ron minnich wrote:
On Thu, Feb 12, 2009 at 8:35 AM, Stefan Reinauer stepan@coresystems.de wrote:
I'm not getting tired of repeating myself.:There should not be a reason to un-cache the upper SMM area, or the problem is elsewhere. There is definitely no issue with any of my SMM code, and there is no other SMM code in the public tree or on the mailing list. So once that problem will occur, we can solve that problem.
I strongly agree. We've got a working cached SMM, from Stefan, so we know that one can cache SMM. For the counterargument that SMM can not be cached, we have a hint from someone that someone else had trouble -- and we have no code to inspect -- sorry, that's not strong enough evidence for me to accept that we need another change to the MTRR code.
Actually my SMM code is uncached, too, because it runs in ASEG. I have a pretty clear idea what to do to get caching of the HSEG/TSEG working, if we really ever need it. (see other mails to the list a few days ago)
Right now the SMM handler only runs when called by ACPI (as in: display darker, display brighter buttoon pressed) or when trying to write to the BIOS without "permission" to do so. No regular calls are made (timer etc), so the impact of making sure we copy to HSEG and running cached from there is probably higher than just running 2 SMM calls uncached in the time from start-up to shut-down.
But let's get SMM for another platform into the tree and see what's best then.
Stefan