Re: [PATCH 2/6] x86/iommu: add common page-table allocator

[Jan Beulich <jbeulich@xxxxxxxx>]

On 27.07.2020 11:37, Durrant, Paul wrote:
> > From: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
> > Sent: 24 July 2020 19:24
> >
> > On 24/07/2020 17:46, Paul Durrant wrote:
> > > --- a/xen/drivers/passthrough/x86/iommu.c
> > > +++ b/xen/drivers/passthrough/x86/iommu.c
> > > @@ -140,11 +140,19 @@ int arch_iommu_domain_init(struct domain *d)
> > >
> > >       spin_lock_init(&hd->arch.mapping_lock);
> > >
> > > +    INIT_PAGE_LIST_HEAD(&hd->arch.pgtables.list);
> > > +    spin_lock_init(&hd->arch.pgtables.lock);
> > > +
> > >       return 0;
> > >   }
> > >
> > >   void arch_iommu_domain_destroy(struct domain *d)
> > >   {
> > > +    struct domain_iommu *hd = dom_iommu(d);
> > > +    struct page_info *pg;
> > > +
> > > +    while ( (pg = page_list_remove_head(&hd->arch.pgtables.list)) )
> > > +        free_domheap_page(pg);
> > Some of those 90 lines saved were the logic to not suffer a watchdog
> > timeout here.
> >
> > To do it like this, it needs plumbing into the relinquish resources path.
> Ok. I does look like there could be other potentially lengthy destruction done 
> off the back of the RCU call. Ought we have the ability to have a restartable 
> domain_destroy()?
I don't see how this would be (easily) feasible. Instead - why do
page tables not get cleaned up already at relinquish_resources time?

Jan