[PATCH 0/6 -mm] cpu-timers patches

[PATCH 0/6 -mm] cpu-timers patches

[Stanislaw Gruszka]

Here are patches for cpu-timers I mainly wrote to not call 
thread_group_cputime() in fastpath_timer_check(), which we
discuss in this thread:

http://thread.gmane.org/gmane.linux.kernel/909349/focus=912327
  
On the way I fix some other bugs (SIGEV_NONE timer, old increase value
in timer_settime()) and do other cleanups and optimizations, patch 5
however increase time when sighand->siglock is held. 

Compared with previous send of patches as RFC, I rebased them to -mm
tree, change some changelogs and do some minor code changes.

[PATCH 1/6 -mm] cpu-timers: simplify RLIMIT_CPU handling
[PATCH 2/6 -mm] cpu-timers: cleanup arm_timer()
[PATCH 3/6 -mm] cpu-timers: return correct previous timer reload value
[PATCH 4/6 -mm] cpu-timers: change SIGEV_NONE timer implementation
[PATCH 5/6 -mm] cpu-timers: assure to not iterate over all threads in fastpath_timer_check()
[PATCH 6/6 -mm] cpu-timers: optimize run_posix_cpu_timers()

Cheers
Stanislaw

[PATCH 1/6 -mm] cpu-timers: simplify RLIMIT_CPU handling

[Stanislaw Gruszka]

Let always set signal->cputime_expires expiration cache when setting
new itimer, POSIX 1.b timer, and RLIMIT_CPU. Since we are initializing
prof_exp expiration cache during fork(), this allow to remove
"RLIMIT_CPU != inf" check from fastpath_timer_check() and do some other
cleanups.

Checked against regression using test cases from:
http://marc.info/?l=linux-kernel&m=123749066504641&w=4
http://marc.info/?l=linux-kernel&m=123811277916642&w=2

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 kernel/posix-cpu-timers.c |   74 ++++++++++++++++-----------------------------
 1 files changed, 26 insertions(+), 48 deletions(-)

diff --git a/kernel/posix-cpu-timers.c b/kernel/posix-cpu-timers.c
index 9931a64..3e31d51 100644
--- a/kernel/posix-cpu-timers.c
+++ b/kernel/posix-cpu-timers.c
@@ -11,19 +11,17 @@
 #include <trace/events/timer.h>
 
 /*
- * Called after updating RLIMIT_CPU to set timer expiration if necessary.
+ * Called after updating RLIMIT_CPU to run cpu timer and update
+ * tsk->signal->cputime_expires expiration cache if necessary. Needs siglock
+ * protection since other code may update expiration cache as well.
  */
 void update_rlimit_cpu(struct task_struct *task, unsigned long rlim_new)
 {
 	cputime_t cputime = secs_to_cputime(rlim_new);
-	struct signal_struct *const sig = task->signal;
 
-	if (cputime_eq(sig->it[CPUCLOCK_PROF].expires, cputime_zero) ||
-	    cputime_gt(sig->it[CPUCLOCK_PROF].expires, cputime)) {
-		spin_lock_irq(&task->sighand->siglock);
-		set_process_cpu_timer(task, CPUCLOCK_PROF, &cputime, NULL);
-		spin_unlock_irq(&task->sighand->siglock);
-	}
+	spin_lock_irq(&task->sighand->siglock);
+	set_process_cpu_timer(task, CPUCLOCK_PROF, &cputime, NULL);
+	spin_unlock_irq(&task->sighand->siglock);
 }
 
 static int check_clock(const clockid_t which_clock)
@@ -564,7 +562,6 @@ static void arm_timer(struct k_itimer *timer, union cpu_time_count now)
 	struct list_head *head, *listpos;
 	struct cpu_timer_list *const nt = &timer->it.cpu;
 	struct cpu_timer_list *next;
-	unsigned long i;
 
