论坛徽章:: 0

电梯直达

1楼 [收藏(0)] [报告]

发表于 2012-01-09 17:29 |只看该作者 |倒序浏览

本帖最后由 cjog 于 2012-01-09 17:31 编辑

main()
{
   char c='x';
   read(0,&c,1);
   printf("go\n");

   printf("come\n");
}

编译生成a.out后,运行gdb a.out
(gdb) b main
Breakpoint 1 at 0x80483e5: file gdb.c, line 3.
(gdb) r
Starting program: /tmp/tx

Breakpoint 1, main () at gdb.c:3
3    char c='x';
(gdb) set $foo=&c
(gdb) watch *$foo //对c所在的地址设置一个观察点
Hardware watchpoint 2: *$foo
(gdb) del 1
(gdb) c
Continuing.
Hardware watchpoint 2: *$foo

Old value = 0 '\0'
New value = 120 'x'
main () at gdb.c:4
4    read(0,&c,1);
(gdb) c
Continuing.    //输入回车

Hardware watchpoint 2: *$foo

Old value = 120 'x'
New value = 10 '\n'
0x00ae3402 in __kernel_vsyscall ()
(gdb)
Continuing.

Program received signal SIGTRAP, Trace/breakpoint trap. //此后，gdb不停收到sigtrap信号，不知何故
0x009f23f3 in __read_nocancel () from /lib/libc.so.6
(gdb)
Continuing.

Program received signal SIGTRAP, Trace/breakpoint trap.
0x009f23f4 in __read_nocancel () from /lib/libc.so.6
(gdb)
Continuing.

Program received signal SIGTRAP, Trace/breakpoint trap.
0x009f23f9 in __read_nocancel () from /lib/libc.so.6
(gdb)

但如果把代码改成
main()
{
   char c='x';

   printf("go\n");
   c++;

   printf("come\n");
}

sigtrap就不会发生。第一个程序，c的值是在kernel里面被改的，第二个程序，c是在用户空间被改的。为什么在kernel里修改观察点的值，就会不断收到sigtrap呢？