posix-timers: sys_timer_create: remove the buggy PF_EXITING check
Oleg Nesterov [Mon, 22 Sep 2008 21:42:47 +0000 (14:42 -0700)]
sys_timer_create() return -EINVAL if the target thread has PF_EXITING.

This doesn't really make sense, the sub-thread can die right after unlock.
And in fact, this is just wrong.  Without SIGEV_THREAD_ID good_sigevent()
returns ->group_leader, and it is very possible that the leader is already
dead.  This is OK, we shouldn't return the error in this case.

Remove this check and the comment.  Note that the "process" was found
under tasklist_lock, it must have ->sighand != NULL.

Also, remove a couple of unneeded initializations.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: mingo@elte.hu
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

kernel/posix-timers.c

index bd9c931..60b2620 100644 (file)
@@ -460,9 +460,9 @@ sys_timer_create(const clockid_t which_clock,
                 timer_t __user * created_timer_id)
 {
        int error = 0;
-       struct k_itimer *new_timer = NULL;
+       struct k_itimer *new_timer;
        int new_timer_id;
-       struct task_struct *process = NULL;
+       struct task_struct *process;
        unsigned long flags;
        sigevent_t event;
        int it_id_set = IT_ID_NOT_SET;
@@ -523,32 +523,12 @@ sys_timer_create(const clockid_t which_clock,
 
                read_lock(&tasklist_lock);
                if ((process = good_sigevent(&event))) {
-                       /*
-                        * We may be setting up this process for another
-                        * thread.  It may be exiting.  To catch this
-                        * case the we check the PF_EXITING flag.  If
-                        * the flag is not set, the siglock will catch
-                        * him before it is too late (in exit_itimers).
-                        *
-                        * The exec case is a bit more invloved but easy
-                        * to code.  If the process is in our thread
-                        * group (and it must be or we would not allow
-                        * it here) and is doing an exec, it will cause
-                        * us to be killed.  In this case it will wait
-                        * for us to die which means we can finish this
-                        * linkage with our last gasp. I.e. no code :)
-                        */
+                       get_task_struct(process);
                        spin_lock_irqsave(&process->sighand->siglock, flags);
-                       if (!(process->flags & PF_EXITING)) {
-                               get_task_struct(process);
-                               new_timer->it_process = process;
-                               list_add(&new_timer->list,
-                                        &process->signal->posix_timers);
-                               spin_unlock_irqrestore(&process->sighand->siglock, flags);
-                       } else {
-                               spin_unlock_irqrestore(&process->sighand->siglock, flags);
-                               process = NULL;
-                       }
+                       new_timer->it_process = process;
+                       list_add(&new_timer->list,
+                               &process->signal->posix_timers);
+                       spin_unlock_irqrestore(&process->sighand->siglock, flags);
                }
                read_unlock(&tasklist_lock);
                if (!process) {