NACK: [SRU][J][PATCH 5/5] cgroup: Reorganize css_set_lock and kernfs path processing

[Koichiro Den]

On Thu, Feb 13, 2025 at 02:55:53PM GMT, Philip Cox wrote:
> From: Michal Koutný <mkoutny at suse.com>
> 
> The commit 74e4b956eb1c incorrectly wrapped kernfs_walk_and_get
> (might_sleep) under css_set_lock (spinlock). css_set_lock is needed by
> __cset_cgroup_from_root to ensure stable cset->cgrp_links but not for
> kernfs_walk_and_get.
> 
> We only need to make sure that the returned root_cgrp won't be freed
> under us. This is given in the case of global root because it is static
> (cgrp_dfl_root.cgrp). When the root_cgrp is lower in the hierarchy, it
> is pinned by cgroup_ns->root_cset (and `current` task cannot switch
> namespace asynchronously so ns_proxy pins cgroup_ns).
> 
> Note this reasoning won't hold for root cgroups in v1 hierarchies,
> therefore create a special-cased helper function just for the default
> hierarchy.
> 
> Fixes: 74e4b956eb1c ("cgroup: Honor caller's cgroup NS when resolving path")
> Reported-by: Dan Carpenter <dan.carpenter at oracle.com>
> Signed-off-by: Michal Koutný <mkoutny at suse.com>
> Signed-off-by: Tejun Heo <tj at kernel.org>
> (backported from commit 46307fd6e27a3f678a1678b02e667678c22aa8cc)
> [philcox: context changes in cset_cgroup_from_root()]
> Signed-off-by: Philip Cox <philip.cox at canonical.com>
> ---
>  kernel/cgroup/cgroup.c | 39 +++++++++++++++++++++++++++------------
>  1 file changed, 27 insertions(+), 12 deletions(-)
> 
> diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
> index da2b61f60bfb8..142fd9bd6bc25 100644
> --- a/kernel/cgroup/cgroup.c
> +++ b/kernel/cgroup/cgroup.c
> @@ -1378,6 +1378,9 @@ static void cgroup_destroy_root(struct cgroup_root *root)
>  	cgroup_free_root(root);
>  }
>  
> +/*
> + * Returned cgroup is without refcount but it's valid as long as cset pins it.
> + */
>  static inline struct cgroup *__cset_cgroup_from_root(struct css_set *cset,
>  					    struct cgroup_root *root)
>  {
> @@ -1389,6 +1392,7 @@ static inline struct cgroup *__cset_cgroup_from_root(struct css_set *cset,
>  		res_cgroup = cset->dfl_cgrp;
>  	} else {
>  		struct cgrp_cset_link *link;
> +		lockdep_assert_held(&css_set_lock);
>  
>  		list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
>  			struct cgroup *c = link->cgrp;
> @@ -1400,6 +1404,7 @@ static inline struct cgroup *__cset_cgroup_from_root(struct css_set *cset,
>  		}
>  	}
>  
> +	BUG_ON(!res_cgroup);

This should not be included. res_cgroup can be NULL if cgroup_mutex is not
acquired.

Commit d23b5c577715 ("cgroup: Make operations on the cgroup root_list RCU
safe") was backported to Jammy via stable upstream updates (from
v5.15.168, LP: #2086242).

>  	return res_cgroup;
>  }
>  
> @@ -1425,18 +1430,32 @@ current_cgns_cgroup_from_root(struct cgroup_root *root)
>  	return res;
>  }
>  
> +/*
> + * Look up cgroup associated with current task's cgroup namespace on the default
> + * hierarchy.
> + *
> + * Unlike current_cgns_cgroup_from_root(), this doesn't need locks:
> + * - Internal rcu_read_lock is unnecessary because we don't dereference any rcu
> + *   pointers.
> + * - css_set_lock is not needed because we just read cset->dfl_cgrp.
> + * - As a bonus returned cgrp is pinned with the current because it cannot
> + *   switch cgroup_ns asynchronously.
> + */
> +static struct cgroup *current_cgns_cgroup_dfl(void)
> +{
> +	struct css_set *cset;
> +
> +	cset = current->nsproxy->cgroup_ns->root_cset;
> +	return __cset_cgroup_from_root(cset, &cgrp_dfl_root);
> +}
> +
>  /* look up cgroup associated with given css_set on the specified hierarchy */
>  static struct cgroup *cset_cgroup_from_root(struct css_set *cset,
>  					    struct cgroup_root *root)
>  {
> -	struct cgroup *res = NULL;
> -
>  	lockdep_assert_held(&css_set_lock);
>  
> -	res = __cset_cgroup_from_root(cset, root);
> -
> -	BUG_ON(!res);
> -	return res;
> +	return __cset_cgroup_from_root(cset, root);
>  }
>  
>  /*
> @@ -6057,9 +6076,7 @@ struct cgroup *cgroup_get_from_id(u64 id)
>  	if (!cgrp)
>  		return ERR_PTR(-ENOENT);
>  
> -	spin_lock_irq(&css_set_lock);
> -	root_cgrp = current_cgns_cgroup_from_root(&cgrp_dfl_root);
> -	spin_unlock_irq(&css_set_lock);
> +	root_cgrp = current_cgns_cgroup_dfl();
>  	if (!cgroup_is_descendant(cgrp, root_cgrp)) {
>  		cgroup_put(cgrp);
>  		return ERR_PTR(-ENOENT);
> @@ -6629,10 +6646,8 @@ struct cgroup *cgroup_get_from_path(const char *path)
>  	struct cgroup *cgrp = ERR_PTR(-ENOENT);
>  	struct cgroup *root_cgrp;
>  
> -	spin_lock_irq(&css_set_lock);
> -	root_cgrp = current_cgns_cgroup_from_root(&cgrp_dfl_root);
> +	root_cgrp = current_cgns_cgroup_dfl();
>  	kn = kernfs_walk_and_get(root_cgrp->kn, path);
> -	spin_unlock_irq(&css_set_lock);
>  	if (!kn)
>  		goto out;
>  
> -- 
> 2.43.0
> 
> 
> -- 
> kernel-team mailing list
> kernel-team at lists.ubuntu.com
> https://lists.ubuntu.com/mailman/listinfo/kernel-team