← Documents Documentation/filesystems/fsverity.rst GitHub 원문 ↗

Linux 6.18.37 · Filesystems

fs-verity: read-only file-based authenticity protection

fs-verity의 Merkle tree, ioctl API, digest·서명 정책, ext4·f2fs·btrfs 구현과 검증 경로를 다루는 전문 번역입니다.

Source pathDocumentation/filesystems/fsverity.rst
Source versionLinux v6.18.37
TranslationDUJINLABS 전문 번역 + 해설

요약·해설과 원문, 전문 번역을 서로 분리했습니다. API 이름, symbol, source path는 원문 표기를 사용합니다.

1. 요약·해설

원문의 핵심 논리와 kernel programming 관점의 보충 설명입니다. 아래의 전문 번역과는 별도로 작성했습니다.

요약·해설

fsverity.rst:1-912

fs-verity는 read-write 파일시스템에 개별 설치·갱신되는 읽기 전용 파일을 Merkle tree로 보호합니다. 상수 시간 file digest 조회와 page-in마다의 부분 검증을 결합하며, digest를 신뢰하는 방법은 사용자 공간·IMA·IPE·내장 서명 정책이 별도로 결정합니다.

운영 설계에서는 fs-verity 자체의 우발 손상 탐지와 digest 인증을 구분해야 합니다. 내장 PKCS#7·X.509 서명은 완전한 인증 정책이 아니고 시스템 전체 keyring·sysctl·algorithm 제약과 kernel attack surface가 있으므로, 단순 형식의 사용자 공간 서명 검증을 우선 비교해야 합니다.

구현은 검증 전 folio를 Uptodate로 표시하지 않고 암호화 파일은 복호화 뒤 평문을 검증합니다. ext4·f2fs는 `i_size` 뒤 metadata를 숨겨 저장하고 btrfs는 btree item을 사용하며, 각 파일시스템의 transaction으로 ENABLE 작업의 원자성을 보장합니다.

fs-verity 신뢰 사슬
data block과 Merkle tree block을 hashroot hash·file size·algorithm을 descriptor에 결속descriptor hash로 fs-verity file digest 계산사용자 공간·IMA·IPE·builtin signature가 digest 인증pagecache·bio 읽기마다 root까지 필요한 경로 검증불일치 시 EIO 또는 SIGBUS로 data 사용 차단

저장된 data에서 최종 정책 결정까지의 연결입니다.

2. 영어 원문 전체

번역 기준이 된 Linux v6.18.37 원문입니다. 줄 번호는 이 버전의 파일 좌표입니다.

