netdev - Re: [Patch net 2/2] net_sched: always call ->destroy when ->init() fails

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]

Date:	Fri, 24 Oct 2014 17:36:09 -0700
From:	Cong Wang <cwang@...pensource.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	Cong Wang <xiyou.wangcong@...il.com>,
	netdev <netdev@...r.kernel.org>,
	David Miller <davem@...emloft.net>,
	Wang Bo <wang.bo116@....com.cn>,
	John Fastabend <john.r.fastabend@...el.com>,
	Eric Dumazet <edumazet@...gle.com>,
	Patrick McHardy <kaber@...sh.net>, Terry Lam <vtlam@...gle.com>
Subject: Re: [Patch net 2/2] net_sched: always call ->destroy when ->init() fails

On Fri, Oct 24, 2014 at 5:14 PM, Eric Dumazet <eric.dumazet@...il.com> wrote:
> On Fri, 2014-10-24 at 16:55 -0700, Cong Wang wrote:
>> In qdisc_create(), when ->init() exists and it fails, we should
>> call ->destroy() to clean up the potentially partially initialized
>> qdisc's. This will also make the ->init() implementation easier.
>>
>
> Why is this patch needed ?
>
> You are adding bugs, its not clear what bug you are fixing.


You missed what Wang Bo reported, we could double free
mq qdisc on failure path, one in mq_init() and one in qdisc_create_dflt().


>
> I really do not like the idea of ->init() relying on a ->destroy() to
> cleanup a failed ->init().
>
> This is not what most management functions do in our stack.
>
> I very much prefer that a function returning an error has no side
> effect, like if it hadnt be called at all.


If you would like to fix all qdisc_create_dflt() callers, yes, and good
luck with even more complex error handling there.

Read the diffstat, if we could fix a bug with less code, why not....

>
>> qdisc_create_dflt() already does that. Most callers of qdisc_create_dflt()
>> simply use noop_qdisc when it fails.
>>
>> And, most of the ->destroy() implementations are already able to clean up
>> partially initialization, we don't need to worry.
>>
>> The following ->init()'s need to catch up:
>> fq_codel_init(), hhf_init(), multiq_init(),  sfq_init(), mq_init(),
>> mqprio_init().
>>
>
> ...
>
>>
>>  static int multiq_dump(struct Qdisc *sch, struct sk_buff *skb)
>> diff --git a/net/sched/sch_qfq.c b/net/sched/sch_qfq.c
>> index 3ec7e88..55ac6a4 100644
>> --- a/net/sched/sch_qfq.c
>> +++ b/net/sched/sch_qfq.c
>> @@ -522,7 +522,6 @@ static int qfq_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
>>       return 0;
>>
>>  destroy_class:
>> -     qdisc_destroy(cl->qdisc);
>>       kfree(cl);
>>       return err;
>>  }
>
>
> Really ? I am calling this a leak of cl->qdisc.
>
>

How? I already added a comment on qdisc_create_dflt(), callers don't
need to clean up
on failure path.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html