[PATCH] support "requeueing" futexes

This addresses a futex related SMP scalability problem of
glibc. A number of regressions have been reported to the NTPL mailing list
when going to many CPUs, for applications that use condition variables and
the pthread_cond_broadcast() API call. Using this functionality, testcode
shows a slowdown from 0.12 seconds runtime to over 237 seconds (!)
runtime, on 4-CPU systems.

pthread condition variables use two futex-backed mutex-alike locks: an
internal one for the glibc CV state itself, and a user-supplied mutex
which the API guarantees to take in certain codepaths. (Unfortunately the
user-supplied mutex cannot be used to protect the CV state, so we've got
to deal with two locks.)

The cause of the slowdown is a 'swarm effect': if lots of threads are
blocked on a condition variable, and pthread_cond_broadcast() is done,
then glibc first does a FUTEX_WAKE on the cv-internal mutex, then down a
mutex_down() on the user-supplied mutex. Ie. a swarm of threads is created
which all race to serialize on the user-supplied mutex. The more threads
are used, the more likely it becomes that the scheduler will balance them
over to other CPUs - where they just schedule, try to lock the mutex, and
go to sleep. This 'swarm effect' is purely technical, a side-effect of
glibc's use of futexes, and the imperfect coupling of the two locks.

the solution to this problem is to not wake up the swarm of threads, but
'requeue' them from the CV-internal mutex to the user-supplied mutex. The
attached patch adds the FUTEX_REQUEUE feature FUTEX_REQUEUE requeues N
threads from futex address A to futex address B.

This way glibc can wake up a single thread (which will take the
user-mutex), and can requeue the rest, with a single system-call.

Ulrich Drepper has implemented FUTEX_REQUEUE support in glibc, and a
number of people have tested it over the past couple of weeks. Here are
the measurements done by Saurabh Desai:

System: 4xPIII 700MHz

 ./cond-perf -r 100 -n 200:        1p       2p         4p
 Default NPTL:                 0.120s   0.211s   237.407s
 requeue NPTL:                 0.124s   0.156s     0.040s

 ./cond-perf -r 1000 -n 100:
 Default NPTL:                 0.276s   0.412s     0.530s
 requeue NPTL:                 0.349s   0.503s     0.550s

 ./pp -v -n 128 -i 1000 -S 32768:
 Default NPTL: 128 games in    1.111s   1.270s    16.894s
 requeue NPTL: 128 games in    1.111s   1.959s     2.426s

 ./pp -v -n 1024 -i 10 -S 32768:
 Default NPTL: 1024 games in   0.181s   0.394s     incompleted 2m+
 requeue NPTL: 1024 games in   0.166s   0.254s     0.341s

the speedup with increasing number of threads is quite significant, in the
128 threads, case it's more than 8 times. In the cond-perf test, on 4 CPUs
it's almost infinitely faster than the 'swarm of threads' catastrophy
triggered by the old code.

include/linux/futex.h

@@ -2,10 +2,16 @@
 #define _LINUX_FUTEX_H
 
 /* Second argument to futex syscall */
+
+
 #define FUTEX_WAIT (0)
 #define FUTEX_WAKE (1)
 #define FUTEX_FD (2)
+#define FUTEX_REQUEUE (3)
+
+
+asmlinkage long sys_futex(u32 __user *uaddr, int op, int val,
+			  struct timespec __user *utime, u32 __user *uaddr2);
 
-extern asmlinkage long sys_futex(u32 __user *uaddr, int op, int val, struct timespec __user *utime);
 
 #endif

kernel/compat.c

@@ -211,20 +211,25 @@ asmlinkage long compat_sys_sigprocmask(int how, compat_old_sigset_t *set,
 	return ret;
 }
 
-extern long do_futex(unsigned long, int, int, unsigned long);
+extern long do_futex(unsigned long uaddr, int op, int val,
+			unsigned long timeout, unsigned long uaddr2, int val2);
 
 asmlinkage long compat_sys_futex(u32 *uaddr, int op, int val,
 		struct compat_timespec *utime)
 {
 	struct timespec t;
 	unsigned long timeout = MAX_SCHEDULE_TIMEOUT;
+	int val2 = 0;
 
 	if ((op == FUTEX_WAIT) && utime) {
 		if (get_compat_timespec(&t, utime))
 			return -EFAULT;
 		timeout = timespec_to_jiffies(&t) + 1;
 	}
-	return do_futex((unsigned long)uaddr, op, val, timeout);
+	if (op == FUTEX_REQUEUE)
+		val2 = (int) utime;
+
+	return do_futex((unsigned long)uaddr, op, val, timeout, uaddr2, val2);
 }
 
 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit *rlim);

kernel/fork.c

@@ -457,7 +457,7 @@ void mm_release(struct task_struct *tsk, struct mm_struct *mm)
 		 * not set up a proper pointer then tough luck.
 		 */
 		put_user(0, tidptr);
-		sys_futex(tidptr, FUTEX_WAKE, 1, NULL);
+		sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL);
 	}
 }
 

kernel/futex.c

@@ -2,6 +2,9 @@
  *  Fast Userspace Mutexes (which I call "Futexes!").
  *  (C) Rusty Russell, IBM 2002
  *
+ *  Generalized futexes, futex requeueing, misc fixes by Ingo Molnar
+ *  (C) Copyright 2003 Red Hat Inc, All Rights Reserved
+ *
  *  Thanks to Ben LaHaise for yelling "hashed waitqueues" loudly
  *  enough at me, Linus for the original (flawed) idea, Matthew
  *  Kirkwood for proof-of-concept implementation.
@@ -9,9 +12,6 @@
  *  "The futexes are also cursed."
  *  "But they come in a choice of three flavours!"
  *
- *  Generalized futexes for every mapping type, Ingo Molnar, 2002
- *
- *
  *  This program is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
  *  the Free Software Foundation; either version 2 of the License, or
@@ -93,19 +93,18 @@ static inline struct list_head *hash_futex(struct page *page, int offset)
 							FUTEX_HASHBITS)];
 }
 
-/* Waiter either waiting in FUTEX_WAIT or poll(), or expecting signal */
-static inline void tell_waiter(struct futex_q *q)
-{
-	wake_up_all(&q->waiters);
-	if (q->filp)
-		send_sigio(&q->filp->f_owner, q->fd, POLL_IN);
-}
-
 /*
  * Get kernel address of the user page and pin it.
  *
  * Must be called with (and returns with) all futex-MM locks held.
  */
+static inline struct page *__pin_page_atomic (struct page *page)
+{
+	if (!PageReserved(page))
+		get_page(page);
+	return page;
+}
+
 static struct page *__pin_page(unsigned long addr)
 {
 	struct mm_struct *mm = current->mm;
@@ -116,11 +115,8 @@ static struct page *__pin_page(unsigned long addr)
 	 * Do a quick atomic lookup first - this is the fastpath.
 	 */
 	page = follow_page(mm, addr, 0);
-	if (likely(page != NULL)) {	
-		if (!PageReserved(page))
-			get_page(page);
-		return page;
-	}
+	if (likely(page != NULL))
+		return __pin_page_atomic(page);
 
 	/*
 	 * No luck - need to fault in the page:
@@ -150,16 +146,11 @@ static struct page *__pin_page(unsigned long addr)
 	return page;
 }
 
