Re: [RFC HACK PATCH] PCI: dwc: layerscape: Hack around enumeration problems with Honeycomb LX2K

[Rob Herring]

On Mon, Dec 14, 2020 at 4:43 AM Daniel Thompson
<daniel.thompson@xxxxxxxxxx> wrote:
>
> On Fri, Dec 11, 2020 at 05:05:58PM +0000, Daniel Thompson wrote:
> > On Fri, Dec 11, 2020 at 08:37:40AM -0600, Rob Herring wrote:
> > > On Fri, Dec 11, 2020 at 6:15 AM Daniel Thompson
> > > >     BTW I noticed many other pcie-designware drivers take advantage
> > > >     of a function called dw_pcie_wait_for_link() in their init paths...
> > > >     but my naive attempts to add it to the layerscape driver results
> > > >     in non-booting systems so I haven't embarrassed myself by including
> > > >     that in the patch!
> > >
> > > You need to look at what's pending for v5.11, because I reworked this
> > > to be more unified. The ordering of init is also possibly changed. The
> > > sequence is now like this:
> > >
> > >         dw_pcie_setup_rc(pp);
> > >         dw_pcie_msi_init(pp);
> > >
> > >         if (!dw_pcie_link_up(pci) && pci->ops->start_link) {
> > >                 ret = pci->ops->start_link(pci);
> > >                 if (ret)
> > >                         goto err_free_msi;
> > >         }
> > >
> > >         /* Ignore errors, the link may come up later */
> > >         dw_pcie_wait_for_link(pci);
> >
> > Thanks. That looks likely to fix it since IIUC dw_pcie_wait_for_link()
> > will end up waiting somewhat like the double check I added to
> > ls_pcie_link_up().
> >
> > I'll take a look at let you know.
>
> Yes. These changes have fixed the enumeration problems for me.
>
> I tested pci/next and I cherry picked your patch series onto v5.10 and
> both are working well.
>
> Given this fixes a bug for me, do you think there is any scope for me
> to whittle down your series into patches for the stable kernels or am
> I likely to find too many extra bits being pulled in?

I think I'd just go the adding a delay route. It's a fairly big series
and depends on my other clean-up done in 5.10. And there's at least
some possibility it regresses some platform given the limited testing
linux-next gets.

Rob