jvm crash 的崩溃日志分析及注意点

生成

1. 生成error 文件的路径：你可以通过参数设置-XX:ErrorFile=/path/hs_error%p.log, 默认是在java运行的当前目录[default: ./hs_err_pid%p.log]

2. 参数-XX:OnError 可以在crash退出的时候执行命令，格式是-XX:OnError=“string”, <string> 可以是命令的集合，用分号做分隔符, 可以用"%p"来取到当前进程的ID.
例如：

// -XX:OnError="pmap %p" // show memory map

// -XX:OnError="gcore %p; dbx - %p" // dump core and launch debugger

在linux中系统会fork出一个子进程去执行shell的命令，因为是用fork可能会内存不够的情况，注意修改你的 /proc/sys/vm/overcommit_memory 参数,不清楚为什么这里不使用vfork

3. -XX:+ShowMessageBoxOnError 参数，当jvm crash的时候在linux里会启动gdb 去分析和调式，适合在测试环境中使用。

什么情况下不会生成error文件

linux 内核在发生OOM的时候会强制kill一些进程, 可以在/var/logs/messages中查找

Error crash 文件的几个重要部分

a. 错误信息概要

# A fatal error has been detected by the Java Runtime Environment: # # SIGSEGV (0xb) at pc=0x0000000000043566, pid=32046, tid=1121192256 # # JRE version: 6.0_17-b04 # Java VM: Java HotSpot(TM) 64-Bit Server VM (14.3-b01 mixed mode linux-amd64 ) # Problematic frame: # C 0x0000000000043566 # # If you would like to submit a bug report, please visit: # http://java.sun.com/webapps/bugreport/crash.jsp # The crash happened outside the Java Virtual Machine in native code. # See problematic frame for where to report the bug.
SIGSEGV 错误的信号类型

pc 就是IP/PC寄存器值也就是执行指令的代码地址

pid 就是进程id

 

  # Problematic frame:
 
 

  # V  [libjvm.so+0x593045]
 

就是导致问题的动态链接库函数的地址

pc 和 +0x593045 指的是同一个地址，只是一个是动态的偏移地址，一个是真实运行的虚拟地址

b.信号信息

Java中在linux 中注册的信号处理函数，中间有2个参数info, ucvoid

static void crash_handler(int sig, siginfo_t* info, void* ucVoid) { // unmask current signal sigset_t newset; sigemptyset(&newset); sigaddset(&newset, sig); sigprocmask(SIG_UNBLOCK, &newset, NULL); VMError err(NULL, sig, NULL, info, ucVoid); err.report_and_die(); }寄存器的信息就保存在 (ucontext_t*)usVoid中

siginfo:si_signo=SIGSEGV: si_errno=0, si_code=1 (SEGV_MAPERR), si_addr=0x0000000000043566信号的详细信息，和si_addr 出错误的内存，这里对应的就是信号siginfo_t的结构体里的si_addr,内核会把导致错误的内存地址保存在用户空间的信号结构体中siginfo_t，这样在进程在注册的信号处理函数中可以拿到导致错误的地址。

c.寄存器信息

Registers: RAX=0x00002aacb5ae5de2, RBX=0x00002aaaaf46aa48, RCX=0x0000000000000219, RDX=0x00002aaaaf46b920 RSP=0x0000000042d3f968, RBP=0x0000000042d3f9c8, RSI=0x0000000042d3f9e8, RDI=0x0000000045aef9b8 R8 =0x0000000000000f80, R9 =0x00002aaab3d30ce8, R10=0x00002aaaab138ea1, R11=0x00002b017ae65110 R12=0x0000000042d3f6f0, R13=0x00002aaaaf46aa48, R14=0x0000000042d3f9e8, R15=0x0000000045aef800 RIP=0x0000000000043566, EFL=0x0000000000010202, CSGSFS=0x0000000000000033, ERR=0x0000000000000014 TRAPNO=0x000000000000000e
寄存器的信息就保存在处理函数参数 (ucontext_t*)usVoid中

