archive/jdk
1
0
Fork 0
mirror of https://github.com/openjdk/jdk.git synced 2025-09-19 10:34:38 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

8007476: assert(the_owner != NULL) failed: Did not find owning Java thread for lock word address

Browse source 
Make deadlock detection a little more robust in the case of being unable to find the JavaThread associated with an object lock.

Reviewed-by: sla, acorn
This commit is contained in:
Daniel D. Daugherty 2013-02-27 15:00:30 -08:00
parent 84144898c1
commit 776841b1bc
4 changed files with 47 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

10
hotspot/src/share/vm/prims/jvmtiEnvBase.cpp
View file

@ -997,13 +997,19 @@ JvmtiEnvBase::get_object_monitor_usage(JavaThread* calling_thread, jobject objec
// move our object at this point. However, our owner value is safe
// since it is either the Lock word on a stack or a JavaThread *.
owning_thread = Threads::owning_thread_from_monitor_owner(owner, !at_safepoint);
assert(owning_thread != NULL, "sanity check");
if (owning_thread != NULL) { // robustness
// Cannot assume (owning_thread != NULL) here because this function
// may not have been called at a safepoint and the owning_thread
// might not be suspended.
if (owning_thread != NULL) {
// The monitor's owner either has to be the current thread, at safepoint
// or it has to be suspended. Any of these conditions will prevent both
// contending and waiting threads from modifying the state of
// the monitor.
if (!at_safepoint && !JvmtiEnv::is_thread_fully_suspended(owning_thread, true, &debug_bits)) {
// Don't worry! This return of JVMTI_ERROR_THREAD_NOT_SUSPENDED
// will not make it back to the JVM/TI agent. The error code will
// get intercepted in JvmtiEnv::GetObjectMonitorUsage() which
// will retry the call via a VM_GetObjectMonitorUsage VM op.
return JVMTI_ERROR_THREAD_NOT_SUSPENDED;
}
HandleMark hm;