Re: [Mactel-linux-users] lost one cpu core when upgraded to RHEL 5.5

[Justin P. m. <jus...@gm...>]

On 04/14/2010 06:49 AM, Geoffrey wrote:
> Justin P. mattock wrote:
>> On 04/13/2010 07:57 AM, Geoffrey wrote:
>>> Geoffrey wrote:
>>>> Anyone out there running 64 bit OS on a macbook pro 4,1. I upgraded
>>>> from 32 bit Red Hat EL 5.4 to 64 bit RHEL 5.4, then again to 64 bit
>>>> RHEL
>>>> 5.5, which is actually still in beta. I don't know if I lost the core
>>>> in the 64 bit 5.4 or not. I see the following from dmesg:
>>>>
>>>> SMP: Allowing 2 CPUs, 0 hotplug CPUs
>>>> .
>>>> .
>>>> SMP alternatives: switching to UP code
>>>> .
>>>> .
>>>> SMP motherboard not detected.
>>>> .
>>>> .
>>>> SMP disabled
>>>
>>> Thought I'd update the list with a recent success:
>>>
>>> For whatever reason, I could not get the initial installation RHEL 5.4
>>> disk to even boot until I stumbled across a link that suggested the
>>> following boot parms:
>>>
>>> acpi=force noapic irqpoll

the acpi=force seems harmless, but the noapic scares me(you need this),
In any case you should be able to bootup without any of these boot 
params which makes me wonder if this is a kernel issue and/or userspace

>>>
>>> I don't know why they worked, but they did. Once installed, I found
>>> that I had to add them to my grub.conf to get the install to boot, but
>>> all was happy. After moving to 64bit, I lost one core of my two core
>>> processor. After much discussion with RH, I revisited these parameters
>>> and found that by removing the noapic, parm, my macbook would still boot
>>> and I got both cores. The weird thing is, I've tried every combination
>>> of these parms before in order to remove them from my boot process in
>>> the past, but my macbook would not boot. There must be something about
>>> the latest RHEL 5.5 kernel that resolved that issue.
>>>
>>> The weird thing is, I did not have this issue on 32bit RHEL 5.4. That
>>> is, I had both cores using all these parms.
>>>
>>
>> this doesn't seem right..
>> you need apic etc... they must be
>> doing something in there configs somewhere
>> (COFNIG_SMP=y if set gives you both cores
>> (but could be wrong)).
>
> No telling, all I know is I have a happy dual core system again with a
> 64 bit OS. ;)
>

That's good news.. if you can, maybe try loading the latest kernel from 
git, then if the system can bootup without any of these boot params
then you know there's a regression in the kernel somewhere(then if you 
have the time do a bisect, and report it to lkml, so there aware of the 
commit that causes this).

Justin P. Mattock