Re: [PATCH v8 17/23] vfio/listener: Bypass readonly region for dirty tracking

["Cédric Le Goater" <clg@redhat.com>]

Hello Zhenzhong,

On 11/28/25 03:08, Duan, Zhenzhong wrote:
> Hi Yi, Cedric,
> 
> Could you also help comment on this patch? This is a pure VFIO migration related optimization, I think it's better to let it go with the "vfio: relax the vIOMMU check" series.
> I'd like to move it in next respin of "vfio: relax the vIOMMU check" series if you think it make sense.

IMO, the "vfio: relax the vIOMMU check" is fine as it is.

I would instead introduce a new series to handle ERRATA_772415
since it is a special case of "intel_iommu: Enable first stage
translation for passthrough device".

So we would have 3 series (in order of appearance on the list) :

1. "vfio: relax the vIOMMU check"
2. "intel_iommu: Enable first stage translation for passthrough
     device" without quirks
3. "vfio: handle ERRATA_772415" with the quirk part, so that's
    patch 17,19,20,21 ?


Series 2 seems the most important, as it sets the foundation
for the other architectures which have a need for nested
IOMMU support (smmu/nvidia). Series 1 is nice to have.
Series 3. is an extension of 2. for broken HW.

For the next iterations (QEMU 11.0), let's get series 2. in
first. I have been including it in my QEMU tree for a while
now I didn't see any regression. Should be fine.

Then, we can merge 1. and 3. through the vfio queue. Shouldn't
be a major task now that we had all these reviews.

How's that ?


btw,

   Reference from 4th Gen Intel Xeon Processor Scalable Family Specification
   Update, Errata Details, SPR17.
   https://edc.intel.com/content/www/us/en/design/products-and-solutions/processors-and-chipsets/eagle-stream/sapphire-rapids-specification-update/

Url is not accessible (for me).

Thanks,

C.