Re: [PATCH] Make cgroup_path() RCU-safe

[Li Zefan <lizf@cn.fujitsu.com>]

> +static void free_cgroup_rcu(struct rcu_head *obj)
> +{
> +	struct cgroup *cgrp = container_of(obj, struct cgroup, rcu_head);
> +	/*
> +	 * Drop the active superblock reference that we took when we
> +	 * created the cgroup
> +	 */
> +	deactivate_super(cgrp->root->sb);
> +
> +	kfree(cgrp);
> +}

I just wrote a different test program, and triggered kernel panic immediately:
	for ((; ;))
	{
		mount -t cgroup -o cpu xxx /mnt
		mkdir /mnt/0
		rmdir /mnt/0
		umount /mnt
	}

I know little about vfs internal, but it seems wrong to call deactivate_super()
here. I tried to call deactivate_super() before call_rcu(), and ran 2 test
programs simultaneously, and my box are working find.

> +
>  static void cgroup_diput(struct dentry *dentry, struct inode *inode)
>  {
>  	/* is dentry a directory ? if so, kfree() associated cgroup */
> @@ -620,11 +632,7 @@ static void cgroup_diput(struct dentry *dentry,
> struct inode *inode)
>  		cgrp->root->number_of_cgroups--;
>  		mutex_unlock(&cgroup_mutex);
> 
> -		/* Drop the active superblock reference that we took when we
> -		 * created the cgroup */
> -		deactivate_super(cgrp->root->sb);
> -
> -		kfree(cgrp);
> +		call_rcu(&cgrp->rcu_head, free_cgroup_rcu);
>  	}
>  	iput(inode);
>  }