linux-kernel - Re: [LKP] [lkp] [xfs] 68a9f5e700: aim7.jobs-per-min -13.6% regression

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: <87eg5w18iu.fsf@yhuang-mobile.sh.intel.com>
Date:	Wed, 10 Aug 2016 16:58:17 -0700
From:	"Huang\, Ying" <ying.huang@...el.com>
To:	Linus Torvalds <torvalds@...ux-foundation.org>
Cc:	Dave Chinner <david@...morbit.com>,
	Wu Fengguang <fengguang.wu@...el.com>,
	Bob Peterson <rpeterso@...hat.com>,
	LKML <linux-kernel@...r.kernel.org>, LKP <lkp@...org>,
	Christoph Hellwig <hch@....de>
Subject: Re: [LKP] [lkp] [xfs] 68a9f5e700: aim7.jobs-per-min -13.6% regression

Hi, Linus,

Linus Torvalds <torvalds@...ux-foundation.org> writes:

> On Wed, Aug 10, 2016 at 4:08 PM, Dave Chinner <david@...morbit.com> wrote:
>>
>> That, to me, says there's a change in lock contention behaviour in
>> the workload (which we know aim7 is good at exposing). i.e. the
>> iomap change shifted contention from a sleeping lock to a spinning
>> lock, or maybe we now trigger optimistic spinning behaviour on a
>> lock we previously didn't spin on at all.
>
> Hmm. Possibly. I reacted to the lower cpu load number, but yeah, I
> could easily imagine some locking primitive difference too.
>
>> We really need instruction level perf profiles to understand
>> this - I don't have a machine with this many cpu cores available
>> locally, so I'm not sure I'm going to be able to make any progress
>> tracking it down in the short term. Maybe the lkp team has more
>> in-depth cpu usage profiles they can share?
>
> Yeah, I've occasionally wanted to see some kind of "top-25 kernel
> functions in the profile" thing. That said, when the load isn't all
> that familiar, the profiles usually are not all that easy to make
> sense of either. But comparing the before and after state might give
> us clues.

I have started perf-profile data collection, will send out the
comparison result soon.

Best Regards,
Huang, Ying