Re: only a single core (out of 8) initialized

Re: only a single core (out of 8) initialized

[Andrew Morton]

On Thu, 4 Jun 2009 19:52:36 -0700 (PDT) david@lang.hm wrote:

> this is 2.6.30-rc7

Were any earlier kernels OK?

>  with a patch added to debug a nv sata driver 
> incompatibility with the Intel SSDs on a tyan motherboard
> 
> on one system it sees all 8 cores, on the other system it only initializes 
> one core.
> 
> can anyone help me spot what's going wrong in the one that's only seeing a 
> single core? If I'm reading things correctly it's seeing both sockets for 
> the NUMA setup, but then only initializing one core on the first socket.

good:
	SMP: Allowing 8 CPUs, 4 hotplug CPUs

bad:	
	SMP: Allowing 1 CPUs, 0 hotplug CPUs

perhaps due to
	ACPI: No APIC-table, disabling MPS

You could have a poke around in arch/x86/kernel/smpboot.c() and find
out what happened here.  I'd be suspecting this code triggered:

        /* no processor from mptable or madt */
        if (!num_processors)
                num_processors = 1;

because ACPI doesn't like that board.

Did you look into updating the BIOS?

Re: only a single core (out of 8) initialized

[david]

On Thu, 4 Jun 2009, Andrew Morton wrote:

> Subject: Re: only a single core (out of 8) initialized
> 
> On Thu, 4 Jun 2009 19:52:36 -0700 (PDT) david@lang.hm wrote:
>
>> this is 2.6.30-rc7
>
> Were any earlier kernels OK?
>
>>  with a patch added to debug a nv sata driver
>> incompatibility with the Intel SSDs on a tyan motherboard
>>
>> on one system it sees all 8 cores, on the other system it only initializes
>> one core.
>>
>> can anyone help me spot what's going wrong in the one that's only seeing a
>> single core? If I'm reading things correctly it's seeing both sockets for
>> the NUMA setup, but then only initializing one core on the first socket.
>
> good:
> 	SMP: Allowing 8 CPUs, 4 hotplug CPUs
>
> bad:
> 	SMP: Allowing 1 CPUs, 0 hotplug CPUs
>
> perhaps due to
> 	ACPI: No APIC-table, disabling MPS
>
> You could have a poke around in arch/x86/kernel/smpboot.c() and find
> out what happened here.  I'd be suspecting this code triggered:
>
>        /* no processor from mptable or madt */
>        if (!num_processors)
>                num_processors = 1;
>
> because ACPI doesn't like that board.
>
> Did you look into updating the BIOS?

this is actually the newer of the two systems. I've already escalated up 
to Tyan about the fact that this board won't see the Intel SSD drives 
(unless you hotplug the drive). I've forwarded this to them as well. their 
website does show a newer bios than what was shipped to me.

David Lang

Re: only a single core (out of 8) initialized

[david]

On Thu, 4 Jun 2009, Andrew Morton wrote:

> On Thu, 4 Jun 2009 19:52:36 -0700 (PDT) david@lang.hm wrote:
>
>> this is 2.6.30-rc7
>
> Were any earlier kernels OK?

sorry, I missed this question.

I was running 2.6.29.1 when I noticed this problem. I haven't tried older 
kernels yet.

David Lang

>>  with a patch added to debug a nv sata driver
>> incompatibility with the Intel SSDs on a tyan motherboard
>>
>> on one system it sees all 8 cores, on the other system it only initializes
>> one core.
>>
>> can anyone help me spot what's going wrong in the one that's only seeing a
>> single core? If I'm reading things correctly it's seeing both sockets for
>> the NUMA setup, but then only initializing one core on the first socket.
>
> good:
> 	SMP: Allowing 8 CPUs, 4 hotplug CPUs
>
> bad:
> 	SMP: Allowing 1 CPUs, 0 hotplug CPUs
>
> perhaps due to
> 	ACPI: No APIC-table, disabling MPS
>
> You could have a poke around in arch/x86/kernel/smpboot.c() and find
> out what happened here.  I'd be suspecting this code triggered:
>
>        /* no processor from mptable or madt */
>        if (!num_processors)
>                num_processors = 1;
>
> because ACPI doesn't like that board.
>
> Did you look into updating the BIOS?
>
>

Re: only a single core (out of 8) initialized

[Robert Hancock]

david@lang.hm wrote:
> On Thu, 4 Jun 2009, Andrew Morton wrote:
> 
>> Subject: Re: only a single core (out of 8) initialized
>>
>> On Thu, 4 Jun 2009 19:52:36 -0700 (PDT) david@lang.hm wrote:
>>
>>> this is 2.6.30-rc7
>>
>> Were any earlier kernels OK?
>>
>>>  with a patch added to debug a nv sata driver
>>> incompatibility with the Intel SSDs on a tyan motherboard
>>>
>>> on one system it sees all 8 cores, on the other system it only 
>>> initializes
>>> one core.
>>>
>>> can anyone help me spot what's going wrong in the one that's only 
>>> seeing a
>>> single core? If I'm reading things correctly it's seeing both sockets 
>>> for
>>> the NUMA setup, but then only initializing one core on the first socket.
>>
>> good:
>>     SMP: Allowing 8 CPUs, 4 hotplug CPUs
>>
>> bad:
>>     SMP: Allowing 1 CPUs, 0 hotplug CPUs
>>
>> perhaps due to
>>     ACPI: No APIC-table, disabling MPS
>>
>> You could have a poke around in arch/x86/kernel/smpboot.c() and find
>> out what happened here.  I'd be suspecting this code triggered:
>>
>>        /* no processor from mptable or madt */
>>        if (!num_processors)
>>                num_processors = 1;
>>
>> because ACPI doesn't like that board.
>>
>> Did you look into updating the BIOS?
> 
> this is actually the newer of the two systems. I've already escalated up 
> to Tyan about the fact that this board won't see the Intel SSD drives 
> (unless you hotplug the drive). I've forwarded this to them as well. 
> their website does show a newer bios than what was shipped to me.

Yeah, it seems like a BIOS problem on that system. It's missing a bunch 
of the ACPI tables that the other system has.