ipc,sem: fine grained locking for semtimedop

Introduce finer grained locking for semtimedop, to handle the common case
of a program wanting to manipulate one semaphore from an array with
multiple semaphores.

If the call is a semop manipulating just one semaphore in an array with
multiple semaphores, only take the lock for that semaphore itself.

If the call needs to manipulate multiple semaphores, or another caller is
in a transaction that manipulates multiple semaphores, the sem_array lock
is taken, as well as all the locks for the individual semaphores.

On a 24 CPU system, performance numbers with the semop-multi
test with N threads and N semaphores, look like this:

	vanilla		Davidlohr's	Davidlohr's +	Davidlohr's +
threads			patches		rwlock patches	v3 patches
10	610652		726325		1783589		2142206
20	341570		365699		1520453		1977878
30	288102		307037		1498167		2037995
40	290714		305955		1612665		2256484
50	288620		312890		1733453		2650292
60	289987		306043		1649360		2388008
70	291298		306347		1723167		2717486
80	290948		305662		1729545		2763582
90	290996		306680		1736021		2757524
100	292243		306700		1773700		3059159

[davidlohr.bueso@hp.com: do not call sem_lock when bogus sma]
[davidlohr.bueso@hp.com: make refcounter atomic]
Signed-off-by: Rik van Riel <riel@redhat.com>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Chegu Vinod <chegu_vinod@hp.com>
Cc: Jason Low <jason.low2@hp.com>
Reviewed-by: Michel Lespinasse <walken@google.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Tested-by: Emmanuel Benisty <benisty.e@gmail.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 171 insertions, 100 deletions

diff --git a/ipc/sem.c b/ipc/sem.c
index f68b61749a85..e78ee3186d1f 100644
--- a/ipc/sem.c
+++ b/ipc/sem.c
@@ -94,6 +94,7 @@
 struct sem {
 	int	semval;		/* current value */
 	int	sempid;		/* pid of last operation */
+	spinlock_t	lock;	/* spinlock for fine-grained semtimedop */
 	struct list_head sem_pending; /* pending single-sop operations */
 };
 
@@ -137,7 +138,6 @@ struct sem_undo_list {
 
 #define sem_ids(ns)	((ns)->ids[IPC_SEM_IDS])
 
-#define sem_unlock(sma)		ipc_unlock(&(sma)->sem_perm)
 #define sem_checkid(sma, semid)	ipc_checkid(&sma->sem_perm, semid)
 
 static int newary(struct ipc_namespace *, struct ipc_params *);
@@ -190,10 +190,89 @@ void __init sem_init (void)
 }
 
 /*
+ * If the request contains only one semaphore operation, and there are
+ * no complex transactions pending, lock only the semaphore involved.
+ * Otherwise, lock the entire semaphore array, since we either have
+ * multiple semaphores in our own semops, or we need to look at
+ * semaphores from other pending complex operations.
+ *
+ * Carefully guard against sma->complex_count changing between zero
+ * and non-zero while we are spinning for the lock. The value of
+ * sma->complex_count cannot change while we are holding the lock,
+ * so sem_unlock should be fine.
+ *
+ * The global lock path checks that all the local locks have been released,
+ * checking each local lock once. This means that the local lock paths
+ * cannot start their critical sections while the global lock is held.
+ */
+static inline int sem_lock(struct sem_array *sma, struct sembuf *sops,
+			      int nsops)
+{
+	int locknum;
+ again:
+	if (nsops == 1 && !sma->complex_count) {
+		struct sem *sem = sma->sem_base + sops->sem_num;
+
+		/* Lock just the semaphore we are interested in. */
+		spin_lock(&sem->lock);
+
+		/*
+		 * If sma->complex_count was set while we were spinning,
+		 * we may need to look at things we did not lock here.
+		 */
+		if (unlikely(sma->complex_count)) {
+			spin_unlock(&sem->lock);
+			goto lock_array;
+		}
+
+		/*
+		 * Another process is holding the global lock on the
+		 * sem_array; we cannot enter our critical section,
+		 * but have to wait for the global lock to be released.
+		 */
+		if (unlikely(spin_is_locked(&sma->sem_perm.lock))) {
+			spin_unlock(&sem->lock);
+			spin_unlock_wait(&sma->sem_perm.lock);
+			goto again;
+		}
+
+		locknum = sops->sem_num;
+	} else {
+		int i;
+		/*
+		 * Lock the semaphore array, and wait for all of the
+		 * individual semaphore locks to go away.  The code
+		 * above ensures no new single-lock holders will enter
+		 * their critical section while the array lock is held.
+		 */
+ lock_array:
+		spin_lock(&sma->sem_perm.lock);
+		for (i = 0; i < sma->sem_nsems; i++) {
+			struct sem *sem = sma->sem_base + i;
+			spin_unlock_wait(&sem->lock);
+		}
+		locknum = -1;
+	}
+	return locknum;
+}
+
+static inline void sem_unlock(struct sem_array *sma, int locknum)
+{
+	if (locknum == -1) {
+		spin_unlock(&sma->sem_perm.lock);
+	} else {
+		struct sem *sem = sma->sem_base + locknum;
+		spin_unlock(&sem->lock);
+	}
+	rcu_read_unlock();
+}
+
+/*
  * sem_lock_(check_) routines are called in the paths where the rw_mutex
  * is not held.
  */
-static inline struct sem_array *sem_obtain_lock(struct ipc_namespace *ns, int id)
+static inline struct sem_array *sem_obtain_lock(struct ipc_namespace *ns,
+			int id, struct sembuf *sops, int nsops, int *locknum)
 {
 	struct kern_ipc_perm *ipcp;
 	struct sem_array *sma;
@@ -205,7 +284,8 @@ static inline struct sem_array *sem_obtain_lock(struct ipc_namespace *ns, int id
 		goto err;
 	}
 
