From bee45bccfd54b55e3bfc17f652e1e87e7b38be5c Mon Sep 17 00:00:00 2001
From: Nick Craig-Wood <nick@craig-wood.com>
Date: Mon, 17 Aug 2026 12:35:14 +0100
Subject: [PATCH] azureblob: fix spurious vfs cache corruption errors during
 chunked reads - fixes #9782

On a ranged download the metadata decoder stored the response's
Content-Length (the length of the range, not the blob) in the object's
size and only corrected it from the Content-Range total afterwards.
Object.Size() is read concurrently by the VFS cache and chunked reader
while a download is in progress, so with --vfs-read-chunk-size a reader
could observe the chunk length (e.g. 67108864 for 64M chunks) as the
object size. The VFS cache then logged

    vfs cache: cached file (N) is unexpectedly larger than the remote
    object (67108864). The cached file is likely corrupted after an
    unclean shutdown; recovering ...

and truncated the read request against the bogus size, breaking
sequential reads of large blobs with --vfs-cache-mode full.

This applies the Content-Range correction before the size is stored so
the range length is never published as the object size.
---
 backend/azureblob/azureblob.go               | 28 +++++-------
 backend/azureblob/azureblob_internal_test.go | 46 ++++++++++++++++++++
 2 files changed, 57 insertions(+), 17 deletions(-)

diff --git a/backend/azureblob/azureblob.go b/backend/azureblob/azureblob.go
index 32b834754..642d34d2e 100644
--- a/backend/azureblob/azureblob.go
+++ b/backend/azureblob/azureblob.go
@@ -19,7 +19,6 @@ import (
 	"path"
 	"slices"
 	"sort"
-	"strconv"
 	"strings"
 	"sync"
 	"sync/atomic"
@@ -51,6 +50,7 @@ import (
 	"github.com/rclone/rclone/lib/pacer"
 	"github.com/rclone/rclone/lib/pool"
 	"github.com/rclone/rclone/lib/readers"
+	"github.com/rclone/rclone/lib/rest"
 	"github.com/rclone/rclone/lib/transferaccounter"
 	"golang.org/x/sync/errgroup"
 )
@@ -2345,6 +2345,16 @@ func (o *Object) decodeMetaDataFromDownloadResponse(info *blob.DownloadStreamRes
 	} else {
 		size = *info.ContentLength
 	}
+	// On a range request Content-Length is the length of the range, not the object, so take the
+	// object's size from the total in Content-Range instead.
+	if info.ContentRange != nil {
+		contentRange, err := rest.ParseContentRange(*info.ContentRange)
+		if err != nil {
+			fs.Debugf(o, "Failed to parse Content-Range %q: %v", *info.ContentRange, err)
+		} else if contentRange.Size >= 0 {
+			size = contentRange.Size
+		}
+	}
 	if isDirectoryMarker(size, metadata, o.remote) {
 		return fs.ErrorNotAFile
 	}
@@ -2378,22 +2388,6 @@ func (o *Object) decodeMetaDataFromDownloadResponse(info *blob.DownloadStreamRes
 	// 	o.accessTier = blob.AccessTier(*info.AccessTier)
 	// }
 	o.setMetadata(metadata)
-
-	// If it was a Range request, the size is wrong, so correct it
-	if info.ContentRange != nil {
-		contentRange := *info.ContentRange
-		slash := strings.IndexRune(contentRange, '/')
-		if slash >= 0 {
-			i, err := strconv.ParseInt(contentRange[slash+1:], 10, 64)
-			if err == nil {
-				o.size = i
-			} else {
-				fs.Debugf(o, "Failed to find parse integer from in %q: %v", contentRange, err)
-			}
-		} else {
-			fs.Debugf(o, "Failed to find length in %q", contentRange)
-		}
-	}
 	o.contentEncoding = info.ContentEncoding
 
 	// If decompressing then size and md5sum are unknown
diff --git a/backend/azureblob/azureblob_internal_test.go b/backend/azureblob/azureblob_internal_test.go
index c234463c2..f8deb0fdb 100644
--- a/backend/azureblob/azureblob_internal_test.go
+++ b/backend/azureblob/azureblob_internal_test.go
@@ -54,6 +54,52 @@ func TestBlockIDCreator(t *testing.T) {
 	assert.ErrorContains(t, bic2.checkID(chunkNumber, got), "random bytes")
 }
 
+func TestDecodeMetaDataFromDownloadResponse(t *testing.T) {
+	pInt64 := func(i int64) *int64 { return &i }
+	newTestObject := func() *Object {
+		return &Object{
+			fs:     &Fs{},
+			remote: "test.bin",
+			size:   -2, // sentinel to check it gets set
+		}
+	}
+
+	t.Run("WholeBlob", func(t *testing.T) {
+		o := newTestObject()
+		info := blob.DownloadStreamResponse{}
+		info.ContentLength = pInt64(12345)
+		require.NoError(t, o.decodeMetaDataFromDownloadResponse(&info))
+		assert.Equal(t, int64(12345), o.size)
+	})
+
+	t.Run("RangeRequest", func(t *testing.T) {
+		// On a ranged download Content-Length is the chunk length so the
+		// size must come from the total in Content-Range
+		o := newTestObject()
+		info := blob.DownloadStreamResponse{}
+		info.ContentLength = pInt64(67108864)
+		info.ContentRange = pString("bytes 67108864-134217727/169721004032")
+		require.NoError(t, o.decodeMetaDataFromDownloadResponse(&info))
+		assert.Equal(t, int64(169721004032), o.size)
+	})
+
+	t.Run("BadContentRange", func(t *testing.T) {
+		o := newTestObject()
+		info := blob.DownloadStreamResponse{}
+		info.ContentLength = pInt64(67108864)
+		info.ContentRange = pString("potato")
+		require.NoError(t, o.decodeMetaDataFromDownloadResponse(&info))
+		assert.Equal(t, int64(67108864), o.size)
+	})
+
+	t.Run("UnknownLength", func(t *testing.T) {
+		o := newTestObject()
+		info := blob.DownloadStreamResponse{}
+		require.NoError(t, o.decodeMetaDataFromDownloadResponse(&info))
+		assert.Equal(t, int64(-1), o.size)
+	})
+}
+
 func TestCopySASTimingConstants(t *testing.T) {
 	require.Greater(t, sasCopyStartSkew, time.Duration(0))
 	require.Greater(t, sasCopyValidity, sasCopyStartSkew)