원문 전체 펼치기
1 .. SPDX-License-Identifier: GPL-2.0
2
3 .. _fsverity:
4
5 =======================================================
6 fs-verity: read-only file-based authenticity protection
7 =======================================================
8
9 Introduction
10 ============
11
12 fs-verity (``fs/verity/``) is a support layer that filesystems can
13 hook into to support transparent integrity and authenticity protection
14 of read-only files. Currently, it is supported by the ext4, f2fs, and
15 btrfs filesystems. Like fscrypt, not too much filesystem-specific
16 code is needed to support fs-verity.
17
18 fs-verity is similar to `dm-verity
19 <https://www.kernel.org/doc/Documentation/admin-guide/device-mapper/verity.rst>`_
20 but works on files rather than block devices. On regular files on
21 filesystems supporting fs-verity, userspace can execute an ioctl that
22 causes the filesystem to build a Merkle tree for the file and persist
23 it to a filesystem-specific location associated with the file.
24
25 After this, the file is made readonly, and all reads from the file are
26 automatically verified against the file's Merkle tree. Reads of any
27 corrupted data, including mmap reads, will fail.
28
29 Userspace can use another ioctl to retrieve the root hash (actually
30 the "fs-verity file digest", which is a hash that includes the Merkle
31 tree root hash) that fs-verity is enforcing for the file. This ioctl
32 executes in constant time, regardless of the file size.
33
34 fs-verity is essentially a way to hash a file in constant time,
35 subject to the caveat that reads which would violate the hash will
36 fail at runtime.
37
38 Use cases
39 =========
40
41 By itself, fs-verity only provides integrity protection, i.e.
42 detection of accidental (non-malicious) corruption.
43
44 However, because fs-verity makes retrieving the file hash extremely
45 efficient, it's primarily meant to be used as a tool to support
46 authentication (detection of malicious modifications) or auditing
47 (logging file hashes before use).
48
49 A standard file hash could be used instead of fs-verity. However,
50 this is inefficient if the file is large and only a small portion may
51 be accessed. This is often the case for Android application package
52 (APK) files, for example. These typically contain many translations,
53 classes, and other resources that are infrequently or even never
54 accessed on a particular device. It would be slow and wasteful to
55 read and hash the entire file before starting the application.
56
57 Unlike an ahead-of-time hash, fs-verity also re-verifies data each
58 time it's paged in. This ensures that malicious disk firmware can't
59 undetectably change the contents of the file at runtime.
60
61 fs-verity does not replace or obsolete dm-verity. dm-verity should
62 still be used on read-only filesystems. fs-verity is for files that
63 must live on a read-write filesystem because they are independently
64 updated and potentially user-installed, so dm-verity cannot be used.
65
66 fs-verity does not mandate a particular scheme for authenticating its
67 file hashes. (Similarly, dm-verity does not mandate a particular
68 scheme for authenticating its block device root hashes.) Options for
69 authenticating fs-verity file hashes include:
70
71 - Trusted userspace code. Often, the userspace code that accesses
72 files can be trusted to authenticate them. Consider e.g. an
73 application that wants to authenticate data files before using them,
74 or an application loader that is part of the operating system (which
75 is already authenticated in a different way, such as by being loaded
76 from a read-only partition that uses dm-verity) and that wants to
77 authenticate applications before loading them. In these cases, this
78 trusted userspace code can authenticate a file's contents by
79 retrieving its fs-verity digest using `FS_IOC_MEASURE_VERITY`_, then
80 verifying a signature of it using any userspace cryptographic
81 library that supports digital signatures.
82
83 - Integrity Measurement Architecture (IMA). IMA supports fs-verity
84 file digests as an alternative to its traditional full file digests.
85 "IMA appraisal" enforces that files contain a valid, matching
86 signature in their "security.ima" extended attribute, as controlled
87 by the IMA policy. For more information, see the IMA documentation.
88
89 - Integrity Policy Enforcement (IPE). IPE supports enforcing access
90 control decisions based on immutable security properties of files,
91 including those protected by fs-verity's built-in signatures.
92 "IPE policy" specifically allows for the authorization of fs-verity
93 files using properties ``fsverity_digest`` for identifying
94 files by their verity digest, and ``fsverity_signature`` to authorize
95 files with a verified fs-verity's built-in signature. For
96 details on configuring IPE policies and understanding its operational
97 modes, please refer to :doc:`IPE admin guide </admin-guide/LSM/ipe>`.
98
99 - Trusted userspace code in combination with `Built-in signature
100 verification`_. This approach should be used only with great care.
101
102 User API
103 ========
104
105 FS_IOC_ENABLE_VERITY
106 --------------------
107
108 The FS_IOC_ENABLE_VERITY ioctl enables fs-verity on a file. It takes
109 in a pointer to a struct fsverity_enable_arg, defined as
110 follows::
111
112 struct fsverity_enable_arg {
113 __u32 version;
114 __u32 hash_algorithm;
115 __u32 block_size;
116 __u32 salt_size;
117 __u64 salt_ptr;
118 __u32 sig_size;
119 __u32 __reserved1;
120 __u64 sig_ptr;
121 __u64 __reserved2[11];
122 };
123
124 This structure contains the parameters of the Merkle tree to build for
125 the file. It must be initialized as follows:
126
127 - ``version`` must be 1.
128 - ``hash_algorithm`` must be the identifier for the hash algorithm to
129 use for the Merkle tree, such as FS_VERITY_HASH_ALG_SHA256. See
130 ``include/uapi/linux/fsverity.h`` for the list of possible values.
131 - ``block_size`` is the Merkle tree block size, in bytes. In Linux
132 v6.3 and later, this can be any power of 2 between (inclusively)
133 1024 and the minimum of the system page size and the filesystem
134 block size. In earlier versions, the page size was the only allowed
135 value.
136 - ``salt_size`` is the size of the salt in bytes, or 0 if no salt is
137 provided. The salt is a value that is prepended to every hashed
138 block; it can be used to personalize the hashing for a particular
139 file or device. Currently the maximum salt size is 32 bytes.
140 - ``salt_ptr`` is the pointer to the salt, or NULL if no salt is
141 provided.
142 - ``sig_size`` is the size of the builtin signature in bytes, or 0 if no
143 builtin signature is provided. Currently the builtin signature is
144 (somewhat arbitrarily) limited to 16128 bytes.
145 - ``sig_ptr`` is the pointer to the builtin signature, or NULL if no
146 builtin signature is provided. A builtin signature is only needed
147 if the `Built-in signature verification`_ feature is being used. It
148 is not needed for IMA appraisal, and it is not needed if the file
149 signature is being handled entirely in userspace.
150 - All reserved fields must be zeroed.
151
152 FS_IOC_ENABLE_VERITY causes the filesystem to build a Merkle tree for
153 the file and persist it to a filesystem-specific location associated
154 with the file, then mark the file as a verity file. This ioctl may
