Re: platform/x86: p2sb: Allow p2sb_bar() calls during PCI device probe

["Daniel Walker (danielwa)" <danielwa@cisco.com>]

On Tue, Nov 19, 2024 at 07:28:16PM +0100, Hans de Goede wrote:
> Hi,
> 
> On 19-Nov-24 3:20 AM, Shinichiro Kawasaki wrote:
> > On Nov 18, 2024 / 17:15, Daniel Walker (danielwa) wrote:
> >> On Mon, Nov 18, 2024 at 05:00:52PM +0100, Hans de Goede wrote:
> >>> Hi,
> >>>
> >>> On 18-Nov-24 4:55 PM, Andy Shevchenko wrote:
> > [...]
> >>>> Hans, there will be no need to fix anything if they implement correct access
> >>>> to the GPIO, i.e. via driver and board code with GPIO lookup tables.
> >>>
> >>> Agreed, still I'm not sure how I feel about us hiding the previously unhidden P2SB.
> >>>
> >>> OTOH I guess it may have only been unhidden in the BIOS to make the hack they
> >>> are using possible in the first place.
> >>
> >> From a flexibility POV I would suggest if you can not hide it if it's not already
> >> hidden by the BIOS that would be better since some company may have a good
> >> reason to make a custom driver or to export the pci device to userspace thru
> >> UIO. The current situation is you can't make a custom driver if p2sb is enable
> >> with this additional patch even if you unhide the device inside the BIOS.
> >>
> >> In our case it seems like we could use the already existing solution with
> >> pinctrl, but others may not be able to do that or may not want to for different
> >> reasons.
> > 
> > I don't have strong opinion about the choice, but I wonder how the p2sb code
> > will be if we keep the unhidden P2SB. I created a trial patch below. If the
> > device is not hidden, it does not call pci_scan_single_device() and
> > pci_stop_and_remove_bus_device(). Instead, it calls pci_get_slot() and
> > pci_dev_put(). I don't have the environment which unhides P2SB. Daniel, if you
> > have time to afford, please try it out.
> 
> Thank you for looking into this.
> 
> Daniel can you give this a try? It should fix the regression you are seeing
> without needing to rework your code (reworking your code to be cleaner
> might still be a good idea though).
> 

The patch restores the device 8086:19dd back to lspci which is the pre commit
5913320eb behavior. So I would say it's working on my system.

Daniel