[PATCH] btrfs: tree-checker: reject file extents with oversized disk_num_bytes

From: Cheng Yongkang

Date: Sun Nov 15 2026 - 23:45:56 EST


check_extent_data_item() only checks that disk_num_bytes of a regular or
preallocated file extent is sector aligned. A crafted image can have a
compressed extent far larger than anything btrfs writes, and reading it
goes wrong in different ways:

- With disk_num_bytes == 4GiB, btrfs_submit_compressed_read() truncates
it to a zero compressed_len and submits an empty bio:

assertion failed: IS_ALIGNED(logical, blocksize) &&
IS_ALIGNED(length, blocksize) && length != 0, in fs/btrfs/bio.c:895

- With a 4MiB zlib extent whose num_bytes and ram_bytes are above 4GiB,
bvec_offset + bv_len wraps in btrfs_decompress_buf2page() and memcpy()
reads past the zlib workspace buffer:

BUG: KASAN: slab-out-of-bounds in btrfs_decompress_buf2page+0x3e0/0x650

Compression and encoded writes never create a compressed extent larger
than BTRFS_MAX_COMPRESSED. An uncompressed extent is at most
BTRFS_MAX_EXTENT_SIZE for a regular write and 256MiB for preallocation.

Add BTRFS_MAX_PREALLOC_EXTENT_SIZE for the preallocation limit, and make
the tree checker reject a file extent whose disk_num_bytes is above
BTRFS_MAX_COMPRESSED if it is compressed, or above
BTRFS_MAX_PREALLOC_EXTENT_SIZE otherwise.

Suggested-by: Qu Wenruo <wqu@xxxxxxxx>
Signed-off-by: Cheng Yongkang <teel4res@xxxxxxxxx>
---
Report: https://lore.kernel.org/linux-btrfs/20261115163724.226922-1-teel4res@xxxxxxxxx/

On v7.3 both images from the report are rejected by the tree checker
with this patch, and 256MiB preallocated extents still pass.

fs/btrfs/fs.h | 3 +++
fs/btrfs/inode.c | 2 +-
fs/btrfs/tree-checker.c | 18 ++++++++++++++++++
3 files changed, 22 insertions(+), 1 deletion(-)

diff --git a/fs/btrfs/fs.h b/fs/btrfs/fs.h
index 2dd167c32d61..c65051a9f018 100644
--- a/fs/btrfs/fs.h
+++ b/fs/btrfs/fs.h
@@ -72,6 +72,9 @@ static_assert(BTRFS_MAX_FOLIO_SIZE > PAGE_SIZE);

#define BTRFS_MAX_EXTENT_SIZE SZ_128M

+/* The maximum size of one extent created by btrfs_prealloc_file_range(). */
+#define BTRFS_MAX_PREALLOC_EXTENT_SIZE SZ_256M
+
/*
* Maximum length to trim in a single iteration to avoid holding device list
* mutex for too long.
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 19b1eba2dd84..3f8889378d6a 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -9175,7 +9175,7 @@ int btrfs_prealloc_file_range(struct inode *inode, int mode,
u64 end = start + num_bytes - 1;

while (num_bytes > 0) {
- cur_bytes = min_t(u64, num_bytes, SZ_256M);
+ cur_bytes = min_t(u64, num_bytes, BTRFS_MAX_PREALLOC_EXTENT_SIZE);
cur_bytes = max(cur_bytes, min_size);
/*
* If we are severely fragmented we could end up with really
diff --git a/fs/btrfs/tree-checker.c b/fs/btrfs/tree-checker.c
index f021ac14babf..6e57f5969bf9 100644
--- a/fs/btrfs/tree-checker.c
+++ b/fs/btrfs/tree-checker.c
@@ -244,6 +244,7 @@ static int check_extent_data_item(struct extent_buffer *leaf,
struct btrfs_file_extent_item *fi;
u32 sectorsize = fs_info->sectorsize;
u32 item_size = btrfs_item_size(leaf, slot);
+ u32 max_disk_num_bytes;
u64 extent_end;
u8 fi_type;

@@ -377,6 +378,23 @@ static int check_extent_data_item(struct extent_buffer *leaf,
CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize)))
return -EUCLEAN;

+ /*
+ * A compressed extent is never larger than BTRFS_MAX_COMPRESSED on
+ * disk. An uncompressed one is at most BTRFS_MAX_EXTENT_SIZE for a
+ * regular write and BTRFS_MAX_PREALLOC_EXTENT_SIZE for preallocation.
+ */
+ if (btrfs_file_extent_compression(leaf, fi) != BTRFS_COMPRESS_NONE)
+ max_disk_num_bytes = BTRFS_MAX_COMPRESSED;
+ else
+ max_disk_num_bytes = BTRFS_MAX_PREALLOC_EXTENT_SIZE;
+ if (unlikely(btrfs_file_extent_disk_num_bytes(leaf, fi) > max_disk_num_bytes)) {
+ file_extent_err(leaf, slot,
+ "invalid disk_num_bytes for file extent, have %llu expect <= %u",
+ btrfs_file_extent_disk_num_bytes(leaf, fi),
+ max_disk_num_bytes);
+ return -EUCLEAN;
+ }
+
/* Catch extent end overflow */
if (unlikely(check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi),
key->offset, &extent_end))) {

base-commit: 3aefd67f08f8ed59ed0fa90619b25856df415031
--
2.34.1