There is code in sys_wait4() that makes a task wait for the children of all
members of the thread group. This code has a race condition when more than
one task in a thread group calls wait. Once it has picked a zomebie task
to return it drops the tasklist lock before it calls release_task(). This
allows another task to select the same zombie to clean up.
My fix for this race condition is to set the zombie task's state to a
different value before the tasklist lock is released. This means no other
waiting task will select it. There could be other ways of marking a task
as already selected for cleanup, but this seemed the simplest.
The patch is appended below. It applies cleanly to 2.4.14
Dave McCracken
======================================================================
Dave McCracken IBM Linux Base Kernel Team 1-512-838-3059
[email protected] T/L 678-3059
-------------------------
--- linux-2.4.13/kernel/exit.c Mon Oct 22 11:27:55 2001
+++ linux-2.4.14-pre8/kernel/exit.c Mon Nov 5 11:56:45 2001
@@ -534,6 +534,9 @@
}
goto end_wait4;
case TASK_ZOMBIE:
+ /* Make sure no other waiter picks this task up */
+ p->state = TASK_DEAD;
+
current->times.tms_cutime += p->times.tms_utime + p->times.tms_cutime;
current->times.tms_cstime += p->times.tms_stime + p->times.tms_cstime;
read_unlock(&tasklist_lock);
--- linux-2.4.13/include/linux/sched.h Tue Oct 23 23:59:06 2001
+++ linux-2.4.14-pre8/include/linux/sched.h Mon Nov 5 12:56:20 2001
@@ -88,6 +88,7 @@
#define TASK_UNINTERRUPTIBLE 2
#define TASK_ZOMBIE 4
#define TASK_STOPPED 8
+#define TASK_DEAD 16
#define __set_task_state(tsk, state_value) \
do { (tsk)->state = (state_value); } while (0)