Validate blob hashes, offsets and lengths from the destination (closes #155)

A blob hash read back from the downloaded snapshot database or the store listing was trusted unchecked. A hostile remote could set a hash such as "aa/../../etc" and have a decrypted blob written outside the cache directory, or feed a short or negative value that panicked a command.

blobDiskCache.path now refuses any key with a path separator, and ReadAt rejects a negative offset or length, bounding so a sum cannot overflow past the check. A new isBlobHash helper gates FetchBlob, shallow and deep verify, and restore: buildBlobIndexes rejects every hash from the snapshot database before any fetch. The blobs/ and metadata/ listings skip a non-conforming name, and short-hash prefixes in log and error text go through a panic-safe shortHash helper.

Model: opus-4-8

internal/vaultik/blobcache.go

@@ -6,13 +6,17 @@ import (
 	"io"
 	"os"
 	"path/filepath"
+	"strings"
 	"sync"
 )
 
 // Sentinel errors for blob cache lookups.
 var (
-	errCacheKeyMissing     = errors.New("key not in cache")
-	errCacheReadBeyondBlob = errors.New("read beyond blob size")
+	errCacheKeyMissing      = errors.New("key not in cache")
+	errCacheReadBeyondBlob  = errors.New("read beyond blob size")
+	errCacheKeyHasSeparator = errors.New(
+		"cache key contains a path separator")
+	errCacheNegativeRead = errors.New("negative offset or length")
 )
 
 // blobCacheFileMode is the permission mode for cached blob files.
@@ -74,6 +78,11 @@ func newBlobDiskCache(maxBytes int64) (*blobDiskCache, error) {
 // Put writes blob data to disk cache. Entries larger than maxBytes are
 // silently skipped.
 func (c *blobDiskCache) Put(key string, data []byte) error {
+	p, err := c.path(key)
+	if err != nil {
+		return err
+	}
+
 	entrySize := int64(len(data))
 
 	c.mu.Lock()
@@ -87,11 +96,12 @@ func (c *blobDiskCache) Put(key string, data []byte) error {
 	if e, ok := c.items[key]; ok {
 		c.unlink(e)
 		c.curBytes -= e.size
-		_ = os.Remove(c.path(key))
+		_ = os.Remove(p)
+
 		delete(c.items, key)
 	}
 
-	err := os.WriteFile(c.path(key), data, blobCacheFileMode)
+	err = os.WriteFile(p, data, blobCacheFileMode)
 	if err != nil {
 		return fmt.Errorf("writing blob to cache: %w", err)
 	}
@@ -119,19 +129,26 @@ func (c *blobDiskCache) Put(key string, data []byte) error {
 // disk without buffering its entire plaintext (which may be tens of GB)
 // in RAM.
 func (c *blobDiskCache) PutFromReader(key string, r io.Reader) (int64, error) {
+	p, err := c.path(key)
+	if err != nil {
+		return 0, err
+	}
+
 	c.mu.Lock()
 	// Remove any prior entry first; we'll re-link after the file is
 	// written successfully.
 	if e, ok := c.items[key]; ok {
 		c.unlink(e)
 		c.curBytes -= e.size
-		_ = os.Remove(c.path(key))
+		_ = os.Remove(p)
+
 		delete(c.items, key)
 	}
 	c.mu.Unlock()
 
+	//nolint:gosec // G304: path() rejects keys with a separator
 	f, err := os.OpenFile(
-		c.path(key), os.O_CREATE|os.O_TRUNC|os.O_WRONLY, blobCacheFileMode)
+		p, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, blobCacheFileMode)
 	if err != nil {
 		return 0, fmt.Errorf("creating cache file: %w", err)
 	}
@@ -140,13 +157,13 @@ func (c *blobDiskCache) PutFromReader(key string, r io.Reader) (int64, error) {
 	closeErr := f.Close()
 
 	if copyErr != nil {
-		_ = os.Remove(c.path(key))
+		_ = os.Remove(p)
 
 		return written, fmt.Errorf("streaming to cache file: %w", copyErr)
 	}
 
 	if closeErr != nil {
-		_ = os.Remove(c.path(key))
+		_ = os.Remove(p)
 
 		return written, fmt.Errorf("closing cache file: %w", closeErr)
 	}
@@ -158,7 +175,7 @@ func (c *blobDiskCache) PutFromReader(key string, r io.Reader) (int64, error) {
 	// floor — but the restore path passes math.MaxInt64 as maxBytes
 	// so this branch is effectively unreachable there.
 	if written > c.maxBytes {
-		_ = os.Remove(c.path(key))
+		_ = os.Remove(p)
 
 		return written, nil
 	}
@@ -181,6 +198,11 @@ func (c *blobDiskCache) PutFromReader(key string, r io.Reader) (int64, error) {
 
 // Get reads a cached blob from disk. Returns data and true on hit.
 func (c *blobDiskCache) Get(key string) ([]byte, bool) {
+	p, err := c.path(key)
+	if err != nil {
+		return nil, false
+	}
+
 	c.mu.Lock()
 	c.getCalls++
 
@@ -195,7 +217,8 @@ func (c *blobDiskCache) Get(key string) ([]byte, bool) {
 	c.pushFront(e)
 	c.mu.Unlock()
 
-	data, err := os.ReadFile(c.path(key))
+	//nolint:gosec // G304: path() rejects keys with a separator
+	data, err := os.ReadFile(p)
 	if err != nil {
 		c.mu.Lock()
 		if e2, ok2 := c.items[key]; ok2 && e2 == e {
@@ -213,6 +236,20 @@ func (c *blobDiskCache) Get(key string) ([]byte, bool) {
 
 // ReadAt reads a slice of a cached blob without loading the entire blob into memory.
 func (c *blobDiskCache) ReadAt(key string, offset, length int64) ([]byte, error) {
+	p, err := c.path(key)
+	if err != nil {
+		return nil, err
+	}
+
+	// offset and length come from a blob_chunks row read back from the
+	// destination. A negative value must be rejected outright; the upper
+	// bound is checked as length > size-offset (a subtraction) so a huge
+	// offset+length cannot overflow int64 and slip past the check.
+	if offset < 0 || length < 0 {
+		return nil, fmt.Errorf("%w: offset=%d length=%d",
+			errCacheNegativeRead, offset, length)
+	}
+
 	c.mu.Lock()
 	c.readAtCalls++
 
@@ -223,7 +260,7 @@ func (c *blobDiskCache) ReadAt(key string, offset, length int64) ([]byte, error)
 		return nil, fmt.Errorf("%w: %q", errCacheKeyMissing, key)
 	}
 
-	if offset+length > e.size {
+	if length > e.size-offset {
 		c.mu.Unlock()
 
 		return nil, fmt.Errorf("%w: offset=%d length=%d size=%d",
@@ -234,7 +271,7 @@ func (c *blobDiskCache) ReadAt(key string, offset, length int64) ([]byte, error)
 	c.pushFront(e)
 	c.mu.Unlock()
 
-	f, err := os.Open(c.path(key))
+	f, err := os.Open(p) //nolint:gosec // G304: path() rejects keys with a separator
 	if err != nil {
 		return nil, err
 	}
@@ -276,7 +313,13 @@ func (c *blobDiskCache) Delete(key string) {
 	c.unlink(e)
 	delete(c.items, key)
 	c.curBytes -= e.size
-	_ = os.Remove(c.path(key))
+
+	// The key is already in the map, so it passed path() when it was
+	// inserted; the error cannot occur here.
+	p, err := c.path(key)
+	if err == nil {
+		_ = os.Remove(p)
+	}
 }
 
 // Keys returns a snapshot of all cached keys. Safe for iteration without
@@ -347,8 +390,18 @@ func (c *blobDiskCache) Close() error {
 	return os.RemoveAll(c.dir)
 }
 
-func (c *blobDiskCache) path(key string) string {
-	return filepath.Join(c.dir, key)
+// path returns the on-disk location of the cache file for key. The key is
+// a blob hash read back from the destination and is not trusted: a value
+// such as "aa/../../../home/u/.profile" would otherwise make filepath.Join
+// escape the cache directory, so a key containing a path separator is
+// refused rather than joined.
+func (c *blobDiskCache) path(key string) (string, error) {
+	if strings.ContainsRune(key, '/') ||
+		strings.ContainsRune(key, filepath.Separator) {
+		return "", fmt.Errorf("%w: %q", errCacheKeyHasSeparator, key)
+	}
+
+	return filepath.Join(c.dir, key), nil
 }
 
 func (c *blobDiskCache) unlink(e *blobDiskCacheEntry) {
@@ -391,5 +444,10 @@ func (c *blobDiskCache) evictLRU() {
 	c.unlink(victim)
 	delete(c.items, victim.key)
 	c.curBytes -= victim.size
-	_ = os.Remove(c.path(victim.key))
+
+	// victim.key was validated by path() on insertion, so this cannot err.
+	p, err := c.path(victim.key)
+	if err == nil {
+		_ = os.Remove(p)
+	}
 }