mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Ming Lei <tom.leiming@gmail.com>
To: Qiliang Yuan <odys.yuan@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>,
	linux-block@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] ublk: drop the device reference outside ublk_ctl_mutex in DEL_DEV
Date: Thu, 1 Oct 2026 09:34:53 -0500	[thread overview]
Message-ID: <ar5vjRR_I4Vkd9h5@fedora-laptop> (raw)
In-Reply-To: <20261001-bug-ublk-del-dev-put-unlocked-v1-1-18e9bb9be727@gmail.com>

On Thu, Oct 01, 2026 at 11:10:47AM +0800, Qiliang Yuan wrote:
> ublk_ctrl_del_dev() drops its reference to the device while holding the
> global ublk_ctl_mutex. When this is the last reference, ublk_cdev_rel()
> frees the tag set, and blk_mq_free_tag_set() waits for the pending SRCU
> callbacks of the tag set with srcu_barrier(). Every DEL_DEV thus waits
> for an SRCU grace period with the mutex held, and deleting devices from
> several threads serializes on it.
> 
> The release path doesn't need ublk_ctl_mutex. The device number is freed
> under ublk_idr_lock, and the last reference is already dropped without
> the mutex when the ublk server still has the char device open at
> DEL_DEV time, from ublk_ch_release_work_fn().
> 
> Drop the reference after unlocking ublk_ctl_mutex.
> 
> ublk null target, 4000 single-queue devices, STOP_DEV + DEL_DEV issued
> from N threads, 16 vCPU KVM guest:
> 
>   threads     before      after
>         1     77.6 s     77.6 s
>         4     20.1 s     19.4 s
>         8     19.3 s      9.0 s
>        16     19.2 s      4.1 s
> 
> Signed-off-by: Qiliang Yuan <odys.yuan@gmail.com>
> ---
>  drivers/block/ublk_drv.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
> index 66eb55e7162e5..5d237aae51b7b 100644
> --- a/drivers/block/ublk_drv.c
> +++ b/drivers/block/ublk_drv.c
> @@ -4901,10 +4901,15 @@ static int ublk_ctrl_del_dev(struct ublk_device **p_ub, bool wait)
>  		set_bit(UB_STATE_DELETED, &ub->state);
>  	}
>  
> -	/* Mark the reference as consumed */
> +	mutex_unlock(&ublk_ctl_mutex);
> +
> +	/*
> +	 * Drop the reference outside ublk_ctl_mutex: if it is the last one,
> +	 * the release frees the tag set, which waits for an SRCU grace period,
> +	 * and holding the mutex would serialize concurrent deletions on it.
> +	 */
>  	*p_ub = NULL;
>  	ublk_put_device(ub);
> -	mutex_unlock(&ublk_ctl_mutex);

Looks good,

Reviewed-by: Ming Lei <tom.leiming@gmail.com>


Thanks,
Ming

  reply	other threads:[~2026-10-01 14:35 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-01  3:10 Qiliang Yuan
2026-10-01 14:34 ` Ming Lei [this message]
2026-10-01 14:43 ` 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 \
    --in-reply-to=ar5vjRR_I4Vkd9h5@fedora-laptop \
    --to=tom.leiming@gmail.com \
    --cc=axboe@kernel.dk \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=odys.yuan@gmail.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

all inboxes | Powered by JetHome®