posix-timers: fix RLIMIT_CPU && fork()
[safe/jmp/linux-2.6] / kernel / posix-cpu-timers.c
index 153dcb2..8e5d9a6 100644 (file)
 #include <linux/kernel_stat.h>
 
 /*
- * Allocate the thread_group_cputime structure appropriately and fill in the
- * current values of the fields.  Called from copy_signal() via
- * thread_group_cputime_clone_thread() when adding a second or subsequent
- * thread to a thread group.  Assumes interrupts are enabled when called.
- */
-int thread_group_cputime_alloc(struct task_struct *tsk)
-{
-       struct signal_struct *sig = tsk->signal;
-       struct task_cputime *cputime;
-
-       /*
-        * If we have multiple threads and we don't already have a
-        * per-CPU task_cputime struct (checked in the caller), allocate
-        * one and fill it in with the times accumulated so far.  We may
-        * race with another thread so recheck after we pick up the sighand
-        * lock.
-        */
-       cputime = alloc_percpu(struct task_cputime);
-       if (cputime == NULL)
-               return -ENOMEM;
-       spin_lock_irq(&tsk->sighand->siglock);
-       if (sig->cputime.totals) {
-               spin_unlock_irq(&tsk->sighand->siglock);
-               free_percpu(cputime);
-               return 0;
-       }
-       sig->cputime.totals = cputime;
-       cputime = per_cpu_ptr(sig->cputime.totals, smp_processor_id());
-       cputime->utime = tsk->utime;
-       cputime->stime = tsk->stime;
-       cputime->sum_exec_runtime = tsk->se.sum_exec_runtime;
-       spin_unlock_irq(&tsk->sighand->siglock);
-       return 0;
-}
-
-/**
- * thread_group_cputime - Sum the thread group time fields across all CPUs.
- *
- * @tsk:       The task we use to identify the thread group.
- * @times:     task_cputime structure in which we return the summed fields.
- *
- * Walk the list of CPUs to sum the per-CPU time fields in the thread group
- * time structure.
- */
-void thread_group_cputime(
-       struct task_struct *tsk,
-       struct task_cputime *times)
-{
-       struct signal_struct *sig;
-       int i;
-       struct task_cputime *tot;
-
-       sig = tsk->signal;
-       if (unlikely(!sig) || !sig->cputime.totals) {
-               times->utime = tsk->utime;
-               times->stime = tsk->stime;
-               times->sum_exec_runtime = tsk->se.sum_exec_runtime;
-               return;
-       }
-       times->stime = times->utime = cputime_zero;
-       times->sum_exec_runtime = 0;
-       for_each_possible_cpu(i) {
-               tot = per_cpu_ptr(tsk->signal->cputime.totals, i);
-               times->utime = cputime_add(times->utime, tot->utime);
-               times->stime = cputime_add(times->stime, tot->stime);
-               times->sum_exec_runtime += tot->sum_exec_runtime;
-       }
-}
-
-/*
  * Called after updating RLIMIT_CPU to set timer expiration if necessary.
  */
 void update_rlimit_cpu(unsigned long rlim_new)
@@ -300,6 +230,71 @@ static int cpu_clock_sample(const clockid_t which_clock, struct task_struct *p,
        return 0;
 }
 
