Re: [PATCH v2 1/8] mmc: sdhci-pxav2: add initial support for PXA168 V1 controller

[Doug Brown <doug@xxxxxxxxxxxxx>]

Hi Adrian,

On 12/22/2022 8:03 AM, Adrian Hunter wrote:
> On 2/12/22 05:13, Doug Brown wrote:
> > Add a new compatible string for the version 1 controller used in the
> > PXA168, along with necessary quirks. Use a separate ops struct in
> > preparation for a silicon bug workaround only necessary on V1.
> >
> > Signed-off-by: Doug Brown <doug@xxxxxxxxxxxxx>
> > ---
> >   drivers/mmc/host/sdhci-pxav2.c | 18 +++++++++++++++++-
> >   1 file changed, 17 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/mmc/host/sdhci-pxav2.c b/drivers/mmc/host/sdhci-pxav2.c
> > index f18906b5575f..2f9fa0ecbddd 100644
> > --- a/drivers/mmc/host/sdhci-pxav2.c
> > +++ b/drivers/mmc/host/sdhci-pxav2.c
> > @@ -101,6 +101,14 @@ static void pxav2_mmc_set_bus_width(struct sdhci_host *host, int width)
> >   	writeb(ctrl, host->ioaddr + SDHCI_HOST_CONTROL);
> >   }
> >   
> > +static const struct sdhci_ops pxav1_sdhci_ops = {
> > +	.set_clock     = sdhci_set_clock,
> > +	.get_max_clock = sdhci_pltfm_clk_get_max_clock,
> > +	.set_bus_width = pxav2_mmc_set_bus_width,
> > +	.reset         = pxav2_reset,
> > +	.set_uhs_signaling = sdhci_set_uhs_signaling,
> > +};
> > +
> >   static const struct sdhci_ops pxav2_sdhci_ops = {
> >   	.set_clock     = sdhci_set_clock,
> >   	.get_max_clock = sdhci_pltfm_clk_get_max_clock,
> > @@ -114,6 +122,9 @@ static const struct of_device_id sdhci_pxav2_of_match[] = {
> >   	{
> >   		.compatible = "mrvl,pxav2-mmc",
> >   	},
> > +	{
> > +		.compatible = "mrvl,pxav1-mmc",
> > +	},
> >   	{},
> >   };
> >   MODULE_DEVICE_TABLE(of, sdhci_pxav2_of_match);
> > @@ -208,7 +219,12 @@ static int sdhci_pxav2_probe(struct platform_device *pdev)
> >   			host->mmc->pm_caps |= pdata->pm_caps;
> >   	}
> >   
> > -	host->ops = &pxav2_sdhci_ops;
> > +	if (match && of_device_is_compatible(dev->of_node, "mrvl,pxav1-mmc")) {
> > +		host->quirks |= SDHCI_QUIRK_NO_BUSY_IRQ | SDHCI_QUIRK_32BIT_DMA_SIZE;
> > +		host->ops = &pxav1_sdhci_ops;
> > +	} else {
> > +		host->ops = &pxav2_sdhci_ops;
> > +	}
>
> It would be better to put the information above in a structure and
> get it with of_device_get_match_data() (instead of of_match_device).
> Also drivers typically assume there is always a match since that
> is the only way the driver ->probe() will get run.

Thanks for all of your great feedback on this series. That makes sense.
I did have one question about this suggestion. There are other parts of
sdhci_pxav2_probe() that don't assume there was a match so that it can
be set up the old way as a platform_device without CONFIG_OF. I was
trying to preserve compatibility by defaulting to pxav2_sdhci_ops if
it was set up as a platform_device. Is it all right if I leave a
fallback in place for that, or should I just end compatibility with the
old way at this point and assume a match in all cases? I don't see any
legacy board files that use this driver.

> >   
> >   	ret = sdhci_add_host(host);
> >   	if (ret)

Thanks,
Doug