Multiple *BAD*gran_size messages

[melal]

My dmesg has multiple *BAD*gran_size messages. As follows from specific of the messages there is an issue at RAM level. I use PAE kernel, memory has been detected completely. More details:

dmesg: http://forums.fedoraforum.org/attach...0&d=1345363807

uname -a

Linux hostname 3.4.7-1.fc16.i686.PAE #1 SMP Mon Jul 30 16:45:15 UTC 2012 i686 i686 i386 GNU/Linux

cat /proc/meminfo

MemTotal:       16492496 kB
MemFree:        14845144 kB
Buffers:           99448 kB
Cached:           234036 kB
SwapCached:            0 kB
Active:           989504 kB
Inactive:         551368 kB
Active(anon):     822892 kB
Inactive(anon):   392528 kB
Active(file):     166612 kB
Inactive(file):   158840 kB
Unevictable:        3256 kB
Mlocked:            3256 kB
HighTotal:      15766764 kB
HighFree:       14304520 kB
LowTotal:         725732 kB
LowFree:          540624 kB
SwapTotal:             0 kB
SwapFree:              0 kB
Dirty:                68 kB
Writeback:             0 kB
AnonPages:       1210580 kB
Mapped:            57052 kB
Shmem:              5768 kB
Slab:              48236 kB
SReclaimable:      31740 kB
SUnreclaim:        16496 kB
KernelStack:        3256 kB
PageTables:        13408 kB
NFS_Unstable:          0 kB
Bounce:                0 kB
WritebackTmp:          0 kB
CommitLimit:     8246248 kB
Committed_AS:    4037808 kB
VmallocTotal:     122880 kB
VmallocUsed:       40236 kB
VmallocChunk:      81012 kB
HardwareCorrupted:     0 kB
AnonHugePages:    704512 kB
HugePages_Total:       0
HugePages_Free:        0
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB
DirectMap4k:       12280 kB
DirectMap2M:      894976 kB

If I understand right something is wrong at MTRR level. Thus here is my cat /proc/mtrr:

reg00: base=0x000000000 (    0MB), size=16384MB, count=1: write-back
reg01: base=0x400000000 (16384MB), size=  512MB, count=1: write-back
reg02: base=0x0e0000000 ( 3584MB), size=  512MB, count=1: uncachable
reg03: base=0x0dc000000 ( 3520MB), size=   64MB, count=1: uncachable
reg04: base=0x0db800000 ( 3512MB), size=    8MB, count=1: uncachable
reg05: base=0x41f800000 (16888MB), size=    8MB, count=1: uncachable
reg06: base=0x41f600000 (16886MB), size=    2MB, count=1: uncachable

Please, help.

[george_toolan]

You can test your memory with Memtest86+, but your BIOS seems to be broken.

If you have 16 GiB of memory you should consider installing a 64 bit operating system, because a (single) 32 bit process isn't allowed to use more than 3 GiB of memory even with a PAE kernel.

[melal]

You can test your memory with Memtest86+, but your BIOS seems to be broken.

If you have 16 GiB of memory you should consider installing a 64 bit operating system, because a (single) 32 bit process isn't allowed to use more than 3 GiB of memory even with a PAE kernel.

Dear george_toolan, thank you very much for your early reply!

I'll try to test the memory by Memtest86+ tomorrow and post a report.

It's not necessary for me to have 3 Gb for a process as this is regular web- mail- proxy-server and there are no processes which can consume so much of RAM per process. PAE-kernel is OK for me, I understood it during installation well. For some reasons I need 32bit OS, but not 64bit, due to some software requirements.

Thanks again!

[melal]

You can test your memory with Memtest86+, but your BIOS seems to be broken.

OK. Memory check shows my RAM is OK.

Updating to newer version of BIOS for my Intel DH77EB motherboard (from v.0053 to v.0075) gave me no positive result, as it follows from dmesg: dmesg_0075.txt - the problem still the same.

Afterwards I paid attention to some dmesg messages:

[    0.000000] mtrr_cleanup: can not find optimal value
[    0.000000] please specify mtrr_gran_size/mtrr_chunk_size

I have decided to enable MTRR sanitizer and specify values for mtrr_gran_size and mtrr_chunk_size by appending to boot options in /boot/grub2/grub.cfg and /etc/default/grub the following parameters:

enable_mtrr_cleanup mtrr_spare_reg_nr=1 mtrr_gran_size=32M mtrr_chunk_size=128M

This MTRR options solved my problem completely, only one dmesg message appears:

[    0.000000] WARNING: BIOS bug: CPU MTRRs don't cover all of memory, losing 31MB of RAM.

Thus 31 Mb from 16Gb of my RAM gone, but for another things I'm happy, dmesg looks quite clean: dmesg_OK.txt and now we can see:

cat /proc/mtrr

reg00: base=0x000000000 (    0MB), size= 2048MB, count=1: write-back
reg01: base=0x080000000 ( 2048MB), size= 1024MB, count=1: write-back
reg02: base=0x0c0000000 ( 3072MB), size=  512MB, count=1: write-back
reg03: base=0x0da000000 ( 3488MB), size=   32MB, count=1: uncachable
reg04: base=0x0dc000000 ( 3520MB), size=   64MB, count=1: uncachable
reg05: base=0x100000000 ( 4096MB), size= 4096MB, count=1: write-back
reg06: base=0x200000000 ( 8192MB), size= 8192MB, count=1: write-back
reg07: base=0x400000000 (16384MB), size=  512MB, count=1: write-back
reg08: base=0x41e000000 (16864MB), size=   32MB, count=1: uncachable
reg09: base=0x0e0000000 ( 3584MB), size=  256MB, count=1: write-combining

Thank you dear george_toolan for trying help me!

The problem is solved.

[OlivFedora]

Hi,

your post helped me a lot, thanks.

I used too those two sites :
http://www.mjmwired.net/kernel/Docum...parameters.txt
and
http://my-fuzzy-logic.de/blog/index....-problems.html

The second one was very helpful to understand a bit more what were those 'gran_size: xM' messages.
The light came with this sentence :
"The sanitizer was not able to find an optimal mtrr setup, but he kindly printed a long list with possible setups for me into the buffer."

So it's easy...
In my case the minimum RAM I could lose was 8M :
[ 0.000000] gran_size: 8M chunk_size: 32M num_reg: 10 lose cover RAM: 8M
[ 0.000000] gran_size: 8M chunk_size: 64M num_reg: 10 lose cover RAM: 8M
[ 0.000000] gran_size: 8M chunk_size: 128M num_reg: 10 lose cover RAM: 8M
[ 0.000000] gran_size: 8M chunk_size: 256M num_reg: 10 lose cover RAM:

My choice is gran_size: 8M chunk_size: 64M , but now the question is : what is the 'optimal' chunk_size ? a big one or a small one ?
By the way, apprently, I'm losing only 2M :
WARNING: BIOS bug: CPU MTRRs don't cover all of memory, losing 2MB of RAM.

And a note for the road : in grub write 8M not 8, it will freeze up everything (thanks to the 'e' command I could modify grub.cfg on the fly)

O'

[jpollard]

I suspect you need a BIOS update...

Depending on the age of the motherboard, it might not handle the memory you have (might be a bit too large per unit). Doesn't mean you can't work around it, but loosing memory is a sign that it exceeds the capacity of the CPU.