 	head = (CPUCLOCK_PERTHREAD(timer->it_clock) ?
 		p->cpu_timers : p->signal->cpu_timers);
@@ -630,20 +627,11 @@ static void arm_timer(struct k_itimer *timer, union cpu_time_count now)
 			default:
 				BUG();
 			case CPUCLOCK_VIRT:
-				if (expires_le(sig->it[CPUCLOCK_VIRT].expires,
-					       exp->cpu))
-					break;
-				sig->cputime_expires.virt_exp = exp->cpu;
-				break;
+				if (expires_gt(sig->cputime_expires.virt_exp, exp->cpu))
+					sig->cputime_expires.virt_exp = exp->cpu;
 			case CPUCLOCK_PROF:
-				if (expires_le(sig->it[CPUCLOCK_PROF].expires,
-					       exp->cpu))
-					break;
-				i = sig->rlim[RLIMIT_CPU].rlim_cur;
-				if (i != RLIM_INFINITY &&
-				    i <= cputime_to_secs(exp->cpu))
-					break;
-				sig->cputime_expires.prof_exp = exp->cpu;
+				if (expires_gt(sig->cputime_expires.prof_exp, exp->cpu))
+					sig->cputime_expires.prof_exp = exp->cpu;
 				break;
 			case CPUCLOCK_SCHED:
 				sig->cputime_expires.sched_exp = exp->sched;
@@ -1386,7 +1374,7 @@ static inline int fastpath_timer_check(struct task_struct *tsk)
 			return 1;
 	}
 
-	return sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY;
+	return 0;
 }
 
 /*
@@ -1452,21 +1440,23 @@ void run_posix_cpu_timers(struct task_struct *tsk)
 }
 
 /*
- * Set one of the process-wide special case CPU timers.
+ * Set one of the process-wide special case CPU timers or RLIMIT_CPU.
  * The tsk->sighand->siglock must be held by the caller.
- * The *newval argument is relative and we update it to be absolute, *oldval
- * is absolute and we update it to be relative.
  */
 void set_process_cpu_timer(struct task_struct *tsk, unsigned int clock_idx,
 			   cputime_t *newval, cputime_t *oldval)
 {
 	union cpu_time_count now;
-	struct list_head *head;
 
 	BUG_ON(clock_idx == CPUCLOCK_SCHED);
 	cpu_timer_sample_group(clock_idx, tsk, &now);
 
 	if (oldval) {
+		/*
+		 * We are setting itimer. The *oldval is absolute and we update
+		 * it to be relative, *newval argument is relative and we update
+		 * it to be absolute.
+		 */
 		if (!cputime_eq(*oldval, cputime_zero)) {
 			if (cputime_le(*oldval, now.cpu)) {
 				/* Just about to fire. */
@@ -1479,33 +1469,21 @@ void set_process_cpu_timer(struct task_struct *tsk, unsigned int clock_idx,
 		if (cputime_eq(*newval, cputime_zero))
 			return;
 		*newval = cputime_add(*newval, now.cpu);
-
-		/*
-		 * If the RLIMIT_CPU timer will expire before the
-		 * ITIMER_PROF timer, we have nothing else to do.
-		 */
-		if (tsk->signal->rlim[RLIMIT_CPU].rlim_cur
-		    < cputime_to_secs(*newval))
-			return;
 	}
 
 	/*
-	 * Check whether there are any process timers already set to fire
-	 * before this one.  If so, we don't have anything more to do.
+	 * Update expiration cache if we are the earliest timer, or eventually
+	 * RLIMIT_CPU limit is earlier than prof_exp cpu timer expire.
 	 */
-	head = &tsk->signal->cpu_timers[clock_idx];
-	if (list_empty(head) ||
-	    cputime_ge(list_first_entry(head,
-				  struct cpu_timer_list, entry)->expires.cpu,
-		       *newval)) {
-		switch (clock_idx) {
-		case CPUCLOCK_PROF:
+	switch (clock_idx) {
+	case CPUCLOCK_PROF:
+		if (expires_gt(tsk->signal->cputime_expires.prof_exp, *newval))
 			tsk->signal->cputime_expires.prof_exp = *newval;
-			break;
-		case CPUCLOCK_VIRT:
+		break;
+	case CPUCLOCK_VIRT:
+		if (expires_gt(tsk->signal->cputime_expires.virt_exp, *newval))
 			tsk->signal->cputime_expires.virt_exp = *newval;
-			break;
-		}
+		break;
 	}
 }
 