-static inline void unpin_page(struct page *page)
-{
-	put_page(page);
-}
-
 /*
  * Wake up all waiters hashed on the physical page that is mapped
  * to this virtual address:
  */
-static int futex_wake(unsigned long uaddr, int offset, int num)
+static inline int futex_wake(unsigned long uaddr, int offset, int num)
 {
 	struct list_head *i, *next, *head;
 	struct page *page;
@@ -181,7 +172,9 @@ static int futex_wake(unsigned long uaddr, int offset, int num)
 		if (this->page == page && this->offset == offset) {
 			list_del_init(i);
 			__detach_vcache(&this->vcache);
-			tell_waiter(this);
+			wake_up_all(&this->waiters);
+			if (this->filp)
+				send_sigio(&this->filp->f_owner, this->fd, POLL_IN);
 			ret++;
 			if (ret >= num)
 				break;
@@ -189,7 +182,7 @@ static int futex_wake(unsigned long uaddr, int offset, int num)
 	}
 
 	unlock_futex_mm();
-	unpin_page(page);
+	put_page(page);
 
 	return ret;
 }
@@ -208,7 +201,9 @@ static void futex_vcache_callback(vcache_t *vcache, struct page *new_page)
 	spin_lock(&futex_lock);
 
 	if (!list_empty(&q->list)) {
+		put_page(q->page);
 		q->page = new_page;
+		__pin_page_atomic(new_page);
 		list_del(&q->list);
 		list_add_tail(&q->list, head);
 	}
@@ -216,6 +211,65 @@ static void futex_vcache_callback(vcache_t *vcache, struct page *new_page)
 	spin_unlock(&futex_lock);
 }
 
+/*
+ * Requeue all waiters hashed on one physical page to another
+ * physical page.
+ */
+static inline int futex_requeue(unsigned long uaddr1, int offset1,
+	unsigned long uaddr2, int offset2, int nr_wake, int nr_requeue)
+{
+	struct list_head *i, *next, *head1, *head2;
+	struct page *page1 = NULL, *page2 = NULL;
+	int ret = 0;
+
+	lock_futex_mm();
+
+	page1 = __pin_page(uaddr1 - offset1);
+	if (!page1)
+		goto out;
+	page2 = __pin_page(uaddr2 - offset2);
+	if (!page2)
+		goto out;
+
+	head1 = hash_futex(page1, offset1);
+	head2 = hash_futex(page2, offset2);
+
+	list_for_each_safe(i, next, head1) {
+		struct futex_q *this = list_entry(i, struct futex_q, list);
+
+		if (this->page == page1 && this->offset == offset1) {
+			list_del_init(i);
+			__detach_vcache(&this->vcache);
+			if (++ret <= nr_wake) {
+				wake_up_all(&this->waiters);
+				if (this->filp)
+					send_sigio(&this->filp->f_owner,
+							this->fd, POLL_IN);
+			} else {
+				put_page(this->page);
+				__pin_page_atomic (page2);
+				list_add_tail(i, head2);
+				__attach_vcache(&this->vcache, uaddr2,
+					current->mm, futex_vcache_callback);
+				this->offset = offset2;
+				this->page = page2;
+				if (ret - nr_wake >= nr_requeue)
+					break;
+			}
+		}
+	}
+
+out:
+	unlock_futex_mm();
+
+	if (page1)
+		put_page(page1);
+	if (page2)
+		put_page(page2);
+
+	return ret;
+}
+
 static inline void __queue_me(struct futex_q *q, struct page *page,
 				unsigned long uaddr, int offset,
 				int fd, struct file *filp)
@@ -252,7 +306,7 @@ static inline int unqueue_me(struct futex_q *q)
 	return ret;
 }
 
-static int futex_wait(unsigned long uaddr,
+static inline int futex_wait(unsigned long uaddr,
 		      int offset,
 		      int val,
 		      unsigned long time)
@@ -273,14 +327,17 @@ static int futex_wait(unsigned long uaddr,
 	}
 	__queue_me(&q, page, uaddr, offset, -1, NULL);
 
