.

go/neo/client/client.go

@@ -446,25 +446,25 @@ func (c *Client) Load(ctx context.Context, xid zodb.Xid) (data *zodb.Buf, serial
 		return nil, 0, err	// XXX err context
 	}
 
-	data = resp.Data
+	buf = resp.DataBuf
 
-	//checksum := sha1.Sum(data.Data)
+	//checksum := sha1.Sum(buf.Data)
 	//if checksum != resp.Checksum {
 	//	return nil, 0, fmt.Errorf("data corrupt: checksum mismatch")
 	//}
 
 	if resp.Compression {
 		// XXX cleanup mess vvv
-		data2 := zodb.BufAlloc(len(data.Data))
-		data2.Data = data2.Data[:0]
-		udata, err := decompress(data.Data, data2.Data)
-		data.Free()
+		buf2 := zodb.BufAlloc(len(buf.Data))
+		buf2.Data = buf2.Data[:0]
+		udata, err := decompress(buf.Data, buf2.Data)
+		buf.Release()
 		if err != nil {
-			data2.Free()
+			buf2.Release()
 			return nil, 0, fmt.Errorf("data corrupt: %v", err)
 		}
-		data2.Data = udata
-		data = data2
+		buf2.Data = udata
+		buf = buf2
 	}
 
 	// reply.NextSerial

go/neo/proto.go

@@ -646,8 +646,7 @@ type AnswerObject struct {
 	NextSerial      zodb.Tid	// XXX but there it is out of sync
 	Compression     bool
 	Checksum        Checksum
-//	Data            []byte          // TODO encode -> separately (for writev)
-	Data            *zodb.Buf	// TODO encode -> separately (for writev)
+	DataBuf         *zodb.Buf	// TODO encode -> separately (for writev)
 	DataSerial      zodb.Tid
 }
 

go/zodb/buffer.go

@@ -22,6 +22,7 @@ package zodb
 
 import (
 	"sync"
+	"sync/atomic"
 
 	"lab.nexedi.com/kirr/go123/xmath"
 )
@@ -29,13 +30,19 @@ import (
 // Buf represents memory buffer.
 //
 // To lower pressure on Go garbage-collector allocate buffers with BufAlloc and
-// free them with Buf.Free.
+// free them with Buf.Release.
 //
 // Custom allocation functions affect only performance, not correctness -
 // everything should work if data buffer is allocated and/or free'ed
 // regular Go/GC-way.
 type Buf struct {
 	Data []byte
+
+	// reference counter.
+	//
+	// NOTE to allow both Bufs created via BufAlloc and via std new Buf is
+	// created with refcnt=0. The real number of references to Buf is thus .refcnt+1
+	refcnt int32
 }
 
 const order0 = 4 // buf sizes start from 2^4 (=16)
@@ -87,13 +94,27 @@ func BufAlloc64(size int64) *Buf {
 	buf := bufPoolv[order].Get().(*Buf)
 	//println("\tlen:", len(buf.Data), "cap:", cap(buf.Data))
 	buf.Data = buf.Data[:size] // leaving cap as is = 2^i
+	buf.refcnt = 0
 	return buf
 }
 
-// Free returns no-longer needed buf to freelist.
+// Release marks buf as no longer used by caller.
+//
+// It decrements buf reference-counter and if it reaches zero returns buf to
+// freelist.
+//
+// The caller must not use buf after call to Release.
 //
-// The caller must not use buf after call to Free.
-func (buf *Buf) Free() {
+// XXX naming? -> Free? -> Decref? -> Unref?
+func (buf *Buf) Release() {
+	rc := atomic.AddInt32(&buf.refcnt, -1)
+	if rc < -1 {
+		panic("Buf: refcnt < 0")
+	}
+	if rc > -1 {
+		return
+	}
+
 	// order = max i: 2^i <= cap
 	//order := bits.Len(uint(cap(buf.Data)))
 	order := xmath.FloorLog2(uint64(cap(buf.Data)))
@@ -110,3 +131,43 @@ func (buf *Buf) Free() {
 
