port-amd64/38612: Some programs die in _lwp_park

[Thomas Klausner <wiz%NetBSD.org@localhost>]

>Number:         38612
>Category:       port-amd64
>Synopsis:       mono program dies in _lwp_park
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-amd64-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu May 08 14:40:00 +0000 2008
>Originator:     Thomas Klausner
>Release:        NetBSD 4.99.62
>Organization:
Curiosity is the very basis of education and if you tell me that 
curiosity killed the cat, I say only that the cat died nobly.
- Arnold Edinborough
>Environment:
        
        
System: NetBSD hiro.nih.at 4.99.62 NetBSD 4.99.62 (HIRO64) #25: Tue Apr 29 
23:04:54 CEST 2008 
wiz%hiro.nih.at@localhost:/disk/wd2a/archive/cvs/src/sys/arch/amd64/compile/obj/HIRO64
 amd64
Architecture: x86_64
Machine: amd64
>Description:
wip/OdfConverter, run on a simple input file (which I can provide
in private mail, if need be), dies with:
(gdb) bt
#0  0x00007f7ffcd32f7a in _lwp_park () from /usr/lib/libc.so.12
#1  0x00007f7ffdc0bac3 in pthread__park () from /usr/lib/libpthread.so.0
#2  0x00007f7ffdc096b8 in pthread_cond_wait () from /usr/lib/libpthread.so.0
#3  0x00000000004c8b45 in _wapi_handle_dump ()
#4  0x00000000004cb033 in _wapi_handle_timedwait_signal_handle ()
#5  0x00000000004da4a4 in WaitForSingleObjectEx ()
#6  0x0000000000473c31 in mono_gc_is_finalizer_thread ()
#7  0x00000000004904a8 in mono_hazard_pointer_get ()
#8  0x00000000004d8ff8 in ExitThread ()
#9  0x00000000004efb24 in GC_start_routine ()
#10 0x00007f7ffdc0b332 in pthread_create () from /usr/lib/libpthread.so.0
#11 0x00007f7ffcd53010 in swapcontext () from /usr/lib/libc.so.12
#12 0x00007f7ffbe00000 in ?? ()
#13 0x0000000111110001 in ?? ()
#14 0x0000000033330003 in ?? ()
#15 0x0000000000000000 in ?? ()

On Mar 17, 2008, Jeremy Reed provided a similar backtrace for wip/firefox:
#0  0x00007f7ff7232f7a in _lwp_park () from /usr/lib/libc.so.12
#1  0x00007f7ffdc0bb03 in pthread__park () from /usr/lib/libpthread.so.0
#2  0x00007f7ffdc0950d in pthread_cond_timedwait ()
   from /usr/lib/libpthread.so.0
#3  0x00007f7ffb724441 in PRP_NakedNotify ()
   from /usr/pkg/lib/firefox/libnspr4.so
#4  0x00007f7ffb724fa9 in PR_WaitCondVar ()
   from /usr/pkg/lib/firefox/libnspr4.so
#5  0x00007f7ffb7250a6 in PR_Wait () from /usr/pkg/lib/firefox/libnspr4.so
#6  0x00007f7ffd5a09f4 in nsThreadPool::Run ()
   from /usr/pkg/lib/firefox/libxul.so
#7  0x00007f7ffd59e8d7 in nsThread::ProcessNextEvent ()
   from /usr/pkg/lib/firefox/libxul.so
#8  0x00007f7ffd566019 in NS_ProcessNextEvent_P ()
   from /usr/pkg/lib/firefox/libxul.so
#9  0x00007f7ffd59ec7d in nsThread::ThreadFunc ()
   from /usr/pkg/lib/firefox/libxul.so
#10 0x00007f7ffb729ebc in PR_JoinThread ()
   from /usr/pkg/lib/firefox/libnspr4.so
#11 0x00007f7ffdc0b39b in pthread_create () from /usr/lib/libpthread.so.0
#12 0x00007f7ff7252f30 in swapcontext () from /usr/lib/libc.so.12
#13 0x00007f7ffdf11000 in ?? ()
#14 0x0000000000000000 in ?? ()

