You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@harmony.apache.org by "Gregory Shimansky (JIRA)" <ji...@apache.org> on 2007/01/10 14:59:27 UTC
[jira] Created: (HARMONY-2963) [drlvm][stress] MegaSpawn test from
HARMONY-2803 hangs in the end on dead lock
[drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
------------------------------------------------------------------------------
Key: HARMONY-2963
URL: https://issues.apache.org/jira/browse/HARMONY-2963
Project: Harmony
Issue Type: Sub-task
Components: DRLVM
Reporter: Gregory Shimansky
When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Updated: (HARMONY-2963) [drlvm][stress] MegaSpawn test from
HARMONY-2803 hangs in the end on dead lock
Posted by "Peter Novodvorsky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Peter Novodvorsky updated HARMONY-2963:
---------------------------------------
Attachment: h2963.diff
This patch includes Salikh's changes and also changed the way vm_attach works with errors. Now it asserts asserts instead of returning errors in vm_attach function. Returning with error in vm_attach is nonsense, because parent process won't understand that and will hang, because it won't be notified that child vm is started.
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: h2963.diff, moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-2963) [drlvm][stress] MegaSpawn test from
HARMONY-2803 hangs in the end on dead lock
Posted by "Peter Novodvorsky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Peter Novodvorsky updated HARMONY-2963:
---------------------------------------
Attachment: move_memory_allocation_to_parent_thread_1.1.patch
updated version of the patch the only relevant patch is the latest one, with 1.1 as a version it includes all previous and fixes for windows build.
I think it's ready for commit.
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: h2963.diff, move_memory_allocation_to_parent_thread.patch, move_memory_allocation_to_parent_thread_1.1.patch, moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-2963) [drlvm][stress] MegaSpawn test
from HARMONY-2803 hangs in the end on dead lock
Posted by "Salikh Zakirov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12463910 ]
Salikh Zakirov commented on HARMONY-2963:
-----------------------------------------
moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
The attached patch provides a fix for incorrect non-daemon thread count modifications
by moving the increment to the place where we are sure the thread is already started (and also matches counter decrement better, because
counter is decremented in jthread_detach() on the target thread).
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Attachments: moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Commented: (HARMONY-2963) [drlvm][stress] MegaSpawn test
from HARMONY-2803 hangs in the end on dead lock
Posted by "Salikh Zakirov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12471358 ]
Salikh Zakirov commented on HARMONY-2963:
-----------------------------------------
Interestingly, moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch reverts the HARMONY-1816 modification.
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-2963) [drlvm][stress] MegaSpawn test from
HARMONY-2803 hangs in the end on dead lock
Posted by "Salikh Zakirov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Salikh Zakirov updated HARMONY-2963:
------------------------------------
Attachment: moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Attachments: moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Commented: (HARMONY-2963) [drlvm][stress] MegaSpawn test
from HARMONY-2803 hangs in the end on dead lock
Posted by "Peter Novodvorsky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12473045 ]
Peter Novodvorsky commented on HARMONY-2963:
--------------------------------------------
this patch really fixes the issue. the reason it reverts HARMONY-1816 modification is that the only functional part of HARMONY-1816 is
- while (lib->nondaemon_thread_count) {
+ while ((!jvmti_thread->daemon && lib->nondaemon_thread_count > 1)
+ || (jvmti_thread->daemon && lib->nondaemon_thread_count > 0)) {
status = hycond_wait(lib->nondaemon_thread_cond, lib->TM_LOCK);
which is not reverted by Salikh's patch.
I've tried to run 1816 test with Salikh's patch and it works ok.
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-2963) [drlvm][stress] MegaSpawn test
from HARMONY-2803 hangs in the end on dead lock
Posted by "Gregory Shimansky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12463556 ]
Gregory Shimansky commented on HARMONY-2963:
--------------------------------------------
To run this test more reliably so that it doesn't crash VM but gets to the end and hang it is better to use single CPU computers. Otherwise different threads usually hit out of C heap condition and VM dies before test gets to shutdown stage.