155 take a long time to execute on large files, and it is interruptible by
156 fatal signals.
157
158 FS_IOC_ENABLE_VERITY checks for write access to the inode. However,
159 it must be executed on an O_RDONLY file descriptor and no processes
160 can have the file open for writing. Attempts to open the file for
161 writing while this ioctl is executing will fail with ETXTBSY. (This
162 is necessary to guarantee that no writable file descriptors will exist
163 after verity is enabled, and to guarantee that the file's contents are
164 stable while the Merkle tree is being built over it.)
165
166 On success, FS_IOC_ENABLE_VERITY returns 0, and the file becomes a
167 verity file. On failure (including the case of interruption by a
168 fatal signal), no changes are made to the file.
169
170 FS_IOC_ENABLE_VERITY can fail with the following errors:
171
172 - ``EACCES``: the process does not have write access to the file
173 - ``EBADMSG``: the builtin signature is malformed
174 - ``EBUSY``: this ioctl is already running on the file
175 - ``EEXIST``: the file already has verity enabled
176 - ``EFAULT``: the caller provided inaccessible memory
177 - ``EFBIG``: the file is too large to enable verity on
178 - ``EINTR``: the operation was interrupted by a fatal signal
179 - ``EINVAL``: unsupported version, hash algorithm, or block size; or
180 reserved bits are set; or the file descriptor refers to neither a
181 regular file nor a directory.
182 - ``EISDIR``: the file descriptor refers to a directory
183 - ``EKEYREJECTED``: the builtin signature doesn't match the file
184 - ``EMSGSIZE``: the salt or builtin signature is too long
185 - ``ENOKEY``: the ".fs-verity" keyring doesn't contain the certificate
186 needed to verify the builtin signature
187 - ``ENOPKG``: fs-verity recognizes the hash algorithm, but it's not
188 available in the kernel as currently configured
189 - ``ENOTTY``: this type of filesystem does not implement fs-verity
190 - ``EOPNOTSUPP``: the kernel was not configured with fs-verity
191 support; or the filesystem superblock has not had the 'verity'
192 feature enabled on it; or the filesystem does not support fs-verity
193 on this file. (See `Filesystem support`_.)
194 - ``EPERM``: the file is append-only; or, a builtin signature is
195 required and one was not provided.
196 - ``EROFS``: the filesystem is read-only
197 - ``ETXTBSY``: someone has the file open for writing. This can be the
198 caller's file descriptor, another open file descriptor, or the file
199 reference held by a writable memory map.
200
201 FS_IOC_MEASURE_VERITY
202 ---------------------
203
204 The FS_IOC_MEASURE_VERITY ioctl retrieves the digest of a verity file.
205 The fs-verity file digest is a cryptographic digest that identifies
206 the file contents that are being enforced on reads; it is computed via
207 a Merkle tree and is different from a traditional full-file digest.
208
209 This ioctl takes in a pointer to a variable-length structure::
210
211 struct fsverity_digest {
212 __u16 digest_algorithm;
213 __u16 digest_size; /* input/output */
214 __u8 digest[];
215 };
216
217 ``digest_size`` is an input/output field. On input, it must be
218 initialized to the number of bytes allocated for the variable-length
219 ``digest`` field.
220
221 On success, 0 is returned and the kernel fills in the structure as
222 follows:
223
224 - ``digest_algorithm`` will be the hash algorithm used for the file
225 digest. It will match ``fsverity_enable_arg::hash_algorithm``.
226 - ``digest_size`` will be the size of the digest in bytes, e.g. 32
227 for SHA-256. (This can be redundant with ``digest_algorithm``.)
228 - ``digest`` will be the actual bytes of the digest.
229
230 FS_IOC_MEASURE_VERITY is guaranteed to execute in constant time,
231 regardless of the size of the file.
232
233 FS_IOC_MEASURE_VERITY can fail with the following errors:
234
235 - ``EFAULT``: the caller provided inaccessible memory
236 - ``ENODATA``: the file is not a verity file
237 - ``ENOTTY``: this type of filesystem does not implement fs-verity
238 - ``EOPNOTSUPP``: the kernel was not configured with fs-verity
239 support, or the filesystem superblock has not had the 'verity'
240 feature enabled on it. (See `Filesystem support`_.)
241 - ``EOVERFLOW``: the digest is longer than the specified
242 ``digest_size`` bytes. Try providing a larger buffer.
243
244 FS_IOC_READ_VERITY_METADATA
245 ---------------------------
246
247 The FS_IOC_READ_VERITY_METADATA ioctl reads verity metadata from a
248 verity file. This ioctl is available since Linux v5.12.
249
250 This ioctl is useful for cases where the verity verification should be
251 performed somewhere other than the currently running kernel.
252
253 One example is a server program that takes a verity file and serves it
254 to a client program, such that the client can do its own fs-verity
255 compatible verification of the file. This only makes sense if the
256 client doesn't trust the server and if the server needs to provide the
257 storage for the client.
258
259 Another example is copying verity metadata when creating filesystem
260 images in userspace (such as with ``mkfs.ext4 -d``).
261
262 This is a fairly specialized use case, and most fs-verity users won't
263 need this ioctl.
264
265 This ioctl takes in a pointer to the following structure::
266
267 #define FS_VERITY_METADATA_TYPE_MERKLE_TREE 1
268 #define FS_VERITY_METADATA_TYPE_DESCRIPTOR 2
269 #define FS_VERITY_METADATA_TYPE_SIGNATURE 3
270
271 struct fsverity_read_metadata_arg {
272 __u64 metadata_type;
273 __u64 offset;
274 __u64 length;
275 __u64 buf_ptr;
276 __u64 __reserved;
277 };
278
279 ``metadata_type`` specifies the type of metadata to read:
280
281 - ``FS_VERITY_METADATA_TYPE_MERKLE_TREE`` reads the blocks of the
282 Merkle tree. The blocks are returned in order from the root level
283 to the leaf level. Within each level, the blocks are returned in
284 the same order that their hashes are themselves hashed.
285 See `Merkle tree`_ for more information.
286
287 - ``FS_VERITY_METADATA_TYPE_DESCRIPTOR`` reads the fs-verity
288 descriptor. See `fs-verity descriptor`_.
289
290 - ``FS_VERITY_METADATA_TYPE_SIGNATURE`` reads the builtin signature
291 which was passed to FS_IOC_ENABLE_VERITY, if any. See `Built-in
292 signature verification`_.
293
294 The semantics are similar to those of ``pread()``. ``offset``
295 specifies the offset in bytes into the metadata item to read from, and
296 ``length`` specifies the maximum number of bytes to read from the
297 metadata item. ``buf_ptr`` is the pointer to the buffer to read into,
298 cast to a 64-bit integer. ``__reserved`` must be 0. On success, the
299 number of bytes read is returned. 0 is returned at the end of the
300 metadata item. The returned length may be less than ``length``, for
301 example if the ioctl is interrupted.
302
303 The metadata returned by FS_IOC_READ_VERITY_METADATA isn't guaranteed
304 to be authenticated against the file digest that would be returned by
305 `FS_IOC_MEASURE_VERITY`_, as the metadata is expected to be used to
306 implement fs-verity compatible verification anyway (though absent a
307 malicious disk, the metadata will indeed match). E.g. to implement
308 this ioctl, the filesystem is allowed to just read the Merkle tree
309 blocks from disk without actually verifying the path to the root node.
310
311 FS_IOC_READ_VERITY_METADATA can fail with the following errors:
312
313 - ``EFAULT``: the caller provided inaccessible memory
314 - ``EINTR``: the ioctl was interrupted before any data was read
315 - ``EINVAL``: reserved fields were set, or ``offset + length``
316 overflowed
317 - ``ENODATA``: the file is not a verity file, or
318 FS_VERITY_METADATA_TYPE_SIGNATURE was requested but the file doesn't
319 have a builtin signature
320 - ``ENOTTY``: this type of filesystem does not implement fs-verity, or
321 this ioctl is not yet implemented on it
322 - ``EOPNOTSUPP``: the kernel was not configured with fs-verity
323 support, or the filesystem superblock has not had the 'verity'
324 feature enabled on it. (See `Filesystem support`_.)
325
326 FS_IOC_GETFLAGS
327 ---------------
328
329 The existing ioctl FS_IOC_GETFLAGS (which isn't specific to fs-verity)
330 can also be used to check whether a file has fs-verity enabled or not.
331 To do so, check for FS_VERITY_FL (0x00100000) in the returned flags.
332
333 The verity flag is not settable via FS_IOC_SETFLAGS. You must use
334 FS_IOC_ENABLE_VERITY instead, since parameters must be provided.
335
336 statx
337 -----
338
339 Since Linux v5.5, the statx() system call sets STATX_ATTR_VERITY if
340 the file has fs-verity enabled. This can perform better than
341 FS_IOC_GETFLAGS and FS_IOC_MEASURE_VERITY because it doesn't require
342 opening the file, and opening verity files can be expensive.
343
344 .. _accessing_verity_files:
345
346 Accessing verity files
347 ======================
348
349 Applications can transparently access a verity file just like a
350 non-verity one, with the following exceptions:
351
352 - Verity files are readonly. They cannot be opened for writing or
353 truncate()d, even if the file mode bits allow it. Attempts to do
354 one of these things will fail with EPERM. However, changes to
355 metadata such as owner, mode, timestamps, and xattrs are still
356 allowed, since these are not measured by fs-verity. Verity files
357 can also still be renamed, deleted, and linked to.
358
359 - Direct I/O is not supported on verity files. Attempts to use direct
360 I/O on such files will fall back to buffered I/O.
361
362 - DAX (Direct Access) is not supported on verity files, because this
363 would circumvent the data verification.
364
365 - Reads of data that doesn't match the verity Merkle tree will fail
366 with EIO (for read()) or SIGBUS (for mmap() reads).
367
368 - If the sysctl "fs.verity.require_signatures" is set to 1 and the
369 file is not signed by a key in the ".fs-verity" keyring, then
370 opening the file will fail. See `Built-in signature verification`_.
371
372 Direct access to the Merkle tree is not supported. Therefore, if a
373 verity file is copied, or is backed up and restored, then it will lose
374 its "verity"-ness. fs-verity is primarily meant for files like
375 executables that are managed by a package manager.
376
377 File digest computation
378 =======================
379
380 This section describes how fs-verity hashes the file contents using a
381 Merkle tree to produce the digest which cryptographically identifies
382 the file contents. This algorithm is the same for all filesystems
383 that support fs-verity.
384
385 Userspace only needs to be aware of this algorithm if it needs to
386 compute fs-verity file digests itself, e.g. in order to sign files.
387
388 .. _fsverity_merkle_tree:
389
390 Merkle tree
391 -----------
392
393 The file contents is divided into blocks, where the block size is
394 configurable but is usually 4096 bytes. The end of the last block is
395 zero-padded if needed. Each block is then hashed, producing the first
396 level of hashes. Then, the hashes in this first level are grouped
397 into 'blocksize'-byte blocks (zero-padding the ends as needed) and
398 these blocks are hashed, producing the second level of hashes. This
399 proceeds up the tree until only a single block remains. The hash of
400 this block is the "Merkle tree root hash".
401
402 If the file fits in one block and is nonempty, then the "Merkle tree
403 root hash" is simply the hash of the single data block. If the file
404 is empty, then the "Merkle tree root hash" is all zeroes.
405
406 The "blocks" here are not necessarily the same as "filesystem blocks".
407
408 If a salt was specified, then it's zero-padded to the closest multiple
409 of the input size of the hash algorithm's compression function, e.g.
410 64 bytes for SHA-256 or 128 bytes for SHA-512. The padded salt is
411 prepended to every data or Merkle tree block that is hashed.
412
413 The purpose of the block padding is to cause every hash to be taken
414 over the same amount of data, which simplifies the implementation and
415 keeps open more possibilities for hardware acceleration. The purpose
416 of the salt padding is to make the salting "free" when the salted hash
417 state is precomputed, then imported for each hash.
418
419 Example: in the recommended configuration of SHA-256 and 4K blocks,
420 128 hash values fit in each block. Thus, each level of the Merkle
421 tree is approximately 128 times smaller than the previous, and for
422 large files the Merkle tree's size converges to approximately 1/127 of
423 the original file size. However, for small files, the padding is
424 significant, making the space overhead proportionally more.
425
426 .. _fsverity_descriptor:
427
428 fs-verity descriptor
429 --------------------
430
431 By itself, the Merkle tree root hash is ambiguous. For example, it
432 can't a distinguish a large file from a small second file whose data
433 is exactly the top-level hash block of the first file. Ambiguities
434 also arise from the convention of padding to the next block boundary.
435
436 To solve this problem, the fs-verity file digest is actually computed
437 as a hash of the following structure, which contains the Merkle tree
438 root hash as well as other fields such as the file size::
439
440 struct fsverity_descriptor {
441 __u8 version; /* must be 1 */
442 __u8 hash_algorithm; /* Merkle tree hash algorithm */
443 __u8 log_blocksize; /* log2 of size of data and tree blocks */
444 __u8 salt_size; /* size of salt in bytes; 0 if none */
445 __le32 __reserved_0x04; /* must be 0 */
446 __le64 data_size; /* size of file the Merkle tree is built over */
447 __u8 root_hash[64]; /* Merkle tree root hash */
448 __u8 salt[32]; /* salt prepended to each hashed block */
449 __u8 __reserved[144]; /* must be 0's */
450 };
451
452 Built-in signature verification
453 ===============================
454
455 CONFIG_FS_VERITY_BUILTIN_SIGNATURES=y adds supports for in-kernel
456 verification of fs-verity builtin signatures.
457
458 **IMPORTANT**! Please take great care before using this feature.
459 It is not the only way to do signatures with fs-verity, and the
460 alternatives (such as userspace signature verification, and IMA
461 appraisal) can be much better. It's also easy to fall into a trap
462 of thinking this feature solves more problems than it actually does.
463
464 Enabling this option adds the following:
465
466 1. At boot time, the kernel creates a keyring named ".fs-verity". The
467 root user can add trusted X.509 certificates to this keyring using
468 the add_key() system call.
469
470 2. `FS_IOC_ENABLE_VERITY`_ accepts a pointer to a PKCS#7 formatted
471 detached signature in DER format of the file's fs-verity digest.
472 On success, the ioctl persists the signature alongside the Merkle
473 tree. Then, any time the file is opened, the kernel verifies the
474 file's actual digest against this signature, using the certificates
475 in the ".fs-verity" keyring. This verification happens as long as the
476 file's signature exists, regardless of the state of the sysctl variable
477 "fs.verity.require_signatures" described in the next item. The IPE LSM
478 relies on this behavior to recognize and label fsverity files
479 that contain a verified built-in fsverity signature.
480
481 3. A new sysctl "fs.verity.require_signatures" is made available.
482 When set to 1, the kernel requires that all verity files have a
483 correctly signed digest as described in (2).
484
485 The data that the signature as described in (2) must be a signature of
486 is the fs-verity file digest in the following format::
487
488 struct fsverity_formatted_digest {
489 char magic[8]; /* must be "FSVerity" */
490 __le16 digest_algorithm;
491 __le16 digest_size;
492 __u8 digest[];
493 };
494
495 That's it. It should be emphasized again that fs-verity builtin
496 signatures are not the only way to do signatures with fs-verity. See
497 `Use cases`_ for an overview of ways in which fs-verity can be used.
498 fs-verity builtin signatures have some major limitations that should
499 be carefully considered before using them:
500
501 - Builtin signature verification does *not* make the kernel enforce
502 that any files actually have fs-verity enabled. Thus, it is not a
503 complete authentication policy. Currently, if it is used, one
504 way to complete the authentication policy is for trusted userspace
505 code to explicitly check whether files have fs-verity enabled with a
506 signature before they are accessed. (With
507 fs.verity.require_signatures=1, just checking whether fs-verity is
508 enabled suffices.) But, in this case the trusted userspace code
509 could just store the signature alongside the file and verify it
510 itself using a cryptographic library, instead of using this feature.
511
512 - Another approach is to utilize fs-verity builtin signature
513 verification in conjunction with the IPE LSM, which supports defining
514 a kernel-enforced, system-wide authentication policy that allows only
515 files with a verified fs-verity builtin signature to perform certain
516 operations, such as execution. Note that IPE doesn't require
517 fs.verity.require_signatures=1.
518 Please refer to :doc:`IPE admin guide </admin-guide/LSM/ipe>` for
519 more details.
520
521 - A file's builtin signature can only be set at the same time that
522 fs-verity is being enabled on the file. Changing or deleting the
523 builtin signature later requires re-creating the file.
524
525 - Builtin signature verification uses the same set of public keys for
526 all fs-verity enabled files on the system. Different keys cannot be
527 trusted for different files; each key is all or nothing.
528
529 - The sysctl fs.verity.require_signatures applies system-wide.
530 Setting it to 1 only works when all users of fs-verity on the system
531 agree that it should be set to 1. This limitation can prevent
532 fs-verity from being used in cases where it would be helpful.
533
534 - Builtin signature verification can only use signature algorithms
535 that are supported by the kernel. For example, the kernel does not
536 yet support Ed25519, even though this is often the signature
537 algorithm that is recommended for new cryptographic designs.
538
539 - fs-verity builtin signatures are in PKCS#7 format, and the public
540 keys are in X.509 format. These formats are commonly used,
541 including by some other kernel features (which is why the fs-verity
542 builtin signatures use them), and are very feature rich.
543 Unfortunately, history has shown that code that parses and handles
544 these formats (which are from the 1990s and are based on ASN.1)
545 often has vulnerabilities as a result of their complexity. This
546 complexity is not inherent to the cryptography itself.
547
548 fs-verity users who do not need advanced features of X.509 and
549 PKCS#7 should strongly consider using simpler formats, such as plain
550 Ed25519 keys and signatures, and verifying signatures in userspace.
551
552 fs-verity users who choose to use X.509 and PKCS#7 anyway should
553 still consider that verifying those signatures in userspace is more
554 flexible (for other reasons mentioned earlier in this document) and
555 eliminates the need to enable CONFIG_FS_VERITY_BUILTIN_SIGNATURES
556 and its associated increase in kernel attack surface. In some cases
557 it can even be necessary, since advanced X.509 and PKCS#7 features
558 do not always work as intended with the kernel. For example, the
559 kernel does not check X.509 certificate validity times.
560
561 Note: IMA appraisal, which supports fs-verity, does not use PKCS#7
562 for its signatures, so it partially avoids the issues discussed
563 here. IMA appraisal does use X.509.
564
565 Filesystem support
566 ==================
567
568 fs-verity is supported by several filesystems, described below. The
569 CONFIG_FS_VERITY kconfig option must be enabled to use fs-verity on
570 any of these filesystems.
571
572 ``include/linux/fsverity.h`` declares the interface between the
573 ``fs/verity/`` support layer and filesystems. Briefly, filesystems
574 must provide an ``fsverity_operations`` structure that provides
575 methods to read and write the verity metadata to a filesystem-specific
576 location, including the Merkle tree blocks and
577 ``fsverity_descriptor``. Filesystems must also call functions in
578 ``fs/verity/`` at certain times, such as when a file is opened or when
579 pages have been read into the pagecache. (See `Verifying data`_.)
580
581 ext4
582 ----
583
584 ext4 supports fs-verity since Linux v5.4 and e2fsprogs v1.45.2.
585
586 To create verity files on an ext4 filesystem, the filesystem must have
587 been formatted with ``-O verity`` or had ``tune2fs -O verity`` run on
588 it. "verity" is an RO_COMPAT filesystem feature, so once set, old
589 kernels will only be able to mount the filesystem readonly, and old
590 versions of e2fsck will be unable to check the filesystem.
591
592 Originally, an ext4 filesystem with the "verity" feature could only be
593 mounted when its block size was equal to the system page size
594 (typically 4096 bytes). In Linux v6.3, this limitation was removed.
595
596 ext4 sets the EXT4_VERITY_FL on-disk inode flag on verity files. It
597 can only be set by `FS_IOC_ENABLE_VERITY`_, and it cannot be cleared.
598
599 ext4 also supports encryption, which can be used simultaneously with
600 fs-verity. In this case, the plaintext data is verified rather than
601 the ciphertext. This is necessary in order to make the fs-verity file
602 digest meaningful, since every file is encrypted differently.
603
604 ext4 stores the verity metadata (Merkle tree and fsverity_descriptor)
605 past the end of the file, starting at the first 64K boundary beyond
606 i_size. This approach works because (a) verity files are readonly,
607 and (b) pages fully beyond i_size aren't visible to userspace but can
608 be read/written internally by ext4 with only some relatively small
609 changes to ext4. This approach avoids having to depend on the
610 EA_INODE feature and on rearchitecturing ext4's xattr support to
611 support paging multi-gigabyte xattrs into memory, and to support
612 encrypting xattrs. Note that the verity metadata *must* be encrypted
613 when the file is, since it contains hashes of the plaintext data.
614
615 ext4 only allows verity on extent-based files.
