Bug 14698 - only 3 GB RAM from 4 GB RAM with 64 bit OS
Summary: only 3 GB RAM from 4 GB RAM with 64 bit OS
Status: CLOSED DOCUMENTED
Alias: None
Product: ACPI
Classification: Unclassified
Component: Other (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: Zhang Rui
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-11-26 20:34 UTC by Alex Fera
Modified: 2010-01-21 07:59 UTC (History)
3 users (show)

See Also:
Kernel Version: 2.6.31-ARCH
Subsystem:
Regression: No
Bisected commit-id:


Attachments
Processor information (1.46 KB, text/plain)
2009-11-26 20:34 UTC, Alex Fera
Details
Module information (4.11 KB, text/plain)
2009-11-26 20:34 UTC, Alex Fera
Details
/proc/ioports (1.30 KB, application/octet-stream)
2009-11-26 20:35 UTC, Alex Fera
Details
/proc/iomem (1.96 KB, text/plain)
2009-11-26 20:37 UTC, Alex Fera
Details
PCI information (26.79 KB, text/plain)
2009-11-26 20:38 UTC, Alex Fera
Details
Memory information (930 bytes, application/octet-stream)
2009-11-26 20:40 UTC, Alex Fera
Details
acpidump (123.38 KB, text/plain)
2009-11-27 07:25 UTC, Alex Fera
Details
dmesg (32.85 KB, text/plain)
2009-11-27 07:26 UTC, Alex Fera
Details
dmidecode (11.96 KB, text/plain)
2009-11-27 07:28 UTC, Alex Fera
Details
.config (93.78 KB, text/plain)
2009-11-27 07:32 UTC, Alex Fera
Details
dmesg output with "processor" and "acpi-cpufreq" disabled (31.38 KB, text/plain)
2009-12-10 14:56 UTC, Alex Fera
Details
/var/log/messages.log with "processor" and "acpi-cpufreq" disabled (40.02 KB, text/plain)
2009-12-10 14:58 UTC, Alex Fera
Details
dmesg output with no modules disabled and with "mem=4096m" (32.24 KB, text/plain)
2009-12-10 15:01 UTC, Alex Fera
Details
/var/log/messages.log with no modules disabled and "mem=4096m" (40.78 KB, text/plain)
2009-12-10 15:02 UTC, Alex Fera
Details
SSDT4 (780 bytes, application/octet-stream)
2009-12-11 07:25 UTC, Alex Fera
Details
SSDT5 (1.34 KB, application/octet-stream)
2009-12-11 07:26 UTC, Alex Fera
Details
SSDT6 (463 bytes, application/octet-stream)
2009-12-11 07:26 UTC, Alex Fera
Details
SSDT7 (141 bytes, application/octet-stream)
2009-12-11 07:26 UTC, Alex Fera
Details
patch: introduce processor.max_pstate (1.10 KB, patch)
2010-01-13 08:07 UTC, Zhang Rui
Details | Diff

Description Alex Fera 2009-11-26 20:34:08 UTC
Created attachment 23952 [details]
Processor information

I get only 3 GB RAM even if I have 4 GB RAM installed.

 When I try to install or run a 64 bit OS, the system reboots, over and over.
 I've observed that the system reboots at the stage "Loading modules".
 If I add the boot parameter "mem=4096m" everything is fine, except that I get only 3 GB RAM. 
 If I add the boot parameter "acpi=off" I get the 4 GB RAM, but after a few minutes the system reboots.
 
Kernel version: Linux version 2.6.31-ARCH (root@architect) (gcc version 4.4.2 (GCC) ) #1 SMP PREEMPT Tue Nov 10 19:01:40 CET 2009
Distribution: Arch Linux (64 bit)
Hardware: Fujitsu Siemens Amilo PI 3540, BIOS Version: 1.09C
Comment 1 Alex Fera 2009-11-26 20:34:41 UTC
Created attachment 23953 [details]
Module information
Comment 2 Alex Fera 2009-11-26 20:35:59 UTC
Created attachment 23954 [details]
/proc/ioports
Comment 3 Alex Fera 2009-11-26 20:37:20 UTC
Created attachment 23955 [details]
/proc/iomem
Comment 4 Alex Fera 2009-11-26 20:38:53 UTC
Created attachment 23956 [details]
PCI information
Comment 5 Alex Fera 2009-11-26 20:40:28 UTC
Created attachment 23957 [details]
Memory information
Comment 6 ykzhao 2009-11-27 00:39:03 UTC
Will you please attach the output of acpidump, dmesg, dmidecode on your box?