-- 
1.6.2.5

[PATCH 2/6 -mm] cpu-timers: cleanup arm_timer()

[Stanislaw Gruszka]

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 kernel/posix-cpu-timers.c |  104 +++++++++++++++------------------------------
 1 files changed, 35 insertions(+), 69 deletions(-)

diff --git a/kernel/posix-cpu-timers.c b/kernel/posix-cpu-timers.c
index 3e31d51..1039501 100644
--- a/kernel/posix-cpu-timers.c
+++ b/kernel/posix-cpu-timers.c
@@ -546,97 +546,63 @@ static inline int expires_gt(cputime_t expires, cputime_t new_exp)
 	       cputime_gt(expires, new_exp);
 }
 
-static inline int expires_le(cputime_t expires, cputime_t new_exp)
-{
-	return !cputime_eq(expires, cputime_zero) &&
-	       cputime_le(expires, new_exp);
-}
 /*
  * Insert the timer on the appropriate list before any timers that
  * expire later.  This must be called with the tasklist_lock held
  * for reading, and interrupts disabled.
  */
-static void arm_timer(struct k_itimer *timer, union cpu_time_count now)
+static void arm_timer(struct k_itimer *timer)
 {
 	struct task_struct *p = timer->it.cpu.task;
 	struct list_head *head, *listpos;
+	struct task_cputime *cputime_expires;
 	struct cpu_timer_list *const nt = &timer->it.cpu;
 	struct cpu_timer_list *next;
 
-	head = (CPUCLOCK_PERTHREAD(timer->it_clock) ?
-		p->cpu_timers : p->signal->cpu_timers);
+	if (CPUCLOCK_PERTHREAD(timer->it_clock)) {
+		head = p->cpu_timers;
+		cputime_expires = &p->cputime_expires;
+	} else {
+		head = p->signal->cpu_timers;
+		cputime_expires = &p->signal->cputime_expires;
+	}
 	head += CPUCLOCK_WHICH(timer->it_clock);
 
 	BUG_ON(!irqs_disabled());
 	spin_lock(&p->sighand->siglock);
 
 	listpos = head;
-	if (CPUCLOCK_WHICH(timer->it_clock) == CPUCLOCK_SCHED) {
-		list_for_each_entry(next, head, entry) {
-			if (next->expires.sched > nt->expires.sched)
-				break;
-			listpos = &next->entry;
-		}
-	} else {
-		list_for_each_entry(next, head, entry) {
-			if (cputime_gt(next->expires.cpu, nt->expires.cpu))
-				break;
-			listpos = &next->entry;
-		}
+	list_for_each_entry(next, head, entry) {
+		if (cpu_time_before(timer->it_clock, nt->expires, next->expires))
+			break;
+		listpos = &next->entry;
 	}
 	list_add(&nt->entry, listpos);
 
 	if (listpos == head) {
+		union cpu_time_count *exp = &nt->expires;
+
 		/*
-		 * We are the new earliest-expiring timer.
-		 * If we are a thread timer, there can always
-		 * be a process timer telling us to stop earlier.
+		 * We are the new earliest-expiring POSIX 1.b timer, hence
+		 * need to update expiration cache. Take into account that
+		 * for process timers we share expiration cache with itimers
+		 * and RLIMIT_CPU and for thread timers with RLIMIT_RTTIME.
 		 */
 
