0001905: 4GB only recognized as 3GB

(0004791)
smooge (developer)
2007-04-14 13:06

Probably not a bug.

Upstream uses a very conservative set of kernel options for detecting available amount of memory on hardware. In order to get around this limitation, you need to run PAE kernel in most cases, and in some cases use the pci=nommconf or acpi=off <or other options>.

(0004793)
igor_313 (reporter)
2007-04-14 13:50

Thanks for the quick response smooge.
Unfortunatelly I tried PAE kernel and it actually freezes when it boots if the remap feature in the BIOS is enabled.

The kernel stops at:
Time: tsc clocksource has been installed OR
Write protecting the kernel read-only data: 382k

I've tried all the combinations of pci=nommconf and acpi=off with both standard and PAE kernel and no luck the highest I get is 3GB.

Do you have any other suggestions the board is INtel chipset P965 with Core 2 Duo 6300.

It looks like I might have to go back to FC6.

Thanks

Igor

(0004794)
igor_313 (reporter)
2007-04-14 17:18

There is definitely an issue how the normal kernel recongizes the RAM on this Asus board. I was able to get the CentOS 5 recognize 4GB of RAM by loading the xen kernel. Unfortunatelly I have the NVIDIA card and NVIDIA does not support the xen enabled kernel. After a little bit of further research I've stumbled across the procedure on how to compile NVIDIA driver with a patch on opensuse site.

Igor

(0004798)
igor_313 (reporter)
2007-04-15 10:44

I had a lot of stability issues with the xen kernel.
The system would simply lockup.
So I'm still down to 3GB and the system is stable.

Please advise if there is anything else that I can try.

Thanks

Igor

(0004799)
smooge (developer)
2007-04-15 13:20

Ok lets see if we have any 'updates' for the boards BIOS. I found a Dell system that was 'acting' the say way, but a BIOS update from them got the box to getting the memory being detected correctly. Doing some research on other systems (mostly AMD boxes but some Intel's) with the newer memory layouts found that the BIOS's will usually come less that optimal.

If that does not fix the problems... then we will need to report upstream and see what the issues they end up finding for this.

Not sure why FC6 is seeing the stuff correctly and CentOS-5 is not.. although I would need to know which FC6 kernel was involved (the 2.6.18's or the later ones with 2.6.20.) If it is the later ones then it will be something that upstream will 'fix' in a later kernel rev.

We will need to know:
CPU, model, speed, etc
Motherboard, model, settings from default, etc
Memory, model, type, speed, etc
Peripheals
Video
Keyboard/Mouse
Drives
etc
BIOS, make, version ,etc

Thanks

(0004801)
igor_313 (reporter)
2007-04-15 13:55

I'm currently running the latest BIOS from ASUS.
The FC6 kernel that was working correctly with this board was 2.6.20.x and I can't tell you if the original 2.6.18 was recognizing it correctly or not since I've upgraded the hardware on it with the kernel already at 2.6.20.

Here are the system specs:

CPU: Intel Core 2 Duo E6300 1.86GHz w/ 2MB
Motherboard: Asus P5B-E, Intel chipset P965, the only changes from the default settings are AHCI for SATA drives and at this point the memory remaping feature is set to off instead of on which results in BIOS reporting 2GB and CenOS5 reporting 3GB, while if its on BIOS reports 4GB and OS reports 2GB.
Memory: DDR2 OCZ PC2-6400 EL Platinum XTC Edition Rev. 2, 4x1GB
Peripherals:
Video: eVGA e-GeForce 7600 GS 256MB PCI-E w/ Passive Heatsink, DVI, HDTV-Out
Keybard/Mouse: MS Comfort Curve 2000 USB keyboard/MS Comfort Optical Mouse 3000 USB
Drives: 3 x Seagate 320GB Barracuda 7200.10 SATA II w/ NCQ, 16MB Cache
TV Tuner: Pinnacle PCTV Studio

BIOS: Latest Asus BIOS, AMI, 1202

Please let me know if you need output from lspci etc.

