2.4.x, can't ptrace a task created with clone() ?
Tested with 4.2.20 and 2.4.21-pre3, after PTRACE_ATTACHing to a process I created with clone(), wait() returns ECHILD . strace of the parent shows
clone(child_stack=0x80599a8, flags=CLONE_VM) = 1292 rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8) = 0 rt_sigaction(SIGCHLD, NULL, {SIG_DFL}, 8) = 0 rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0nanosleep({5, 0}, {5, 0}) = 0ptrace(PTRACE_ATTACH, 1292, 0, 0) = 0
--- SIGCHLD (Child exited) ---wait4(-1, 0xbffffbf0, 0, NULL) = -1 ECHILD (No child processes)write(2, "waitpid: No child processes\n", 28) = 28
If I replace the clone in my test program with a fork, everything works as I expected. Is this a bug, or are my expectations at fault? if the latter, how _do_ I attach to one of my cloned children?
Here's the test program I've been using:
#include <sched.h>
#include <sys/ptrace.h>
#include <linux/user.h>
#include <stdio.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/wait.h>
void child(void *d)
{
int foo;
fprintf(stderr,"child stack is somewhere around 0x%x\n",&foo);
while(1){
fprintf(stderr, "still here\n");
sleep(3);
}
}main()
{
pid_t kid;
int waitr;
struct user_regs_struct regs;
kid=clone(child,SIGSTKSZ*8+malloc(SIGSTKSZ*8), CLONE_VM ,0);
/* kid=fork(); */
if(kid<0) perror("clone");
if(kid==0) child(0);
sleep(5);
if(kid>0) {
if(ptrace(PTRACE_ATTACH,kid,0,0)) perror("PTRACE_ATTACH");
if(wait(&waitr)==-1) {
perror("waitpid");
exit(1);
}
if(ptrace(PTRACE_GETREGS,kid,0,®s)) perror("PTRACE_GETREGS");
fprintf(stderr,"child stack pointer is 0x%x\n",regs.esp);
if(ptrace(PTRACE_DETACH,kid,0,0)) perror("PTRACE_DETACH");
kill(kid,15); }
}
libc 2.3.1, in case that makes a difference. Let me know if there's anything else you wanted to ask
-dan
Signature · 1 line
http://www.cliki.net/ - Link farm for free CL-on-Unix resources
Quoted 15 lines · Daniel Barlow
Tested with 4.2.20 and 2.4.21-pre3, after
PTRACE_ATTACHingto a process I created withclone(),wait()returns ECHILD . strace of the parent shows
clone(child_stack=0x80599a8, flags=CLONE_VM)= 1292rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8)= 0rt_sigaction(SIGCHLD, NULL, {SIG_DFL}, 8)= 0rt_sigprocmask(SIG_SETMASK, [], NULL, 8)= 0nanosleep({5, 0}, {5, 0})= 0ptrace(PTRACE_ATTACH, 1292, 0, 0)= 0
--- SIGCHLD (Child exited) ---wait4(-1, 0xbffffbf0, 0, NULL)= -1 ECHILD (No child processes)write(2, "waitpid: No child processes\n", 28)= 28
Quoted 3 lines
If I replace the clone in my test program with a fork, everything works as I expected. Is this a bug, or are my expectations at fault? if the latter, how _do_ I attach to one of my cloned children?
You did attach. But if you want to wait for a cloned child, you need to use __WCLONE or __WALL.
Signature · 2 lines
Daniel Jacobowitz MontaVista Software Debian GNU/Linux Developer