[JVMTI]Run error with SetFieldModificationWatch on every field in SPECJBB2005
Daniel D. Daugherty
daniel.daugherty at oracle.com
Mon Jul 12 10:21:06 PDT 2010
Li jian,
I'm redirecting this e-mail thread to serviceability-dev at ...
as this is a JVM/TI question. hotspot-dev at ... is named by Bcc and
will drop off after this e-mail...
This crash appears to be using a very old version of OpenJDK6.
The OpenJDK6 project is currently at B21 and this crash is from
B12 (Mon Oct 27 16:16:00 EDT 2008).
Can you please retry your experiment with something more recent?
Dan
On 7/10/2010 7:48 PM, 李健 wrote:
> Hi all,
> I've run OpenJDK1.6 on Fedora10 with a JVMTI which watch every
> field access. The field modification callback method simply return and
> do nothing. But OpenJDK randomly throw a bug like this:
>
> /Start User Threads
> started user thread for Warehouse 1
> started user thread for Warehouse 2
> started user thread for Warehouse 3
> Timing Measurement began Sun Jul 11 09:10:48 CST 2010 for 4 minutes
> #
> # An unexpected error has been detected by Java Runtime Environment:
> #
> # SIGSEGV (0xb) at pc=0x00f0ec91, pid=10295, tid=5340048
> #
> # Java VM: OpenJDK Server VM (1.6.0_0-b12 mixed mode linux-x86)
> # Distribution: Custom build (Mon Oct 27 16:16:00 EDT 2008)
> # Problematic frame:
> # V [libjvm.so+0x471c91]
> #
> # An error report file with more information is saved as:
> # /media/_home/lijian/specjbb2005/dist/hs_err_pid10295.log
> #
> # If you would like to submit a bug report, please include
> # instructions how to reproduce the bug and visit:
> # //http://java.sun.com/webapps/bugreport/crash.jsp/
> /#
> Aborted
> /
> Any one can help me with this internal bug? I'll be very appreciate!
> / /
>
> Here's my JVMTI source code:
>
> /static void JNICALL
> FieldModifyCallBack(jvmtiEnv *jvmti_env,
> JNIEnv* jni_env,
> jthread thread,
> jmethodID method,
> jlocation location,
> jclass field_klass,
> jobject object,
> jfieldID field,
> char signature_type,
> jvalue new_value)/
> /{/
> / //do nothing/
> /}/
>
> /static void JNICALL
> ClassPrepareCallback(jvmtiEnv *jvmti_env,
> JNIEnv* jni_env,
> jthread thread,
> jclass klass)
> {
> jint field_num;
> jfieldID * fieldIDs;
> jvmtiError error = jvmti_env->GetClassFields(klass,
> &field_num, &fieldIDs);/
> / if(error != JVMTI_ERROR_NONE)
> fprintf(stderr, "!!!%d\n", error);
> for(int i=0; i< field_num;i++)
> {/
> / error =
> jvmti_env->SetFieldModificationWatch(klass,fieldIDs[i]);
> if(error != JVMTI_ERROR_NONE)
> fprintf(stderr, "!!!\n");
> }
> error = jvmti_env->Deallocate((unsigned char*)fieldIDs);
> if(error != JVMTI_ERROR_NONE)
> fprintf(stderr, "!!!\n");/
> /}/
>
> /JNIEXPORT jint JNICALL Agent_OnLoad(JavaVM *jvm, char *options, void
> *reserved)
> {
> static jvmtiEnv *jvmti = NULL;
> static jvmtiCapabilities capa;
> jvmtiError error;
>
> jint ret = (jvm)->GetEnv((void **)(&jvmti), JVMTI_VERSION_1_0);
> (void)memset(&capa, 0, sizeof(jvmtiCapabilities));
> capa.can_generate_field_modification_events = 1;
> error = jvmti->AddCapabilities(&capa);
> if(error != JVMTI_ERROR_NONE)
> fprintf(stderr, "!!!\n");/
> / error = jvmti->SetEventNotificationMode(JVMTI_ENABLE,
> JVMTI_EVENT_CLASS_PREPARE, (jthread)NULL);
> if(error != JVMTI_ERROR_NONE)
> fprintf(stderr, "!!!\n");
> error = jvmti->SetEventNotificationMode(JVMTI_ENABLE,
> JVMTI_EVENT_FIELD_MODIFICATION, NULL);
> if(error != JVMTI_ERROR_NONE)
> fprintf(stderr, "!!!\n");
>
> jvmtiEventCallbacks callbacks;
> (void)memset(&callbacks, 0, sizeof(callbacks));/
> / callbacks.ClassPrepare= &ClassPrepareCallback;
> callbacks.FieldModification = &FieldModifyCallBack;
> error =
> jvmti->SetEventCallbacks(&callbacks,(jint)sizeof(callbacks));
> if(error != JVMTI_ERROR_NONE)
> fprintf(stderr, "!!!\n");
>
> return JNI_OK;
> }
> /
> /JNIEXPORT void JNICALL Agent_OnUnload(JavaVM *jvm)
> {/
> / return;
> }/
>
> Here's my openJDK version:
>
> /[root at localhost dist]# java -versionjava version "1.6.0_0"
> IcedTea6 1.3.1 (6b12-Fedora-10) Runtime Environment (build 1.6.0_0-b12)
> OpenJDK Server VM (build 1.6.0_0-b12, mixed mode)/
> //
> The attachment is /hs_err_pid10295.log/
>
>
> Thanks a lot!
> --
> Li jian
> China
More information about the serviceability-dev
mailing list