616
617 f2fs
618 ----
619
620 f2fs supports fs-verity since Linux v5.4 and f2fs-tools v1.11.0.
621
622 To create verity files on an f2fs filesystem, the filesystem must have
623 been formatted with ``-O verity``.
624
625 f2fs sets the FADVISE_VERITY_BIT on-disk inode flag on verity files.
626 It can only be set by `FS_IOC_ENABLE_VERITY`_, and it cannot be
627 cleared.
628
629 Like ext4, f2fs stores the verity metadata (Merkle tree and
630 fsverity_descriptor) past the end of the file, starting at the first
631 64K boundary beyond i_size. See explanation for ext4 above.
632 Moreover, f2fs supports at most 4096 bytes of xattr entries per inode
633 which usually wouldn't be enough for even a single Merkle tree block.
634
635 f2fs doesn't support enabling verity on files that currently have
636 atomic or volatile writes pending.
637
638 btrfs
639 -----
640
641 btrfs supports fs-verity since Linux v5.15. Verity-enabled inodes are
642 marked with a RO_COMPAT inode flag, and the verity metadata is stored
643 in separate btree items.
644
645 Implementation details
646 ======================
647
648 Verifying data
649 --------------
650
651 fs-verity ensures that all reads of a verity file's data are verified,
652 regardless of which syscall is used to do the read (e.g. mmap(),
653 read(), pread()) and regardless of whether it's the first read or a
654 later read (unless the later read can return cached data that was
655 already verified). Below, we describe how filesystems implement this.
656
657 Pagecache
658 ~~~~~~~~~
659
660 For filesystems using Linux's pagecache, the ``->read_folio()`` and
661 ``->readahead()`` methods must be modified to verify folios before
662 they are marked Uptodate. Merely hooking ``->read_iter()`` would be
663 insufficient, since ``->read_iter()`` is not used for memory maps.
664
665 Therefore, fs/verity/ provides the function fsverity_verify_blocks()
666 which verifies data that has been read into the pagecache of a verity
667 inode. The containing folio must still be locked and not Uptodate, so
668 it's not yet readable by userspace. As needed to do the verification,
669 fsverity_verify_blocks() will call back into the filesystem to read
670 hash blocks via fsverity_operations::read_merkle_tree_page().
671
672 fsverity_verify_blocks() returns false if verification failed; in this
673 case, the filesystem must not set the folio Uptodate. Following this,
674 as per the usual Linux pagecache behavior, attempts by userspace to
675 read() from the part of the file containing the folio will fail with
676 EIO, and accesses to the folio within a memory map will raise SIGBUS.
677
678 In principle, verifying a data block requires verifying the entire
679 path in the Merkle tree from the data block to the root hash.
680 However, for efficiency the filesystem may cache the hash blocks.
681 Therefore, fsverity_verify_blocks() only ascends the tree reading hash
682 blocks until an already-verified hash block is seen. It then verifies
683 the path to that block.
684
685 This optimization, which is also used by dm-verity, results in
686 excellent sequential read performance. This is because usually (e.g.
687 127 in 128 times for 4K blocks and SHA-256) the hash block from the
688 bottom level of the tree will already be cached and checked from
689 reading a previous data block. However, random reads perform worse.
690
691 Block device based filesystems
692 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
693
694 Block device based filesystems (e.g. ext4 and f2fs) in Linux also use
695 the pagecache, so the above subsection applies too. However, they
696 also usually read many data blocks from a file at once, grouped into a
697 structure called a "bio". To make it easier for these types of
698 filesystems to support fs-verity, fs/verity/ also provides a function
699 fsverity_verify_bio() which verifies all data blocks in a bio.
700
701 ext4 and f2fs also support encryption. If a verity file is also
702 encrypted, the data must be decrypted before being verified. To
703 support this, these filesystems allocate a "post-read context" for
704 each bio and store it in ``->bi_private``::
705
706 struct bio_post_read_ctx {
707 struct bio *bio;
708 struct work_struct work;
709 unsigned int cur_step;
710 unsigned int enabled_steps;
711 };
712
713 ``enabled_steps`` is a bitmask that specifies whether decryption,
714 verity, or both is enabled. After the bio completes, for each needed
715 postprocessing step the filesystem enqueues the bio_post_read_ctx on a
716 workqueue, and then the workqueue work does the decryption or
717 verification. Finally, folios where no decryption or verity error
718 occurred are marked Uptodate, and the folios are unlocked.
719
720 On many filesystems, files can contain holes. Normally,
721 ``->readahead()`` simply zeroes hole blocks and considers the
722 corresponding data to be up-to-date; no bios are issued. To prevent
723 this case from bypassing fs-verity, filesystems use
724 fsverity_verify_blocks() to verify hole blocks.
725
726 Filesystems also disable direct I/O on verity files, since otherwise
727 direct I/O would bypass fs-verity.
728
729 Userspace utility
730 =================
731
732 This document focuses on the kernel, but a userspace utility for
733 fs-verity can be found at:
734
735 https://git.kernel.org/pub/scm/fs/fsverity/fsverity-utils.git
736
737 See the README.md file in the fsverity-utils source tree for details,
738 including examples of setting up fs-verity protected files.
739
740 Tests
741 =====
742
743 To test fs-verity, use xfstests. For example, using `kvm-xfstests
744 <https://github.com/tytso/xfstests-bld/blob/master/Documentation/kvm-quickstart.md>`_::
745
746 kvm-xfstests -c ext4,f2fs,btrfs -g verity
747
748 FAQ
749 ===
750
751 This section answers frequently asked questions about fs-verity that
752 weren't already directly answered in other parts of this document.
753
754 :Q: Why isn't fs-verity part of IMA?
755 :A: fs-verity and IMA (Integrity Measurement Architecture) have
756 different focuses. fs-verity is a filesystem-level mechanism for
757 hashing individual files using a Merkle tree. In contrast, IMA
758 specifies a system-wide policy that specifies which files are
759 hashed and what to do with those hashes, such as log them,
760 authenticate them, or add them to a measurement list.
761
762 IMA supports the fs-verity hashing mechanism as an alternative
763 to full file hashes, for those who want the performance and
764 security benefits of the Merkle tree based hash. However, it
765 doesn't make sense to force all uses of fs-verity to be through
766 IMA. fs-verity already meets many users' needs even as a
767 standalone filesystem feature, and it's testable like other
768 filesystem features e.g. with xfstests.
769
770 :Q: Isn't fs-verity useless because the attacker can just modify the
771 hashes in the Merkle tree, which is stored on-disk?
772 :A: To verify the authenticity of an fs-verity file you must verify
773 the authenticity of the "fs-verity file digest", which
774 incorporates the root hash of the Merkle tree. See `Use cases`_.
775
776 :Q: Isn't fs-verity useless because the attacker can just replace a
777 verity file with a non-verity one?
778 :A: See `Use cases`_. In the initial use case, it's really trusted
779 userspace code that authenticates the files; fs-verity is just a
780 tool to do this job efficiently and securely. The trusted
