Open MPI logo

Hardware Locality Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Hardware Locality Development mailing list

Subject: Re: [hwloc-devel] [hwloc-users] hwloc 1.5, freebsd and linux output on the same hardware
From: Sebastian Kuzminsky (seb_at_[hidden])
Date: 2012-10-11 12:33:31


The modification of the default cpuset (to exclude CPUs 0-9) is done by our
local installation. I assume the vanilla FreeBSD 7.3 that we're based on
does not monkey with the default cpuset.

I believe modifying your own cpuset is not a privileged operation in
FreeBSD. The cpuset executable is not setuid root and is usable by regular
users.

I think you can move the lstopo process to cpuset 0 (which includes all
CPUs) by calling cpuset_setid(), and that should take care of the issue.

On Thu, Oct 11, 2012 at 10:13 AM, Brice Goglin <Brice.Goglin_at_[hidden]>wrote:

> (forwarding your mail to the list, and replying)
>
> Good to know that it works, thanks for testing. I released 1.5.1rc1 today,
> it should work fine as well.
>
> You say that the cpuset does not contain 0-9 by default on this hardware.
> This is something specific to your installation, I guess? Nothing that
> happens by default on a default FreeBSD install, right?
>
> Is the cpuset-modification a root-only operation on FreeBSD? If so lstopo
> wouldn't be able to expand the cpuset at startup.
>
> lstopo has a --whole-system option to ignore such limitations.
> Unfortunately the x86 backend that hwloc uses on FreeBSD requires that we
> bind to each individual core to get its locality information, so that won't
> help unless lstopo can indeed remove the cpuset first.
>
> Brice
>
>
>
> -------- Message original -------- Sujet: Re: [hwloc-users] hwloc 1.5,
> freebsd and linux output on the same hardware Date : Thu, 11 Oct 2012
> 10:02:55 -0600 De : Sebastian Kuzminsky <seb_at_[hidden]><seb_at_[hidden]> Pour :
> Brice Goglin <Brice.Goglin_at_[hidden]> <Brice.Goglin_at_[hidden]>
>
>
> Ok, it's our fault, sort of.
>
> We use cpusets, and by default on this hardware CPUs 0-9 are denied to
> most processes (including lstopo). If I explicitly change the cpuset of
> lstopo to include CPUs 0-9, it runs correctly.
>
> Maybe lstopo should expand its cpuset to be fully inclusive at startup?
> I'll be happy to test patches if you want.
>
>
>
>
> On Thu, Oct 11, 2012 at 9:27 AM, Sebastian Kuzminsky <
> seb_at_[hidden]> wrote:
>
>> On Thu, Oct 11, 2012 at 2:39 AM, Brice Goglin <Brice.Goglin_at_[hidden]>wrote:
>>
>>> Le 06/10/2012 01:03, Sebastian Kuzminsky a écrit :
>>> > Hm. It may be that we're doing something funny and reserving those
>>> > CPUs. I'll run some tests on Monday and get back to you.
>>>
>>> (replying outside of the hwloc mailing list)
>>>
>>> Did you have a chance to look at this?
>>> I am about to release a hwloc 1.5.1rc1. It'd be good to get your
>>> feedback before we do the final 1.5.1 (likely next week).
>>>
>>
>> Oops, thanks for the reminder... I got side tracked on other stuff,
>> but I have some time this morning to devote to this. I'll get back to you
>> soon.
>>
>>
>> --
>> Sebastian Kuzminsky
>> Sr Software Engineer, Linerate Systems
>>
>>
>
>
> --
> Sebastian Kuzminsky
> Sr Software Engineer, Linerate Systems
>
>

-- 
Sebastian Kuzminsky
Sr Software Engineer, Linerate Systems