Reference: Message-ID: 
<Pine.NEB.4.64.0803170837070.6082%tx.reedmedia.net@localhost>

Looking at top during the OdfConverter execution, it doesn't seem
to be a memory problem, since it uses about 40MB RAM (on a 2GB machine).
ktrace -di/kdump output ends with:
 15317      1 OdfConverter RET   write 1040/0x410
 15317      1 OdfConverter CALL  lseek(6,0,0,1)
 15317      1 OdfConverter RET   lseek 2580/0xa14
 15317      1 OdfConverter CALL  lseek(6,0,0x612,0)
 15317      1 OdfConverter RET   lseek 1554/0x612
 15317      1 OdfConverter CALL  write(6,0x7f7ffb850000,0xc)
 15317      1 OdfConverter GIO   fd 6 wrote 12 bytes
       "\M^L\M-H\M-mC\M-h\^C\0\0\M^_\^N\0\0"
 15317      1 OdfConverter RET   write 12/0xc
 15317      1 OdfConverter CALL  lseek(6,0,0xa14,0)
 15317      1 OdfConverter RET   lseek 2580/0xa14
 15317      1 OdfConverter CALL  read(3,0x2579ef0,0x10)
 15317      1 OdfConverter GIO   fd 3 read 16 bytes
       "\M^^\M-xF\M-Bw\M-$\M-Q\M-i\M-t\M-.\M-`\^R\M-k\M-mG\M-m"
 15317      1 OdfConverter RET   read 16/0x10
 15317      1 OdfConverter CALL  read(3,0x2579e90,0x10)
 15317      1 OdfConverter GIO   fd 3 read 16 bytes
       "\M-5Y\M-0\M-;,\M-t\M^?A\M^O7\M-7\^C\M-x\^T\M-H\M-C"
 15317      1 OdfConverter RET   read 16/0x10
 15317      1 OdfConverter PSIG  SIGSEGV caught handler=0x4f2450 mask=(11): code
=SEGV_MAPERR, addr=0x0, trap=6)
 15317      1 OdfConverter CALL  write(2,0x138e966,0xd)
 15317      1 OdfConverter GIO   fd 2 wrote 13 bytes
       "Stacktrace:\n\n"
 15317      1 OdfConverter RET   write 13/0xd
 15317      1 OdfConverter CALL  __sigaction_sigtramp(SIGABRT,0x7f7fffffba20,0,0
x7f7ffcd7f660,2)
 15317      1 OdfConverter RET   __sigaction_sigtramp 0
 15317      1 OdfConverter CALL  __sigprocmask14(3,0x7f7fffffba00,0)
 15317      1 OdfConverter RET   __sigprocmask14 0
 15317      1 OdfConverter CALL  write(6,0x7f7ffb850000,0x26)
 15317      1 OdfConverter GIO   fd 6 wrote 38 bytes
       "PK\^C\^D\^T\0\0\0\b\0\M-K\M^C\M-(8\0\0\0\0\0\0\0\0\0\0\0\0\b\0\0\0meta\
        .xml"
 15317      1 OdfConverter RET   write 38/0x26
 15317      1 OdfConverter CALL  getpid
 15317      1 OdfConverter RET   getpid 15317/0x3bd5, 13985/0x36a1
 15317      1 OdfConverter CALL  kill(0x3bd5, SIGABRT)
 15317      1 OdfConverter RET   kill 0
 15317      3 OdfConverter RET   _lwp_park -1 errno 4 Interrupted system call
 15317      3 OdfConverter PSIG  SIGABRT SIG_DFL
 15317      2 OdfConverter RET   nanosleep -1 errno 4 Interrupted system call
 15317      3 OdfConverter NAMI  "OdfConverter.core"

Please let me know if you need further information.

Probably this problem is amd64 specific, since commenters in the firefox
thread say that they can't reproduce it on i386, and kefren@ couldn't
reproduce the problem (with the same input file) on i386/4.99.55.

Please note that not all input files crash OdfConverter.
>How-To-Repeat:
cd /usr/pkgsrc/wip/OdfConverter
make install
OdfConverter /I input.docx /O output.odt
>Fix:
Not provided.

>Unformatted: