On Wed, 08 Jun 2011 14:40:37 +1000
Benjamin Herrenschmidt <benh@xxxxxxxxxxxxxxxxxxx> wrote:
> Hi folks !
>
> I'm trying to refactor/cleanup our pseries "EEH" code (the IBM specific
> error handling/recovery thingy), for the sake of general cleanness but
> also because I will have to significantly extend it soon to deal with a
> new/different platform.
>
> Part of the thing that I'd like to get rid of is the way it walks the
> device-tree at boot time to do "things" to every PCI device (like
> enabling the EEH functionality) before the PCI probing happens.
>
> This code is duplicated with hotplug hooks used when hot-plugging new
> devices, and isn't going to be suitable for platforms that may not
> represent all PCI devices in the OFW device-tree.
>
> However, I -do- need to tickle the firmware/hardware for each device
> before it is probed.
>
> Thus would there be any objection to add a hook, possibly in
> pci_scan_device(), something along the lines of:
>
> static struct pci_dev *pci_scan_device(struct pci_bus *bus, int devfn)
> {
> struct pci_dev *dev;
> u32 l;
> int delay = 1;
>
> + if (pcibios_scan_device_hook(bus, devfn))
> + return NULL;
> +
> if (pci_bus_read_config_dword(bus, devfn, PCI_VENDOR_ID, &l))
> return NULL;
>
> I want to poke you guys first before I actually do the patch & refactor
> everything on my side :-)
So basically an arch hook in place of a quirk for everything? Yeah
that's probably clearer, and maybe there's some code in a catch all
quirk that could be moved there on other arches as well.
--
Jesse Barnes, Intel Open Source Technology Center
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
