Skip to contentSkip to conversation

memory beyond4GB invisible to the system even though CONFIG_HIGHMEM64G=y

DISCUSSION James Kirin 3 participants 10 messages Jan 14, 2009 – Jan 15, 2009

lore
Share & tools
lore
#0 lore raw

Hello,

I am running kernel 2.6.28 on a computer with 8GB RAM. The BIOS detects all 8GB, yet

jk@as12:~$ free
             total       used       free     shared    buffers     cached
Mem:       3622504     799696    2822808          0      90936     369604
-/+ buffers/cache:     339156    3283348
Swap:      6867776          0    6867776

I compiled the kernel with the CONFIG_HIGHMEM64G=y and CONFIG_RESOURCES_64BIT=y options.

I have attached the kernel config, dmesg output and /proc/cpuinfo.

How can I use all the memory on this computer?

Thank you in advance for any help,

JamesK

#1 lore raw
Quoted 8 lines

I am running kernel 2.6.28 on a computer with 8GB RAM. The BIOS detects all 8GB, yet

jk@as12:~$ free
             total       used       free     shared    buffers     cached
Mem:       3622504     799696    2822808          0      90936     369604
-/+ buffers/cache:     339156    3283348
Swap:      6867776          0    6867776

The message in your kernel log:

[    0.000000] WARNING: BIOS bug: CPU MTRRs don't cover all of memory, losing 4096MB of RAM.
[    0.000000] ------------[ cut here ]------------
[    0.000000] WARNING: at arch/x86/kernel/cpu/mtrr/main.c:1662 mtrr_trim_uncached_memory+0x283/0x2a4()

is a clue. You might want to look for a BIOS update for your motherboard.

- R.

#2 lore raw

Thank you for your reply, Roland. Unfortunately, this continues happening after having installed the most recent version of the BIOS available from Dell.

I understand this is a problem in the BIOS and not the kernel, but is there a way to have the kernel work around this bug in the BIOS and make that memory accessible? Eg,

http://www.gossamer-threads.com/lists/linux/kernel/991748#991748

The difference seems to be that the BIOS on this machine I am using does list 8GB memory available.

Thanks for any help,

James

Quoted 20 lines · Roland Dreier
Quoted 8 lines

I am running kernel 2.6.28 on a computer with 8GB RAM. The BIOS detects all 8GB, yet

jk@as12:~$ free
             total       used       free     shared    buffers     cached
Mem:       3622504     799696    2822808          0      90936     369604
-/+ buffers/cache:     339156    3283348
Swap:      6867776          0    6867776

The message in your kernel log:

[    0.000000] WARNING: BIOS bug: CPU MTRRs don't cover all of memory, losing 4096MB of RAM.
[    0.000000] ------------[ cut here ]------------
[    0.000000] WARNING: at arch/x86/kernel/cpu/mtrr/main.c:1662 mtrr_trim_uncached_memory+0x283/0x2a4()

is a clue. You might want to look for a BIOS update for your motherboard.

- R.

#3 lore raw
Quoted 9 lines · James Kirin

Thank you for your reply, Roland. Unfortunately, this continues happening after having installed the most recent version of the BIOS available from Dell.

I understand this is a problem in the BIOS and not the kernel, but is there a way to have the kernel work around this bug in the BIOS and make that memory accessible? Eg,

http://www.gossamer-threads.com/lists/linux/kernel/991748#991748

please do
1. boot linux with "disable_mtrr_trim" in grub.conf etc
2. after booting input:
echo "base=0x100000000 size=0x20000000 type=write-back" >/proc/mtrr

later you could put that line in one scripts and call it from inittab

# grep mtrrfixup /etc/inittab
mtrr:345:once:/root/mtrrfixup.sh

YH

#4 lore raw
Quoted 4 lines

please do
1. boot linux with "disable_mtrr_trim" in grub.conf etc
2. after booting input:
echo "base=0x100000000 size=0x20000000 type=write-back" >/proc/mtrr

Is there any way we can have the kernel do this automatically if there are any free mtrr slots? I would guess this type of BIOS bug is pretty common (desktop boards not validated with > 4GB of RAM), and having the kernel throw away 4GB of memory unless the user does some fancy scripting is not very friendly (not to mention the fact that things are likely to be very slow until the mtrr fixup script runs).

- R.

#5 lore raw
Quoted 11 lines · Roland Dreier
Quoted 4 lines

please do
1. boot linux with "disable_mtrr_trim" in grub.conf etc
2. after booting input:
echo "base=0x100000000 size=0x20000000 type=write-back" >/proc/mtrr

Is there any way we can have the kernel do this automatically if there are any free mtrr slots? I would guess this type of BIOS bug is pretty common (desktop boards not validated with > 4GB of RAM), and having the kernel throw away 4GB of memory unless the user does some fancy scripting is not very friendly (not to mention the fact that things are likely to be very slow until the mtrr fixup script runs).

could do it. aka make mtrr and e820 consistent.
it depends which one that we trust mtrr or e820.

