[PATCH v2] block: Skip destroyed blkg when restart in blkg_destroy_all()

[PATCH v2] block: Skip destroyed blkg when restart in blkg_destroy_all()

[Tao Su]

Kernel hang in blkg_destroy_all() when total blkg greater than
BLKG_DESTROY_BATCH_SIZE, because of not removing destroyed blkg in
blkg_list. So the size of blkg_list is same after destroying a
batch of blkg, and the infinite 'restart' occurs.

Since blkg should stay on the queue list until blkg_free_workfn(),
skip destroyed blkg when restart a new round, which will solve this
kernel hang issue and satisfy the previous will to restart.

Reported-by: Xiangfei Ma <xiangfeix.ma@intel.com>
Tested-by: Xiangfei Ma <xiangfeix.ma@intel.com>
Tested-by: Farrah Chen <farrah.chen@intel.com>
Signed-off-by: Yu Kuai <yukuai1@huaweicloud.com>
Signed-off-by: Tao Su <tao1.su@linux.intel.com>
---
v2:
- change 'directly remove destroyed blkg' to 'skip destroyed blkg'

v1:
- https://lore.kernel.org/all/20230425075911.839539-1-tao1.su@linux.intel.com/

 block/blk-cgroup.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index bd50b55bdb61..75bad5d60c9f 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -528,6 +528,9 @@ static void blkg_destroy_all(struct gendisk *disk)
 	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
 		struct blkcg *blkcg = blkg->blkcg;
 
+		if (hlist_unhashed(&blkg->blkcg_node))
+			continue;
+
 		spin_lock(&blkcg->lock);
 		blkg_destroy(blkg);
 		spin_unlock(&blkcg->lock);
-- 
2.34.1

Re: [PATCH v2] block: Skip destroyed blkg when restart in blkg_destroy_all()

[Yu Kuai]

Hi,

在 2023/04/28 12:51, Tao Su 写道:
> Kernel hang in blkg_destroy_all() when total blkg greater than
> BLKG_DESTROY_BATCH_SIZE, because of not removing destroyed blkg in
> blkg_list. So the size of blkg_list is same after destroying a
> batch of blkg, and the infinite 'restart' occurs.
> 
> Since blkg should stay on the queue list until blkg_free_workfn(),
> skip destroyed blkg when restart a new round, which will solve this
> kernel hang issue and satisfy the previous will to restart.

Please add a fix tag:

Fixes: f1c006f1c685 ("blk-cgroup: synchronize pd_free_fn() from 
blkg_free_workfn() and blkcg_deactivate_policy()")
> 
> Reported-by: Xiangfei Ma <xiangfeix.ma@intel.com>
> Tested-by: Xiangfei Ma <xiangfeix.ma@intel.com>
> Tested-by: Farrah Chen <farrah.chen@intel.com>
> Signed-off-by: Yu Kuai <yukuai1@huaweicloud.com>

You can remove this tag, and feel free to add:

Suggested-and-reviewed-by: Yu Kuai <yukuai3@huawei.com>

Thanks,
Kuai
> Signed-off-by: Tao Su <tao1.su@linux.intel.com>
> ---
> v2:
> - change 'directly remove destroyed blkg' to 'skip destroyed blkg'
> 
> v1:
> - https://lore.kernel.org/all/20230425075911.839539-1-tao1.su@linux.intel.com/
> 
>   block/blk-cgroup.c | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
> index bd50b55bdb61..75bad5d60c9f 100644
> --- a/block/blk-cgroup.c
> +++ b/block/blk-cgroup.c
> @@ -528,6 +528,9 @@ static void blkg_destroy_all(struct gendisk *disk)
>   	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
>   		struct blkcg *blkcg = blkg->blkcg;
>   
> +		if (hlist_unhashed(&blkg->blkcg_node))
> +			continue;
> +
>   		spin_lock(&blkcg->lock);
>   		blkg_destroy(blkg);
>   		spin_unlock(&blkcg->lock);
> 

Re: [PATCH v2] block: Skip destroyed blkg when restart in blkg_destroy_all()

[Tao Su]

