Collect -h crashes JVM

Question

I installed Sun Studio Express 07/08 on my Niagara2 machine. When I ran collect with options -j and -h to measure cache misses of Java programs, the JVM crashed with segmentation fault. Without -h option, it works fine. But with any HW counters and with any Java programs, JVM crashes.
# uname -sri
SunOS 5.10 SUNW,SPARC-Enterprise-T5220
# /disk2/StudioExpress/SUNWspro/bin/collect -V
collect: Sun Ceres Analyzer 7.7 SunOS_sparc 2008/07/10
# /disk2/StudioExpress/SUNWspro/bin/collect -j on -h ecdm,10007 -d /tmp /opt/jdk1.6.0_07/bin/java -jar /opt/jdk1.6.0_07/demo/jfc/Java2D/Java2Demo.jar
Creating experiment database /tmp/test.139.er ...
Segmentation Fault (core dumped)
# dbx /opt/jdk1.6.0_07/bin/java core
For information about new features see `help changes'
To remove this message, put `dbxenv suppress_startup_message 7.6' in your .dbxrc
Reading java
core file header read successfully
Reading ld.so.1
Reading libcollector.so
Reading libthread.so.1
Reading libjli.so
Reading libdl.so.1
Reading libc.so.1
Reading libc_psr.so.1
Reading libcpc.so.1
Reading libpctx.so.1
Reading libnvpair.so.1
Reading libdevinfo.so.1
Reading libproc.so.1
Reading libnsl.so.1
Reading libsec.so.1
Reading libgen.so.1
Reading librtld_db.so.1
Reading libelf.so.1
Reading libctf.so.1
Reading libavl.so.1
Reading libjvm.so
Reading libsocket.so.1
Reading libsched.so.1
Reading libm.so.1
Reading libCrun.so.1
Reading libdoor.so.1
Reading libm.so.2
Reading libscf.so.1
Reading libuutil.so.1
Reading libmd_psr.so.1
Reading libmp.so.2
t@1 (l@1) program terminated by signal SEGV (no mapping at the fault address)
0xff336aac: update_map_segments+0x068c: ld [%l3], %i4
(dbx) where
current thread: t@1
=>[1] update_map_segments(0x13, 0xfe7b7fc0, 0xff353aa0, 0xff369b18, 0x0, 0xfde81eb8), at 0xff336aac
[2] dlopen(0xff031490, 0x800, 0x938, 0xffbfef74, 0x0, 0x102), at 0xff337eec
[3] LoadJavaVM(0xffbfef74, 0xffbfef6c, 0xffbff374, 0x10e0a8, 0x10e4f0, 0x2a7f4), at 0x170c8
[4] main(0x0, 0xffbff7e4, 0x42e, 0x5, 0x14, 0xff2c0140), at 0x116fc
(dbx) quit
dbx: internal warning: td_ta_clear_event() failed -- debugger service failed
dbx: internal warning: td_ta_sync_tracking_enable(0) failed -- debugger service failed
Is this a bug, or I did something wrong?

Answer 1

Thank you for the report! Unfortunately I don't have access to a Niagara-2 system right now.
I tried to run a similar test on Niagara-1 system, but it did not help to reproduce the problem.
Probably it is important to use JDK 1.6.0_07 (I have JDK 1.6.0_06 on my Niagara-1 system),
or the problem is caused by "-h ecdm,10007" option. This type of HWC is not available on
Niagara-1 system, only these 2 are available:
Well-known HW counters available for profiling:
insts[1],9999991 (`Instructions Executed', alias for Instr_cnt; events)
Raw HW counters available for profiling:
Instr_cnt[1],1000003 (events)
I'll try to find a Niagara-2 system and let you know if I see the same problem.
BTW, can you try to run the same test with JDK 1.6.0_06?
Thanks,
Nik