Speed up fat32 by replacing clusters map with a slice

filesystem/fat32/common_test.go

@@ -467,7 +467,7 @@ func testReadFilesystemData() (info *testFSInfo, err error) {
 				rootDirCluster: 2,         // root is at cluster 2
 				size:           sizeInBytes,
 				maxCluster:     numClusters,
-				clusters:       make(map[uint32]uint32),
+				clusters:       make([]uint32, numClusters+1),
 			}
 		case inClusters && len(clusterLineMatch) > 4:
 			start, err := strconv.Atoi(clusterLineMatch[1])

filesystem/fat32/fat32.go

@@ -228,17 +228,16 @@ func Create(f util.File, size, start, blocksize int64, volumeLabel string) (*Fil
 	fatSecondaryStart := uint64(fatPrimaryStart) + uint64(fatSize)
 	maxCluster := fatSize / 4
 	rootDirCluster := uint32(2)
+	clusters := make([]uint32, maxCluster+1)
+	clusters[rootDirCluster] = eocMarker
 	fat := table{
 		fatID:          fatID,
 		eocMarker:      eocMarker,
 		unusedMarker:   unusedMarker,
 		size:           fatSize,
 		rootDirCluster: rootDirCluster,
-		clusters: map[uint32]uint32{
-			// when we start, there is just one directory with a single cluster
-			rootDirCluster: eocMarker,
-		},
-		maxCluster: maxCluster,
+		clusters:       clusters,
+		maxCluster:     maxCluster,
 	}
 
 	// where does our data start?
@@ -694,10 +693,9 @@ func (fs *FileSystem) getClusterList(firstCluster uint32) ([]uint32, error) {
 	// first, get the chain of clusters
 	complete := false
 	cluster := firstCluster
-	clusters := fs.table.clusters
 
 	// do we even have a valid cluster?
-	if _, ok := clusters[cluster]; !ok {
+	if cluster > fs.table.maxCluster || fs.table.clusters[cluster] == 0 {
 		return nil, fmt.Errorf("invalid start cluster: %d", cluster)
 	}
 
@@ -706,11 +704,13 @@ func (fs *FileSystem) getClusterList(firstCluster uint32) ([]uint32, error) {
 		// save the current cluster
 		clusterList = append(clusterList, cluster)
 		// get the next cluster
-		newCluster := clusters[cluster]
+		newCluster := fs.table.clusters[cluster]
 		// if it is EOC, we are done
 		switch {
 		case fs.table.isEoc(newCluster):
 			complete = true
+		case newCluster > fs.table.maxCluster:
+			return nil, fmt.Errorf("invalid cluster chain at %d", newCluster)
 		case cluster < 2:
 			return nil, fmt.Errorf("invalid cluster chain at %d", cluster)
 		}
@@ -924,6 +924,10 @@ func (fs *FileSystem) readDirWithMkdir(p string, doMake bool) (*Directory, []*di
 // returns the indexes of clusters to be used in order. If the new size is smaller than
 // the original size, will shrink the chain.
 func (fs *FileSystem) allocateSpace(size uint64, previous uint32) ([]uint32, error) {
+	if previous > fs.table.maxCluster {
+		return nil, fmt.Errorf("invalid cluster chain at %d", previous)
+	}
+
 	var (
 		clusters             []uint32
 		err                  error
@@ -961,12 +965,11 @@ func (fs *FileSystem) allocateSpace(size uint64, previous uint32) ([]uint32, err
 	}
 
 	// get a list of allocated clusters, so we can know which ones are unallocated and therefore allocatable
-	allClusters := fs.table.clusters
 	maxCluster := fs.table.maxCluster
 
 	if extraClusterCount > 0 {
 		for i := uint32(2); i < maxCluster && len(allocated) < extraClusterCount; i++ {
-			if _, ok := allClusters[i]; !ok {
+			if fs.table.clusters[i] == 0 {
 				// these become the same at this point
 				allocated = append(allocated, i)
 			}
@@ -982,12 +985,12 @@ func (fs *FileSystem) allocateSpace(size uint64, previous uint32) ([]uint32, err
 
 		// extend the chain and fill them in
 		if previous > 0 {
-			allClusters[previous] = allocated[0]
+			fs.table.clusters[previous] = allocated[0]
 		}
 		for i := 0; i < lastAlloc; i++ {
-			allClusters[allocated[i]] = allocated[i+1]
+			fs.table.clusters[allocated[i]] = allocated[i+1]
 		}
-		allClusters[allocated[lastAlloc]] = fs.table.eocMarker
+		fs.table.clusters[allocated[lastAlloc]] = fs.table.eocMarker
 
 		// update the FSIS
 		lastAllocatedCluster = allocated[len(allocated)-1]
@@ -1003,13 +1006,21 @@ func (fs *FileSystem) allocateSpace(size uint64, previous uint32) ([]uint32, err
 		}
 		deallocated = clusters[lastAlloc+1:]
 
+		if uint32(lastAlloc) > fs.table.maxCluster || clusters[lastAlloc] > fs.table.maxCluster { //nolint:gosec // lastAlloc is >= 0
+			return nil, fmt.Errorf("invalid cluster chain at %d", lastAlloc)
+		}
+
 		// mark last allocated one as EOC
-		allClusters[clusters[lastAlloc]] = fs.table.eocMarker
+		fs.table.clusters[clusters[lastAlloc]] = fs.table.eocMarker
 
 		// unmark all of the unused ones
 		lastAllocatedCluster = fs.fsis.lastAllocatedCluster
 		for _, cl := range deallocated {
-			allClusters[cl] = fs.table.unusedMarker
+			if cl > fs.table.maxCluster {
+				return nil, fmt.Errorf("invalid cluster chain at %d", cl)
+			}
+
+			fs.table.clusters[cl] = fs.table.unusedMarker
 			if cl == lastAllocatedCluster {
 				lastAllocatedCluster--
 			}

filesystem/fat32/fat32_internal_test.go

@@ -17,6 +17,15 @@ import (
 	in that case, the dataStart is relative to partition, not to disk, so need to read the offset correctly
 */
 
+func clustersFromMap(m map[uint32]uint32, maxCluster uint32) []uint32 {
+	clusters := make([]uint32, maxCluster+1)
+	for k, v := range m {
+		clusters[k] = v
+	}
+
+	return clusters
+}
+
 func getValidFat32FSFull() *FileSystem {
 	fs := getValidFat32FSSmall()
 	fs.table = *getValidFat32Table()
@@ -25,11 +34,12 @@ func getValidFat32FSFull() *FileSystem {
 
 func getValidFat32FSSmall() *FileSystem {
 	eoc := uint32(0xffffffff)
+	maxCluster := uint32(128)
 	fs := &FileSystem{
 		table: table{
 			rootDirCluster: 2,
 			size:           512,
-			maxCluster:     128,
+			maxCluster:     maxCluster,
 			eocMarker:      eoc,
 			/*
 				 map:
@@ -40,8 +50,9 @@ func getValidFat32FSSmall() *FileSystem {
 					 11
 					 15
 					 16-broken
+					 17-broken
 			*/
-			clusters: map[uint32]uint32{
+			clusters: clustersFromMap(map[uint32]uint32{
 				2:  eoc,
 				3:  4,
 				4:  5,
@@ -53,8 +64,9 @@ func getValidFat32FSSmall() *FileSystem {
 				9:  11,
 				11: eoc,
 				15: eoc,
-				16: 0,
-			},
+				16: 1,
+				17: 999,
+			}, maxCluster),
 		},
 		bytesPerCluster: 512,
 		dataStart:       178176,
@@ -80,6 +92,7 @@ func getValidFat32FSSmall() *FileSystem {
 	}
 	return fs
 }
+
 func TestFat32GetClusterList(t *testing.T) {
 	fs := getValidFat32FSSmall()
 
@@ -178,6 +191,7 @@ func TestFat32AllocateSpace(t *testing.T) {
 				 11
 				 15
 				 16-broken
+				 17-broken
 		// recall that 512 bytes per cluster here
 	*/
 	tests := []struct {
@@ -189,9 +203,11 @@ func TestFat32AllocateSpace(t *testing.T) {
 		{500, 2, []uint32{2}, nil},
 		{600, 2, []uint32{2, 12}, nil},
 		{2000, 2, []uint32{2, 12, 13, 14}, nil},
-		{2000, 0, []uint32{12, 13, 14, 17}, nil},
+		{2000, 0, []uint32{12, 13, 14, 18}, nil},
 		{200000000000, 0, nil, fmt.Errorf("no space left on device")},
 		{200000000000, 2, nil, fmt.Errorf("no space left on device")},
+		{2000, 17, nil, fmt.Errorf("unable to get cluster list: invalid cluster chain at 999")},
+		{2000, 999, nil, fmt.Errorf("invalid cluster chain at 999")},
 	}
 	for _, tt := range tests {
 		// reset for each test

filesystem/fat32/table.go

@@ -2,15 +2,15 @@ package fat32
 
 import (
 	"encoding/binary"
-	"reflect"
+	"slices"
 )
 
 // table a FAT32 table
 type table struct {
 	fatID          uint32
 	eocMarker      uint32
 	unusedMarker   uint32
-	clusters       map[uint32]uint32
+	clusters       []uint32
 	rootDirCluster uint32
 	size           uint32
 	maxCluster     uint32
@@ -28,7 +28,7 @@ func (t *table) equal(a *table) bool {
 		t.rootDirCluster == a.rootDirCluster &&
 		t.size == a.size &&
 		t.maxCluster == a.maxCluster &&
-		reflect.DeepEqual(t.clusters, a.clusters)
+		slices.Equal(a.clusters, t.clusters)
 }
 
 /*
@@ -37,12 +37,14 @@ func (t *table) equal(a *table) bool {
 */
 
 func tableFromBytes(b []byte) *table {
+	maxCluster := uint32(len(b) / 4) //nolint:gosec // b isn't big enough to cause an overflow
+
 	t := table{
 		fatID:          binary.LittleEndian.Uint32(b[0:4]),
 		eocMarker:      binary.LittleEndian.Uint32(b[4:8]),
-		size:           uint32(len(b)),
-		clusters:       map[uint32]uint32{},
-		maxCluster:     uint32(len(b) / 4),
+		size:           uint32(len(b)), //nolint:gosec // b isn't big enough to cause an overflow
+		clusters:       make([]uint32, maxCluster+1),
+		maxCluster:     maxCluster,
 		rootDirCluster: 2, // always 2 for FAT32
 	}
 	// just need to map the clusters in
@@ -71,10 +73,7 @@ func (t *table) bytes() []byte {
 	for i := uint32(2); i < numClusters; i++ {
 		bStart := i * 4
 		bEnd := bStart + 4
-		val := uint32(0)
-		if cluster, ok := t.clusters[i]; ok {
-			val = cluster
-		}
+		val := t.clusters[i]
 		binary.LittleEndian.PutUint32(b[bStart:bEnd], val)
 	}
 

filesystem/fat32/table_internal_test.go

@@ -3,6 +3,7 @@ package fat32
 import (
 	"bytes"
 	"os"
+	"slices"
 	"testing"
 
 	"github.com/diskfs/go-diskfs/util"
@@ -16,13 +17,10 @@ const (
 
 func getValidFat32Table() *table {
 	// make a duplicate, in case someone modifies what we return
-	var t = &table{}
+	t := &table{}
 	*t = *fsInfo.table
 	// and because the clusters are copied by reference
-	t.clusters = make(map[uint32]uint32)
-	for k, v := range fsInfo.table.clusters {
-		t.clusters[k] = v
-	}
+	t.clusters = slices.Clone(t.clusters)
 
 	return t
 }