[PATCH v4 0/3] BMG PCIe Gen5 downgrade attributes and usage

[Raag Jadav]

On Thu, May 01, 2025 at 01:10:49AM +0530, Vivi, Rodrigo wrote:
> On Tue, 2025-04-29 at 08:38 +0300, Raag Jadav wrote:
> > On Mon, Apr 28, 2025 at 04:09:05PM -0400, Rodrigo Vivi wrote:
> > > On Mon, Apr 28, 2025 at 04:00:44PM -0400, Rodrigo Vivi wrote:
> > > > On Mon, Apr 28, 2025 at 01:12:38PM +0300, Raag Jadav wrote:
> > > > > On Fri, Apr 25, 2025 at 07:36:23PM +0530, Raag Jadav wrote:
> > > > > > This series exposes sysfs attributes for BMG PCIe Gen5
> > > > > > downgrade and
> > > > > > documents their usage.
> > > > > 
> > > > > Anything I can do to move this forward?
> > > > 
> > > > I almost push it here, but then I noticed that it is
> > > > gen5_downgrade.
> > > > Hadn't we agreed to follow what spec says so?
> > > > 
> > > > "to then automatically persist the Gen4 downgrade flag in Flash"
> > > > "Write Gen4 Downgrade bit to MRC Flash File"
> > > > 
> > > > == Applying  PCIe Gen4 Downgrade ==
> > > > 
> > > > Although I see that there are some mentions calling "Gen5
> > > > downgrade", "Gen4 downgrade" seems to be the most used term in
> > > > the specs, specially when calling bits and
> > > > sections names...
> > 
> > Which is what I followed until we had a change of preference over gen
> > definition.
> > https://lore.kernel.org/intel-xe/34b33d3135fc24302db2764ce86a641e7c49054f.camel@intel.com/
> > 
> > > Because of the inconsistencies and our back and forth here and to
> > > get prepared
> > > for future cases where we might need to downgrade from gen6 to
> > > gen5, the current
> > > Architecture recommendation is to simply go with
> > > 
> > > so /sys/bus/pci/devices/<bdf>/pcie_gen_downgrade_{status,capable}
> > 
> > I really like Lucas' proposal, which is also consistent with similar
> > existing
> > attributes.
> > 
> > /sys/bus/pci/devices/<bdf>/auto_link_downgrade_capable
> 
> I'm sorry for the delay in the response, it took a while to get
> confirmation. We can go with this one. or with the one you sent
> already. Your call. Just let me know that I push the version you tell
> and tell the architects to update the spec.

I'm okay with either one. But since it was Lucas' idea, let's hear his
thoughts.

Lucas, can we move forward with auto_link_downgrade_capable?

Raag