On Fri, Apr 28, 2023 at 02:12:06PM +0800, Yu Kuai wrote:
> Hi,
> 
> 在 2023/04/28 12:51, Tao Su 写道:
> > Kernel hang in blkg_destroy_all() when total blkg greater than
> > BLKG_DESTROY_BATCH_SIZE, because of not removing destroyed blkg in
> > blkg_list. So the size of blkg_list is same after destroying a
> > batch of blkg, and the infinite 'restart' occurs.
> > 
> > Since blkg should stay on the queue list until blkg_free_workfn(),
> > skip destroyed blkg when restart a new round, which will solve this
> > kernel hang issue and satisfy the previous will to restart.
> 
> Please add a fix tag:
> 
> Fixes: f1c006f1c685 ("blk-cgroup: synchronize pd_free_fn() from
> blkg_free_workfn() and blkcg_deactivate_policy()")
> > 
> > Reported-by: Xiangfei Ma <xiangfeix.ma@intel.com>
> > Tested-by: Xiangfei Ma <xiangfeix.ma@intel.com>
> > Tested-by: Farrah Chen <farrah.chen@intel.com>
> > Signed-off-by: Yu Kuai <yukuai1@huaweicloud.com>
> 
> You can remove this tag, and feel free to add:
> 
> Suggested-and-reviewed-by: Yu Kuai <yukuai3@huawei.com>

Got it, thanks!

Tao

> 
> Thanks,
> Kuai
> > Signed-off-by: Tao Su <tao1.su@linux.intel.com>
> > ---
> > v2:
> > - change 'directly remove destroyed blkg' to 'skip destroyed blkg'
> > 
> > v1:
> > - https://lore.kernel.org/all/20230425075911.839539-1-tao1.su@linux.intel.com/
> > 
> >   block/blk-cgroup.c | 3 +++
> >   1 file changed, 3 insertions(+)
> > 
> > diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
> > index bd50b55bdb61..75bad5d60c9f 100644
> > --- a/block/blk-cgroup.c
> > +++ b/block/blk-cgroup.c
> > @@ -528,6 +528,9 @@ static void blkg_destroy_all(struct gendisk *disk)
> >   	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
> >   		struct blkcg *blkcg = blkg->blkcg;
> > +		if (hlist_unhashed(&blkg->blkcg_node))
> > +			continue;
> > +
> >   		spin_lock(&blkcg->lock);
> >   		blkg_destroy(blkg);
> >   		spin_unlock(&blkcg->lock);
> > 
> 

Re: [PATCH v2] block: Skip destroyed blkg when restart in blkg_destroy_all()

[Jens Axboe]

On Fri, 28 Apr 2023 12:51:49 +0800, Tao Su wrote:
> Kernel hang in blkg_destroy_all() when total blkg greater than
> BLKG_DESTROY_BATCH_SIZE, because of not removing destroyed blkg in
> blkg_list. So the size of blkg_list is same after destroying a
> batch of blkg, and the infinite 'restart' occurs.
> 
> Since blkg should stay on the queue list until blkg_free_workfn(),
> skip destroyed blkg when restart a new round, which will solve this
> kernel hang issue and satisfy the previous will to restart.
> 
> [...]

Applied, thanks!

[1/1] block: Skip destroyed blkg when restart in blkg_destroy_all()
      commit: 8176080d59e6d4ff9fc97ae534063073b4f7a715

Best regards,
-- 
Jens Axboe

Re: [PATCH v2] block: Skip destroyed blkg when restart in blkg_destroy_all()

[Tao Su]

On Fri, Apr 28, 2023 at 11:24:20AM -0600, Jens Axboe wrote:
> 
> On Fri, 28 Apr 2023 12:51:49 +0800, Tao Su wrote:
> > Kernel hang in blkg_destroy_all() when total blkg greater than
> > BLKG_DESTROY_BATCH_SIZE, because of not removing destroyed blkg in
> > blkg_list. So the size of blkg_list is same after destroying a
> > batch of blkg, and the infinite 'restart' occurs.
> > 
> > Since blkg should stay on the queue list until blkg_free_workfn(),
> > skip destroyed blkg when restart a new round, which will solve this
> > kernel hang issue and satisfy the previous will to restart.
> > 
> > [...]
> 
> Applied, thanks!

Axboe, thanks!

Tao

> 
> [1/1] block: Skip destroyed blkg when restart in blkg_destroy_all()
>       commit: 8176080d59e6d4ff9fc97ae534063073b4f7a715
> 
> Best regards,
> -- 
> Jens Axboe
> 
> 
>