You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@harmony.apache.org by "Aleksey Ignatenko (JIRA)" <ji...@apache.org> on 2007/06/18 14:15:26 UTC
[jira] Created: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
[drlvm][jit] assert in JIT while unwinding stack (during enumeration)
---------------------------------------------------------------------
Key: HARMONY-4215
URL: https://issues.apache.org/jira/browse/HARMONY-4215
Project: Harmony
Issue Type: Bug
Components: DRLVM
Environment: Win 2003
Reporter: Aleksey Ignatenko
2 reliablity tests failed on assertion in jit:
1. CharsetSyncCacheTest
2. ZlibTest
svn rev 548320
Call stack:
jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
03bf33ef()
harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Resolved: (HARMONY-4215) [drlvm][thread][reliability] assert
in JIT while unwinding stack (during enumeration)
Posted by "Gregory Shimansky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Gregory Shimansky resolved HARMONY-4215.
----------------------------------------
Resolution: Fixed
Assignee: Gregory Shimansky (was: weldon washburn)
Patch to exclude list is applied at 613164. Closing the bug as it is slit into parts.
> [drlvm][thread][reliability] assert in JIT while unwinding stack (during enumeration)
> -------------------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Assignee: Gregory Shimansky
> Attachments: 2rel_tests_fail.dmp, exclude_list.patch, H4215_tm_races.patch, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Aleksey Ignatenko (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12505800 ]
Aleksey Ignatenko commented on HARMONY-4215:
--------------------------------------------
CharsetSyncCacheTest crash was reproduced on release build
ZlibTest crash was reproduced on debug build
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Gregory Shimansky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12518452 ]
Gregory Shimansky commented on HARMONY-4215:
--------------------------------------------
In addition to the above stack trace I get this one
> msvcr71d.dll!_assert(const char * expr=0x01498bc8, const char * filename=0x01498b60, unsigned int lineno=160) Line 295 C
gc_gen.dll!collector_trace_rootsets(Collector * collector=0x015e2540) Line 160 + 0x1c bytes C++
gc_gen.dll!nongen_forward_pool(Collector * collector=0x015e2540) Line 222 + 0x9 bytes C++
gc_gen.dll!collector_thread_func(void * arg=0x015e2540) Line 186 + 0x9 bytes C++
hythr.dll!thread_start_proc(void * arg=0x00426a88) Line 710 + 0x9 bytes C
hythr.dll!_threadstartex(void * ptd=0x004266f8) Line 241 + 0xd bytes C
kernel32.dll!77e66063()
in about as many cases as the one above. The object in *ref is NULL.
Both cases look like problems with enumeration.
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Aleksey Ignatenko (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Aleksey Ignatenko updated HARMONY-4215:
---------------------------------------
Attachment: 2rel_tests_fail.dmp
2rel_tests_fail.dmp - minidump file showing the crash.
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Pavel Ozhdikhin (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Pavel Ozhdikhin updated HARMONY-4215:
-------------------------------------
Attachment: util.Date_int.out.err
Today I ran into the util.Date smoke test failure on the interpreter with very similar stack trace (please see attached). So, this failure is not pertained to the mentioned reliability test only.
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Yu-Nan He (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12510305 ]
Yu-Nan He commented on HARMONY-4215:
------------------------------------
Hi, I ran ZlibTest on JET and found different symptom.
I found that the root set was changed after the "stop the world" root set enumeration ( vm_enumerate_root_set_all_threads() ).
The root set is enumerated after threads are suspended, so root set should not be changed after enumeration.
It looks like the some thread modifies the root set during "stop the world".
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][thread][reliability]
assert in JIT while unwinding stack (during enumeration)
Posted by "Andrey Yakushev (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12560362#action_12560362 ]
Andrey Yakushev commented on HARMONY-4215:
------------------------------------------
I agree, let's close this unfocused issue, and open test specific.
Pavel, could you also please correct reliability suite exclude list in order to remove tests which are currently pass, and change issue number in comments for others.
