It's really great what has become of coreboot!! Most visible to me (so far) is Kconfig which makes things a lot less scary than they used to be and Geode-LX VGA support.
The attached patch calls init() again.
Indeed, it does. Excellent, thanks a lot!!! :)
OK. Hopefully you're almost there.
Myles Watson wrote:
I think that the value that was left out of devicetree.cb should move to the mainboard init() or into the device that's being configured
(cs5536.)
In fact I was already wondering whether the whole code in mainboard init() is actually at the right place. That's one of the things I meant when saying "bring the 2 older board's code up to scratch first". The init() code does 3 things:
- set up misc CS5536 GPIOs
- configure the EC, which is a device of the IT8712F SIO. However config
isn't done via the usual SIO ports 2E/2F but via the EC's own I/O adresses.
- set misc IT8712 GPIOs, via the Simple-I/O port (with the config value
that was dropped from devicetree.cb)
Other parts of the SIO config however, e.g. assigning the EC and Simple- I/O base addresses is done in devicetree.cb whereas setting up which SIO GPIOs are outputs and which are mapped for Simple-I/O is done in romstage.c because parts of it are better be done early.
Yes. Since the device tree isn't available until RAM is initialized, early things can't live there.
This mess makes me think if there is a cleaner way and which parts belong where.
Great.
It makes sense to use values in the device tree to parameterize device code for specific mainboards, but I don't see the point of parameterizing mainboard-specific code from the device tree.
I on the other hand was intending to separate the configuration _data_ from the _code_ that writes it into the chip. The customers will still have to customize coreboot for their purpose. Some might want to move the COM1 from 3F8 to 3E8, because they have added an external RS232 on PC/104 (ISA). And some might disable LPT because they need the IRQ for something else. With a standard BIOS they can do it in the BIOS Setup. Here they can edit a common config _data_ file devicetree.cb. They don't have to mess with e.g. the superio/ite/it8712f/superio.c _code_. And switching COM1 from RS232 to RS485 mode can be done in the standard BIOS Setup, too. That's why I placed this mainboard config value in devicetree.cb, too, rather than having customers mess with the code in mainboard.c. I had guessed that this is the purpose why there is a struct mainboard_config in the mainboard's chip.h, just as there is a struct superio_..._config in all superio's chip.h. In what way is configuring hardware implemented inside a SIO chip so much different from configuring hardware implemented as a couple of discrete logic gates on the mainboard? Am I making any sense?? :)
Yes. You're right that the separation is important.
There is second mainboard-specific config value (SMC_CONFIG), which I'd have prefered in devicetree.cb, too. But I couldn't get this to work 2 years ago, because the code is run early from romstage.c ...
I think you need Kconfig variables for both of those things. The device tree isn't for options. It's for immutable properties of the mainboard.
I'd suggest that your options depend on the motherboard.
Thanks, Myles