Re: [PATCH 2/2] PCI: generic: Add set_msi_parent callback

[Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>]

On Tue, Nov 11, 2014 at 11:52:27AM +0000, Liviu Dudau wrote:
> On Tue, Nov 11, 2014 at 11:24:24AM +0000, Will Deacon wrote:
> > Hi Suravee,
> > 
> > On Mon, Nov 10, 2014 at 07:24:40PM +0000, suravee.suthikulpanit@amd.com wrote:
> > > From: Suravee Suthikulpanit <Suravee.Suthikulpanit@amd.com>
> > > 
> > > This patch implement set_msi_parent callback for PCI generic host controller.
> > > 
> > > Cc: Bjorn Helgass <bhelgaas@google.com>
> > > Cc: Liviu Dudau <liviu.dudau@arm.com>
> > > Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> > > Signed-off-by: Suravee Suthikulpanit <Suravee.Suthikulpanit@amd.com>
> > > ---
> > >  drivers/pci/host/pci-host-generic.c | 14 ++++++++++++++
> > >  1 file changed, 14 insertions(+)
> > > 
> > > diff --git a/drivers/pci/host/pci-host-generic.c b/drivers/pci/host/pci-host-generic.c
> > > index 036ab1b..f9bb97a 100644
> > > --- a/drivers/pci/host/pci-host-generic.c
> > > +++ b/drivers/pci/host/pci-host-generic.c
> > > @@ -42,6 +42,7 @@ struct gen_pci {
> > >  	struct pci_host_bridge			host;
> > >  	struct gen_pci_cfg_windows		cfg;
> > >  	struct list_head			resources;
> > > +	struct msi_chip				*mchip;
> > >  };
> > >  
> > >  #ifdef CONFIG_ARM
> > > @@ -128,9 +129,19 @@ static int gen_pci_config_write(struct pci_bus *bus, unsigned int devfn,
> > >  	return PCIBIOS_SUCCESSFUL;
> > >  }
> > >  
> > > +static int gen_pci_set_msi_parent(struct pci_bus *bus)
> > > +{
> > > +	struct gen_pci *pci = bus_to_gen_pci(bus);
> > > +
> > > +	bus->msi = pci->mchip;
> > > +
> > > +	return PCIBIOS_SUCCESSFUL;
> > > +}
> > 
> > I don't think this makes much sense a host controller callback. Why can't
> > bus->msi be set in generic code?
> 
> Because of the current way in which a bus gets created and them immediately used for
> scanning devices if you use pci_scan_root_bus(). Alternative is to use pci_create_root_bus()
> and increase your code size with the body of pci_scan_root_bus().
> 
> Solution is to have pci_host_bridge holding the msi_chip pointer and that gets created
> before root bus, with pci_scan_root_bus() now having all the info needed to do successful
> setup of scanned devices.

Why can't we add a hook like pci_bus_assign_domain_nr(), say:

pci_bus_msi_init()

in pci_create_root_bus() that does what Suravee wants in a generic way ?

What am I missing ?

Lorenzo

> 
> Best regards,
> Liviu
> 
> > 
> > >  static struct pci_ops gen_pci_ops = {
> > >  	.read	= gen_pci_config_read,
> > >  	.write	= gen_pci_config_write,
> > > +	.set_msi_parent = gen_pci_set_msi_parent,
> > >  };
> > >  
> > >  static const struct of_device_id gen_pci_of_match[] = {
> > > @@ -313,6 +324,9 @@ static int gen_pci_probe(struct platform_device *pdev)
> > >  		return err;
> > >  	}
> > >  
> > > +	pci->mchip = of_pci_find_msi_chip_by_node(of_parse_phandle(np,
> > > +						  "msi-parent", 0));
> > 
> > This bit should be in the generic of_pci.c code and not duplicated for
> > each host controller.
> > 
> > Will
> > 
> 
> -- 
> ====================
> | I would like to |
> | fix the world,  |
> | but they're not |
> | giving me the   |
>  \ source code!  /
>   ---------------
>     ¯\_(ツ)_/¯