-		if (CPUCLOCK_PERTHREAD(timer->it_clock)) {
-			union cpu_time_count *exp = &nt->expires;
-
-			switch (CPUCLOCK_WHICH(timer->it_clock)) {
-			default:
-				BUG();
-			case CPUCLOCK_PROF:
-				if (expires_gt(p->cputime_expires.prof_exp,
-					       exp->cpu))
-					p->cputime_expires.prof_exp = exp->cpu;
-				break;
-			case CPUCLOCK_VIRT:
-				if (expires_gt(p->cputime_expires.virt_exp,
-					       exp->cpu))
-					p->cputime_expires.virt_exp = exp->cpu;
-				break;
-			case CPUCLOCK_SCHED:
-				if (p->cputime_expires.sched_exp == 0 ||
-				    p->cputime_expires.sched_exp > exp->sched)
-					p->cputime_expires.sched_exp =
-								exp->sched;
-				break;
-			}
-		} else {
-			struct signal_struct *const sig = p->signal;
-			union cpu_time_count *exp = &timer->it.cpu.expires;
-
-			/*
-			 * For a process timer, set the cached expiration time.
-			 */
-			switch (CPUCLOCK_WHICH(timer->it_clock)) {
-			default:
-				BUG();
-			case CPUCLOCK_VIRT:
-				if (expires_gt(sig->cputime_expires.virt_exp, exp->cpu))
-					sig->cputime_expires.virt_exp = exp->cpu;
-			case CPUCLOCK_PROF:
-				if (expires_gt(sig->cputime_expires.prof_exp, exp->cpu))
-					sig->cputime_expires.prof_exp = exp->cpu;
-				break;
-			case CPUCLOCK_SCHED:
-				sig->cputime_expires.sched_exp = exp->sched;
-				break;
-			}
+		switch (CPUCLOCK_WHICH(timer->it_clock)) {
+		case CPUCLOCK_PROF:
+			if (expires_gt(cputime_expires->prof_exp, exp->cpu))
+				cputime_expires->prof_exp = exp->cpu;
+			break;
+		case CPUCLOCK_VIRT:
+			if (expires_gt(cputime_expires->virt_exp, exp->cpu))
+				cputime_expires->virt_exp = exp->cpu;
+			break;
+		case CPUCLOCK_SCHED:
+			if (cputime_expires->sched_exp == 0 ||
+			    cputime_expires->sched_exp > exp->sched)
+				cputime_expires->sched_exp = exp->sched;
+			break;
 		}
 	}
 
@@ -818,7 +784,7 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
 	if (new_expires.sched != 0 &&
 	    (timer->it_sigev_notify & ~SIGEV_THREAD_ID) != SIGEV_NONE &&
 	    cpu_time_before(timer->it_clock, val, new_expires)) {
-		arm_timer(timer, val);
+		arm_timer(timer);
 	}
 
 	read_unlock(&tasklist_lock);
@@ -1282,7 +1248,7 @@ void posix_cpu_timer_schedule(struct k_itimer *timer)
 	/*
 	 * Now re-arm for the new expiry time.
 	 */
-	arm_timer(timer, now);
+	arm_timer(timer);
 
 out_unlock:
 	read_unlock(&tasklist_lock);
-- 
1.6.2.5

[PATCH 3/6 -mm] cpu-timers: return correct previous timer reload value

[Stanislaw Gruszka]

According POSIX we need to correctly set old timer it_interval value
when user request that in timer_settime(). Tested using below program.

 #include <sys/time.h>
 #include <signal.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <time.h>
 #include <unistd.h>
 #include <assert.h>

 int main(void)
 {
	struct sigaction act;
	struct sigevent evt = { };
	timer_t tid;
	struct itimerspec spec, u_spec, k_spec;

	evt.sigev_notify = SIGEV_SIGNAL;
	evt.sigev_signo = SIGPROF;
	assert(timer_create(CLOCK_PROCESS_CPUTIME_ID, &evt,  &tid) == 0);

	spec.it_value.tv_sec = 1;
	spec.it_value.tv_nsec = 2;
	spec.it_interval.tv_sec = 3;
	spec.it_interval.tv_nsec = 4;
	u_spec = spec;
	assert(timer_settime(tid, 0, &spec,  NULL) == 0);

	spec.it_value.tv_sec = 5;
	spec.it_value.tv_nsec = 6;
	spec.it_interval.tv_sec = 7;
	spec.it_interval.tv_nsec = 8;
	assert(timer_settime(tid, 0, &spec,  &k_spec) == 0);