> [drlvm][thread][reliability] assert in JIT while unwinding stack (during enumeration)
> -------------------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Assignee: weldon washburn
> Attachments: 2rel_tests_fail.dmp, H4215_tm_races.patch, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Aleksey Ignatenko (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12508752 ]
Aleksey Ignatenko commented on HARMONY-4215:
--------------------------------------------
Some other reliablity tests failed with the similar crash on release build on 06/27/2007:
api.kernel.thread.Calculation.CalcTest, api.kernel.thread.LifeSimulationTest.Life, api.kernel.thread.ThreadLocalTest.ThreadLocalTest, api.text.getSentenceInstance, api.zip.ZlibTest,
Call Satck:
Windows reported exception: 0xc0000005
Registers:
EAX: 0x01b18c00, EBX: 0x00000000, ECX: 0x000002f5, EDX: 0x00000000
ESI: 0x01b18ef4, EDI: 0x01941788, ESP: 0x0013f3d0, EBP: 0x0064c940
EIP: 0x016d11a9
Stack trace:
0: ?? (??:-1)
1: java/lang/VMMemoryManager.runGC()V (VMMemoryManager.java:-2)
2: java/lang/Runtime.gc()V (Runtime.java:679)
3: java/lang/System.gc()V (System.java:107)
4: org/apache/harmony/test/reliability/share/LongRunningTestStarter.test([Ljava/lang/String;)I (??:-1)
5: org/apache/harmony/test/reliability/share/ReliabilityTestStarter.test([Ljava/lang/String;)I (??:-1)
6: org/apache/harmony/test/reliability/share/ReliabilityTestStarter.main([Ljava/lang/String;)V (??:-1)
LTend of stack traceGT
]]></system-out>
</testsuite>
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Aleksey Ignatenko (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12508754 ]
Aleksey Ignatenko commented on HARMONY-4215:
--------------------------------------------
Here is additional stack trace on ThreadLocalTest:
Windows reported exception: 0xc0000005
Registers:
EAX: 0x01b07400, EBX: 0x00000000, ECX: 0x0000026c, EDX: 0x00000000
ESI: 0x01b0766b, EDI: 0x01941270, ESP: 0x0013f360, EBP: 0x0064c980
EIP: 0x016d11a9
Stack trace:
0: sync_stack_push (c:\cc-bt2.0\2.0\build\checkouts\drlvm\vm\gc_gen\src\utils\sync_stack.h:129)
1: java/lang/VMMemoryManager.runGC()V (VMMemoryManager.java:-2)
2: java/lang/Runtime.gc()V (Runtime.java:679)
3: java/lang/System.gc()V (System.java:107)
4: org/apache/harmony/test/reliability/share/LongRunningTestStarter.test([Ljava/lang/String;)I (??:-1)
5: org/apache/harmony/test/reliability/share/ReliabilityTestStarter.test([Ljava/lang/String;)I (??:-1)
6: org/apache/harmony/test/reliability/share/ReliabilityTestStarter.main([Ljava/lang/String;)V (??:-1)
LTend of stack traceGT
]]></system-out>
</testsuite>
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Pavel Ozhdikhin (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Pavel Ozhdikhin updated HARMONY-4215:
-------------------------------------
Attachment: test.zip
I've attached the script which runs the ZlibTest test in a loop. Just unzip test.zip, update JAVA definition in run-test.sh and run it.
Please note that the test parameters in the script are more agressive than in the test suite to speed up reproducing.
With this script the issue is stably reproducible on the interpreter:
...
The GC did not provide gc_add_weak_root_set_entry()
OK
6
The GC did not provide gc_add_weak_root_set_entry()
OK
7
The GC did not provide gc_add_weak_root_set_entry()
OK
Windows reported exception: 0xc0000005
Registers:
EAX: 0x00000020, EBX: 0x7ffd5000, ECX: 0xddddddbd, EDX: 0xddddddbd
ESI: 0x00001000, EDI: 0xdddddddc, ESP: 0x0013de34, EBP: 0x0013de60
EIP: 0x77e6b268
Stack trace:
0: IsBadReadPtr (??:-1)
1: _CrtIsValidPointer (f:\vs70builds\3077\vc\crtbld\crt\src\dbgheap.c:1710)
2: _CrtIsValidHeapPointer (f:\vs70builds\3077\vc\crtbld\crt\src\dbgheap.c:1741)
3: _free_dbg_lk (f:\vs70builds\3077\vc\crtbld\crt\src\dbgheap.c:1132)
4: _free_dbg (f:\vs70builds\3077\vc\crtbld\crt\src\dbgheap.c:1070)
5: _freefls (f:\vs70builds\3077\vc\crtbld\crt\src\tidtable.c:361)
6: FlsFree (??:-1)
7: _mtterm (f:\vs70builds\3077\vc\crtbld\crt\src\tidtable.c:220)
8: _CRT_INIT (f:\vs70builds\3077\vc\crtbld\crt\src\dllcrt0.c:205)
9: _DllMainCRTStartup (f:\vs70builds\3077\vc\crtbld\crt\src\dllcrt0.c:312)
10: LdrInitializeThunk (??:-1)
11: RtlEnumerateGenericTableWithoutSplaying (??:-1)
12: GetEnvironmentVariableA (??:-1)
13: ExitProcess (??:-1)
14: __crtExitProcess (f:\vs70builds\3077\vc\crtbld\crt\src\crt0dat.c:464)
15: doexit (f:\vs70builds\3077\vc\crtbld\crt\src\crt0dat.c:414)
16: _exit (f:\vs70builds\3077\vc\crtbld\crt\src\crt0dat.c:311)
