netdev - Re: [PATCH net 1/3] net/sched: cls_u32: No longer copy tcf

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: <39d1a433-d44f-1a90-e943-8e9f046bdf80@mojatatu.com>
Date: Fri, 21 Jul 2023 15:58:58 -0300
From: Pedro Tammela <pctammela@...atatu.com>
To: M A Ramdhan <ramdhan@...rlabs.sg>, valis <sec@...is.email>
Cc: netdev@...r.kernel.org, jhs@...atatu.com, xiyou.wangcong@...il.com,
 jiri@...nulli.us, davem@...emloft.net, edumazet@...gle.com, kuba@...nel.org,
 pabeni@...hat.com, victor@...atatu.com, billy@...rlabs.sg
Subject: Re: [PATCH net 1/3] net/sched: cls_u32: No longer copy tcf_result on
 update to avoid use-after-free

On 21/07/2023 15:04, M A Ramdhan wrote:
> On Sat, Jul 22, 2023 at 12:51 AM valis <sec@...is.email> wrote:
>>
>> When u32_change() is called on an existing filter, the whole
>> tcf_result struct is always copied into the new instance of the filter.
>>
>> This causes a problem when updating a filter bound to a class,
>> as tcf_unbind_filter() is always called on the old instance in the
>> success path, decreasing filter_cnt of the still referenced class
>> and allowing it to be deleted, leading to a use-after-free.
>>
>> Fix this by no longer copying the tcf_result struct from the old filter.
>>
>> Fixes: de5df63228fc ("net: sched: cls_u32 changes to knode must appear atomic to readers")
>> Reported-by: valis <sec@...is.email>
>> Reported-by: M A Ramdhan <ramdhan@...rlabs.sg>
>> Signed-off-by: valis <sec@...is.email>
>> Cc: stable@...r.kernel.org
>> ---
>>   net/sched/cls_u32.c | 1 -
>>   1 file changed, 1 deletion(-)
>>
>> diff --git a/net/sched/cls_u32.c b/net/sched/cls_u32.c
>> index 5abf31e432ca..19aa60d1eea7 100644
>> --- a/net/sched/cls_u32.c
>> +++ b/net/sched/cls_u32.c
>> @@ -826,7 +826,6 @@ static struct tc_u_knode *u32_init_knode(struct net *net, struct tcf_proto *tp,
>>
>>          new->ifindex = n->ifindex;
>>          new->fshift = n->fshift;
>> -       new->res = n->res;
>>          new->flags = n->flags;
>>          RCU_INIT_POINTER(new->ht_down, ht);
>>
>> --
>> 2.30.2
>>
> Hi,
> 
> We also thought it's also the correct fixes,
> but we're not sure because it will always remove the already bound
> qdisc class when we change the filter, even tho we never specify
> the new TCA_U32_CLASSID in the new filter.

The user should always explicitly tell the classid. Some other 
classifiers are already behaving like this, these were just wrong.

> 
> I also look at the implementation of cls_tcindex and cls_rsvp which still copy
> the old tcf_result, but don't call the tcf_unbind_filter when changing
> the filter.

Both were deprecated and removed as they were beyond saving.

> 
> If it's the intended behaviour, then I'm good with this patch.
> 
> Thanks & Regards,
> M A Ramdhan