 #define PRT(val) printf(#val ":\t%d/%d\n", (int) u_spec.val, (int) k_spec.val)
	PRT(it_value.tv_sec);
	PRT(it_value.tv_nsec);
	PRT(it_interval.tv_sec);
	PRT(it_interval.tv_nsec);

	return 0;
 }

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 kernel/posix-cpu-timers.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/kernel/posix-cpu-timers.c b/kernel/posix-cpu-timers.c
index 1039501..e9aa4dc 100644
--- a/kernel/posix-cpu-timers.c
+++ b/kernel/posix-cpu-timers.c
@@ -675,7 +675,7 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
 			struct itimerspec *new, struct itimerspec *old)
 {
 	struct task_struct *p = timer->it.cpu.task;
-	union cpu_time_count old_expires, new_expires, val;
+	union cpu_time_count old_expires, new_expires, old_incr, val;
 	int ret;
 
 	if (unlikely(p == NULL)) {
@@ -706,6 +706,7 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
 	BUG_ON(!irqs_disabled());
 
 	ret = 0;
+	old_incr = timer->it.cpu.incr;
 	spin_lock(&p->sighand->siglock);
 	old_expires = timer->it.cpu.expires;
 	if (unlikely(timer->it.cpu.firing)) {
@@ -821,7 +822,7 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
  out:
 	if (old) {
 		sample_to_timespec(timer->it_clock,
-				   timer->it.cpu.incr, &old->it_interval);
+				   old_incr, &old->it_interval);
 	}
 	return ret;
 }
-- 
1.6.2.5

[PATCH 4/6 -mm] cpu-timers: change SIGEV_NONE timer implementation

[Stanislaw Gruszka]

When user will setup timer without associated signal and process does
not use any other cpu timers and does not exit, tsk->signal->cputimer
is enabled and running forever, change that.

I used below program to check patch does not break current user space
visible behavior.

 #include <sys/time.h>
 #include <signal.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 #include <time.h>
 #include <unistd.h>
 #include <assert.h>

 void consume_cpu(void)
 {
	int i = 0;
	int count = 0;

	for(i=0; i<100000000; i++)
		count++;
 }

 int main(void)
 {
	int i;
	struct sigaction act;
	struct sigevent evt = { };
	timer_t tid;
	struct itimerspec spec = { };

	evt.sigev_notify = SIGEV_NONE;
	assert(timer_create(CLOCK_PROCESS_CPUTIME_ID, &evt,  &tid) == 0);

	spec.it_value.tv_sec = 10;
	assert(timer_settime(tid, 0, &spec,  NULL) == 0);

	for (i = 0; i < 30; i++) {
		consume_cpu();
		memset(&spec, 0, sizeof(spec));
		assert(timer_gettime(tid, &spec) == 0);
		printf("%lu.%09lu\n",
			(unsigned long) spec.it_value.tv_sec,
			(unsigned long) spec.it_value.tv_nsec);
	}

	assert(timer_delete(tid) == 0);
	return 0;
 }

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 kernel/posix-cpu-timers.c |   28 ++++++----------------------
 1 files changed, 6 insertions(+), 22 deletions(-)

diff --git a/kernel/posix-cpu-timers.c b/kernel/posix-cpu-timers.c
index e9aa4dc..faa0b42 100644
--- a/kernel/posix-cpu-timers.c
+++ b/kernel/posix-cpu-timers.c
@@ -614,7 +614,12 @@ static void arm_timer(struct k_itimer *timer)
  */
 static void cpu_timer_fire(struct k_itimer *timer)
 {
-	if (unlikely(timer->sigq == NULL)) {
+	if ((timer->it_sigev_notify & ~SIGEV_THREAD_ID) == SIGEV_NONE) {
+		/*
+		 * User don't want any signal.
+		 */
+		timer->it.cpu.expires.sched = 0;
+	} else if (unlikely(timer->sigq == NULL)) {
 		/*
 		 * This a special case for clock_nanosleep,
 		 * not a normal timer from sys_timer_create.
@@ -783,7 +788,6 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
 	 */
 	timer->it.cpu.expires = new_expires;
 	if (new_expires.sched != 0 &&
-	    (timer->it_sigev_notify & ~SIGEV_THREAD_ID) != SIGEV_NONE &&
 	    cpu_time_before(timer->it_clock, val, new_expires)) {
 		arm_timer(timer);
 	}
@@ -808,7 +812,6 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
 	timer->it_overrun = -1;
 
