[coreboot] Kconfig vs. devicetree vs. CMOS policy for options?

[Scott Duplichan]

]Peter Stuge wrote:
]
]Reading this I think that there should be a Kconfig option to choose
]if the chipset should be set up as SATA IDE or AHCI.
]
]Talking to a lot of visitors at LinuxTag it is absolutely clear that
]this is an example of what should actually be an NVRAM option.
]
]Do we have some policy for where to place an option? I don't think we
]do. Do we want to create one?
]
]The purpose is to have a perfectly streamlined user experience across
]all different mainboards. Of course all boards don't support all
]options, but when two different boards *do* support an option, that
]option must be in the same place, working the same way.
]
]The balance between compile time options and NVRAM options is not so
]easy. :\

Hello Peter,

Right now, the sata controller is hard-coded to use the AHCI software
interface, and the IDE controller is hidden. I think for the most
part, AHCI mode should be OK for every use. But certainly at least a
compile option for the IDE software interface is needed. Would a 
kconfig option make sense? I am not familiar with coreboot nvram 
use.

In what situation is AHCI undesirable? For me, the answer is older
editions of windows. If you happen to want to test windows xp setup
using a standard setup CD, windows will not find the drives because
it has no AHCI support. The standard solution is the F6 floppy method
of adding an AHCI driver, but lack of floppy support on new boards 
makes this method difficult. I use the http://www.nliteos.com/ tool
to make a custom setup CD. But this method requires a new custom CD
for each chipset. The ability to disable AHCI is certainly a good
feature to have when doing a quick test of an older OS. The OS I use
for my own development machine is windows server 2003 x64 edition.
This OS has no in-box AHCI driver, so I am familiar with the hassle.

Thanks,
Scott