Skip to contentSkip to conversation

btrfs: support fsverity

PATCH Superseded Boris Burkov 3 participants 15 messages Mar 5, 2021 – Mar 16, 2021

Subsystems (2)

v2 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v2 cover btrfs: support fsverity

This patchset provides support for fsverity in btrfs.

At a high level, we store the verity descriptor and Merkle tree data in the file system btree with the file's inode as the objectid, and direct reads/writes to those items to implement the generic fsverity interface required by fs/verity/.

The first patch is a preparatory patch which adds a notion of compat_flags to the btrfs_inode and inode_item in order to allow enabling verity on a file without making the file system unmountable for older kernels. (It runs afoul of the leaf corruption check otherwise)

The second patch is the bulk of the fsverity implementation. It implements the fsverity interface and adds verity checks for the typical file reading case.

The third patch cleans up the corner cases in readpage, covering inline extents, preallocated extents, and holes.

The fourth patch handles direct io of a veritied file by falling back to buffered io.

The fifth patch handles crashes mid-verity enable via orphan items

changes for v2:
Patch 1: Unchanged.
Patch 2: Return EFBIG if Merkle data past s_maxbytes. Added special descriptor item for encryption and to handle ERANGE case for get_verity_descriptor. Improved function comments. Rebased onto subpage read patches -- modified end_page_read to do verity check before marking the page uptodate. Changed from full compat to ro_compat; merged sysfs feature here.
Patch 3: Rebased onto subpage read patches.
Patch 4: Unchanged.
Patch 5: Used to be sysfs feature, now a new patch that handles orphaned verity data.

Boris Burkov (4):
btrfs: add compat_flags to btrfs_inode_item
btrfs: check verity for reads of inline extents and holes
btrfs: fallback to buffered io for verity files
btrfs: verity metadata orphan items

Chris Mason (1):
btrfs: initial fsverity support

 fs/btrfs/Makefile               |   1 +
 fs/btrfs/btrfs_inode.h          |   2 +
 fs/btrfs/ctree.h                |  25 +-
 fs/btrfs/delayed-inode.c        |   2 +
 fs/btrfs/extent_io.c            |  53 +--
 fs/btrfs/file.c                 |   9 +
 fs/btrfs/inode.c                |  25 +-
 fs/btrfs/ioctl.c                |  21 +-
 fs/btrfs/super.c                |   1 +
 fs/btrfs/sysfs.c                |   6 +
 fs/btrfs/tree-log.c             |   1 +
 fs/btrfs/verity.c               | 658 ++++++++++++++++++++++++++++++++
 include/uapi/linux/btrfs.h      |   2 +-
 include/uapi/linux/btrfs_tree.h |  22 +-
 14 files changed, 792 insertions(+), 36 deletions(-)
 create mode 100644 fs/btrfs/verity.c
Signature · 1 line
2.24.1
#1 lore raw

PATCH v2 1/5 btrfs: add compat_flags to btrfs_inode_item

The tree checker currently rejects unrecognized flags when it reads btrfs_inode_item. Practically, this means that adding a new flag makes the change backwards incompatible if the flag is ever set on a file.

Take up one of the 4 reserved u64 fields in the btrfs_inode_item as a new "compat_flags". These flags are zero on inode creation in btrfs and mkfs and are ignored by an older kernel, so it should be safe to use them in this way.

---
 fs/btrfs/btrfs_inode.h          | 1 +
 fs/btrfs/ctree.h                | 2 ++
 fs/btrfs/delayed-inode.c        | 2 ++
 fs/btrfs/inode.c                | 3 +++
 fs/btrfs/ioctl.c                | 7 ++++---
 fs/btrfs/tree-log.c             | 1 +
 include/uapi/linux/btrfs_tree.h | 7 ++++++-
 7 files changed, 19 insertions(+), 4 deletions(-)
fs/btrfs/btrfs_inode.h +1 -0
index 28e202e89660..8b95932f25a8 100644
@@ -191,6 +191,7 @@ struct btrfs_inode {
 
 	/* flags field from the on disk inode */
 	u32 flags;
+	u64 compat_flags;
 
 	/*
 	 * Counters to keep track of the number of extent item's we may use due
fs/btrfs/ctree.h +2 -0
index 40ec3393d2a1..e65df997a2c6 100644
@@ -1761,6 +1761,7 @@ BTRFS_SETGET_FUNCS(inode_gid, struct btrfs_inode_item, gid, 32);
 BTRFS_SETGET_FUNCS(inode_mode, struct btrfs_inode_item, mode, 32);
 BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 64);
 BTRFS_SETGET_FUNCS(inode_flags, struct btrfs_inode_item, flags, 64);
+BTRFS_SETGET_FUNCS(inode_compat_flags, struct btrfs_inode_item, compat_flags, 64);
 BTRFS_SETGET_STACK_FUNCS(stack_inode_generation, struct btrfs_inode_item,
 			 generation, 64);
 BTRFS_SETGET_STACK_FUNCS(stack_inode_sequence, struct btrfs_inode_item,
@@ -1778,6 +1779,7 @@ BTRFS_SETGET_STACK_FUNCS(stack_inode_gid, struct btrfs_inode_item, gid, 32);
 BTRFS_SETGET_STACK_FUNCS(stack_inode_mode, struct btrfs_inode_item, mode, 32);
 BTRFS_SETGET_STACK_FUNCS(stack_inode_rdev, struct btrfs_inode_item, rdev, 64);
 BTRFS_SETGET_STACK_FUNCS(stack_inode_flags, struct btrfs_inode_item, flags, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_compat_flags, struct btrfs_inode_item, compat_flags, 64);
 BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_timespec, sec, 64);
 BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_timespec, nsec, 32);
 BTRFS_SETGET_STACK_FUNCS(stack_timespec_sec, struct btrfs_timespec, sec, 64);
fs/btrfs/delayed-inode.c +2 -0
index ec0b50b8c5d6..6ea9d27b9c9d 100644
@@ -1733,6 +1733,7 @@ static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
 	btrfs_set_stack_inode_transid(inode_item, trans->transid);
 	btrfs_set_stack_inode_rdev(inode_item, inode->i_rdev);
 	btrfs_set_stack_inode_flags(inode_item, BTRFS_I(inode)->flags);
+	btrfs_set_stack_inode_compat_flags(inode_item, BTRFS_I(inode)->compat_flags);
 	btrfs_set_stack_inode_block_group(inode_item, 0);
 
 	btrfs_set_stack_timespec_sec(&inode_item->atime,
@@ -1791,6 +1792,7 @@ int btrfs_fill_inode(struct inode *inode, u32 *rdev)
 	inode->i_rdev = 0;
 	*rdev = btrfs_stack_inode_rdev(inode_item);
 	BTRFS_I(inode)->flags = btrfs_stack_inode_flags(inode_item);
+	BTRFS_I(inode)->compat_flags = btrfs_stack_inode_compat_flags(inode_item);
 
 	inode->i_atime.tv_sec = btrfs_stack_timespec_sec(&inode_item->atime);
 	inode->i_atime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->atime);
fs/btrfs/inode.c +3 -0
index 4f2f1e932751..4529e974c582 100644
@@ -3627,6 +3627,7 @@ static int btrfs_read_locked_inode(struct inode *inode,
 
 	BTRFS_I(inode)->index_cnt = (u64)-1;
 	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
+	BTRFS_I(inode)->compat_flags = btrfs_inode_compat_flags(leaf, inode_item);
 
 cache_index:
 	/*
@@ -3793,6 +3794,7 @@ static void fill_inode_item(struct btrfs_trans_handle *trans,
 	btrfs_set_token_inode_transid(&token, item, trans->transid);
 	btrfs_set_token_inode_rdev(&token, item, inode->i_rdev);
 	btrfs_set_token_inode_flags(&token, item, BTRFS_I(inode)->flags);
+	btrfs_set_token_inode_compat_flags(&token, item, BTRFS_I(inode)->compat_flags);
 	btrfs_set_token_inode_block_group(&token, item, 0);
 }
 
