netdev - Re: [PATCH net-next 0/3] sch_fq: segment too big GSO packets

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:	Tue, 25 Nov 2014 08:31:34 -0800
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Yang Yingliang <yangyingliang@...wei.com>
Cc:	netdev@...r.kernel.org, davem@...emloft.net
Subject: Re: [PATCH net-next 0/3] sch_fq: segment too big GSO packets

On Tue, 2014-11-25 at 21:24 +0800, Yang Yingliang wrote:
> As the TODO says: "maybe segment the too big skb, as in commit
> e43ac79a4bc ("sch_tbf: segment too big GSO packets")" in fq_dequeue(),
> this patchset segment the GSO packets that are too big.
> 
> Sometimes a GSO packet is too big at a low rate. This patchset check
> the packet before it's enqueued, if the GSO packet cost more than 125ms
> to send, it will be segmented, then enqueue the segments one by one.
> Because of the segment, the qlen may be bigger than limit in some condition.
> My way is that let the packet in if qlen is smaller than limit before
> it's segmented.

We extended the 125ms value to 750ms, back in July.

I will upstream this change, now it gave us good experimental results.

There is no point segmenting packets with 2 MSS, as fq quantum is 2 MSS
by default.

Your patches are the wrong way to handle this.

We have SO_MAX_PACING_RATE, and try to cook GSO packets of the right
size, instead of segmenting them later.


--
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