linux-kernel - Re: [PATCH 1/4] blk-mq-debugfs: warn about possible deadlock

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]

Message-ID: <aO4LGJ6I49dydw2J@fedora>
Date: Tue, 14 Oct 2025 16:34:32 +0800
From: Ming Lei <ming.lei@...hat.com>
To: Yu Kuai <yukuai1@...weicloud.com>
Cc: nilay@...ux.ibm.com, tj@...nel.org, josef@...icpanda.com,
	axboe@...nel.dk, cgroups@...r.kernel.org,
	linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
	yi.zhang@...wei.com, yangerkun@...wei.com, johnny.chenyi@...wei.com,
	"yukuai (C)" <yukuai3@...wei.com>
Subject: Re: [PATCH 1/4] blk-mq-debugfs: warn about possible deadlock

On Tue, Oct 14, 2025 at 04:21:30PM +0800, Yu Kuai wrote:
> Hi,
> 
> 在 2025/10/14 16:06, Ming Lei 写道:
> > On Tue, Oct 14, 2025 at 10:21:46AM +0800, Yu Kuai wrote:
> > > Creating new debugfs entries can trigger fs reclaim, hence we can't do
> > > this with queue freezed, meanwhile, other locks that can be held while
> > > queue is freezed should not be held as well.
> > > 
> > > Signed-off-by: Yu Kuai <yukuai3@...wei.com>
> > > ---
> > >   block/blk-mq-debugfs.c | 31 ++++++++++++++++++++++++-------
> > >   1 file changed, 24 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
> > > index 4896525b1c05..66864ed0b77f 100644
> > > --- a/block/blk-mq-debugfs.c
> > > +++ b/block/blk-mq-debugfs.c
> > > @@ -608,9 +608,23 @@ static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
> > >   	{},
> > >   };
> > > -static void debugfs_create_files(struct dentry *parent, void *data,
> > > +static void debugfs_create_files(struct request_queue *q, struct dentry *parent,
> > > +				 void *data,
> > >   				 const struct blk_mq_debugfs_attr *attr)
> > >   {
> > > +	/*
> > > +	 * Creating new debugfs entries with queue freezed has the rist of
> > > +	 * deadlock.
> > > +	 */
> > > +	WARN_ON_ONCE(q->mq_freeze_depth != 0);
> > > +	/*
> > > +	 * debugfs_mutex should not be nested under other locks that can be
> > > +	 * grabbed while queue is freezed.
> > > +	 */
> > > +	lockdep_assert_not_held(&q->elevator_lock);
> > > +	lockdep_assert_not_held(&q->rq_qos_mutex);
> > 
> > ->rq_qos_mutex use looks one real mess, in blk-cgroup.c, it is grabbed after
> > queue is frozen. However, inside block/blk-rq-qos.c, the two are re-ordered,
> > maybe we need to fix order between queue freeze and q->rq_qos_mutex first?
> > Or move on by removing the above line?
> 
> Yeah, I see this reoder as well, and I tried to fix this in the other
> thread for blkg configuration.
> 
> - queue is freezed by new helper blkg_conf_start(), and unfreezed after
>   blkg_conf_end(), rq_qos_add() is now called between them.
> 
> And for wbt, there are two cases:
>  - for blk-sysfs, queue is alredy freezed before rq_qos_add() as well;
>  - for wbt_enable_default(), this looks still problemaic, we should fix
>    the reorder seperatly.
> 
> Perhaps, should I fix this simple problem first, and then rebase the
> thread to convert queue_lock to blkcg_mtuex?

As I mentioned, if you want to move on with patchset first, the line of
`lockdep_assert_not_held(&q->rq_qos_mutex);` shouldn't be added until
->rq_qos_mutex vs. freeze queue order is finalized.


Thanks,
Ming