Re: [Xen-devel] [PATCH v8 3/4] xen/arm: Make gic-v2 code handle hip04-d01 platform

[Frediano Ziglio <freddy77@xxxxxxxxx>]

2015-03-03 15:42 GMT+00:00 Julien Grall <julien.grall@xxxxxxxxxx>:
> On 03/03/15 15:36, Frediano Ziglio wrote:
>>>
>>> Hello Frediano,
>>>
>>> On 03/03/15 11:19, Frediano Ziglio wrote:
>>>> The GIC in this platform is mainly compatible with the standard
>>>> GICv2 beside:
>>>> - ITARGET is extended to 16 bit to support 16 CPUs;
>>>> - SGI mask is extended to support 16 CPUs;
>>>> - maximum supported interrupt is 510;
>>>
>>> 510 is not a multiple of 32. Is it normal?
>>>
>>> This will result to having nr_lines = 512. What happen is we are trying
>>> to access IRQ 510 and 511?
>>>
>>
>> I don't know. I think it's the same reason why in xen/arch/arm/gic.c the 
>> limit for irq is 1021 and not 1024 (see "if ( likely(irq >= 16 && irq < 
>> 1021) )" line)
>
> IRQ 1021-1023 are reserved by the GIC as spurious interrupt.
>
> If I understand correctly what you say, IRQ 510-511 may be considered
> for spurious interrupt?
>
> If so, the check (irq >= 16 && irq < 1021) needs to be changed.
>
>>> Also, is it possible to have GICH.VirtualID >= 510?
>>>
>>
>> I think so, GICH have the same interface of normal GICv2.
>
> But some offsets are different... so I'd like a confirmation based on
> some spec.
>

You can find spec at https://github.com/hisilicon/boards/tree/master/D01/docs.

Yes, VirtualID >= 510 can be used.

> For instance on GICv2 if we use some VirtualID (1021-1023) the behavior
> is unpredictable.
>

Same for our GIC.

> So if you have the a similar things on your board we may need to
> restrict the number of VirtualID in order to avoid introduce a possible
> host denial from a guest.
>

Frediano

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