I've managed to get a bit more information re the kernel 2.6.24.x crashes I have when physical mem is used up.. but first to recap:-
I have an acer 5620 core 2 duo running the x86_64 version of fedora 8 with 2GB Ram and 2GB swap. It works fine with kernel 2.6.23.15-137.fc8, but when I try any of the 2.6.24.xx-xx.fc8 kernels the machine crashes when pyhsical ram fills up. Swap is not used.
* I have checked swap space with badblocks, and all reports fine * Swap is on
...$ cat /proc/swaps Filename Type Size Used Priority /dev/mapper/VolGroup00-LogVol01 partition 2031608 0 -1
...$ cat /proc/sys/vm/swappiness 60
at the point it crashes top (via ssh) looks like this:-
top - 14:13:36 up 29 min, 4 users, load average: 1.32, 0.56, 0.38 Tasks: 162 total, 1 running, 161 sleeping, 0 stopped, 0 zombie Cpu(s): 18.2%us, 6.0%sy, 0.0%ni, 43.0%id, 32.1%wa, 0.0%hi, 0.7%si, 0.0%st Mem: 2061304k total, 2042936k used, 18368k free, 44764k buffers Swap: 2031608k total, 0k used, 2031608k free, 679356k cached
if I run stress from the command line I do get a stack dump I have only been able to capture a bit of that by copying it by hand:-
(*new info) ---- quote ---- CR2 ffffffff00000001
BUG sleeping function called from invalid context at kernel/rwsem.c:21 in_atomic():0, irqs_disabled():1 Pid 3920, comm: stress Tainted: P D 2.6.24.3-50.fc8 #1
Call Trace: [<ffffffff81268b17>] down_read +0x15/0x23 [<ffffffff8105e091>] acct_collect +0x42/0x18e [<ffffffff8103a632>] do_exit + 0x217/0x76b [<ffffffff8126b666>] do_page_fault +0x5c3/0x691 ... ---- end quote ----
Any Ideas what might be happening? Surely it can't just be my machine???
Cheers, Peter.
On 02/04/2008, Peter McNeil peter@mcneils.net wrote: [snip]
Any Ideas what might be happening? Surely it can't just be my machine???
Well, if you give a little more detail about the test you're running I can try and reproduce it on similar hardware - what's "stress" ?
J.
Jonathan Underwood wrote:
On 02/04/2008, Peter McNeil peter@mcneils.net wrote: [snip]
Any Ideas what might be happening? Surely it can't just be my machine???
Well, if you give a little more detail about the test you're running I can try and reproduce it on similar hardware - what's "stress" ?
J.
Stress is a stress testing program, however *anything* that fills up physical memory and needs swap causes the problem for me (I'm a java developer so my IDE and a few other programs can do that).
You may have memhog installed, if you have say 2GB memory then try memhog 2g and it will begin filling up memory, that should do it.
Cheers, Peter.
On 03/04/2008, Peter McNeil peter@mcneils.net wrote:
Jonathan Underwood wrote:
On 02/04/2008, Peter McNeil peter@mcneils.net wrote: [snip]
Any Ideas what might be happening? Surely it can't just be my
machine???
Well, if you give a little more detail about the test you're running I can try and reproduce it on similar hardware - what's "stress" ?
J.
Stress is a stress testing program, however *anything* that fills up physical memory and needs swap causes the problem for me (I'm a java developer so my IDE and a few other programs can do that).
You may have memhog installed, if you have say 2GB memory then try memhog 2g and it will begin filling up memory, that should do it.
Am currently sat at a x86_64 pentium D machine with 1GB ram. I did memhog 1g, which didn't trigger the bug you see. I also did stress --vm 32, which clobbered the machine while I went for coffee, but also didn't trigger the bug. This is with kernel 2.6.24.4-64.fc8. Tomorrow I can try on a pentium core 2 duo machine with 2G ram.
Jonathan Underwood wrote:
On 03/04/2008, Peter McNeil peter@mcneils.net wrote:
Jonathan Underwood wrote:
On 02/04/2008, Peter McNeil peter@mcneils.net wrote: [snip]
Any Ideas what might be happening? Surely it can't just be my
machine???
Well, if you give a little more detail about the test you're running I can try and reproduce it on similar hardware - what's "stress" ?
J.
Stress is a stress testing program, however *anything* that fills up physical memory and needs swap causes the problem for me (I'm a java developer so my IDE and a few other programs can do that).
You may have memhog installed, if you have say 2GB memory then try memhog 2g and it will begin filling up memory, that should do it.
Am currently sat at a x86_64 pentium D machine with 1GB ram. I did memhog 1g, which didn't trigger the bug you see. I also did stress --vm 32, which clobbered the machine while I went for coffee, but also didn't trigger the bug. This is with kernel 2.6.24.4-64.fc8. Tomorrow I can try on a pentium core 2 duo machine with 2G ram.
Thanks for that, just installed and tried 2.6.24-64.fc8, and the same error is triggered on my machine.
On 04/04/2008, Peter McNeil peter@mcneils.net wrote:
Thanks for that, just installed and tried 2.6.24-64.fc8, and the same error is triggered on my machine.
FWIW, I just tried both memhog 2g and memhog 3g on a core 2 duo machine with 2GB ram, and didn't trigger the bug. This doesn't mean you're not seeing a genuine bug however - you should probably still report it upstream.
J.