v4:
  - Update comment and commit logs for both patches.

 v3:
  - Drop v2 patch 2 as it may not be needed.
  - Replace css_tryget() with percpu_ref_is_zero() in patch 1 as
    suggested by Tejun.
  - Expand comment on patch 2 to elaborate the reason for this patch.

 v2:
  - Remove unnecessary rcu_read_{lock|unlock} from
    cgroup_rstat_css_cpu_flush() in patch 3.

It was found that blkcg_destroy_blkgs() may be called with all blkcg
references gone. This may potentially cause user-after-free and so should
be fixed. The second patch flushes rstat when calling blkcg_destroy_blkgs().

Waiman Long (2):
  bdi, blk-cgroup: Fix potential UAF of blkcg
  blk-cgroup: Flush stats at blkgs destruction path

 block/blk-cgroup.c     | 23 +++++++++++++++++++++++
 include/linux/cgroup.h |  1 +
 kernel/cgroup/rstat.c  | 18 ++++++++++++++++++
 mm/backing-dev.c       |  8 ++++++--
 4 files changed, 48 insertions(+), 2 deletions(-)

-- 
2.31.1

On 12/14/22 22:31, Waiman Long wrote:
>   v4:
>    - Update comment and commit logs for both patches.
>
>   v3:
>    - Drop v2 patch 2 as it may not be needed.
>    - Replace css_tryget() with percpu_ref_is_zero() in patch 1 as
>      suggested by Tejun.
>    - Expand comment on patch 2 to elaborate the reason for this patch.
>
>   v2:
>    - Remove unnecessary rcu_read_{lock|unlock} from
>      cgroup_rstat_css_cpu_flush() in patch 3.
>
> It was found that blkcg_destroy_blkgs() may be called with all blkcg
> references gone. This may potentially cause user-after-free and so should
> be fixed. The second patch flushes rstat when calling blkcg_destroy_blkgs().
>
> Waiman Long (2):
>    bdi, blk-cgroup: Fix potential UAF of blkcg
>    blk-cgroup: Flush stats at blkgs destruction path
>
>   block/blk-cgroup.c     | 23 +++++++++++++++++++++++
>   include/linux/cgroup.h |  1 +
>   kernel/cgroup/rstat.c  | 18 ++++++++++++++++++
>   mm/backing-dev.c       |  8 ++++++--
>   4 files changed, 48 insertions(+), 2 deletions(-)
>
Ping!

Any comments on these patches.

Thanks,
Longman