On 5/27/10 3:50 AM, Keith Hui wrote:
A question. How is one supposed to make use of this script?
$ cd src/cpu/intel/microcode $ sh update-microcode.sh
Then you cd to your CPU model directory and check what's there.
As part of my effort to re-enable L2 cache on SECC Pentium 2/3s I'd also split off model_63x, model_65x and model_67x (patch submitted earlier and fell into limbo) and introduce all available microcode to the family.
If a model_635 directory is there, the script will copy a model 635 microcode there. Next it would try model_63x, then model_6xx, then delete the file.
I'd suggest that you wipe old existing microcode files before running the script so you see which ones are outdated and which ones are new.
Actually I would like to see one folder and one microcode update file for one family of CPUs. Less overhead on so many fronts - number of files, compiler processing and license headers.
The files are all automatically generated (now) so it's not really an issue. The way it's done now with single files allows you to comment in/out single microcodes.
Generally, mid term I want to move microcode to CBFS and wipe it completely from the coreboot source tree. Instead coreboot should be able to parse an intel microcode file specified in Kconfig, or it could download it automatically and use it. This way we don't have to carry around potentially troublesome binary hunks.
Stefan