 	bufPoolv[order].Put(buf)
 }
+
+// Incref increments buf's reference counter by 1.
+func (buf *Buf) Incref() {
+	atomic.AddInt32(&buf.refcnt, +1)
+}
+
+// XRelease releases buf it is != nil.
+func (buf *Buf) XRelease() {
+	if buf != nil {
+		buf.Release()
+	}
+}
+
+// XIncref increments buf's reference counter by 1 if buf != nil.
+func (buf *Buf) XIncref() {
+	if buf != nil {
+		buf.Incref()
+	}
+}
+
+
+// Len returns buf's len.
+//
+// it works even if buf=nil similarly to len() on nil []byte slice.
+func (buf *Buf) Len() int {
+	if buf != nil {
+		return len(buf.Data)
+	}
+	return 0
+}
+
+// Cap returns buf's cap.
+//
+// it works even if buf=nil similarly to len() on nil []byte slice.
+func (buf *Buf) Cap() int {
+	if buf != nil {
+		return cap(buf.Data)
+	}
+	return 0
+}

go/zodb/buffer_test.go

@@ -25,11 +25,22 @@ import (
 	"unsafe"
 )
 
+//go:linkname runtime_procPin runtime.procPin
+//go:linkname runtime_procUnpin runtime.procUnpin
+func runtime_procPin() int
+func runtime_procUnpin()
+
+
 func sliceDataPtr(b []byte) unsafe.Pointer {
 	return unsafe.Pointer((*reflect.SliceHeader)(unsafe.Pointer(&b)).Data)
 }
 
 func TestBufAllocFree(t *testing.T) {
+	// sync.Pool uses per-P free-lists. We check that after release we will
+	// allocate released object. This works only if P is not changed underneath.
+	runtime_procPin()
+	defer runtime_procUnpin()
+
 	for i := uint(0); i < 25; i++ {
 		size := 1<<i - 1
 		xcap := 1<<i
@@ -48,22 +59,56 @@ func TestBufAllocFree(t *testing.T) {
 			t.Errorf("%v: cap=%v  ; want %v", i, cap(buf.Data), xcap)
 		}
 
+		checkref := func(rc int32) {
+			t.Helper()
+			if buf.refcnt != rc {
+				t.Errorf("%v: refcnt=%v  ; want %v", i, buf.refcnt, rc)
+			}
+		}
+
+		checkref(0)
+
 		// free and allocate another buf -> it must be it
 		data := buf.Data
-		buf.Free()
+		buf.Release()
+		checkref(-1)
 		buf2 := BufAlloc(size)
 
-		// from pool -> it must be the same
-		if int(i) < order0 + len(bufPoolv) {
-			if !(buf2 == buf && sliceDataPtr(buf2.Data) == sliceDataPtr(data)) {
-				t.Errorf("%v: buffer not reused on free/realloc", i)
-			}
-
 		// not from pool - memory won't be reused
-		} else {
+		if int(i) >= order0 + len(bufPoolv) {
 			if buf2 == buf || sliceDataPtr(buf2.Data) == sliceDataPtr(data) {
 				t.Errorf("%v: buffer reused but should not", i)
 			}
+			continue
+		}
+
+		// from pool -> it must be the same
+		if !(buf2 == buf && sliceDataPtr(buf2.Data) == sliceDataPtr(data)) {
+			t.Errorf("%v: buffer not reused on free/realloc", i)
+		}
+		checkref(0)
+
+		// add more ref and release original buf - it must stay alive
+		buf.Incref()
+		checkref(1)
+		buf.Release()
+		checkref(0)
+
+		// another alloc must be different
+		buf2 = BufAlloc(size)
+		checkref(0)
+
+		if buf2 == buf || sliceDataPtr(buf2.Data) == sliceDataPtr(data) {
+			t.Errorf("%v: buffer reused but should not", i)
+		}
+
+		// release buf again -> should go to pool
+		buf.Release()
+		checkref(-1)
+		buf2 = BufAlloc(size)
+		if !(buf2 == buf && sliceDataPtr(buf2.Data) == sliceDataPtr(data)) {
+			t.Errorf("%v: buffer not reused on free/realloc", i)
 		}
+		checkref(0)
 	}
 }

go/zodb/buffer_test.s

+// empty .s so `go build` does not use -complete for go:linkname to work

go/zodb/storage/cache.go

@@ -96,8 +96,8 @@ type revCacheEntry struct {
 	// case when loadBefore with tid > cache.before was called.
 	before zodb.Tid
 
-	// loading result: object (data, serial) or error
-	data   []byte
+	// loading result: object (buf, serial) or error
+	buf    *zodb.Buf
 	serial zodb.Tid
 	err    error
 
@@ -108,7 +108,7 @@ type revCacheEntry struct {
 // StorLoader represents loading part of a storage.
 // XXX -> zodb?
 type StorLoader interface {
-	Load(ctx context.Context, xid zodb.Xid) (data []byte, serial zodb.Tid, err error)
+	Load(ctx context.Context, xid zodb.Xid) (buf *zodb.Buf, serial zodb.Tid, err error)
 }
 
 // lock order: Cache.mu   > oidCacheEntry
@@ -149,7 +149,7 @@ func (c *Cache) SetSizeMax(sizeMax int) {
 // Load loads data from database via cache.
 //
 // If data is already in cache - cached content is returned.
-func (c *Cache) Load(ctx context.Context, xid zodb.Xid) (data []byte, serial zodb.Tid, err error) {
+func (c *Cache) Load(ctx context.Context, xid zodb.Xid) (buf *zodb.Buf, serial zodb.Tid, err error) {
 	rce, rceNew := c.lookupRCE(xid)
 
 	// rce is already in cache - use it
@@ -177,7 +177,8 @@ func (c *Cache) Load(ctx context.Context, xid zodb.Xid) (data []byte, serial zod
 		}
 	}
 
-	return rce.data, rce.serial, nil
+	rce.buf.XIncref()
+	return rce.buf, rce.serial, nil
 }
 
 // Prefetch arranges for data to be eventually present in cache.
@@ -293,19 +294,20 @@ func (c *Cache) lookupRCE(xid zodb.Xid) (rce *revCacheEntry, rceNew bool) {
 // loading completion is signalled by closing rce.ready.
 func (c *Cache) loadRCE(ctx context.Context, rce *revCacheEntry, oid zodb.Oid) {
 	oce := rce.parent
-	data, serial, err := c.loader.Load(ctx, zodb.Xid{
+	buf, serial, err := c.loader.Load(ctx, zodb.Xid{
 		Oid:  oid,
 		XTid: zodb.XTid{Tid: rce.before, TidBefore: true},
 	})
 
-	// normalize data/serial if it was error
+	// normalize buf/serial if it was error
 	if err != nil {
 		// XXX err == canceled? -> ?
-		data = nil
+		buf.XRelease()
+		buf = nil
 		serial = 0
 	}
 	rce.serial = serial
-	rce.data = data
+	rce.buf = buf
 	rce.err = err
 	// verify db gives serial < before
 	if rce.serial >= rce.before {
@@ -313,7 +315,7 @@ func (c *Cache) loadRCE(ctx context.Context, rce *revCacheEntry, oid zodb.Oid) {
 	}
 
 	close(rce.ready)
-	δsize := len(rce.data)
+	δsize := rce.buf.Len()
 
 	// merge rce with adjacent entries in parent
 	// ( e.g. loadBefore(3) and loadBefore(4) results in the same data loaded if
@@ -339,9 +341,10 @@ func (c *Cache) loadRCE(ctx context.Context, rce *revCacheEntry, oid zodb.Oid) {
 	if i + 1 < len(oce.rcev) {
 		rceNext := oce.rcev[i+1]
 		if rceNext.loaded() && tryMerge(rce, rceNext, rce, oid) {
-			// not δsize -= len(rce.data)
-			// tryMerge can change rce.data if consistency is broken
+			// not δsize -= len(rce.buf.Data)
+			// tryMerge can change rce.buf if consistency is broken
 			δsize = 0
+			rce.buf.XRelease()
 			rce = rceNext
 			rceDropped = true
 		}
@@ -355,8 +358,9 @@ func (c *Cache) loadRCE(ctx context.Context, rce *revCacheEntry, oid zodb.Oid) {
 		if rcePrev.loaded() && tryMerge(rcePrev, rce, rce, oid) {
 			rcePrevDropped = rcePrev
 			if rcePrev.accounted {
-				δsize -= len(rcePrev.data)
+				δsize -= rcePrev.buf.Len()
 			}
+			rcePrev.buf.XRelease()
 		}
 	}
 
@@ -506,8 +510,9 @@ func (c *Cache) gc() {
 		i := oce.find(rce)
 		if i != -1 {	// rce could be already deleted by e.g. merge
 			oce.deli(i)
-			c.size -= len(rce.data)
-			//fmt.Printf("gc: free %d bytes\n", len(rce.data))
+			c.size -= rce.buf.Len()
+			//fmt.Printf("gc: free %d bytes\n", rce.buf.Len()))
+			rce.buf.XRelease()
 		}
 		oce.Unlock()
 
@@ -632,6 +637,7 @@ func errDB(oid zodb.Oid, format string, argv ...interface{}) error {
 // errDB marks rce with database inconsistency error
 func (rce *revCacheEntry) errDB(oid zodb.Oid, format string, argv ...interface{}) {
 	rce.err = errDB(oid, format, argv...)
-	rce.data = nil
+	rce.buf.XRelease()
+	rce.buf = nil
 	rce.serial = 0
 }

go/zodb/storage/cache_test.go

@@ -50,9 +50,27 @@ type tOidData struct {
 	err    error    // e.g. io error
 }
 
-func (stor *tStorage) Load(_ context.Context, xid zodb.Xid) (data []byte, serial zodb.Tid, err error) {
+// create new buffer with specified content copied there.
+func mkbuf(data []byte) *zodb.Buf {
+	buf := zodb.BufAlloc(len(data))
+	copy(buf.Data, data)
+	return buf
+}
+
+// check whether buffers hold same data or both are nil.
+//
+// NOTE we ignore refcnt here
+func bufSame(buf1, buf2 *zodb.Buf) bool {
+	if buf1 == nil {
+		return (buf2 == nil)
+	}
+
+	return reflect.DeepEqual(buf1.Data, buf2.Data)
+}
+
+func (stor *tStorage) Load(_ context.Context, xid zodb.Xid) (buf *zodb.Buf, serial zodb.Tid, err error) {
 	//fmt.Printf("> load(%v)\n", xid)
-	//defer func() { fmt.Printf("< %v, %v, %v\n", data, serial, err) }()
+	//defer func() { fmt.Printf("< %v, %v, %v\n", buf.XData(), serial, err) }()
 	tid := xid.Tid
 	if !xid.TidBefore {
 		tid++		// XXX overflow?
@@ -85,7 +103,7 @@ func (stor *tStorage) Load(_ context.Context, xid zodb.Xid) (data []byte, serial
 	if e != nil {
 		s = 0 // obey protocol of returning 0 with error
 	}
-	return datav[i].data, s, e
+	return mkbuf(datav[i].data), s, e
 }
 
 var ioerr = errors.New("input/output error")
@@ -142,15 +160,17 @@ func TestCache(t *testing.T) {
 		},
 	}
 
+	b := mkbuf
+
 	c := NewCache(tstor, 100 /* > Σ all data */)
 	ctx := context.Background()
 
-	checkLoad := func(xid zodb.Xid, data []byte, serial zodb.Tid, err error) {
+	checkLoad := func(xid zodb.Xid, buf *zodb.Buf, serial zodb.Tid, err error) {
 		t.Helper()
 		bad := &bytes.Buffer{}
-		d, s, e := c.Load(ctx, xid)
-		if !reflect.DeepEqual(data, d) {
-			fmt.Fprintf(bad, "data:\n%s\n", pretty.Compare(data, d))
+		b, s, e := c.Load(ctx, xid)
+		if !bufSame(buf, b) {
+			fmt.Fprintf(bad, "buf:\n%s\n", pretty.Compare(buf, b))
 		}
 		if serial != s {
 			fmt.Fprintf(bad, "serial:\n%s\n", pretty.Compare(serial, s))
@@ -164,7 +184,7 @@ func TestCache(t *testing.T) {
 		}
 	}
 
