Skip to contentSkip to conversation

btrfs: tree-checker: reject file extents with oversized disk_num_bytes

PATCH Superseded Cheng Yongkang 2 participants 4 messages Nov 17 Btrfs file system

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 btrfs: tree-checker: reject file extents with oversized disk_num_bytes

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 on disk or BTRFS_MAX_UNCOMPRESSED in memory. Uncompressed extents are limited to 256MiB since v3.9, but older kernels could create larger preallocated extents.

Reject a compressed file extent whose ram_bytes is above BTRFS_MAX_UNCOMPRESSED or whose disk_num_bytes is above BTRFS_MAX_COMPRESSED. For an uncompressed extent, only reject a disk_num_bytes above U32_MAX, which should never be reached.

---
Changes in v3:
- Relax the uncompressed limit from 256MiB to U32_MAX, so preallocated extents created before v3.9 are still accepted.
- Also check ram_bytes of compressed extents against BTRFS_MAX_UNCOMPRESSED.

v2: https://lore.kernel.org/linux-btrfs/[email protected]/
v1: https://lore.kernel.org/linux-btrfs/[email protected]/

 fs/btrfs/tree-checker.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
fs/btrfs/tree-checker.c +25 -0
index f021ac14babf..e5d17bf4051b 100644
@@ -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 +378,30 @@ static int check_extent_data_item(struct extent_buffer *leaf,
 		     CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize)))
 		return -EUCLEAN;
 
+	/*
+	 * Compressed extents are at most 128K both on disk and uncompressed.
+	 * Uncompressed extents should never get close to U32_MAX.
+	 */
+	if (btrfs_file_extent_compression(leaf, fi) != BTRFS_COMPRESS_NONE) {
+		if (unlikely(btrfs_file_extent_ram_bytes(leaf, fi) > BTRFS_MAX_UNCOMPRESSED)) {
+			file_extent_err(leaf, slot,
+		"invalid ram_bytes for compressed file extent, have %llu expect <= %u",
+					btrfs_file_extent_ram_bytes(leaf, fi),
+					BTRFS_MAX_UNCOMPRESSED);
+			return -EUCLEAN;
+		}
+		max_disk_num_bytes = BTRFS_MAX_COMPRESSED;
+	} else {
+		max_disk_num_bytes = U32_MAX;
+	}
+	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

Signature · 1 line
2.34.1
#1 lore raw
Quoted 36 lines · Cheng Yongkang

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 on disk or BTRFS_MAX_UNCOMPRESSED in memory. Uncompressed extents are limited to 256MiB since v3.9, but older kernels could create larger preallocated extents.

Reject a compressed file extent whose ram_bytes is above BTRFS_MAX_UNCOMPRESSED or whose disk_num_bytes is above BTRFS_MAX_COMPRESSED. For an uncompressed extent, only reject a disk_num_bytes above U32_MAX, which should never be reached.

---
Changes in v3:
- Relax the uncompressed limit from 256MiB to U32_MAX, so preallocated extents created before v3.9 are still accepted.
- Also check ram_bytes of compressed extents against BTRFS_MAX_UNCOMPRESSED.

Quoted 2 lines

v2: https://lore.kernel.org/linux-btrfs/[email protected]/
v1: https://lore.kernel.org/linux-btrfs/[email protected]/

I don't know which agent you're using, but I have seen several ones (not only from you) doing this linking-to-older-versions behavior.

It doesn't make much sense.

Let your agent to learn from patches of known developers of each involved subsystem, other than follow whatever weird built-in prompts.

And the full changelog would make more sense, other than letting the reviewer to jump to your previous version for history.

