Update go.etcd.io/bbolt to v1.3.2

Relevant changes:

- ectd-io/bbolt#139 update the freelist readIDs
- etcd-io/bbolt#140 add getFreePageIDs
- etcd-io/bbolt#141 use segregated hashmap to boost the freelist allocate and release performance

Signed-off-by: Sebastiaan van Stijn <github@gone.nl>

vendor.conf

@@ -40,7 +40,7 @@ github.com/containerd/ttrpc f02858b1457c5ca3aaec3a0803eb0d59f96e41d6
 github.com/syndtr/gocapability db04d3cc01c8b54962a58ec7e491717d06cfcc16
 gotest.tools v2.1.0
 github.com/google/go-cmp v0.1.0
-go.etcd.io/bbolt v1.3.1-etcd.8
+go.etcd.io/bbolt v1.3.2
 
 # cri dependencies
 github.com/containerd/cri 4dd6735020f5596dd41738f8c4f5cb07fa804c5e # master

vendor/go.etcd.io/bbolt/README.md

@@ -929,6 +929,7 @@ Below is a list of public, open source projects that use Bolt:
 * [ipxed](https://github.com/kelseyhightower/ipxed) - Web interface and api for ipxed.
 * [Ironsmith](https://github.com/timshannon/ironsmith) - A simple, script-driven continuous integration (build - > test -> release) tool, with no external dependencies
 * [Kala](https://github.com/ajvb/kala) - Kala is a modern job scheduler optimized to run on a single node. It is persistent, JSON over HTTP API, ISO 8601 duration notation, and dependent jobs.
+* [Key Value Access Langusge (KVAL)](https://github.com/kval-access-language) - A proposed grammar for key-value datastores offering a bbolt binding.
 * [LedisDB](https://github.com/siddontang/ledisdb) - A high performance NoSQL, using Bolt as optional storage.
 * [lru](https://github.com/crowdriff/lru) - Easy to use Bolt-backed Least-Recently-Used (LRU) read-through cache with chainable remote stores.
 * [mbuckets](https://github.com/abhigupta912/mbuckets) - A Bolt wrapper that allows easy operations on multi level (nested) buckets.

vendor/go.etcd.io/bbolt/db.go

@@ -43,6 +43,16 @@ var defaultPageSize = os.Getpagesize()
 // The time elapsed between consecutive file locking attempts.
 const flockRetryTimeout = 50 * time.Millisecond
 
+// FreelistType is the type of the freelist backend
+type FreelistType string
+
+const (
+	// FreelistArrayType indicates backend freelist type is array
+	FreelistArrayType = FreelistType("array")
+	// FreelistMapType indicates backend freelist type is hashmap
+	FreelistMapType = FreelistType("hashmap")
+)
+
 // DB represents a collection of buckets persisted to a file on disk.
 // All data access is performed through transactions which can be obtained through the DB.
 // All the functions on DB will return a ErrDatabaseNotOpen if accessed before Open() is called.
@@ -70,6 +80,13 @@ type DB struct {
 	// re-sync during recovery.
 	NoFreelistSync bool
 
+	// FreelistType sets the backend freelist type. There are two options. Array which is simple but endures
+	// dramatic performance degradation if database is large and framentation in freelist is common.
+	// The alternative one is using hashmap, it is faster in almost all circumstances
+	// but it doesn't guarantee that it offers the smallest page id available. In normal case it is safe.
+	// The default type is array
+	FreelistType FreelistType
+
 	// 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
@@ -169,6 +186,7 @@ func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
 	db.NoGrowSync = options.NoGrowSync
 	db.MmapFlags = options.MmapFlags
 	db.NoFreelistSync = options.NoFreelistSync
+	db.FreelistType = options.FreelistType
 
 	// Set default values for later DB operations.
 	db.MaxBatchSize = DefaultMaxBatchSize
@@ -283,7 +301,7 @@ func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
 // concurrent accesses being made to the freelist.
 func (db *DB) loadFreelist() {
 	db.freelistLoad.Do(func() {
-		db.freelist = newFreelist()
+		db.freelist = newFreelist(db.FreelistType)
 		if !db.hasSyncedFreelist() {
 			// Reconstruct free list by scanning the DB.
 			db.freelist.readIDs(db.freepages())
@@ -291,7 +309,7 @@ func (db *DB) loadFreelist() {
 			// Read free list from freelist page.
 			db.freelist.read(db.page(db.meta().freelist))
 		}
-		db.stats.FreePageN = len(db.freelist.ids)
+		db.stats.FreePageN = db.freelist.free_count()
 	})
 }
 
