netdev - Re: [PATCH] net-sched: add packet length pointer parameter for qdisc

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20080731154055.GA2532@ami.dom.local>
Date:	Thu, 31 Jul 2008 17:40:56 +0200
From:	Jarek Poplawski <jarkao2@...il.com>
To:	Jussi Kivilinna <jussi.kivilinna@...et.fi>
Cc:	David Miller <davem@...emloft.net>, netdev@...r.kernel.org,
	Patrick McHardy <kaber@...sh.net>
Subject: Re: [PATCH] net-sched: add packet length pointer parameter for
	qdisc_enqueue

On Thu, Jul 31, 2008 at 05:49:19PM +0300, Jussi Kivilinna wrote:
> Quoting "Jarek Poplawski" <jarkao2@...il.com>:
>
>> On Thu, Jul 31, 2008 at 12:04:57PM +0300, Jussi Kivilinna wrote:
>>> Quoting "Jarek Poplawski" <jarkao2@...il.com>:
>>>
>>>> On 30-07-2008 20:55, Jussi Kivilinna wrote:
>>>>> Pass packet length to caller through pointer so that length is
>>>>> available to caller even if inner qdisc frees skb.
>> ...
>>>> As I've written before, IMHO using an skb after enqueuing should be
>>>> avoided unless we refcounted it or the returned code is clear enough,
>>>> so the idea of this patch could be right to me. But, I guess, you are
>>>> changing here the way it's done: the size is calculated before the
>>>> current enqueing instead of after the last one.
>>>
>>> Doh, you're right. qdisc->enqueue should pass packet length too.
>>>
>>>>
>>>> BTW, since I don't really get this "stab" idea enough, I wonder how
>>>> it is expected to be used: with a top qdisc, a leaf one or some
>>>> summing?
>>>>
>>>
>>> With top and/or leaf qdisc, inner stab overrides upper.
>>
>> Probably this could be called a bit ugly, but another way could be
>> simply updating these .bstas.bytes in ->dequeue() with a difference
>> between calculated and skb->len?
>>
>
> That would work, but HFSC uses packet length in enqueue for something  
> else too:
> 	if (cl->qdisc->q.qlen == 1)
> 		set_active(cl, qdisc_pkt_len(skb));
>
> Could this bit be moved to dequeue?

I think, Patrick should look at this.

On the other hand, it seems reading such an skb after enqueuing with
NET_XMIT_SUCCESS could be quite safe after a patch I'm currently
working on, because there will be no more overloading in case of
TC_ACT_STOLEN etc. More problematic would be keeping stats exactly
after NET_XMIT_CN, but it's usually skipped now. So maybe it would be
better to wait with these changes a bit?

Jarek P.
--
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