Re: [PATCH 2/2] PCI: qcom: Sort variants by Qcom IP rev

[Bjorn Helgaas <helgaas@kernel.org>]

On Tue, Jul 26, 2022 at 02:45:34PM +0200, Johan Hovold wrote:
> On Fri, Jul 22, 2022 at 10:49:19AM -0500, Bjorn Helgaas wrote:
> > From: Bjorn Helgaas <bhelgaas@google.com>
> > 
> > Previously the variant resource structs, ops, etc., were in no obvious
> > order (mostly but not consistently in *Synopsys* IP rev order, which is not
> > reflected in the naming).
> > 
> > Reorder them in order of the struct and function names.  No functional
> > change intended.
> > 
> > Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
> > ---
> >  drivers/pci/controller/dwc/pcie-qcom.c | 732 ++++++++++++-------------
> >  1 file changed, 366 insertions(+), 366 deletions(-)
> > 
> > diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c
> > index c27e3494179f..d0237d821323 100644
> > --- a/drivers/pci/controller/dwc/pcie-qcom.c
> > +++ b/drivers/pci/controller/dwc/pcie-qcom.c
> 
> Moving code around like this makes code forensics harder as it messes up
> git blame. At least the callbacks appears to be grouped by IP version
> currently, so not sure how much you gain from moving the callbacks
> around.

The existing hodge-podge is sloppy and makes code reading harder for
everybody.  If we want them grouped by IP version, they should be
*named* by IP version.

> > -static const struct qcom_pcie_cfg sc8180x_cfg = {
> > -	.ops = &ops_1_9_0,
> > -	.has_tbu_clk = true,
> > -};
> > -
> >  static const struct qcom_pcie_cfg ipq6018_cfg = {
> >  	.ops = &ops_2_9_0,
> >  };
> 
> But this bit I disagree with. Why sort the SoCs configurations by IP
> revision, when what you typically need is to look them up by name?

Makes sense.

> Also note that this conflicts with my sc8280xp-support and IP-revision
> series:
> 
> 	https://lore.kernel.org/all/20220714071348.6792-1-johan+linaro@kernel.org/
> 
> The result of applying that series is that these structs are renamed
> after the IP revision (and sorted alphabetically) so the end-result is
> similar.
> 
> Could you consider dropping this patch, or at least the struct
> qcom_pcie_cfg bits, and applying the above series for 5.20?

I dropped it for now.  We can see how it shakes out after your series,
but not sure I'll get to it for this cycle.

Bjorn