linux-kernel - Re: SNB PCI root information

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

Date:	Fri, 22 Jun 2012 09:14:06 +0200
From:	Brice Goglin <brice.goglin@...il.com>
To:	Yinghai Lu <yinghai@...nel.org>
CC:	Bjorn Helgaas <bhelgaas@...gle.com>,
	Ulrich Drepper <drepper@...il.com>, jbarnes@...tuousgeek.org,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	lenb@...nel.org, x86@...nel.org, linux-pci@...r.kernel.org
Subject: Re: SNB PCI root information

Le 21/06/2012 21:24, Yinghai Lu a écrit :
> On Wed, Jun 20, 2012 at 10:56 PM, Brice Goglin <brice.goglin@...il.com> wrote:
>>> I suspected that could be other problem. intel system before
>>> sandbridge does not have IIO.
>>> they will have ioh instead, and one ioh would connect to two cpu sockets.
>>> but _PXM for root bus in dsdt only can return one value. Aka it is
>>> acpi spec limitation.
>> The machines I am talking about have two IOHs (one connected to each
>> socket). I only have an old 2.6.27 boot log at hand, I'll ask my admins
>> for more. Do you want the 2.6.27 dmesg anyway?
>>
> log from 2.6.27 should be good enough for this.
>
> BIOS _PXM return 0 for both pci root buses?
>

Here's dmesg.
I can't say for sure whether _PXM returns 0 since I don't know how to
read all this :) But Linux puts the first socket cpumap in
/sys/bus/pci/devices/*/local_cpus (and that's wrong according to the
motherboard manual and according to the performance we see).

Brice


View attachment "dmesg.txt" of type "text/plain" (195498 bytes)