storage/cmdq: O(1) copy-on-write btree clones and atomic refcount GC …

…policy

All commits from cockroachdb#32165 except the last one.

This change introduces O(1) btree cloning and a new copy-on-write scheme,
essentially giving the btree an immutable API (for which I took inspiration
from https://docs.rs/crate/im/). This is made efficient by the second part
of the change - a new garbage collection policy for btrees. Nodes are now
reference counted atomically and freed into global `sync.Pools` when they
are no longer referenced.

One of the main ideas in cockroachdb#31997 is to treat the btrees backing the command
queue as immutable structures. In doing so, we adopt a copy-on-write scheme.
Trees are cloned under lock and then accessed concurrently. When future
writers want to modify the tree, they can do so by cloning any nodes that
they touch. This commit provides this functionality in a much more elegant
manner than 6994347. Instead of giving each node a "copy-on-write context",
we instead give each node a reference count. We then use the following rule:
1. trees with exclusive ownership (refcount == 1) over a node can modify
   it in-place.
2. trees without exclusive ownership over a node must clone the node
   in order to modify it. Once cloned, the tree will now have exclusive
   ownership over that node. When cloning the node, the reference count
   of all of the node's children must be incremented.

In following the simple rules, we end up with a really nice property -
trees gain more and more "ownership" as they make modifications, meaning
that subsequent modifications are much less likely to need to clone nodes.
Essentially, we transparently incorporates the idea of local mutations
(e.g. Clojure's transients or Haskell's ST monad) without any external
API needed.

Even better, reference counting internal nodes ties directly into the
new GC policy, which allows us to recycle old nodes and make the copy-on-write
scheme zero-allocation in almost all cases. When a node's reference count
drops to 0, we simply toss it into a `sync.Pool`. We keep two separate
pools - one for leaf nodes and one for non-leaf nodes. This wasn't possible
with the previous "copy-on-write context" approach.

The atomic reference counting does have an effect on benchmarks, but
its not a big one (single/double digit ns) and is negligible compared to
the speedup observed in cockroachdb#32165.
```
name                             old time/op  new time/op  delta
BTreeInsert/count=16-4           73.2ns ± 4%  84.4ns ± 4%  +15.30%  (p=0.008 n=5+5)
BTreeInsert/count=128-4           152ns ± 4%   167ns ± 4%   +9.89%  (p=0.008 n=5+5)
BTreeInsert/count=1024-4          250ns ± 1%   263ns ± 2%   +5.21%  (p=0.008 n=5+5)
BTreeInsert/count=8192-4          381ns ± 1%   394ns ± 2%   +3.36%  (p=0.008 n=5+5)
BTreeInsert/count=65536-4         720ns ± 6%   746ns ± 1%     ~     (p=0.119 n=5+5)
BTreeDelete/count=16-4            127ns ±15%   131ns ± 9%     ~     (p=0.690 n=5+5)
BTreeDelete/count=128-4           182ns ± 8%   192ns ± 8%     ~     (p=0.222 n=5+5)
BTreeDelete/count=1024-4          323ns ± 3%   340ns ± 4%   +5.20%  (p=0.032 n=5+5)
BTreeDelete/count=8192-4          532ns ± 2%   556ns ± 1%   +4.55%  (p=0.008 n=5+5)
BTreeDelete/count=65536-4        1.15µs ± 2%  1.22µs ± 7%     ~     (p=0.222 n=5+5)
BTreeDeleteInsert/count=16-4      166ns ± 4%   174ns ± 3%   +4.70%  (p=0.032 n=5+5)
BTreeDeleteInsert/count=128-4     370ns ± 2%   383ns ± 1%   +3.57%  (p=0.008 n=5+5)
BTreeDeleteInsert/count=1024-4    548ns ± 3%   575ns ± 5%   +4.89%  (p=0.032 n=5+5)
BTreeDeleteInsert/count=8192-4    775ns ± 1%   789ns ± 1%   +1.86%  (p=0.016 n=5+5)
BTreeDeleteInsert/count=65536-4  2.20µs ±22%  2.10µs ±18%     ~     (p=0.841 n=5+5)
```

We can see how important the GC and memory re-use policy is by comparing
the following few benchmarks. Specifically, notice the difference in
operation speed and allocation count in `BenchmarkBTreeDeleteInsertCloneEachTime`
between the tests that `Reset` old clones (allowing nodes to be freed into
`sync.Pool`s) and the tests that don't `Reset` old clones.
```
name                                                      time/op
BTreeDeleteInsert/count=16-4                               198ns ±28%
BTreeDeleteInsert/count=128-4                              375ns ± 3%
BTreeDeleteInsert/count=1024-4                             577ns ± 2%
BTreeDeleteInsert/count=8192-4                             798ns ± 1%
BTreeDeleteInsert/count=65536-4                           2.00µs ±13%
BTreeDeleteInsertCloneOnce/count=16-4                      173ns ± 2%
BTreeDeleteInsertCloneOnce/count=128-4                     379ns ± 2%
BTreeDeleteInsertCloneOnce/count=1024-4                    584ns ± 4%
BTreeDeleteInsertCloneOnce/count=8192-4                    800ns ± 2%
BTreeDeleteInsertCloneOnce/count=65536-4                  2.04µs ±32%
BTreeDeleteInsertCloneEachTime/reset=false/count=16-4      535ns ± 8%
BTreeDeleteInsertCloneEachTime/reset=false/count=128-4    1.29µs ± 1%
BTreeDeleteInsertCloneEachTime/reset=false/count=1024-4   2.22µs ± 5%
BTreeDeleteInsertCloneEachTime/reset=false/count=8192-4   2.55µs ± 5%
BTreeDeleteInsertCloneEachTime/reset=false/count=65536-4  5.89µs ±20%
BTreeDeleteInsertCloneEachTime/reset=true/count=16-4       240ns ± 1%
BTreeDeleteInsertCloneEachTime/reset=true/count=128-4      610ns ± 4%
BTreeDeleteInsertCloneEachTime/reset=true/count=1024-4    1.20µs ± 2%
BTreeDeleteInsertCloneEachTime/reset=true/count=8192-4    1.69µs ± 1%
BTreeDeleteInsertCloneEachTime/reset=true/count=65536-4   3.52µs ±18%

name                                                      alloc/op
BTreeDeleteInsert/count=16-4                               0.00B
BTreeDeleteInsert/count=128-4                              0.00B
BTreeDeleteInsert/count=1024-4                             0.00B
BTreeDeleteInsert/count=8192-4                             0.00B
BTreeDeleteInsert/count=65536-4                            0.00B
BTreeDeleteInsertCloneOnce/count=16-4                      0.00B
