> Date: Sun, 2 May 2021 15:46:30 +0000
> From: Visa Hankala <v...@hankala.org>
>
> On Sun, May 02, 2021 at 05:28:24PM +0200, Mark Kettenis wrote:
> > > Date: Sun, 2 May 2021 14:21:29 +0000
> > > From: Visa Hankala <v...@hankala.org>
> > >
> > > Zynq-7000 has a variant of Arasan SD controller that is not recognized
> > > by the fdt glue. The diff below fixes this.
> > >
> > > The controller's capabilities register lacks the base clock frequency.
> > > The attachment glue has to provide this parameter.
> > >
> > > OK?
> > >
> > > Index: dev/fdt/sdhc_fdt.c
> > > ===================================================================
> > > RCS file: src/sys/dev/fdt/sdhc_fdt.c,v
> > > retrieving revision 1.15
> > > diff -u -p -r1.15 sdhc_fdt.c
> > > --- dev/fdt/sdhc_fdt.c 5 Apr 2021 09:31:45 -0000 1.15
> > > +++ dev/fdt/sdhc_fdt.c 2 May 2021 14:07:45 -0000
> > > @@ -126,6 +126,7 @@ sdhc_fdt_match(struct device *parent, vo
> > > struct fdt_attach_args *faa = aux;
> > >
> > > return (OF_is_compatible(faa->fa_node, "arasan,sdhci-5.1") ||
> > > + OF_is_compatible(faa->fa_node, "arasan,sdhci-8.9a") ||
> > > OF_is_compatible(faa->fa_node, "brcm,bcm2711-emmc2") ||
> > > OF_is_compatible(faa->fa_node, "brcm,bcm2835-sdhci") ||
> > > OF_is_compatible(faa->fa_node, "marvell,armada-3700-sdhci") ||
> > > @@ -232,6 +233,11 @@ sdhc_fdt_attach(struct device *parent, s
> > > sc->sc.sc_flags |= SDHC_F_NODDR50;
> > > }
> > >
> > > + if (OF_is_compatible(faa->fa_node, "arasan,sdhci-8.9a")) {
> > > + freq = clock_get_frequency(faa->fa_node, "clk_xin");
> > > + sc->sc.sc_clkbase = (freq + 500) / 1000;
> >
> > Is there a particular reason why you are trying to round to the
> > nearest kHz here? We don't do that for any of the other cases where
> > we set sc_clkbase, and I'd prefer not to have any special cases that
> > aren't necessary.
>
> In my case, the rounding makes a displayed value look more sensible.
>
> sdhc0 at simplebus0
> sdhc0: SDHC 2.0, 100 MHz base clock
> sdmmc0 at sdhc0: 4-bit, sd high-speed, mmc high-speed, dma
>
> The clock computations use frequency 33333333 Hz as input. This value
> is multiplied and divided, giving 99999990 Hz (which already has some
> truncation). If this was truncated to MHz, the result would be 99 MHz.
>
> However, if the rounding is not wanted, I can leave it out.
I'd prefer it that way. ok kettenis@ with that change.
This will potentially result in a slight overclocking of the bus, but
it will be smaller than 1 kHz on a minimal clock rate of 400 kHz.
This should be well within the tolerances of the SD card
specification, so I don't expect any issues.
If we want to fix this, I'd suggest we change sc_clkbase to be in Hz
rather than kHz and maybe do some rounding when we print the value in
MHz. That way we avoid double rounding. But it isn't very important.
> > > + }
> > > +
> > > if (OF_is_compatible(faa->fa_node, "brcm,bcm2711-emmc2"))
> > > sc->sc.sc_flags |= SDHC_F_NOPWR0;
> > >
>
