Re: [Xen-devel] [RFC XEN PATCH v3 08/11] xen: arm: vgic: don't fail if IRQ is already connected

[Julien Grall <julien@xxxxxxx>]

Hi,

On 15/11/2019 20:10, Stewart Hildebrand wrote:
> There are some IRQs that happen to have multiple "interrupts = < ... >;"
> properties with the same IRQ in the device tree. For example:
>
> interrupts = <0 123 4>,
>               <0 123 4>,
>               <0 123 4>,
>               <0 123 4>,
>               <0 123 4>;
>
> In this case it seems that we are invoking vgic_connect_hw_irq multiple
> times for the same IRQ.
>
> Rework the checks to allow booting in this scenario.
>
> I have not seen any cases where the pre-existing p->desc is any different from
> the new desc, so BUG() out if they're different for now.
>
> Signed-off-by: Stewart Hildebrand <stewart.hildebrand@xxxxxxxxxxxxxxx>
>
> ---
> v3: new patch
>
> I tested on Xilinx Zynq UltraScale+ with the old vGIC. I have not fully
> tested with CONFIG_NEW_VGIC. This hack only became necessary after
> introducing the PPI series, and I'm not entirely sure what the reason
> is for that.
>
> I'm also unsure if BUG()ing out is the right thing to do in case of
> desc != p->desc, or what conditions would even trigger this? Is this
> function exposed to guests?
This can happen with PPIs as the desc is per-CPU. If you migrate the 
vCPU to another pCPU, you will likely hit the BUG() below if the guest 
disabled the interrupt.
But I don't think we should call vgic_connect_hw_irq() multiples time on 
the same IRQ. The upper layer should take care of such issue (such as 
setup_guest_irq()).
Cheers,

--
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel