Uploaded image for project: 'JDK'
  1. JDK
  2. JDK-4504213

Solaris jdk 1.3.1_01 hotspot core dumps when loading pre compiled JSPs

    Details

    • Subcomponent:
    • CPU:
      sparc
    • OS:
      solaris_8

      Description


      Solaris jdk 1.3.1_01 HotSpot (java_g) core dumps when loading pre compiled JSPs,
      with following error:
      #
      # HotSpot Virtual Machine Error, assertion failure
      # Please report this error at
      # http://java.sun.com/cgi-bin/bugreport.cgi
      #
      # assert(is_within_bounds(which), "index out of bounds")
      #
      # Error ID: /export/d0/jdk131-update/hotspot/ws/solsparc/hotspot1.3.1fcs/src/share
      /vm/oops/typeArrayOop.hpp, 37
       [ Patched ]
      #
      # Problematic Thread: prio=5 tid=0x332f10 nid=0x13 runnable
      #
      Dumping core....
      Abort - core dumped


      Stack trace (only partial listed in here) is as follows:
      current thread: t@17
      =>[1] __sigprocmask(0x0, 0xd627fc38, 0x0, 0x0, 0x0, 0x0), at 0xff369ab8
        [2] _resetsig(0xff36c408, 0x0, 0x0, 0xd6281d78, 0xff37e000, 0x0), at 0xff35e4fc
        [3] _sigon(0xd6281d78, 0xff3859a8, 0x6, 0xd627fd0c, 0xd6281d78, 0xd627fd50), at 0xff35dc9c
        [4] _thrp_kill(0x0, 0x11, 0x6, 0xff37e000, 0x11, 0xff2bc480), at 0xff360cb0
        [5] raise(0x6, 0x0, 0x0, 0xffffffff, 0xff2bc3ec, 0xd627fe68), at 0xff24ae60
        [6] abort(0xff2b8000, 0xd627fe68, 0x0, 0xfffffff8, 0x4, 0xd627fe89), at 0xff23563c
        [7] os::abort(0x1, 0xfe518a28, 0xd628070c, 0x100, 0x0, 0x0), at 0xfe09d0e8
        [8] report_error(0x1, 0xfe561774, 0x25, 0xfe518878, 0xfe51888c, 0xfe56175c), at 0xfdd9f300
        [9] report_assertion_failure(0xfe56175c, 0xfe561774, 0x25, 0xfe5617d4, 0xd6280810, 0xd628080c), at 0xfdd9e6d0
        [10] typeArrayOopDesc::byte_at_addr(0xd9545cc8, 0x0, 0x8, 0x4, 0x1, 0xd6280a44), at 0xfdf07d14
        [11] jni_GetByteArrayElements(0x2e4bb0, 0xd6280b50, 0x0, 0xd62808ec, 0xd62808e8, 0xd62808e4), at 0xfdefaf58
        [12] JNIEnv_::GetByteArrayElements(0x2e4bb0, 0xd6280b50, 0x0, 0x34, 0x2e4b18, 0x0), at 0xd351d890
        [13] Java_weblogic_db_oci_OciColumn_setValueWithString(0x2e4bb0, 0xd6280b54, 0xd6280b50, 0x0, 0x0, 0x0), at 0xd351c25c
        [14] 0xced9c(0xd9545c58, 0xb6, 0xd6280bdc, 0xe3f3c, 0x0, 0x0), at 0xced9b
      <..snip..>

      HotSpot (java_g) crashes during load tests with just 10-15 users. Running a test suite
      which invokes pre-compiled JSPs in a Weblogic 6.1 environment crashes HotSpot, as the
      number of JSPs invoked increases (not a whole lot of JSPs, jvm size is about 400+ MB).
      The same problem is observed irrespective of using the "client" or "server" hotspot
      versions. This same problem is observed with Solaris jdk 1.3.0 & Solaris jdk 1.3.1.

      Attached to this bug report are tar files. Each tar file has following info, but from
      different runs of java_g 1.3.1_01 (using different hotspot options):
      stack & threads output from core file using dbx
      exact error from hotspot
      weblogic log file (only for a few tests)

      These attached files contain consistently similar info, from running jdk 1.3.1_01
      java_g, with the following hotspot options:
      file test1_client_13101_javag_int.tar.Z
      run with java_g options: -client -ms256m -mx512m -Xint
      file test2_server_13101_javag_int.tar.Z
      run with java_g options: -server -ms256m -mx512m -Xint
      file test3_client_13101_javag_permsize.tar.Z
      run with java_g options: -client -ms256m -mx512m
      -XX:MaxPermSize=128m
      file test4_server_13101_javag_permsize.tar.Z
      run with java_g options: -server -ms256m -mx512m
      -XX:MaxPermSize=128m

      (non standard option -XX:MaxPermSize=128m is being used per info from BEA systems,
      documented at http://e-docs.bea.com/wls/platforms/sun/index.html)

      System Configuration:
      Sol 8, HW 220R (currently using 1 out of 2 CPUs), RAM 512 MB
      3 220R's - 1 for BEA WL app server, 1 for webserver, 1 for Oracle 8.1.7
      Weblogic 6.1 (moved from WL 6.0), using type2 native weblogic db driver
      jdk 1.3.1_01 (moved from jdk 1.3.0 & jdk 1.3.1)
      jvm options: -ms256m -mx512m
      (also used -Xint or -XX:MaxPermSize=128m in a few tests)

      Interestingly enough, customer does not see this problem of HotSpot crash on a 2 CPU
      system. Instead, the jvm gets larger and larger (and thus slower and slower) to
      a point where it is unuseable (cannot compile new JSPs), and throws an
      'Not enough space' error:
      Compilation of
      '/opt/home/bea/wlserver6.0sp1/config/netilla/applications
      /.wl_temp_do_not_delete/WEB-INF/
      _tmp_war_engapp01_engapp01_netapp/jsp_servlet/_jsp/
      _serviceMgmt/_selectProfile.java'
              failed:
               java.io.IOException: Not enough space
               at java.lang.UNIXProcess.forkAndExec(Native Method)
               at java.lang.UNIXProcess.<init>(UNIXProcess.java:54)
               at java.lang.Runtime.execInternal(Native Method)
               at java.lang.Runtime.exec(Runtime.java:551)
               at java.lang.Runtime.exec(Runtime.java:477)
               at java.lang.Runtime.exec(Runtime.java:443)
               at weblogic.utils.Executable.exec(Executable.java:144)
      <..snip..>

        Attachments

          Issue Links

            Activity

              People

              • Assignee:
                xjiangsunw Xuyang Jiang (Inactive)
                Reporter:
                duke J. Duke (Inactive)
              • Votes:
                0 Vote for this issue
                Watchers:
                0 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved:
                  Imported:
                  Indexed: