[PATCH] run_posix_cpu_timers: remove a bogus BUG_ON()
authorOleg Nesterov <oleg@tv-sign.ru>
Thu, 15 Jun 2006 16:11:43 +0000 (20:11 +0400)
committerLinus Torvalds <torvalds@g5.osdl.org>
Sat, 17 Jun 2006 17:52:13 +0000 (10:52 -0700)
do_exit() clears ->it_##clock##_expires, but nothing prevents
another cpu to attach the timer to exiting process after that.
arm_timer() tries to protect against this race, but the check
is racy.

After exit_notify() does 'write_unlock_irq(&tasklist_lock)' and
before do_exit() calls 'schedule() local timer interrupt can find
tsk->exit_state != 0. If that state was EXIT_DEAD (or another cpu
does sys_wait4) interrupted task has ->signal == NULL.

At this moment exiting task has no pending cpu timers, they were
cleanuped in __exit_signal()->posix_cpu_timers_exit{,_group}(),
so we can just return from irq.

John Stultz recently confirmed this bug, see

http://marc.theaimsgroup.com/?l=linux-kernel&m=115015841413687

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
kernel/exit.c
kernel/posix-cpu-timers.c

index e95b932822109e6bd517eee9a67876fedc91b426..e06d0c10a24e2f4e054d17c21d032beadf8d1ba6 100644 (file)
@@ -881,14 +881,6 @@ fastcall NORET_TYPE void do_exit(long code)
 
        tsk->flags |= PF_EXITING;
 
-       /*
-        * Make sure we don't try to process any timer firings
-        * while we are already exiting.
-        */
-       tsk->it_virt_expires = cputime_zero;
-       tsk->it_prof_expires = cputime_zero;
-       tsk->it_sched_expires = 0;
-
        if (unlikely(in_atomic()))
                printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
                                current->comm, current->pid,
index 9d9169aa2e243f8920f305d82e99124bc61d4945..4882bf1e094a0727b5d23c31ef2ce9afd9516446 100644 (file)
@@ -1288,30 +1288,30 @@ void run_posix_cpu_timers(struct task_struct *tsk)
 
 #undef UNEXPIRED
 
-       BUG_ON(tsk->exit_state);
-
        /*
         * Double-check with locks held.
         */
        read_lock(&tasklist_lock);
-       spin_lock(&tsk->sighand->siglock);
+       if (likely(tsk->signal != NULL)) {
+               spin_lock(&tsk->sighand->siglock);
 
-       /*
-        * Here we take off tsk->cpu_timers[N] and tsk->signal->cpu_timers[N]
-        * all the timers that are firing, and put them on the firing list.
-        */
-       check_thread_timers(tsk, &firing);
-       check_process_timers(tsk, &firing);
+               /*
+                * Here we take off tsk->cpu_timers[N] and tsk->signal->cpu_timers[N]
+                * all the timers that are firing, and put them on the firing list.
+                */
+               check_thread_timers(tsk, &firing);
+               check_process_timers(tsk, &firing);
 
-       /*
-        * We must release these locks before taking any timer's lock.
-        * There is a potential race with timer deletion here, as the
-        * siglock now protects our private firing list.  We have set
-        * the firing flag in each timer, so that a deletion attempt
-        * that gets the timer lock before we do will give it up and
-        * spin until we've taken care of that timer below.
-        */
-       spin_unlock(&tsk->sighand->siglock);
+               /*
+                * We must release these locks before taking any timer's lock.
+                * There is a potential race with timer deletion here, as the
+                * siglock now protects our private firing list.  We have set
+                * the firing flag in each timer, so that a deletion attempt
+                * that gets the timer lock before we do will give it up and
+                * spin until we've taken care of that timer below.
+                */
+               spin_unlock(&tsk->sighand->siglock);
+       }
        read_unlock(&tasklist_lock);
 
        /*