Fix issues while truncating the file (#2003)

Co-authored-by: Vikas Bhansali <64532198+vibhansa-msft@users.noreply.github.com>
Co-authored-by: vibhansa <vibhansa@microsoft.com>

Author: syeleti-msft

.golangci.yml

@@ -54,4 +54,4 @@ formatters:
     paths:
       - third_party$
       - builtin$
-      - examples$
+      - examples$

CHANGELOG.md

@@ -3,6 +3,10 @@
 - Fail file open operation if the file being downloaded by file-cache can not fit in available disk space (either configured by user or computed implicitly by blobfuse). User application will receive ENOSPC (no space left on device) in response to file open call.
 - Mount will fail if FNS account is mounted as HNS account.
 - Redirect Stack trace to log file (WORK_DIR/mount_path.pid.trace) instead of console in case of panic for better debuggability.
+- Truncating the file in file_cache resulting in OOM panic by go-runtime in some scenarios.
+- Open file error(No BlockList error) in block_cache when file is truncated before to less than 256MiB.
+- Prevent reusing the same block ID in truncate operation which could lead to issues.
+
 
 ## 2.5.0 (2025-07-17)
 **Bug Fixes**

azure-pipeline-templates/verbose-tests.yml

@@ -221,6 +221,20 @@ steps:
 
 #--------------------------------------- Tests: End to end tests with different File Cache configurations ------------------------------------------
 
+  - template: e2e-tests-spcl.yml
+    parameters:
+      conf_template: azure_key_directio.yaml
+      config_file: ${{ parameters.config }}
+      adls: ${{ parameters.adls }}
+      account_name: ${{ parameters.account_name }}
+      account_key: ${{ parameters.account_key }}
+      account_type: ${{ parameters.account_type }}
+      idstring: "${{ parameters.service }} with File-cache"
+      distro_name: ${{ parameters.distro_name }}
+      quick_test: false
+      verbose_log: ${{ parameters.verbose_log }}
+      clone: false
+
   - template: e2e-tests-spcl.yml
     parameters:
       conf_template: azure_key_lru_purge.yaml

common/types.go

@@ -154,12 +154,11 @@ type LogConfig struct {
 	Tag         string // logging tag which can be either blobfuse2 or bfusemon
 }
 
-// Flags for blocks
+// Flags for block
 const (
 	BlockFlagUnknown uint16 = iota
 	DirtyBlock
 	TruncatedBlock
-	RemovedBlocks
 )
 
 type Block struct {
@@ -181,14 +180,11 @@ func (block *Block) Truncated() bool {
 	return block.Flags.IsSet(TruncatedBlock)
 }
 
-func (block *Block) Removed() bool {
-	return block.Flags.IsSet(RemovedBlocks)
-}
-
 // Flags for block offset list
 const (
-	BolFlagUnknown uint16 = iota
-	SmallFile
+	BlobFlagUnknown     uint16 = iota
+	BlobFlagHasNoBlocks        // set if the blob does not have any blocks
+	BlobFlagBlockListModified
 )
 
 // list that holds blocks containing ids and corresponding offsets
@@ -200,13 +196,39 @@ type BlockOffsetList struct {
 	Mtime         time.Time
 }
 
-// Dirty : Handle is dirty or not
-func (bol *BlockOffsetList) SmallFile() bool {
-	return bol.Flags.IsSet(SmallFile)
+func (bol *BlockOffsetList) HasNoBlocks() bool {
+	return len(bol.BlockList) == 0
+}
+
+func (bol *BlockOffsetList) IsBlockListModified() bool {
+	return bol.Flags.IsSet(BlobFlagBlockListModified)
+}
+
+func (bol *BlockOffsetList) ValidateBlockListAgainstFileSize(fileSize int64) bool {
+	if bol.HasNoBlocks() {
+		return fileSize == 0
+	}
+	if bol.BlockList[len(bol.BlockList)-1].EndIndex != fileSize {
+		return false
+	}
+	return true
+}
+
+func (bol *BlockOffsetList) HasAllBlocksWithSameBlockSize() (blockSize int64, ok bool) {
+	if bol.HasNoBlocks() {
+		return 0, true
+	}
+	blockSize = bol.BlockList[0].EndIndex - bol.BlockList[0].StartIndex
+	for _, blk := range bol.BlockList {
+		if blk.EndIndex-blk.StartIndex != blockSize {
+			return 0, false
+		}
+	}
+	return blockSize, true
 }
 
 // return true if item found and index of the item
-func (bol BlockOffsetList) BinarySearch(offset int64) (bool, int) {
+func (bol *BlockOffsetList) BinarySearch(offset int64) (bool, int) {
 	lowerBound := 0
 	size := len(bol.BlockList)
 	higherBound := size - 1

component/attr_cache/attr_cache.go

@@ -494,7 +494,7 @@ func (ac *AttrCache) TruncateFile(options internal.TruncateFileOptions) error {
 		// no need to truncate the name of the file
 		value, found := ac.cacheMap[options.Name]
 		if found && value.valid() && value.exists() {
-			value.setSize(options.Size)
+			value.setSize(options.NewSize)
 		}
 		// todo: invalidating path here rather than updating with etag
 		// due to some changes that are required in az storage comp which

component/attr_cache/attr_cache_test.go

@@ -792,7 +792,7 @@ func (suite *attrCacheTestSuite) TestTruncateFile() {
 	path := "a"
 	size := 1024
 
-	options := internal.TruncateFileOptions{Name: path, Size: int64(size)}
+	options := internal.TruncateFileOptions{Name: path, NewSize: int64(size)}
 
 	// Error
 	suite.mock.EXPECT().TruncateFile(options).Return(errors.New("Failed to truncate a file"))

component/azstorage/azstorage.go

@@ -498,11 +498,11 @@ func (az *AzStorage) GetFileBlockOffsets(options internal.GetFileBlockOffsetsOpt
 }
 
 func (az *AzStorage) TruncateFile(options internal.TruncateFileOptions) error {
-	log.Trace("AzStorage::TruncateFile : %s to %d bytes", options.Name, options.Size)
-	err := az.storage.TruncateFile(options.Name, options.Size)
+	log.Trace("AzStorage::TruncateFile : %s to %d bytes", options.Name, options.NewSize)
+	err := az.storage.TruncateFile(options)
 
 	if err == nil {
-		azStatsCollector.PushEvents(truncateFile, options.Name, map[string]any{size: options.Size})
+		azStatsCollector.PushEvents(truncateFile, options.Name, map[string]any{size: options.NewSize})
 		azStatsCollector.UpdateStats(stats_manager.Increment, truncateFile, (int64)(1))
 	}
 	return err