@@ -8844,6 +8846,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 	ei->defrag_bytes = 0;
 	ei->disk_i_size = 0;
 	ei->flags = 0;
+	ei->compat_flags = 0;
 	ei->csum_bytes = 0;
 	ei->index_cnt = (u64)-1;
 	ei->dir_index = 0;
fs/btrfs/ioctl.c +4 -3
index a8c60d46d19c..d3fb065312f5 100644
@@ -102,8 +102,9 @@ static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
  * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
  * ioctl.
  */
-static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags)
+static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
 {
+	unsigned int flags = binode->flags;
 	unsigned int iflags = 0;
 
 	if (flags & BTRFS_INODE_SYNC)
@@ -156,7 +157,7 @@ void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
 {
 	struct btrfs_inode *binode = BTRFS_I(file_inode(file));
-	unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags);
+	unsigned int flags = btrfs_inode_flags_to_fsflags(binode);
 
 	if (copy_to_user(arg, &flags, sizeof(flags)))
 		return -EFAULT;
@@ -228,7 +229,7 @@ static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 
 	inode_lock(inode);
 	fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
-	old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
+	old_fsflags = btrfs_inode_flags_to_fsflags(binode);
 
 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
 	if (ret)
fs/btrfs/tree-log.c +1 -0
index 2f1acc9aea9e..c436b4517452 100644
@@ -3942,6 +3942,7 @@ static void fill_inode_item(struct btrfs_trans_handle *trans,
 	btrfs_set_token_inode_transid(&token, item, trans->transid);
 	btrfs_set_token_inode_rdev(&token, item, inode->i_rdev);
 	btrfs_set_token_inode_flags(&token, item, BTRFS_I(inode)->flags);
+	btrfs_set_token_inode_compat_flags(&token, item, BTRFS_I(inode)->compat_flags);
 	btrfs_set_token_inode_block_group(&token, item, 0);
 }
 
include/uapi/linux/btrfs_tree.h +6 -1
index 58d7cff9afb1..ae25280316bd 100644
@@ -574,11 +574,16 @@ struct btrfs_inode_item {
 	/* modification sequence number for NFS */
 	__le64 sequence;
 
+	/*
+	 * flags which aren't checked for corruption at mount
+	 * and can be added in a backwards compatible way
+	 */
+	__le64 compat_flags;
 	/*
 	 * a little future expansion, for more than this we can
 	 * just grow the inode item and version it
 	 */
-	__le64 reserved[4];
+	__le64 reserved[3];
 	struct btrfs_timespec atime;
 	struct btrfs_timespec ctime;
 	struct btrfs_timespec mtime;
Signature · 1 line
2.24.1
maintainer
#2 lore raw

Re: [PATCH v2 1/5] btrfs: add compat_flags to btrfs_inode_item

Quoted 10 lines · Boris Burkov

The tree checker currently rejects unrecognized flags when it reads btrfs_inode_item. Practically, this means that adding a new flag makes the change backwards incompatible if the flag is ever set on a file.

Take up one of the 4 reserved u64 fields in the btrfs_inode_item as a new "compat_flags". These flags are zero on inode creation in btrfs and mkfs and are ignored by an older kernel, so it should be safe to use them in this way.

How does this interact with the RO_COMPAT filesystem flag which is also being added?

- Eric

#3 lore raw

Re: [PATCH v2 1/5] btrfs: add compat_flags to btrfs_inode_item

Quoted 10 lines · Eric Biggers
Quoted 9 lines · Boris Burkov

The tree checker currently rejects unrecognized flags when it reads btrfs_inode_item. Practically, this means that adding a new flag makes the change backwards incompatible if the flag is ever set on a file.

Take up one of the 4 reserved u64 fields in the btrfs_inode_item as a new "compat_flags". These flags are zero on inode creation in btrfs and mkfs and are ignored by an older kernel, so it should be safe to use them in this way.

Quoted 6 lines
Quoted 1 line

How does this interact with the RO_COMPAT filesystem flag which is also being added?

- Eric

This allows us to mount at all -- without it, when mounting on an older system, the logic in fs/btrfs/tree-checker.c would see an unexpected inode flag value and reject the filesystem entirely.

The RO_COMPAT flag represents the decision to mount RO rather than RW when a verity file is present (though it suffers from the fact that it gets set on verity enable, but not unset when the last verity file is deleted)

I'm now thinking that ro_compat inode flags might be an interesting option where the tree-checker could indicate we have to mount read only if it sees an unrecognized value.

#4 lore raw

PATCH v2 2/5 btrfs: initial fsverity support

In reply to Boris Burkov

From: Chris Mason <[email protected]>

Add support for fsverity in btrfs. To support the generic interface in fs/verity, we add two new item types in the fs tree for inodes with verity enabled. One stores the per-file verity descriptor and the other stores the Merkle tree data itself.

Verity checking is done at the end of IOs to ensure each page is checked before it is marked uptodate.

Verity relies on PageChecked for the Merkle tree data itself to avoid re-walking up shared paths in the tree. For this reason, we need to cache the Merkle tree data. Since the file is immutable after verity is turned on, we can cache it at an index past EOF.

Use the new inode compat_flags to store verity on the inode item, so that we can enable verity on a file, then rollback to an older kernel and still mount the file system and read the file. Since we can't safely write the file anymore without ruining the invariants of the Merkle tree, we mark a ro_compat flag on the file system when a file has verity enabled.

---
 fs/btrfs/Makefile               |   1 +
 fs/btrfs/btrfs_inode.h          |   1 +
 fs/btrfs/ctree.h                |  23 +-
 fs/btrfs/extent_io.c            |  27 +-
 fs/btrfs/file.c                 |   6 +
 fs/btrfs/inode.c                |   7 +
 fs/btrfs/ioctl.c                |  14 +-
 fs/btrfs/super.c                |   1 +
 fs/btrfs/sysfs.c                |   6 +
 fs/btrfs/verity.c               | 598 ++++++++++++++++++++++++++++++++
 include/uapi/linux/btrfs.h      |   2 +-
 include/uapi/linux/btrfs_tree.h |  15 +
 12 files changed, 690 insertions(+), 11 deletions(-)
 create mode 100644 fs/btrfs/verity.c
fs/btrfs/Makefile +1 -0
index b634c42115ea..ff719d2bba1a 100644
@@ -34,6 +34,7 @@ btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o
 btrfs-$(CONFIG_BTRFS_FS_CHECK_INTEGRITY) += check-integrity.o
 btrfs-$(CONFIG_BTRFS_FS_REF_VERIFY) += ref-verify.o
 btrfs-$(CONFIG_BLK_DEV_ZONED) += zoned.o
+btrfs-$(CONFIG_FS_VERITY) += verity.o
 
 btrfs-$(CONFIG_BTRFS_FS_RUN_SANITY_TESTS) += tests/free-space-tests.o \
 	tests/extent-buffer-tests.o tests/btrfs-tests.o \
fs/btrfs/btrfs_inode.h +1 -0
index 8b95932f25a8..c0b0537bd50e 100644
@@ -51,6 +51,7 @@ enum {
 	 * the file range, inode's io_tree).
 	 */
 	BTRFS_INODE_NO_DELALLOC_FLUSH,
+	BTRFS_INODE_VERITY_IN_PROGRESS,
 };
 
 /* in memory btrfs inode */
fs/btrfs/ctree.h +20 -3
index e65df997a2c6..f73ba7794dda 100644
@@ -279,9 +279,10 @@ struct btrfs_super_block {
 #define BTRFS_FEATURE_COMPAT_SAFE_SET		0ULL
 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR		0ULL
 
