Re: Question about domain_init (v5.3-v5.7)

[Lu Baolu]

Hi Jerry,

On 12/1/20 1:50 AM, Jerry Snitselaar wrote:
> Lu Baolu @ 2020-11-26 19:12 MST:
>
> > Hi Jerry,
> >
> > On 11/27/20 5:35 AM, Jerry Snitselaar wrote:
> > > Lu Baolu @ 2020-11-26 04:01 MST:
> > >
> > > > Hi Jerry,
> > > >
> > > > On 2020/11/26 4:27, Jerry Snitselaar wrote:
> > > > > Is there a reason we check the requested guest address width against
> > > > > the
> > > > > iommu's mgaw, instead of the agaw that we already know for the iommu?
> > > > > I've run into a case with a new system where the mgaw reported is 57,
> > > > > but if they set PAE to 46 instead of 52 in the bios, then sagaw reports
> > > > > the highest supported agaw is 48 and the domain_init code fails here. In
> > > >
> > > > Isn't this a platform bug? If it's too late to fix it in the BIOS, you
> > > > maybe have to add a platform specific quirk to set mgaw to the highest
> > > > supported agaw?
> > > >
> > > > Best regards,
> > > > baolu
> > > Is there somewhere you can point me to that discusses how they
> > > should be
> > > setting the mgaw? I misunderstood when I previously asked you about
> > > whether the mgaw could be a value that was greater than any of sagaw.
> > > If it is a bios issue, then they should fix it there.
> >
> > MGAW indicates the max gpa width supported by 2nd translation. The VT-d
> > spec requires that this value must be at least equal to the host
> > physical addressibility. According to this, BIOS is good, right?
> >
> > For this failure case, domain_init() just wants to find a suitable agaw
> > for the private domain. I think it makes sense to check against
> > iommu->agaw instead of cap_mgaw.
> >
> > Best regards,
> > baolu
>
>  From this bit in the spec about MGAW:
>
>      Guest addressability for a given DMA request is limited to the
>      minimum of the value reported through this field and the adjusted
>      guest address width of the corresponding page-table structure.
>      (Adjusted guest address widths supported by hardware are reported
>      through the SAGAW field).
>
> That does suggest it should be adjusted down to the sagaw value in this case, yes?
> Just want to make sure I'm understanding it correctly.

Yes. I think so.

Best regards,
baolu

> > > > > other places like prepare_domain_attach_device, the dmar domain agaw
> > > > > gets adjusted down to the iommu agaw. The agaw of the iommu gets
> > > > > determined based off what is reported for sagaw. I'm wondering if it
> > > > > can't instead do:
> > > > > ---
> > > > >     drivers/iommu/intel-iommu.c | 4 ++--
> > > > >     1 file changed, 2 insertions(+), 2 deletions(-)
> > > > > diff --git a/drivers/iommu/intel-iommu.c
> > > > > b/drivers/iommu/intel-iommu.c
> > > > > index 6ca5c92ef2e5..a8e41ec36d9e 100644
> > > > > --- a/drivers/iommu/intel-iommu.c
> > > > > +++ b/drivers/iommu/intel-iommu.c
> > > > > @@ -1862,8 +1862,8 @@ static int domain_init(struct dmar_domain *domain, struct intel_iommu *iommu,
> > > > >     	domain_reserve_special_ranges(domain);
> > > > >     	/* calculate AGAW */
> > > > > -	if (guest_width > cap_mgaw(iommu->cap))
> > > > > -	        guest_width = cap_mgaw(iommu->cap);
> > > > > +	if (guest_width > agaw_to_width(iommu->agaw))
> > > > > +	        guest_width = agaw_to_width(iommu->agaw);
> > > > >     	domain->gaw = guest_width;
> > > > >     	adjust_width = guestwidth_to_adjustwidth(guest_width);
> > > > >     	agaw = width_to_agaw(adjust_width);
> > > > > --
> > > > > 2.27.0
> > > > >
> > > > > Thoughts? With the former code the ehci device for the ilo fails when
> > > > > trying to get a private domain.
> > > > > Thanks,
> > > > > Jerry