Am 06.04.2012 17:27, schrieb Marc Jones:
Can you be more descriptive to how it fails? Does it hang on that instruction?
That change might also break on future CPUs (if they finally manage to make the TPM stuff secure, so that's a big if)
Patrick
* Patrick Georgi patrick@georgi-clan.de [120406 17:32]:
Am 06.04.2012 17:27, schrieb Marc Jones:
Can you be more descriptive to how it fails? Does it hang on that instruction?
That change might also break on future CPUs (if they finally manage to make the TPM stuff secure, so that's a big if)
How so?
Guess I'll have to work making that a CONFIG option.
I'm not sure why it would be breaking anything, but it just doesn't make sense. From Volume 3: "Because all internal cache lines are invalid following reset initialization, it is not necessary to invalidate the cache before enabling caching." It is pretty weird, I downloaded one random Sandybridge ROM image from ECS, and I did not see the WBINVD before the far jmp.
Also, caches *should* remain valid after INIT, but this change will invalidate them unconditionally (unless INIT follows a different path in coreboot)
Tom
On Fri, Apr 6, 2012 at 11:26 AM, Stefan Reinauer stefan.reinauer@coreboot.org wrote:
- Patrick Georgi patrick@georgi-clan.de [120406 17:32]:
Am 06.04.2012 17:27, schrieb Marc Jones:
Can you be more descriptive to how it fails? Does it hang on that instruction?
That change might also break on future CPUs (if they finally manage to make the TPM stuff secure, so that's a big if)
How so?
Guess I'll have to work making that a CONFIG option.
-- coreboot mailing list: coreboot@coreboot.org http://www.coreboot.org/mailman/listinfo/coreboot
Am 06.04.2012 20:26, schrieb Stefan Reinauer:
That change might also break on future CPUs (if they finally manage to make the TPM stuff secure, so that's a big if)
How so?
Load top x KB into cache, let the CPU measure the data from cache into a PCR, run the code from cache (to avoid TOCTOU issues).
Patrick