harmony crash vmState=0x0005ffff #20546
Comments
|
@zl-wang fyi |
|
Issue Number: 20546 |
|
See also #20567 |
|
http://vmfarm.rtp.raleigh.ibm.com/job_output.php?id=95899244 |
|
Not sure if this is the same issue, but the vmstate and mode is the same. http://vmfarm.rtp.raleigh.ibm.com/job_output.php?id=95965401 100x grinder failed 5/100 |
pshipton
added this to the Release 0.49 (Java 8, 11, 17, 21, 23) January refresh milestone
|
@hzongaro pls take a look |
|
http://vmfarm.rtp.raleigh.ibm.com/job_output.php?id=95980123 http://vmfarm.rtp.raleigh.ibm.com/job_output.php?id=95999705 |
|
@pshipton, sorry for the basic question, but it's been a long time since I looked at a failure reported in vmfarm. Where should I be looking for the Linux x86 builds that you mentioned in #20546 (comment) and #20546 (comment)? |
|
The builds are retired and no longer on disk, you can find them in artifactory. |
|
Looking at the x86 Linux core from http://vmfarm.rtp.raleigh.ibm.com/job_output.php?id=95965401, the problem appears to be the same as that described by @IBMJimmyk in #20567 (comment) |
|
AIX core files can be found under http://vmfarm.rtp.raleigh.ibm.com/etc/cores/tmp/ |
|
@pshipton, do you know where I might find the debug image for that build? |
|
We have https://na.artifactory.swg-devops.com/ui/native/sys-rt-vmfarm-generic-local/R29/acceptance/jvmxa6480/81617/jvmxa6480.zip which appear to contain debug. |
|
Though the root cause of the problem is the same as that reported in #20567, the location where it occurs is different. That debug build helped move me further in investigating the problem. I was unable to tie addresses reported in the core file directly to line numbers in the source code, but I managed to piece together where exactly the problem occurred by looking for method calls that weren't virtual. It looks like the problem in this case occurs at the call to |
|
@IBMJimmyk @zl-wang Can we verify this one with the fix for #20567 ? |
|
Based on Henry's comments, my fix for #20567 would not help. My understanding is This is similar to the problem I saw but is taking a different path. I think the solution might also be to add error checking for invalid startPCs and take appropriate action. I would need to look at the code in more detail to determine what the best way to handle an error would be. It might be possible for |
|
http://vmfarm.rtp.raleigh.ibm.com/job_output.php?id=96726917 |
|
@IBMJimmyk, I know you were thinking about a fix for this. Do you think it can make it into 0.49 or should this move out to 0.51? |
|
We need a good justification to move a regression/blocker out. This failure has been occurring frequently and also occurred in a 25_01 build. |
|
I just talked to @dsouzai and it looks like this issue might be fixed by this recently merged PR: It reverts a problematic PR that was messing with the extra field and was originally merged in on Nov 7. I think this is just before these Harmony -3 start PC problems showed up (This issue was opened on Nov 8). The problematic PR was causing a bad start PC to cause problems in other places as well. I am currently in the middle of trying to see if I can verify that reverting the change will fix the problem. |
|
@IBMJimmyk do you have any conclusions? Something seems to have fixed the problem as I haven't been seeing it in the head stream builds any more. We should get the fix backported asap. |
|
I opened #20825; the original issue I had fixed wasn't tagged for any release, so I never double delivered. |
|
It does seem like the (now reverted) change that modified the extra field can potentially cause the problem in this issue. So the PR that reverts it should fix this issue as well. |
|
Issue Number: 20546 |
Internal build
[Linux PPC 64bit] 80 Load_Level_2.harmony.5mins.Mode112
-Xgcpolicy:gencon -Xjit:count=0 -Xnocompressedrefsrhel7p8vm14
vmState [0x5ffff]: {J9VMSTATE_JIT} {Illegal optimization number}
30x grinder failed 1/30
The text was updated successfully, but these errors were encountered: