Hi,
a collegue has a 8GB quad-core Intel machine that shows similar symptoms.
The interesting thing is that Ubuntu 7.10 runs just fine on it.
# cat /proc/mtrr
reg00: base=0x00000000 ( 0MB), size=2048MB: write-back, count=1
reg01: base=0x80000000 (2048MB), size=1024MB: write-back, count=1
reg02: base=0xc0000000 (3072MB), size= 256MB: write-back, count=1
reg03: base=0xcfc00000 (3324MB), size= 4MB: uncachable, count=1
reg04: base=0x100000000 (4096MB), size=4096MB: write-back, count=1
reg05: base=0x200000000 (8192MB), size= 512MB: write-back, count=1
reg06: base=0x220000000 (8704MB), size= 128MB: write-back, count=1
dmesg on his machine shows this from e820:
[ 0.000000] BIOS-provided physical RAM map:
[ 0.000000] BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
[ 0.000000] BIOS-e820: 000000000009fc00 - 00000000000a0000 (reserved)
[ 0.000000] BIOS-e820: 00000000000e0000 - 0000000000100000 (reserved)
[ 0.000000] BIOS-e820: 0000000000100000 - 00000000cfa13000 (usable)
[ 0.000000] BIOS-e820: 00000000cfa13000 - 00000000cfa15000 (reserved)
[ 0.000000] BIOS-e820: 00000000cfa15000 - 00000000cfafe000 (usable)
[ 0.000000] BIOS-e820: 00000000cfafe000 - 00000000cfbe5000 (ACPI NVS)
[ 0.000000] BIOS-e820: 00000000cfbe5000 - 00000000cfbea000 (usable)
[ 0.000000] BIOS-e820: 00000000cfbea000 - 00000000cfbf3000 (ACPI data)
[ 0.000000] BIOS-e820: 00000000cfbf3000 - 00000000cfbf4000 (usable)
[ 0.000000] BIOS-e820: 00000000cfbf4000 - 00000000cfbff000 (ACPI data)
[ 0.000000] BIOS-e820: 00000000cfbff000 - 00000000cfc00000 (usable)
[ 0.000000] BIOS-e820: 00000000cfc00000 - 00000000d0000000 (reserved)
[ 0.000000] BIOS-e820: 00000000f0000000 - 00000000f8000000 (reserved)
[ 0.000000] BIOS-e820: 00000000fff00000 - 0000000100000000 (reserved)
[ 0.000000] BIOS-e820: 0000000100000000 - 000000022c000000 (usable)
So, situation is clear: memory is cached up to 8192 + 512 + 128MB but
usable memory is 8192 + 512 + 192MB. The top 64MB is uncached.
I remember a patch on LKML detecting this situation and limiting end_pfn.
Maybe the Ubuntu kernel has this patch.
Best regards,
Zoltán Böszörményi
Frank Cox írta:
On Mon, 18 Feb 2008 12:44:43 -0500
Bill Davidsen <davidsen@xxxxxxx> wrote:
Looking at your mtrr output, reg5 and reg6 seem to point to the missing
memory, but it's not being used. I'm just not enough of a memory
hardware expert to quite understand what's happening.
I managed to find some discussion of this issue (or one very much like it) on
lkml.org. It sounds like this is a common problem on certain Intel
motherboards, and it sounds like this is known to Intel as well.
I guess there's no fix for it at the moment, other than feeding the kernel a
mem= parameter on boot.