-	unlock_futex_mm();
-
-	/* Page is pinned, but may no longer be in this address space. */
+	/*
+	 * Page is pinned, but may no longer be in this address space.
+	 * It cannot schedule, so we access it with the spinlock held.
+	 */
 	if (get_user(curval, (int *)uaddr) != 0) {
+		unlock_futex_mm();
 		ret = -EFAULT;
 		goto out;
 	}
 	if (curval != val) {
+		unlock_futex_mm();
 		ret = -EWOULDBLOCK;
 		goto out;
 	}
@@ -288,13 +345,15 @@ static int futex_wait(unsigned long uaddr,
 	 * The get_user() above might fault and schedule so we
 	 * cannot just set TASK_INTERRUPTIBLE state when queueing
 	 * ourselves into the futex hash. This code thus has to
-	 * rely on the FUTEX_WAKE code doing a wakeup after removing
+	 * rely on the futex_wake() code doing a wakeup after removing
 	 * the waiter from the list.
 	 */
 	add_wait_queue(&q.waiters, &wait);
 	set_current_state(TASK_INTERRUPTIBLE);
-	if (!list_empty(&q.list))
+	if (!list_empty(&q.list)) {
+		unlock_futex_mm();
 		time = schedule_timeout(time);
+	}
 	set_current_state(TASK_RUNNING);
 	/*
 	 * NOTE: we don't remove ourselves from the waitqueue because
@@ -310,7 +369,7 @@ static int futex_wait(unsigned long uaddr,
 	/* Were we woken up anyway? */
 	if (!unqueue_me(&q))
 		ret = 0;
-	unpin_page(page);
+	put_page(q.page);
 
 	return ret;
 }
@@ -320,7 +379,7 @@ static int futex_close(struct inode *inode, struct file *filp)
 	struct futex_q *q = filp->private_data;
 
 	unqueue_me(q);
-	unpin_page(q->page);
+	put_page(q->page);
 	kfree(filp->private_data);
 	return 0;
 }
@@ -416,11 +475,12 @@ static int futex_fd(unsigned long uaddr, int offset, int signal)
 	page = NULL;
 out:
 	if (page)
-		unpin_page(page);
+		put_page(page);
 	return ret;
 }
 
-long do_futex(unsigned long uaddr, int op, int val, unsigned long timeout)
+long do_futex(unsigned long uaddr, int op, int val, unsigned long timeout,
+		unsigned long uaddr2, int val2)
 {
 	unsigned long pos_in_page;
 	int ret;
@@ -442,23 +502,45 @@ long do_futex(unsigned long uaddr, int op, int val, unsigned long timeout)
 		/* non-zero val means F_SETOWN(getpid()) & F_SETSIG(val) */
 		ret = futex_fd(uaddr, pos_in_page, val);
 		break;
+	case FUTEX_REQUEUE:
+	{
+		unsigned long pos_in_page2 = uaddr2 % PAGE_SIZE;
+
+		/* Must be "naturally" aligned */
+		if (pos_in_page2 % sizeof(u32))
+			return -EINVAL;
+
+		ret = futex_requeue(uaddr, pos_in_page, uaddr2, pos_in_page2,
+				    val, val2);
+		break;
+	}
 	default:
-		ret = -EINVAL;
+		ret = -ENOSYS;
 	}
 	return ret;
 }
 
-asmlinkage long sys_futex(u32 __user *uaddr, int op, int val, struct timespec __user *utime)
+
+asmlinkage long sys_futex(u32 __user *uaddr, int op, int val,
+			  struct timespec __user *utime, u32 __user *uaddr2)
 {
 	struct timespec t;
 	unsigned long timeout = MAX_SCHEDULE_TIMEOUT;
+	int val2 = 0;
 
 	if ((op == FUTEX_WAIT) && utime) {
 		if (copy_from_user(&t, utime, sizeof(t)) != 0)
 			return -EFAULT;
 		timeout = timespec_to_jiffies(&t) + 1;
 	}
-	return do_futex((unsigned long)uaddr, op, val, timeout);
+	/*
+	 * requeue parameter in 'utime' if op == FUTEX_REQUEUE.
+	 */
+	if (op == FUTEX_REQUEUE)
+		val2 = (int) utime;
+
+	return do_futex((unsigned long)uaddr, op, val, timeout,
+			(unsigned long)uaddr2, val2);
 }
 
 static struct super_block *