Re: [PATCH] iommu/amd: Fix "Guest Virtual APIC Table Root Pointer" configuration in IRTE

[Kishon VijayAbraham I]

Hi Joao,

On 4/3/2023 3:41 PM, Joao Martins wrote:
> On 03/04/2023 05:29, Suthikulpanit, Suravee wrote:
> > Kishon,
> >
> > On 3/31/2023 1:17 PM, Kishon Vijay Abraham I wrote:
> > > commit b9c6ff94e43a ("iommu/amd: Re-factor guest virtual APIC
> > > (de-)activation code") while refactoring guest virtual APIC
> > > activation/de-activation code, stored information for activate/de-activate
> > > in "struct amd_ir_data". It used 32-bit integer data type for storing the
> > > "Guest Virtual APIC Table Root Pointer" (ga_root_ptr), though the
> > > "ga_root_ptr" is actually a 40-bit field in IRTE (Interrupt Remapping
> > > Table Entry).
> > >
> > > This causes interrupts from PCIe devices to not reach the guest in the case
> > > of PCIe passthrough with SME (Secure Memory Encryption) enabled as _SME_
> > > bit in the "ga_root_ptr" is lost before writing it to the IRTE.
> > >
> > > Fix it by using 64-bit data type for storing the "ga_root_ptr".
> > >
> > > Fixes: b9c6ff94e43a ("iommu/amd: Re-factor guest virtual APIC (de-)activation
> > > code")
> > > Cc: stable@xxxxxxxxxxxxxxx # v5.4+
> > > Reviewed-by: Suravee Suthikulpanit <suravee.suthikulpanit@xxxxxxx>
> > > Signed-off-by: Kishon Vijay Abraham I <kvijayab@xxxxxxx>
> >
> > Please also add
> > Reported-by: Alejandro Jimenez <alejandro.j.jimenez@xxxxxxxxxx>
> > ---
> >   drivers/iommu/amd/amd_iommu_types.h | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/iommu/amd/amd_iommu_types.h b/drivers/iommu/amd/amd_iommu_types.h
> > index 3d684190b4d5..990614b8a1fe 100644
> > --- a/drivers/iommu/amd/amd_iommu_types.h
> > +++ b/drivers/iommu/amd/amd_iommu_types.h
> > @@ -1001,7 +1001,7 @@ struct amd_ir_data {
> >   	 */
> >   	struct irq_cfg *cfg;
> >   	int ga_vector;
> > -	int ga_root_ptr;
> > +	u64 ga_root_ptr;
> >   	int ga_tag;
> >   };
>
> Nice catch. While at it I noticed the following:
>
> * ga_tag: The conversion of uint (in kvm as they use proper u32 for gatag) to
> int will miss a bit thus at least VM IDs with the msb set will be stripped from
> vCPU wake ups? Fortunately the 24 bits for VCPU ID wouldn't be affected. But
> ga_tag should probably be a u32.
>
> * ga_vector: this in principle should be u8 per spec, maybe it is OK to stay as
> is to prevent compiler warnings as IOMMU intremap struct vcpu_data is using a u32.

Thank you for reviewing. Will add fix for other data types in my v2 of 
this patch.

Regards,
Kishon