Then Eric need to fix the MTRR for AMD...?
-----Original Message----- From: Li-Ta Lo [mailto:ollie@lanl.gov] Sent: Friday, January 14, 2005 6:58 AM To: YhLu Cc: Sagiv Yefet; LinuxBIOS Subject: RE: Running with VGA
On Thu, 2005-01-13 at 22:20, YhLu wrote:
The problem is no body touch that range. Even init_ecc_memory will skip
the
first 1M range.
No. The problem is amd/early_mtrr and amd/mtrr are wrong. The amd/early_mtrr does not enable MEM forwarding of A0000-BFFFF to main memory and amd/mtrr is called to late (it doesn't check if A0000-BFFFF should be forward to IO or MEM neither).
I fixed it a long time ago. Eric's CPU code just wipe them out.
Ollie
On Fri, 2005-01-14 at 10:36, YhLu wrote:
Then Eric need to fix the MTRR for AMD...?
He uses TOP_MEM instead of MTRR for caching in early_mtrr. I don't know how to fix this execpt rewrite the whole early_mtrr.
Probably we should add a dev_post() pass and put the ROM post into device_op::post?
Ollie
-----Original Message----- From: Li-Ta Lo [mailto:ollie@lanl.gov] Sent: Friday, January 14, 2005 6:58 AM To: YhLu Cc: Sagiv Yefet; LinuxBIOS Subject: RE: Running with VGA
On Thu, 2005-01-13 at 22:20, YhLu wrote:
The problem is no body touch that range. Even init_ecc_memory will skip
the
first 1M range.
No. The problem is amd/early_mtrr and amd/mtrr are wrong. The amd/early_mtrr does not enable MEM forwarding of A0000-BFFFF to main memory and amd/mtrr is called to late (it doesn't check if A0000-BFFFF should be forward to IO or MEM neither).
I fixed it a long time ago. Eric's CPU code just wipe them out.
Ollie