Re: [PATCH 4/7] xen/arm: gic: Use the correct CPU ID

[Julien Grall <julien.grall@linaro.org>]

On 09/10/2013 04:52 PM, Ian Campbell wrote:
> On Tue, 2013-09-10 at 16:42 +0100, Julien Grall wrote:
>> On 09/09/2013 02:28 PM, Ian Campbell wrote:
>>> On Fri, 2013-08-30 at 14:30 +0100, Julien Grall wrote:
>>>> The GIC mapping of CPU interfaces does not necessarily match the logical
>>>> CPU numbering.
>>>>
>>>> When Xen wants to send an SGI to specific CPU, it needs to use the GIC CPU ID.
>>>> It can be retrieved from ITARGETSR0, in fact when this field is read, the GIC
>>>> will return a value that corresponds only to the processor reading the register.
>>>> So Xen can use the PPI 0 to initialize the mapping.
>>>>
>>>> Signed-off-by: Julien Grall <julien.grall@linaro.org>
>>>> ---
>>>>  xen/arch/arm/gic.c |   35 ++++++++++++++++++++++++++++-------
>>>>  1 file changed, 28 insertions(+), 7 deletions(-)
>>>>
>>>> diff --git a/xen/arch/arm/gic.c b/xen/arch/arm/gic.c
>>>> index cadc258..4f3a8a5 100644
>>>> --- a/xen/arch/arm/gic.c
>>>> +++ b/xen/arch/arm/gic.c
>>>> @@ -57,6 +57,27 @@ static DEFINE_PER_CPU(uint64_t, lr_mask);
>>>>  
>>>>  static unsigned nr_lrs;
>>>>  
>>>> +/* The GIC mapping of CPU interfaces does not necessarily match the
>>>> + * logical CPU numbering. Let's use mapping as returned by the GIC
>>>> + * itself
>>>> + */
>>>> +#define NR_GIC_CPU_IF 8
>>>> +static u8 gic_cpu_map[NR_GIC_CPU_IF] __read_mostly = {0xff};
>>>
>>> Isn't this mapping from logical cpus ids to gic cpus ids? IOW the size
>>> of this array is the wrong way around?
>>
>> This array maps a logical cpu id (the index) to a gic cpus id. The size
>> if correct, because Xen will allocate logical cpu id from 0 to 7.
> 
> Oh right, yes.
> 
> I think NR_CPUS (which should be 8 given the other limitations) is the
> right size for this array.
> 
>>
>>> Is what you want here is a per-cpu variable?
>>
>> Does per-cpu variable allow any cpu to retrieve data of another cpu? I
>> didn't find a such function.
> 
> per_cpu(variable, cpu) will do it.

Is it fast ? I think this solution is slower, because the mapping is
used often used, mainly when an SGI is sent to another cpu.

>>
>>>> +
>>>> +static unsigned int gic_cpu_mask(const cpumask_t *cpumask)
>>>> +{
>>>> +    unsigned int cpu;
>>>> +    unsigned int mask = 0;
>>>> +
>>>> +    for_each_cpu(cpu, cpumask)
>>>> +    {
>>>> +        ASSERT(cpu < NR_GIC_CPU_IF);
>>>> +        mask |= gic_cpu_map[cpu];
>>>
>>> Further to above, can't cpu here be e.g. 0x100 for an a7 on a big.ITTLE
>>> system?
>>
>> 0x100 is the hardware CPU id. Xen will allocate a logical cpu id from 0.
> 
> Right.
> 
> 


-- 
Julien Grall