[PATCH] RCU signal handling

RCU tasklist_lock and RCU signal handling: send signals RCU-read-locked
instead of tasklist_lock read-locked.  This is a scalability improvement on
SMP and a preemption-latency improvement under PREEMPT_RCU.

Signed-off-by: Paul E. McKenney <paulmck@us.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: William Irwin <wli@holomorphy.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

8 files changed, 143 insertions, 31 deletions

diff --git a/fs/exec.c b/fs/exec.c
index e75a9548da8e..e9650cd22a3b 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -760,7 +760,7 @@ no_thread_group:
 		spin_lock(&oldsighand->siglock);
 		spin_lock(&newsighand->siglock);
 
-		current->sighand = newsighand;
+		rcu_assign_pointer(current->sighand, newsighand);
 		recalc_sigpending();
 
 		spin_unlock(&newsighand->siglock);
@@ -768,7 +768,7 @@ no_thread_group:
 		write_unlock_irq(&tasklist_lock);
 
 		if (atomic_dec_and_test(&oldsighand->count))
-			kmem_cache_free(sighand_cachep, oldsighand);
+			sighand_free(oldsighand);
 	}
 
 	BUG_ON(!thread_group_leader(current));
diff --git a/include/linux/sched.h b/include/linux/sched.h
index a74662077d60..a6af77e9b4cf 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -34,6 +34,7 @@
 #include <linux/percpu.h>
 #include <linux/topology.h>
 #include <linux/seccomp.h>
+#include <linux/rcupdate.h>
 
 #include <linux/auxvec.h>	/* For AT_VECTOR_SIZE */
 
@@ -350,8 +351,16 @@ struct sighand_struct {
 	atomic_t		count;
 	struct k_sigaction	action[_NSIG];
 	spinlock_t		siglock;
+	struct rcu_head		rcu;
 };
 
+extern void sighand_free_cb(struct rcu_head *rhp);
+
+static inline void sighand_free(struct sighand_struct *sp)
+{
+	call_rcu(&sp->rcu, sighand_free_cb);
+}
+
 /*
  * NOTE! "signal_struct" does not have it's own
  * locking, because a shared signal_struct always
@@ -844,6 +853,7 @@ struct task_struct {
 	int cpuset_mems_generation;
 #endif
 	atomic_t fs_excl;	/* holding fs exclusive resources */
+	struct rcu_head rcu;
 };
 
 static inline pid_t process_group(struct task_struct *tsk)
@@ -867,8 +877,26 @@ static inline int pid_alive(struct task_struct *p)
 extern void free_task(struct task_struct *tsk);
 extern void __put_task_struct(struct task_struct *tsk);
 #define get_task_struct(tsk) do { atomic_inc(&(tsk)->usage); } while(0)
-#define put_task_struct(tsk) \
-do { if (atomic_dec_and_test(&(tsk)->usage)) __put_task_struct(tsk); } while(0)
+
+static inline int get_task_struct_rcu(struct task_struct *t)
+{
+	int oldusage;
+
+	do {
+		oldusage = atomic_read(&t->usage);
+		if (oldusage == 0)
+			return 0;
+	} while (cmpxchg(&t->usage.counter, oldusage, oldusage+1) != oldusage);
+	return 1;
+}
+
+extern void __put_task_struct_cb(struct rcu_head *rhp);
+
+static inline void put_task_struct(struct task_struct *t)
+{
+	if (atomic_dec_and_test(&t->usage))
+		call_rcu(&t->rcu, __put_task_struct_cb);
+}
 
 /*
  * Per process flags
diff --git a/kernel/exit.c b/kernel/exit.c
index ee515683b92d..c73a7eb26de3 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -72,7 +72,6 @@ repeat:
 		__ptrace_unlink(p);
 	BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
 	__exit_signal(p);
-	__exit_sighand(p);
 	/*
 	 * Note that the fastpath in sys_times depends on __exit_signal having
 	 * updated the counters before a task is removed from the tasklist of
diff --git a/kernel/fork.c b/kernel/fork.c
index fb8572a42297..7fe3adfa65cb 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -743,6 +743,14 @@ int unshare_files(void)
 