在X86架构下：

void os::print_context(outputStream *st, void *context) { if (context == NULL) return; ucontext_t *uc = (ucontext_t*)context; st->print_cr("Registers:"); #ifdef AMD64 st->print( "RAX=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RAX]); st->print(", RBX=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RBX]); st->print(", RCX=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RCX]); st->print(", RDX=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RDX]); st->cr(); st->print( "RSP=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RSP]); st->print(", RBP=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RBP]); st->print(", RSI=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RSI]); st->print(", RDI=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RDI]); st->cr(); st->print( "R8 =" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_R8]); st->print(", R9 =" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_R9]); st->print(", R10=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_R10]); st->print(", R11=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_R11]); st->cr(); st->print( "R12=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_R12]); st->print(", R13=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_R13]); st->print(", R14=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_R14]); st->print(", R15=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_R15]); st->cr(); st->print( "RIP=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_RIP]); st->print(", EFL=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_EFL]); st->print(", CSGSFS=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_CSGSFS]); st->print(", ERR=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_ERR]); st->cr(); st->print(" TRAPNO=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_TRAPNO]); #else st->print( "EAX=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_EAX]); st->print(", EBX=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_EBX]); st->print(", ECX=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_ECX]); st->print(", EDX=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_EDX]); st->cr(); st->print( "ESP=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_UESP]); st->print(", EBP=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_EBP]); st->print(", ESI=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_ESI]); st->print(", EDI=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_EDI]); st->cr(); st->print( "EIP=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_EIP]); st->print(", CR2=" INTPTR_FORMAT, uc->uc_mcontext.cr2); st->print(", EFLAGS=" INTPTR_FORMAT, uc->uc_mcontext.gregs[REG_EFL]); #endif // AMD64 st->cr(); st->cr(); intptr_t *sp = (intptr_t *)os::Linux::ucontext_get_sp(uc); st->print_cr("Top of Stack: (sp=" PTR_FORMAT ")", sp); print_hex_dump(st, (address)sp, (address)(sp + 8*sizeof(intptr_t)), sizeof(intptr_t)); st->cr(); // Note: it may be unsafe to inspect memory near pc. For example, pc may // point to garbage if entry point in an nmethod is corrupted. Leave // this at the end, and hope for the best. address pc = os::Linux::ucontext_get_pc(uc); st->print_cr("Instructions: (pc=" PTR_FORMAT ")", pc); print_hex_dump(st, pc - 16, pc + 16, sizeof(char)); }寄存器的信息在分析出错的时候是非常重要的

打印出执行附近的部分机器码

Instructions: (pc=0x00007f48f14ef51a) 0x00007f48f14ef4fa: 90 90 55 48 89 e5 48 81 ec 98 9f 00 00 48 89 bd 0x00007f48f14ef50a: f8 5f ff ff 48 89 b5 f0 5f ff ff b8 00 00 00 00 0x00007f48f14ef51a: c7 00 01 00 00 00 c6 85 00 60 ff ff ff c9 c3 90 0x00007f48f14ef52a: 90 90 90 90 90 90 55 48 89 e5 53 48 8d 1d 94 00
在instruction 部分中会打印出部分的机器码

格式是

地址：机器码
第一种使用udis库里带的udcli工具来反汇编

命令：

echo '90 90 55 48 89 e5 48 81 ec 98 9f 00 00 48 89 bd' | udcli -intel -x -64 -o 0x00007f48f14ef4fa
显示出对应的汇编

第二种可以用

objectdump -d -C libjvm.so >> jvmsodisass.dump

查找偏移地址 0x593045， 就是当时的执行的汇编，然后结合上下文，源码猜测出问题的语句。

d.寄存器对应的内存的值

RAX=0x0000000000000000 is an unknown value RBX=0x000000041a07d1e8 is an oop {method} - klass: {other class} RCX=0x0000000000000000 is an unknown value RDX=0x0000000040111800 is a thread RSP=0x0000000041261b88 is pointing into the stack for thread: 0x0000000040111800 RBP=0x000000004126bb20 is pointing into the stack for thread: 0x0000000040111800 RSI=0x000000004126bb80 is pointing into the stack for thread: 0x0000000040111800 RDI=0x00000000401119d0 is an unknown value R8 =0x0000000040111c40 is an unknown value R9 =0x00007f48fcc8b550: <offset 0xa85550> in /usr/java/jdk1.6.0_30/jre/lib/amd64/server/libjvm.so at 0x00007f48fc206000 R10=0x00007f48f8ca7d41 is an Interpreter codelet method entry point (kind = native) [0x00007f48f8ca7ae0, 0x00007f48f8ca8320] 2112 bytes R11=0x00007f48fc98f270: <offset 0x789270> in /usr/java/jdk1.6.0_30/jre/lib/amd64/server/libjvm.so at 0x00007f48fc206000 R12=0x0000000000000000 is an unknown value R13=0x000000041a07d1e8 is an oop {method} - klass: {other class} R14=0x000000004126bb88 is pointing into the stack for thread: 0x0000000040111800 R15=0x0000000040111800 is a thread jvm 会通过寄存器的值对找对应的对象，也是一个比较好的参考

e. 其他的信息

error 里面还有一些线程信息，还有当时内存映像信息，这些都可以作为分析的部分参考

crash 报告可以大概的反应出一个当时的情况，特别是在没有core dump的时候，是比较有助于帮助分析的，但如果有core dump的话，最终还是core dump能快速准确的发现问题原因。