-#define BTRFS_FEATURE_COMPAT_RO_SUPP			\
-	(BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE |	\
-	 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID)
+#define BTRFS_FEATURE_COMPAT_RO_SUPP				\
+	(BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE |		\
+	 BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID |	\
+	 BTRFS_FEATURE_COMPAT_RO_VERITY)
 
 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET	0ULL
 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR	0ULL
@@ -1480,6 +1481,11 @@ do {                                                                   \
 	 BTRFS_INODE_COMPRESS |						\
 	 BTRFS_INODE_ROOT_ITEM_INIT)
 
+/*
+ * Inode compat flags
+ */
+#define BTRFS_INODE_VERITY		(1 << 0)
+
 struct btrfs_map_token {
 	struct extent_buffer *eb;
 	char *kaddr;
@@ -3735,6 +3741,17 @@ static inline int btrfs_defrag_cancelled(struct btrfs_fs_info *fs_info)
 
 #define in_range(b, first, len) ((b) >= (first) && (b) < (first) + (len))
 
+/* verity.c */
+extern const struct fsverity_operations btrfs_verityops;
+int btrfs_drop_verity_items(struct btrfs_inode *inode);
+BTRFS_SETGET_FUNCS(verity_descriptor_encryption, struct btrfs_verity_descriptor_item,
+		   encryption, 8);
+BTRFS_SETGET_FUNCS(verity_descriptor_size, struct btrfs_verity_descriptor_item, size, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_verity_descriptor_encryption, struct btrfs_verity_descriptor_item,
+			 encryption, 8);
+BTRFS_SETGET_STACK_FUNCS(stack_verity_descriptor_size, struct btrfs_verity_descriptor_item,
+			 size, 64);
+
 /* Sanity test specific functions */
 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
 void btrfs_test_destroy_inode(struct inode *inode);
fs/btrfs/extent_io.c +22 -5
index 4dfb3ead1175..0d6f8b33830f 100644
@@ -13,6 +13,7 @@
 #include <linux/pagevec.h>
 #include <linux/prefetch.h>
 #include <linux/cleancache.h>
+#include <linux/fsverity.h>
 #include "extent_io.h"
 #include "extent-io-tree.h"
 #include "extent_map.h"
@@ -2861,15 +2862,28 @@ static void begin_page_read(struct btrfs_fs_info *fs_info, struct page *page)
 	btrfs_subpage_start_reader(fs_info, page, page_offset(page), PAGE_SIZE);
 }
 
-static void end_page_read(struct page *page, bool uptodate, u64 start, u32 len)
+static int end_page_read(struct page *page, bool uptodate, u64 start, u32 len)
 {
-	struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb);
+	int ret = 0;
+	struct inode *inode = page->mapping->host;
+	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
 
 	ASSERT(page_offset(page) <= start &&
 		start + len <= page_offset(page) + PAGE_SIZE);
 
 	if (uptodate) {
-		btrfs_page_set_uptodate(fs_info, page, start, len);
+		/*
+		 * buffered reads of a file with page alignment will issue a
+		 * 0 length read for one page past the end of file, so we must
+		 * explicitly skip checking verity on that page of zeros.
+		 */
+		if (!PageError(page) && !PageUptodate(page) &&
+		    start < i_size_read(inode) &&
+		    fsverity_active(inode) &&
+		    fsverity_verify_page(page) != true)
+			ret = -EIO;
+		else
+			btrfs_page_set_uptodate(fs_info, page, start, len);
 	} else {
 		btrfs_page_clear_uptodate(fs_info, page, start, len);
 		btrfs_page_set_error(fs_info, page, start, len);
@@ -2877,12 +2891,13 @@ static void end_page_read(struct page *page, bool uptodate, u64 start, u32 len)
 
 	if (fs_info->sectorsize == PAGE_SIZE)
 		unlock_page(page);
-	else if (is_data_inode(page->mapping->host))
+	else if (is_data_inode(inode))
 		/*
 		 * For subpage data, unlock the page if we're the last reader.
 		 * For subpage metadata, page lock is not utilized for read.
 		 */
 		btrfs_subpage_end_reader(fs_info, page, start, len);
+	return ret;
 }
 
 /*
@@ -3019,7 +3034,9 @@ static void end_bio_extent_readpage(struct bio *bio)
 		bio_offset += len;
 
 		/* Update page status and unlock */
-		end_page_read(page, uptodate, start, len);
+		ret = end_page_read(page, uptodate, start, len);
+		if (ret)
+			uptodate = 0;
 		endio_readpage_release_extent(&processed, BTRFS_I(inode),
 					      start, end, uptodate);
 	}
fs/btrfs/file.c +6 -0
index 2c282664c4b8..75f1406021e5 100644
@@ -16,6 +16,7 @@
 #include <linux/btrfs.h>
 #include <linux/uio.h>
 #include <linux/iversion.h>
+#include <linux/fsverity.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -3588,7 +3589,12 @@ static loff_t btrfs_file_llseek(struct file *file, loff_t offset, int whence)
 
 static int btrfs_file_open(struct inode *inode, struct file *filp)
 {
+	int ret;
 	filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC;
+
+	ret = fsverity_file_open(inode, filp);
+	if (ret)
+		return ret;
 	return generic_file_open(inode, filp);
 }
 
fs/btrfs/inode.c +7 -0
index 4529e974c582..9291f6633bc6 100644
@@ -32,6 +32,7 @@
 #include <linux/sched/mm.h>
 #include <linux/iomap.h>
 #include <asm/unaligned.h>
+#include <linux/fsverity.h>
 #include "misc.h"
 #include "ctree.h"
 #include "disk-io.h"
@@ -5399,7 +5400,9 @@ void btrfs_evict_inode(struct inode *inode)
 
 	trace_btrfs_inode_evict(inode);
 
+
 	if (!root) {
+		fsverity_cleanup_inode(inode);
 		clear_inode(inode);
 		return;
 	}
@@ -5482,6 +5485,7 @@ void btrfs_evict_inode(struct inode *inode)
 	 * to retry these periodically in the future.
 	 */
 	btrfs_remove_delayed_node(BTRFS_I(inode));
+	fsverity_cleanup_inode(inode);
 	clear_inode(inode);
 }
 
@@ -9026,6 +9030,7 @@ static int btrfs_getattr(const struct path *path, struct kstat *stat,
 	struct inode *inode = d_inode(path->dentry);
 	u32 blocksize = inode->i_sb->s_blocksize;
 	u32 bi_flags = BTRFS_I(inode)->flags;
+	u32 bi_compat_flags = BTRFS_I(inode)->compat_flags;
 
 	stat->result_mask |= STATX_BTIME;
 	stat->btime.tv_sec = BTRFS_I(inode)->i_otime.tv_sec;
@@ -9038,6 +9043,8 @@ static int btrfs_getattr(const struct path *path, struct kstat *stat,
 		stat->attributes |= STATX_ATTR_IMMUTABLE;
 	if (bi_flags & BTRFS_INODE_NODUMP)
 		stat->attributes |= STATX_ATTR_NODUMP;
+	if (bi_compat_flags & BTRFS_INODE_VERITY)
+		stat->attributes |= STATX_ATTR_VERITY;
 
 	stat->attributes_mask |= (STATX_ATTR_APPEND |
 				  STATX_ATTR_COMPRESSED |
fs/btrfs/ioctl.c +12 -2
index d3fb065312f5..cf91d8f0d479 100644
@@ -26,6 +26,7 @@
 #include <linux/btrfs.h>
 #include <linux/uaccess.h>
 #include <linux/iversion.h>
+#include <linux/fsverity.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "export.h"
@@ -105,6 +106,7 @@ static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
 {
 	unsigned int flags = binode->flags;
+	unsigned int compat_flags = binode->compat_flags;
 	unsigned int iflags = 0;
 
 	if (flags & BTRFS_INODE_SYNC)
@@ -121,6 +123,8 @@ static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
 		iflags |= FS_DIRSYNC_FL;
 	if (flags & BTRFS_INODE_NODATACOW)
 		iflags |= FS_NOCOW_FL;
+	if (compat_flags & BTRFS_INODE_VERITY)
+		iflags |= FS_VERITY_FL;
 
 	if (flags & BTRFS_INODE_NOCOMPRESS)
 		iflags |= FS_NOCOMP_FL;
@@ -148,10 +152,12 @@ void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
 		new_fl |= S_NOATIME;
 	if (binode->flags & BTRFS_INODE_DIRSYNC)
 		new_fl |= S_DIRSYNC;
+	if (binode->compat_flags & BTRFS_INODE_VERITY)
+		new_fl |= S_VERITY;
 
 	set_mask_bits(&inode->i_flags,
-		      S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
-		      new_fl);
+		      S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
+		      S_VERITY, new_fl);
 }
 
