Add the SDHCI Address Decoder registers and routines for ARMADA 38X.

[loos]

Summary

Add the SDHCI Address Decoder registers and routines for ARMADA 38X.

Sponsored by: Rubicon Communications, LLC (Netgate)

Test Plan

Tested on ClearFog Pro.

[mw_semihalf.com]

sys/arm/mv/mv_common.c

2090 ↗

(On Diff #28035)

Hi,

I know e.g. usb3 is done exactly same, but I think the only place we should use the ifdef is here:

static int
decode_win_sdhci_valid(void)
{
#ifdef SOC_MV_ARMADA38X

return (decode_win_can_cover_ddr(MV_WIN_SDHCI_MAX));

#endif

/* Satisfy platforms not equipped with this controller */
return (1);

}

The registers definition are not under ifdef, so e.g. kirkwood should compile easily. _setup and _dump functions execution is depending on the compatible string, so it can be safe as well. In further perspective, I think also _valid check should depend on the compatible, so that we don't have to add dummy stuff. What do you think?

[loos]

Remove the #ifdef for decode_win_sdhci_valid() as requested by Marcin at Semihalf.

sys/arm/mv/mv_common.c

2090 ↗

(On Diff #28035)

Fixed in the latest update. Thanks!

[mw_semihalf.com]

sys/arm/mv/mv_common.c

669 ↗

(On Diff #28162)

Thanks for the update. However, if you move WIN_REG_BASE_IDX_ macros out of #if defined(SOC_MV_ARMADA38X)...

2128 ↗

(On Diff #28162)

... you will be able to remove ifdef around decode_win_sdhci_setup/dump and the dummy functions below. We won't break Kirkwood or ArmadaXP compilation, and because the routines won't execute on those platforms, because they require compatible string match.

[loos]

Move the MACROs definition outside the #ifdefs and remove the now unnecessary #ifdefs.

Spotted by: [e-mail]