412fe2de4b38f88f5ef1d59645720acbaf2af4750c150cc78901c0b9.patch

From b1619fa897b35d45477bfd38e52d93f3bdd2bf11 Mon Sep 17 00:00:00 2001
From: Peter Oskolkov <posk@google.com>
Date: Wed, 22 Jul 2020 16:45:36 -0700
Subject: [PATCH] FROMLIST: futex: introduce FUTEX_SWAP operation

As Paul Turner presented at LPC in 2013 ...
- pdf: http://pdxplumbers.osuosl.org/2013/ocw//system/presentations/1653/original/LPC%20-%20User%20Threading.pdf
- video: https://www.youtube.com/watch?v=KXuZi9aeGTw

... Google has developed an M:N userspace threading subsystem backed
by Google-private SwitchTo Linux Kernel API (page 17 in the pdf referenced
above). This subsystem provides latency-sensitive services at Google with
fine-grained user-space control/scheduling over what is running when,
and this subsystem is used widely internally (called schedulers or fibers).

This patchset is the first step to open-source this work. As explained
in the linked pdf and video, SwitchTo API has three core operations: wait,
resume, and swap (=switch). So this patchset adds a FUTEX_SWAP operation
that, in addition to FUTEX_WAIT and FUTEX_WAKE, will provide a foundation
on top of which user-space threading libraries can be built.

Another common use case for FUTEX_SWAP is message passing a-la RPC
between tasks: task/thread T1 prepares a message,
wakes T2 to work on it, and waits for the results; when T2 is done, it
wakes T1 and waits for more work to arrive. Currently the simplest
way to implement this is

a. T1: futex-wake T2, futex-wait
b. T2: wakes, does what it has been woken to do
c. T2: futex-wake T1, futex-wait

With FUTEX_SWAP, steps a and c above can be reduced to one futex operation
that runs 5-10 times faster.

Patches in this patchset:

Patch 1: (this patch) introduce FUTEX_SWAP futex operation that,
         internally, does wake + wait. The purpose of this patch is
         to work out the API.
Patch 2: a first rough attempt to make FUTEX_SWAP faster than
         what wake + wait can do.
Patch 3: a selftest that can also be used to benchmark FUTEX_SWAP vs
         FUTEX_WAKE + FUTEX_WAIT.

Tested: see patch 3 in this patchset.

Signed-off-by: Peter Oskolkov <posk@google.com>
(am from https://lore.kernel.org/patchwork/patch/1277905/)
(also found at https://lore.kernel.org/r/20200722234538.166697-2-posk@posk.io)

BUG=b:170967073
TEST=eve-kernelnext and manually verification with selftests

Change-Id: I0f693816689477a5f7252ff361df4dd96ef4af88
Signed-off-by: Guenter Roeck <groeck@chromium.org>
---
 include/uapi/linux/futex.h |  2 +
 kernel/futex.c             | 89 +++++++++++++++++++++++++++++++-------
 2 files changed, 76 insertions(+), 15 deletions(-)

diff --git a/include/uapi/linux/futex.h b/include/uapi/linux/futex.h
index 235e5b2facaa..6d0fb8e032d8 100644
--- a/include/uapi/linux/futex.h
+++ b/include/uapi/linux/futex.h
@@ -22,6 +22,7 @@
 #define FUTEX_WAIT_REQUEUE_PI	11
 #define FUTEX_CMP_REQUEUE_PI	12
 #define FUTEX_LOCK_PI2		13
+#define FUTEX_SWAP		14
 
 #define FUTEX_PRIVATE_FLAG	128
 #define FUTEX_CLOCK_REALTIME	256
@@ -42,6 +43,7 @@
 					 FUTEX_PRIVATE_FLAG)
 #define FUTEX_CMP_REQUEUE_PI_PRIVATE	(FUTEX_CMP_REQUEUE_PI | \
 					 FUTEX_PRIVATE_FLAG)
+#define FUTEX_SWAP_PRIVATE		(FUTEX_SWAP | FUTEX_PRIVATE_FLAG)
 
 /*
  * Support for robust futexes: the kernel cleans up held futexes at
diff --git a/kernel/futex.c b/kernel/futex.c
index 2ecb07575055..006188c09afe 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -1180,7 +1180,7 @@ static int handle_exit_race(u32 __user *uaddr, u32 uval,
 	 *  tsk->futex_state =               } else {
 	 *	FUTEX_STATE_DEAD;              if (tsk->futex_state !=
 	 *					  FUTEX_STATE_DEAD)
-	 *				         return -EAGAIN;
+	 *					 return -EAGAIN;
 	 *				       return -ESRCH; <--- FAIL
 	 *				     }
 	 *
@@ -1586,16 +1586,16 @@ double_unlock_hb(struct futex_hash_bucket *hb1, struct futex_hash_bucket *hb2)
 }
 
 /*
- * Wake up waiters matching bitset queued on this futex (uaddr).
+ * Prepare wake queue matching bitset queued on this futex (uaddr).
  */
 static int
-futex_wake(u32 __user *uaddr, unsigned int flags, int nr_wake, u32 bitset)
+prepare_wake_q(u32 __user *uaddr, unsigned int flags, int nr_wake, u32 bitset,
+	       struct wake_q_head *wake_q)
 {
 	struct futex_hash_bucket *hb;
 	struct futex_q *this, *next;
 	union futex_key key = FUTEX_KEY_INIT;
 	int ret;
-	DEFINE_WAKE_Q(wake_q);
 
 	if (!bitset)
 		return -EINVAL;
@@ -1623,14 +1623,28 @@ futex_wake(u32 __user *uaddr, unsigned int flags, int nr_wake, u32 bitset)
 			if (!(this->bitset & bitset))
 				continue;
 
-			mark_wake_futex(&wake_q, this);
+			mark_wake_futex(wake_q, this);
 			if (++ret >= nr_wake)
 				break;
 		}
 	}
 
 	spin_unlock(&hb->lock);
+	return ret;
+}
+
+/*
+ * Wake up waiters matching bitset queued on this futex (uaddr).
+ */
+static int
+futex_wake(u32 __user *uaddr, unsigned int flags, int nr_wake, u32 bitset)
+{
+	int ret;
+	DEFINE_WAKE_Q(wake_q);
+
+	ret = prepare_wake_q(uaddr, flags, nr_wake, bitset, &wake_q);
 	wake_up_q(&wake_q);
+
 	return ret;
 }
 
@@ -2571,9 +2585,12 @@ static int fixup_owner(u32 __user *uaddr, struct futex_q *q, int locked)
  * @hb:		the futex hash bucket, must be locked by the caller
  * @q:		the futex_q to queue up on
  * @timeout:	the prepared hrtimer_sleeper, or null for no timeout
+ * @next:	if present, wake next and hint to the scheduler that we'd
+ *		prefer to execute it locally.
  */
 static void futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q,
-				struct hrtimer_sleeper *timeout)
+				struct hrtimer_sleeper *timeout,
+				struct task_struct *next)
 {
 	/*
 	 * The task state is guaranteed to be set before another task can
@@ -2598,10 +2615,26 @@ static void futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q,
 		 * flagged for rescheduling. Only call schedule if there
 		 * is no timeout, or if it has yet to expire.
 		 */
-		if (!timeout || timeout->task)
+		if (!timeout || timeout->task) {
+			if (next) {
+				/*
+				 * wake_up_process() below will be replaced
+				 * in the next patch with
+				 * wake_up_process_prefer_current_cpu().
+				 */
+				wake_up_process(next);
+				put_task_struct(next);
+				next = NULL;
+			}
 			freezable_schedule();
+		}
 	}
 	__set_current_state(TASK_RUNNING);
+
+	if (next) {
+		wake_up_process(next);
+		put_task_struct(next);
+	}
 }
 
 /**
@@ -2677,7 +2710,7 @@ static int futex_wait_setup(u32 __user *uaddr, u32 val, unsigned int flags,
 }
 
 static int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
-		      ktime_t *abs_time, u32 bitset)
+		      ktime_t *abs_time, u32 bitset, struct task_struct *next)
 {
 	struct hrtimer_sleeper timeout, *to;
 	struct restart_block *restart;
@@ -2701,7 +2734,8 @@ static int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
 		goto out;
 
 	/* queue_me and wait for wakeup, timeout, or a signal. */
-	futex_wait_queue_me(hb, &q, to);
+	futex_wait_queue_me(hb, &q, to, next);
+	next = NULL;
 
 	/* If we were woken (and unqueued), we succeeded, whatever. */
 	ret = 0;
@@ -2733,6 +2767,10 @@ static int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
 	ret = set_restart_fn(restart, futex_wait_restart);
 
 out:
+	if (next) {
+		wake_up_process(next);
+		put_task_struct(next);
+	}
 	if (to) {
 		hrtimer_cancel(&to->timer);
 		destroy_hrtimer_on_stack(&to->timer);
@@ -2740,7 +2778,6 @@ static int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
 	return ret;
 }
 
-
 static long futex_wait_restart(struct restart_block *restart)
 {
 	u32 __user *uaddr = restart->futex.uaddr;
@@ -2752,10 +2789,29 @@ static long futex_wait_restart(struct restart_block *restart)
 	}
 	restart->fn = do_no_restart_syscall;
 
-	return (long)futex_wait(uaddr, restart->futex.flags,
-				restart->futex.val, tp, restart->futex.bitset);
+	return (long)futex_wait(uaddr, restart->futex.flags, restart->futex.val,
+				tp, restart->futex.bitset, NULL);
 }
 
+static int futex_swap(u32 __user *uaddr, unsigned int flags, u32 val,
+		      ktime_t *abs_time, u32 __user *uaddr2)
+{
+	u32 bitset = FUTEX_BITSET_MATCH_ANY;
+	struct task_struct *next = NULL;
+	DEFINE_WAKE_Q(wake_q);
+	int ret;
+
+	ret = prepare_wake_q(uaddr2, flags, 1, bitset, &wake_q);
+	if (ret < 0)
+		return ret;
+	if (!wake_q_empty(&wake_q)) {
+		/* At most one wakee can be present. Pull it out. */
+		next = container_of(wake_q.first, struct task_struct, wake_q);
+		next->wake_q.next = NULL;
+	}
+
+	return futex_wait(uaddr, flags, val, abs_time, bitset, next);
+}
 
 /*
  * Userspace tried a 0 -> TID atomic transition of the futex value
@@ -3216,7 +3272,7 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 	}
 
 	/* Queue the futex_q, drop the hb lock, wait for wakeup. */
-	futex_wait_queue_me(hb, &q, to);
+	futex_wait_queue_me(hb, &q, to, NULL);
 
 	spin_lock(&hb->lock);
 	ret = handle_early_requeue_pi_wakeup(hb, &q, &key2, to);
@@ -3727,7 +3783,7 @@ long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout,
 		val3 = FUTEX_BITSET_MATCH_ANY;
 		fallthrough;
 	case FUTEX_WAIT_BITSET:
-		return futex_wait(uaddr, flags, val, timeout, val3);
+		return futex_wait(uaddr, flags, val, timeout, val3, NULL);
 	case FUTEX_WAKE:
 		val3 = FUTEX_BITSET_MATCH_ANY;
 		fallthrough;
@@ -3754,6 +3810,8 @@ long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout,
 					     uaddr2);
 	case FUTEX_CMP_REQUEUE_PI:
 		return futex_requeue(uaddr, flags, uaddr2, val, val2, &val3, 1);
+	case FUTEX_SWAP:
+		return futex_swap(uaddr, flags, val, timeout, uaddr2);
 	}
 	return -ENOSYS;
 }
@@ -3766,6 +3824,7 @@ static __always_inline bool futex_cmd_has_timeout(u32 cmd)
 	case FUTEX_LOCK_PI2:
 	case FUTEX_WAIT_BITSET:
 	case FUTEX_WAIT_REQUEUE_PI:
+	case FUTEX_SWAP:
 		return true;
 	}
 	return false;
@@ -3778,7 +3837,7 @@ futex_init_timeout(u32 cmd, u32 op, struct timespec64 *ts, ktime_t *t)
 		return -EINVAL;
 
 	*t = timespec64_to_ktime(*ts);
-	if (cmd == FUTEX_WAIT)
+	if (cmd == FUTEX_WAIT || cmd == FUTEX_SWAP)
 		*t = ktime_add_safe(ktime_get(), *t);
 	else if (cmd != FUTEX_LOCK_PI && !(op & FUTEX_CLOCK_REALTIME))
 		*t = timens_ktime_to_host(CLOCK_MONOTONIC, *t);
-- 
2.32.0.93.g670b81a890-goog