 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
@@ -5035,6 +5041,10 @@ long btrfs_ioctl(struct file *file, unsigned int
 		return btrfs_ioctl_get_subvol_rootref(file, argp);
 	case BTRFS_IOC_INO_LOOKUP_USER:
 		return btrfs_ioctl_ino_lookup_user(file, argp);
+	case FS_IOC_ENABLE_VERITY:
+		return fsverity_ioctl_enable(file, (const void __user *)argp);
+	case FS_IOC_MEASURE_VERITY:
+		return fsverity_ioctl_measure(file, argp);
 	}
 
 	return -ENOTTY;
fs/btrfs/super.c +1 -0
index f8435641b912..be7dfde7e4cf 100644
@@ -1339,6 +1339,7 @@ static int btrfs_fill_super(struct super_block *sb,
 	sb->s_op = &btrfs_super_ops;
 	sb->s_d_op = &btrfs_dentry_operations;
 	sb->s_export_op = &btrfs_export_ops;
+	sb->s_vop = &btrfs_verityops;
 	sb->s_xattr = btrfs_xattr_handlers;
 	sb->s_time_gran = 1;
 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
fs/btrfs/sysfs.c +6 -0
index 6eb1c50fa98c..e8de0055c61e 100644
@@ -267,6 +267,9 @@ BTRFS_FEAT_ATTR_INCOMPAT(raid1c34, RAID1C34);
 #ifdef CONFIG_BTRFS_DEBUG
 BTRFS_FEAT_ATTR_INCOMPAT(zoned, ZONED);
 #endif
+#ifdef CONFIG_FS_VERITY
+BTRFS_FEAT_ATTR_COMPAT_RO(verity, VERITY);
+#endif
 
 static struct attribute *btrfs_supported_feature_attrs[] = {
 	BTRFS_FEAT_ATTR_PTR(mixed_backref),
@@ -284,6 +287,9 @@ static struct attribute *btrfs_supported_feature_attrs[] = {
 	BTRFS_FEAT_ATTR_PTR(raid1c34),
 #ifdef CONFIG_BTRFS_DEBUG
 	BTRFS_FEAT_ATTR_PTR(zoned),
+#endif
+#ifdef CONFIG_FS_VERITY
+	BTRFS_FEAT_ATTR_PTR(verity),
 #endif
 	NULL
 };
fs/btrfs/verity.c new file +598 -0

598 lines not on this page · raw message

include/uapi/linux/btrfs.h +1 -1
index 5df73001aad4..fa21c8aac78d 100644
@@ -288,6 +288,7 @@ struct btrfs_ioctl_fs_info_args {
  * first mount when booting older kernel versions.
  */
 #define BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE_VALID	(1ULL << 1)
+#define BTRFS_FEATURE_COMPAT_RO_VERITY		(1ULL << 2)
 
 #define BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF	(1ULL << 0)
 #define BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL	(1ULL << 1)
@@ -308,7 +309,6 @@ struct btrfs_ioctl_fs_info_args {
 #define BTRFS_FEATURE_INCOMPAT_METADATA_UUID	(1ULL << 10)
 #define BTRFS_FEATURE_INCOMPAT_RAID1C34		(1ULL << 11)
 #define BTRFS_FEATURE_INCOMPAT_ZONED		(1ULL << 12)
-
 struct btrfs_ioctl_feature_flags {
 	__u64 compat_flags;
 	__u64 compat_ro_flags;
include/uapi/linux/btrfs_tree.h +15 -0
index ae25280316bd..2be57416f886 100644
@@ -118,6 +118,14 @@
 #define BTRFS_INODE_REF_KEY		12
 #define BTRFS_INODE_EXTREF_KEY		13
 #define BTRFS_XATTR_ITEM_KEY		24
+
+/*
+ * fsverity has a descriptor per file, and then
+ * a number of sha or csum items indexed by offset in to the file.
+ */
+#define BTRFS_VERITY_DESC_ITEM_KEY	36
+#define BTRFS_VERITY_MERKLE_ITEM_KEY	37
+
 #define BTRFS_ORPHAN_ITEM_KEY		48
 /* reserve 2-15 close to the inode for later flexibility */
 
@@ -996,4 +1004,11 @@ struct btrfs_qgroup_limit_item {
 	__le64 rsv_excl;
 } __attribute__ ((__packed__));
 
+struct btrfs_verity_descriptor_item {
+	/* size of the verity descriptor in bytes */
+	__le64 size;
+	__le64 reserved[2];
+	__u8 encryption;
+} __attribute__ ((__packed__));
+
 #endif /* _BTRFS_CTREE_H_ */
Signature · 1 line
2.24.1
#5 lore raw

Re: [PATCH v2 2/5] btrfs: initial fsverity support

I love your patch! Yet something to improve:

Hi Boris,

I love your patch! Yet something to improve:

[auto build test ERROR on v5.12-rc2]
[also build test ERROR on next-20210305]
[cannot apply to kdave/for-next]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Boris-Burkov/btrfs-support-fsverity/20210307-105111
base:    a38fd8748464831584a19438cbb3082b5a2dab15
config: i386-randconfig-m021-20210307 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
reproduce (this is a W=1 build):
        # https://github.com/0day-ci/linux/commit/230ae9030c4636229de25a9013d979fc8e04f459
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Boris-Burkov/btrfs-support-fsverity/20210307-105111
        git checkout 230ae9030c4636229de25a9013d979fc8e04f459
        # save the attached .config to linux build tree
        make W=1 ARCH=i386 

If you fix the issue, kindly add following tag as appropriate

All errors (new ones prefixed by >>):

fs/btrfs/super.c: In function 'btrfs_fill_super':

>> fs/btrfs/super.c:1342:6: error: 'struct super_block' has no member named 's_vop'; did you mean 's_op'?
    1342 |  sb->s_vop = &btrfs_verityops;
         |      ^~~~~
         |      s_op

vim +1342 fs/btrfs/super.c

1328
1329 static int btrfs_fill_super(struct super_block *sb,
1330 struct btrfs_fs_devices *fs_devices,
1331 void *data)
1332 {
1333 struct inode *inode;
1334 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1335 int err;
1336
1337 sb->s_maxbytes = MAX_LFS_FILESIZE;
1338 sb->s_magic = BTRFS_SUPER_MAGIC;
1339 sb->s_op = &btrfs_super_ops;
1340 sb->s_d_op = &btrfs_dentry_operations;
1341 sb->s_export_op = &btrfs_export_ops;

> 1342		sb->s_vop = &btrfs_verityops;