-	spin_lock(&ipcp->lock);
+	sma = container_of(ipcp, struct sem_array, sem_perm);
+	*locknum = sem_lock(sma, sops, nsops);
 
 	/* ipc_rmid() may have already freed the ID while sem_lock
 	 * was spinning: verify that the structure is still valid
@@ -213,7 +293,7 @@ static inline struct sem_array *sem_obtain_lock(struct ipc_namespace *ns, int id
 	if (!ipcp->deleted)
 		return container_of(ipcp, struct sem_array, sem_perm);
 
-	spin_unlock(&ipcp->lock);
+	sem_unlock(sma, *locknum);
 	sma = ERR_PTR(-EINVAL);
 err:
 	rcu_read_unlock();
@@ -230,17 +310,6 @@ static inline struct sem_array *sem_obtain_object(struct ipc_namespace *ns, int
 	return container_of(ipcp, struct sem_array, sem_perm);
 }
 
-static inline struct sem_array *sem_lock_check(struct ipc_namespace *ns,
-						int id)
-{
-	struct kern_ipc_perm *ipcp = ipc_lock_check(&sem_ids(ns), id);
-
-	if (IS_ERR(ipcp))
-		return ERR_CAST(ipcp);
-
-	return container_of(ipcp, struct sem_array, sem_perm);
-}
-
 static inline struct sem_array *sem_obtain_object_check(struct ipc_namespace *ns,
 							int id)
 {
@@ -254,21 +323,21 @@ static inline struct sem_array *sem_obtain_object_check(struct ipc_namespace *ns
 
 static inline void sem_lock_and_putref(struct sem_array *sma)
 {
-	ipc_lock_by_ptr(&sma->sem_perm);
+	rcu_read_lock();
+	sem_lock(sma, NULL, -1);
 	ipc_rcu_putref(sma);
 }
 
 static inline void sem_getref_and_unlock(struct sem_array *sma)
 {
-	ipc_rcu_getref(sma);
-	ipc_unlock(&(sma)->sem_perm);
+	WARN_ON_ONCE(!ipc_rcu_getref(sma));
+	sem_unlock(sma, -1);
 }
 
 static inline void sem_putref(struct sem_array *sma)
 {
-	ipc_lock_by_ptr(&sma->sem_perm);
-	ipc_rcu_putref(sma);
-	ipc_unlock(&(sma)->sem_perm);
+	sem_lock_and_putref(sma);
+	sem_unlock(sma, -1);
 }
 
 /*
@@ -276,9 +345,9 @@ static inline void sem_putref(struct sem_array *sma)
  */
 static inline void sem_getref(struct sem_array *sma)
 {
-	spin_lock(&(sma)->sem_perm.lock);
-	ipc_rcu_getref(sma);
-	ipc_unlock(&(sma)->sem_perm);
+	sem_lock(sma, NULL, -1);
+	WARN_ON_ONCE(!ipc_rcu_getref(sma));
+	sem_unlock(sma, -1);
 }
 
 static inline void sem_rmid(struct ipc_namespace *ns, struct sem_array *s)
@@ -371,15 +440,17 @@ static int newary(struct ipc_namespace *ns, struct ipc_params *params)
 
 	sma->sem_base = (struct sem *) &sma[1];
 
