linux-kernel - RE: Regression in reading /proc/stat in the newer kernels with large SMP and NUMA configurations

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:	Mon, 17 Oct 2011 19:05:09 +0100
From:	"Seger, Mark" <mark.seger@...com>
To:	"Oberman, Laurence (HAS GSE)" <Laurence.Oberman@...com>,
	Eric Dumazet <eric.dumazet@...il.com>
CC:	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"Cabaniols, Sebastien" <Sebastien.Cabaniols@...com>
Subject: RE: Regression in reading /proc/stat in the newer kernels with
 large SMP and NUMA configurations

>Just FYI, the same behavior is seen on multiple hardware platforms so it
>is definitely kernel changes. The earlier kernel on the DL785 platform
>was returning reads within 60 microseconds.
>I have measured it on multiple platforms.
>
>In this case we have a DL980 with HT enabled so you will see 128 CPUS.
>If I shut off HT, it does not make a difference.

Given that there have only been a couple of responses to this I can't help but wonder if this hasn't really gotten anyone's attention yet or if perhaps I'm just overreacting.  I'd claim this breaks key Linux utilities by making them have a significantly heavier footprint than they used to have.  Historically people would always run top whenever they wanted to knowing it had minimal impact on the system and now I'm not so sure that will be the case anymore, at least not on big numa boxes.

I can assure you if someone wants to report cpu stats every tenth of a second it will more definitely have an impact.

-mark