-	checkRCE := func(rce *revCacheEntry, before, serial zodb.Tid, data []byte, err error) {
+	checkRCE := func(rce *revCacheEntry, before, serial zodb.Tid, buf *zodb.Buf, err error) {
 		t.Helper()
 		bad := &bytes.Buffer{}
 		if rce.before != before {
@@ -173,8 +193,8 @@ func TestCache(t *testing.T) {
 		if rce.serial != serial {
 			fmt.Fprintf(bad, "serial:\n%s\n", pretty.Compare(serial, rce.serial))
 		}
-		if !reflect.DeepEqual(rce.data, data) {
-			fmt.Fprintf(bad, "data:\n%s\n", pretty.Compare(data, rce.data))
+		if !bufSame(rce.buf, buf) {
+			fmt.Fprintf(bad, "buf:\n%s\n", pretty.Compare(buf, rce.buf))
 		}
 		if !reflect.DeepEqual(rce.err, err) {
 			fmt.Fprintf(bad, "err:\n%s\n", pretty.Compare(err, rce.err))
@@ -209,7 +229,7 @@ func TestCache(t *testing.T) {
 					debug.Sprint(h), debug.Sprint(hp))
 			}
 			rce := h.rceFromInLRU()
-			size += len(rce.data)
+			size += rce.buf.Len()
 			mruv = append(mruv, rce)
 		}
 		if !reflect.DeepEqual(mruv, mruvOk) {
@@ -253,24 +273,24 @@ func TestCache(t *testing.T) {
 	checkMRU(0, rce1_b4)
 
 	// load <6 -> new rce entry with data
-	checkLoad(xidlt(1,6), hello, 4, nil)
+	checkLoad(xidlt(1,6), b(hello), 4, nil)
 	ok1(len(oce1.rcev) == 2)
 	rce1_b6 := oce1.rcev[1]
-	checkRCE(rce1_b6, 6, 4, hello, nil)
+	checkRCE(rce1_b6, 6, 4, b(hello), nil)
 	checkOCE(1, rce1_b4, rce1_b6)
 	checkMRU(5, rce1_b6, rce1_b4)
 
 	// load <5 -> <5 merged with <6
-	checkLoad(xidlt(1,5), hello, 4, nil)
+	checkLoad(xidlt(1,5), b(hello), 4, nil)
 	checkOCE(1, rce1_b4, rce1_b6)
 	checkMRU(5, rce1_b6, rce1_b4)
 
 	// load <7 -> <6 merged with <7
-	checkLoad(xidlt(1,7), hello, 4, nil)
+	checkLoad(xidlt(1,7), b(hello), 4, nil)
 	ok1(len(oce1.rcev) == 2)
 	rce1_b7 := oce1.rcev[1]
 	ok1(rce1_b7 != rce1_b6)
-	checkRCE(rce1_b7, 7, 4, hello, nil)
+	checkRCE(rce1_b7, 7, 4, b(hello), nil)
 	checkOCE(1, rce1_b4, rce1_b7)
 	checkMRU(5, rce1_b7, rce1_b4)
 
@@ -291,19 +311,19 @@ func TestCache(t *testing.T) {
 	checkMRU(5, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
 	// load <11 -> new data rce, not merged with ioerr @<10
-	checkLoad(xidlt(1,11), world, 10, nil)
+	checkLoad(xidlt(1,11), b(world), 10, nil)
 	ok1(len(oce1.rcev) == 5)
 	rce1_b11 := oce1.rcev[4]
-	checkRCE(rce1_b11, 11, 10, world, nil)
+	checkRCE(rce1_b11, 11, 10, b(world), nil)
 	checkOCE(1, rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b11)
 	checkMRU(12, rce1_b11, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
 	// load <12 -> <11 merged with <12
-	checkLoad(xidlt(1,12), world, 10, nil)
+	checkLoad(xidlt(1,12), b(world), 10, nil)
 	ok1(len(oce1.rcev) == 5)
 	rce1_b12 := oce1.rcev[4]
 	ok1(rce1_b12 != rce1_b11)
-	checkRCE(rce1_b12, 12, 10, world, nil)
+	checkRCE(rce1_b12, 12, 10, b(world), nil)
 	checkOCE(1, rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b12)
 	checkMRU(12, rce1_b12, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
@@ -315,7 +335,7 @@ func TestCache(t *testing.T) {
 	rce1_b16, new16 := c.lookupRCE(xidlt(1,16))
 	ok1(new16)
 	rce1_b16.serial = 10
-	rce1_b16.data = world
+	rce1_b16.buf = mkbuf(world)
 	// here: first half of loadRCE(<16) before close(<16.ready)
 	checkOCE(1, rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b12, rce1_b16)
 	ok1(!rce1_b16.loaded())
@@ -338,9 +358,9 @@ func TestCache(t *testing.T) {
 	//  <16 did not yet took oce lock so c.size is temporarily reduced and
 	//  <16 is not yet on LRU list)
 	c.loadRCE(ctx, rce1_b14, 1)
-	checkRCE(rce1_b14, 14, 10, world, nil)
-	checkRCE(rce1_b16, 16, 10, world, nil)
-	checkRCE(rce1_b12, 12, 10, world, nil)
+	checkRCE(rce1_b14, 14, 10, b(world), nil)
+	checkRCE(rce1_b16, 16, 10, b(world), nil)
+	checkRCE(rce1_b12, 12, 10, b(world), nil)
 	checkOCE(1, rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b16)
 	checkMRU(5 /*was 12*/, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
@@ -359,7 +379,7 @@ func TestCache(t *testing.T) {
 