1343 sb->s_xattr = btrfs_xattr_handlers;
1344 sb->s_time_gran = 1;
1345 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1346 sb->s_flags |= SB_POSIXACL;
1347 #endif
1348 sb->s_flags |= SB_I_VERSION;
1349 sb->s_iflags |= SB_I_CGROUPWB;
1350
1351 err = super_setup_bdi(sb);
1352 if (err) {
1353 btrfs_err(fs_info, "super_setup_bdi failed");
1354 return err;
1355 }
1356
1357 err = open_ctree(sb, fs_devices, (char *)data);
1358 if (err) {
1359 btrfs_err(fs_info, "open_ctree failed");
1360 return err;
1361 }
1362
1363 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1364 if (IS_ERR(inode)) {
1365 err = PTR_ERR(inode);
1366 goto fail_close;
1367 }
1368
1369 sb->s_root = d_make_root(inode);
1370 if (!sb->s_root) {
1371 err = -ENOMEM;
1372 goto fail_close;
1373 }
1374
1375 cleancache_init_fs(sb);
1376 sb->s_flags |= SB_ACTIVE;
1377 return 0;
1378
1379 fail_close:
1380 close_ctree(fs_info);
1381 return err;
1382 }
1383

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

maintainer
#6 lore raw

Re: [PATCH v2 2/5] btrfs: initial fsverity support

In reply to Boris Burkov
Quoted 15 lines · Boris Burkov
fs/btrfs/extent_io.c +7 -2
+static int end_page_read(struct page *page, bool uptodate, u64 start, u32 len)
 {
-	struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb);
+	int ret = 0;
+	struct inode *inode = page->mapping->host;
+	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
 
 	ASSERT(page_offset(page) <= start &&
 		start + len <= page_offset(page) + PAGE_SIZE);
 
 	if (uptodate) {
-		btrfs_page_set_uptodate(fs_info, page, start, len);
+		/*
+		 * buffered reads of a file with page alignment will issue a
+		 * 0 length read for one page past the end of file, so we must

fs/btrfs/extent_io.c static void begin_page_read(struct btrfs_fs_info *fs_info, struct page *page) · line 2887

Quoted 10 lines
fs/btrfs/extent_io.c +9 -0
+		 * explicitly skip checking verity on that page of zeros.
+		 */
+		if (!PageError(page) && !PageUptodate(page) &&
+		    start < i_size_read(inode) &&
+		    fsverity_active(inode) &&
+		    fsverity_verify_page(page) != true)
+			ret = -EIO;
+		else
+			btrfs_page_set_uptodate(fs_info, page, start, len);
 	} else {

'fsverity_verify_page(page) != true' is better written as '!fsverity_verify_page(page)'.

Quoted 8 lines

+/*
+ * Just like ext4, we cache the merkle tree in pages after EOF in the page
+ * cache. Unlike ext4, we're storing these in dedicated btree items and
+ * not just shoving them after EOF in the file. This means we'll need to
+ * do extra work to encrypt them once encryption is supported in btrfs,
+ * but btrfs has a lot of careful code around i_size and it seems better
+ * to make a new key type than try and adjust all of our expectations
+ * for i_size.

Quoted 8 lines

+ *
+ * fs verity items are stored under two different key types on disk.
+ *
+ * The descriptor items:
+ * [ inode objectid, BTRFS_VERITY_DESC_ITEM_KEY, offset ]
+ *
+ * At offset 0, we store a btrfs_verity_descriptor_item which tracks the
+ * size of the descriptor item and some extra data for encryption.

Is the separate size field really needed? It seems like the type of thing that would be redundant with information that btrfs already stores about the tree items. Having two sources of truth is error-prone.

Quoted 9 lines
fs/btrfs/verity.c new file +9 -0
+/*
+ * Insert and write inode items with a given key type and offset.
+ * @inode: The inode to insert for.
+ * @key_type: The key type to insert.
+ * @offset: The item offset to insert at.
+ * @src: Source data to write.
+ * @len: Length of source data to write.
+ * @max_item_size: Break up the write into items of this size at most.
+ *                 Useful for small leaf size file systems.

fs/btrfs/verity.c line 151

Quoted 10 lines
fs/btrfs/verity.c new file +10 -0
+ *
+ * Write len bytes from src into items of up to max_item_size length.
+ * The inserted items will have key <ino, key_type, offset + off> where
+ * off is consecutively increasing from 0 up to the last item ending at
+ * offset + len.
+ *
+ * Returns 0 on success and a negative error code on failure.
+ */
+static int write_key_bytes(struct btrfs_inode *inode, u8 key_type, u64 offset,
+			   const char *src, u64 len)

The max_item_size parameter is documented but doesn't exist.

Quoted 15 lines
fs/btrfs/verity.c new file +15 -0
+/*
+ * fsverity op that gets the struct fsverity_descriptor.
+ * fsverity does a two pass setup for reading the descriptor, in the first pass
+ * it calls with buf_size = 0 to query the size of the descriptor,
+ * and then in the second pass it actually reads the descriptor off
+ * disk.
+ */
+static int btrfs_get_verity_descriptor(struct inode *inode, void *buf,
+				       size_t buf_size)
+{
+	size_t true_size;
+	ssize_t ret = 0;
+	struct btrfs_verity_descriptor_item item;
+
+	memset(&item, 0, sizeof(item));

fs/btrfs/verity.c line 467

Quoted 10 lines
fs/btrfs/verity.c new file +10 -0
+	ret = read_key_bytes(BTRFS_I(inode), BTRFS_VERITY_DESC_ITEM_KEY,
+			     0, (char *)&item, sizeof(item), NULL);
+	if (ret < 0)
+		return ret;
+
+	true_size = btrfs_stack_verity_descriptor_size(&item);
+	if (!buf_size)
+		return true_size;
+	if (buf_size < true_size)
+		return -ERANGE;

It would be good to validate that true_size <= INT_MAX, as it's returned it an 'int'.

fs/btrfs/verity.c line 477

Quoted 10 lines
fs/btrfs/verity.c new file +10 -0
+
+	ret = read_key_bytes(BTRFS_I(inode),
+			     BTRFS_VERITY_DESC_ITEM_KEY, 1,
+			     buf, buf_size, NULL);
+	if (ret < 0)
+		return ret;
+	if (ret != buf_size)
+		return -EIO;
+
+	return buf_size;

Shouldn't this part use true_size, not buf_size?

fs/btrfs/verity.c line 586

Quoted 12 lines
fs/btrfs/verity.c new file +12 -0
+/*
+ * fsverity op that writes a merkle tree block into the btree in 1k chunks.
+ */
+static int btrfs_write_merkle_tree_block(struct inode *inode, const void *buf,
+					u64 index, int log_blocksize)
+{
+	u64 start = index << log_blocksize;
+	u64 len = 1 << log_blocksize;
+	unsigned long mapping_index = get_verity_mapping_index(inode, index);
+
+	if (mapping_index > inode->i_sb->s_maxbytes >> PAGE_SHIFT)
+		return -EFBIG;

I don't think this overflow check will work correctly, as 'mapping_index' can overflow a ULONG_MAX before the overflow check is done.

include/uapi/linux/btrfs_tree.h struct btrfs_qgroup_limit_item { · line 1012

Quoted 6 lines
include/uapi/linux/btrfs_tree.h +6 -0
+struct btrfs_verity_descriptor_item {
+	/* size of the verity descriptor in bytes */
+	__le64 size;
+	__le64 reserved[2];
+	__u8 encryption;
+} __attribute__ ((__packed__));

Should these reserved fields be strictly validated? Currently they appear to be ignored.