17: Java_java_lang_VMExecutionEngine_exit (c:\users\pozhdikh\harmony\trunk\working_vm\vm\vmcore\src\kernel_classes\native\java_lang_vmexecutionengine.cpp:54)
18: invokeJNI (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interp_native_ia32.cpp:50)
19: interpreterInvokeStaticNative (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interp_native_ia32.cpp:359)
20: interpreterInvokeStatic (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:3312)
21: Opcode_INVOKESTATIC (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:2104)
22: java/lang/VMExecutionEngine.exit(IZ)V (VMExecutionEngine.java:72)
23: interpreterInvokeStatic (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:3337)
24: Opcode_INVOKESTATIC (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:2104)
25: java/lang/Runtime.exit(I)V (Runtime.java:464)
26: interpreterInvoke (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:3427)
27: interpreterInvokeVirtual (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:3509)
28: Opcode_INVOKEVIRTUAL (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:2060)
29: java/lang/System.exit(I)V (System.java:100)
30: interpreterInvokeStatic (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:3337)
31: Opcode_INVOKESTATIC (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:2104)
32: org/apache/harmony/test/reliability/api/zip/ZlibTest.main([Ljava/lang/String;)V (??:-1)
33: interpreter_execute_method (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interpreter.cpp:3230)
34: JIT_execute_method (c:\users\pozhdikh\harmony\trunk\working_vm\vm\interpreter\src\interp_exports.cpp:166)
35: DrlEMImpl::executeMethod (c:\users\pozhdikh\harmony\trunk\working_vm\vm\em\src\drlemimpl.cpp:509)
36: ExecuteMethod (c:\users\pozhdikh\harmony\trunk\working_vm\vm\em\src\em_intf.cpp:43)
37: vm_execute_java_method_array (c:\users\pozhdikh\harmony\trunk\working_vm\vm\vmcore\src\jit\ini.cpp:60)
38: call_static_method_no_ref_result (c:\users\pozhdikh\harmony\trunk\working_vm\vm\vmcore\src\jni\jni_method.cpp:1154)
39: CallStaticVoidMethodA (c:\users\pozhdikh\harmony\trunk\working_vm\vm\vmcore\src\jni\jni_method.cpp:1562)
40: CallStaticVoidMethodV (c:\users\pozhdikh\harmony\trunk\working_vm\vm\vmcore\src\jni\jni_method.cpp:1547)
41: CallStaticVoidMethod (c:\users\pozhdikh\harmony\trunk\working_vm\vm\vmcore\src\jni\jni_method.cpp:1537)
42: main_runJavaMain (c:\users\pozhdikh\harmony\trunk\working_classlib\modules\luni\src\main\native\launcher\shared\main.c:1321)
43: invocation (c:\users\pozhdikh\harmony\trunk\working_classlib\modules\luni\src\main\native\launcher\shared\main.c:742)
44: gpProtectedMain (c:\users\pozhdikh\harmony\trunk\working_classlib\modules\luni\src\main\native\launcher\shared\main.c:391)
45: main (c:\users\pozhdikh\harmony\trunk\working_classlib\modules\luni\src\main\native\launcher\shared\cmain.c:146)
46: mainCRTStartup (f:\vs70builds\3077\vc\crtbld\crt\src\crtexe.c:398)
47: IsProcessorFeaturePresent (??:-1)
<end of stack trace>
bash-2.05b$
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Pavel Ozhdikhin (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12507112 ]
Pavel Ozhdikhin commented on HARMONY-4215:
------------------------------------------
Unpinned objects are copied to unmanaged temporary memory space so pinning is not the cause here. Need to continue investigation...
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Pavel Afremov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12513537 ]
Pavel Afremov commented on HARMONY-4215:
----------------------------------------
In JIRA HARMONY-4264 was very similar problem with EIP==0. So patch from JIRA HARMONY-4264 may be helpful.
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Aleksey Ignatenko (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12506117 ]
Aleksey Ignatenko commented on HARMONY-4215:
--------------------------------------------
to reproduce it on ZlibTest test one needs to run
ant -f run.xml run.api.zip
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Xiao-Feng Li (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12510313 ]
Xiao-Feng Li commented on HARMONY-4215:
---------------------------------------
Yu-Nan's finding is very important. I think this is confirmed partially by other people's comments that this failure is more subtle than expected. It happens actually with a big group of test cases. We probably want some threading guys to examine if this is possible for a mutator to modify the runtime stack after stop-the-world.
In current DRLVM design, this is possible, but, it by design should only happen in safe region. The safe region means the mutator may modify the stack but never modifies root set entries. I am afraid this design principle is not 100% complied by our code (specifically, the native code).
Do we have a good way to verify all the safe regions are well-formed? This helps not only to this bug, but also the overall DRLVM quality. Currently in my understanding only the assertion for suspension enable and disable helps to check safe region, but it does not help to check if a root set entry is modified.
Btw, in my tests, both GCv4.1 and GCv5 have failures (with different symptoms).
Thanks.
xiaofeng
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Pavel Ozhdikhin (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12508133 ]
Pavel Ozhdikhin commented on HARMONY-4215:
------------------------------------------
A brief summary according to my findings:
The bug is reproducible on all JITs and interpreter.
It is reproducible with gc_gen and not reproducible with gc_cc.
It is not reproducible on J9.
We need an expert in GC or VM to take a deeper look here.
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "weldon washburn (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12527662 ]
weldon washburn commented on HARMONY-4215:
------------------------------------------
Evgeuni,
Thanks for the analysis. It is good work. A few of your discoveries are new to me. And a few of the below items actually hit on the key thread state design issues documented in Harmony-4716. Inline comments are below.
I want to share with you my findings during HARMONY-4215 investigation. In total I found 8 issues in TM component. By fixing/workarounding these issues I managed to pass HARMONY-4215. I'm not sure if all of them are unknown.
1) Problem state: New threads may appear and switch to suspend disabled state (so performing dangerous operations) during GC cycle. That may happen because hythread_suspend_all() releases the global lock which guards attachment of new threads.
[Washburn, Weldon] I did not realize this. We have to find out if this will cause deadlock. This looks like it should be part of the Thread State Transition project.
Possible solution: Keep global lock until hythread_resume_all() is called.
2) Problem state: Race condition in vm_attach(). Last action of jthread_allocate_vm_thread is to associate native with vm_thread. From that moment this thread becomes available for GC enumeration. But by that time it's not yet fully initialized. For example M2N frame is not allocated and initialized. Possible solution: Native and VM thread association should be last action with atomic effect.
[Washburn, Weldon] I think we just fixed this bug in a recent Thread Lifecycle commit.
3) Problem state: Race condition in hythread_suspend_disable(). This one was mentioned by Weldon on the list. But only one side of the problem was discussed. To guarantee correct work of suspension machinery two conditions should be met. The first one is that suspender thread MUST observe exactly the same value of disable_count as it is assigned before suspendee thread reads suspend_request flag. The second one is that suspendee thread MUST increment disable_count before it reads suspend_request flag.
Possible solution: Add memory read/write barrier between write operation to disable_count and read operation from suspend_request.
[Washburn, Weldon] This is a major concern. My current thoughts are covered in JIRA H4716. It looks like this one should be part of the Thread State Transition project. Briefly there are two big concerns. One, does SFENCE opcode make a guarantee that a write will appear in the coherency domain before the next instruction is retired? I think the answer is yes but the Intel documentation is vague. Two, what is the performance impact of an SFENCE? We tried to model exactly the above with lockcmpxchg but found the performance unacceptable. Also note the proposed patch uses apr_memory_rw_barrier() which apparently is defined to be __asm {lock add [esp], 0 } In other words, ...rw_barrier does not use SFENCE. My guess is that while a lockcmpxchg is a superset of SFENCE, it has a performance penalty.
4) Problem state: Potentially illegal transition from suspend disabled to suspend enabled mode in hythread_global_lock. When some thread is trying to acquire TM's global lock in disabled state it will be quietly switched in enabled state if GC wants to happen. But such transition is illegal if thread already performed "unsafe" operations and is not ready for GC to happen. This issue affects overall stability in a negative way and can be hard to identify.
Possible solution: I think the best choice here to follow the rule that TM's global lock should be grabbed before switching to disabled state.
[Washburn, Weldon] hmm... Maybe. There could be deadlock and performance impact. Note that code outside the Thread Manager that misuses Thread Manager interfaces, while absolutely undesirable, is not really a Thread Manager internals problem. In other words, this is definitely a serious problem that needs to be addressed seperately from the Thread Manager problems.
5) Problem state: Race condition in hythread_global_lock. At the very end of this function disable_count is assigned to its original value. Another thread may send suspend request to the current and check disable_count (which is zero) just after current thread have read suspend_request (read value is zero) flag and changed disable_count. Thus suspender thread thinks that it successfully suspended target thread and target thread in the same time enters unsafe region.
Possible solution: Use hythread_set_suspend_disable to change the value of disable_count which takes care about described situation. Unfortunately, this solution is not perfect because it opens small window when deadlock can happened. This is exactly the same type of deadlock which current procedure is trying to avoid. So the best solution will be the same as described in item 4.
[Washburn, Weldon] This looks like it should be part of the Thread State Transition project.
6) Problem state: Race condition in thread_safe_point_impl(). The problem is exactly the same as described in item 5. The very last assignment to disable_count can be noted by suspender thread to late and the thread will enter unsafe region while suspender thinks it's ok to procced.
Possible solution: Use hythread_set_suspend_disable to change the value of disable_count.
[Washburn, Weldon] Yes. This is a key problem that Thread State Transition project intends to address. Please see H4716.
7) Problem state: Race condition in hythread_set_suspend_disable. See item 2 for description and solution.
8) Problem state: Minor concern regarding thread attach/detach procedure. It seems to be reasonable from stability POV to hold TM's global lock during the whole attach/detach period. That can help to avoid nasty race conditions like mentioned in item 2. So in the attached patch the problem 2. is fixed/workarounded this way.
[Washburn, Weldon] hmm.... I think the code already does what is suggested. I have created a patch for vm shutdown. Let's wait and see what bug(s) remain once we get Thread Lifecycle and vm shutdown patches committed.
I didn't test attached patch thoroughly. I've only checked that it fixes HARMONY-4215. I also didn't check performance impact. Washburn, Weldon] OK.
So don't take it as final rather just as a demonstration of problems.
[Washburn, Weldon] Good idea. Your work has been very helpful. Hopefully we fix these bugs during Q4 when the Thread State Transition project get going.
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Assignee: weldon washburn
> Attachments: 2rel_tests_fail.dmp, H4215_tm_races.patch, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4215) [drlvm][thread][reliability] assert
in JIT while unwinding stack (during enumeration)
Posted by "Pavel Rebriy (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Pavel Rebriy updated HARMONY-4215:
----------------------------------
Attachment: exclude_list.patch
Here is an updated exclude list as asked.
> [drlvm][thread][reliability] assert in JIT while unwinding stack (during enumeration)
> -------------------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Assignee: weldon washburn
> Attachments: 2rel_tests_fail.dmp, exclude_list.patch, H4215_tm_races.patch, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Yu-Nan He (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12519841 ]
Yu-Nan He commented on HARMONY-4215:
------------------------------------
Please try check_root_set() under Win2003 IA32.
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Evgueni Brevnov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Evgueni Brevnov updated HARMONY-4215:
-------------------------------------
Attachment: H4215_tm_races.patch
I want to share with you my findings during HARMONY-4215 investigation. In total I found 8 issues in TM component. By fixing/workarounding these issues I managed to pass HARMONY-4215. I'm not sure if all of them are unknown.
1) Problem state: New threads may appear and switch to suspend disabled state (so performing dangerous operations) during GC cycle. That may happen because hythread_suspend_all() releases the global lock which guards attachment of new threads.
Possible solution: Keep global lock until hythread_resume_all() is called.
2) Problem state: Race condition in vm_attach(). Last action of jthread_allocate_vm_thread is to associate native with vm_thread. From that moment this thread becomes available for GC enumeration. But by that time it's not yet fully initialized. For example M2N frame is not allocated and initialized. Possible solution: Native and VM thread association should be last action with atomic effect.
3) Problem state: Race condition in hythread_suspend_disable(). This one was mentioned by Weldon on the list. But only one side of the problem was discussed. To guarantee correct work of suspension machinery two conditions should be met. The first one is that suspender thread MUST observe exactly the same value of disable_count as it is assigned before suspendee thread reads suspend_request flag. The second one is that suspendee thread MUST increment disable_count before it reads suspend_request flag.
Possible solution: Add memory read/write barrier between write operation to disable_count and read operation from suspend_request.
4) Problem state: Potentially illegal transition from suspend disabled to suspend enabled mode in hythread_global_lock. When some thread is trying to acquire TM's global lock in disabled state it will be quietly switched in enabled state if GC wants to happen. But such transition is illegal if thread already performed "unsafe" operations and is not ready for GC to happen. This issue affects overall stability in a negative way and can be hard to identify.
Possible solution: I think the best choice here to follow the rule that TM's global lock should be grabbed before switching to disabled state.
5) Problem state: Race condition in hythread_global_lock. At the very end of this function disable_count is assigned to its original value. Another thread may send suspend request to the current and check disable_count (which is zero) just after current thread have read suspend_request (read value is zero) flag and changed disable_count. Thus suspender thread thinks that it successfully suspended target thread and target thread in the same time enters unsafe region.
Possible solution: Use hythread_set_suspend_disable to change the value of disable_count which takes care about described situation. Unfortunately, this solution is not perfect because it opens small window when deadlock can happened. This is exactly the same type of deadlock which current procedure is trying to avoid. So the best solution will be the same as described in item 4.
6) Problem state: Race condition in thread_safe_point_impl(). The problem is exactly the same as described in item 5. The very last assignment to disable_count can be noted by suspender thread to late and the thread will enter unsafe region while suspender thinks it's ok to procced.
Possible solution: Use hythread_set_suspend_disable to change the value of disable_count.
7) Problem state: Race condition in hythread_set_suspend_disable. See item 2 for description and solution.
8) Problem state: Minor concern regarding thread attach/detach procedure. It seems to be reasonable from stability POV to hold TM's global lock during the whole attach/detach period. That can help to avoid nasty race conditions like mentioned in item 2. So in the attached patch the problem 2. is fixed/workarounded this way.
I didn't test attached patch thoroughly. I've only checked that it fixes HARMONY-4215. I also didn't check performance impact. So don't take it as final rather just as a demonstration of problems.
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, H4215_tm_races.patch, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Yu-Nan He (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12519838 ]
Yu-Nan He commented on HARMONY-4215:
------------------------------------
I wrote a small function to check the root set before tracing.
void check_root_set(GC* gc)
{
GC_Metadata* metadata = gc->metadata;
pool_iterator_init(metadata->gc_rootset_pool);
Vector_Block* root_set = pool_iterator_next(metadata->gc_rootset_pool);
while(root_set){
POINTER_SIZE_INT* iter = vector_block_iterator_init(root_set);
while(!vector_block_iterator_end(root_set,iter)){
REF *p_ref = (REF *)*iter;
/* root ref cann't be NULL */
assert(*p_ref);
iter = vector_block_iterator_advance(root_set, iter);
}
root_set = pool_iterator_next(metadata->gc_rootset_pool);
}
}
(You can place this function in gc_metadata.cpp)
If you want to check the root set, please call this function after enumeration.
In gc_common.cpp :
......
gc_reset_rootset(gc);
vm_enumerate_root_set_all_threads();
gc_copy_interior_pointer_table_to_rootset();
gc_set_rootset(gc);
check_root_set(gc); <-----function call is placed here.
......
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4215) [drlvm][thread][reliability] assert
in JIT while unwinding stack (during enumeration)
Posted by "Gregory Shimansky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Gregory Shimansky updated HARMONY-4215:
---------------------------------------
Estimated Complexity: Guru (was: Advanced)
Summary: [drlvm][thread][reliability] assert in JIT while unwinding stack (during enumeration) (was: [drlvm][reliability] assert in JIT while unwinding stack (during enumeration))
Marking as threading specific bug based on the above evaluation.
> [drlvm][thread][reliability] assert in JIT while unwinding stack (during enumeration)
> -------------------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Assignee: weldon washburn
> Attachments: 2rel_tests_fail.dmp, H4215_tm_races.patch, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "weldon washburn (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12521858 ]
weldon washburn commented on HARMONY-4215:
------------------------------------------
It seems we need to narrow down what might be causing the problem. I have some ideas on how to proceed.
1)
If it is a race condition, typically each run will report different results. This is because the OS is scheduling threads slightly differently from run-to-run and the threads access the coherent caches at slightly different timings. Question -- does "assert(*p_ref):" always hit at the same spot or does it vary from run to run?
2)
One way to determine if there is a bug in thread suspend/resume that is causing problems is to add a sleep(Zmsec) immediately after the Thread Manager suspends the threads. If the sleep() makes the problem go away, it might be caused by Thread Manager erroneously thinking a thread is properly suspended when in fact it is not. A good place to add the diagnostic sleep() is immedately after hythread_suspend_all() in stop_the_world_root_set_enum.cpp::stop_the_world_root_set_enumeration(). This is line 95.
Picking a value of "Z" might be tricky. Too big might cause the app to fail or disturb the execution patterms too much. Too little might not be effective. This will take some experimentation.
It would be great if someone could run the above tests and report the results. Right now, I am out of time.
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][thread][reliability]
assert in JIT while unwinding stack (during enumeration)
Posted by "Pavel Rebriy (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12560358#action_12560358 ]
Pavel Rebriy commented on HARMONY-4215:
---------------------------------------
I've closely looked at the bug and checked tests mention in it on Windows 2003 32-bits platform.
The results are below. There are 7 tests related to the bug:
api.kernel.thread.Calculation.CalcTest
api.kernel.thread.ExcptHandlerTest.ExcptHandlerTest
api.kernel.thread.LifeSimulationTest.Life
api.kernel.thread.ThreadLocalTest.ThreadLocalTest
api.nio.charset.CharsetSyncCacheTest
api.text.getSentenceInstance
api.zip.ZlibTest
I've checked tests with release revision 612461 on Win32:
- 3 tests passed:
api.kernel.thread.ThreadLocalTest.ThreadLocalTest
api.text.getSentenceInstance
api.zip.ZlibTest
- 2 tests failed with "java.lang.OutOfMemoryError: Failed to create new thread" (5 failures from 5 runs):
api.kernel.thread.Calculation.CalcTest
api.kernel.thread.LifeSimulationTest.Life
This failure is described in HARMONY-5334.
