public inbox for [email protected]
 help / color / mirror / Atom feed
From: Jens Axboe <[email protected]>
To: Pavel Begunkov <[email protected]>, [email protected]
Subject: Re: [PATCH 3/6] io_uring: fix racy overflow count reporting
Date: Thu, 30 Jul 2020 11:41:57 -0600	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

On 7/30/20 11:33 AM, Pavel Begunkov wrote:
> On 30/07/2020 20:18, Jens Axboe wrote:
>> On 7/30/20 9:43 AM, Pavel Begunkov wrote:
>>> All ->cq_overflow modifications should be under completion_lock,
>>> otherwise it can report a wrong number to the userspace. Fix it in
>>> io_uring_cancel_files().
>>>
>>> Signed-off-by: Pavel Begunkov <[email protected]>
>>> ---
>>>  fs/io_uring.c | 3 +--
>>>  1 file changed, 1 insertion(+), 2 deletions(-)
>>>
>>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>>> index 11f4ab87e08f..6e2322525da6 100644
>>> --- a/fs/io_uring.c
>>> +++ b/fs/io_uring.c
>>> @@ -7847,10 +7847,9 @@ static void io_uring_cancel_files(struct io_ring_ctx *ctx,
>>>  				clear_bit(0, &ctx->cq_check_overflow);
>>>  				ctx->rings->sq_flags &= ~IORING_SQ_CQ_OVERFLOW;
>>>  			}
>>> -			spin_unlock_irq(&ctx->completion_lock);
>>> -
>>>  			WRITE_ONCE(ctx->rings->cq_overflow,
>>>  				atomic_inc_return(&ctx->cached_cq_overflow));
>>> +			spin_unlock_irq(&ctx->completion_lock);
>>
>> Torn writes? Not sure I see what the issue here, can you expand?
> 
> No, just off-by-one(many). E.g.
> 
> let: cached_overflow = 0;
> 
>         CPU 1                   |               CPU 2
> ====================================================================
> t = ++cached_overflow // t == 1 |
>                                 | t2 = ++cached_overflow // t2 == 2
>                                 | WRITE_ONCE(cq_overflow, t2)
> WRITE_ONCE(cq_overflow, t1) 	|
> 
> 
> So, ctx->rings->cq_overflow == 1, but ctx->cached_cq_overflow == 2.
> A minor problem and easy to fix.

Ah yes, good point.

-- 
Jens Axboe


  reply	other threads:[~2020-07-30 17:42 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-30 15:43 [PATCH 0/6] 5.9 small cleanups/fixes Pavel Begunkov
2020-07-30 15:43 ` [PATCH 1/6] io_uring: de-unionise io_kiocb Pavel Begunkov
2020-07-30 15:43 ` [PATCH 2/6] io_uring: deduplicate __io_complete_rw() Pavel Begunkov
2020-07-30 15:43 ` [PATCH 3/6] io_uring: fix racy overflow count reporting Pavel Begunkov
2020-07-30 17:18   ` Jens Axboe
2020-07-30 17:33     ` Pavel Begunkov
2020-07-30 17:41       ` Jens Axboe [this message]
2020-07-30 15:43 ` [PATCH 4/6] io_uring: fix stalled deferred requests Pavel Begunkov
2020-07-30 15:43 ` [PATCH 5/6] io_uring: consolidate *_check_overflow accounting Pavel Begunkov
2020-07-30 15:43 ` [PATCH 6/6] io_uring: get rid of atomic FAA for cq_timeouts Pavel Begunkov
2020-07-30 18:08 ` [PATCH 0/6] 5.9 small cleanups/fixes Jens Axboe
2020-07-31 14:28   ` Pavel Begunkov
2020-07-31 15:10     ` Jens Axboe

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 \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /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