 	if (new_expires.sched != 0 &&
-	    (timer->it_sigev_notify & ~SIGEV_THREAD_ID) != SIGEV_NONE &&
 	    !cpu_time_before(timer->it_clock, val, new_expires)) {
 		/*
 		 * The designated time already passed, so we notify
@@ -882,25 +885,6 @@ void posix_cpu_timer_get(struct k_itimer *timer, struct itimerspec *itp)
 		read_unlock(&tasklist_lock);
 	}
 
-	if ((timer->it_sigev_notify & ~SIGEV_THREAD_ID) == SIGEV_NONE) {
-		if (timer->it.cpu.incr.sched == 0 &&
-		    cpu_time_before(timer->it_clock,
-				    timer->it.cpu.expires, now)) {
-			/*
-			 * Do-nothing timer expired and has no reload,
-			 * so it's as if it was never set.
-			 */
-			timer->it.cpu.expires.sched = 0;
-			itp->it_value.tv_sec = itp->it_value.tv_nsec = 0;
-			return;
-		}
-		/*
-		 * Account for any expirations and reloads that should
-		 * have happened.
-		 */
-		bump_cpu_timer(timer, now);
-	}
-
 	if (unlikely(clear_dead)) {
 		/*
 		 * We've noticed that the thread is dead, but
-- 
1.6.2.5

[PATCH 5/6 -mm] cpu-timers: assure to not iterate over all threads in fastpath_timer_check()

[Stanislaw Gruszka]

Spread p->sighand->siglock locking scope to assure fastpath_timer_check()
never iterate over each threads. Without locking there is small possibility
that signal->cputimer will stop running while we write values to
signal->cputime_expires.

Calling thread_group_cputime() from fastpath_timer_check() is not only
bad because it is slow, also it is racy with __exit_signal() which can
lead to invalid signal->{s,u}time values.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 kernel/posix-cpu-timers.c |   14 +++++++-------
 1 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/kernel/posix-cpu-timers.c b/kernel/posix-cpu-timers.c
index faa0b42..67c877e 100644
--- a/kernel/posix-cpu-timers.c
+++ b/kernel/posix-cpu-timers.c
@@ -549,7 +549,7 @@ static inline int expires_gt(cputime_t expires, cputime_t new_exp)
 /*
  * Insert the timer on the appropriate list before any timers that
  * expire later.  This must be called with the tasklist_lock held
- * for reading, and interrupts disabled.
+ * for reading, interrupts disabled and p->sighand->siglock taken.
  */
 static void arm_timer(struct k_itimer *timer)
 {
@@ -568,9 +568,6 @@ static void arm_timer(struct k_itimer *timer)
 	}
 	head += CPUCLOCK_WHICH(timer->it_clock);
 
-	BUG_ON(!irqs_disabled());
-	spin_lock(&p->sighand->siglock);
-
 	listpos = head;
 	list_for_each_entry(next, head, entry) {
 		if (cpu_time_before(timer->it_clock, nt->expires, next->expires))
@@ -605,8 +602,6 @@ static void arm_timer(struct k_itimer *timer)
 			break;
 		}
 	}
-
-	spin_unlock(&p->sighand->siglock);
 }
 
 /*
@@ -719,7 +714,6 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
 		ret = TIMER_RETRY;
 	} else
 		list_del_init(&timer->it.cpu.entry);
-	spin_unlock(&p->sighand->siglock);
 
