Merge pull request #8210 from gyuho/bbolt

*: use 'coreos/bbolt' (replace 'boltdb/bolt')

bill-of-materials.json

@@ -27,19 +27,19 @@
 		]
 	},
 	{
-		"project": "github.com/boltdb/bolt",
+		"project": "github.com/cockroachdb/cmux",
 		"licenses": [
 			{
-				"type": "MIT License",
+				"type": "Apache License 2.0",
 				"confidence": 1
 			}
 		]
 	},
 	{
-		"project": "github.com/cockroachdb/cmux",
+		"project": "github.com/coreos/bbolt",
 		"licenses": [
 			{
-				"type": "Apache License 2.0",
+				"type": "MIT License",
 				"confidence": 1
 			}
 		]

cmd/vendor/github.com/boltdb/bolt/bolt_amd64.go

@@ -1,7 +0,0 @@
-package bolt
-
-// maxMapSize represents the largest mmap size supported by Bolt.
-const maxMapSize = 0xFFFFFFFFFFFF // 256TB
-
-// maxAllocSize is the size used when creating array pointers.
-const maxAllocSize = 0x7FFFFFFF

cmd/vendor/github.com/boltdb/bolt/bolt_arm.go

@@ -1,7 +0,0 @@
-package bolt
-
-// maxMapSize represents the largest mmap size supported by Bolt.
-const maxMapSize = 0x7FFFFFFF // 2GB
-
-// maxAllocSize is the size used when creating array pointers.
-const maxAllocSize = 0xFFFFFFF

cmd/vendor/github.com/boltdb/bolt/bolt_386.go → cmd/vendor/github.com/coreos/bbolt/bolt_386.go

@@ -5,3 +5,6 @@ const maxMapSize = 0x7FFFFFFF // 2GB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0xFFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_amd64.go

@@ -0,0 +1,10 @@
+package bolt
+
+// maxMapSize represents the largest mmap size supported by Bolt.
+const maxMapSize = 0xFFFFFFFFFFFF // 256TB
+
+// maxAllocSize is the size used when creating array pointers.
+const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_arm.go

@@ -0,0 +1,28 @@
+package bolt
+
+import "unsafe"
+
+// maxMapSize represents the largest mmap size supported by Bolt.
+const maxMapSize = 0x7FFFFFFF // 2GB
+
+// maxAllocSize is the size used when creating array pointers.
+const maxAllocSize = 0xFFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned bool
+
+func init() {
+	// Simple check to see whether this arch handles unaligned load/stores
+	// correctly.
+
+	// ARM9 and older devices require load/stores to be from/to aligned
+	// addresses. If not, the lower 2 bits are cleared and that address is
+	// read in a jumbled up order.
+
+	// See http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.faqs/ka15414.html
+
+	raw := [6]byte{0xfe, 0xef, 0x11, 0x22, 0x22, 0x11}
+	val := *(*uint32)(unsafe.Pointer(uintptr(unsafe.Pointer(&raw)) + 2))
+
+	brokenUnaligned = val != 0x11222211
+}

cmd/vendor/github.com/boltdb/bolt/bolt_arm64.go → cmd/vendor/github.com/coreos/bbolt/bolt_arm64.go

@@ -7,3 +7,6 @@ const maxMapSize = 0xFFFFFFFFFFFF // 256TB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/boltdb/bolt/bolt_ppc64.go → cmd/vendor/github.com/coreos/bbolt/bolt_ppc64.go

@@ -7,3 +7,6 @@ const maxMapSize = 0xFFFFFFFFFFFF // 256TB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/boltdb/bolt/bolt_ppc64le.go → cmd/vendor/github.com/coreos/bbolt/bolt_ppc64le.go

@@ -7,3 +7,6 @@ const maxMapSize = 0xFFFFFFFFFFFF // 256TB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/boltdb/bolt/bolt_s390x.go → cmd/vendor/github.com/coreos/bbolt/bolt_s390x.go

@@ -7,3 +7,6 @@ const maxMapSize = 0xFFFFFFFFFFFF // 256TB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/boltdb/bolt/bolt_windows.go → cmd/vendor/github.com/coreos/bbolt/bolt_windows.go