Thanks

Igor

(0006569)
tdereus (reporter)
2007-12-18 06:51

I have just come across a similar issue whilst running XEN in Centos 5. Centos is only recognising 2.5G of the 4G memory available.

Motherboard - Super X7DBE
2 x Intel Xeon 5365 DC 3.0Ghz.667FSB/4MB/LGA6/Pass
4 x RM/EC1GD2FB533 1.0GB 533Mhz DDR2 ECC FBDIMM CL4. Kingston
4 x HD-S2/WD250YS 250GD WD 2500YS HDD. SATAII,7200rpm,16MB

cat /proc/meminfo | head -4
MemTotal: 2555904 kB
MemFree: 873544 kB
Buffers: 87308 kB
Cached: 980024 kB

(0006832)
natediggs (reporter)
2008-02-07 15:15

I am experiencing the same issue with a AMD64 setup running the Xen kernel. I have 8GBs of memory which is correctly detected by the non-Xen kernel. When I choose to boot the Xen kernel then I only get 3.3gigs of memory detected.

There is an open Fedora bug tracking the same issue.
https://bugzilla.redhat.com/show_bug.cgi?id=250299 [^]

The problem is with Grub. It doesn't map the memory correctly.

(0006833)
toracat (developer)
2008-02-07 15:24

The issue with grub and the xen kernel is also being discussed in the forum thread:

http://www.centos.org/modules/newbb/viewtopic.php?topic_id=12491&forum=38 [^]

(0006835)
toracat (developer)
2008-02-07 16:14

I can offer the patched grub if you wish to give it a try. Is your system 32-bit?

Akemi

(0006836)
toracat (developer)
2008-02-07 16:18

I needed to modify the patch file provided in the upstream bugzilla (minor line number diff) to apply it to the grub-0.97-13 srpm.

--- grub-0.97/stage2/common.c 2008-02-07 07:56:16.000000000 -0800
+++ grub-0.97/stage2/common.c.new 2008-02-07 07:59:06.000000000 -0800
@@ -142,7 +142,8 @@
init_bios_info (void)
{
#ifndef STAGE1_5
- unsigned long cont, memtmp, addr;
+ unsigned long memtmp, addr;
+ volatile unsigned long cont;
int drive;
#endif

(0006837)
natediggs (reporter)
2008-02-07 17:27

I am running the 64 bit version.

(0006839)
toracat (developer)
2008-02-07 17:49

For the 64-bit version, it will take some time for me to build. I have yet to set up mock on a CentOS-4 machine to build it. Hopefully someone else can get to it (patched grub) before I do ?

(0006858)
natediggs (reporter)
2008-02-08 22:47

Would it be possible to get a 64bit package with the patch that solves this issue?

(0006860)
jhughes@hughesjr.com (administrator)
2008-02-09 15:07

http://www.centos.org/modules/newbb/viewtopic.php?topic_id=12491&start=0#forumpost41509 [^]

That link has an explanation ...

You can build with mock on centos4 (setup for building centos-5 pacakges) or install glibc-devel.i386 and glibc.i686 on x86_64 to build this package

(0006898)
toracat (developer)
2008-02-15 16:19

There is another person who successfully fixed the problem by using grub with the patch applied. He also made the srpm and x86_64 binary available:

ftp://ftp.zouric.com/public/grub-0.97-14.x86_64.rpm [^]

ftp://ftp.zouric.com/public/grub-0.97-14.src.rpm [^]

He has reported this at:

https://bugzilla.redhat.com/show_bug.cgi?id=250299 [^]

(0007244)
toracat (developer)
2008-05-09 17:23

The grub issue is now being tracked upstream in this bugzilla:

https://bugzilla.redhat.com/show_bug.cgi?id=445893 [^]

(0007264)
lovecraft (reporter)
2008-05-15 02:10

Hi there,
I've got a Dell 1900 with CentOS 5.1 64 bit and 4 GB ram.
Only 3 GB are recognized with kernel-xen-2.6.18-53.1.19.el5.
I've downloaded & installed grub-0.97-14.x86_64.rpm and did grub-install /dev/sda.
But it did not help and I still got 3 GB recognized.
Can you please help me to fix this.
Thank you in advance.