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Updated: (HARMONY-2963) [drlvm][stress] MegaSpawn test from
HARMONY-2803 hangs in the end on dead lock
Posted by "Peter Novodvorsky (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Peter Novodvorsky updated HARMONY-2963:
---------------------------------------
Attachment: move_memory_allocation_to_parent_thread.patch
This patch moves all memory allocation for initial thread structures from child thread to parent. It also removes assertions.
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: h2963.diff, move_memory_allocation_to_parent_thread.patch, moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-2963) [drlvm][stress] MegaSpawn test
from HARMONY-2803 hangs in the end on dead lock
Posted by "Salikh Zakirov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12463883 ]
Salikh Zakirov commented on HARMONY-2963:
-----------------------------------------
Reproduced the hang of MegaSpawn on single processor Windows XP.
The two threads are deadlocked on a shutdown:
Thread 1: jthread_wait_for_all_nondaemon_threads();
Thread 2: wait on java/lang/FinalizerThread$FinalizerWorkLock
Thread library has lib->nondaemon_thread_count = 230, meaning that
threads terminated by OOM did not correctly decreased non-daemon thread count.
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Closed: (HARMONY-2963) [drlvm][stress] MegaSpawn test from
HARMONY-2803 hangs in the end on dead lock
Posted by "weldon washburn (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
weldon washburn closed HARMONY-2963.
------------------------------------
Resolution: Fixed
build, build test pass on winxp32 and linux32 rhel 4.0 w/ gcc 4.0.2
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: h2963.diff, move_memory_allocation_to_parent_thread.patch, move_memory_allocation_to_parent_thread_1.1.patch, moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-2963) [drlvm][stress] MegaSpawn test
from HARMONY-2803 hangs in the end on dead lock
Posted by "Naveen Neelakantam (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12475550 ]
Naveen Neelakantam commented on HARMONY-2963:
---------------------------------------------
not that any additional confirmation is necessary but, MegaSpawn no longer hangs for me on rhel4. I ran it 100 times successfully in both debug and release modes.
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: h2963.diff, move_memory_allocation_to_parent_thread.patch, move_memory_allocation_to_parent_thread_1.1.patch, moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-2963) [drlvm][stress] MegaSpawn test
from HARMONY-2803 hangs in the end on dead lock
Posted by "Salikh Zakirov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12463911 ]
Salikh Zakirov commented on HARMONY-2963:
-----------------------------------------
moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch passed tests on WinXP (uniprocessor) and Linux SUSE9 (ht)
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Attachments: moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Assigned: (HARMONY-2963) [drlvm][stress] MegaSpawn test from
HARMONY-2803 hangs in the end on dead lock
Posted by "weldon washburn (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
weldon washburn reassigned HARMONY-2963:
----------------------------------------
Assignee: weldon washburn
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Commented: (HARMONY-2963) [drlvm][stress] MegaSpawn test
from HARMONY-2803 hangs in the end on dead lock
Posted by "weldon washburn (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-2963?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12474707 ]
weldon washburn commented on HARMONY-2963:
------------------------------------------
which patches are relevant? All three??
> [drlvm][stress] MegaSpawn test from HARMONY-2803 hangs in the end on dead lock
> ------------------------------------------------------------------------------
>
> Key: HARMONY-2963
> URL: https://issues.apache.org/jira/browse/HARMONY-2963
> Project: Harmony
> Issue Type: Sub-task
> Components: DRLVM
> Reporter: Gregory Shimansky
> Assigned To: weldon washburn
> Attachments: h2963.diff, move_memory_allocation_to_parent_thread.patch, moved-non-daemon-thread-count-modification-from-start-to-wrapper_proc.patch
>
>
> When running MegaSpawn test from HARMONY-2803 on windows and linux, if this test doesn't kill VM, it hangs at the end in some dead lock of 2 or more threads. This bug doesn't seem to be related to the stressing nature of the test, so it should be something different. It is likely that this is the same IO dead lock which makes stress.Mix to hang when it is running from "build test".
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.