[9fans] multi-processor support

[9fans] multi-processor support

[Deepak Chawla]

[-- Attachment #1: Type: text/plain, Size: 173 bytes --]

My Vbox VM has 2 CPUs, but /dev/sysstat only has one line entry for CPU ID
0. I'm running Erik's 9atom on the VM. Do I need to run anything to start
the other processor?

[-- Attachment #2: Type: text/html, Size: 200 bytes --]

Re: [9fans] multi-processor support

[erik quanstrom]

> My Vbox VM has 2 CPUs, but /dev/sysstat only has one line entry for CPU ID
> 0. I'm running Erik's 9atom on the VM. Do I need to run anything to start
> the other processor?

does your plan9.ini have *nomp=1?

- erik

Re: [9fans] multi-processor support

[Deepak Chawla]

[-- Attachment #1: Type: text/plain, Size: 422 bytes --]

I do. I'll set it to 0.

BTW, whats with the * in the beginning of some of the names?


On Wed, Feb 20, 2013 at 11:27 PM, erik quanstrom <quanstro@quanstro.net>wrote:

> > My Vbox VM has 2 CPUs, but /dev/sysstat only has one line entry for CPU
> ID
> > 0. I'm running Erik's 9atom on the VM. Do I need to run anything to start
> > the other processor?
>
> does your plan9.ini have *nomp=1?
>
> - erik
>
>

[-- Attachment #2: Type: text/html, Size: 860 bytes --]

Re: [9fans] multi-processor support

[cinap_lenrek]

* denotes that the key is for the kernel and will not show
up in the enviroment. sometimes, theres a difference between
a key having the value 0 and a key not being present. often,
the kernel will just check for the presence of a key and ignore
its value. but this is not the case for *nomp :)

--
cinap

Re: [9fans] multi-processor support

[Deepak Chawla]

[-- Attachment #1.1: Type: text/plain, Size: 671 bytes --]

I tried *nomp=0, but still only saw one CPU. I removed the line altogether,
now the kernel panics (see attachment). Any way to override the kernel
params in plan9.ini?

Also, is there a way to mount the plan9 internal partitions (9fat, fossil
etc) on Linux?




On Thu, Feb 21, 2013 at 12:46 AM, <cinap_lenrek@gmx.de> wrote:

> * denotes that the key is for the kernel and will not show
> up in the enviroment. sometimes, theres a difference between
> a key having the value 0 and a key not being present. often,
> the kernel will just check for the presence of a key and ignore
> its value. but this is not the case for *nomp :)
>
> --
> cinap
>
>

[-- Attachment #1.2: Type: text/html, Size: 1048 bytes --]

[-- Attachment #2: Plan9-multiproc-fail.jpg --]
[-- Type: image/jpeg, Size: 101807 bytes --]

Re: [9fans] multi-processor support

[Gorka Guardiola]

[-- Attachment #1: Type: text/plain, Size: 40 bytes --]

Have you ticked?

Enable the IO APIC

[-- Attachment #2: Type: text/html, Size: 66 bytes --]

Re: [9fans] multi-processor support

[David du Colombier]

> Also, is there a way to mount the plan9 internal partitions
> (9fat, fossil etc) on Linux?

You could do it easily, using 9vx:

% disk/partfs plan9.img
% disk/fdisk -p /dev/sdXX/data >/dev/sdXX/ctl
% disk/prep -p /dev/sdXX/plan9 >/dev/sd00/ctl
% fossil/fossil -m 20 -f /dev/sdXX/fossil -c 'srv -AWP fossil' -c 'srv
-p fscons'
% mount /srv/fossil /n/fossil

Re: [9fans] multi-processor support

[erik quanstrom]

> Also, is there a way to mount the plan9 internal partitions (9fat, fossil
> etc) on Linux?

if you just want to rescue your plan9.ini, then boot the cd in that image
and edit it directly.  you can toss in new kernels that way, too.

if you get one kernel limping along in one vm and set up a nat for your
vms, you can pxe boot test kernels.  it makes life quite a bit nicer.

i know that this works for vmware fusion.  should probablly work for
workstation, too.

> I tried *nomp=0, but still only saw one CPU. I removed the line altogether,
> now the kernel panics (see attachment). Any way to override the kernel
> params in plan9.ini?

it appears that the lapic clock is reasonable, but what is not reasonable is the
cpu clock is 1/10th the frequency printed on the cpu0 line.

i think this may be because the emulation is confusing the 8253
timing.  if this is true, then the solution is to change this line

archmp.c:79: 	if(cpuserver && m->havetsc)

to remove the cpu server bit.  i removed this test from 9atom kernels,
since i didn't understand how the distinction could be helpful, and
it seemed like the seed of a very long debugging session starting with
the puzzle "why does the cpu kernel work on this hardware, but the
terminal not?"

- erik