-	for (i = 0; i < nsems; i++)
+	for (i = 0; i < nsems; i++) {
 		INIT_LIST_HEAD(&sma->sem_base[i].sem_pending);
+		spin_lock_init(&sma->sem_base[i].lock);
+	}
 
 	sma->complex_count = 0;
 	INIT_LIST_HEAD(&sma->sem_pending);
 	INIT_LIST_HEAD(&sma->list_id);
 	sma->sem_nsems = nsems;
 	sma->sem_ctime = get_seconds();
-	sem_unlock(sma);
+	sem_unlock(sma, -1);
 
 	return sma->sem_perm.id;
 }
@@ -818,7 +889,7 @@ static void freeary(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
 
 	/* Remove the semaphore set from the IDR */
 	sem_rmid(ns, sma);
-	sem_unlock(sma);
+	sem_unlock(sma, -1);
 
 	wake_up_sem_queue_do(&tasks);
 	ns->used_sems -= sma->sem_nsems;
@@ -947,7 +1018,6 @@ static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum,
 	struct sem_array *sma;
 	struct sem* curr;
 	int err;
-	int nsems;
 	struct list_head tasks;
 	int val;
 #if defined(CONFIG_64BIT) && defined(__BIG_ENDIAN)
@@ -958,31 +1028,39 @@ static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum,
 	val = arg;
 #endif
 
-	sma = sem_lock_check(ns, semid);
-	if (IS_ERR(sma))
-		return PTR_ERR(sma);
+	if (val > SEMVMX || val < 0)
+		return -ERANGE;
 
 	INIT_LIST_HEAD(&tasks);
-	nsems = sma->sem_nsems;
 
-	err = -EACCES;
-	if (ipcperms(ns, &sma->sem_perm, S_IWUGO))
-		goto out_unlock;
+	rcu_read_lock();
+	sma = sem_obtain_object_check(ns, semid);
+	if (IS_ERR(sma)) {
+		rcu_read_unlock();
+		return PTR_ERR(sma);
+	}
+
+	if (semnum < 0 || semnum >= sma->sem_nsems) {
+		rcu_read_unlock();
+		return -EINVAL;
+	}
+
+
+	if (ipcperms(ns, &sma->sem_perm, S_IWUGO)) {
+		rcu_read_unlock();
+		return -EACCES;
+	}
 
 	err = security_sem_semctl(sma, SETVAL);
-	if (err)
-		goto out_unlock;
+	if (err) {
+		rcu_read_unlock();
+		return -EACCES;
+	}
 
-	err = -EINVAL;
-	if(semnum < 0 || semnum >= nsems)
-		goto out_unlock;
+	sem_lock(sma, NULL, -1);
 
 	curr = &sma->sem_base[semnum];
 
-	err = -ERANGE;
-	if (val > SEMVMX || val < 0)
-		goto out_unlock;
-
 	assert_spin_locked(&sma->sem_perm.lock);
 	list_for_each_entry(un, &sma->list_id, list_id)
 		un->semadj[semnum] = 0;
@@ -992,11 +1070,9 @@ static int semctl_setval(struct ipc_namespace *ns, int semid, int semnum,
 	sma->sem_ctime = get_seconds();
 	/* maybe some queued-up processes were waiting for this */
 	do_smart_update(sma, NULL, 0, 0, &tasks);
-	err = 0;
-out_unlock:
-	sem_unlock(sma);
+	sem_unlock(sma, -1);
 	wake_up_sem_queue_do(&tasks);
-	return err;
+	return 0;
 }
 
 static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
@@ -1051,16 +1127,16 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
 
 			sem_lock_and_putref(sma);
 			if (sma->sem_perm.deleted) {
-				sem_unlock(sma);
+				sem_unlock(sma, -1);
 				err = -EIDRM;
 				goto out_free;
 			}
-		}
+		} else
+			sem_lock(sma, NULL, -1);
 