- 1 test failed as described in HARMONY-5400 (5 failures from 5 runs):
api.nio.charset.CharsetSyncCacheTest
- 1 test failed as described in HARMONY-5402 (3 failures from 5 runs):
api.kernel.thread.ExcptHandlerTest.ExcptHandlerTest
On debug version all tests passed except CharsetSyncCacheTest.
Summarize investigation I can say this bug is out-of-date. Some tests which was excluded due to this bug work fine for others corresponding JIRAs were created.
I've suggested to close the bug and change exclude list for reliability suite.
> [drlvm][thread][reliability] assert in JIT while unwinding stack (during enumeration)
> -------------------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Assignee: weldon washburn
> Attachments: 2rel_tests_fail.dmp, H4215_tm_races.patch, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Pavel Ozhdikhin (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12506801 ]
Pavel Ozhdikhin commented on HARMONY-4215:
------------------------------------------
One more stack trace from the failure of ZlibTest:
harmonyvm.dll!_NMSG_WRITE(int rterrnum=10) Line 195 C
harmonyvm.dll!abort() Line 44 + 0x7 C
> harmonyvm.dll!ReleasePrimitiveArrayCritical(JNIEnv_External * jni_env=0x02d42420, _jobject * array=0x04e2f624, void * carray=0x02fc6458, int mode=0) Line 1315 C++
hyarchive.dll!Java_java_util_zip_Deflater_deflateImpl(const JNINativeInterface_ * * env=0x02d42420, _jobject * recv=0x04e2f620, _jobject * buf=0x04e2f624, int off=0, int len=1024, __int64 handle=65024096, int flushParm=4) Line 209 C
033f522a()
harmonyvm.dll!get_vm_thread(HyThread * thr=0x04e2f650) Line 193 + 0xb C++
harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000000) Line 194 + 0x7 C++
The interesting thing that in ReleasePrimitiveArrayCritical at the point of failure array_clss->get_name() contains the string "java/util/zip/Inflater". That means array_clss was changed between these 2 lines in the same method:
...
assert(array_clss->get_name()->bytes[0]=='[');
switch (array_clss->get_name()->bytes[1]) {
...
It looks like the problem is that gc_gen does not implement object pinning (gc_pin_object()) which is used in GetPrimitiveArrayCritical/ReleasePrimitiveArrayCritical pair in the method Java_java_util_zip_Deflater_deflateImpl().
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Assigned: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "weldon washburn (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
weldon washburn reassigned HARMONY-4215:
----------------------------------------
Assignee: weldon washburn
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Assignee: weldon washburn
> Attachments: 2rel_tests_fail.dmp, H4215_tm_races.patch, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Gregory Shimansky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Gregory Shimansky updated HARMONY-4215:
---------------------------------------
Summary: [drlvm][reliability] assert in JIT while unwinding stack (during enumeration) (was: [drlvm][jit] assert in JIT while unwinding stack (during enumeration))
Looking at the above evaluation I think the bug should not be assigned to JIT category. The tests crash in many places and require additional analyzes to find out what component is to be blamed.
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][reliability] assert in JIT
while unwinding stack (during enumeration)
Posted by "Gregory Shimansky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12517190 ]
Gregory Shimansky commented on HARMONY-4215:
--------------------------------------------
I the few last days when I ran this test in the loop I have seen only the following types of failures. All the times there is a crash on an assertion check in GC (note obj pointer is bad):
> gc_gen.dll!get_obj_info_raw(Partial_Reveal_Object * obj=0xdeadbeef) Line 164 + 0x23 C++
gc_gen.dll!collector_forward_object(Collector * collector=0x016d2ca0, Partial_Reveal_Object * p_obj=0xdeadbeef) Line 35 + 0x9 C++
gc_gen.dll!forward_object(Collector * collector=0x016d2ca0, Partial_Reveal_Object * * p_ref=0x0501f6f8) Line 96 + 0xd C++
gc_gen.dll!trace_object(Collector * collector=0x016d2ca0, Partial_Reveal_Object * * p_ref=0x0501f6f8) Line 125 + 0xd C++
gc_gen.dll!collector_trace_rootsets(Collector * collector=0x016d2ca0) Line 180 + 0xd C++
gc_gen.dll!nongen_forward_pool(Collector * collector=0x016d2ca0) Line 222 + 0x9 C++
gc_gen.dll!collector_thread_func(void * arg=0x016d2ca0) Line 186 + 0x9 C++
hythr.dll!thread_start_proc(void * arg=0x00382f38) Line 710 + 0x9 C
hythr.dll!_threadstartex(void * ptd=0x004267b8) Line 241 + 0xd C
kernel32.dll!77e66063()
This happens about once in 100 runs on win32 for me. I haven't seen any other types of crashes mentioned in this bug report.
