setuid() hang in multithread application

nyquist892004

hi, 各位大侠

在Freescale P1020RDB上，gcc-4.3.74 + linux .32 + eglibc-2.8.74, 我遇到了这样一个问题:
     libc setuid() function hang in multithread application， 无论是main thread还是之后fork出的thread，只要有多threads，就一定会hang。


单独运行是就hang住了，但用strace跟踪就不会hang。  

发现在sys_setuid()中的确是看到system call是成功了的，但thread是hang在了do_futex()中。


我的测试程序如下:
/*
*        a subthread to setuid(0)
*/
void* tfunc_setuid(void *arg)
{
        int i = 0;

                setuid(0);

        return NULL;
}

int main(int args, char **argv)
{
        pthread_t tid_setuid = 0;
        int i = 0;

        pthread_create(&tid_setuid, NULL, tfunc_setuid, NULL);

        pthread_join(tid_setuid, NULL);

        return 0;
}


"strace"的输出如下:

clone(Process 7456 attached
child_stack=0x4882e030, flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_
THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTID, pare
nt_tidptr=0x4882e4f8, tls=0x48835930, child_tidptr=0x4882e4f = 7456
[pid  7455] futex(0x4882e4f8, FUTEX_WAIT, 7456, NULL <unfinished ...>
[pid  7456] get_robust_list(0x4882e500, 0xc, 0x4882e4f = 0
[pid  7456] tgkill(7455, 7455, SIGRT_1) = 0
[pid  7455] <... futex resumed> )       = ? ERESTARTSYS (To be restarted)
[pid  7456] futex(0x4882dd48, FUTEX_WAIT_PRIVATE, 1, NULL <unfinished ...>
[pid  7455] --- SIGRT_1 (Unknown signal 33) @ 0 (0) ---
[pid  7455] setuid(0)                   = 0                                                                                # 这里一个setuid()，为什么main thread会调用setuid(0)  ???
[pid  7455] futex(0x4882dd48, FUTEX_WAKE_PRIVATE, 1 <unfinished ...>
[pid  7456] <... futex resumed> )       = 0
[pid  7455] <... futex resumed> )       = 1
[pid  7456] tgkill(7455, 7455, SIGRT_1 <unfinished ...>
[pid  7455] futex(0x4801f608, FUTEX_WAKE_PRIVATE, 1 <unfinished ...>
[pid  7456] <... tgkill resumed> )      = 0
[pid  7455] <... futex resumed> )       = 0
[pid  7456] futex(0x4882dd48, FUTEX_WAIT_PRIVATE, 1, NULL <unfinished ...>
[pid  7455] futex(0x4882dd48, FUTEX_WAKE_PRIVATE, 1 <unfinished ...>
[pid  7456] <... futex resumed> )       = -1 EAGAIN (Resource temporarily unavai
lable)
[pid  7455] <... futex resumed> )       = 0
[pid  7456] setuid(0 <unfinished ...>                                                                            # 这里是sub thread的setuid(0)
[pid  7455] rt_sigreturn(0 <unfinished ...>
[pid  7456] <... setuid resumed> )      = 0
[pid  7455] <... rt_sigreturn resumed> ) = -1216537848
[pid  7456] exit(0)                     = ?
Process 7456 detached
--- SIGRT_1 (Unknown signal 33) @ 0 (0) ---
syscall_268436712(0, 0x4882dd60, 0x100004e8, 0xbfac32a0, 0x4882e4f8, 0x48835930,
0x4882dd38, 0xffffe2e1, 0x4801f474, 0x100350, 0x10018854, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0x4882e6ac, 0x4801f410, 0xffed230, 0xffed1f8, 0x4882e4b0, 0xff
cc8a4, 0x2d900, 0) = -1 (errno 3
futex(0x4882dd48, FUTEX_WAKE_PRIVATE, 1) = 0
futex(0x4801f608, FUTEX_WAKE_PRIVATE, 1) = 0
rt_sigreturn(0)                         = -1216537848
futex(0x4882e4f8, FUTEX_WAIT, 7456, NULL) = -1 EAGAIN (Resource temporarily unav
ailable)
exit_group(0)                           = ?   



我不理解的地方有:
     从strace的输出来看，为什么main thread和sub thread都有setuid()，而在C code里我只在sub thread里调用了setuid() ?
     为什么会hang，是不是这个eglic在thread和signal上有什么问题?  在strace下这个测试程序是不会hang的，是不是和程序处于TASK_TRACED有关系。