- Eric

#7 lore raw

Re: [PATCH v2 2/5] btrfs: initial fsverity support

Quoted 47 lines · Eric Biggers
Quoted 15 lines · Boris Burkov
fs/btrfs/extent_io.c +7 -2
+static int end_page_read(struct page *page, bool uptodate, u64 start, u32 len)
 {
-	struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb);
+	int ret = 0;
+	struct inode *inode = page->mapping->host;
+	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
 
 	ASSERT(page_offset(page) <= start &&
 		start + len <= page_offset(page) + PAGE_SIZE);
 
 	if (uptodate) {
-		btrfs_page_set_uptodate(fs_info, page, start, len);
+		/*
+		 * buffered reads of a file with page alignment will issue a
+		 * 0 length read for one page past the end of file, so we must

fs/btrfs/extent_io.c static void begin_page_read(struct btrfs_fs_info *fs_info, struct page *page) · line 2887

Quoted 10 lines
fs/btrfs/extent_io.c +9 -0
+		 * explicitly skip checking verity on that page of zeros.
+		 */
+		if (!PageError(page) && !PageUptodate(page) &&
+		    start < i_size_read(inode) &&
+		    fsverity_active(inode) &&
+		    fsverity_verify_page(page) != true)
+			ret = -EIO;
+		else
+			btrfs_page_set_uptodate(fs_info, page, start, len);
 	} else {

'fsverity_verify_page(page) != true' is better written as '!fsverity_verify_page(page)'.

Quoted 8 lines

+/*
+ * Just like ext4, we cache the merkle tree in pages after EOF in the page
+ * cache. Unlike ext4, we're storing these in dedicated btree items and
+ * not just shoving them after EOF in the file. This means we'll need to
+ * do extra work to encrypt them once encryption is supported in btrfs,
+ * but btrfs has a lot of careful code around i_size and it seems better
+ * to make a new key type than try and adjust all of our expectations
+ * for i_size.

Quoted 8 lines

+ *
+ * fs verity items are stored under two different key types on disk.
+ *
+ * The descriptor items:
+ * [ inode objectid, BTRFS_VERITY_DESC_ITEM_KEY, offset ]
+ *
+ * At offset 0, we store a btrfs_verity_descriptor_item which tracks the
+ * size of the descriptor item and some extra data for encryption.

Quoted 3 lines

Is the separate size field really needed? It seems like the type of thing that would be redundant with information that btrfs already stores about the tree items. Having two sources of truth is error-prone.

For the case where the user supplied signature is absent or fits in one 1K btrfs verity descriptor item, I believe you are right, and we could infer that from the generic item size. Assuming arbitrary signature lengths, we could still probably do it by searching for the last item and computing the size from its key offset and item size. Though now that we have a special meaning for offset 0, that gets a bit messier.

I believe we will need the special item no matter what for the eventual encryption case, so I figured it wasn't a big deal to include the size rather than try to come up with some computation to get it. I like the argument about redundant representation, and saving 8 bytes per verity enabled file isn't nothing either. I will see if I can compute it efficiently without adding the extra field.

Quoted 53 lines
Quoted 10 lines
fs/btrfs/verity.c new file +9 -0
+/*
+ * Insert and write inode items with a given key type and offset.
+ * @inode: The inode to insert for.
+ * @key_type: The key type to insert.
+ * @offset: The item offset to insert at.
+ * @src: Source data to write.
+ * @len: Length of source data to write.
+ * @max_item_size: Break up the write into items of this size at most.
+ *                 Useful for small leaf size file systems.

fs/btrfs/verity.c line 151

Quoted 10 lines
fs/btrfs/verity.c new file +10 -0
+ *
+ * Write len bytes from src into items of up to max_item_size length.
+ * The inserted items will have key <ino, key_type, offset + off> where
+ * off is consecutively increasing from 0 up to the last item ending at
+ * offset + len.
+ *
+ * Returns 0 on success and a negative error code on failure.
+ */
+static int write_key_bytes(struct btrfs_inode *inode, u8 key_type, u64 offset,
+			   const char *src, u64 len)

The max_item_size parameter is documented but doesn't exist.

Quoted 15 lines
fs/btrfs/verity.c new file +15 -0
+/*
+ * fsverity op that gets the struct fsverity_descriptor.
+ * fsverity does a two pass setup for reading the descriptor, in the first pass
+ * it calls with buf_size = 0 to query the size of the descriptor,
+ * and then in the second pass it actually reads the descriptor off
+ * disk.
+ */
+static int btrfs_get_verity_descriptor(struct inode *inode, void *buf,
+				       size_t buf_size)
+{
+	size_t true_size;
+	ssize_t ret = 0;
+	struct btrfs_verity_descriptor_item item;
+
+	memset(&item, 0, sizeof(item));

fs/btrfs/verity.c line 467

Quoted 10 lines
fs/btrfs/verity.c new file +10 -0
+	ret = read_key_bytes(BTRFS_I(inode), BTRFS_VERITY_DESC_ITEM_KEY,
+			     0, (char *)&item, sizeof(item), NULL);
+	if (ret < 0)
+		return ret;
+
+	true_size = btrfs_stack_verity_descriptor_size(&item);
+	if (!buf_size)
+		return true_size;
+	if (buf_size < true_size)
+		return -ERANGE;

It would be good to validate that true_size <= INT_MAX, as it's returned it an 'int'.

Quoted 12 lines

fs/btrfs/verity.c line 477

Quoted 10 lines
fs/btrfs/verity.c new file +10 -0
+
+	ret = read_key_bytes(BTRFS_I(inode),
+			     BTRFS_VERITY_DESC_ITEM_KEY, 1,
+			     buf, buf_size, NULL);
+	if (ret < 0)
+		return ret;
+	if (ret != buf_size)
+		return -EIO;
+
+	return buf_size;

Shouldn't this part use true_size, not buf_size?

I felt this was clunky when writing it. If I'm not mistaken, it doesn't really matter what we return since the actual key line is the ret != buf_size check above.

I think it would be reasonable to allow too-large buffers and check against/return true_size. If true_size is somehow wrong, then presumably the signature check itself will still fail.

Quoted 16 lines

fs/btrfs/verity.c line 586

Quoted 13 lines
fs/btrfs/verity.c new file +12 -0
+/*
+ * fsverity op that writes a merkle tree block into the btree in 1k chunks.
+ */
+static int btrfs_write_merkle_tree_block(struct inode *inode, const void *buf,
+					u64 index, int log_blocksize)
+{
+	u64 start = index << log_blocksize;
+	u64 len = 1 << log_blocksize;
+	unsigned long mapping_index = get_verity_mapping_index(inode, index);
+
+	if (mapping_index > inode->i_sb->s_maxbytes >> PAGE_SHIFT)
+		return -EFBIG;

I don't think this overflow check will work correctly, as 'mapping_index' can overflow a ULONG_MAX before the overflow check is done.

Oof, I had a bit of trouble with this, so it's not too surprising I got it wrong. I misunderstood the maxbytes logic for systems with 32 bit longs, and obviously only tested it my machine with 64 bit longs.

Quoted 11 lines

include/uapi/linux/btrfs_tree.h struct btrfs_qgroup_limit_item { · line 1012

Quoted 7 lines
include/uapi/linux/btrfs_tree.h +6 -0
+struct btrfs_verity_descriptor_item {
+	/* size of the verity descriptor in bytes */
+	__le64 size;
+	__le64 reserved[2];
+	__u8 encryption;
+} __attribute__ ((__packed__));

Should these reserved fields be strictly validated? Currently they appear to be ignored.

Quoted 1 line

- Eric

Thanks for the review,
Boris

maintainer
#8 lore raw

Re: [PATCH v2 2/5] btrfs: initial fsverity support

Quoted 11 lines · Boris Burkov
Quoted 3 lines

Is the separate size field really needed? It seems like the type of thing that would be redundant with information that btrfs already stores about the tree items. Having two sources of truth is error-prone.