BTreeDeleteInsertCloneOnce/count=128-4                     0.00B
BTreeDeleteInsertCloneOnce/count=1024-4                    0.00B
BTreeDeleteInsertCloneOnce/count=8192-4                    0.00B
BTreeDeleteInsertCloneOnce/count=65536-4                   1.00B ± 0%
BTreeDeleteInsertCloneEachTime/reset=false/count=16-4       288B ± 0%
BTreeDeleteInsertCloneEachTime/reset=false/count=128-4      897B ± 0%
BTreeDeleteInsertCloneEachTime/reset=false/count=1024-4   1.61kB ± 1%
BTreeDeleteInsertCloneEachTime/reset=false/count=8192-4   1.47kB ± 0%
BTreeDeleteInsertCloneEachTime/reset=false/count=65536-4  2.40kB ±12%
BTreeDeleteInsertCloneEachTime/reset=true/count=16-4       0.00B
BTreeDeleteInsertCloneEachTime/reset=true/count=128-4      0.00B
BTreeDeleteInsertCloneEachTime/reset=true/count=1024-4     0.00B
BTreeDeleteInsertCloneEachTime/reset=true/count=8192-4     0.00B
BTreeDeleteInsertCloneEachTime/reset=true/count=65536-4    0.00B

name                                                      allocs/op
BTreeDeleteInsert/count=16-4                                0.00
BTreeDeleteInsert/count=128-4                               0.00
BTreeDeleteInsert/count=1024-4                              0.00
BTreeDeleteInsert/count=8192-4                              0.00
BTreeDeleteInsert/count=65536-4                             0.00
BTreeDeleteInsertCloneOnce/count=16-4                       0.00
BTreeDeleteInsertCloneOnce/count=128-4                      0.00
BTreeDeleteInsertCloneOnce/count=1024-4                     0.00
BTreeDeleteInsertCloneOnce/count=8192-4                     0.00
BTreeDeleteInsertCloneOnce/count=65536-4                    0.00
BTreeDeleteInsertCloneEachTime/reset=false/count=16-4       1.00 ± 0%
BTreeDeleteInsertCloneEachTime/reset=false/count=128-4      2.00 ± 0%
BTreeDeleteInsertCloneEachTime/reset=false/count=1024-4     3.00 ± 0%
BTreeDeleteInsertCloneEachTime/reset=false/count=8192-4     3.00 ± 0%
BTreeDeleteInsertCloneEachTime/reset=false/count=65536-4    4.40 ±14%
BTreeDeleteInsertCloneEachTime/reset=true/count=16-4        0.00
BTreeDeleteInsertCloneEachTime/reset=true/count=128-4       0.00
BTreeDeleteInsertCloneEachTime/reset=true/count=1024-4      0.00
BTreeDeleteInsertCloneEachTime/reset=true/count=8192-4      0.00
BTreeDeleteInsertCloneEachTime/reset=true/count=65536-4     0.00
```

Release note: None

pkg/storage/cmdq/interval_btree.go

@@ -18,6 +18,8 @@ import (
 	"bytes"
 	"sort"
 	"strings"
+	"sync"
+	"sync/atomic"
 	"unsafe"
 
 	"github.com/cockroachdb/cockroach/pkg/roachpb"
@@ -106,21 +108,121 @@ func upperBound(c *cmd) keyBound {
 }
 
 type leafNode struct {
-	max   keyBound
+	ref   int32
 	count int16
 	leaf  bool
+	max   keyBound
 	cmds  [maxCmds]*cmd
 }
 
-func newLeafNode() *node {
-	return (*node)(unsafe.Pointer(&leafNode{leaf: true}))
-}
-
 type node struct {
 	leafNode
 	children [maxCmds + 1]*node
 }
 
+func leafToNode(ln *leafNode) *node {
+	return (*node)(unsafe.Pointer(ln))
+}
+
+func nodeToLeaf(n *node) *leafNode {
+	return (*leafNode)(unsafe.Pointer(n))
+}
+
+var leafPool = sync.Pool{
+	New: func() interface{} {
+		return new(leafNode)
+	},
+}
+
+var nodePool = sync.Pool{
+	New: func() interface{} {
+		return new(node)
+	},
+}
+
+func newLeafNode() *node {
+	n := leafToNode(leafPool.Get().(*leafNode))
+	n.leaf = true
+	n.ref = 1
+	return n
+}
+
+func newNode() *node {
+	n := nodePool.Get().(*node)
+	n.ref = 1
+	return n
+}
+
+// mut creates and returns a mutable node reference. If the node is not shared
+// with any other trees then it can be modified in place. Otherwise, it must be
+// cloned to ensure unique ownership. In this way, we enforce a copy-on-write
+// policy which transparently incorporates the idea of local mutations, like
+// Clojure's transients or Haskell's ST monad, where nodes are only copied
+// during the first time that they are modified between Clone operations.
+//
+// When a node is cloned, the provided pointer will be redirected to the new
+// mutable node.
+func mut(n **node) *node {
+	if atomic.LoadInt32(&(*n).ref) == 1 {
+		// Exclusive ownership. Can mutate in place.
+		return *n
+	}
+	// If we do not have unique ownership over the node then we
+	// clone it to gain unique ownership. After doing so, we can
+	// release our reference to the old node.
+	c := (*n).clone()
+	(*n).decRef(true /* recursive */)
+	*n = c
+	return *n
+}
+
+// incRef acquires a reference to the node.
+func (n *node) incRef() {
+	atomic.AddInt32(&n.ref, 1)
+}
+
+// decRef releases a reference to the node. If requested, the method
+// will recurse into child nodes and decrease their refcounts as well.
+func (n *node) decRef(recursive bool) {
+	if atomic.AddInt32(&n.ref, -1) > 0 {
+		// Other references remain. Can't free.
+		return
+	}
+	// Clear and release node into memory pool.
+	if n.leaf {
+		ln := nodeToLeaf(n)
+		*ln = leafNode{}
+		leafPool.Put(ln)
+	} else {
+		// Release child references first, if requested.
+		if recursive {
+			for i := int16(0); i <= n.count; i++ {
+				n.children[i].decRef(true /* recursive */)
+			}
+		}
+		*n = node{}
+		nodePool.Put(n)
+	}
+}
+
+// clone creates a clone of the receiver with a single reference count.
+func (n *node) clone() *node {
+	var c *node
+	if n.leaf {
+		c = newLeafNode()
+		*nodeToLeaf(c) = *nodeToLeaf(n)
+	} else {
+		c = newNode()
+		*c = *n
+		// Increase refcount of each child.
+		for i := int16(0); i <= c.count; i++ {
+			c.children[i].incRef()
+		}
+	}
+	c.ref = 1
+	return c
+}
+
 func (n *node) insertAt(index int, c *cmd, nd *node) {
 	if index < int(n.count) {
 		copy(n.cmds[index+1:n.count+1], n.cmds[index:n.count])
@@ -246,7 +348,7 @@ func (n *node) split(i int) (*cmd, *node) {
 	if n.leaf {
 		next = newLeafNode()
 	} else {
-		next = &node{}
+		next = newNode()
 	}
 	next.count = n.count - int16(i+1)
 	copy(next.cmds[:], n.cmds[i+1:n.count])
@@ -286,7 +388,7 @@ func (n *node) insert(c *cmd) (replaced, newBound bool) {
 		return false, n.adjustUpperBoundOnInsertion(c, nil)
 	}
 	if n.children[i].count >= maxCmds {
-		splitcmd, splitNode := n.children[i].split(maxCmds / 2)
+		splitcmd, splitNode := mut(&n.children[i]).split(maxCmds / 2)
 		n.insertAt(i, splitcmd, splitNode)
 
 		switch cmp := cmp(c, n.cmds[i]); {
@@ -299,7 +401,7 @@ func (n *node) insert(c *cmd) (replaced, newBound bool) {
 			return true, false
 		}
 	}
-	replaced, newBound = n.children[i].insert(c)
+	replaced, newBound = mut(&n.children[i]).insert(c)
 	if newBound {
 		newBound = n.adjustUpperBoundOnInsertion(c, nil)
 	}
@@ -316,7 +418,7 @@ func (n *node) removeMax() *cmd {
 		n.adjustUpperBoundOnRemoval(out, nil)
 		return out
 	}
-	child := n.children[n.count]
+	child := mut(&n.children[n.count])
 	if child.count <= minCmds {
 		n.rebalanceOrMerge(int(n.count))
 		return n.removeMax()
@@ -336,12 +438,12 @@ func (n *node) remove(c *cmd) (out *cmd, newBound bool) {
 		}
 		return nil, false
 	}
-	child := n.children[i]
-	if child.count <= minCmds {
+	if n.children[i].count <= minCmds {
 		// Child not large enough to remove from.
 		n.rebalanceOrMerge(i)
 		return n.remove(c)
 	}
+	child := mut(&n.children[i])
 	if found {
 		// Replace the cmd being removed with the max cmd in our left child.
 		out = n.cmds[i]
@@ -389,8 +491,8 @@ func (n *node) rebalanceOrMerge(i int) {
 		//                  v
 		//                  a
 		//
-		left := n.children[i-1]
-		child := n.children[i]
+		left := mut(&n.children[i-1])
+		child := mut(&n.children[i])
 		xCmd, grandChild := left.popBack()
 		yCmd := n.cmds[i-1]
 		child.pushFront(yCmd, grandChild)
@@ -428,8 +530,8 @@ func (n *node) rebalanceOrMerge(i int) {
 		//              v
 		//              a
 		//
-		right := n.children[i+1]
-		child := n.children[i]
+		right := mut(&n.children[i+1])
+		child := mut(&n.children[i])
 		xCmd, grandChild := right.popFront()
 		yCmd := n.cmds[i]
 		child.pushBack(yCmd, grandChild)
@@ -464,7 +566,9 @@ func (n *node) rebalanceOrMerge(i int) {
 		if i >= int(n.count) {
 			i = int(n.count - 1)
 		}
-		child := n.children[i]
+		child := mut(&n.children[i])
+		// Make mergeChild mutable, bumping the refcounts on its children if necessary.
+		_ = mut(&n.children[i+1])
 		mergeCmd, mergeChild := n.removeAt(i)
 		child.cmds[child.count] = mergeCmd
 		copy(child.cmds[child.count+1:], mergeChild.cmds[:mergeChild.count])
@@ -474,6 +578,7 @@ func (n *node) rebalanceOrMerge(i int) {
 		child.count += mergeChild.count + 1
 
 		child.adjustUpperBoundOnInsertion(mergeCmd, mergeChild)
+		mergeChild.decRef(false /* recursive */)
 	}
 }
 
@@ -547,25 +652,39 @@ type btree struct {
 	length int
 }
 
-// Reset removes all cmds from the btree.
+// Reset removes all cmds from the btree. In doing so, it allows memory
+// held by the btree to be recycled. Failure to call this method before
+// letting a btree be GCed is safe in that it won't cause a memory leak,
+// but it will prevent btree nodes from being efficiently re-used.
 func (t *btree) Reset() {
-	t.root = nil
+	if t.root != nil {
+		t.root.decRef(true /* recursive */)
+		t.root = nil
+	}
 	t.length = 0
 }
 
-// Silent unused warning.
-var _ = (*btree).Reset
+// Clone clones the btree, lazily.
+func (t *btree) Clone() btree {
+	c := *t
+	if c.root != nil {
+		c.root.incRef()
+	}
+	return c
+}
 
 // Delete removes a cmd equal to the passed in cmd from the tree.
 func (t *btree) Delete(c *cmd) {
 	if t.root == nil || t.root.count == 0 {
 		return
 	}
-	if out, _ := t.root.remove(c); out != nil {
+	if out, _ := mut(&t.root).remove(c); out != nil {
 		t.length--
 	}
 	if t.root.count == 0 && !t.root.leaf {
+		old := t.root
 		t.root = t.root.children[0]
+		old.decRef(false /* recursive */)
 	}
 }
 
@@ -575,16 +694,16 @@ func (t *btree) Set(c *cmd) {
 	if t.root == nil {
 		t.root = newLeafNode()
 	} else if t.root.count >= maxCmds {
-		splitcmd, splitNode := t.root.split(maxCmds / 2)
-		newRoot := &node{}
+		splitcmd, splitNode := mut(&t.root).split(maxCmds / 2)
+		newRoot := newNode()
 		newRoot.count = 1
 		newRoot.cmds[0] = splitcmd
 		newRoot.children[0] = t.root
 		newRoot.children[1] = splitNode
 		newRoot.max = newRoot.findUpperBound()
 		t.root = newRoot
 	}
-	if replaced, _ := t.root.insert(c); !replaced {
+	if replaced, _ := mut(&t.root).insert(c); !replaced {
 		t.length++
 	}
 }