lkml.org 
[lkml]   [2020]   [Aug]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH] net: Fix potential memory leak in proto_register()
Date
Hi all:
David Miller <davem@davemloft.net> wrote:
>From: Cong Wang <xiyou.wangcong@gmail.com>
>Date: Tue, 11 Aug 2020 16:02:51 -0700
>
>>> @@ -3406,6 +3406,16 @@ static void sock_inuse_add(struct net *net,
>>> int val) } #endif
>>>
>>> +static void tw_prot_cleanup(struct timewait_sock_ops *twsk_prot) {
>>> + if (!twsk_prot)
>>> + return;
>>> + kfree(twsk_prot->twsk_slab_name);
>>> + twsk_prot->twsk_slab_name = NULL;
>>> + kmem_cache_destroy(twsk_prot->twsk_slab);
>>
>> Hmm, are you sure you can free the kmem cache name before
>> kmem_cache_destroy()? To me, it seems kmem_cache_destroy() frees the
>> name via slab_kmem_cache_release() via kfree_const().
>> With your patch, we have a double-free on the name?
>>
>> Or am I missing anything?
>
>Yep, there is a double free here.
>
>Please fix this.

Many thanks for both of you to point this issue out. But I'am not really understand, could you please explain it more?
As far as I can see, the double free path is:
1. kfree(twsk_prot->twsk_slab_name)
2. kmem_cache_destroy
--> shutdown_memcg_caches
--> shutdown_cache
--> slab_kmem_cache_release
--> kfree_const(s->name)
But twsk_prot->twsk_slab_name is allocated from kasprintf via kmalloc_track_caller while twsk_prot->twsk_slab->name is allocated
via kstrdup_const. So I think twsk_prot->twsk_slab_name and twsk_prot->twsk_slab->name point to different memory, and there is no double free.

Or am I missing anything?

By the way, req_prot_cleanup() do the same things, i.e. free the slab_name before involve kmem_cache_destroy(). If there is a double
free, so as here?

Thanks.

\
 
 \ /
  Last update: 2020-08-12 11:23    [W:0.034 / U:30.264 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site