Re: [PATCH] PCI: Add ACS errata for Pericom PI7C9X2G404 switch

[Bjorn Helgaas <helgaas@kernel.org>]

On Thu, Sep 09, 2021 at 06:08:33PM +1000, Nathan Rossi wrote:
> On Thu, 9 Sept 2021 at 08:24, Bjorn Helgaas <helgaas@kernel.org> wrote:
> >
> > [+cc Alex, beginning of thread:
> > https://lore.kernel.org/r/20210903034029.306816-1-nathan@nathanrossi.com]
> >
> > On Mon, Sep 06, 2021 at 04:01:20PM +1000, Nathan Rossi wrote:
> > > On Fri, 3 Sept 2021 at 16:18, Lukas Wunner <lukas@wunner.de> wrote:
> > > >
> > > > On Fri, Sep 03, 2021 at 03:40:29AM +0000, Nathan Rossi wrote:
> > > > > The Pericom PI7C9X2G404 PCIe switch has an errata for ACS P2P Request
> > > > > Redirect behaviour when used in the cut-through forwarding mode. The
> > > > > recommended work around for this issue is to use the switch in store and
> > > > > forward mode.
> >
> > Is there a URL for this erratum?  What is the issue?  Does the switch
> 
> Unfortunately the document is not public, it was provided under a NDA.
> However with that said, there is very little additional information in
> the document itself compared to the information provided in the commit
> message/code comments here. The only other information in the document
> that may be applicable is that the whole document is for a number of
> Pericom switch models, however I do not have access to the other two
> switch models and thus cannot validate if this fixup would also apply
> to them.
> 
> For reference the models with PCI IDs:
> - PI7C9X2G404 - 12d8:2404
> - PI7C9X2G304 - 12d8:2304
> - PI7C9X2G303 - 12d8:2303

I assume that running all these models in store and forward mode is
safe, even if it's not the highest-performance config.  If so, I'd
prefer to include all the documented Device IDs in the quirk so people
don't have to stumble over them before we can fix them.

If somebody complains about the performance and can verify that a
device *doesn't* need the quirk, we can remove it then.

Bjorn