UVM related crash

[Zach Wegner <zwegner%gmail.com@localhost>]

Hello all,

I've gotten a crash a couple times just now, when I'm trying to run
some CPU intensive tasks. I'm building a test system that is running
on four cores. When starting up the system, it runs at 100% for about
a minute or so, and then the computer suddenly dies. After the first
time, looking at the dmesg output, I saw a bunch of output like:

UVM: pid XXXX, uid 1000 killed: out of swap

...for each of the processes that I was running in the test. The
processes don't consume much memory at all (maybe 12M a piece), but
there are some processes that are frequently stopped/restarted (about
8 every 10 seconds, which shouldn't be too much at all). So I started
the same tests again, while watching top. The memory fluctuated at
about 3900 (on a machine with 4G ram), and the machine died again
about 30 seconds into it. Dmesg says pretty much the same thing,
including this, which I assume is from savecore:


uvm_fault(0xffffffff80c31640, 0xffffffff80e56000, 1) -> e
fatal page fault in supervisor mode
trap type 6 code 0 rip ffffffff8041b83d cs 8 rflags 10a06 cr2  ffffffff80e56900
cpl 7 rsp ffff80004e4ef770
panic: trap
Begin traceback...
uvm_fault(0xffff80004e619460, 0x0, 1) -> e
fatal page fault in supervisor mode
trap type 6 code 0 rip ffffffff804f97cc cs 8 rflags 10246 cr2  0 cpl 7 rsp ffff8
0004e4ef310
panic: trap
Faulted in mid-traceback; aborting...
dump to dev 0,17 not possible

I assume this is somehow related to the processes I'm running, as this
hasn't happened before. But I'm not doing anything weird--the
processes have pthreads, fork/exec, and some stdio, that's about
everything that involves the OS. What should I do?

Thanks in advance,
Zach