E.g: https://lore.kernel.org/linux-btrfs/036f030792d3b497db41f1e7e376e085e094d006.1794588939.git.fdmanana@suse.com/

Quoted 17 lines
  fs/btrfs/tree-checker.c | 25 +++++++++++++++++++++++++
  1 file changed, 25 insertions(+)
fs/btrfs/tree-checker.c +1 -0
index f021ac14babf..e5d17bf4051b 100644
@@ -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;
 

fs/btrfs/tree-checker.c static int check_extent_data_item(struct extent_buffer *leaf, · line 386

Quoted 9 lines
fs/btrfs/tree-checker.c +6 -0
@@ -377,6 +378,30 @@ static int check_extent_data_item(struct extent_buffer *leaf,
  		     CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize)))
  		return -EUCLEAN;
 
+	/*
+	 * Compressed extents are at most 128K both on disk and uncompressed.
+	 * Uncompressed extents should never get close to U32_MAX.
+	 */
+	if (btrfs_file_extent_compression(leaf, fi) != BTRFS_COMPRESS_NONE) {
+		if (unlikely(btrfs_file_extent_ram_bytes(leaf, fi) > BTRFS_MAX_UNCOMPRESSED)) {

I think the ram bytes check should apply to both compressed and uncompressed extents.

For compressed, it's MAX_UNCOMPRESSED, and for uncompressed it's U32_MAX.

fs/btrfs/tree-checker.c static int check_extent_data_item(struct extent_buffer *leaf, · line 397

Quoted 11 lines
fs/btrfs/tree-checker.c +11 -0
+			file_extent_err(leaf, slot,
+		"invalid ram_bytes for compressed file extent, have %llu expect <= %u",
+					btrfs_file_extent_ram_bytes(leaf, fi),
+					BTRFS_MAX_UNCOMPRESSED);
+			return -EUCLEAN;
+		}
+		max_disk_num_bytes = BTRFS_MAX_COMPRESSED;
+	} else {
+		max_disk_num_bytes = U32_MAX;
+	}
+	if (unlikely(btrfs_file_extent_disk_num_bytes(leaf, fi) > max_disk_num_bytes)) {

Just like what you did for disk_num_bytes.

Quoted 12 lines
fs/btrfs/tree-checker.c +7 -0
+		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 lore raw
Quoted 4 lines · Qu Wenruo

I don't know which agent you're using, but I have seen several ones (not only from you) doing this linking-to-older-versions behavior.

It doesn't make much sense.

[...]

Quoted 2 lines

And the full changelog would make more sense, other than letting the reviewer to jump to your previous version for history.

The links came from Documentation/process/submitting-patches.rst, which recommends adding lore links to previous versions after the "---" line. Its example also lists the changes of every version, though, and I only listed the v3 changes. That was my mistake.

I'll follow the btrfs style from now on: the full changelog of all versions, without links to the older ones.

Quoted 4 lines

I think the ram bytes check should apply to both compressed and uncompressed extents.

For compressed, it's MAX_UNCOMPRESSED, and for uncompressed it's U32_MAX.

Will do in v4, with max_ram_bytes picked the same way as max_disk_num_bytes.

Thanks,
Cheng Yongkang

#3 lore raw
Quoted 17 lines · Cheng Yongkang
Quoted 4 lines · Qu Wenruo

I don't know which agent you're using, but I have seen several ones (not only from you) doing this linking-to-older-versions behavior.

It doesn't make much sense.

[...]

Quoted 2 lines

And the full changelog would make more sense, other than letting the reviewer to jump to your previous version for history.

The links came from Documentation/process/submitting-patches.rst, which recommends adding lore links to previous versions after the "---" line. Its example also lists the changes of every version, though, and I only listed the v3 changes. That was my mistake.

I'll follow the btrfs style from now on: the full changelog of all versions, without links to the older ones.

Quoted 7 lines
Quoted 4 lines

I think the ram bytes check should apply to both compressed and uncompressed extents.

For compressed, it's MAX_UNCOMPRESSED, and for uncompressed it's U32_MAX.

Will do in v4, with max_ram_bytes picked the same way as max_disk_num_bytes.

And Sashiko has some reviews too:

https://sashiko.dev/#/patchset/20261117023841.467366-1-teel4res%40gmail.com

Quoted 3 lines

Thanks,
Cheng Yongkang