7

Customer reported an error in one of our programs caused by division by zero. We have only this VLM line:

kernel: myprog[16122] trap divide error rip:79dd99 rsp:2b6d2ea40450 error:0

I do not believe there is core file for that.

I searched through the Internet to find how I can tell the line of the program that caused this division by zero, but so far I am failing.

I understand that 16122 is pid of the program, so that will not help me.

I suspect that rsp:2b6d2ea40450 has something to do with the address of the line that caused the error (0x2b6d2ea40450) but is that true?

If it is then how can I translate it to a physical approximate location in the source assuming I can load debug version of myprog into gdb, and then request to show the context around this address...

Any, any help will be greatly appreciated!

Jolta
  • 2,620
  • 1
  • 29
  • 42
Grzegorz
  • 3,207
  • 3
  • 20
  • 43

1 Answers1

14

rip is the instruction pointer, rsp is the stack pointer. The stack pointer is not too useful unless you have a core image or a running process.

You can use either addr2line or the disassemble command in gdb to see the line that got the error, based on the ip.

$ cat divtest.c
main()
{
    int a, b;

    a = 1; b = a/0;
}

$ ./divtest
Floating point exception (core dumped)
$ dmesg|tail -1
[ 6827.463256] traps: divtest[3255] trap divide error ip:400504 sp:7fff54e81330
    error:0 in divtest[400000+1000]

$ addr2line -e divtest 400504
./divtest.c:5

$ gdb divtest
(gdb) disass /m 0x400504
Dump of assembler code for function main:
2       {
   0x00000000004004f0 :     push   %rbp
   0x00000000004004f1 :     mov    %rsp,%rbp

3               int a, b;
4       
5               a = 1; b = a/0;
   0x00000000004004f4 :     movl   $0x1,-0x4(%rbp)
   0x00000000004004fb :    mov    -0x4(%rbp),%eax
   0x00000000004004fe :    mov    $0x0,%ecx
   0x0000000000400503 :    cltd   
   0x0000000000400504 :    idiv   %ecx
   0x0000000000400506 :    mov    %eax,-0x8(%rbp)

6       }
   0x0000000000400509 :    pop    %rbp
   0x000000000040050a :    retq   

End of assembler dump.
Mark Plotnick
  • 9,598
  • 1
  • 24
  • 40