For the case where the user supplied signature is absent or fits in one 1K btrfs verity descriptor item, I believe you are right, and we could infer that from the generic item size. Assuming arbitrary signature lengths, we could still probably do it by searching for the last item and computing the size from its key offset and item size. Though now that we have a special meaning for offset 0, that gets a bit messier.

Quoted 6 lines

I believe we will need the special item no matter what for the eventual encryption case, so I figured it wasn't a big deal to include the size rather than try to come up with some computation to get it. I like the argument about redundant representation, and saving 8 bytes per verity enabled file isn't nothing either. I will see if I can compute it efficiently without adding the extra field.

To be clear, I'm not necessarily saying "remove the size field". But if you keep it, it needs a proper explanation and careful consideration of what to do if it doesn't match the actual size.

I don't think 8 bytes per file matters, especially when the Merkle tree blocks are zero-padded to 4096 bytes anyway.

Quoted 23 lines
Quoted 10 lines

fs/btrfs/verity.c line 467

Quoted 5 lines
fs/btrfs/verity.c new file +5 -0
+	true_size = btrfs_stack_verity_descriptor_size(&item);
+	if (!buf_size)
+		return true_size;
+	if (buf_size < true_size)
+		return -ERANGE;

It would be good to validate that true_size <= INT_MAX, as it's returned it an 'int'.

Quoted 12 lines

fs/btrfs/verity.c line 477

Quoted 10 lines
fs/btrfs/verity.c new file +10 -0
+
+	ret = read_key_bytes(BTRFS_I(inode),
+			     BTRFS_VERITY_DESC_ITEM_KEY, 1,
+			     buf, buf_size, NULL);
+	if (ret < 0)
+		return ret;
+	if (ret != buf_size)
+		return -EIO;
+
+	return buf_size;

Shouldn't this part use true_size, not buf_size?

Quoted 7 lines

I felt this was clunky when writing it. If I'm not mistaken, it doesn't really matter what we return since the actual key line is the ret != buf_size check above.

I think it would be reasonable to allow too-large buffers and check against/return true_size. If true_size is somehow wrong, then presumably the signature check itself will still fail.

fsverity_operations::get_verity_descriptor() is supposed to behave like getxattr(). That means that if the buffer provided is too large, the function shouldn't fail but rather just read and return the actual size.

That means using true_size above.

I probably should have gone with something harder to get wrong, but getxattr() semantics seemed "simple".

- Eric

maintainer
#9 lore raw

Re: [PATCH v2 2/5] btrfs: initial fsverity support

In reply to Boris Burkov
Quoted 44 lines · Boris Burkov
fs/btrfs/verity.c new file +44 -0
+/*
+ * fsverity op that ends enabling verity.
+ * fsverity calls this when it's done with all of the pages in the file
+ * and all of the merkle items have been inserted.  We write the
+ * descriptor and update the inode in the btree to reflect its new life
+ * as a verity file.
+ */
+static int btrfs_end_enable_verity(struct file *filp, const void *desc,
+				  size_t desc_size, u64 merkle_tree_size)
+{
+	struct btrfs_trans_handle *trans;
+	struct inode *inode = file_inode(filp);
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_verity_descriptor_item item;
+	int ret;
+
+	if (desc != NULL) {
+		/* write out the descriptor item */
+		memset(&item, 0, sizeof(item));
+		btrfs_set_stack_verity_descriptor_size(&item, desc_size);
+		ret = write_key_bytes(BTRFS_I(inode),
+				      BTRFS_VERITY_DESC_ITEM_KEY, 0,
+				      (const char *)&item, sizeof(item));
+		if (ret)
+			goto out;
+		/* write out the descriptor itself */
+		ret = write_key_bytes(BTRFS_I(inode),
+				      BTRFS_VERITY_DESC_ITEM_KEY, 1,
+				      desc, desc_size);
+		if (ret)
+			goto out;
+
+		/* update our inode flags to include fs verity */
+		trans = btrfs_start_transaction(root, 1);
+		if (IS_ERR(trans)) {
+			ret = PTR_ERR(trans);
+			goto out;
+		}
+		BTRFS_I(inode)->compat_flags |= BTRFS_INODE_VERITY;
+		btrfs_sync_inode_flags_to_i_flags(inode);
+		ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
+		btrfs_end_transaction(trans);
+	}
+

fs/btrfs/verity.c line 441

Quoted 10 lines
fs/btrfs/verity.c new file +10 -0
+out:
+	if (desc == NULL || ret) {
+		/* If we failed, drop all the verity items */
+		drop_verity_items(BTRFS_I(inode), BTRFS_VERITY_DESC_ITEM_KEY);
+		drop_verity_items(BTRFS_I(inode), BTRFS_VERITY_MERKLE_ITEM_KEY);
+	} else
+		btrfs_set_fs_compat_ro(root->fs_info, VERITY);
+	clear_bit(BTRFS_INODE_VERITY_IN_PROGRESS, &BTRFS_I(inode)->runtime_flags);
+	return ret;
+}

If enabling verity failed, I think you also need to call truncate_inode_pages(inode->i_mapping, inode->i_size)
to remove the cached Merkle tree pages from the page cache. Otherwise they can be exposed to userspace if the file is later extended. I recently fixed this same problem for ext4 and f2fs:
https://lkml.kernel.org/linux-f2fs-devel/[email protected]/T/#u

- Eric

#10 lore raw

PATCH v2 3/5 btrfs: check verity for reads of inline extents and holes

In reply to Boris Burkov

The majority of reads receive a verity check after the bio is complete as the page is marked uptodate. However, there is a class of reads which are handled with btrfs logic in readpage, rather than by submitting a bio. Specifically, these are inline extents, preallocated extents, and holes. Tweak readpage so that if it is going to mark such a page uptodate, it first checks verity on it.

Now if a veritied file has corruption to this class of EXTENT_DATA items, it will be detected at read time.

There is one annoying edge case that requires checking for start <
last_byte: if userspace reads to the end of a file with page aligned size and then tries to keep reading (as cat does), the buffered read code will try to read the page past the end of the file, and expects it to be filled with 0s and marked uptodate. That bogus page is not part of the data hashed by verity, so we have to ignore it.

---
 fs/btrfs/extent_io.c | 26 +++++++-------------------
 1 file changed, 7 insertions(+), 19 deletions(-)
fs/btrfs/extent_io.c +7 -19
index 0d6f8b33830f..653b066f3043 100644
@@ -2201,18 +2201,6 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
 	return bitset;
 }
 
-/*
- * helper function to set a given page up to date if all the
- * extents in the tree for that page are up to date
- */
-static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
-{
-	u64 start = page_offset(page);
-	u64 end = start + PAGE_SIZE - 1;
-	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
-		SetPageUptodate(page);
-}
-
 int free_io_failure(struct extent_io_tree *failure_tree,
 		    struct extent_io_tree *io_tree,
 		    struct io_failure_record *rec)
@@ -3427,14 +3415,14 @@ int btrfs_do_readpage(struct page *page, struct extent_map **em_cached,
 					    &cached, GFP_NOFS);
 			unlock_extent_cached(tree, cur,
 					     cur + iosize - 1, &cached);
-			end_page_read(page, true, cur, iosize);
+			ret = end_page_read(page, true, cur, iosize);
 			break;
 		}
 		em = __get_extent_map(inode, page, pg_offset, cur,
 				      end - cur + 1, em_cached);
 		if (IS_ERR_OR_NULL(em)) {
 			unlock_extent(tree, cur, end);
-			end_page_read(page, false, cur, end + 1 - cur);
+			ret = end_page_read(page, false, cur, end + 1 - cur);
 			break;
 		}
 		extent_offset = cur - em->start;
@@ -3515,9 +3503,10 @@ int btrfs_do_readpage(struct page *page, struct extent_map **em_cached,
 
