Re: [PATCH v3 0/3] PCI: of: update endpoint ranges dynamically

[Andrea della Porta <andrea.porta@suse.com>]

Hi,

On 10:49 Wed 14 Oct     , Andrea della Porta wrote:
> Hi Alex,
> 
> On 17:30 Tue 13 Oct     , Alex Elder wrote:
> > On 10/9/26 8:41 AM, Andrea della Porta wrote:
> > > Hi Alex,
> > > 
> > > On 10:55 Thu 08 Oct     , Alex Elder wrote:
> > > > On 10/8/26 10:45 AM, Andrea della Porta wrote:
> > > > > After a quick test, it seems that it's not working out of the
> > > > > box on my RPi5. Let me investigate a bit to see what's wrong.
> > > 
> > > I stand corrected: it works perfectly. So:
> > > 
> > > Tested-by: Andrea della Porta <andrea.porta@suse.com>
> > 
> > Sorry for the delay.  I'll add the above to v4.
> > 
> > > I think it should even be possible to tweak the static RPi5
> > > DT in order to make it work dynamically with your patch. Can
> > 
> > I suspected that your static definitions *happen* to work,
> > but I never took the time to confirm that and understand
> > why.
> 
> As per your design and as expected, the static range is preserved
> since the conditionals make the function exit if the device is a
> bridge or it already has a ranges property,o it all works as before.
> Right now the RP1 is acknowledged as a PCI bridge.
> what I meant is that we can maybe amend the RPi5 DT getting rid
> of the static ranges and tweaking the reg offset in the subperipheral
> nodes. I will do some experiments with this, but not now in order
> to not clash with current downstream DT. Definitely a thing I'd
> like to try in teh long run, though.

After some experiments, invoking of_pci_make_dev_node() as quirks
for RP1 PCI endpoint and getting rid of the ranges property in the
endpoint node makes everything working fine, which is very nice.

Regards,
Andrea

> 
> Cheers,
> Andrea
> 
> > 
> > > you please provide the relevant PCI DT nodes (from PCI controller
> > > to pci-ep-bus endpoint) from the output of:
> > > 
> > > dtc -I fs -O dts /sys/firmware/devicetree/base/
> > > 
> > > taken on your reference board?
> > 
> > I'll send you that in a private message.
> > 
> > Thanks again.
> > 
> > 					-Alex
> > 
> > > 
> > > Many thanks for your work,
> > > Andrea
> > > 
> > > > 
> > > > THANK YOU FOR TESTING THIS!!!
> > > > 
> > > > My hope was that the recent change (to avoid doing any
> > > > updating of ranges property unless it was found to be
> > > > "empty") would avoid any trouble with the RPi5.
> > > > 
> > > > I'm eager to hear what more you find.  Please let me
> > > > know if I can be of any other help.
> > > > 
> > > > 					-Alex
> > > 
> >