-		spin_lock(&sma->sem_perm.lock);
 		for (i = 0; i < sma->sem_nsems; i++)
 			sem_io[i] = sma->sem_base[i].semval;
-		sem_unlock(sma);
+		sem_unlock(sma, -1);
 		err = 0;
 		if(copy_to_user(array, sem_io, nsems*sizeof(ushort)))
 			err = -EFAULT;
@@ -1071,7 +1147,10 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
 		int i;
 		struct sem_undo *un;
 
-		ipc_rcu_getref(sma);
+		if (!ipc_rcu_getref(sma)) {
+			rcu_read_unlock();
+			return -EIDRM;
+		}
 		rcu_read_unlock();
 
 		if(nsems > SEMMSL_FAST) {
@@ -1097,7 +1176,7 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
 		}
 		sem_lock_and_putref(sma);
 		if (sma->sem_perm.deleted) {
-			sem_unlock(sma);
+			sem_unlock(sma, -1);
 			err = -EIDRM;
 			goto out_free;
 		}
@@ -1124,7 +1203,7 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
 		goto out_wakeup;
 	}
 
-	spin_lock(&sma->sem_perm.lock);
+	sem_lock(sma, NULL, -1);
 	curr = &sma->sem_base[semnum];
 
 	switch (cmd) {
@@ -1143,7 +1222,7 @@ static int semctl_main(struct ipc_namespace *ns, int semid, int semnum,
 	}
 
 out_unlock:
-	sem_unlock(sma);
+	sem_unlock(sma, -1);
 out_wakeup:
 	wake_up_sem_queue_do(&tasks);
 out_free:
@@ -1211,11 +1290,11 @@ static int semctl_down(struct ipc_namespace *ns, int semid,
 
 	switch(cmd){
 	case IPC_RMID:
-		ipc_lock_object(&sma->sem_perm);
+		sem_lock(sma, NULL, -1);
 		freeary(ns, ipcp);
 		goto out_up;
 	case IPC_SET:
-		ipc_lock_object(&sma->sem_perm);
+		sem_lock(sma, NULL, -1);
 		err = ipc_update_perm(&semid64.sem_perm, ipcp);
 		if (err)
 			goto out_unlock;
@@ -1228,7 +1307,7 @@ static int semctl_down(struct ipc_namespace *ns, int semid,
 	}
 
 out_unlock:
-	sem_unlock(sma);
+	sem_unlock(sma, -1);
 out_up:
 	up_write(&sem_ids(ns).rw_mutex);
 	return err;
@@ -1340,8 +1419,7 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
 	struct sem_array *sma;
 	struct sem_undo_list *ulp;
 	struct sem_undo *un, *new;
-	int nsems;
-	int error;
+	int nsems, error;
 
 	error = get_undo_list(&ulp);
 	if (error)
@@ -1363,7 +1441,11 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
 	}
 
 	nsems = sma->sem_nsems;
-	ipc_rcu_getref(sma);
+	if (!ipc_rcu_getref(sma)) {
+		rcu_read_unlock();
+		un = ERR_PTR(-EIDRM);
+		goto out;
+	}
 	rcu_read_unlock();
 
 	/* step 2: allocate new undo structure */
@@ -1376,7 +1458,7 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
 	/* step 3: Acquire the lock on semaphore array */
 	sem_lock_and_putref(sma);
 	if (sma->sem_perm.deleted) {
-		sem_unlock(sma);
+		sem_unlock(sma, -1);
 		kfree(new);
 		un = ERR_PTR(-EIDRM);
 		goto out;
@@ -1404,7 +1486,7 @@ static struct sem_undo *find_alloc_undo(struct ipc_namespace *ns, int semid)
 success:
 	spin_unlock(&ulp->lock);
 	rcu_read_lock();
-	sem_unlock(sma);
+	sem_unlock(sma, -1);
 out:
 	return un;
 }
@@ -1444,7 +1526,7 @@ SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
 	struct sembuf fast_sops[SEMOPM_FAST];
 	struct sembuf* sops = fast_sops, *sop;
 	struct sem_undo *un;
-	int undos = 0, alter = 0, max;
+	int undos = 0, alter = 0, max, locknum;
 	struct sem_queue queue;
 	unsigned long jiffies_left = 0;
 	struct ipc_namespace *ns;
@@ -1488,22 +1570,23 @@ SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
 			alter = 1;
 	}
 
