linux-kernel - Re: [PATCH] block: don't decrement flush request refcount if it's state is idle in flush_end

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Sun, 8 Aug 2021 16:44:46 +0800
From:   Ming Lei <ming.lei@...hat.com>
To:     Yu Kuai <yukuai3@...wei.com>
Cc:     axboe@...nel.dk, linux-block@...r.kernel.org,
        linux-kernel@...r.kernel.org, yi.zhang@...wei.com
Subject: Re: [PATCH] block: don't decrement flush request refcount if it's
 state is idle in flush_end_io()

On Sun, Aug 08, 2021 at 03:03:30PM +0800, Yu Kuai wrote:
> flush_end_io() currently decrement request refcount unconditionally.
> However, it's possible that the request is already idle and it's
> refcount is zero since that flush_end_io() can be called concurrently.
> 
> For example, nbd_clear_que() can be called concurrently with normal
> io completion or io timeout.
> 
> Thus check idle before decrement to avoid refcount_t underflow
> warning.
> 
> Signed-off-by: Yu Kuai <yukuai3@...wei.com>
> ---
>  block/blk-flush.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/block/blk-flush.c b/block/blk-flush.c
> index 1002f6c58181..9b65dc43702c 100644
> --- a/block/blk-flush.c
> +++ b/block/blk-flush.c
> @@ -222,7 +222,8 @@ static void flush_end_io(struct request *flush_rq, blk_status_t error)
>  	/* release the tag's ownership to the req cloned from */
>  	spin_lock_irqsave(&fq->mq_flush_lock, flags);
>  
> -	if (!refcount_dec_and_test(&flush_rq->ref)) {
> +	if (blk_mq_rq_state(flush_rq) == MQ_RQ_IDLE ||
> +	    !refcount_dec_and_test(&flush_rq->ref)) {

It is driver's bug to finish one already ended request, so please
fix the driver instead.

Thanks,
Ming