Re: [PATCH v2 1/3] blk-cgroup: fix race between policy activation and blkg destruction

[Zheng Qixing <zhengqixing@huawei.com>]

>> diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
>> index 3cffb68ba5d8..600f8c5843ea 100644
>> --- a/block/blk-cgroup.c
>> +++ b/block/blk-cgroup.c
>> @@ -1596,6 +1596,8 @@ int blkcg_activate_policy(struct gendisk *disk, const struct blkcg_policy *pol)
>>    
>>    	if (queue_is_mq(q))
>>    		memflags = blk_mq_freeze_queue(q);
>> +
>> +	mutex_lock(&q->blkcg_mutex);
>>    retry:
>>    	spin_lock_irq(&q->queue_lock);
>>    
>> @@ -1658,6 +1660,7 @@ int blkcg_activate_policy(struct gendisk *disk, const struct blkcg_policy *pol)
>>    
>>    	spin_unlock_irq(&q->queue_lock);
>>    out:
>> +	mutex_unlock(&q->blkcg_mutex);
>>    	if (queue_is_mq(q))
>>    		blk_mq_unfreeze_queue(q, memflags);
>>    	if (pinned_blkg)
> Can you also protect blkg_destroy_all() will blkcg_mutex as well? Then all access for q->blkg_list will
> be protected.
Why does blkg_destroy_all() also need blkcg_mutex?

After finishing ->pd_offline_fn() for blkgs and scheduling 
blkg_free_workfn() in blkg_destroy(),
blkg_destroy_all() clears the corresponding policy bit in q->blkcg_pols 
to avoid duplicate policy
teardown in blkcg_deactivate_policy().