The problem occurs with 1GB, 2GB or 4GB
of memory installed for sure. I do not know about the HW memory hole. Can you
explain what that is?
Alan Mimms, Senior Architect
F5 Networks, Inc.
v: 509-343-3524 f: 509-343-3501
From: Lu, Yinghai
[mailto:yinghai.lu@amd.com]
Sent: Wednesday, October 04, 2006
12:05 PM
To: Alan Mimms;
linuxbios@linuxbios.org; Andi Kleen
Subject: RE: [LinuxBIOS] ACPI NVS
and last 64k-ish area of RAM?
What’s your total RAM installed? 4G
or more. With HW memory hole enable?
YH
From:
linuxbios-bounces@linuxbios.org [mailto:linuxbios-bounces@linuxbios.org] On Behalf Of Alan Mimms
Sent: Wednesday, October 04, 2006
11:28 AM
To: linuxbios@linuxbios.org
Subject: [LinuxBIOS] ACPI NVS and
last 64k-ish area of RAM?
We have AMD dual Opteron hardware with AMD 8131+8111
chipsets attached. Using LinuxBIOS, we have a slight problem, that
APPEARS to be related to the last 64Kbytes of RAM. Our kboot based environment,
running in 32 bit instruction set, seems to randomly crash, and the implicated
area of memory is this last 64KB.
When we run a commercial BIOS on nearly identical
hardware, we see that that BIOS has created in the E820 table an ACPI
Non-Volatile-Storage area covering this last 64KB. LinuxBIOS is NOT doing
that; LinuxBIOS is treating all of the space as simple USABLE space.
In trying to figure this out, we have used the AMD HDT
tool to read the last 64KB. We (SOMETIMES) the system crashes when we read
this area using HDT.
Can someone please explain what this area is for and why
it’s strange to read even using a hardware debugging tool? Is it
REALLY in use for the ACPI NVS, and can we simply tell Linux to ignore it (map
it out) by creating an entry in E820 table so it won’t be used (we
don’t use ACPI suspend/resume)?
Thanks very much for any information.
Alan Mimms, Senior Architect
F5 Networks, Inc.
v: 509-343-3524 f: 509-343-3501