netdev - Re: Strange latency spikes/TX network stalls on Sun Fire X4150(x86) and e1000e

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:	Wed, 6 Jun 2012 11:21:40 -0700
From:	Tom Herbert <therbert@...gle.com>
To:	Hiroaki SHIMODA <shimoda.hiroaki@...il.com>
Cc:	Jesse Brandeburg <jesse.brandeburg@...el.com>,
	Eric Dumazet <eric.dumazet@...il.com>,
	Denys Fedoryshchenko <denys@...p.net.lb>,
	netdev@...r.kernel.org, e1000-devel@...ts.sourceforge.net,
	jeffrey.t.kirsher@...el.com, davem@...emloft.net
Subject: Re: Strange latency spikes/TX network stalls on Sun Fire X4150(x86)
 and e1000e

I'm not exactly sure what the exact effect of WTHRESH is here.  Does
the device coalesce 5 completions regardless of size?  Would the
problem be avoided if bql limit_min were MTU, or could same issue be
hit with larger that 64 byte packets?

Tom

On Wed, Jun 6, 2012 at 10:19 AM, Hiroaki SHIMODA
<shimoda.hiroaki@...il.com> wrote:
> On Wed, 6 Jun 2012 09:26:35 -0700
> Jesse Brandeburg <jesse.brandeburg@...el.com> wrote:
>
>> On Wed, 6 Jun 2012 Hiroaki SHIMODA <shimoda.hiroaki@...il.com> wrote:
>> > Sorry for long delay. I'll post.
>> > (I have no idea how to fix this problem as keeping TXDCTL.WTHRESH to 5)
>>
>> I don't like changing WTHRESH wholesale because making the global change
>> to WTHRESH on e1000e just to fix this one bug (likely specific to a
>> particular chip/hardware) will adversely effect performance on many
>> models supported by e1000e not demonstrating any problem.  We could
>> possibly check something in for just ESB2LAN (S5000 chipset).
>>
>> Other people (tom herbert) with this same chipset have been unable to
>> reproduce this issue right?
>
> I understand your performance concern.
>
> The affected chip would be e1000_82571, e1000_82572, e1000_82574
> and e1000_80003es2lan which have FLAG2_DMA_BURST bit in
> adapter->flags2.
> Anyway, I have no objection intel guys NACK to my patch and
> provide right fix. But in that case please consider 82574L chip too
> which I observed similar behaviour.
>
> Thanks.
--
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