Re: [Xen-devel] [PATCH 05/11] xen/arm: vpl011: Initialize nr_spis in vgic_init in Xen to atleast 1

[Julien Grall <julien.grall@xxxxxxx>]

Hi Bhupinder,

On 03/16/2017 10:31 AM, Bhupinder Thakur wrote:
> On 16 March 2017 at 13:54, Julien Grall <julien.grall@xxxxxxx> wrote:
> > > The other option is to reserve a SPI for pl011 at compile time and use
> > > that value. Let me know.
> >
> > Whilst I am ok to have the pl011 SPI number hardcoded, I don't like the
> > approach taken in this patch because the toolstack is in charge of the guest
> > layout (interrupt, memory...) and not the hypervisor.
> >
> > The values are hardcoded today because we decided to do a fix layout for
> > simplicity. It is likely to be changed in the future.
> >
> > The toolstack knows how much memory the user requested, the list of devices
> > available... So it is the goal of the toolstack to bump the number of SPIs
> > before creating the domain if a PL011 will be exposed.
> >
> > Also, the interaction between the pl011 and the parameter "irqs" in the
> > domain configuration file will need to be documented. By that I mean
> > explaining from which number the SPIs will be allocated when choosing a
> > pl011 enabling.
> >
> > Note the probably want to allow the user to choose the pl011 IRQ and MMIO
> > region. If he doesn't provide any, we would use the default value.
> >
> > Cheers,
> We can follow the convention that when pl011 is enabled then the last
> irq in the "irqs" list will be used as the pl011 irq. I believe that
> the irqs mentioned in the "irqs" list will be reserved by the
> hypervisor when the domain is created and we need not allocate a SPI
> separately as we are doing currently.
I don't understand what you mean here. If you speak about the xl 
parameter "irqs", it is a list of physical IRQs and not virtual IRQs. 
Furthermore, we would like to keep to avoid a normal user to do more 
than specifying "pl011" or else on in the guest configuration file.
So could you clarify what you mean?

Cheers,

--
Julien Grall

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