 			set_extent_uptodate(tree, cur, cur + iosize - 1,
 					    &cached, GFP_NOFS);
+
 			unlock_extent_cached(tree, cur,
 					     cur + iosize - 1, &cached);
-			end_page_read(page, true, cur, iosize);
+			ret = end_page_read(page, true, cur, iosize);
 			cur = cur + iosize;
 			pg_offset += iosize;
 			continue;
@@ -3525,9 +3514,8 @@ int btrfs_do_readpage(struct page *page, struct extent_map **em_cached,
 		/* the get_extent function already copied into the page */
 		if (test_range_bit(tree, cur, cur_end,
 				   EXTENT_UPTODATE, 1, NULL)) {
-			check_page_uptodate(tree, page);
 			unlock_extent(tree, cur, cur + iosize - 1);
-			end_page_read(page, true, cur, iosize);
+			ret = end_page_read(page, true, cur, iosize);
 			cur = cur + iosize;
 			pg_offset += iosize;
 			continue;
@@ -3537,7 +3525,7 @@ int btrfs_do_readpage(struct page *page, struct extent_map **em_cached,
 		 */
 		if (block_start == EXTENT_MAP_INLINE) {
 			unlock_extent(tree, cur, cur + iosize - 1);
-			end_page_read(page, false, cur, iosize);
+			ret = end_page_read(page, false, cur, iosize);
 			cur = cur + iosize;
 			pg_offset += iosize;
 			continue;
@@ -3555,7 +3543,7 @@ int btrfs_do_readpage(struct page *page, struct extent_map **em_cached,
 			*bio_flags = this_bio_flag;
 		} else {
 			unlock_extent(tree, cur, cur + iosize - 1);
-			end_page_read(page, false, cur, iosize);
+			ret = end_page_read(page, false, cur, iosize);
 			goto out;
 		}
 		cur = cur + iosize;
Signature · 1 line
2.24.1
#11 lore raw

PATCH v2 4/5 btrfs: fallback to buffered io for verity files

In reply to Boris Burkov

Reading the contents with direct IO would circumvent verity checks, so fallback to buffered reads. For what it's worth, this is how ext4 handles it as well.

---
 fs/btrfs/file.c | 3 +++
 1 file changed, 3 insertions(+)
fs/btrfs/file.c +3 -0
index 75f1406021e5..c948c30e132c 100644
@@ -3623,6 +3623,9 @@ static ssize_t btrfs_direct_read(struct kiocb *iocb, struct iov_iter *to)
 	struct inode *inode = file_inode(iocb->ki_filp);
 	ssize_t ret;
 
+	if (fsverity_active(inode))
+		return 0;
+
 	if (check_direct_read(btrfs_sb(inode->i_sb), to, iocb->ki_pos))
 		return 0;
 
Signature · 1 line
2.24.1
#12 lore raw

PATCH v2 5/5 btrfs: verity metadata orphan items

In reply to Boris Burkov

If we don't finish creating fsverity metadata for a file, or fail to clean up already created metadata after a failure, we could leak the verity items.

To address this issue, we use the orphan mechanism. When we start enabling verity on a file, we also add an orphan item for that inode. When we are finished, we delete the orphan. However, if we are interrupted midway, the orphan will be present at mount and we can cleanup the half-formed verity state.

One thing to note is that this is a resurrection of using orphans to signal orphaned metadata that isn't the inode itself. This makes the comment discussing deprecating that concept a bit messy in full context.

---
 fs/btrfs/inode.c  | 15 +++++++--
 fs/btrfs/verity.c | 80 +++++++++++++++++++++++++++++++++++++++++------
 2 files changed, 83 insertions(+), 12 deletions(-)
fs/btrfs/inode.c +13 -2
index 9291f6633bc6..7c00863ba8a3 100644
@@ -3419,7 +3419,9 @@ int btrfs_orphan_cleanup(struct btrfs_root *root)
 
 		/*
 		 * If we have an inode with links, there are a couple of
-		 * possibilities. Old kernels (before v3.12) used to create an
+		 * possibilities:
+		 *
+		 * 1. Old kernels (before v3.12) used to create an
 		 * orphan item for truncate indicating that there were possibly
 		 * extent items past i_size that needed to be deleted. In v3.12,
 		 * truncate was changed to update i_size in sync with the extent
@@ -3432,13 +3434,22 @@ int btrfs_orphan_cleanup(struct btrfs_root *root)
 		 * slim, and it's a pain to do the truncate now, so just delete
 		 * the orphan item.
 		 *
+		 * 2. We were halfway through creating fsverity metadata for the
+		 * file. In that case, the orphan item represents incomplete
+		 * fsverity metadata which must be cleaned up with
+		 * btrfs_drop_verity_items.
+		 *
 		 * It's also possible that this orphan item was supposed to be
 		 * deleted but wasn't. The inode number may have been reused,
 		 * but either way, we can delete the orphan item.
 		 */
 		if (ret == -ENOENT || inode->i_nlink) {
-			if (!ret)
+			if (!ret) {
+				ret = btrfs_drop_verity_items(BTRFS_I(inode));
 				iput(inode);
+				if (ret)
+					goto out;
+			}
 			trans = btrfs_start_transaction(root, 1);
 			if (IS_ERR(trans)) {
 				ret = PTR_ERR(trans);
fs/btrfs/verity.c +70 -10

110 lines not on this page · raw message

Signature · 1 line
2.24.1
maintainer
#13 lore raw
In reply to Boris Burkov
Quoted 16 lines · Boris Burkov

This patchset provides support for fsverity in btrfs.

At a high level, we store the verity descriptor and Merkle tree data in the file system btree with the file's inode as the objectid, and direct reads/writes to those items to implement the generic fsverity interface required by fs/verity/.

The first patch is a preparatory patch which adds a notion of compat_flags to the btrfs_inode and inode_item in order to allow enabling verity on a file without making the file system unmountable for older kernels. (It runs afoul of the leaf corruption check otherwise)

The second patch is the bulk of the fsverity implementation. It implements the fsverity interface and adds verity checks for the typical file reading case.

Quoted 8 lines

The third patch cleans up the corner cases in readpage, covering inline extents, preallocated extents, and holes.

The fourth patch handles direct io of a veritied file by falling back to buffered io.

The fifth patch handles crashes mid-verity enable via orphan items

Can you include information about how this was tested?

Also, fsverity-utils works with btrfs as-is, correct?

- Eric

#14 lore raw
Quoted 20 lines · Eric Biggers
Quoted 19 lines · Boris Burkov

This patchset provides support for fsverity in btrfs.

At a high level, we store the verity descriptor and Merkle tree data in the file system btree with the file's inode as the objectid, and direct reads/writes to those items to implement the generic fsverity interface required by fs/verity/.

The first patch is a preparatory patch which adds a notion of compat_flags to the btrfs_inode and inode_item in order to allow enabling verity on a file without making the file system unmountable for older kernels. (It runs afoul of the leaf corruption check otherwise)

The second patch is the bulk of the fsverity implementation. It implements the fsverity interface and adds verity checks for the typical file reading case.

The third patch cleans up the corner cases in readpage, covering inline extents, preallocated extents, and holes.

Quoted 7 lines
Quoted 5 lines

The fourth patch handles direct io of a veritied file by falling back to buffered io.

The fifth patch handles crashes mid-verity enable via orphan items

Can you include information about how this was tested?

Right now, I'm testing it with the btrfs xfstest I added as well as a one-off script that corrupts regular extent data. I'm still working on integrating the btrfs specifics with the generic verity xfstests, and how to test verity+compression without hacks.

Quoted 3 lines

Also, fsverity-utils works with btrfs as-is, correct?

As far as I know, yes. I've tested using both the rpm packaged by Fedora and building from source.

Quoted 1 line

- Eric