Re: [Xen-devel] [PATCH] xen/arm: minor improvement in smp_send_call_function_mask()

[Anup Patel <anup.patel@xxxxxxxxxx>]

On 20 August 2014 20:37, Julien Grall <julien.grall@xxxxxxxxxx> wrote:
>
>
> On 20/08/14 01:14, Anup Patel wrote:
>>>
>>> Is it necessary? What happen if Xen tries to send an SGI with an empty
>>> mask?
>>
>>
>> If Xen tries to send SGI with empty mask then target_mask will be empty
>> hence cpumask_weight(&target_mask) will return 0
>
>
> The GIC documentation says:
> "If this field is 0x00 when TargetListFilter is 0b00 , the Distributor does
> not forward the interrupt to any
> CPU interface."
>
> TargetListFilter == 0b00 => Forward interrupt to a specific list of CPUs.
>
>
>>>
>>> AFAIU, the function cpumask_weight is complex so if we can avoid it, it
>>> would be better.
>>
>>
>> Can you explain more about how cpumask_weight is complex ??
>
>
> cpumask_weight contains a loop and multiple addition. I doubt the case where
> the mask only contains the current cpu happens often.
>
> As the GIC won't forward the interrupt if the list of CPUs is empty, I don't
> think it's worth to add this check.
>
>
>> Other alternative is to use "cpumask_first(&target_mask) != NR_CPUS".
>
>
> The best alternative would be cpumask_empty.

All three cpumask_empty(), cpumask_first(), and cpumask_weight()
are O(N) where N is number of bits in cpumask.

It really does not make much difference which of these operation
is chosen.

Since empty target list is fine with GIC Distributor, I will drop the check.

--
Anup

>
> Regards,
>
> --
> Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel