Re: [PATCH v7 2/5] xen/arm: Enable the existing x86 virtual PCI support for ARM

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Fri, 15 Oct 2021, Bertrand Marquis wrote:
> Hi,
> 
> > On 15 Oct 2021, at 16:10, Julien Grall <julien@xxxxxxx> wrote:
> > 
> > Hi Bertrand,
> > 
> > On 15/10/2021 14:59, Bertrand Marquis wrote:
> >> diff --git a/xen/drivers/passthrough/pci.c b/xen/drivers/passthrough/pci.c
> >> index 3aa8c3175f..082892c8a2 100644
> >> --- a/xen/drivers/passthrough/pci.c
> >> +++ b/xen/drivers/passthrough/pci.c
> >> @@ -766,7 +766,21 @@ int pci_add_device(u16 seg, u8 bus, u8 devfn,
> >>          list_add(&pdev->domain_list, &hardware_domain->pdev_list);
> >>      }
> >>      else
> >> +    {
> >> +#ifdef CONFIG_ARM
> >> +        /*
> >> +         * On ARM PCI devices discovery will be done by Dom0. Add vpci 
> >> handler
> >> +         * when Dom0 inform XEN to add the PCI devices in XEN.
> >> +         */
> >> +        ret = vpci_add_handlers(pdev);
> > 
> > Sorry for only noticing it now. Looking at the last staging
> > vpci_add_handlers() is annotated with __hwdom_init. On Arm, __hwdom_init 
> > means the function will disappear after boot.
> > 
> > However, pci_add_device() can be called from a physdev op. So I think we 
> > would need to drop __hwdom_init. I can't seem to find this change in this 
> > series. Did I miss anything?
> 
> Good catch and not this is not in the serie.
> 
> Can we consider that a bug so that I can send a new patch or should I send a 
> v8 ?
 
We don't typically do that, but I could make the change on commit, or
merge a second patch from you with this one on commit, after I run all
the gitlab-ci tests.

(I still have to read the series but FYI)