kexec for powerpc32

kexec for powerpc32

[Christian Kujau]

[please CC me as I am not subscribed]

Hi,

when building a kernel for powerpc32 (PowerBook6,8 / 7447A) I can select
CONFIG_KEXEC. However, after booting I can't convince kexec (from 
kexec-tools, latest git checkout) to load the kernel:

# /opt/kexec-tools/sbin/kexec -l /boot/2.6/zImage --append="root=/dev/hda6" 
get_memory_ranges(): Unsupported platform
Could not get memory layout

Can anybody comment on wether or not kexec is supported on this platform? 
If not, is this a limitation of kexec-tools or the kernel? If it's the 
latter, can't the option be hidden/disabled in Kconfig for this platform?

Then again, from looking at the Kconfig files, I see that the "74xx" CPU 
is among quite a few CPU types and so PPC_BOOK3S_32 is selected, and then
ALTIVEC selects 6xx which in turn selects PPC_BOOK3S - so, I guess 
!PPC_BOOK3S for KEXEC is not an option here?

Thanks,
Christian.
-- 
BOFH excuse #69:

knot in cables caused data stream to become twisted and kinked

Re: kexec for powerpc32

[Anton Vorontsov]

Hi,

On Wed, Mar 31, 2010 at 05:18:18AM -0700, Christian Kujau wrote:
> [please CC me as I am not subscribed]
> 
> Hi,
> 
> when building a kernel for powerpc32 (PowerBook6,8 / 7447A) I can select
> CONFIG_KEXEC. However, after booting I can't convince kexec (from 
> kexec-tools, latest git checkout) to load the kernel:
> 
> # /opt/kexec-tools/sbin/kexec -l /boot/2.6/zImage --append="root=/dev/hda6" 
> get_memory_ranges(): Unsupported platform
> Could not get memory layout

Kernel has all needed for kexec, but kexec-tools are broken for
powerpc32.

http://www.mail-archive.com/linuxppc-dev@lists.ozlabs.org/msg22498.html

I've just asked around, and it seems that Maxim (Cc'ed) will start
working on reviving ppc32 support real soon.

Thanks,

-- 
Anton Vorontsov
email: cbouatmailru@gmail.com
irc://irc.freenode.net/bd2

Re: kexec for powerpc32

[Christian Kujau]

On Wed, 31 Mar 2010 at 17:23, Anton Vorontsov wrote:
> Kernel has all needed for kexec, but kexec-tools are broken for
> powerpc32.
> 
> http://www.mail-archive.com/linuxppc-dev@lists.ozlabs.org/msg22498.html

Oh :-\

> I've just asked around, and it seems that Maxim (Cc'ed) will start
> working on reviving ppc32 support real soon.

Well, in my case the powerpc32 machine is far from being a "production 
system", but I'm glad to hear that he's working on it!

Thanks for the response,
Christian.
-- 
BOFH excuse #292:

We ran out of dial tone and we're and waiting for the phone company to deliver another bottle.

Re: kexec for powerpc32

[Maxim Uvarov]

Christian Kujau wrote:
> On Wed, 31 Mar 2010 at 17:23, Anton Vorontsov wrote:
>   
>> Kernel has all needed for kexec, but kexec-tools are broken for
>> powerpc32.
>>
>> http://www.mail-archive.com/linuxppc-dev@lists.ozlabs.org/msg22498.html
>>     
>
> Oh :-\
>
>   
>> I've just asked around, and it seems that Maxim (Cc'ed) will start
>> working on reviving ppc32 support real soon.
>>     
>
> Well, in my case the powerpc32 machine is far from being a "production 
> system", but I'm glad to hear that he's working on it!
>
> Thanks for the response,
> Christian.
>   
Yes,  the plan was to push upstream all MontaVista's powerpc kexec patches.
But also some work here is required in understanding how far community 
code from our,
port, cleanup and test patches. I hope that I will get to that task in 
this month.

Best regards,
Maxim Uvarov.

poweroff/suspend problems on PowerMac G5

[Bolko Maass]

Hi everybody!

Hope this list is even the right place to bring this up, but I've exhausted everything else I could think of, so here it goes.

The problem is that my PowerMac G5 Quad machine ("PowerMac 11,2", Gentoo, linux-2.6.34-rc2) will not power down if suspend infrastructure (CONFIG_HIBERNATION) is in the kernel. It will stop all services and halt, but never power down. This is when actually trying to suspend as well as when just trying to shut down the machine. Reboot works fine, so I suspect that the kernel doesn't actually crash.

