Re: zombie with CLONE_THREAD

[Andreas Schwab]

Roland McGrath <roland@xxxxxxxxxx> writes:

> Are you saying that if the ptracer dies, it can leave some threads in limbo?
> I think that case is supposed to work because forget_original_parent will
> move all the threads ptrace'd by the dying tracer process to be ptrace'd by
> init, which will then clean up their zombies as previously described.

Here is the test case, run it with "strace -f ./clone".  When the bug
happens then strace is stuck waiting for it's traced child that just died,
but you may have to try a few times before it happens.

#include <stdio.h>
#include <signal.h>
#include <sched.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/mman.h>

extern int __clone (int (*__fn) (void *__arg), void *__child_stack,
		    int __flags, void *__arg, ...);
extern int __clone2 (int (*__fn) (void *__arg), void *__child_stack_base,
		     size_t __child_stack_size, int __flags, void *__arg, ...);

static int thread (void *arg)
{
  write (2, "thread\n", sizeof ("thread\n"));
  *(volatile int *) 0;
  return 0;
}

#define STACK_SIZE 1024 * 1024
#define CLONE_FLAGS CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM

int
main (void)
{
  void *stack = mmap (0, STACK_SIZE, PROT_READ|PROT_WRITE,
		      MAP_ANON|MAP_PRIVATE, -1, 0);
  pid_t pid;

#ifdef __ia64__
  pid = __clone2 (thread, stack, STACK_SIZE - 64, CLONE_FLAGS, 0);
#else
  pid = __clone (thread, stack + STACK_SIZE - 64, CLONE_FLAGS, 0);
#endif
  printf ("pid = %d\n", pid);
  sleep (1);
  return 0;
}

	    

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@xxxxxxx
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."