Am Mittwoch, den 07.10.2009, 17:52 +0200 schrieb Maximilian Thuermer:
Hi all,
seems as if the switch to CBFS has broken the Tyan S2912 boards. It reports that CONFIG_XIP_ROM_BASE needs to be a multiple of CONFIG_XIP_ROM_SIZE. The fix in failovercalculation.ld seems to be pretty straight forward,
If you have a fix for inclusion to newconfig, we'll be happy to merge it, of course!
however after having read some posts on the mailing list, it seemed to me as if the whole Config.lb scheme of those boards are outdated (they still contain the failover sections in contrast to CBFS being incompatible with this).
CBFS is not incompatible to failover (or vice versa). It might not work, but that would be a bug to fix.
The Kconfig effort (which is separate from CBFS) does not implement failover, or any fallback/rescue mechanism at all at this time.
In addition to this, the boot process of fixed boards (with failover still enabled) hangs after finishing CAR which seems to fit into the picture (sorry, no log attached).
After finishing CAR? That would be very early. There are a couple of places (loading coreboot_ram and loading the payload) that can take a considerable amount of time (several minutes in some cases) on certain AMD boards. Maybe the board is simply very busy with copying the next stage? Of course this must be fixed, but first we'll have to know what to fix.
Does anyone have a thought on this?
Please post a log :-)
Regards, Patrick