At least one other person on the #gentoo-powerpc channel on freenode has reported the same problem and that unsetting CONFIG_HIBERNATION will fix it.

Please advise on how to debug something like this, as this is really beyond my expertise.

Thanks!

Re: poweroff/suspend problems on PowerMac G5

[Anton Vorontsov]

Hi,

On Thu, Apr 01, 2010 at 02:56:42PM +0200, Bolko Maass wrote:
> Hi everybody!
> 
> Hope this list is even the right place to bring this up, but I've
> exhausted everything else I could think of, so here it goes.
> 
> The problem is that my PowerMac G5 Quad machine ("PowerMac 11,2",
> Gentoo, linux-2.6.34-rc2) will not power down if suspend
> infrastructure (CONFIG_HIBERNATION) is in the kernel.
...
> It will
> stop all services and halt, but never power down. This is when
> actually trying to suspend

So CONFIG_HIBERNATION breaks suspend? Hm. How do you try
to suspend? echo mem > /sys/power/state?

> as well as when just trying to shut
> down the machine.

How do try to shut down the machine? halt/poweroff commands?

>From a brief look, I don't see how CONFIG_HIBERNATION may change
halt/poweroff code flow in the kernel...

Can you show us a boot log of working and not working kernels?
And .config files you use.

..maybe for some reason PMU/SMU modules didn't get loaded?

-- 
Anton Vorontsov
email: cbouatmailru@gmail.com
irc://irc.freenode.net/bd2

Re: poweroff/suspend problems on PowerMac G5

[Benjamin Herrenschmidt]

On Thu, 2010-04-01 at 19:14 +0400, Anton Vorontsov wrote:

> So CONFIG_HIBERNATION breaks suspend? Hm. How do you try
> to suspend? echo mem > /sys/power/state?

Nah, suspend isn't supported on these ( ... yet, I suppose I could
implement it one of these days if nouveau can bring back the video
card).

I think his problem is that CONFIG_HIB* is breaking shutdown.

> > as well as when just trying to shut
> > down the machine.
> 
> How do try to shut down the machine? halt/poweroff commands?
> 
> >From a brief look, I don't see how CONFIG_HIBERNATION may change
> halt/poweroff code flow in the kernel...
> 
> Can you show us a boot log of working and not working kernels?
> And .config files you use.
> 
> ..maybe for some reason PMU/SMU modules didn't get loaded?

Nah, I doubt it. Smells more like the old crap with broken cpu unplug on
shutdown, no ?

I have one of these guys at work, I can give it a try next week.

Cheers,
Ben.

Re: poweroff/suspend problems on PowerMac G5

[Romain Goyet]

On Fri, Apr 2, 2010 at 12:49 AM, Benjamin Herrenschmidt
<benh@kernel.crashing.org> wrote:
> On Thu, 2010-04-01 at 19:14 +0400, Anton Vorontsov wrote:
>
>> So CONFIG_HIBERNATION breaks suspend? Hm. How do you try
>> to suspend? echo mem > /sys/power/state?
>
> Nah, suspend isn't supported on these ( ... yet, I suppose I could
> implement it one of these days if nouveau can bring back the video
> card).
>
> I think his problem is that CONFIG_HIB* is breaking shutdown.
>
>> > as well as when just trying to shut
>> > down the machine.
>>
>> How do try to shut down the machine? halt/poweroff commands?
>>
>> >From a brief look, I don't see how CONFIG_HIBERNATION may change
>> halt/poweroff code flow in the kernel...
>>
>> Can you show us a boot log of working and not working kernels?
>> And .config files you use.
>>
>> ..maybe for some reason PMU/SMU modules didn't get loaded?
>
> Nah, I doubt it. Smells more like the old crap with broken cpu unplug on
> shutdown, no ?

I'm using a QuadG5 under Debian PPC right now, and indeed, it seems
like I have a similar issue (i.e. it takes *forever* for the machine
to halt). In my case it does end up powering itself off, but sometimes
it even goes in a state where all the fan are running at top-speed.
And indeed, I remember having some message in the console about CPUs
"unplugging" themselves.

>
> I have one of these guys at work, I can give it a try next week.
>

Great. While you're at it, is there any chance you could try to get
"server_mode" to work for this machine (or any SMU-based machine for
that matter). You know, so that the machine reboots after a power
failure :-)

Thanks a lot Ben !

  - Romain

> Cheers,
> Ben.
>
>
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/linuxppc-dev
>