Uploaded image for project: 'JDK'
  1. JDK
  2. JDK-8223961

ZGC: Unexpected behaviour due to ZMetronome::wait_for_tick() oversleeping

    XMLWordPrintable

    Details

    • Type: Bug
    • Status: Resolved
    • Priority: P4
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 13
    • Component/s: hotspot
    • Labels:
    • Subcomponent:
      gc
    • Resolved In Build:
      b22
    • CPU:
      generic
    • OS:
      generic

      Description

      ## Symptom

      If the zgc threads are blocked (or unable to be scheduled) for some time, the update of "_nticks" in ZMetronome::wait_for_tick() may become incorrect.

      For example, zgc threads can be blocked at a breakpoint when debugging with GDB.
      A lot of unexpected gc+stats logs might be observed due to the incorrect "_nticks".

      ## Reproduce

      - Running script
      ---------------------------------------------------------
      gdb --args \
      ${JDK}/bin/java \
        -Xlog:gc* \
        -XX:+UnlockExperimentalVMOptions \
        -XX:-UseG1GC -XX:+UseZGC \
        SomeJavaAppWithGCs
      ---------------------------------------------------------

      - Set a GDB breakpoint
      ---------------------------------------------------------
      (gdb) b ZDriver::concurrent_mark()
      Function "ZDriver::concurrent_mark()" not defined.
      Make breakpoint pending on future shared library load? (y or [n]) y
      Breakpoint 1 (ZDriver::concurrent_mark()) pending.
      (gdb) handle SIGSEGV nostop noprint
      Signal Stop Print Pass to program Description
      SIGSEGV No No Yes Segmentation fault
      (gdb) r
      ---------------------------------------------------------

      - Do some check at the breakpoint (or just wait for some time), and then continue
      ---------------------------------------------------------
      Breakpoint 1, ZDriver::concurrent_mark (this=0x7ffff004f3c0) at /home/fool/Workspace/jdk/src/hotspot/share/gc/z/zDriver.cpp:278
      278 ZStatTimer timer(ZPhaseConcurrentMark);
      (gdb)
      (gdb) do some check here, or just wait for some time (e.g., 60s)
          ... after some time ...
      (gdb) c
          ... a lot of unexpected gc+stats logs were printed ...
      ---------------------------------------------------------

      ## Fix
      ---------------------------------------------------------
      diff -r b0b20413d853 src/hotspot/share/gc/z/zMetronome.cpp
      --- a/src/hotspot/share/gc/z/zMetronome.cpp Wed May 15 11:53:47 2019 +0200
      +++ b/src/hotspot/share/gc/z/zMetronome.cpp Wed May 15 20:15:39 2019 +0800
      @@ -57,6 +57,8 @@
           if (!_stopped && timeout_ms > 0) {
             // Wait
             ml.wait(timeout_ms);
      + } else if (timeout_ms < -1 * (int64_t)_interval_ms) {
      + _nticks += -timeout_ms / _interval_ms;
           } else {
             // Tick
             return !_stopped;
      ---------------------------------------------------------

        Attachments

          Activity

            People

            Assignee:
            jiefu Jie Fu
            Reporter:
            jiefu Jie Fu
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

              Dates

              Created:
              Updated:
              Resolved: