[Gc] Any reason the collector might fail in GC_unmap()?

Talbot, George Gtalbot at ansarisbio.com
Mon Dec 14 11:41:39 PST 2009


Hey folks,

Any idea why I might get an ABORT() in GC_unmap?  I’m configured for 64-bit Linux, parallel collection on, memory unmapping on.

Here’s a backtrace…not sure what’s going on here.  Just added a log message just before the abort() to try to print errno, start_addr, and len in the mmap() to see if I can get more information in case I can get this to happen again.

[Switching to Thread 1083730256 (LWP 28920)]

Breakpoint 1, 0x00002b2fdf1ce0b0 in abort () from /lib/libc.so.6
(gdb) bt
#0  0x00002b2fdf1ce0b0 in abort () from /lib/libc.so.6
#1  0x0000000000553852 in GC_abort (
    msg=0x57278c "mmap(...PROT_NONE...) failed") at misc.c:1259
#2  0x0000000000555273 in GC_unmap (start=0x2aaad7a27000 "��U", bytes=4096)
    at os_dep.c:2245
#3  0x0000000000559182 in GC_unmap_old () at allchblk.c:409
#4  0x000000000054950b in GC_finish_collection () at alloc.c:905
#5  0x0000000000548bb2 in GC_try_to_collect_inner (
    stop_func=0x54853f <GC_never_stop_func>) at alloc.c:447
#6  0x0000000000549718 in GC_try_to_collect_general (stop_func=0,
    force_unmap=0) at alloc.c:947
#7  0x00000000005497a1 in GC_gcollect () at alloc.c:973
#8  0x0000000000490902 in Client::createfile (this=0x2aaaf99eb500,
    path=0x2aab249f9018 "data/GC_collect", uid=4308, gid=10007, flags=34881,
    mode=33188, cookie=3376865) at master/Client.cpp:1586
#9  0x0000000000491c02 in Client::process_packet (this=0x2aaaf99eb500,
    pkt=0x2aab249f9000 "\217", len=40) at master/Client.cpp:2846
#10 0x0000000000480ee6 in PacketWorker::process_input (this=0x2aaaf99eb500)
    at common/Worker.cpp:152
#11 0x0000000000481577 in WorkerBase::thread_function (
    arg=<value optimized out>) at common/Worker.cpp:90
#12 0x0000000000557b5a in GC_inner_start_routine (sb=0x409860f0,
    arg=0x2aaadd55d180) at pthread_support.c:1216
---Type <return> to continue, or q <return> to quit---q

If you can shed any light on this, I’d appreciate it.

Thanks.

--
George T. Talbot
gtalbot at locuspharma.com




More information about the Gc mailing list