Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <e736ad09-e29d-7a76-6823-55e14fec87c1@suse.cz>
Date:   Mon, 19 Sep 2022 14:03:15 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.2.2
Subject: Re: [PATCH] mm/slab_common: fix possiable double free of kmem_cache
To:     Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc:     Feng Tang <feng.tang@intel.com>, Christoph Lameter <cl@linux.com>,
        Pekka Enberg <penberg@kernel.org>,
        David Rientjes <rientjes@google.com>,
        Joonsoo Kim <iamjoonsoo.kim@lge.com>,
        Roman Gushchin <roman.gushchin@linux.dev>,
        Andrew Morton <akpm@linux-foundation.org>,
        Waiman Long <longman@redhat.com>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, kasan-dev@googlegroups.com
References: <20220919031241.1358001-1-feng.tang@intel.com>
 <e38cc728-f5e5-86d1-d6a1-c3e99cc02239@suse.cz> <YyhY7RBLxCEuSHp9@hyeyoo>
Content-Language: en-US
From:   Vlastimil Babka <vbabka@suse.cz>
In-Reply-To: <YyhY7RBLxCEuSHp9@hyeyoo>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 9/19/22 13:56, Hyeonggon Yoo wrote:
> On Mon, Sep 19, 2022 at 11:12:38AM +0200, Vlastimil Babka wrote:
>> On 9/19/22 05:12, Feng Tang wrote:
>> > When doing slub_debug test, kfence's 'test_memcache_typesafe_by_rcu'
>> > kunit test case cause a use-after-free error:
>> >
> 
> If I'm not mistaken, I think the subject should be:
> s/double free/use after free/g

Well, it's both AFAICS. By the initial use-after-free we can read a wrong
s->flags that was modified since we freed for the first time, and it can
lead to another kmem_cache_release() which is basically a double free.

>> >   BUG: KASAN: use-after-free in kobject_del+0x14/0x30
>> >   Read of size 8 at addr ffff888007679090 by task kunit_try_catch/261
>> > 
>> >   CPU: 1 PID: 261 Comm: kunit_try_catch Tainted: G    B            N 6.0.0-rc5-next-20220916 #17
>> >   Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.15.0-1 04/01/2014
>> >   Call Trace:
>> >    <TASK>
>> >    dump_stack_lvl+0x34/0x48
>> >    print_address_description.constprop.0+0x87/0x2a5
>> >    print_report+0x103/0x1ed
>> >    kasan_report+0xb7/0x140
>> >    kobject_del+0x14/0x30
>> >    kmem_cache_destroy+0x130/0x170
>> >    test_exit+0x1a/0x30
>> >    kunit_try_run_case+0xad/0xc0
>> >    kunit_generic_run_threadfn_adapter+0x26/0x50
>> >    kthread+0x17b/0x1b0
>> >    </TASK>
>> > 
>> > The cause is inside kmem_cache_destroy():
>> > 
>> > kmem_cache_destroy
>> >     acquire lock/mutex
>> >     shutdown_cache
>> >         schedule_work(kmem_cache_release) (if RCU flag set)
>> >     release lock/mutex
>> >     kmem_cache_release (if RCU flag set)
>> 
>> 				      ^ not set
>> 
>> I've fixed that up.
>> 
>> > 
>> > in some certain timing, the scheduled work could be run before
>> > the next RCU flag checking which will get a wrong state.
>> > 
>> > Fix it by caching the RCU flag inside protected area, just like 'refcnt'
> 
> Very nice catch, thanks!
> 
> Otherwise (and with Vlastimil's fix):
> 
> Looks good to me.
> Reviewed-by: Hyeonggon Yoo <42.hyeyoo@gmail.com>
> 
>> > 
>> > Signed-off-by: Feng Tang <feng.tang@intel.com>
>> 
>> Thanks!
>> 
>> > ---
>> > 
>> > note:
>> > 
>> > The error only happens on linux-next tree, and not in Linus' tree,
>> > which already has Waiman's commit:
>> > 0495e337b703 ("mm/slab_common: Deleting kobject in kmem_cache_destroy()
>> > without holding slab_mutex/cpu_hotplug_lock")
>> 
>> Actually that commit is already in Linus' rc5 too, so I will send your fix
>> this week too. Added a Fixes: 0495e337b703 (...) too.
>> 
>> >  mm/slab_common.c | 5 ++++-
>> >  1 file changed, 4 insertions(+), 1 deletion(-)
>> > 
>> > diff --git a/mm/slab_common.c b/mm/slab_common.c
>> > index 07b948288f84..ccc02573588f 100644
>> > --- a/mm/slab_common.c
>> > +++ b/mm/slab_common.c
>> > @@ -475,6 +475,7 @@ void slab_kmem_cache_release(struct kmem_cache *s)
>> >  void kmem_cache_destroy(struct kmem_cache *s)
>> >  {
>> >  	int refcnt;
>> > +	bool rcu_set;
>> >  
>> >  	if (unlikely(!s) || !kasan_check_byte(s))
>> >  		return;
>> > @@ -482,6 +483,8 @@ void kmem_cache_destroy(struct kmem_cache *s)
>> >  	cpus_read_lock();
>> >  	mutex_lock(&slab_mutex);
>> >  
>> > +	rcu_set = s->flags & SLAB_TYPESAFE_BY_RCU;
>> > +
>> >  	refcnt = --s->refcount;
>> >  	if (refcnt)
>> >  		goto out_unlock;
>> > @@ -492,7 +495,7 @@ void kmem_cache_destroy(struct kmem_cache *s)
>> >  out_unlock:
>> >  	mutex_unlock(&slab_mutex);
>> >  	cpus_read_unlock();
>> > -	if (!refcnt && !(s->flags & SLAB_TYPESAFE_BY_RCU))
>> > +	if (!refcnt && !rcu_set)
>> >  		kmem_cache_release(s);
>> >  }
>> >  EXPORT_SYMBOL(kmem_cache_destroy);
>> 
>