(0008561)
toracat (developer)
2009-01-12 20:33

The upstream bug tracker has moved to:

https://bugzilla.redhat.com/show_bug.cgi?id=448758 [^]

I have made the patched grub packages available from:

http://centos.toracat.org/grub-0.97/CentOS-5/ [^]

for both i386 and x86_64

(0008577)
prschaffner (reporter)
2009-01-15 16:06

> I have made the patched grub packages available from:

Need to bump the release up for yum/rpm to see it as an update.

Phil

(0008578)
toracat (developer)
2009-01-15 16:11

Will do. It was from the CentOS-5.1 days.

(0008579)
toracat (developer)
2009-01-15 17:36

grub-0.97-13.2.ay1 uploaded to http://centos.toracat.org/grub-0.97/CentOS-5/ [^]

(0010457)
toracat (developer)
2009-12-04 12:19

Updated grub to match the version in CentOS 5.4 ( grub-0.97-13.5.bz250299 ) at:

http://centos.toracat.org/grub-0.97/CentOS-5/ [^]

Notes
(0004791) smooge (developer) 2007-04-14 13:06	Probably not a bug. Upstream uses a very conservative set of kernel options for detecting available amount of memory on hardware. In order to get around this limitation, you need to run PAE kernel in most cases, and in some cases use the pci=nommconf or acpi=off <or other options>.

(0004793) igor_313 (reporter) 2007-04-14 13:50	Thanks for the quick response smooge. Unfortunatelly I tried PAE kernel and it actually freezes when it boots if the remap feature in the BIOS is enabled. The kernel stops at: Time: tsc clocksource has been installed OR Write protecting the kernel read-only data: 382k I've tried all the combinations of pci=nommconf and acpi=off with both standard and PAE kernel and no luck the highest I get is 3GB. Do you have any other suggestions the board is INtel chipset P965 with Core 2 Duo 6300. It looks like I might have to go back to FC6. Thanks Igor

(0004794) igor_313 (reporter) 2007-04-14 17:18	There is definitely an issue how the normal kernel recongizes the RAM on this Asus board. I was able to get the CentOS 5 recognize 4GB of RAM by loading the xen kernel. Unfortunatelly I have the NVIDIA card and NVIDIA does not support the xen enabled kernel. After a little bit of further research I've stumbled across the procedure on how to compile NVIDIA driver with a patch on opensuse site. Igor

(0004798) igor_313 (reporter) 2007-04-15 10:44	I had a lot of stability issues with the xen kernel. The system would simply lockup. So I'm still down to 3GB and the system is stable. Please advise if there is anything else that I can try. Thanks Igor

(0004799) smooge (developer) 2007-04-15 13:20	Ok lets see if we have any 'updates' for the boards BIOS. I found a Dell system that was 'acting' the say way, but a BIOS update from them got the box to getting the memory being detected correctly. Doing some research on other systems (mostly AMD boxes but some Intel's) with the newer memory layouts found that the BIOS's will usually come less that optimal. If that does not fix the problems... then we will need to report upstream and see what the issues they end up finding for this. Not sure why FC6 is seeing the stuff correctly and CentOS-5 is not.. although I would need to know which FC6 kernel was involved (the 2.6.18's or the later ones with 2.6.20.) If it is the later ones then it will be something that upstream will 'fix' in a later kernel rev. We will need to know: CPU, model, speed, etc Motherboard, model, settings from default, etc Memory, model, type, speed, etc Peripheals Video Keyboard/Mouse Drives etc BIOS, make, version ,etc Thanks