> [drlvm][reliability] assert in JIT while unwinding stack (during enumeration)
> -----------------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp, test.zip, util.Date_int.out.err
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4215) [drlvm][jit] assert in JIT while
unwinding stack (during enumeration)
Posted by "Pavel Ozhdikhin (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12506477 ]
Pavel Ozhdikhin commented on HARMONY-4215:
------------------------------------------
I've reproduced the failure running ZlibTest. The analysis in debugger showed that the JitFrameContext passed to JIT from VM contains eip=0, but JIT can not unwind stack frame with zero eip.
I ran the test with si trace enabled I've got the access violation failure in VM with the following stack trace:
> harmonyvm.dll!oh_enumerate_handles(ObjectHandles * hs=0x10c48333) Line 312 + 0x15 C++
harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02e80fd8) Line 330 + 0x1b C++
harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02e50b80) Line 224 + 0x9 C++
harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
gc_gen.dll!gc_reclaim_heap(GC * gc=0x01554d98, unsigned int gc_cause=3) Line 295 + 0x8 C++
gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02e25d30, JNIEnv_External * __formal=0x02e25d30) Line 138 + 0x8 C++
036dafcf()
harmonyvm.dll!get_vm_thread(HyThread * thr=0x03cef9ec) Line 193 + 0xb C++
harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> [drlvm][jit] assert in JIT while unwinding stack (during enumeration)
> ---------------------------------------------------------------------
>
> Key: HARMONY-4215
> URL: https://issues.apache.org/jira/browse/HARMONY-4215
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: Win 2003
> Reporter: Aleksey Ignatenko
> Attachments: 2rel_tests_fail.dmp
>
>
> 2 reliablity tests failed on assertion in jit:
> 1. CharsetSyncCacheTest
> 2. ZlibTest
> svn rev 548320
> Call stack:
> jitrino.dll!_assert(const char * expr=0x01ba9f08, const char * filename=0x01ba9eb0, unsigned int lineno=0x000000ac) Line 295 C
> > jitrino.dll!Jitrino::Ia32::StackInfo::read(Jitrino::MethodDesc * pMethodDesc=0x03bef720, unsigned int eip=0x00000000, bool isFirst=false) Line 172 + 0x1a C++
> jitrino.dll!Jitrino::Ia32::RuntimeInterface::unwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 40 C++
> jitrino.dll!Jitrino::Jitrino::UnwindStack(Jitrino::MethodDesc * methodDesc=0x03bef720, JitFrameContext * context=0x02de90d4, bool isFirst=false) Line 280 + 0x1e C++
> jitrino.dll!JIT_unwind_stack_frame(void * jit=0x012a6a48, Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 362 + 0x18 C++
> harmonyvm.dll!Dll_JIT::unwind_stack_frame(Method * method=0x02cb3668, JitFrameContext * context=0x02de90d4) Line 94 + 0x14 C++
> harmonyvm.dll!si_goto_previous(StackIterator * si=0x02de90d0, bool over_popped=false) Line 315 + 0x32 C++
> harmonyvm.dll!vm_enumerate_root_set_single_thread_on_stack(StackIterator * si=0x02de90d0) Line 339 + 0xb C++
> harmonyvm.dll!vm_enumerate_thread(VM_thread * thread=0x02d7e088) Line 224 + 0x9 C++
> harmonyvm.dll!stop_the_world_root_set_enumeration() Line 110 + 0xc C++
> harmonyvm.dll!vm_enumerate_root_set_all_threads() Line 150 C++
> gc_gen.dll!gc_reclaim_heap(GC * gc=0x014d5de8, unsigned int gc_cause=0x00000003) Line 295 + 0x8 C++
> gc_gen.dll!gc_force_gc() Line 138 + 0xd C++
> harmonyvm.dll!Java_java_lang_VMMemoryManager_runGC(JNIEnv_External * __formal=0x02d9e240, JNIEnv_External * __formal=0x02d9e240) Line 138 + 0x8 C++
> 03bf33ef()
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x03bef9ec) Line 193 + 0xb C++
> harmonyvm.dll!get_vm_thread(HyThread * thr=0x00000001) Line 194 + 0x7 C++
> To reproduce onee needs to run these tests with 30 min cycle (set <property name="test.duration" value="1800" /> parameter in build.xml of reliablity tests).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.