[Letux-kernel] jz4730 Ethernet

[H. Nikolaus Schaller]

Hi Paul,

> Am 09.03.2021 um 21:43 schrieb Paul Boddie <paul at boddie.org.uk>:
> 
> On Tuesday, 9 March 2021 17:46:24 CET H. Nikolaus Schaller wrote:
>> Hi Paul,
>> have you had success building and booting an SD card?
> 
> No, not yet. I tried putting the new U-Boot in the FAT partition as uImage 
> alongside the uncompressed version as u-boot.bin, also adding the bootargs.scr 
> and boot.scr files, deploying the kernel as Image, and having the device tree 
> files inside the ingenic directory, but I still didn't get it to boot.
> 
> I probably won't get to look at this for a few days now, and when I do, I 
> guess I'll just copy your binaries for the above things to the card to see if 
> there is something fundamentally wrong.

It is very likely only a small difference. Maybe just 1 bit :)

> 
> [...]
> 
>> I have pushed a new version (same link):
>> 
>> * based on v5.12-rc2
>> * based on latest fixes for CI20
>> * includes patches for pinctrl to support MII (based on patches by Yanjie
>> * code made compile (with remaining warnings)by fixing some things and
>> commenting out some parts
>> 
>> It loads but then:
>> 
>> [   53.148964] genirq: Flags mismatch irq 19. 00000003 (eth0) vs. 00000003
>> (matrix-keypad)
>> [   53.157718] eth0: unable to get IRQ 19 .
> 
> What is the matrix-keypad part of this about?

It appears as if the kernel does assign some matrix-keypad gpio the same
interrupt number as the hard coded ethernet driver thinks.

I have added of_irq_get() and now this is gone and I've pushed an update.

> 
> [...]
> 
>> What needs to be fixed:
>> - missing: get IRQ from device tree
>> - missing: get hardware address of MII controller from device tree
>> - needs working DMA (at least the driver allocates DMA memory)
>> - needs substitute for a background thread to detect cable plug/unplug
>> - needs to resolve all FIXME
>> 
>> Still a lot of work until all gaps for the jz4730 are closed:
>> 
>> * futex (or what ever the sysbench and X11 issue is)
> 
> I should investigate LL/SC and see if it generates illegal instruction errors 
> under L4Re.

I did build and run the futex example from https://man7.org/linux/man-pages/man2/futex.2.html
but it shows no anomalies.

> 
>> * i2c
>> * sound
>> * dma
>> * mii
>> * dma based mmc
>> 
>> But we are closing them one-by-one and the list has become quite short :)
> 
> Yes, it looks like progress is being made.

Indeed.

BR,
Nikolaus