linux-kernel - Re: [PATCHv3] mm/slab: fix race with dereferencing NULL ptr in alloc_calls

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20160203094420.GH21016@esperanza>
Date:	Wed, 3 Feb 2016 12:44:20 +0300
From:	Vladimir Davydov <vdavydov@...tuozzo.com>
To:	Dmitry Safonov <dsafonov@...tuozzo.com>
CC:	<akpm@...ux-foundation.org>, <cl@...ux.com>, <penberg@...nel.org>,
	<rientjes@...gle.com>, <iamjoonsoo.kim@....com>,
	<linux-mm@...ck.org>, <linux-kernel@...r.kernel.org>,
	<0x7f454c46@...il.com>
Subject: Re: [PATCHv3] mm/slab: fix race with dereferencing NULL ptr in
 alloc_calls_show

On Wed, Feb 03, 2016 at 10:52:13AM +0300, Dmitry Safonov wrote:
...
> diff --git a/mm/slab_common.c b/mm/slab_common.c
> index b50aef0..2bfc0b1 100644
> --- a/mm/slab_common.c
> +++ b/mm/slab_common.c
> @@ -451,6 +451,8 @@ EXPORT_SYMBOL(kmem_cache_create);
>  static int shutdown_cache(struct kmem_cache *s,
>  		struct list_head *release, bool *need_rcu_barrier)
>  {
> +	sysfs_slab_remove(s);
> +

shutdown_cache is called with slab_mutex held. slab_attr_store may take
the mutex. So placing sysfs_slab_remove here introduces a potential
deadlock.

>  	if (__kmem_cache_shutdown(s) != 0)
>  		return -EBUSY;
>  
> @@ -468,13 +470,8 @@ static void release_caches(struct list_head *release, bool need_rcu_barrier)
>  	if (need_rcu_barrier)
>  		rcu_barrier();
>  
> -	list_for_each_entry_safe(s, s2, release, list) {
> -#ifdef SLAB_SUPPORTS_SYSFS
> -		sysfs_slab_remove(s);
> -#else
> +	list_for_each_entry_safe(s, s2, release, list)
>  		slab_kmem_cache_release(s);
> -#endif
> -	}
>  }
>  
>  #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
> @@ -614,6 +611,7 @@ void memcg_destroy_kmem_caches(struct mem_cgroup *memcg)
>  	list_for_each_entry_safe(s, s2, &slab_caches, list) {
>  		if (is_root_cache(s) || s->memcg_params.memcg != memcg)
>  			continue;
> +

Please remove this hunk.

>  		/*
>  		 * The cgroup is about to be freed and therefore has no charges
>  		 * left. Hence, all its caches must be empty by now.
> diff --git a/mm/slub.c b/mm/slub.c
> index 2e1355a..b6a68b7 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -5296,11 +5296,6 @@ static void memcg_propagate_slab_attrs(struct kmem_cache *s)
>  #endif
>  }
>  
> -static void kmem_cache_release(struct kobject *k)
> -{
> -	slab_kmem_cache_release(to_slab(k));
> -}
> -
>  static const struct sysfs_ops slab_sysfs_ops = {
>  	.show = slab_attr_show,
>  	.store = slab_attr_store,
> @@ -5308,7 +5303,6 @@ static const struct sysfs_ops slab_sysfs_ops = {
>  
>  static struct kobj_type slab_ktype = {
>  	.sysfs_ops = &slab_sysfs_ops,
> -	.release = kmem_cache_release,

I surmise this will resurrect the bug that was fixed by 41a212859a4dd
("slub: use sysfs'es release mechanism for kmem_cache").

Thanks,
Vladimir