+void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times)
+{
+       struct sighand_struct *sighand;
+       struct signal_struct *sig;
+       struct task_struct *t;
+
+       *times = INIT_CPUTIME;
+
+       rcu_read_lock();
+       sighand = rcu_dereference(tsk->sighand);
+       if (!sighand)
+               goto out;
+
+       sig = tsk->signal;
+
+       t = tsk;
+       do {
+               times->utime = cputime_add(times->utime, t->utime);
+               times->stime = cputime_add(times->stime, t->stime);
+               times->sum_exec_runtime += t->se.sum_exec_runtime;
+
+               t = next_thread(t);
+       } while (t != tsk);
+
+       times->utime = cputime_add(times->utime, sig->utime);
+       times->stime = cputime_add(times->stime, sig->stime);
+       times->sum_exec_runtime += sig->sum_sched_runtime;
+out:
+       rcu_read_unlock();
+}
+
+static void update_gt_cputime(struct task_cputime *a, struct task_cputime *b)
+{
+       if (cputime_gt(b->utime, a->utime))
+               a->utime = b->utime;
+
+       if (cputime_gt(b->stime, a->stime))
+               a->stime = b->stime;
+
+       if (b->sum_exec_runtime > a->sum_exec_runtime)
+               a->sum_exec_runtime = b->sum_exec_runtime;
+}
+
+void thread_group_cputimer(struct task_struct *tsk, struct task_cputime *times)
+{
+       struct thread_group_cputimer *cputimer = &tsk->signal->cputimer;
+       struct task_cputime sum;
+       unsigned long flags;
+
+       spin_lock_irqsave(&cputimer->lock, flags);
+       if (!cputimer->running) {
+               cputimer->running = 1;
+               /*
+                * The POSIX timer interface allows for absolute time expiry
+                * values through the TIMER_ABSTIME flag, therefore we have
+                * to synchronize the timer to the clock every time we start
+                * it.
+                */
+               thread_group_cputime(tsk, &sum);
+               update_gt_cputime(&cputimer->cputime, &sum);
+       }
+       *times = cputimer->cputime;
+       spin_unlock_irqrestore(&cputimer->lock, flags);
+}
+
 /*
  * Sample a process (thread group) clock for the given group_leader task.
  * Must be called with tasklist_lock held for reading.
@@ -311,7 +306,7 @@ static int cpu_clock_sample_group(const clockid_t which_clock,
        struct task_cputime cputime;
 
        thread_group_cputime(p, &cputime);
-       switch (which_clock) {
+       switch (CPUCLOCK_WHICH(which_clock)) {
        default:
                return -EINVAL;
        case CPUCLOCK_PROF:
@@ -527,7 +522,7 @@ void posix_cpu_timers_exit_group(struct task_struct *tsk)
 {
        struct task_cputime cputime;
 
-       thread_group_cputime(tsk, &cputime);
+       thread_group_cputimer(tsk, &cputime);
        cleanup_timers(tsk->signal->cpu_timers,
                       cputime.utime, cputime.stime, cputime.sum_exec_runtime);
 }
@@ -686,6 +681,33 @@ static void cpu_timer_fire(struct k_itimer *timer)
 }
 
 /*
+ * Sample a process (thread group) timer for the given group_leader task.
+ * Must be called with tasklist_lock held for reading.
+ */
+static int cpu_timer_sample_group(const clockid_t which_clock,
+                                 struct task_struct *p,
+                                 union cpu_time_count *cpu)
+{
+       struct task_cputime cputime;
+
+       thread_group_cputimer(p, &cputime);
+       switch (CPUCLOCK_WHICH(which_clock)) {
+       default:
+               return -EINVAL;
+       case CPUCLOCK_PROF:
+               cpu->cpu = cputime_add(cputime.utime, cputime.stime);
+               break;
+       case CPUCLOCK_VIRT:
+               cpu->cpu = cputime.utime;
+               break;
+       case CPUCLOCK_SCHED:
+               cpu->sched = cputime.sum_exec_runtime + task_delta_exec(p);
+               break;
+       }
+       return 0;
+}
+
+/*
  * Guts of sys_timer_settime for CPU timers.
  * This is called with the timer locked and interrupts disabled.
  * If we return TIMER_RETRY, it's necessary to release the timer's lock
@@ -746,7 +768,7 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
        if (CPUCLOCK_PERTHREAD(timer->it_clock)) {
                cpu_clock_sample(timer->it_clock, p, &val);
        } else {
-               cpu_clock_sample_group(timer->it_clock, p, &val);
+               cpu_timer_sample_group(timer->it_clock, p, &val);
        }
 
        if (old) {
@@ -894,7 +916,7 @@ void posix_cpu_timer_get(struct k_itimer *timer, struct itimerspec *itp)
                        read_unlock(&tasklist_lock);
                        goto dead;
                } else {
-                       cpu_clock_sample_group(timer->it_clock, p, &now);
+                       cpu_timer_sample_group(timer->it_clock, p, &now);
                        clear_dead = (unlikely(p->exit_state) &&
                                      thread_group_empty(p));
                }
@@ -1034,6 +1056,19 @@ static void check_thread_timers(struct task_struct *tsk,
        }
 }
 
