On 10/16/10 5:48 PM, Keith Hui wrote: >> We want microcode updates to be as close as possible to upstream > How upstream are we talking about? :P These updates come from Intel as > one big .h file. ;-)
I investigated the possibility to merge the intel cpu drivers in that regard and looked at pulling in the complete intel microcode update file in. However, it's far too big in size for that and there's really no gain in pulling in a lot of blob that never gets used. If you look at other BIOSes you will find out that they carefully select which microcode to include, and they do that for a good reason, as do we. > But a bit of scheduling headache then awaits. How much CPU init do we > need to do before we start drawing things in from the cbfs? Right now we don't do any. > Is there > anything that cannot be done until microcode updates are loaded? Is > cpu_dev_ops.init() an appropriate place to access cbfs? If not, from > where? That's where we load microcode right now, and I suggest not changing that unless there is a reason to do so. Stefan -- coreboot mailing list: [email protected] http://www.coreboot.org/mailman/listinfo/coreboot
