###@###.### 2003-06-19

A GC operation starts after AsyncGetCallTrace() makes it past
its 'is GC running' check. The GC relocates a methodID that we
are accessing during the stack walk and we crash.

Here is a stack trace snippet for the JavaThread calling
AsyncGetCallTrace():

THREAD t@12

  ---- called from signal handler with signal 11 (SIGSEGV) ------
  [10] jniIdSupport::to_jmethod_id_or_null(0xd674a9a8), at 0xdf8a872f
  [11] methodOopDesc::find_jmethod_id_or_null(0xd674a9a8), at 0xdf99b0f8
  [12] forte_fill_call_trace_given_top(0x81f3428, 0xd22df1bc, 0x32, 0xd22df568,
0xda86b536, 0xd22df570), at 0xdf84b927
  [13] AsyncGetCallTrace(0xd22df1bc, 0x32, 0xd22df27c), at 0xdf84bbbb
=>[14] profhandler(sig = 29, siginfo = 0xd22df47c, ucontext = 0xd22df27c), line
138 in "b4757672.c"
  [15] __sighndlr(0x1d, 0xd22df47c, 0xd22df27c, 0xdd241008), at 0xdfb92963
  [16] call_user_handler(0x1d, 0xd22df47c, 0xd22df27c), at 0xdfb8e0b6
  [17] sigacthandler(), at 0xdfb8e1e0
  ---- called from signal handler with signal 29 (SIGPROF) ------
  [18] __lwp_mutex_unlock(0x8069238), at 0xdfaeb80c
  [19] Mutex::unlock(0x8069238), at 0xdf6f257d
  [20] SafepointSynchronize::block(0x81f3428), at 0xdf79ca2d
  [21] Runtime1::resolve_static_call(), at 0xdf80a9fe
  [22] 0xda86b536(), at 0xda86b535

Here is a stack trace snippet for the VMThread doing a GC:

THREAD t@2

t@2(l@2) stopped in ConstantPoolCacheEntry::adjust_pointers at 0xdf7c3917
0xdf7c3917: adjust_pointers+0x0027: andl $0x4000000,%eax
current thread: t@2
=>[1] ConstantPoolCacheEntry::adjust_pointers(0xd6703398), at 0xdf7c3917
  [2] constantPoolCacheKlass::oop_adjust_pointers(0xd6400c48, 0xd6703328), at 0x
df7c38da
  [3] CompactibleSpace::adjust_pointers(0x80de818), at 0xdf7c2f0a
  [4] AdjustPointersClosure::do_space(0xdd34fc00, 0x80de818), at 0xdf7c2e6b
  [5] OneContigSpaceCardGeneration::space_iterate(0x80de708, 0xdd34fc00, 0x1), a
t 0xdf7c2e44
  [6] Generation::adjust_pointers(0x80de708), at 0xdf7c2e0c
  [7] GenMarkSweep::mark_sweep_phase3(0x1), at 0xdf856400
  [8] GenMarkSweep::invoke_at_safepoint(0x1, 0x80de254, 0x0), at 0xdf855e62
  [9] OneContigSpaceCardGeneration::collect(0x80de208, 0x1, 0x0, 0x0, 0x0, 0x0),
 at 0xdf7c01b0
  [10] TenuredGeneration::collect(0x80de208, 0x1, 0x0, 0x0, 0x0, 0x0), at 0xdf9d
8902
  [11] GenCollectedHeap::do_collection(0x80dc7e8, 0x1, 0x0, 0x0, 0x0, 0x0, 0x1,
0xd224fbfc), at 0xdf7a81c9
  [12] GenCollectedHeap::do_full_collection(0x80dc7e8, 0x0, 0x1, 0xd224fbfc), at
 0xdf7cabb9
  [13] VM_GenCollectFull::doit(0xd224fbe0), at 0xdf7cab87
  [14] VM_Operation::evaluate(0xd224fbe0), at 0xdf79c76c
  [15] VMThread::evaluate_operation(0x81031b0, 0xd224fbe0), at 0xdf79c65b
  [16] VMThread::loop(0x81031b0), at 0xdf73a401
  [17] VMThread::run(0x81031b0), at 0xdf73a096

The complete stack trace is attached as threads.log.287.


###@###.### 2004-01-28

This can also happen after the VMThread has finished doing a GC when
the target thread just happens to get a chance to run again:

  [10] sigacthandler(0xfeff1800, 0xef67ea68, 0xef67e7b0, 0xff396000, 0xef67ea68,
 0x4), at 0xff37fccc
  ---- called from signal handler with signal 4 (SIGILL) ------
  [11] 0xf4401148(0xf4bcfef0, 0x0, 0x0, 0x0, 0x0, 0x0), at 0xf4401147
  [12] jniIdPrivate::id_for_or_null(k = 0xf4bcfef0, index = 1), line 389 in "jni
Id.cpp"
  [13] jniIdSupport::to_jmethod_id_or_null(method_oop = 0xf4bcfe88), line 531 in
 "jniId.cpp"
  [14] methodOopDesc::find_jmethod_id_or_null(this = 0xf4bcfe88), line 740 in "m
ethodOop.cpp"
  [15] forte_fill_call_trace_given_top(thd = 0x5dd3d8, trace = 0xef67f024, depth
 = 50, top_frame = CLASS), line 773 in "forte.cpp"
  [16] AsyncGetCallTrace(trace = 0xef67f024, depth = 50, ucontext = 0xef67f1f8),
 line 913 in "forte.cpp"
  [17] profhandler(sig = 29, siginfo = 0xef67f4b0, ucontext = 0xef67f1f8), line
171 in "b4757672.c"
  [18] __sighndlr(0x1d, 0xef67f4b0, 0xef67f1f8, 0xff0113e0, 0x0, 0x0), at 0xff38
4cc8
  [19] call_user_handler(0xfeff1800, 0x2b, 0xff397b20, 0xef67f1f8, 0xef67f4b0, 0
x1d), at 0xff37fb00
  [20] sigacthandler(0xfeff1800, 0xef67f4b0, 0xef67f1f8, 0xff396000, 0xef67f4b0,
 0x1d), at 0xff37fccc
  ---- called from signal handler with signal 29 (SIGPROF) ------
  [21] _libc_poll(0x0, 0x0, 0x1e, 0xe1a, 0xf041e320, 0x0), at 0xff31d348
  [22] _ti_poll(0x0, 0xfeff1800, 0x1e, 0x0, 0x0, 0xef67f618), at 0xff37da7c
  [23] os_sleep(millis = 30LL, interruptible = 1), line 2207 in "os_solaris.cpp"
  [24] os::sleep(thread = 0x5dd3d8, millis = 30LL, interruptible = 1), line 2287
 in "os_solaris.cpp"
  [25] JVM_Sleep(env = 0x5dd4ac, threadClass = 0xef67f7d4, millis = 30LL), line
2468 in "jvm.cpp"

In this crash, the VMThread was just waiting for something to do.
The complete thread dump is attached as threads.log.48.