JDK-6487381 : Additional path for 5.0 jvm crash on exhaustion of CodeBuffer
  • Type: Bug
  • Component: hotspot
  • Sub-Component: compiler
  • Affected Version: 5.0u6
  • Priority: P3
  • Status: Resolved
  • Resolution: Fixed
  • OS: solaris_8
  • CPU: generic
  • Submitted: 2006-10-27
  • Updated: 2010-04-03
  • Resolved: 2007-01-25
The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.
Other
1.4.2_15Fixed
Related Reports
Relates :  
Relates :  
Relates :  
Relates :  
Description
Solaris 8 java 1.5.0_06 with options:
-Xmn1152m -Xms2944m -Xmx2944m -XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:MaxPermSize=384m

----------------- t@16 -----------------
0xff3691f0      __sigprocmask + 0x8
0xff35dd0c      _sigon + 0xd0
0xff360d4c      _thrp_kill + 0xf8
0xff24bce0      raise + 0x40
0xff235984      abort + 0x100
0xfe264724      void os::abort(int) + 0x58
0xfe2f08f0      void VMError::report_and_die() + 0xc84
0xfe035b24      void report_fatal_vararg(const char*,int,const char*,...) + 0x44
0xfddb4b70      CodeBuffer::CodeBuffer #Nvariant 1(int,int,int,int,int,int,BufferBlob*,relocInfo*,RelocateBuffer*,int,OopRecorder*,const char*,int,int) + 0xbc
0xfde1b140      void CodeBuffer::resize(int,int,int,int) + 0x94
0xfde26de8      void Compile::Fill_buffer() + 0xc64
0xfde24574      void Compile::Output() + 0x9f0
0xfde1a408      void Compile::Code_Gen() + 0x384
0xfde8276c      Compile::Compile(ciEnv*,C2Compiler*,ciMethod*,int,int) + 0xc14
0xfde786f8      void C2Compiler::compile_method(ciEnv*,ciMethod*,int) + 0xb0
0xfde793b4      void CompileBroker::invoke_compiler_on_method(CompileTask*) + 0x4cc
0xfdf35f20      void CompileBroker::compiler_thread_loop() + 0x44c
0xfdede29c      void JavaThread::run() + 0x1d8
0xfe264240      void*_start(void*) + 0x208
0xff36b01c      _thread_start + 0x40

(dbx) frame 9
0xfe035b24: report_fatal_vararg+0x0044: call     report_and_die ! 0xfe2efc6c
(dbx) x 0xfe336047/s
0xfe336047: __RTTI__1nNExceptionBlob_+0x03a7:   "/BUILD_AREA/jdk1.5.0_06/hotspot/src/share/vm/asm/codeBuffer.cpp"
(dbx) print 0x60
96
(dbx) x 0xfe336087/s
0xfe336087: __RTTI__1nNExceptionBlob_+0x03e7:   "CodeCache: no room for %s"
(dbx) x  0xfe336030/s
0xfe336030: __RTTI__1nNExceptionBlob_+0x0390:   "CodeBuffer constructor"

Same as
http://cheesypoof.uk:8080/Saucer/ErrorID?id=434F44452255464645520E4350500060
Error ID 434F44452255464645520E4350500060 = codeBuffer.cpp, line 96

which is http://cheesypoof.uk.sun.com/lxr2/source/ws/hotspot/src/share/vm/asm/codeBuffer.cpp?v=Java_1.5.0_06#096

I can also see that
Thread t@52: is deep and all compiled
Thread t@24: is deep and all compiled
Thread t@21: is deep and all compiled

Heap
par new generation   total 1178560K, used 1065004K [0x28000000, 0x70000000, 0x70000000)
 eden space 1177472K,  90% used [0x28000000, 0x6900b128, 0x6fde0000)
 from space 1088K,   0% used [0x6fde0000, 0x6fde0000, 0x6fef0000)
 to   space 1088K,   0% used [0x6fef0000, 0x6fef0000, 0x70000000)
concurrent mark-sweep generation total 1835008K, used 546506K [0x70000000, 0xe0000000, 0xe0000000)
concurrent-mark-sweep perm gen total 126464K, used 126206K [0xe0000000, 0xe7b80000, 0xf8000000)

According to 4925292 and looking at perm space I would suggest in the meantime to use:

-XX:CodeCacheMinimumFreeSpace=2M -XX:+ReservedCodeCacheSize=64M    (defaults are free 512k and size 32m)
-XX:PermSize=128m -XX:MaxPermSize=384m
-XX:+CMSPermGenSweepingEnabled

but this is 1.5.0_06 so it shouldn't happen any more a crash for a small CodeCache ? Is it a regression of 4925292 ?
(Obviously it should be better to use /usr/lib/lwp/libthread.so.1 always )

More info in /net/cores.central/cores/10916493/sunup10916493/coreinfo_83a38a37_20060925

Comments
EVALUATION VM Crash due to exhaustion of CodeBuffer space, need to provide a softer landing. See bug 4925292 for mechanism.
27-10-2006