@@ -1005,6 +1023,13 @@ type Options struct {
 	// under normal operation, but requires a full database re-sync during recovery.
 	NoFreelistSync bool
 
+	// FreelistType sets the backend freelist type. There are two options. Array which is simple but endures
+	// dramatic performance degradation if database is large and framentation in freelist is common.
+	// The alternative one is using hashmap, it is faster in almost all circumstances
+	// but it doesn't guarantee that it offers the smallest page id available. In normal case it is safe.
+	// The default type is array
+	FreelistType FreelistType
+
 	// Open database in read-only mode. Uses flock(..., LOCK_SH |LOCK_NB) to
 	// grab a shared lock (UNIX).
 	ReadOnly bool
@@ -1036,6 +1061,7 @@ type Options struct {
 var DefaultOptions = &Options{
 	Timeout:      0,
 	NoGrowSync:   false,
+	FreelistType: FreelistArrayType,
 }
 
 // Stats represents statistics about the database.

vendor/go.etcd.io/bbolt/freelist.go

@@ -14,22 +14,54 @@ type txPending struct {
 	lastReleaseBegin txid   // beginning txid of last matching releaseRange
 }
 
+// pidSet holds the set of starting pgids which have the same span size
+type pidSet map[pgid]struct{}
+
 // 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 {
+	freelistType   FreelistType                // freelist type
 	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.
+	freemaps       map[uint64]pidSet           // key is the size of continuous pages(span), value is a set which contains the starting pgids of same size
+	forwardMap     map[pgid]uint64             // key is start pgid, value is its span size
+	backwardMap    map[pgid]uint64             // key is end pgid, value is its span size
+	allocate       func(txid txid, n int) pgid // the freelist allocate func
+	free_count     func() int                  // the function which gives you free page number
+	mergeSpans     func(ids pgids)             // the mergeSpan func
+	getFreePageIDs func() []pgid               // get free pgids func
+	readIDs        func(pgids []pgid)          // readIDs func reads list of pages and init the freelist
 }
 
 // newFreelist returns an empty, initialized freelist.
-func newFreelist() *freelist {
-	return &freelist{
+func newFreelist(freelistType FreelistType) *freelist {
+	f := &freelist{
+		freelistType: freelistType,
 		allocs:       make(map[pgid]txid),
 		pending:      make(map[txid]*txPending),
 		cache:        make(map[pgid]bool),
+		freemaps:     make(map[uint64]pidSet),
+		forwardMap:   make(map[pgid]uint64),
+		backwardMap:  make(map[pgid]uint64),
 	}
+
+	if freelistType == FreelistMapType {
+		f.allocate = f.hashmapAllocate
+		f.free_count = f.hashmapFreeCount
+		f.mergeSpans = f.hashmapMergeSpans
+		f.getFreePageIDs = f.hashmapGetFreePageIDs
+		f.readIDs = f.hashmapReadIDs
+	} else {
+		f.allocate = f.arrayAllocate
+		f.free_count = f.arrayFreeCount
+		f.mergeSpans = f.arrayMergeSpans
+		f.getFreePageIDs = f.arrayGetFreePageIDs
+		f.readIDs = f.arrayReadIDs
+	}
+
+	return f
 }
 
 // size returns the size of the page after serialization.
@@ -47,8 +79,8 @@ func (f *freelist) count() int {
 	return f.free_count() + f.pending_count()
 }
 
-// free_count returns count of free pages
-func (f *freelist) free_count() int {
+// arrayFreeCount returns count of free pages(array version)
+func (f *freelist) arrayFreeCount() int {
 	return len(f.ids)
 }
 
@@ -69,12 +101,12 @@ func (f *freelist) copyall(dst []pgid) {
 		m = append(m, txp.ids...)
 	}
 	sort.Sort(m)
-	mergepgids(dst, f.ids, m)
+	mergepgids(dst, f.getFreePageIDs(), m)
 }
 
