FreeBSD pl/java problem : backend hangs

[Achilleas Mantzios]

Hello community,

I am running postgresql 9.2.2 with 
openjdk version "1.6.0_32"
OpenJDK Runtime Environment (build 1.6.0_32-b25)
OpenJDK 64-Bit Server VM (build 20.0-b12, mixed mode)
On FreeBSD 8.3-RELEASE

I am having trouble even running simple tests like : 
CREATE FUNCTION getsysprop(VARCHAR) 
  RETURNS VARCHAR 
  AS 'java.lang.System.getProperty' 
  LANGUAGE java; 

SELECT getsysprop('user.home'); 

When i do that, the session hangs, and have to kill -9 the backend.

Searching with truss got :

gettimeofday({1359983241.596250 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596275 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596300 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596327 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596352 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596378 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596403 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596431 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596458 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596482 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596507 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596532 },0x0)           = 0 (0x0)
gettimeofday({1359983241.596556 },0x0)           = 0 (0x0)
_umtx_op(0x7fffffffc7e8,0x3,0x1,0x0,0x0,0x1)     = 0 (0x0)
sigprocmask(SIG_BLOCK,SIGHUP|SIGINT|SIGQUIT|SIGABRT|SIGEMT|SIGKILL|SIGSYS|SIGPIPE|SIGALRM|SIGTERM|SIGURG|SIGSTOP|SIGTSTP|SIGCONT|SIGCHLD|SIGTTIN|SIGTTOU|SIGIO|SIGXCPU|SIGXFSZ|SIGVTALRM|SIGPROF|SIGWINCH|SIGINFO|SIGUSR1|SIGUSR2,0x0) = 0 (0x0)
sigprocmask(SIG_SETMASK,0x0,0x0)                 = 0 (0x0)
sigprocmask(SIG_BLOCK,SIGHUP|SIGINT|SIGQUIT|SIGABRT|SIGEMT|SIGKILL|SIGSYS|SIGPIPE|SIGALRM|SIGTERM|SIGURG|SIGSTOP|SIGTSTP|SIGCONT|SIGCHLD|SIGTTIN|SIGTTOU|SIGIO|SIGXCPU|SIGXFSZ|SIGVTALRM|SIGPROF|SIGWINCH|SIGINFO|SIGUSR1|SIGUSR2,0x0) = 0 (0x0)
sigprocmask(SIG_SETMASK,0x0,0x0)                 = 0 (0x0)

while ktrace gave :

 98910 initial thread CALL  gettimeofday(0x7fffffffc900,0)
 98910 initial thread RET   gettimeofday 0
 98910 initial thread CALL  gettimeofday(0x7fffffffc900,0)
 98910 initial thread RET   gettimeofday 0
 98910 initial thread CALL  gettimeofday(0x7fffffffc900,0)
 98910 initial thread RET   gettimeofday 0
 98910 initial thread CALL  gettimeofday(0x7fffffffc900,0)
 98910 initial thread RET   gettimeofday 0
 98910 initial thread CALL  gettimeofday(0x7fffffffc940,0)
 98910 initial thread RET   gettimeofday 0
 98910 initial thread CALL  _umtx_op(0x7fffffffc7e8,0x3,0x1,0,0)
 98910 initial thread RET   _umtx_op 0
 98910 initial thread CALL  sigprocmask(SIG_BLOCK,0x7fffffffc780,0x8016167d8)
 98910 initial thread RET   sigprocmask 0
 98910 initial thread CALL  sigprocmask(SIG_SETMASK,0x8016167d8,0)
 98910 initial thread RET   sigprocmask 0
 98910 initial thread CALL  sigprocmask(SIG_BLOCK,0x7fffffffc6e0,0x8016167d8)
 98910 initial thread RET   sigprocmask 0
 98910 initial thread CALL  sigprocmask(SIG_SETMASK,0x8016167d8,0)
 98910 initial thread RET   sigprocmask 0
 98910 initial thread CALL  _umtx_op(0x8f8dc4280,0x11,0,0,0)

I found some relevant problems with FreeBSD systems hanging after an _umtx_op but all those point 
to threading problems that now are solved. I found nothing modern about _umtx_op.

I know FreeBSD is not a "Tier-1" platform for postgresql (nor java), but this is my main platform and i'd like
to stick with this.

Is there anything i could do to get closer to make pl/java work with FreeBSD?

Also read this http://postgresql.1045698.n5.nabble.com/Java-VM-stalls-during-quot-install-jar-quot-td3290754.html thread.
Very close to my situation. JNI standalone also worked for me.

Any hints/pointers ?

-
Achilleas Mantzios
IT DEV
IT DEPT