Re: [PATCH V8 5/9] pci, acpi: add acpi hook to assign domain number.

[Tomasz Nowicki]

Hi Bjorn,

On 08.06.2016 02:15, Bjorn Helgaas wrote:
> Hi Tomasz,
>
> On Mon, May 30, 2016 at 05:14:18PM +0200, Tomasz Nowicki wrote:
> > PCI core code provides a config option (CONFIG_PCI_DOMAINS_GENERIC)
> > that allows assigning the PCI bus domain number generically by
> > relying on device tree bindings, and falling back to a simple counter
> > when the respective DT properties (ie "linux,pci-domain") are not
> > specified in the host bridge device tree node.
> >
> > In a similar way, when a system is booted through ACPI, architectures
> > that are selecting CONFIG_PCI_DOMAINS_GENERIC (ie ARM64) require kernel
> > hooks to retrieve the domain number so that the PCI bus domain number
> > set-up can be handled seamlessly with DT and ACPI in generic core code
> > when CONFIG_PCI_DOMAINS_GENERIC is selected.
> >
> > Since currently it is not possible to retrieve a pointer to the PCI
> > host bridge ACPI device backing the host bridge from core PCI code
> > (which would allow retrieving the domain number in an arch agnostic
> > way through the ACPI _SEG method), an arch specific ACPI hook has to
> > be declared and implemented by all arches that rely on
> > CONFIG_PCI_DOMAINS_GENERIC to retrieve the domain number and set it
> > up in core PCI code.
> >
> > For the aforementioned reasons, this patch introduces a dummy
> > acpi_pci_bus_domain_nr() hook in preparation for per-arch implementation
> > of the same to retrieve the domain number on a per-arch basis when
> > the system boots through ACPI.
> >
> > For the sake of code clarity the current code implementing generic
> > domain number assignment (ie pci_bus_assign_domain_nr(), selected by
> > CONFIG_PCI_DOMAINS_GENERIC) is reshuffled so that the code implementing
> > the DT domain assignment function is stubbed out into a corresponding
> > helper, so that DT and ACPI functions are clearly separated in
> > preparation for arches acpi_pci_bus_domain_nr() implementations.
> >
> > Signed-off-by: Tomasz Nowicki <tn@xxxxxxxxxxxx>
> > Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@xxxxxxx>
> > ---
> >   drivers/pci/pci.c   | 11 +++++++++--
> >   include/linux/pci.h |  1 +
> >   2 files changed, 10 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > index eb431b5..2b52178 100644
> > --- a/drivers/pci/pci.c
> > +++ b/drivers/pci/pci.c
> > @@ -7,6 +7,7 @@
> >    *	Copyright 1997 -- 2000 Martin Mares <mj@xxxxxx>
> >    */
> >
> > +#include <linux/acpi.h>
> >   #include <linux/kernel.h>
> >   #include <linux/delay.h>
> >   #include <linux/init.h>
> > @@ -4941,7 +4942,7 @@ int pci_get_new_domain_nr(void)
> >   }
> >
> >   #ifdef CONFIG_PCI_DOMAINS_GENERIC
> > -void pci_bus_assign_domain_nr(struct pci_bus *bus, struct device *parent)
> > +static int of_pci_bus_domain_nr(struct device *parent)
>
> Can we do a little cleanup before this patch?
>
>    - pci_bus_assign_domain_nr() is only used inside drivers/pci, so
>      maybe we move the prototype to drivers/pci/pci.h?

pci_bus_assign_domain_nr() goes with pci_domain_nr() as an option for 
CONFIG_PCI_DOMAINS_GENERIC. pci_domain_nr() is used commonly outside 
drivers/pci so we would need to split these calls then, thus personally 
I think it would be better to keep both in inclue/linux/pci.h

>    - I don't really like the style of calling a function that
>      internally assigns bus->domain_nr.  Could we do something like
>      this instead?
>
>      int pci_bus_domain_nr(...)
>      {
>        ...
>        return domain;
>      }
>
>      ... pci_create_root_bus(...)
>      {
>        ...
>        b->domain_nr = pci_bus_domain_nr(...);


We can. I do not see much difference between pci_bus_domain_nr() and 
pci_domain_nr() which we already have so how about calling it 
pci_bus_find_domain_nr instead? Lorenzo, any strong preference for it?

> That would be two new patches, if this makes sense.
>
> And this patch would only rename pci_bus_assign_domain_nr() to
> of_pci_bus_domain_nr() and add the pci_bus_domain_nr() wrapper.

Giving that we would keep prototypes in inclue/linux/pci.h

1. First patch would rename pci_bus_assign_domain_nr() to 
pci_bus_find_domain_nr() and it would return domain number so that we 
could do:
     ... pci_create_root_bus(...)
     {
       ...
       b->domain_nr = pci_bus_domain_nr(...);
       ...
     }

2. Second patch would transform pci_bus_find_domain_nr() to be the 
wrapper for of_pci_bus_domain_nr()

3. Third patch would add stub definition, the ARM64 definition and the 
new call acpi_pci_bus_domain_nr() in pci_bus_find_domain_nr() wrapper.

Does this plan sound reasonable?

> >   {
> >   	static int use_dt_domains = -1;
> >   	int domain = -1;
> > @@ -4985,7 +4986,13 @@ void pci_bus_assign_domain_nr(struct pci_bus *bus, struct device *parent)
> >   		domain = -1;
> >   	}
> >
> > -	bus->domain_nr = domain;
> > +	return domain;
> > +}
> > +
> > +void pci_bus_assign_domain_nr(struct pci_bus *bus, struct device *parent)
> > +{
> > +	bus->domain_nr = acpi_disabled ? of_pci_bus_domain_nr(parent) :
> > +					 acpi_pci_bus_domain_nr(bus);
> >   }
> >   #endif
> >   #endif
> > diff --git a/include/linux/pci.h b/include/linux/pci.h
> > index 12349de..bba4053 100644
> > --- a/include/linux/pci.h
> > +++ b/include/linux/pci.h
> > @@ -1390,6 +1390,7 @@ static inline int pci_domain_nr(struct pci_bus *bus)
> >   {
> >   	return bus->domain_nr;
> >   }
> > +static inline int acpi_pci_bus_domain_nr(struct pci_bus *bus) { return -1; }
>
> I would split the addition of acpi_pci_bus_domain_nr() to a separate
> patch and include the ARM64 definition in that same patch.  That patch
> would only add this stub definition, the ARM64 definition, and the new
> call in pci_bus_domain_nr().

OK

Thanks,
Tomasz