 	/*
 	 * We need to sample the current value to convert the new
@@ -773,6 +767,7 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
 		 * disable this firing since we are already reporting
 		 * it as an overrun (thanks to bump_cpu_timer above).
 		 */
+		spin_unlock(&p->sighand->siglock);
 		read_unlock(&tasklist_lock);
 		goto out;
 	}
@@ -792,6 +787,7 @@ int posix_cpu_timer_set(struct k_itimer *timer, int flags,
 		arm_timer(timer);
 	}
 
+	spin_unlock(&p->sighand->siglock);
 	read_unlock(&tasklist_lock);
 
 	/*
@@ -1205,6 +1201,7 @@ void posix_cpu_timer_schedule(struct k_itimer *timer)
 			goto out;
 		}
 		read_lock(&tasklist_lock); /* arm_timer needs it.  */
+		spin_lock(&p->sighand->siglock);
 	} else {
 		read_lock(&tasklist_lock);
 		if (unlikely(p->signal == NULL)) {
@@ -1225,6 +1222,7 @@ void posix_cpu_timer_schedule(struct k_itimer *timer)
 			clear_dead_task(timer, now);
 			goto out_unlock;
 		}
+		spin_lock(&p->sighand->siglock);
 		cpu_timer_sample_group(timer->it_clock, p, &now);
 		bump_cpu_timer(timer, now);
 		/* Leave the tasklist_lock locked for the call below.  */
@@ -1233,7 +1231,9 @@ void posix_cpu_timer_schedule(struct k_itimer *timer)
 	/*
 	 * Now re-arm for the new expiry time.
 	 */
+	BUG_ON(!irqs_disabled());
 	arm_timer(timer);
+	spin_unlock(&p->sighand->siglock);
 
 out_unlock:
 	read_unlock(&tasklist_lock);
-- 
1.6.2.5

[PATCH 6/6 -mm] cpu-timers: optimize run_posix_cpu_timers()

[Stanislaw Gruszka]

We can optimize and simplify things taking into account signal->cputimer
is always running when we have configured any process wide cpu timer.

In check_process_timers(), we don't have to check if new updated value
of signal->cputime_expires is smaller, since we maintain new first
expiration time ({prof,virt,sched}_expires) in code flow and all other
writes to expiration cache are protected by sighand->siglock .

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 kernel/posix-cpu-timers.c |   80 ++++++++++++++++----------------------------
 1 files changed, 29 insertions(+), 51 deletions(-)

diff --git a/kernel/posix-cpu-timers.c b/kernel/posix-cpu-timers.c
index 67c877e..ffba8f4 100644
--- a/kernel/posix-cpu-timers.c
+++ b/kernel/posix-cpu-timers.c
@@ -1000,16 +1000,9 @@ static void stop_process_timers(struct signal_struct *sig)
 	struct thread_group_cputimer *cputimer = &sig->cputimer;
 	unsigned long flags;
 
-	if (!cputimer->running)
-		return;
-
 	spin_lock_irqsave(&cputimer->lock, flags);
 	cputimer->running = 0;
 	spin_unlock_irqrestore(&cputimer->lock, flags);
-
-	sig->cputime_expires.prof_exp = cputime_zero;
-	sig->cputime_expires.virt_exp = cputime_zero;
-	sig->cputime_expires.sched_exp = 0;
 }
 
 static u32 onecputick;
@@ -1046,6 +1039,23 @@ static void check_cpu_itimer(struct task_struct *tsk, struct cpu_itimer *it,
 	}
 }
 
+/**
+ * task_cputime_zero - Check a task_cputime struct for all zero fields.
+ *
+ * @cputime:	The struct to compare.
+ *
+ * Checks @cputime to see if all fields are zero.  Returns true if all fields
+ * are zero, false if any field is nonzero.
+ */
+static inline int task_cputime_zero(const struct task_cputime *cputime)
+{
+	if (cputime_eq(cputime->utime, cputime_zero) &&
+	    cputime_eq(cputime->stime, cputime_zero) &&
+	    cputime->sum_exec_runtime == 0)
+		return 1;
+	return 0;
+}
+
 /*
  * 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
@@ -1062,19 +1072,6 @@ static void check_process_timers(struct task_struct *tsk,
 	struct task_cputime cputime;
 
