[Letux-kernel] X1600 / LX16 support - here: adding MMC

[H. Nikolaus Schaller]

> Am 08.02.2024 um 12:42 schrieb Paul Boddie <paul at boddie.org.uk>:
> 
> On Thursday, 8 February 2024 12:03:50 CET H. Nikolaus Schaller wrote:
>> 
>> Indeed. It is a patch "hack: disable DMA for mmc" in feature branch
>> letux/jz4730-v12
>> 
>> https://git.goldelico.com/?p=letux-kernel.git;a=commit;h=bc37b81a83b350a2897
>> 604b7e2591863b26673cd
>> 
>> Otherwise we can't boot the jz4730...
> 
> I will eventually revisit the JZ4730 with Linux in this area because I have 
> been able to get DMA working on the JZ4730 in my L4Re-based system. I just 
> need to demonstrate it working with MMC first, although there may be obvious 
> issues in the Linux driver that I might be able to identify already.
> 
> I briefly looked at the MMC driver last night, but I didn't get it to do more 
> than report initialisation with DMA enabled, as previously shown. Weirdly, if

I got his now (with rootwait):

[    0.000203] input: gpio_keys as /devices/platform/gpio_keys/input/input0
[    0.000203] jz4740-mmc 13450000.mmc: Ingenic SD/MMC card driver registered
[    0.000203] jz4740-mmc 13450000.mmc: Using DMA, 4-bit mode
[    0.000203] clk: Not disabling unused clocks
[    0.000203] jz4740-mmc 13460000.mmc: Ingenic SD/MMC card driver registered
[    0.000203] jz4740-mmc 13460000.mmc: Using DMA, 4-bit mode
[    0.000203] Waiting for root device /dev/mmcblk0p2...
[    0.000203] jz4740-mmc 13460000.mmc: no support for card's volts
[    0.000203] mmc1: error -22 whilst initialising SDIO card

that is -EINVAL

[    0.000203] mmc0: host does not support reading read-only switch, assuming write-enable
[    0.000203] mmc0: error -145 whilst initialising SD card

that is -ETIMEDOUT. So the card is not responding.

I also tried with a second µSD in the second SD slot but it was not recognised
at all (same error -22).

What I also got was one -EIO

[    0.000203] mmc0: error -5 whilst initialising SD card

> I added the vmmc-supply property along with a regulator, it wouldn't report 
> initialisation at all, only doing so on MMC1 which didn't have the property 
> specified, so I imagine there is something I don't understand about the MMC 
> framework. (Again, the documentation is not very helpful, and it also seems 
> that voltage-range is supported by some drivers and not others.)
> 
> I do wonder if I should check that the clock management is working. That code 
> isn't entirely finished because of the different divider computation that the 
> X1600 seems to need.

Yes, clock setup might be a reason.

I now plan to trace this -EINVAL on mmc1...

BR,
Nikolaus