781 userspace code will consider non-verity files to be inauthentic.
782
783 :Q: Why does the Merkle tree need to be stored on-disk? Couldn't you
784 store just the root hash?
785 :A: If the Merkle tree wasn't stored on-disk, then you'd have to
786 compute the entire tree when the file is first accessed, even if
787 just one byte is being read. This is a fundamental consequence of
788 how Merkle tree hashing works. To verify a leaf node, you need to
789 verify the whole path to the root hash, including the root node
790 (the thing which the root hash is a hash of). But if the root
791 node isn't stored on-disk, you have to compute it by hashing its
792 children, and so on until you've actually hashed the entire file.
793
794 That defeats most of the point of doing a Merkle tree-based hash,
795 since if you have to hash the whole file ahead of time anyway,
796 then you could simply do sha256(file) instead. That would be much
797 simpler, and a bit faster too.
798
799 It's true that an in-memory Merkle tree could still provide the
800 advantage of verification on every read rather than just on the
801 first read. However, it would be inefficient because every time a
802 hash page gets evicted (you can't pin the entire Merkle tree into
803 memory, since it may be very large), in order to restore it you
804 again need to hash everything below it in the tree. This again
805 defeats most of the point of doing a Merkle tree-based hash, since
806 a single block read could trigger re-hashing gigabytes of data.
807
808 :Q: But couldn't you store just the leaf nodes and compute the rest?
809 :A: See previous answer; this really just moves up one level, since
810 one could alternatively interpret the data blocks as being the
811 leaf nodes of the Merkle tree. It's true that the tree can be
812 computed much faster if the leaf level is stored rather than just
813 the data, but that's only because each level is less than 1% the
814 size of the level below (assuming the recommended settings of
815 SHA-256 and 4K blocks). For the exact same reason, by storing
816 "just the leaf nodes" you'd already be storing over 99% of the
817 tree, so you might as well simply store the whole tree.
818
819 :Q: Can the Merkle tree be built ahead of time, e.g. distributed as
820 part of a package that is installed to many computers?
821 :A: This isn't currently supported. It was part of the original
822 design, but was removed to simplify the kernel UAPI and because it
823 wasn't a critical use case. Files are usually installed once and
824 used many times, and cryptographic hashing is somewhat fast on
825 most modern processors.
826
827 :Q: Why doesn't fs-verity support writes?
828 :A: Write support would be very difficult and would require a
829 completely different design, so it's well outside the scope of
830 fs-verity. Write support would require:
831
832 - A way to maintain consistency between the data and hashes,
833 including all levels of hashes, since corruption after a crash
834 (especially of potentially the entire file!) is unacceptable.
835 The main options for solving this are data journalling,
836 copy-on-write, and log-structured volume. But it's very hard to
837 retrofit existing filesystems with new consistency mechanisms.
838 Data journalling is available on ext4, but is very slow.
839
840 - Rebuilding the Merkle tree after every write, which would be
841 extremely inefficient. Alternatively, a different authenticated
842 dictionary structure such as an "authenticated skiplist" could
843 be used. However, this would be far more complex.
844
845 Compare it to dm-verity vs. dm-integrity. dm-verity is very
846 simple: the kernel just verifies read-only data against a
847 read-only Merkle tree. In contrast, dm-integrity supports writes
848 but is slow, is much more complex, and doesn't actually support
849 full-device authentication since it authenticates each sector
850 independently, i.e. there is no "root hash". It doesn't really
851 make sense for the same device-mapper target to support these two
852 very different cases; the same applies to fs-verity.
853
854 :Q: Since verity files are immutable, why isn't the immutable bit set?
855 :A: The existing "immutable" bit (FS_IMMUTABLE_FL) already has a
856 specific set of semantics which not only make the file contents
857 read-only, but also prevent the file from being deleted, renamed,
858 linked to, or having its owner or mode changed. These extra
859 properties are unwanted for fs-verity, so reusing the immutable
860 bit isn't appropriate.
861
862 :Q: Why does the API use ioctls instead of setxattr() and getxattr()?
863 :A: Abusing the xattr interface for basically arbitrary syscalls is
864 heavily frowned upon by most of the Linux filesystem developers.
865 An xattr should really just be an xattr on-disk, not an API to
866 e.g. magically trigger construction of a Merkle tree.
867
868 :Q: Does fs-verity support remote filesystems?
869 :A: So far all filesystems that have implemented fs-verity support are
870 local filesystems, but in principle any filesystem that can store
871 per-file verity metadata can support fs-verity, regardless of
872 whether it's local or remote. Some filesystems may have fewer
873 options of where to store the verity metadata; one possibility is
874 to store it past the end of the file and "hide" it from userspace
875 by manipulating i_size. The data verification functions provided
876 by ``fs/verity/`` also assume that the filesystem uses the Linux
877 pagecache, but both local and remote filesystems normally do so.
878
879 :Q: Why is anything filesystem-specific at all? Shouldn't fs-verity
880 be implemented entirely at the VFS level?
881 :A: There are many reasons why this is not possible or would be very
882 difficult, including the following:
883
884 - To prevent bypassing verification, folios must not be marked
885 Uptodate until they've been verified. Currently, each
886 filesystem is responsible for marking folios Uptodate via
887 ``->readahead()``. Therefore, currently it's not possible for
888 the VFS to do the verification on its own. Changing this would
889 require significant changes to the VFS and all filesystems.
890
891 - It would require defining a filesystem-independent way to store
892 the verity metadata. Extended attributes don't work for this
893 because (a) the Merkle tree may be gigabytes, but many
894 filesystems assume that all xattrs fit into a single 4K
895 filesystem block, and (b) ext4 and f2fs encryption doesn't
896 encrypt xattrs, yet the Merkle tree *must* be encrypted when the
897 file contents are, because it stores hashes of the plaintext
898 file contents.
899
900 So the verity metadata would have to be stored in an actual
901 file. Using a separate file would be very ugly, since the
902 metadata is fundamentally part of the file to be protected, and
903 it could cause problems where users could delete the real file
904 but not the metadata file or vice versa. On the other hand,
905 having it be in the same file would break applications unless
906 filesystems' notion of i_size were divorced from the VFS's,
907 which would be complex and require changes to all filesystems.
908
909 - It's desirable that FS_IOC_ENABLE_VERITY uses the filesystem's
910 transaction mechanism so that either the file ends up with
911 verity enabled, or no changes were made. Allowing intermediate
912 states to occur after a crash may cause problems.
913

