Proper way to access > 4GB on LMCE 32 bit kernel?

[tsoukas]

Dear all,
Checking out an 8GB RAM server on a LMCE 10.04 w/RC, dmesg prints the following:

Code: [Select]

[    0.000000] Initializing HighMem for node 0 (0002effe:0007f800)
[    0.000000] Memory: 2005284k/2088960k available (4699k kernel code, 46060k reserved, 2128k data, 664k init, 1283128k highmem)
This does not appear correct, (shouldnt 32bit see up to 4GB RAM?) and it seriously cripples a large system... (at a point, it was using 5GB of swap..)
I did some research and the only solution appears to use a PAE-enable kernel.

So I tried installing one with:

Code: [Select]

sudo apt-get install linux-headers-server linux-image-server linux-server
And indeed, I could see the entire memory:

Code: [Select]

dcerouter_1032148:/var/log/pluto# top
top - 17:39:06 up  2:29,  2 users,  load average: 0.64, 0.71, 0.72
Tasks: 273 total,   1 running, 272 sleeping,   0 stopped,   0 zombie
Cpu(s): 13.1%us,  5.2%sy,  0.6%ni, 68.6%id, 12.0%wa,  0.0%hi,  0.5%si,  0.0%st
Mem:   8201900k total,  5062680k used,  3139220k free,   314256k buffers
Swap:  5907448k total,        0k used,  5907448k free,  3420992k cached
The only problem so far is that HAL kept dying, which is not really convenient (for an LMCE server, that is..)
Any recommendations? Should HAL die because of this could there be another issue with it?

Thanks,
Ted

[tschak909]

the best thing that can be done, is to send us a core dump.

The core dumps live in /usr/pluto/coredump/ and have a filename like core_HAL_xxxxxxx where xxxxx is the process ID of the crashed process..

You can then go to /usr/pluto/bin and invoke gdb as such:

gdb /usr/pluto/bin/HAL /usr/pluto/coredump/core_HAL_xxxxxxxx where xxxxx is the process ID of the crashed process.

You can then type bt at the (gdb) prompt, and at least get a backtrace.

Cut and paste this into a pastebin, and give it to us.

Thanks,
-Thom

[darkwizard864]

 if you use any other one expect 2.6.32-45-generic-pae or lower you have hal crash