 	// (<17 loads but not yet takes oce lock)
 	rce1_b17.serial = 16
-	rce1_b17.data = zz
+	rce1_b17.buf = mkbuf(zz)
 	close(rce1_b17.ready)
 	ok1(rce1_b17.loaded())
 	checkOCE(1, rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b16, rce1_b17, rce1_b18)
@@ -367,8 +387,8 @@ func TestCache(t *testing.T) {
 
 	// (<18 loads and takes oce lock first - merge <17 with <18)
 	c.loadRCE(ctx, rce1_b18, 1)
-	checkRCE(rce1_b18, 18, 16, zz, nil)
-	checkRCE(rce1_b17, 17, 16, zz, nil)
+	checkRCE(rce1_b18, 18, 16, b(zz), nil)
+	checkRCE(rce1_b17, 17, 16, b(zz), nil)
 	checkOCE(1, rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b16, rce1_b18)
 	checkMRU(14, rce1_b18, rce1_b16, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
@@ -377,15 +397,15 @@ func TestCache(t *testing.T) {
 	ok1(len(oce1.rcev) == 6)
 	rce1_b20 := oce1.rcev[5]
 	ok1(rce1_b20 != rce1_b18)
-	checkRCE(rce1_b20, 20, 16, zz, nil)
+	checkRCE(rce1_b20, 20, 16, b(zz), nil)
 	checkOCE(1,  rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b16, rce1_b20)
 	checkMRU(14, rce1_b20, rce1_b16, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
 	// load =20 -> new <21
-	checkLoad(xideq(1,20), www, 20, nil)
+	checkLoad(xideq(1,20), b(www), 20, nil)
 	ok1(len(oce1.rcev) == 7)
 	rce1_b21 := oce1.rcev[6]
-	checkRCE(rce1_b21, 21, 20, www, nil)
+	checkRCE(rce1_b21, 21, 20, b(www), nil)
 	checkOCE(1,  rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b16, rce1_b20, rce1_b21)
 	checkMRU(17, rce1_b21, rce1_b20, rce1_b16, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
@@ -394,7 +414,7 @@ func TestCache(t *testing.T) {
 	ok1(len(oce1.rcev) == 7)
 	rce1_b22 := oce1.rcev[6]
 	ok1(rce1_b22 != rce1_b21)
-	checkRCE(rce1_b22, 22, 20, www, nil)
+	checkRCE(rce1_b22, 22, 20, b(www), nil)
 	checkOCE(1,  rce1_b4, rce1_b7, rce1_b8, rce1_b10, rce1_b16, rce1_b20, rce1_b22)
 	checkMRU(17, rce1_b22, rce1_b20, rce1_b16, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
@@ -468,10 +488,10 @@ func TestCache(t *testing.T) {
 	// ---- verify how LRU changes for in-cache loads ----
 	checkMRU(17, rce1_b9, rce1_b22, rce1_b20, rce1_b16, rce1_b10, rce1_b8, rce1_b7, rce1_b4)
 
-	checkLoad(xidlt(1,7), hello, 4, nil)
+	checkLoad(xidlt(1,7), b(hello), 4, nil)
 	checkMRU(17, rce1_b7, rce1_b9, rce1_b22, rce1_b20, rce1_b16, rce1_b10, rce1_b8, rce1_b4)
 
-	checkLoad(xidlt(1,16), world, 10, nil)
+	checkLoad(xidlt(1,16), b(world), 10, nil)
 	checkMRU(17, rce1_b16, rce1_b7, rce1_b9, rce1_b22, rce1_b20, rce1_b10, rce1_b8, rce1_b4)
 
 
@@ -517,7 +537,7 @@ func TestCache(t *testing.T) {
 	ok1(len(oce1.rcev) == 4)
 	rce1_b20_2 := oce1.rcev[3]
 	ok1(rce1_b20_2 != rce1_b20)
-	checkRCE(rce1_b20_2, 20, 16, zz, nil)
+	checkRCE(rce1_b20_2, 20, 16, b(zz), nil)
 	checkOCE(1,  rce1_b7, rce1_b9, rce1_b16, rce1_b20_2)
 	checkMRU(14, rce1_b20_2, rce1_b16, rce1_b7, rce1_b9)
 
@@ -531,7 +551,7 @@ func TestCache(t *testing.T) {
 	// - loaded  <78 (big, size=10)
 	ok1(len(oce1.rcev) == 2)
 	rce1_b78 := oce1.rcev[1]
-	checkRCE(rce1_b78, 78, 77, big, nil)
+	checkRCE(rce1_b78, 78, 77, b(big), nil)
 	checkOCE(1,  rce1_b20_2, rce1_b78)
 	checkMRU(12, rce1_b78, rce1_b20_2)
 
@@ -544,7 +564,7 @@ func TestCache(t *testing.T) {
 	// and still loading works (because even if though rce's are evicted
 	// they stay live while someone user waits and uses it)
 
-	checkLoad(xidlt(1,5), hello, 4, nil)
+	checkLoad(xidlt(1,5), b(hello), 4, nil)
 	tc.Expect(gcstart, gcfinish)
 	checkOCE(1)
 	checkMRU(0)

go/zodb/storage/fs1/filestorage.go

@@ -369,7 +369,7 @@ func (zi *zIter) NextData() (*zodb.DataInfo, error) {
 	// - need to use separate dh because of this
 	zi.dhLoading = zi.iter.Datah
 	if zi.dataBuf != nil {
-		zi.dataBuf.Free()
+		zi.dataBuf.Release()
 		zi.dataBuf = nil
 	}
 	zi.dataBuf, err = zi.dhLoading.LoadData(zi.iter.R)

go/zodb/storage/fs1/format.go

@@ -685,7 +685,7 @@ func (dh *DataHeader) LoadData(r io.ReaderAt) (*zodb.Buf, error) {
 	buf := zodb.BufAlloc64(dh.DataLen)
 	_, err := r.ReadAt(buf.Data, dh.Pos + DataHeaderSize)
 	if err != nil {
-		buf.Free()
+		buf.Release()
 		return nil, dh.err("read data", noEOF(err))	// XXX recheck
 	}
 

go/zodb/storage/fs1/fs1tools/dump.go

@@ -194,7 +194,7 @@ func (d *DumperFsDump) DumpTxn(buf *xfmt.Buffer, it *fs1.Iter) error {
 		}
 
 		buf .S("\n")
-		dbuf.Free()
+		dbuf.Release()
 	}
 
 	return nil

go/zodb/zodbtools/catobj.go

@@ -40,7 +40,7 @@ func Catobj(ctx context.Context, w io.Writer, stor zodb.IStorage, xid zodb.Xid)
 	}
 
 	_, err = w.Write(buf.Data)	// NOTE deleted data are returned as err by Load
-	buf.Free()
+	buf.Release()
 	return err		// XXX err ctx ?
 }
 
@@ -61,7 +61,7 @@ func Dumpobj(ctx context.Context, w io.Writer, stor zodb.IStorage, xid zodb.Xid,
 
 	d := dumper{W: w, HashOnly: hashOnly}
 	err = d.DumpData(&objInfo)
-	buf.Free()
+	buf.Release()
 	return err
 }