-// allocate returns the starting page id of a contiguous list of pages of a given size.
+// arrayAllocate 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(txid txid, n int) pgid {
+func (f *freelist) arrayAllocate(txid txid, n int) pgid {
 	if len(f.ids) == 0 {
 		return 0
 	}
@@ -160,8 +192,7 @@ func (f *freelist) release(txid txid) {
 			delete(f.pending, tid)
 		}
 	}
-	sort.Sort(m)
-	f.ids = pgids(f.ids).merge(m)
+	f.mergeSpans(m)
 }
 
 // releaseRange moves pending pages allocated within an extent [begin,end] to the free list.
@@ -194,8 +225,7 @@ func (f *freelist) releaseRange(begin, end txid) {
 			delete(f.pending, tid)
 		}
 	}
-	sort.Sort(m)
-	f.ids = pgids(f.ids).merge(m)
+	f.mergeSpans(m)
 }
 
 // rollback removes the pages from a given pending tx.
@@ -222,8 +252,7 @@ func (f *freelist) rollback(txid txid) {
 	}
 	// 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)
+	f.mergeSpans(m)
 }
 
 // freed returns whether a given page is in the free list.
@@ -249,21 +278,25 @@ func (f *freelist) read(p *page) {
 		f.ids = nil
 	} else {
 		ids := ((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[idx : idx+count]
-		f.ids = make([]pgid, len(ids))
-		copy(f.ids, ids)
 
+		// copy the ids, so we don't modify on the freelist page directly
+		idsCopy := make([]pgid, count)
+		copy(idsCopy, ids)
 		// Make sure they're sorted.
-		sort.Sort(pgids(f.ids))
-	}
+		sort.Sort(pgids(idsCopy))
 
-	// Rebuild the page cache.
+		f.readIDs(idsCopy)
+	}
+}
+
+// arrayReadIDs initializes the freelist from a given list of ids.
+func (f *freelist) arrayReadIDs(ids []pgid) {
+	f.ids = ids
 	f.reindex()
 }
 
-// read initializes the freelist from a given list of ids.
-func (f *freelist) readIDs(ids []pgid) {
-	f.ids = ids
-	f.reindex()
+func (f *freelist) arrayGetFreePageIDs() []pgid {
+	return f.ids
 }
 
 // write writes the page ids onto a freelist page. All free and pending ids are
@@ -307,22 +340,20 @@ func (f *freelist) reload(p *page) {
 	// Check each page in the freelist and build a new available freelist
 	// with any pages not in the pending lists.
 	var a []pgid
-	for _, id := range f.ids {
+	for _, id := range f.getFreePageIDs() {
 		if !pcache[id] {
 			a = append(a, id)
 		}
 	}
-	f.ids = a
 
-	// Once the available list is rebuilt then rebuild the free cache so that
-	// it includes the available and pending free pages.
-	f.reindex()
+	f.readIDs(a)
 }
 
 // reindex rebuilds the free cache based on available and pending free lists.
 func (f *freelist) reindex() {
-	f.cache = make(map[pgid]bool, len(f.ids))
-	for _, id := range f.ids {
+	ids := f.getFreePageIDs()
+	f.cache = make(map[pgid]bool, len(ids))
+	for _, id := range ids {
 		f.cache[id] = true
 	}
 	for _, txp := range f.pending {
@@ -331,3 +362,9 @@ func (f *freelist) reindex() {
 		}
 	}
 }
+
+// arrayMergeSpans try to merge list of pages(represented by pgids) with existing spans but using array
+func (f *freelist) arrayMergeSpans(ids pgids) {
+	sort.Sort(ids)
+	f.ids = pgids(f.ids).merge(ids)
+}

vendor/go.etcd.io/bbolt/freelist_hmap.go

@@ -0,0 +1,178 @@
+package bbolt
+
+import "sort"
+
+// hashmapFreeCount returns count of free pages(hashmap version)
+func (f *freelist) hashmapFreeCount() int {
+	// use the forwardmap to get the total count
+	count := 0
+	for _, size := range f.forwardMap {
+		count += int(size)
+	}
+	return count
+}
+
+// hashmapAllocate serves the same purpose as arrayAllocate, but use hashmap as backend
+func (f *freelist) hashmapAllocate(txid txid, n int) pgid {
+	if n == 0 {
+		return 0
+	}
+
+	// if we have a exact size match just return short path
+	if bm, ok := f.freemaps[uint64(n)]; ok {
+		for pid := range bm {
+			// remove the span
+			f.delSpan(pid, uint64(n))
+
+			f.allocs[pid] = txid
+
+			for i := pgid(0); i < pgid(n); i++ {
+				delete(f.cache, pid+pgid(i))
+			}
+			return pid
+		}
+	}
+
+	// lookup the map to find larger span
+	for size, bm := range f.freemaps {
+		if size < uint64(n) {
+			continue
+		}
+
+		for pid := range bm {
+			// remove the initial
+			f.delSpan(pid, uint64(size))
+
+			f.allocs[pid] = txid
+
+			remain := size - uint64(n)
+
+			// add remain span
+			f.addSpan(pid+pgid(n), remain)
+
+			for i := pgid(0); i < pgid(n); i++ {
+				delete(f.cache, pid+pgid(i))
+			}
+			return pid
+		}
+	}
+
+	return 0
+}
+
+// hashmapReadIDs reads pgids as input an initial the freelist(hashmap version)
+func (f *freelist) hashmapReadIDs(pgids []pgid) {
+	f.init(pgids)
+
+	// Rebuild the page cache.
+	f.reindex()
+}
+
+// hashmapGetFreePageIDs returns the sorted free page ids
+func (f *freelist) hashmapGetFreePageIDs() []pgid {
+	count := f.free_count()
+	if count == 0 {
+		return nil
+	}
+
+	m := make([]pgid, 0, count)
+	for start, size := range f.forwardMap {
+		for i := 0; i < int(size); i++ {
+			m = append(m, start+pgid(i))
+		}
+	}
+	sort.Sort(pgids(m))
+
+	return m
+}
+
+// hashmapMergeSpans try to merge list of pages(represented by pgids) with existing spans
+func (f *freelist) hashmapMergeSpans(ids pgids) {
+	for _, id := range ids {
+		// try to see if we can merge and update
+		f.mergeWithExistingSpan(id)
+	}
+}
+
+// mergeWithExistingSpan merges pid to the existing free spans, try to merge it backward and forward
+func (f *freelist) mergeWithExistingSpan(pid pgid) {
+	prev := pid - 1
+	next := pid + 1
+
+	preSize, mergeWithPrev := f.backwardMap[prev]
+	nextSize, mergeWithNext := f.forwardMap[next]
+	newStart := pid
+	newSize := uint64(1)
+
+	if mergeWithPrev {
+		//merge with previous span
+		start := prev + 1 - pgid(preSize)
+		f.delSpan(start, preSize)
+
+		newStart -= pgid(preSize)
+		newSize += preSize
+	}
+
+	if mergeWithNext {
+		// merge with next span
+		f.delSpan(next, nextSize)
+		newSize += nextSize
+	}
+
+	f.addSpan(newStart, newSize)
+}
+
+func (f *freelist) addSpan(start pgid, size uint64) {
+	f.backwardMap[start-1+pgid(size)] = size
+	f.forwardMap[start] = size
+	if _, ok := f.freemaps[size]; !ok {
+		f.freemaps[size] = make(map[pgid]struct{})
+	}
+
+	f.freemaps[size][start] = struct{}{}
+}
+
+func (f *freelist) delSpan(start pgid, size uint64) {
+	delete(f.forwardMap, start)
+	delete(f.backwardMap, start+pgid(size-1))
+	delete(f.freemaps[size], start)
+	if len(f.freemaps[size]) == 0 {
+		delete(f.freemaps, size)
+	}
+}
+
+// initial from pgids using when use hashmap version
+// pgids must be sorted
+func (f *freelist) init(pgids []pgid) {
+	if len(pgids) == 0 {
+		return
+	}
+
+	size := uint64(1)
+	start := pgids[0]
+
+	if !sort.SliceIsSorted([]pgid(pgids), func(i, j int) bool { return pgids[i] < pgids[j] }) {
+		panic("pgids not sorted")
+	}
+
+	f.freemaps = make(map[uint64]pidSet)
+	f.forwardMap = make(map[pgid]uint64)
+	f.backwardMap = make(map[pgid]uint64)
+
+	for i := 1; i < len(pgids); i++ {
+		// continuous page
+		if pgids[i] == pgids[i-1]+1 {
+			size++
+		} else {
+			f.addSpan(start, size)
+
+			size = 1
+			start = pgids[i]
+		}
+	}
+
+	// init the tail
+	if size != 0 && start != 0 {
+		f.addSpan(start, size)
+	}
+}