Re: [PATCH] irqchipi/gic-v4: Ensure accessing the correct RD when and writing INVLPIR

[Marc Zyngier <maz@kernel.org>]

On Tue, 16 May 2023 13:01:59 +0100,
Kunkun Jiang <jiangkunkun@huawei.com> wrote:
> 
> Hi Marc,
> 
> On 2023/5/16 18:15, Marc Zyngier wrote:
> > On Thu, 13 Apr 2023 04:57:17 +0100,
> > Kunkun Jiang <jiangkunkun@huawei.com> wrote:
> >>> Wouldn't it be nice if irq_to_cpuid() could work out whether it is
> >>> dealing with a LPI or a VLPI like it does today, but also directly
> >>> with a VPE? We could then use the same code as derect_lpi_inv(). I
> >>> came up with this the hack below, which is totally untested as I don't
> >>> have access to GICv4.1 HW.
> >>> 
> >>> Could you give it a spin?
> >> Nice, I will test it as soon as possible.
> > Did you ever managed to test this?
> Sorry,I've only been coordinating the GICv4.1 environment in the last
> few days. I have tested it with GICv4 many times, and it doesn't recur.
> However, another call trace occurs with GICv4.1 when the device driver
> is loaded in Guest. I haven't found out why. Maybe you can help analyze it.

I also went back to my patch, and it is a bit bogus (it doesn't even
compile to start with). I've now posted new version[1] that should at
least fix the bug you initially reported.

Can you please test it and reply to it?

Thanks,

	M.

[1] https://lore.kernel.org/r/20230617073242.3199746-1-maz@kernel.org

-- 
Without deviation from the norm, progress is not possible.