On Wed, Apr 20, 2022 at 10:01:36AM +0800, yukuai (C) wrote:
> 在 2022/04/20 9:46, Williams, Dan J 写道:
> > On Fri, 2022-03-18 at 21:01 +0800, Ming Lei wrote:
> > > In the whole lifetime of blkcg_gq instance, ->q will be referred, such
> > > as, ->pd_free_fn() is called in blkg_free, and throtl_pd_free() still
> > > may touch the request queue via &tg->service_queue.pending_timer which
> > > is handled by throtl_pending_timer_fn(), so it is reasonable to grab
> > > request queue's refcnt by blkcg_gq instance.
> > >
> > > Previously blkcg_exit_queue() is called from blk_release_queue, and it
> > > is hard to avoid the use-after-free. But recently commit 1059699f87eb ("block:
> > > move blkcg initialization/destroy into disk allocation/release handler")
> > > is merged to for-5.18/block, it becomes simple to fix the issue by simply
> > > grabbing request queue's refcnt.
> >
> > This patch, upstream as commit 0a9a25ca7843 ("block: let blkcg_gq grab
> > request queue's refcnt") causes the nvdimm unit tests to spam messages
> > like:
> >
> > [ 51.439133] debugfs: Directory 'pmem2' with parent 'block' already present!
> > [ 52.095679] debugfs: Directory 'pmem3' with parent 'block' already present!
> > [ 52.505613] block device autoloading is deprecated and will be removed.
> > [ 52.791693] debugfs: Directory 'pmem2' with parent 'block' already present!
> > [ 53.240314] debugfs: Directory 'pmem3' with parent 'block' already present!
> > [ 53.373472] debugfs: Directory 'pmem3' with parent 'block' already present!
> > [ 53.688876] nd_pmem btt2.0: No existing arenas
> > [ 53.773097] debugfs: Directory 'pmem2s' with parent 'block' already present!
> > [ 53.884493] debugfs: Directory 'pmem2s' with parent 'block' already present!
> > [ 54.042946] debugfs: Directory 'pmem2s' with parent 'block' already present!
> > [ 54.195954] debugfs: Directory 'pmem2s' with parent 'block' already present!
> > [ 54.383989] debugfs: Directory 'pmem2' with parent 'block' already present!
> > [ 54.577206] nd_pmem btt3.0: No existing arenas
> >
> Hi,
>
> I saw the same warning in our test, and I posted a patch to fix the
> problem:
>
> https://patchwork.kernel.org/project/linux-block/patch/20220415035607.1836495-1-yukuai3@xxxxxxxxxx/
>
> The root cause is not relate to the above commit, see details in
> the commit message.
Yeah, I am pretty sure the issue of 'debugfs: Directory 'XXXXX' with parent 'block' already
present!' is one block layer long-term issue.
Commit 0a9a25ca7843 ("block: let blkcg_gq grab request queue's refcnt") justs causes queue
released a bit late, then the issue becomes easier to trigger.
Thanks,
Ming
