Re: [PATCH 3/7] iommu/vtd: Do not wipe out the page table NID when devices detach

[Jason Gunthorpe <jgg@nvidia.com>]

On Tue, Jun 10, 2025 at 02:41:11PM +0800, Baolu Lu wrote:
> On 6/10/25 03:58, Jason Gunthorpe wrote:
> > The NID is used to control which NUMA node memory for the page table is
> > allocated it from. It should be a permanent property of the page table
> > when it was allocated and not change during attach/detach of devices.
> > 
> > Signed-off-by: Jason Gunthorpe<jgg@nvidia.com>
> > ---
> >   drivers/iommu/intel/iommu.c | 1 -
> >   1 file changed, 1 deletion(-)
> > 
> > diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c
> > index ceb960a796e2ba..72b0769c391008 100644
> > --- a/drivers/iommu/intel/iommu.c
> > +++ b/drivers/iommu/intel/iommu.c
> > @@ -1391,7 +1391,6 @@ void domain_detach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
> >   	if (--info->refcnt == 0) {
> >   		ida_free(&iommu->domain_ida, info->did);
> >   		xa_erase(&domain->iommu_array, iommu->seq_id);
> > -		domain->nid = NUMA_NO_NODE;
> 
> It appears that this is a fix.

Hard to say, but sure it could be..

> Fixes: 7c204426b818 ("iommu/vt-d: Add domain_alloc_paging support")

Arguably it is before this, the NID should not have been cleared once
set by attach. The above only fixed things to have the NID set from
the creation side..

I'd pick:

Fixes: ba949f4cd4c3 ("iommu/vt-d: Refactor iommu information of each domain")

As the patch that introduced the NUMA_NO_NODE on the detach flow in
the first place..

But that would encourage people to back port this quite far, and I
don't think that's really a good idea since it isn't something someone
cares about, and maybe there is good reason the above had it..

So I'd leave it with no fixes line..

Thanks,
Jason