It will be great if you can attach the .config file.

Thanks.
Comment 7 Alex Fera 2009-11-27 07:25:08 UTC
Created attachment 23960 [details]
acpidump
Comment 8 Alex Fera 2009-11-27 07:26:34 UTC
Created attachment 23961 [details]
dmesg
Comment 9 Alex Fera 2009-11-27 07:28:55 UTC
Created attachment 23962 [details]
dmidecode
Comment 10 Alex Fera 2009-11-27 07:32:45 UTC
Created attachment 23963 [details]
.config
Comment 11 ykzhao 2009-11-30 12:45:00 UTC
Will you please try to change the working mode of SATA controller in BIOS option and see whether the issue still exists?

thanks.
Comment 12 Alex Fera 2009-12-01 14:52:21 UTC
I don't have an option like that in BIOS.
Comment 13 ykzhao 2009-12-03 13:07:03 UTC
Hi, Alex
    DO you mean that you can't change the SATA mode in BIOS?
    Do have an opportunity to try the 32-bit OS and see whether the issue still exists?
    Can you confirm whether this issue exists on windows?

thanks.
   Yakui
Comment 14 Alex Fera 2009-12-03 13:28:09 UTC
Hi, Yakui
  Yes, I can't change the SATA mode in BIOS.
  I have tried a 32-bit OS and I get only 3 GB RAM, but if I enable PAE then I get 4 GB RAM and everything is ok. But I want a 64-bit OS.
  The same issue exists in 64-bit Windows (the system reboots and I cannot install the OS)
Comment 15 Alex Fera 2009-12-03 17:37:09 UTC
Hi, again!
  I found on FujitsuSiemens forum that other people have the same problem with Amilo PI3540. Apparently FujitsuSiemens designed the Amilo Pi3540 with an incorrect memory controller! The device reports wronge values to the OS. Do you think that's the problem?
Comment 16 Alex Fera 2009-12-03 18:02:56 UTC
Retraction, apparently I was wrong:
"The issue isn't with the memory controller hardware. The chipset is a just a standard Intel GM45, if the flaw was in the chipset it would affect all laptops, from all manufacturers, using the GM45, not just the Pi series from FSC.
It's most likely an issue with remapping memory over the 4GB boundary, as reducing the system memory to less than 4GB solves the issue."
A BIOS problem?
Comment 17 Zhang Rui 2009-12-04 03:04:39 UTC
so the kernel reboots if you boot w/o mem=4g?
hmm, I think we should fix this problem first.

> I've observed that the system reboots at the stage "Loading modules".
it would be good if you could catch the screenshot when the kernel reboots.
or get the kernel log via serial console.
Comment 18 Alex Fera 2009-12-04 09:29:00 UTC
No, the system reboots (at the stage "Loading modules") if I do not add "mem=4096m", but if I add "mem=4096m" then I get 3 GB of RAM from 4 GB of RAM available.
Also if I add "acpi=off" I get the 4 GB of RAM, but after a few minutes the system reboots unexpectedly.

> it would be good if you could catch the screenshot when the kernel reboots.
or get the kernel log via serial console.
But if the system reboots how could I catch a screenshot?
Comment 19 Zhang Rui 2009-12-07 06:24:41 UTC
(In reply to comment #18)
> No, the system reboots (at the stage "Loading modules") if I do not add
> "mem=4096m",

well, this is the answer I want to get.

> but if I add "mem=4096m" then I get 3 GB of RAM from 4 GB of RAM
> available.

that's right. the last 1024m memory is mapped to address
"BIOS-e820: 0000000100000000 - 0000000140000000 (usable)"
if you use mem=4096m, Linux kernel  will not be aware of this piece of memory.

> > it would be good if you could catch the screenshot when the kernel reboots.
> or get the kernel log via serial console.
> But if the system reboots how could I catch a screenshot?

is serial console available? anyway, the more information when the system reboots we get, the more easily we debug this problem.
"Loading modules" is not clear enough to me for now. :(
Comment 20 Alex Fera 2009-12-07 13:05:18 UTC
I managed to isolate the modules that cause the reboot: 
"processor" and "acpi-cpufreq".
With this two modules disabled and without using any boot parameters the system doesn't reboots and I have 4 GB of RAM.
Comment 21 Alex Fera 2009-12-07 13:11:35 UTC
>I managed to isolate the modules that cause the reboot: 
>"processor" and "acpi-cpufreq".
>With this two modules disabled and without using any boot parameters the
>system
>doesn't reboots and I have 4 GB of RAM.

If I try to load one of this modules, after the system has booted, the system instantly reboots!
Comment 22 Alex Fera 2009-12-07 21:37:39 UTC
Update: the computer still reboots, but after 5-10 minutes.
Comment 23 Zhang Rui 2009-12-08 08:31:04 UTC
please boot with "idle=poll" and see if it helps.
Comment 24 Alex Fera 2009-12-08 09:04:07 UTC
> please boot with "idle=poll" and see if it helps.
It doesn't help. It still reboots.
Comment 25 Zhang Rui 2009-12-08 09:09:33 UTC
please attach the dmesg output after system boot.
and do you see anything unusual when the system reboot?
hmm, attach the /var/log/message file would be helpful after a system boot.
Comment 26 Alex Fera 2009-12-08 12:19:51 UTC
Post the dmesg output and /var/log/message with "processor" and "acpi-cpufreq" modules disabled?
Comment 27 Zhang Rui 2009-12-09 01:32:41 UTC
yes.
1. boot with these modules and get the dmesg output,
2. wait until the machine reboots.
3. feel free to boot with "mem=4g" this time and attach the/var/log/message file, as well as the dmesg got in step1.
Comment 28 Alex Fera 2009-12-10 14:56:02 UTC
Created attachment 24133 [details]
dmesg output with "processor" and "acpi-cpufreq" disabled
Comment 29 Alex Fera 2009-12-10 14:58:57 UTC
Created attachment 24134 [details]
/var/log/messages.log with "processor" and "acpi-cpufreq" disabled
Comment 30 Alex Fera 2009-12-10 15:01:01 UTC
Created attachment 24135 [details]
dmesg output with no modules disabled and with "mem=4096m"
Comment 31 Alex Fera 2009-12-10 15:02:14 UTC
Created attachment 24136 [details]
/var/log/messages.log with no modules disabled and "mem=4096m"
Comment 32 Zhang Rui 2009-12-11 01:36:15 UTC
(In reply to comment #21)
> >I managed to isolate the modules that cause the reboot: 
> >"processor" and "acpi-cpufreq".
> >With this two modules disabled and without using any boot parameters the
> system
> >doesn't reboots and I have 4 GB of RAM.
> 
> If I try to load one of this modules, after the system has booted, the system
> instantly reboots!

please attach all the files under /sys/firmware/acpi/tables/dynamic/.
For example, you can get the content of the file by:
cat /sys/firmware/acpi/tables/dynamic/SSDT4 > /tmp/ssdt.dat
and attach the ssdt.dat here.
Comment 33 Alex Fera 2009-12-11 07:25:30 UTC
Created attachment 24152 [details]
SSDT4
Comment 34 Alex Fera 2009-12-11 07:26:07 UTC
Created attachment 24153 [details]
SSDT5
Comment 35 Alex Fera 2009-12-11 07:26:30 UTC
Created attachment 24154 [details]
SSDT6
Comment 36 Alex Fera 2009-12-11 07:26:54 UTC
Created attachment 24155 [details]
SSDT7
Comment 37 Zhang Rui 2009-12-14 01:44:48 UTC
does the system reboot if you disable "acpi-cpufreq" driver only?
Comment 38 Alex Fera 2009-12-14 05:16:55 UTC
> does the system reboot if you disable "acpi-cpufreq" driver only?
Yes. I need to disable both drivers.
Comment 39 Zhang Rui 2010-01-13 08:07:14 UTC
Created attachment 24536 [details]
patch: introduce processor.max_pstate

please apply this patch, set CONFIG_ACPI_PROCESSOR=y, rebuild your kernel,
and then reboot with kernel parameter processor.max_pstate=0.
does the problem still exist in this test?
Comment 40 Alex Fera 2010-01-13 18:41:45 UTC
Well, the system boots and I get the 4 GB of RAM, but the system reboots after a few minutes.
Comment 41 Zhang Rui 2010-01-19 07:35:33 UTC
please verify if this is a duplicate of bug http://bugzilla.novell.com/show_bug.cgi?id=436895
and check if there is any new BIOS release for you.
Comment 42 Alex Fera 2010-01-19 21:46:36 UTC
http://bugzilla.novell.com/show_bug.cgi?id=436895
Yes, it's duplicate.
There is no BIOS update available. I guess I'll have to wait for a BIOS update.
Comment 43 Zhang Rui 2010-01-21 07:59:57 UTC
Right.
I'll close this bug because it's a BIOS bug and we can do nothing in Linux kernel.
let's wait for the new BIOS release.
please re-open it if there is anything that you think we could help. :)

Note You need to log in before you can comment on or make changes to this bug.