 EXPORT_SYMBOL(unshare_files);
 
+void sighand_free_cb(struct rcu_head *rhp)
+{
+	struct sighand_struct *sp;
+
+	sp = container_of(rhp, struct sighand_struct, rcu);
+	kmem_cache_free(sighand_cachep, sp);
+}
+
 static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk)
 {
 	struct sighand_struct *sig;
@@ -752,7 +760,7 @@ static inline int copy_sighand(unsigned long clone_flags, struct task_struct * t
 		return 0;
 	}
 	sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
-	tsk->sighand = sig;
+	rcu_assign_pointer(tsk->sighand, sig);
 	if (!sig)
 		return -ENOMEM;
 	spin_lock_init(&sig->siglock);
diff --git a/kernel/pid.c b/kernel/pid.c
index edba31c681ac..1acc07246991 100644
--- a/kernel/pid.c
+++ b/kernel/pid.c
@@ -136,7 +136,7 @@ struct pid * fastcall find_pid(enum pid_type type, int nr)
 	struct hlist_node *elem;
 	struct pid *pid;
 
-	hlist_for_each_entry(pid, elem,
+	hlist_for_each_entry_rcu(pid, elem,
 			&pid_hash[type][pid_hashfn(nr)], pid_chain) {
 		if (pid->nr == nr)
 			return pid;
@@ -150,15 +150,15 @@ int fastcall attach_pid(task_t *task, enum pid_type type, int nr)
 
 	task_pid = &task->pids[type];
 	pid = find_pid(type, nr);
+	task_pid->nr = nr;
 	if (pid == NULL) {
-		hlist_add_head(&task_pid->pid_chain,
-				&pid_hash[type][pid_hashfn(nr)]);
 		INIT_LIST_HEAD(&task_pid->pid_list);
+		hlist_add_head_rcu(&task_pid->pid_chain,
+				   &pid_hash[type][pid_hashfn(nr)]);
 	} else {
 		INIT_HLIST_NODE(&task_pid->pid_chain);
-		list_add_tail(&task_pid->pid_list, &pid->pid_list);
+		list_add_tail_rcu(&task_pid->pid_list, &pid->pid_list);
 	}
-	task_pid->nr = nr;
 
 	return 0;
 }
@@ -170,20 +170,20 @@ static fastcall int __detach_pid(task_t *task, enum pid_type type)
 
 	pid = &task->pids[type];
 	if (!hlist_unhashed(&pid->pid_chain)) {
-		hlist_del(&pid->pid_chain);
 
-		if (list_empty(&pid->pid_list))
+		if (list_empty(&pid->pid_list)) {
 			nr = pid->nr;
-		else {
+			hlist_del_rcu(&pid->pid_chain);
+		} else {
 			pid_next = list_entry(pid->pid_list.next,
 						struct pid, pid_list);
 			/* insert next pid from pid_list to hash */
-			hlist_add_head(&pid_next->pid_chain,
-				&pid_hash[type][pid_hashfn(pid_next->nr)]);
+			hlist_replace_rcu(&pid->pid_chain,
+					  &pid_next->pid_chain);
 		}
 	}
 
-	list_del(&pid->pid_list);
+	list_del_rcu(&pid->pid_list);
 	pid->nr = 0;
 
 	return nr;
diff --git a/kernel/rcupdate.c b/kernel/rcupdate.c
index c9afc61240e4..0a669bd2f6d1 100644
--- a/kernel/rcupdate.c
+++ b/kernel/rcupdate.c
@@ -35,6 +35,7 @@
 #include <linux/init.h>
 #include <linux/spinlock.h>
 #include <linux/smp.h>
+#include <linux/rcupdate.h>
 #include <linux/interrupt.h>
 #include <linux/sched.h>
 #include <asm/atomic.h>
diff --git a/kernel/sched.c b/kernel/sched.c
index 6f46c94cc29e..92733091154c 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -176,6 +176,13 @@ static unsigned int task_timeslice(task_t *p)
 #define task_hot(p, now, sd) ((long long) ((now) - (p)->last_ran)	\
 				< (long long) (sd)->cache_hot_time)
 
+void __put_task_struct_cb(struct rcu_head *rhp)
+{
+	__put_task_struct(container_of(rhp, struct task_struct, rcu));
+}
+
+EXPORT_SYMBOL_GPL(__put_task_struct_cb);
+
 /*
  * These are the runqueue data structures:
  */
diff --git a/kernel/signal.c b/kernel/signal.c
index d7611f189ef7..64737c72dadd 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -329,13 +329,20 @@ void __exit_sighand(struct task_struct *tsk)
 	/* Ok, we're done with the signal handlers */
 	tsk->sighand = NULL;
 	if (atomic_dec_and_test(&sighand->count))
-		kmem_cache_free(sighand_cachep, sighand);
+		sighand_free(sighand);
 }
 
 void exit_sighand(struct task_struct *tsk)
 {
 	write_lock_irq(&tasklist_lock);
-	__exit_sighand(tsk);
+	rcu_read_lock();
+	if (tsk->sighand != NULL) {
+		struct sighand_struct *sighand = rcu_dereference(tsk->sighand);
+		spin_lock(&sighand->siglock);
+		__exit_sighand(tsk);
+		spin_unlock(&sighand->siglock);
+	}
+	rcu_read_unlock();
 	write_unlock_irq(&tasklist_lock);
 }
 
@@ -345,12 +352,14 @@ void exit_sighand(struct task_struct *tsk)
 void __exit_signal(struct task_struct *tsk)
 {
 	struct signal_struct * sig = tsk->signal;
-	struct sighand_struct * sighand = tsk->sighand;
+	struct sighand_struct * sighand;
 
 	if (!sig)
 		BUG();
 	if (!atomic_read(&sig->count))
 		BUG();
+	rcu_read_lock();
+	sighand = rcu_dereference(tsk->sighand);
 	spin_lock(&sighand->siglock);
 	posix_cpu_timers_exit(tsk);
 	if (atomic_dec_and_test(&sig->count)) {
@@ -358,6 +367,7 @@ void __exit_signal(struct task_struct *tsk)
 		if (tsk == sig->curr_target)
 			sig->curr_target = next_thread(tsk);
 		tsk->signal = NULL;
+		__exit_sighand(tsk);
 		spin_unlock(&sighand->siglock);
 		flush_sigqueue(&sig->shared_pending);
 	} else {
@@ -389,9 +399,11 @@ void __exit_signal(struct task_struct *tsk)
 		sig->nvcsw += tsk->nvcsw;
 		sig->nivcsw += tsk->nivcsw;
 		sig->sched_time += tsk->sched_time;
+		__exit_sighand(tsk);
 		spin_unlock(&sighand->siglock);
 		sig = NULL;	/* Marker for below.  */
 	}
+	rcu_read_unlock();
 	clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
 	flush_sigqueue(&tsk->pending);
 	if (sig) {
@@ -1080,18 +1092,28 @@ void zap_other_threads(struct task_struct *p)
 }
 
 /*
- * Must be called with the tasklist_lock held for reading!
+ * Must be called under rcu_read_lock() or with tasklist_lock read-held.
  */
 int group_send_sig_info(int sig, struct siginfo *info, struct task_struct *p)
 {
 	unsigned long flags;
+	struct sighand_struct *sp;
 	int ret;
 
+retry:
 	ret = check_kill_permission(sig, info, p);
-	if (!ret && sig && p->sighand) {
-		spin_lock_irqsave(&p->sighand->siglock, flags);
+	if (!ret && sig && (sp = p->sighand)) {
+		if (!get_task_struct_rcu(p))
+			return -ESRCH;
+		spin_lock_irqsave(&sp->siglock, flags);
+		if (p->sighand != sp) {
+			spin_unlock_irqrestore(&sp->siglock, flags);
+			put_task_struct(p);
+			goto retry;
+		}
 		ret = __group_send_sig_info(sig, info, p);
-		spin_unlock_irqrestore(&p->sighand->siglock, flags);
+		spin_unlock_irqrestore(&sp->siglock, flags);
+		put_task_struct(p);
 	}
 
 	return ret;
@@ -1136,14 +1158,21 @@ int
 kill_proc_info(int sig, struct siginfo *info, pid_t pid)
 {
 	int error;
+	int acquired_tasklist_lock = 0;
 	struct task_struct *p;
 
-	read_lock(&tasklist_lock);
+	rcu_read_lock();
+	if (unlikely(sig_kernel_stop(sig) || sig == SIGCONT)) {
+		read_lock(&tasklist_lock);
+		acquired_tasklist_lock = 1;
+	}
 	p = find_task_by_pid(pid);
 	error = -ESRCH;
 	if (p)
 		error = group_send_sig_info(sig, info, p);
-	read_unlock(&tasklist_lock);
+	if (unlikely(acquired_tasklist_lock))
+		read_unlock(&tasklist_lock);
+	rcu_read_unlock();
 	return error;
 }
 
@@ -1355,16 +1384,54 @@ send_sigqueue(int sig, struct sigqueue *q, struct task_struct *p)
 {
 	unsigned long flags;
 	int ret = 0;
+	struct sighand_struct *sh;
 
 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
-	read_lock(&tasklist_lock);
+
+	/*
+	 * The rcu based delayed sighand destroy makes it possible to
+	 * run this without tasklist lock held. The task struct itself
+	 * cannot go away as create_timer did get_task_struct().
+	 *
+	 * We return -1, when the task is marked exiting, so
+	 * posix_timer_event can redirect it to the group leader
+	 */
+	rcu_read_lock();
 
 	if (unlikely(p->flags & PF_EXITING)) {
 		ret = -1;
 		goto out_err;
 	}
 
-	spin_lock_irqsave(&p->sighand->siglock, flags);
+retry:
+	sh = rcu_dereference(p->sighand);
+
+	spin_lock_irqsave(&sh->siglock, flags);
+	if (p->sighand != sh) {
+		/* We raced with exec() in a multithreaded process... */
+		spin_unlock_irqrestore(&sh->siglock, flags);
+		goto retry;
+	}
+
+	/*
+	 * We do the check here again to handle the following scenario:
+	 *
+	 * CPU 0		CPU 1
+	 * send_sigqueue
+	 * check PF_EXITING
+	 * interrupt		exit code running
+	 *			__exit_signal
+	 *			lock sighand->siglock
+	 *			unlock sighand->siglock
+	 * lock sh->siglock
+	 * add(tsk->pending) 	flush_sigqueue(tsk->pending)
+	 *
+	 */
+
+	if (unlikely(p->flags & PF_EXITING)) {
+		ret = -1;
+		goto out;
+	}
 
 	if (unlikely(!list_empty(&q->list))) {
 		/*
@@ -1388,9 +1455,9 @@ send_sigqueue(int sig, struct sigqueue *q, struct task_struct *p)
 		signal_wake_up(p, sig == SIGKILL);
 
 out:
-	spin_unlock_irqrestore(&p->sighand->siglock, flags);
+	spin_unlock_irqrestore(&sh->siglock, flags);
 out_err:
-	read_unlock(&tasklist_lock);
+	rcu_read_unlock();
 
 	return ret;
 }
@@ -1402,7 +1469,9 @@ send_group_sigqueue(int sig, struct sigqueue *q, struct task_struct *p)
 	int ret = 0;
 
 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
+
 	read_lock(&tasklist_lock);
+	/* Since it_lock is held, p->sighand cannot be NULL. */
 	spin_lock_irqsave(&p->sighand->siglock, flags);
 	handle_stop_signal(sig, p);
 
@@ -1436,7 +1505,7 @@ send_group_sigqueue(int sig, struct sigqueue *q, struct task_struct *p)
 out:
 	spin_unlock_irqrestore(&p->sighand->siglock, flags);
 	read_unlock(&tasklist_lock);
-	return(ret);
+	return ret;
 }
 
 /*