From owner-freebsd-java@FreeBSD.ORG Mon Feb 7 23:07:33 2011 Return-Path: Delivered-To: freebsd-java@freebsd.org Received: from [127.0.0.1] (freefall.freebsd.org [IPv6:2001:4f8:fff6::28]) by hub.freebsd.org (Postfix) with ESMTP id 77782106564A; Mon, 7 Feb 2011 23:07:33 +0000 (UTC) (envelope-from jkim@FreeBSD.org) From: Jung-uk Kim To: freebsd-java@freebsd.org Date: Mon, 7 Feb 2011 18:07:18 -0500 User-Agent: KMail/1.6.2 References: <201101261721.58069.jkim@FreeBSD.org> <201102032111.13479.jkim@FreeBSD.org> <201102041932.43679.jkim@FreeBSD.org> In-Reply-To: <201102041932.43679.jkim@FreeBSD.org> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <201102071807.21551.jkim@FreeBSD.org> Cc: Pieter de Goeje , Mike Jakubik Subject: Re: [CFT] Update OpenJDK6 to b21 X-BeenThere: freebsd-java@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Porting Java to FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 07 Feb 2011 23:07:33 -0000 On Friday 04 February 2011 07:32 pm, Jung-uk Kim wrote: > On Thursday 03 February 2011 09:11 pm, Jung-uk Kim wrote: > > On Thursday 03 February 2011 04:24 pm, Jung-uk Kim wrote: > > > On Thursday 03 February 2011 01:09 pm, Mike Jakubik wrote: > > > > On Wed, 2011-02-02 at 13:54 -0500, Jung-uk Kim wrote: > > > > > Yeah, I actually fixed that. Some patches were missing in > > > > > b20 ports. 'jstack -m' should work, too. :-) > > > > > > > > > > Thanks for testing! > > > > > > > > Compiled OK and working with my applications OK. The jstack > > > > -m option you mention however produces the following results > > > > here: > > > > > > > > Attaching to process ID 84046, please wait... > > > > Debugger attached successfully. > > > > Server compiler detected. > > > > JVM version is 19.0-b09 > > > > sun.jvm.hotspot.debugger.DebuggerException: > > > > sun.jvm.hotspot.debugger.DebuggerException: get_thread_regs > > > > failed for a lwp > > > > at sun.jvm.hotspot.debugger.bsd.BsdDebuggerLocal > > > > $BsdDebuggerLocalWorkerThread.execute(BsdDebuggerLocal.java:1 > > > >52 ) at > > > > sun.jvm.hotspot.debugger.bsd.BsdDebuggerLocal.getThreadIntege > > > >rR eg is terSet(BsdDebuggerLocal.java:466) at > > > > sun.jvm.hotspot.debugger.bsd.BsdThread.getContext(BsdThread.j > > > >av a: 65 ) at > > > > sun.jvm.hotspot.runtime.bsd_amd64.BsdAMD64JavaThreadPDAccess. > > > >ge tC ur rentFrameGuess(BsdAMD64JavaThreadPDAccess.java:92) at > > > > sun.jvm.hotspot.runtime.JavaThread.getCurrentFrameGuess(JavaT > > > >hr ea d. java:256) at > > > > sun.jvm.hotspot.runtime.JavaThread.getLastJavaVFrameDbg(JavaT > > > >hr ea d. java:218) at > > > > sun.jvm.hotspot.tools.PStack.initJFrameCache(PStack.java:208) > > > > at sun.jvm.hotspot.tools.PStack.run(PStack.java:66) > > > > at sun.jvm.hotspot.tools.PStack.run(PStack.java:53) > > > > at sun.jvm.hotspot.tools.PStack.run(PStack.java:48) > > > > at sun.jvm.hotspot.tools.JStack.run(JStack.java:60) > > > > at sun.jvm.hotspot.tools.Tool.start(Tool.java:221) > > > > at sun.jvm.hotspot.tools.JStack.main(JStack.java:86) > > > > at sun.reflect.NativeMethodAccessorImpl.invoke0(Native > > > > Method) at > > > > sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAcces > > > >so rI mp l.java:57) at > > > > sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMet > > > >ho dA cc essorImpl.java:43) at > > > > java.lang.reflect.Method.invoke(Method.java:616) > > > > at sun.tools.jstack.JStack.runJStackTool(JStack.java:136) > > > > at sun.tools.jstack.JStack.main(JStack.java:102) > > > > > > Yeah, I noticed that, too. It seems it's getting bogus thread > > > IDs. It'll take some time for me to figure out why, though. :-( > > > > Now I know why but it is a very complicated problem. :-( > > > > Basically, the culprit was difference between Linux and BSD, > > mostly _thread_id and ptrace(2). Linux has gettid(2), which > > returns a (32-bit) pid_t type, and it is used for _thread_id. > > Linux code simply calls ptrace(PTRACE_GETREGS, ...) with it and > > it works fine. However, BSD does not have gettid(), so we used > > pthread_t instead. Alas, we copied process_get_lwp_regs() from > > Linux-specific file. It does not work for us because _thread_id > > is a pthread_t type and ptrace() does not understand it. > > Unfortunately, there is no easy way to map pthread_t to a lwpid_t > > value, either. More worse, it gets truncated to 32-bit on amd64. > > Grrr... > > To prove my idea, I just implemented libthr function > pthread_getthreadid_np(): > > http://docs.freebsd.org/cgi/mid.cgi?201102041409.12314.jkim > > Updated ports to support this new feature: > > http://people.freebsd.org/~jkim/ports-openjdk6-b21_3.tar.bz2 > > Since it is extremely experimental, it is turned off by default, of > course. If you dare to try it out, you must update libthr with the > above libthr patch and forcefully define > HAVE_PTHREAD_GETTHREADID_NP in the ports Makefile. Now jstack can > trace Java threads properly for me. :-) Please try this: http://people.freebsd.org/~jkim/ports-openjdk6-b21_4.tar.bz2 FYI, (simplified) libthr patch mentioned above was committed on HEAD. However, when pthread_getthreadid_np(3) is not available, now it uses (undocumented) thr_self() to get the same info. So, there is no need to patch libthr any more. :-) Thanks! Jung-uk Kim