+	INIT_LIST_HEAD(&tasks);
+
 	if (undos) {
+		/* On success, find_alloc_undo takes the rcu_read_lock */
 		un = find_alloc_undo(ns, semid);
 		if (IS_ERR(un)) {
 			error = PTR_ERR(un);
 			goto out_free;
 		}
-	} else
+	} else {
 		un = NULL;
+		rcu_read_lock();
+	}
 
-	INIT_LIST_HEAD(&tasks);
-
-	rcu_read_lock();
 	sma = sem_obtain_object_check(ns, semid);
 	if (IS_ERR(sma)) {
-		if (un)
-			rcu_read_unlock();
+		rcu_read_unlock();
 		error = PTR_ERR(sma);
 		goto out_free;
 	}
@@ -1534,23 +1617,9 @@ SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
 	 * "un" itself is guaranteed by rcu.
 	 */
 	error = -EIDRM;
-	ipc_lock_object(&sma->sem_perm);
-	if (un) {
-		if (un->semid == -1) {
-			rcu_read_unlock();
-			goto out_unlock_free;
-		} else {
-			/*
-			 * rcu lock can be released, "un" cannot disappear:
-			 * - sem_lock is acquired, thus IPC_RMID is
-			 *   impossible.
-			 * - exit_sem is impossible, it always operates on
-			 *   current (or a dead task).
-			 */
-
-			rcu_read_unlock();
-		}
-	}
+	locknum = sem_lock(sma, sops, nsops);
+	if (un && un->semid == -1)
+		goto out_unlock_free;
 
 	error = try_atomic_semop (sma, sops, nsops, un, task_tgid_vnr(current));
 	if (error <= 0) {
@@ -1591,7 +1660,7 @@ SYSCALL_DEFINE4(semtimedop, int, semid, struct sembuf __user *, tsops,
 
 sleep_again:
 	current->state = TASK_INTERRUPTIBLE;
-	sem_unlock(sma);
+	sem_unlock(sma, locknum);
 
 	if (timeout)
 		jiffies_left = schedule_timeout(jiffies_left);
@@ -1613,7 +1682,7 @@ sleep_again:
 		goto out_free;
 	}
 
-	sma = sem_obtain_lock(ns, semid);
+	sma = sem_obtain_lock(ns, semid, sops, nsops, &locknum);
 
 	/*
 	 * Wait until it's guaranteed that no wakeup_sem_queue_do() is ongoing.
@@ -1652,7 +1721,7 @@ sleep_again:
 	unlink_queue(sma, &queue);
 
 out_unlock_free:
-	sem_unlock(sma);
+	sem_unlock(sma, locknum);
 out_wakeup:
 	wake_up_sem_queue_do(&tasks);
 out_free:
@@ -1716,8 +1785,7 @@ void exit_sem(struct task_struct *tsk)
 		struct sem_array *sma;
 		struct sem_undo *un;
 		struct list_head tasks;
-		int semid;
-		int i;
+		int semid, i;
 
 		rcu_read_lock();
 		un = list_entry_rcu(ulp->list_proc.next,
@@ -1726,23 +1794,26 @@ void exit_sem(struct task_struct *tsk)
 			semid = -1;
 		 else
 			semid = un->semid;
-		rcu_read_unlock();
 
-		if (semid == -1)
+		if (semid == -1) {
+			rcu_read_unlock();
 			break;
+		}
 
-		sma = sem_lock_check(tsk->nsproxy->ipc_ns, un->semid);
-
+		sma = sem_obtain_object_check(tsk->nsproxy->ipc_ns, un->semid);
 		/* exit_sem raced with IPC_RMID, nothing to do */
-		if (IS_ERR(sma))
+		if (IS_ERR(sma)) {
+			rcu_read_unlock();
 			continue;
+		}
 
+		sem_lock(sma, NULL, -1);
 		un = __lookup_undo(ulp, semid);
 		if (un == NULL) {
 			/* exit_sem raced with IPC_RMID+semget() that created
 			 * exactly the same semid. Nothing to do.
 			 */
-			sem_unlock(sma);
+			sem_unlock(sma, -1);
 			continue;
 		}
 
@@ -1782,7 +1853,7 @@ void exit_sem(struct task_struct *tsk)
 		/* maybe some queued-up processes were waiting for this */
 		INIT_LIST_HEAD(&tasks);
 		do_smart_update(sma, NULL, 0, 1, &tasks);
-		sem_unlock(sma);
+		sem_unlock(sma, -1);
 		wake_up_sem_queue_do(&tasks);
 
 		kfree_rcu(un, rcu);