YH

#6 lore raw
In reply to Yinghai Lu

Hi Yinghai,

Quoted 4 lines

please do
1. boot linux with "disable_mtrr_trim" in grub.conf etc
2. after booting input:
echo "base=0x100000000 size=0x20000000 type=write-back" >/proc/mtrr

Thank you for your suggestion.

Unfortunately, booting with disable_mtrr_trim leads to a kernel panic:

[<c0103abb>] kernel_thread_helper+0x7/0x1c
===========
Code: 0c 31 d2 89 e8 e8 c2 f4 ff ff 89 c3 eb 8e 8b 85 cc 00 00 00 89 44 24 10 83
bd ec 00 01 00 04 0f 84 cf 00 00 00 8b 9d f0 00 01 00 <8b> 43 10 3b 43 0c 0f 84
d5 00 00 00 8b 53 10 8b 85 ec 00 01 00
EIP: [<c026bc2f] hid_parser_main+0xcf/0x290 SS:ESP 0068:f7c2bd7c
---[ end trace 34615b445589fbb6 ]---
Kernel panic - not syncing: Attempted to kill init!

I am attaching the kernel config file. The system boots fine if I remove "disable_mtrr_trim" from lilo.conf (and rerun lilo).

Thank you for any help,

James

#7 lore raw

I tried disabing CONFIG_MTRR_SANITIZER and CONFIG_X86_PAT in the kernel, and now a different kernel panic occurs:

[... ] Write protecting the kernel text: 3040k
[...] Write protecting the kernel read-only data: 1352k [... ] khelper used greatest stack depth: 6724 bytes left [ ...] Kernel panic - not syncing: No init found. Try passing init= option to kernel.

Is an "init=" option necessary when using disable_mtrr_trim?

Thanks in advance,

James

Quoted 27 lines · James Kirin

Hi Yinghai,

Quoted 4 lines

please do
1. boot linux with "disable_mtrr_trim" in grub.conf etc
2. after booting input:
echo "base=0x100000000 size=0x20000000 type=write-back" >/proc/mtrr

Thank you for your suggestion.

Unfortunately, booting with disable_mtrr_trim leads to a kernel panic:

[<c0103abb>] kernel_thread_helper+0x7/0x1c
===========
Code: 0c 31 d2 89 e8 e8 c2 f4 ff ff 89 c3 eb 8e 8b 85 cc 00 00 00 89 44 24 10 83
bd ec 00 01 00 04 0f 84 cf 00 00 00 8b 9d f0 00 01 00 <8b> 43 10 3b 43 0c 0f 84
d5 00 00 00 8b 53 10 8b 85 ec 00 01 00
EIP: [<c026bc2f] hid_parser_main+0xcf/0x290 SS:ESP 0068:f7c2bd7c
---[ end trace 34615b445589fbb6 ]---
Kernel panic - not syncing: Attempted to kill init!

I am attaching the kernel config file. The system boots fine if I remove "disable_mtrr_trim" from lilo.conf (and rerun lilo).

Thank you for any help,

James

#8 lore raw
Quoted 10 lines · James Kirin

I tried disabing CONFIG_MTRR_SANITIZER and CONFIG_X86_PAT in the kernel, and now a different kernel panic occurs:

[... ] Write protecting the kernel text: 3040k
[...] Write protecting the kernel read-only data: 1352k [... ] khelper used greatest stack depth: 6724 bytes left [ ...] Kernel panic - not syncing: No init found. Try passing init= option to kernel.

Is an "init=" option necessary when using disable_mtrr_trim?

that is not related...

YH

#9 lore raw
In reply to James Kirin
Quoted 20 lines · James Kirin

Hi Yinghai,

Quoted 4 lines

please do
1. boot linux with "disable_mtrr_trim" in grub.conf etc
2. after booting input:
echo "base=0x100000000 size=0x20000000 type=write-back" >/proc/mtrr

Thank you for your suggestion.

Unfortunately, booting with disable_mtrr_trim leads to a kernel panic:

[<c0103abb>] kernel_thread_helper+0x7/0x1c
===========
Code: 0c 31 d2 89 e8 e8 c2 f4 ff ff 89 c3 eb 8e 8b 85 cc 00 00 00 89 44 24 10 83
bd ec 00 01 00 04 0f 84 cf 00 00 00 8b 9d f0 00 01 00 <8b> 43 10 3b 43 0c 0f 84
d5 00 00 00 8b 53 10 8b 85 ec 00 01 00
EIP: [<c026bc2f] hid_parser_main+0xcf/0x290 SS:ESP 0068:f7c2bd7c
---[ end trace 34615b445589fbb6 ]---
Kernel panic - not syncing: Attempted to kill init!
Quoted 3 lines

I am attaching the kernel config file. The system boots fine if I remove "disable_mtrr_trim" from lilo.conf (and rerun lilo).

can you try mem=4g , .., and then mem=6g ...

YH