+static void stop_process_timers(struct task_struct *tsk)
+{
+       struct thread_group_cputimer *cputimer = &tsk->signal->cputimer;
+       unsigned long flags;
+
+       if (!cputimer->running)
+               return;
+
+       spin_lock_irqsave(&cputimer->lock, flags);
+       cputimer->running = 0;
+       spin_unlock_irqrestore(&cputimer->lock, flags);
+}
+
 /*
  * Check for any per-thread CPU timers that have fired and move them
  * off the tsk->*_timers list onto the firing list.  Per-thread timers
@@ -1057,13 +1092,15 @@ static void check_process_timers(struct task_struct *tsk,
            sig->rlim[RLIMIT_CPU].rlim_cur == RLIM_INFINITY &&
            list_empty(&timers[CPUCLOCK_VIRT]) &&
            cputime_eq(sig->it_virt_expires, cputime_zero) &&
-           list_empty(&timers[CPUCLOCK_SCHED]))
+           list_empty(&timers[CPUCLOCK_SCHED])) {
+               stop_process_timers(tsk);
                return;
+       }
 
        /*
         * Collect the current process totals.
         */
-       thread_group_cputime(tsk, &cputime);
+       thread_group_cputimer(tsk, &cputime);
        utime = cputime.utime;
        ptime = cputime_add(utime, cputime.stime);
        sum_sched_runtime = cputime.sum_exec_runtime;
@@ -1234,7 +1271,7 @@ void posix_cpu_timer_schedule(struct k_itimer *timer)
                        clear_dead_task(timer, now);
                        goto out_unlock;
                }
-               cpu_clock_sample_group(timer->it_clock, p, &now);
+               cpu_timer_sample_group(timer->it_clock, p, &now);
                bump_cpu_timer(timer, now);
                /* Leave the tasklist_lock locked for the call below.  */
        }
@@ -1308,9 +1345,10 @@ static inline int task_cputime_expired(const struct task_cputime *sample,
  */
 static inline int fastpath_timer_check(struct task_struct *tsk)
 {
-       struct signal_struct *sig = tsk->signal;
+       struct signal_struct *sig;
 
-       if (unlikely(!sig))
+       /* tsk == current, ensure it is safe to use ->signal/sighand */
+       if (unlikely(tsk->exit_state))
                return 0;
 
        if (!task_cputime_zero(&tsk->cputime_expires)) {
@@ -1323,14 +1361,17 @@ static inline int fastpath_timer_check(struct task_struct *tsk)
                if (task_cputime_expired(&task_sample, &tsk->cputime_expires))
                        return 1;
        }
+
+       sig = tsk->signal;
        if (!task_cputime_zero(&sig->cputime_expires)) {
                struct task_cputime group_sample;
 
-               thread_group_cputime(tsk, &group_sample);
+               thread_group_cputimer(tsk, &group_sample);
                if (task_cputime_expired(&group_sample, &sig->cputime_expires))
                        return 1;
        }
-       return 0;
+
+       return sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY;
 }
 
 /*
@@ -1408,7 +1449,7 @@ void set_process_cpu_timer(struct task_struct *tsk, unsigned int clock_idx,
        struct list_head *head;
 
        BUG_ON(clock_idx == CPUCLOCK_SCHED);
-       cpu_clock_sample_group(clock_idx, tsk, &now);
+       cpu_timer_sample_group(clock_idx, tsk, &now);
 
        if (oldval) {
                if (!cputime_eq(*oldval, cputime_zero)) {