@@ -89,7 +89,7 @@ func flock(db *DB, mode os.FileMode, exclusive bool, timeout time.Duration) erro
 func funlock(db *DB) error {
 	err := unlockFileEx(syscall.Handle(db.lockfile.Fd()), 0, 1, 0, &syscall.Overlapped{})
 	db.lockfile.Close()
-	os.Remove(db.path+lockExt)
+	os.Remove(db.path + lockExt)
 	return err
 }
 

cmd/vendor/github.com/boltdb/bolt/bucket.go → cmd/vendor/github.com/coreos/bbolt/bucket.go

@@ -130,9 +130,17 @@ func (b *Bucket) Bucket(name []byte) *Bucket {
 func (b *Bucket) openBucket(value []byte) *Bucket {
 	var child = newBucket(b.tx)
 
+	// If unaligned load/stores are broken on this arch and value is
+	// unaligned simply clone to an aligned byte array.
+	unaligned := brokenUnaligned && uintptr(unsafe.Pointer(&value[0]))&3 != 0
+
+	if unaligned {
+		value = cloneBytes(value)
+	}
+
 	// If this is a writable transaction then we need to copy the bucket entry.
 	// Read-only transactions can point directly at the mmap entry.
-	if b.tx.writable {
+	if b.tx.writable && !unaligned {
 		child.bucket = &bucket{}
 		*child.bucket = *(*bucket)(unsafe.Pointer(&value[0]))
 	} else {
@@ -167,9 +175,8 @@ func (b *Bucket) CreateBucket(key []byte) (*Bucket, error) {
 	if bytes.Equal(key, k) {
 		if (flags & bucketLeafFlag) != 0 {
 			return nil, ErrBucketExists
-		} else {
-			return nil, ErrIncompatibleValue
 		}
+		return nil, ErrIncompatibleValue
 	}
 
 	// Create empty, inline bucket.
@@ -329,6 +336,28 @@ func (b *Bucket) Delete(key []byte) error {
 	return nil
 }
 
+// Sequence returns the current integer for the bucket without incrementing it.
+func (b *Bucket) Sequence() uint64 { return b.bucket.sequence }
+
+// SetSequence updates the sequence number for the bucket.
+func (b *Bucket) SetSequence(v uint64) error {
+	if b.tx.db == nil {
+		return ErrTxClosed
+	} else if !b.Writable() {
+		return ErrTxNotWritable
+	}
+
+	// Materialize the root node if it hasn't been already so that the
+	// bucket will be saved during commit.
+	if b.rootNode == nil {
+		_ = b.node(b.root, nil)
+	}
+
+	// Increment and return the sequence.
+	b.bucket.sequence = v
+	return nil
+}
+
 // NextSequence returns an autoincrementing integer for the bucket.
 func (b *Bucket) NextSequence() (uint64, error) {
 	if b.tx.db == nil {

cmd/vendor/github.com/boltdb/bolt/db.go → cmd/vendor/github.com/coreos/bbolt/db.go

@@ -8,6 +8,7 @@ import (
 	"os"
 	"runtime"
 	"runtime/debug"
+	"sort"
 	"strings"
 	"sync"
 	"time"
@@ -61,6 +62,11 @@ type DB struct {
 	// THIS IS UNSAFE. PLEASE USE WITH CAUTION.
 	NoSync bool
 
+	// When true, skips syncing freelist to disk. This improves the database
+	// write performance under normal operation, but requires a full database
+	// re-sync during recovery.
+	NoFreelistSync bool
+
 	// When true, skips the truncate call when growing the database.
 	// Setting this to true is only safe on non-ext3/ext4 systems.
 	// Skipping truncation avoids preallocation of hard drive space and
@@ -156,6 +162,7 @@ func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
 	}
 	db.NoGrowSync = options.NoGrowSync
 	db.MmapFlags = options.MmapFlags
+	db.NoFreelistSync = options.NoFreelistSync
 
 	// Set default values for later DB operations.
 	db.MaxBatchSize = DefaultMaxBatchSize
@@ -232,9 +239,14 @@ func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
 		return nil, err
 	}
 