(0004801) igor_313 (reporter) 2007-04-15 13:55	I'm currently running the latest BIOS from ASUS. The FC6 kernel that was working correctly with this board was 2.6.20.x and I can't tell you if the original 2.6.18 was recognizing it correctly or not since I've upgraded the hardware on it with the kernel already at 2.6.20. Here are the system specs: CPU: Intel Core 2 Duo E6300 1.86GHz w/ 2MB Motherboard: Asus P5B-E, Intel chipset P965, the only changes from the default settings are AHCI for SATA drives and at this point the memory remaping feature is set to off instead of on which results in BIOS reporting 2GB and CenOS5 reporting 3GB, while if its on BIOS reports 4GB and OS reports 2GB. Memory: DDR2 OCZ PC2-6400 EL Platinum XTC Edition Rev. 2, 4x1GB Peripherals: Video: eVGA e-GeForce 7600 GS 256MB PCI-E w/ Passive Heatsink, DVI, HDTV-Out Keybard/Mouse: MS Comfort Curve 2000 USB keyboard/MS Comfort Optical Mouse 3000 USB Drives: 3 x Seagate 320GB Barracuda 7200.10 SATA II w/ NCQ, 16MB Cache TV Tuner: Pinnacle PCTV Studio BIOS: Latest Asus BIOS, AMI, 1202 Please let me know if you need output from lspci etc. Thanks Igor

(0006569) tdereus (reporter) 2007-12-18 06:51	I have just come across a similar issue whilst running XEN in Centos 5. Centos is only recognising 2.5G of the 4G memory available. Motherboard - Super X7DBE 2 x Intel Xeon 5365 DC 3.0Ghz.667FSB/4MB/LGA6/Pass 4 x RM/EC1GD2FB533 1.0GB 533Mhz DDR2 ECC FBDIMM CL4. Kingston 4 x HD-S2/WD250YS 250GD WD 2500YS HDD. SATAII,7200rpm,16MB cat /proc/meminfo \| head -4 MemTotal: 2555904 kB MemFree: 873544 kB Buffers: 87308 kB Cached: 980024 kB

(0006832) natediggs (reporter) 2008-02-07 15:15	I am experiencing the same issue with a AMD64 setup running the Xen kernel. I have 8GBs of memory which is correctly detected by the non-Xen kernel. When I choose to boot the Xen kernel then I only get 3.3gigs of memory detected. There is an open Fedora bug tracking the same issue. https://bugzilla.redhat.com/show_bug.cgi?id=250299 [^] The problem is with Grub. It doesn't map the memory correctly.

(0006833) toracat (developer) 2008-02-07 15:24	The issue with grub and the xen kernel is also being discussed in the forum thread: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=12491&forum=38 [^]

(0006835) toracat (developer) 2008-02-07 16:14	I can offer the patched grub if you wish to give it a try. Is your system 32-bit? Akemi

(0006836) toracat (developer) 2008-02-07 16:18	I needed to modify the patch file provided in the upstream bugzilla (minor line number diff) to apply it to the grub-0.97-13 srpm. --- grub-0.97/stage2/common.c 2008-02-07 07:56:16.000000000 -0800 +++ grub-0.97/stage2/common.c.new 2008-02-07 07:59:06.000000000 -0800 @@ -142,7 +142,8 @@ init_bios_info (void) { #ifndef STAGE1_5 - unsigned long cont, memtmp, addr; + unsigned long memtmp, addr; + volatile unsigned long cont; int drive; #endif

(0006837) natediggs (reporter) 2008-02-07 17:27	I am running the 64 bit version.

(0006839) toracat (developer) 2008-02-07 17:49	For the 64-bit version, it will take some time for me to build. I have yet to set up mock on a CentOS-4 machine to build it. Hopefully someone else can get to it (patched grub) before I do ?

(0006858) natediggs (reporter) 2008-02-08 22:47	Would it be possible to get a 64bit package with the patch that solves this issue?

