On Fri, Aug 25, 2023 at 02:15:27PM +0000, Sripada, Radhakrishna wrote:
> I was trying this as a solution for the Pipe fault errors. However, I still 
> see
> The pipe fault errors which do not occur all the time.

We should avoid overloading CI with tests. But if needed because we cannot
reproduce locally but only on CI, please use the try-bot list instead of
this one.

And in the very last case where this list needs to be used, please use
another prefix like CI or HAX, but not the topic/core-for-ci, otherwise
we will think you are already asking to get that merged there.

> 
> Will update the explanation in my follow up patches.

> 
> Thanks,
> Radhakrishna Sripada
> 
> > -----Original Message-----
> > From: Jani Nikula <[email protected]>
> > Sent: Thursday, August 24, 2023 11:54 PM
> > To: Sripada, Radhakrishna <[email protected]>; intel-
> > [email protected]
> > Cc: Raj, Ashok <[email protected]>
> > Subject: Re: [Intel-gfx] [topic/core-for-ci v2] iommu/vt-d: Check domain 
> > flags
> > before setting snp bit in page-control
> > 
> > On Thu, 24 Aug 2023, Radhakrishna Sripada <[email protected]>
> > wrote:
> > > From: Ashok Raj <[email protected]>
> > >
> > 
> > The *why* goes here, along with a link to a gitlab issue.
> > 
> > Please don't expect topic/core-for-ci to have lower standards than any
> > other branches. That's not the case. On the contrary, you'll need the
> > *additional* justification for the commit being in topic/core-for-ci,
> > and the gitlab issue.
> > 
> > 
> > BR,
> > Jani.
> > 
> > 
> > > Signed-off-by: Ashok Raj <[email protected]>
> > > Signed-off-by: Radhakrishna Sripada <[email protected]>
> > > ---
> > >  drivers/iommu/intel/iommu.c | 2 +-
> > >  drivers/iommu/intel/pasid.c | 2 +-
> > >  2 files changed, 2 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c
> > > index 5c8c5cdc36cf..71da6f818e96 100644
> > > --- a/drivers/iommu/intel/iommu.c
> > > +++ b/drivers/iommu/intel/iommu.c
> > > @@ -2150,7 +2150,7 @@ __domain_mapping(struct dmar_domain *domain,
> > unsigned long iov_pfn,
> > >   if ((prot & (DMA_PTE_READ|DMA_PTE_WRITE)) == 0)
> > >           return -EINVAL;
> > >
> > > - attr = prot & (DMA_PTE_READ | DMA_PTE_WRITE | DMA_PTE_SNP);
> > > + attr = prot & (DMA_PTE_READ | DMA_PTE_WRITE);
> > >   attr |= DMA_FL_PTE_PRESENT;
> > >   if (domain->use_first_level) {
> > >           attr |= DMA_FL_PTE_XD | DMA_FL_PTE_US |
> > DMA_FL_PTE_ACCESS;
> > > diff --git a/drivers/iommu/intel/pasid.c b/drivers/iommu/intel/pasid.c
> > > index c5d479770e12..a057ecf84d82 100644
> > > --- a/drivers/iommu/intel/pasid.c
> > > +++ b/drivers/iommu/intel/pasid.c
> > > @@ -538,7 +538,7 @@ int intel_pasid_setup_first_level(struct intel_iommu
> > *iommu,
> > >   if (flags & PASID_FLAG_FL5LP)
> > >           pasid_set_flpm(pte, 1);
> > >
> > > - if (flags & PASID_FLAG_PAGE_SNOOP)
> > > + if ((flags & PASID_FLAG_PAGE_SNOOP) && ecap_sc_support(iommu-
> > >ecap))
> > >           pasid_set_pgsnp(pte);
> > >
> > >   pasid_set_domain_id(pte, did);
> > 
> > --
> > Jani Nikula, Intel Open Source Graphics Center

Reply via email to