Adar Dembo has posted comments on this change. ( http://gerrit.cloudera.org:8080/9835 )
Change subject: KUDU-2384. Don't collect thread stacks when running under debugger ...................................................................... Patch Set 3: > > That said, the test in question (TestCreateWithFailedDirs) already resets > > env_inject_eio to 0 when it finishes; if it didn't, we'd also see a warning > > that we couldn't clean up the test's temp dir. But I guess here the issue > > isn't with env_inject_eio, but with env_inject_eio_globs? We don't > > typically reset that one. > > Yea, the issue is with env_inject_eio_globs which is a string and thus > particularly unsafe to read. The odd thing, though, is that we only read > env_inject_eio_globs when env_inject_eio is non-zero, but it appears we're > still reading it. So perhaps we aren't resetting env_inject_eio properly in > all cases or somesuch. I think it's a real data race: right before the main thread (T1) resets env_inject_eio, the kernel stack watchdog thread (T2) gets scheduled and reads the non-zero gflag value. Then T1 is rescheduled and resets to 0, but it's too late: when T2 is rescheduled, it will read env_inject_eio_globs. -- To view, visit http://gerrit.cloudera.org:8080/9835 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: Ib17ffb9514e9b440b8d97843374d2a60913a398b Gerrit-Change-Number: 9835 Gerrit-PatchSet: 3 Gerrit-Owner: Todd Lipcon <t...@apache.org> Gerrit-Reviewer: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Tidy Bot Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-Comment-Date: Fri, 30 Mar 2018 16:51:06 +0000 Gerrit-HasComments: No