(0006860) jhughes@hughesjr.com (administrator) 2008-02-09 15:07	http://www.centos.org/modules/newbb/viewtopic.php?topic_id=12491&start=0#forumpost41509 [^] That link has an explanation ... You can build with mock on centos4 (setup for building centos-5 pacakges) or install glibc-devel.i386 and glibc.i686 on x86_64 to build this package

(0006898) toracat (developer) 2008-02-15 16:19	There is another person who successfully fixed the problem by using grub with the patch applied. He also made the srpm and x86_64 binary available: ftp://ftp.zouric.com/public/grub-0.97-14.x86_64.rpm [^] ftp://ftp.zouric.com/public/grub-0.97-14.src.rpm [^] He has reported this at: https://bugzilla.redhat.com/show_bug.cgi?id=250299 [^]

(0007244) toracat (developer) 2008-05-09 17:23	The grub issue is now being tracked upstream in this bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=445893 [^]

(0007264) lovecraft (reporter) 2008-05-15 02:10	Hi there, I've got a Dell 1900 with CentOS 5.1 64 bit and 4 GB ram. Only 3 GB are recognized with kernel-xen-2.6.18-53.1.19.el5. I've downloaded & installed grub-0.97-14.x86_64.rpm and did grub-install /dev/sda. But it did not help and I still got 3 GB recognized. Can you please help me to fix this. Thank you in advance.

(0008561) toracat (developer) 2009-01-12 20:33	The upstream bug tracker has moved to: https://bugzilla.redhat.com/show_bug.cgi?id=448758 [^] I have made the patched grub packages available from: http://centos.toracat.org/grub-0.97/CentOS-5/ [^] for both i386 and x86_64

(0008577) prschaffner (reporter) 2009-01-15 16:06	> I have made the patched grub packages available from: Need to bump the release up for yum/rpm to see it as an update. Phil

(0008578) toracat (developer) 2009-01-15 16:11	Will do. It was from the CentOS-5.1 days.

(0008579) toracat (developer) 2009-01-15 17:36	grub-0.97-13.2.ay1 uploaded to http://centos.toracat.org/grub-0.97/CentOS-5/ [^]

(0010457) toracat (developer) 2009-12-04 12:19	Updated grub to match the version in CentOS 5.4 ( grub-0.97-13.5.bz250299 ) at: http://centos.toracat.org/grub-0.97/CentOS-5/ [^]

Issue History
Date Modified	Username	Field	Change
2007-04-14 12:59	igor_313	New Issue
2007-04-14 12:59	igor_313	Status	new => assigned
2007-04-14 12:59	igor_313	Assigned To	=> kbsingh@karan.org
2007-04-14 13:06	smooge	Note Added: 0004791
2007-04-14 13:50	igor_313	Note Added: 0004793
2007-04-14 17:18	igor_313	Note Added: 0004794
2007-04-15 10:44	igor_313	Note Added: 0004798
2007-04-15 13:20	smooge	Note Added: 0004799
2007-04-15 13:55	igor_313	Note Added: 0004801
2007-12-18 06:51	tdereus	Note Added: 0006569
2008-02-07 15:15	natediggs	Note Added: 0006832
2008-02-07 15:24	toracat	Note Added: 0006833
2008-02-07 16:14	toracat	Note Added: 0006835
2008-02-07 16:18	toracat	Note Added: 0006836
2008-02-07 17:27	natediggs	Note Added: 0006837
2008-02-07 17:49	toracat	Note Added: 0006839
2008-02-08 22:47	natediggs	Note Added: 0006858
2008-02-09 15:07	jhughes@hughesjr.com	Note Added: 0006860
2008-02-15 16:19	toracat	Note Added: 0006898
2008-03-27 12:40	tru	Issue Monitored: tru
2008-05-09 17:23	toracat	Note Added: 0007244
2008-05-15 02:10	lovecraft	Note Added: 0007264
2009-01-12 20:33	toracat	Note Added: 0008561
2009-01-15 16:06	prschaffner	Note Added: 0008577
2009-01-15 16:11	toracat	Note Added: 0008578
2009-01-15 17:36	toracat	Note Added: 0008579
2009-12-04 01:59	toracat	Relationship added	related to 0002244
2009-12-04 01:59	toracat	Relationship added	related to 0004048
2009-12-04 12:19	toracat	Note Added: 0010457