[KUDU-598] SIGSEGV in TSAN of cbtree-test.TestConcurrentInsert

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: M5
• Fix Version/s: 0.7.0
• Component/s: tablet
• Target Version/s: 0.7.0
• Code Review: http://gerrit.cloudera.org:8080/1859
• Assignee: Todd Lipcon
• Reporter: Adar Dembo
• Created: 20/Jan/15 16:46
• Updated: 21/Jan/16 20:42
• Resolved: 21/Jan/16 20:42

Description

Happened in http://sandbox.jenkins.cloudera.com/job/kudu-gerrit/7089/BUILD_TYPE=TSAN,label=kudu-gerrit-slaves.

[ RUN      ] TestCBTree.TestConcurrentInsert
*** Aborted at 1421790484 (unix time) try "date -d @1421790484" if you are using GNU date ***
PC: @     0x7f1b8a08dd48 __tsan_read8
*** SIGSEGV (@0x0) received by PID 23490 (TID 0x7f1b82b01cc0) from PID 0; stack trace: ***
    @     0x7f1b8a04a89a rtl_sigaction() at ??:0
    @     0x7f1b8763a500 (unknown) at ??:0
    @     0x7f1b8a08dd48 __tsan_read8 at ??:0
    @     0x7f1b8a0c090a kudu::tablet::btree::ValueSlice::as_slice() at ??:0
    @     0x7f1b8a0c08aa kudu::tablet::btree::LeafNode<>::Get() at ??:0
    @     0x7f1b8a0ae24a kudu::tablet::btree::CBTree<>::GetCopy() at ??:0
    @     0x7f1b8a0ad9f3 kudu::tablet::btree::VerifyGet<>() at ??:0
    @     0x7f1b8a0c3276 kudu::tablet::btree::VerifyRange<>() at ??:0
    @     0x7f1b8a0aee74 kudu::tablet::btree::InsertAndVerify<>() at ??:0
    @     0x7f1b8a0c2c6d boost::_bi::list5<>::operator()<>() at ??:0
    @     0x7f1b8a0c2b84 boost::_bi::bind_t<>::operator()() at ??:0
    @     0x7f1b8a0c2b51 boost::detail::thread_data<>::run() at ??:0
    @     0x7f1b88833d47 (unknown) at ??:0
    @     0x7f1b8a040225 __tsan::cur_thread_placeholder at ??:0
    @     0x7f1b87632851 (unknown) at ??:0
    @     0x7f1b869dc94d (unknown) at ??:0
    @                0x0 (unknown)

It follows with a bunch of TSAN warnings about signal-unsafe calls inside a signal, pointing at kudu::tablet::btree::LeafNode<kudu::tablet::btree::SmallFanoutTraits>::Get(unsigned long, kudu::Slice*, kudu::Slice*) const /data1/jenkins-workspace/kudu-gerrit/BUILD_TYPE/TSAN/label/kudu-gerrit-slaves/src/kudu/tablet/concurrent_btree.h:731 (cbtree-test+0x0000000d38a9) as the line that triggered the SIGSEGV.

Seems related to KUDU-381 or KUDU-400, but not exactly the same.