On Tue, Jun 15, 2021 at 09:58:03PM +0800, Ming Lei wrote:
> On Tue, Jun 15, 2021 at 7:52 PM Christoph Hellwig <hch@xxxxxx> wrote:
> >
> > On Mon, Jun 14, 2021 at 02:40:58PM +0200, Bruno Goncalves wrote:
> > > Hi,
> > >
> > > We've noticed a kernel oops during the stress-ng test on aarch64 more log
> > > details on [1]. Christoph, do you think this could be related to the recent
> > > blk_cleanup_disk changes [2]?
> >
> > It doesn't really look very related. Any chance you could bisect it?
>
> It should be the wrong order between freeing tagset and cleanup disk:
>
> static void loop_remove(struct loop_device *lo)
> {
> ...
> blk_mq_free_tag_set(&lo->tag_set);
> blk_cleanup_disk(lo->lo_disk);
> ...
> }
Indeed. Something like this should fix the issue:
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index 9a48b3f9a15c..e0c4de392eab 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -2172,8 +2172,8 @@ static int loop_add(struct loop_device **l, int i)
static void loop_remove(struct loop_device *lo)
{
del_gendisk(lo->lo_disk);
- blk_mq_free_tag_set(&lo->tag_set);
blk_cleanup_disk(lo->lo_disk);
+ blk_mq_free_tag_set(&lo->tag_set);
mutex_destroy(&lo->lo_mutex);
kfree(lo);
}