 	/*
-	 * Don't sample the current process CPU clocks if there are no timers.
-	 */
-	if (list_empty(&timers[CPUCLOCK_PROF]) &&
-	    cputime_eq(sig->it[CPUCLOCK_PROF].expires, cputime_zero) &&
-	    sig->rlim[RLIMIT_CPU].rlim_cur == RLIM_INFINITY &&
-	    list_empty(&timers[CPUCLOCK_VIRT]) &&
-	    cputime_eq(sig->it[CPUCLOCK_VIRT].expires, cputime_zero) &&
-	    list_empty(&timers[CPUCLOCK_SCHED])) {
-		stop_process_timers(sig);
-		return;
-	}
-
-	/*
 	 * Collect the current process totals.
 	 */
 	thread_group_cputimer(tsk, &cputime);
@@ -1161,18 +1158,11 @@ static void check_process_timers(struct task_struct *tsk,
 		}
 	}
 
-	if (!cputime_eq(prof_expires, cputime_zero) &&
-	    (cputime_eq(sig->cputime_expires.prof_exp, cputime_zero) ||
-	     cputime_gt(sig->cputime_expires.prof_exp, prof_expires)))
-		sig->cputime_expires.prof_exp = prof_expires;
-	if (!cputime_eq(virt_expires, cputime_zero) &&
-	    (cputime_eq(sig->cputime_expires.virt_exp, cputime_zero) ||
-	     cputime_gt(sig->cputime_expires.virt_exp, virt_expires)))
-		sig->cputime_expires.virt_exp = virt_expires;
-	if (sched_expires != 0 &&
-	    (sig->cputime_expires.sched_exp == 0 ||
-	     sig->cputime_expires.sched_exp > sched_expires))
-		sig->cputime_expires.sched_exp = sched_expires;
+	sig->cputime_expires.prof_exp = prof_expires;
+	sig->cputime_expires.virt_exp = virt_expires;
+	sig->cputime_expires.sched_exp = sched_expires;
+	if (task_cputime_zero(&sig->cputime_expires))
+		stop_process_timers(sig);
 }
 
 /*
@@ -1245,23 +1235,6 @@ out:
 }
 
 /**
- * task_cputime_zero - Check a task_cputime struct for all zero fields.
- *
- * @cputime:	The struct to compare.
- *
- * Checks @cputime to see if all fields are zero.  Returns true if all fields
- * are zero, false if any field is nonzero.
- */
-static inline int task_cputime_zero(const struct task_cputime *cputime)
-{
-	if (cputime_eq(cputime->utime, cputime_zero) &&
-	    cputime_eq(cputime->stime, cputime_zero) &&
-	    cputime->sum_exec_runtime == 0)
-		return 1;
-	return 0;
-}
-
-/**
  * task_cputime_expired - Compare two task_cputime entities.
  *
  * @sample:	The task_cputime structure to be checked for expiration.
@@ -1317,7 +1290,7 @@ static inline int fastpath_timer_check(struct task_struct *tsk)
 	}
 
 	sig = tsk->signal;
-	if (!task_cputime_zero(&sig->cputime_expires)) {
+	if (sig->cputimer.running) {
 		struct task_cputime group_sample;
 
 		thread_group_cputimer(tsk, &group_sample);
@@ -1354,7 +1327,12 @@ void run_posix_cpu_timers(struct task_struct *tsk)
 	 * put them on the firing list.
 	 */
 	check_thread_timers(tsk, &firing);
-	check_process_timers(tsk, &firing);
+	/*
+	 * If there are any active process wide timers (POSIX 1.b, itimers,
+	 * RLIMIT_CPU) cputimer must be running.
+	 */
+	if (tsk->signal->cputimer.running)
+		check_process_timers(tsk, &firing);
 
 	/*
 	 * We must release these locks before taking any timer's lock.
-- 
1.6.2.5