --cc=joseph.qi@linux.alibaba.com \
    --cc=xiaoguang.wang@linux.alibaba.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).