Re: [Qemu-arm] [kvm-unit-tests PATCH v4 03/11] arm/arm64: smp: support more than 8 cpus

[Andre Przywara]

Hi,

On 09/11/16 11:57, Andrew Jones wrote:
> On Wed, Nov 09, 2016 at 11:12:03AM +0000, Andre Przywara wrote:
> [...]
>>> diff --git a/lib/arm/setup.c b/lib/arm/setup.c
>>> index 7e7b39f11dde..b6e2d5815e72 100644
>>> --- a/lib/arm/setup.c
>>> +++ b/lib/arm/setup.c
>>> @@ -24,12 +24,22 @@ extern unsigned long stacktop;
>>>  extern void io_init(void);
>>>  extern void setup_args_progname(const char *args);
>>>  
>>> -u32 cpus[NR_CPUS] = { [0 ... NR_CPUS-1] = (~0U) };
>>> +u64 cpus[NR_CPUS] = { [0 ... NR_CPUS-1] = (~0U) };
>>
>> This should be ~0UL.
> 
> Indeed. Thanks.
> 
>> Also I think the type should be unsigned long to match the types used
>> everywhere else.
> 
> I'll change mpidr_to_cpu to return u64 instead of unsigned long.

I am not sure this is the right direction. unsigned long is really the
natural type for MPIDR, since this is a system register with exactly the
native register size.
I think we use it this way in the kernel.

Cheers,
Andre

> Actually I think Alex suggested that. I'm not sure why I haven't
> done it...
> 
>>
>>>  int nr_cpus;
>>>  
>>>  struct mem_region mem_regions[NR_MEM_REGIONS];
>>>  phys_addr_t __phys_offset, __phys_end;
>>>  
>>> +int mpidr_to_cpu(unsigned long mpidr)
>>> +{
>>> +   int i;
>>> +
>>> +   for (i = 0; i < nr_cpus; ++i)
>>> +           if (cpus[i] == (mpidr & MPIDR_HWID_BITMASK))
>>> +                   return i;
>>> +   return -1;
>>> +}
>>> +
>>>  static void cpu_set(int fdtnode __unused, u32 regval, void *info __unused)
>>
>> I guess this needs to be extended as well, including
>> dt_for_each_cpu_node() to cope with 64-bit reg properties in the CPU
>> node (where the upper word is not 0).
>> But this is not really crucial atm, so can be fixed in a follow-up patch.
> 
> Yeah, I'll do it as a followup series, because it'll affect powerpc too.
> 
> drew
>