On Wed, Oct 30, 2024 at 03:46:14PM +0100, Christoph Hellwig wrote:
> On Wed, Oct 30, 2024 at 08:42:36PM +0800, Ming Lei wrote:
> > commit f1be1788a32e ("block: model freeze & enter queue as lock for
> > supporting lockdep") tries to apply lockdep for verifying freeze &
> > unfreeze. However, the verification is only done the outmost freeze and
> > unfreeze. This way is actually not correct because q->mq_freeze_depth
> > still may drop to zero on other task instead of the freeze owner task.
>
> Well, that's how non-owner functions work in general.
>
> > Fix this issue by always verifying the last unfreeze lock on the owner
> > task context, and freeze lock is still verified on the outmost one.
> >
> > Fixes: f1be1788a32e ("block: model freeze & enter queue as lock for supporting lockdep")
>
> What does this actually fix vs just improving coverage? Because the
> hacks in here look pretty horrible and I'd be much happier if we didn't
> have them.
task A task B
blk_mq_freeze_queue()
blk_mq_freeze_queue()
blk_mq_unfreeze_queue()
blk_mq_unfreeze_queue()
freeze_queue is verified on task A, but unfreeze_queue is verified on
task B, this way is definitely wrong.
This patch moves unfreeze_queue verification on task A.
Thanks,
Ming
