[Bug libc/4737] fork is not async-signal-safe

["llucax at gmail dot com" <sourceware-bugzilla@sourceware.org>]

http://sourceware.org/bugzilla/show_bug.cgi?id=4737

--- Comment #20 from Leandro Lucarella <llucax at gmail dot com> 2011-10-13 17:52:35 UTC ---
Created attachment 5987
  --> http://sourceware.org/bugzilla/attachment.cgi?id=5987
Test case for what could be a related problem

I'm having a problem with a rare interaction between fork() and malloc() that
looks like it could be related to this issue.

The test case attached is the most reduced test case I could came up with. This
is based on some trickery done to implement a concurrent garbage collector.

The program eventually hangs and this is the backtrace I get:

Thread 3 (Thread 0x7ffff7028700 (LWP 10504)):
#0  __lll_lock_wait_private () at
../nptl/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:97
#1  0x00007ffff78a837e in _L_lock_36 () from /lib/x86_64-linux-gnu/libc.so.6
#2  0x00007ffff78a067c in ptmalloc_lock_all () at arena.c:288
#3  0x00007ffff78d7635 in __libc_fork () at
../nptl/sysdeps/unix/sysv/linux/x86_64/../fork.c:95
#4  0x0000000000400c3b in do_fork ()
#5  0x00007ffff7bc4d8c in start_thread (arg=0x7ffff7028700) at
pthread_create.c:304
#6  0x00007ffff791004d in clone () at
../sysdeps/unix/sysv/linux/x86_64/clone.S:112
#7  0x0000000000000000 in ?? ()

Thread 2 (Thread 0x7ffff7829700 (LWP 10503)):
#0  0x00007ffff785e084 in do_sigsuspend (set=<value optimized out>) at
../sysdeps/unix/sysv/linux/sigsuspend.c:63
#1  __sigsuspend (set=<value optimized out>) at
../sysdeps/unix/sysv/linux/sigsuspend.c:78
#2  0x0000000000400bc1 in signal_handler ()
#3  <signal handler called>
#4  0x00007ffff78a222f in _int_free (av=0x7ffff7bb91c0, p=0x603240) at
malloc.c:4780
#5  0x00007ffff78a68e3 in __libc_free (mem=<value optimized out>) at
malloc.c:3738
#6  0x0000000000400d3e in do_malloc ()
#7  0x00007ffff7bc4d8c in start_thread (arg=0x7ffff7829700) at
pthread_create.c:304
#8  0x00007ffff791004d in clone () at
../sysdeps/unix/sysv/linux/x86_64/clone.S:112
#9  0x0000000000000000 in ?? ()

Thread 1 (Thread 0x7ffff7fd7720 (LWP 10500)):
#0  0x00007ffff7bc606d in pthread_join (threadid=140737345918720,
thread_return=0x0) at pthread_join.c:89
#1  0x0000000000400f82 in main ()


Is this related to the same lock that prevents fork() from being async-safe? If
this is the case, I think it won't be considered a bug, right?

TIA

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.