-	// Read in the freelist.
-	db.freelist = newFreelist()
-	db.freelist.read(db.page(db.meta().freelist))
+	if db.NoFreelistSync {
+		db.freelist = newFreelist()
+		db.freelist.readIDs(db.freepages())
+	} else {
+		// Read in the freelist.
+		db.freelist = newFreelist()
+		db.freelist.read(db.page(db.meta().freelist))
+	}
 
 	// Mark the database as opened and return.
 	return db, nil
@@ -526,21 +538,36 @@ func (db *DB) beginRWTx() (*Tx, error) {
 	t := &Tx{writable: true}
 	t.init(db)
 	db.rwtx = t
+	db.freePages()
+	return t, nil
+}
 
-	// Free any pages associated with closed read-only transactions.
-	var minid txid = 0xFFFFFFFFFFFFFFFF
-	for _, t := range db.txs {
-		if t.meta.txid < minid {
-			minid = t.meta.txid
-		}
+// freePages releases any pages associated with closed read-only transactions.
+func (db *DB) freePages() {
+	// Free all pending pages prior to earliest open transaction.
+	sort.Sort(txsById(db.txs))
+	minid := txid(0xFFFFFFFFFFFFFFFF)
+	if len(db.txs) > 0 {
+		minid = db.txs[0].meta.txid
 	}
 	if minid > 0 {
 		db.freelist.release(minid - 1)
 	}
-
-	return t, nil
+	// Release unused txid extents.
+	for _, t := range db.txs {
+		db.freelist.releaseRange(minid, t.meta.txid-1)
+		minid = t.meta.txid + 1
+	}
+	db.freelist.releaseRange(minid, txid(0xFFFFFFFFFFFFFFFF))
+	// Any page both allocated and freed in an extent is safe to release.
 }
 
+type txsById []*Tx
+
+func (t txsById) Len() int           { return len(t) }
+func (t txsById) Swap(i, j int)      { t[i], t[j] = t[j], t[i] }
+func (t txsById) Less(i, j int) bool { return t[i].meta.txid < t[j].meta.txid }
+
 // removeTx removes a transaction from the database.
 func (db *DB) removeTx(tx *Tx) {
 	// Release the read lock on the mmap.
@@ -552,7 +579,10 @@ func (db *DB) removeTx(tx *Tx) {
 	// Remove the transaction.
 	for i, t := range db.txs {
 		if t == tx {
-			db.txs = append(db.txs[:i], db.txs[i+1:]...)
+			last := len(db.txs) - 1
+			db.txs[i] = db.txs[last]
+			db.txs[last] = nil
+			db.txs = db.txs[:last]
 			break
 		}
 	}
@@ -823,7 +853,7 @@ func (db *DB) meta() *meta {
 }
 
 // allocate returns a contiguous block of memory starting at a given page.
-func (db *DB) allocate(count int) (*page, error) {
+func (db *DB) allocate(txid txid, count int) (*page, error) {
 	// Allocate a temporary buffer for the page.
 	var buf []byte
 	if count == 1 {
@@ -835,7 +865,7 @@ func (db *DB) allocate(count int) (*page, error) {
 	p.overflow = uint32(count - 1)
 
 	// Use pages from the freelist if they are available.
-	if p.id = db.freelist.allocate(count); p.id != 0 {
+	if p.id = db.freelist.allocate(txid, count); p.id != 0 {
 		return p, nil
 	}
 
@@ -890,6 +920,38 @@ func (db *DB) IsReadOnly() bool {
 	return db.readOnly
 }
 
+func (db *DB) freepages() []pgid {
+	tx, err := db.beginTx()
+	defer func() {
+		err = tx.Rollback()
+		if err != nil {
+			panic("freepages: failed to rollback tx")
+		}
+	}()
+	if err != nil {
+		panic("freepages: failed to open read only tx")
+	}
+
+	reachable := make(map[pgid]*page)
+	nofreed := make(map[pgid]bool)
+	ech := make(chan error)
+	go func() {
+		for e := range ech {
+			panic(fmt.Sprintf("freepages: failed to get all reachable pages (%v)", e))
+		}
+	}()
+	tx.checkBucket(&tx.root, reachable, nofreed, ech)
+	close(ech)
+
+	var fids []pgid
+	for i := pgid(2); i < db.meta().pgid; i++ {
+		if _, ok := reachable[i]; !ok {
+			fids = append(fids, i)
+		}
+	}
+	return fids
+}
+
 // Options represents the options that can be set when opening a database.
 type Options struct {
 	// Timeout is the amount of time to wait to obtain a file lock.
@@ -900,6 +962,10 @@ type Options struct {
 	// Sets the DB.NoGrowSync flag before memory mapping the file.
 	NoGrowSync bool
 
+	// Do not sync freelist to disk. This improves the database write performance
+	// under normal operation, but requires a full database re-sync during recovery.
+	NoFreelistSync bool
+
 	// Open database in read-only mode. Uses flock(..., LOCK_SH |LOCK_NB) to
 	// grab a shared lock (UNIX).
 	ReadOnly bool
@@ -952,7 +1018,7 @@ func (s *Stats) Sub(other *Stats) Stats {
 	diff.PendingPageN = s.PendingPageN
 	diff.FreeAlloc = s.FreeAlloc
 	diff.FreelistInuse = s.FreelistInuse
-	diff.TxN = other.TxN - s.TxN
+	diff.TxN = s.TxN - other.TxN
 	diff.TxStats = s.TxStats.Sub(&other.TxStats)
 	return diff
 }

cmd/vendor/github.com/boltdb/bolt/freelist.go → cmd/vendor/github.com/coreos/bbolt/freelist.go

@@ -6,25 +6,41 @@ import (
 	"unsafe"
 )
 
+
+// txPending holds a list of pgids and corresponding allocation txns
+// that are pending to be freed.
+type txPending struct {
+	ids              []pgid
+	alloctx          []txid // txids allocating the ids
+	lastReleaseBegin txid   // beginning txid of last matching releaseRange
+}
+
 // freelist represents a list of all pages that are available for allocation.
 // It also tracks pages that have been freed but are still in use by open transactions.
 type freelist struct {
-	ids     []pgid          // all free and available free page ids.
-	pending map[txid][]pgid // mapping of soon-to-be free page ids by tx.
-	cache   map[pgid]bool   // fast lookup of all free and pending page ids.
+	ids     []pgid              // all free and available free page ids.
+	allocs  map[pgid]txid       // mapping of txid that allocated a pgid.
+	pending map[txid]*txPending // mapping of soon-to-be free page ids by tx.
+	cache   map[pgid]bool       // fast lookup of all free and pending page ids.
 }
 
 // newFreelist returns an empty, initialized freelist.
 func newFreelist() *freelist {
 	return &freelist{
-		pending: make(map[txid][]pgid),
+		allocs:  make(map[pgid]txid),
+		pending: make(map[txid]*txPending),
 		cache:   make(map[pgid]bool),
 	}
 }
 
 // size returns the size of the page after serialization.
 func (f *freelist) size() int {
-	return pageHeaderSize + (int(unsafe.Sizeof(pgid(0))) * f.count())
+	n := f.count()
+	if n >= 0xFFFF {
+		// The first element will be used to store the count. See freelist.write.
+		n++
+	}
+	return pageHeaderSize + (int(unsafe.Sizeof(pgid(0))) * n)
 }
 
 // count returns count of pages on the freelist
@@ -40,27 +56,26 @@ func (f *freelist) free_count() int {
 // pending_count returns count of pending pages
 func (f *freelist) pending_count() int {
 	var count int
-	for _, list := range f.pending {
-		count += len(list)
+	for _, txp := range f.pending {
+		count += len(txp.ids)
 	}
 	return count
 }
 
-// all returns a list of all free ids and all pending ids in one sorted list.
-func (f *freelist) all() []pgid {
-	m := make(pgids, 0)
-
-	for _, list := range f.pending {
-		m = append(m, list...)
+// copyall copies into dst a list of all free ids and all pending ids in one sorted list.
+// f.count returns the minimum length required for dst.
+func (f *freelist) copyall(dst []pgid) {
+	m := make(pgids, 0, f.pending_count())
+	for _, txp := range f.pending {
+		m = append(m, txp.ids...)
 	}
-
 	sort.Sort(m)
-	return pgids(f.ids).merge(m)
+	mergepgids(dst, f.ids, m)
 }
 
 // allocate returns the starting page id of a contiguous list of pages of a given size.
 // If a contiguous block cannot be found then 0 is returned.
-func (f *freelist) allocate(n int) pgid {
+func (f *freelist) allocate(txid txid, n int) pgid {
 	if len(f.ids) == 0 {
 		return 0
 	}
@@ -93,7 +108,7 @@ func (f *freelist) allocate(n int) pgid {
 			for i := pgid(0); i < pgid(n); i++ {
 				delete(f.cache, initial+i)
 			}
-
+			f.allocs[initial] = txid
 			return initial
 		}
 
@@ -110,28 +125,73 @@ func (f *freelist) free(txid txid, p *page) {
 	}
 
 	// Free page and all its overflow pages.
-	var ids = f.pending[txid]
+	txp := f.pending[txid]
+	if txp == nil {
+		txp = &txPending{}
+		f.pending[txid] = txp
+	}
+	allocTxid, ok := f.allocs[p.id]
+	if ok {
+		delete(f.allocs, p.id)
+	} else if (p.flags & (freelistPageFlag | metaPageFlag)) != 0 {
+		// Safe to claim txid as allocating since these types are private to txid.
+		allocTxid = txid
+	}
+
 	for id := p.id; id <= p.id+pgid(p.overflow); id++ {
 		// Verify that page is not already free.
 		if f.cache[id] {
 			panic(fmt.Sprintf("page %d already freed", id))
 		}
-
 		// Add to the freelist and cache.
-		ids = append(ids, id)
+		txp.ids = append(txp.ids, id)
+		txp.alloctx = append(txp.alloctx, allocTxid)
 		f.cache[id] = true
 	}
-	f.pending[txid] = ids
 }
 
 // release moves all page ids for a transaction id (or older) to the freelist.
 func (f *freelist) release(txid txid) {
 	m := make(pgids, 0)
-	for tid, ids := range f.pending {
+	for tid, txp := range f.pending {
 		if tid <= txid {
 			// Move transaction's pending pages to the available freelist.
 			// Don't remove from the cache since the page is still free.
-			m = append(m, ids...)
+			m = append(m, txp.ids...)
+			delete(f.pending, tid)
+		}
+	}
+	sort.Sort(m)
+	f.ids = pgids(f.ids).merge(m)
+}
+
+// releaseRange moves pending pages allocated within an extent [begin,end] to the free list.
+func (f *freelist) releaseRange(begin, end txid) {
+	if begin > end {
+		return
+	}
+	var m pgids
+	for tid, txp := range f.pending {
+		if tid < begin || tid > end {
+			continue
+		}
+		// Don't recompute freed pages if ranges haven't updated.
+		if txp.lastReleaseBegin == begin {
+			continue
+		}
+		for i := 0; i < len(txp.ids); i++ {
+			if atx := txp.alloctx[i]; atx < begin || atx > end {
+				continue
+			}
+			m = append(m, txp.ids[i])
+			txp.ids[i] = txp.ids[len(txp.ids)-1]
+			txp.ids = txp.ids[:len(txp.ids)-1]
+			txp.alloctx[i] = txp.alloctx[len(txp.alloctx)-1]
+			txp.alloctx = txp.alloctx[:len(txp.alloctx)-1]
+			i--
+		}
+		txp.lastReleaseBegin = begin
+		if len(txp.ids) == 0 {
 			delete(f.pending, tid)
 		}
 	}
@@ -142,12 +202,29 @@ func (f *freelist) release(txid txid) {
 // rollback removes the pages from a given pending tx.
 func (f *freelist) rollback(txid txid) {
 	// Remove page ids from cache.
-	for _, id := range f.pending[txid] {
-		delete(f.cache, id)
+	txp := f.pending[txid]
+	if txp == nil {
+		return
 	}
-
-	// Remove pages from pending list.
+	var m pgids
+	for i, pgid := range txp.ids {
+		delete(f.cache, pgid)
+		tx := txp.alloctx[i]
+		if tx == 0 {
+			continue
+		}
+		if tx != txid {
+			// Pending free aborted; restore page back to alloc list.
+			f.allocs[pgid] = tx
+		} else {
+			// Freed page was allocated by this txn; OK to throw away.
+			m = append(m, pgid)
+		}
+	}
+	// Remove pages from pending list and mark as free if allocated by txid.
 	delete(f.pending, txid)
+	sort.Sort(m)
+	f.ids = pgids(f.ids).merge(m)
 }
 
 // freed returns whether a given page is in the free list.
@@ -181,27 +258,33 @@ func (f *freelist) read(p *page) {
 	f.reindex()
 }
 
+// read initializes the freelist from a given list of ids.
+func (f *freelist) readIDs(ids []pgid) {
+	f.ids = ids
+	f.reindex()
+}
+
 // write writes the page ids onto a freelist page. All free and pending ids are
 // saved to disk since in the event of a program crash, all pending ids will
 // become free.
 func (f *freelist) write(p *page) error {
 	// Combine the old free pgids and pgids waiting on an open transaction.
-	ids := f.all()
 
 	// Update the header flag.
 	p.flags |= freelistPageFlag
 
 	// The page.count can only hold up to 64k elements so if we overflow that
 	// number then we handle it by putting the size in the first element.
-	if len(ids) == 0 {
-		p.count = uint16(len(ids))
-	} else if len(ids) < 0xFFFF {
-		p.count = uint16(len(ids))
-		copy(((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[:], ids)
+	lenids := f.count()
+	if lenids == 0 {
+		p.count = uint16(lenids)
+	} else if lenids < 0xFFFF {
+		p.count = uint16(lenids)
+		f.copyall(((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[:])
 	} else {
 		p.count = 0xFFFF
-		((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[0] = pgid(len(ids))
-		copy(((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[1:], ids)
+		((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[0] = pgid(lenids)
+		f.copyall(((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[1:])
 	}
 
 	return nil
@@ -213,8 +296,8 @@ func (f *freelist) reload(p *page) {
 
 	// Build a cache of only pending pages.
 	pcache := make(map[pgid]bool)
-	for _, pendingIDs := range f.pending {
-		for _, pendingID := range pendingIDs {
+	for _, txp := range f.pending {
+		for _, pendingID := range txp.ids {
 			pcache[pendingID] = true
 		}
 	}
@@ -236,12 +319,12 @@ func (f *freelist) reload(p *page) {
 
 // reindex rebuilds the free cache based on available and pending free lists.
 func (f *freelist) reindex() {
-	f.cache = make(map[pgid]bool)
+	f.cache = make(map[pgid]bool, len(f.ids))
 	for _, id := range f.ids {
 		f.cache[id] = true
 	}
-	for _, pendingIDs := range f.pending {
-		for _, pendingID := range pendingIDs {
+	for _, txp := range f.pending {
+		for _, pendingID := range txp.ids {
 			f.cache[pendingID] = true
 		}
 	}

cmd/vendor/github.com/boltdb/bolt/page.go → cmd/vendor/github.com/coreos/bbolt/page.go

@@ -145,12 +145,33 @@ func (a pgids) merge(b pgids) pgids {
 	// Return the opposite slice if one is nil.
 	if len(a) == 0 {
 		return b
-	} else if len(b) == 0 {
+	}
+	if len(b) == 0 {
 		return a
 	}
+	merged := make(pgids, len(a)+len(b))
+	mergepgids(merged, a, b)
+	return merged
+}
+
+// mergepgids copies the sorted union of a and b into dst.
+// If dst is too small, it panics.
+func mergepgids(dst, a, b pgids) {
+	if len(dst) < len(a)+len(b) {
+		panic(fmt.Errorf("mergepgids bad len %d < %d + %d", len(dst), len(a), len(b)))
+	}
+	// Copy in the opposite slice if one is nil.
+	if len(a) == 0 {
+		copy(dst, b)
+		return
+	}
+	if len(b) == 0 {
+		copy(dst, a)
+		return
+	}
 
-	// Create a list to hold all elements from both lists.
-	merged := make(pgids, 0, len(a)+len(b))
+	// Merged will hold all elements from both lists.
+	merged := dst[:0]
 
 	// Assign lead to the slice with a lower starting value, follow to the higher value.
 	lead, follow := a, b
@@ -172,7 +193,5 @@ func (a pgids) merge(b pgids) pgids {
 	}
 
 	// Append what's left in follow.
-	merged = append(merged, follow...)
-
-	return merged
+	_ = append(merged, follow...)
 }

cmd/vendor/github.com/boltdb/bolt/tx.go → cmd/vendor/github.com/coreos/bbolt/tx.go

@@ -169,26 +169,9 @@ func (tx *Tx) Commit() error {
 	// Free the old root bucket.
 	tx.meta.root.root = tx.root.root
 
-	opgid := tx.meta.pgid
-
-	// Free the freelist and allocate new pages for it. This will overestimate
-	// the size of the freelist but not underestimate the size (which would be bad).
-	tx.db.freelist.free(tx.meta.txid, tx.db.page(tx.meta.freelist))
-	p, err := tx.allocate((tx.db.freelist.size() / tx.db.pageSize) + 1)
-	if err != nil {
-		tx.rollback()
-		return err
-	}
-	if err := tx.db.freelist.write(p); err != nil {
-		tx.rollback()
-		return err
-	}
-	tx.meta.freelist = p.id
-
-	// If the high water mark has moved up then attempt to grow the database.
-	if tx.meta.pgid > opgid {
-		if err := tx.db.grow(int(tx.meta.pgid+1) * tx.db.pageSize); err != nil {
-			tx.rollback()
+	if !tx.db.NoFreelistSync {
+		err := tx.commitFreelist()
+		if err != nil {
 			return err
 		}
 	}
@@ -235,6 +218,33 @@ func (tx *Tx) Commit() error {
 	return nil
 }
 
+func (tx *Tx) commitFreelist() error {
+	opgid := tx.meta.pgid
+
+	// Free the freelist and allocate new pages for it. This will overestimate
+	// the size of the freelist but not underestimate the size (which would be bad).
+	tx.db.freelist.free(tx.meta.txid, tx.db.page(tx.meta.freelist))
+	p, err := tx.allocate((tx.db.freelist.size() / tx.db.pageSize) + 1)
+	if err != nil {
+		tx.rollback()
+		return err
+	}
+	if err := tx.db.freelist.write(p); err != nil {
+		tx.rollback()
+		return err
+	}
+	tx.meta.freelist = p.id
+	// If the high water mark has moved up then attempt to grow the database.
+	if tx.meta.pgid > opgid {
+		if err := tx.db.grow(int(tx.meta.pgid+1) * tx.db.pageSize); err != nil {
+			tx.rollback()
+			return err
+		}
+	}
+
+	return nil
+}
+
 // Rollback closes the transaction and ignores all previous updates. Read-only
 // transactions must be rolled back and not committed.
 func (tx *Tx) Rollback() error {
@@ -381,7 +391,9 @@ func (tx *Tx) Check() <-chan error {
 func (tx *Tx) check(ch chan error) {
 	// Check if any pages are double freed.
 	freed := make(map[pgid]bool)
-	for _, id := range tx.db.freelist.all() {
+	all := make([]pgid, tx.db.freelist.count())
+	tx.db.freelist.copyall(all)
+	for _, id := range all {
 		if freed[id] {
 			ch <- fmt.Errorf("page %d: already freed", id)
 		}
@@ -392,8 +404,10 @@ func (tx *Tx) check(ch chan error) {
 	reachable := make(map[pgid]*page)
 	reachable[0] = tx.page(0) // meta0
 	reachable[1] = tx.page(1) // meta1
-	for i := uint32(0); i <= tx.page(tx.meta.freelist).overflow; i++ {
-		reachable[tx.meta.freelist+pgid(i)] = tx.page(tx.meta.freelist)
+	if !tx.DB().NoFreelistSync {
+		for i := uint32(0); i <= tx.page(tx.meta.freelist).overflow; i++ {
+			reachable[tx.meta.freelist+pgid(i)] = tx.page(tx.meta.freelist)
+		}
 	}
 
 	// Recursively check buckets.
@@ -451,7 +465,7 @@ func (tx *Tx) checkBucket(b *Bucket, reachable map[pgid]*page, freed map[pgid]bo
 
 // allocate returns a contiguous block of memory starting at a given page.
 func (tx *Tx) allocate(count int) (*page, error) {
-	p, err := tx.db.allocate(count)
+	p, err := tx.db.allocate(tx.meta.txid, count)
 	if err != nil {
 		return nil, err
 	}

etcdctl/ctlv3/command/snapshot_command.go

@@ -27,7 +27,7 @@ import (
 	"reflect"
 	"strings"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 	"github.com/coreos/etcd/etcdserver"
 	"github.com/coreos/etcd/etcdserver/etcdserverpb"
 	"github.com/coreos/etcd/etcdserver/membership"

glide.lock

@@ -1,5 +1,5 @@
-hash: 1c01233ff3c05df62210db3ec7bd51114f9b22401187e8e749a3ee5ea99f8594
-updated: 2017-06-22T14:26:04.148865868-07:00
+hash: 4151c7de891aaf3c611392ecd302c88fd7b6ea01f494bdb0900eb2b2009c2072
+updated: 2017-07-05T14:33:35.042371004-07:00
 imports:
 - name: github.com/beorn7/perks
   version: 4c0e84591b9aa9e6dcfdf3e020114cd81f89d5f9
@@ -7,10 +7,10 @@ imports:
   - quantile
 - name: github.com/bgentry/speakeasy
   version: 4aabc24848ce5fd31929f7d1e4ea74d3709c14cd
-- name: github.com/boltdb/bolt
-  version: 583e8937c61f1af6513608ccc75c97b6abdf4ff9
 - name: github.com/cockroachdb/cmux
   version: 112f0506e7743d64a6eb8fedbcff13d9979bbf92
+- name: github.com/coreos/bbolt
+  version: ad39960eb40bb33c9bda31bed2eaf4fdda15efe6
 - name: github.com/coreos/go-semver
   version: 8ab6407b697782a06568d4b7f1db25550ec2e4c6
   subpackages:

glide.yaml

@@ -2,8 +2,8 @@ package: github.com/coreos/etcd
 import:
 - package: github.com/bgentry/speakeasy
   version: v0.1.0
-- package: github.com/boltdb/bolt
-  version: v1.3.0
+- package: github.com/coreos/bbolt
+  version: ad39960eb40bb33c9bda31bed2eaf4fdda15efe6
 - package: github.com/cockroachdb/cmux
   version: 112f0506e7743d64a6eb8fedbcff13d9979bbf92
 - package: github.com/coreos/go-semver

mvcc/backend/backend.go

@@ -25,7 +25,7 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 	"github.com/coreos/pkg/capnslog"
 )
 

mvcc/backend/backend_test.go

@@ -22,7 +22,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 func TestBackendClose(t *testing.T) {

mvcc/backend/batch_tx.go

@@ -22,7 +22,7 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 type BatchTx interface {

mvcc/backend/batch_tx_test.go

@@ -19,7 +19,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 func TestBatchTxPut(t *testing.T) {

mvcc/backend/config_default.go

@@ -16,7 +16,7 @@
 
 package backend
 
-import "github.com/boltdb/bolt"
+import bolt "github.com/coreos/bbolt"
 
 var boltOpenOptions *bolt.Options = nil
 

mvcc/backend/config_linux.go

@@ -17,7 +17,7 @@ package backend
 import (
 	"syscall"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 // syscall.MAP_POPULATE on linux 2.6.23+ does sequential read-ahead
@@ -27,7 +27,8 @@ import (
 // (https://github.com/torvalds/linux/releases/tag/v2.6.23), mmap might
 // silently ignore this flag. Please update your kernel to prevent this.
 var boltOpenOptions = &bolt.Options{
-	MmapFlags: syscall.MAP_POPULATE,
+	MmapFlags:      syscall.MAP_POPULATE,
+	NoFreelistSync: true,
 }
 
 func (bcfg *BackendConfig) mmapSize() int { return int(bcfg.MmapSize) }

mvcc/backend/config_windows.go

@@ -16,7 +16,7 @@
 
 package backend
 
-import "github.com/boltdb/bolt"
+import bolt "github.com/coreos/bbolt"
 
 var boltOpenOptions *bolt.Options = nil
 

mvcc/backend/read_tx.go

@@ -19,7 +19,7 @@ import (
 	"math"
 	"sync"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 // safeRangeBucket is a hack to avoid inadvertently reading duplicate keys;

tools/etcd-dump-db/backend.go

@@ -18,7 +18,7 @@ import (
 	"fmt"
 	"path/filepath"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 	"github.com/coreos/etcd/mvcc"
 	"github.com/coreos/etcd/mvcc/backend"
 )