3. 한국어 전문 번역

영어 원문의 문단 순서와 의미를 유지한 전체 번역입니다. 코드, 함수명, symbol과 URL은 원문 표기를 유지합니다.

개요와 인증·감사 사용 사례

1-104

fs-verity(`fs/verity/`)는 파일시스템이 연결하여 읽기 전용 파일의 무결성과 진본성을 투명하게 보호할 수 있게 하는 지원 계층입니다. 현재 ext4, f2fs, btrfs가 지원하며 fscrypt와 마찬가지로 파일시스템별 코드는 많지 않습니다.

dm-verity가 블록 장치에 적용되는 것과 달리 fs-verity는 개별 파일에 적용됩니다. 지원 파일시스템의 일반 파일에 ioctl을 실행하면 파일시스템이 파일의 Merkle tree를 만들고 그 파일에 연결된 파일시스템별 위치에 영속화합니다.

활성화가 끝난 파일은 읽기 전용이 되고 이후 모든 읽기를 Merkle tree와 자동 대조합니다. 손상된 데이터의 `read()`뿐 아니라 `mmap()` 읽기도 실패합니다. 별도 ioctl은 tree root hash 자체가 아니라 이를 포함하는 fs-verity file digest를 반환하며, 파일 크기와 무관하게 상수 시간에 실행됩니다.

따라서 fs-verity는 런타임에서 digest를 위반하는 읽기를 실패시키는 조건 아래 파일을 상수 시간에 해시하는 수단으로 볼 수 있습니다.

fs-verity 자체만으로는 우발적이고 비악의적인 손상을 탐지하는 무결성 보호를 제공합니다. 그러나 파일 hash 조회가 매우 효율적이므로 실제 주된 목적은 악의적 변경을 탐지하는 인증이나 사용 전 hash를 기록하는 감사의 기반이 되는 것입니다.

일반적인 전체 파일 hash도 쓸 수 있지만 큰 파일에서 일부만 읽을 때 비효율적입니다. Android APK처럼 번역·class·resource 중 장치에서 거의 또는 전혀 접근하지 않는 데이터가 많은 파일을 시작 전에 전부 읽고 hash하면 느리고 낭비가 큽니다. fs-verity는 필요한 data block만 검증합니다.

사전 계산 hash와 달리 데이터가 page-in될 때마다 다시 검증하므로 악의적인 disk firmware가 실행 중 파일 내용을 몰래 바꾸는 것도 탐지합니다.

fs-verity는 dm-verity를 대체하지 않습니다. 읽기 전용 파일시스템에는 계속 dm-verity를 사용해야 합니다. fs-verity는 독립적으로 갱신되거나 사용자가 설치하여 read-write 파일시스템에 있어야 하는 개별 파일을 보호합니다.

file digest의 인증 방식은 강제하지 않습니다. 신뢰하는 사용자 공간 코드는 `FS_IOC_MEASURE_VERITY`로 digest를 얻고 원하는 digital signature 라이브러리로 서명을 검증할 수 있습니다. 운영체제가 이미 dm-verity partition에서 인증되어 로드한 application loader나 data file을 검증하는 application이 대표적인 예입니다.

IMA는 전통적인 전체 파일 digest 대신 fs-verity digest를 지원하며, IMA policy가 제어하는 `security.ima` xattr의 일치하는 유효 서명을 IMA appraisal로 강제할 수 있습니다.

IPE는 변경 불가능한 보안 속성에 따른 접근 결정을 강제합니다. `fsverity_digest`로 digest에 따라 파일을 식별하고 `fsverity_signature`로 검증된 내장 fs-verity 서명을 가진 파일을 허가할 수 있습니다. 구성과 동작 모드는 `IPE admin guide </admin-guide/LSM/ipe>`를 참고합니다.

신뢰하는 사용자 공간과 fs-verity의 내장 서명 검증을 함께 쓸 수도 있지만 매우 신중해야 합니다. 내장 서명은 가능한 여러 인증 방식 가운데 하나이며 완전한 인증 정책을 자동으로 제공하지 않습니다.

fs-verity 보호 흐름
읽기 가능한 완성 파일 준비ENABLE_VERITY가 Merkle tree와 descriptor 생성파일을 읽기 전용 verity 상태로 전환MEASURE_VERITY로 상수 시간 file digest 조회사용자 공간·IMA·IPE·내장 서명 중 정책에 맞게 인증각 page-in에서 Merkle 경로를 재검증하고 손상 읽기 차단

파일 설치에서 런타임 검증까지의 기본 수명 주기입니다.

인증 방식 선택
방식검증 위치특징
신뢰 사용자 공간애플리케이션·loader서명 형식과 키 범위를 자유롭게 설계
IMA appraisal커널 IMA policy`security.ima` 서명 강제
IPE커널 LSM policy`fsverity_digest`·`fsverity_signature` 속성 사용
내장 서명fs-verity 커널 코드PKCS#7·X.509, 사용 전 한계 검토 필수

fs-verity digest를 신뢰 정책에 연결하는 주요 방식입니다.

.. SPDX-License-Identifier: GPL-2.0

.. _fsverity:

=======================================================
fs-verity: read-only file-based authenticity protection
=======================================================

Introduction
============

fs-verity (``fs/verity/``) is a support layer that filesystems can
hook into to support transparent integrity and authenticity protection
of read-only files.  Currently, it is supported by the ext4, f2fs, and
btrfs filesystems.  Like fscrypt, not too much filesystem-specific
code is needed to support fs-verity.

fs-verity is similar to `dm-verity
<https://www.kernel.org/doc/Documentation/admin-guide/device-mapper/verity.rst>`_
but works on files rather than block devices.  On regular files on
filesystems supporting fs-verity, userspace can execute an ioctl that
causes the filesystem to build a Merkle tree for the file and persist
it to a filesystem-specific location associated with the file.

After this, the file is made readonly, and all reads from the file are
automatically verified against the file's Merkle tree.  Reads of any
corrupted data, including mmap reads, will fail.

Userspace can use another ioctl to retrieve the root hash (actually
the "fs-verity file digest", which is a hash that includes the Merkle
tree root hash) that fs-verity is enforcing for the file.  This ioctl
executes in constant time, regardless of the file size.

fs-verity is essentially a way to hash a file in constant time,
subject to the caveat that reads which would violate the hash will
fail at runtime.

Use cases
=========

By itself, fs-verity only provides integrity protection, i.e.
detection of accidental (non-malicious) corruption.

However, because fs-verity makes retrieving the file hash extremely
efficient, it's primarily meant to be used as a tool to support
authentication (detection of malicious modifications) or auditing
(logging file hashes before use).

A standard file hash could be used instead of fs-verity.  However,
this is inefficient if the file is large and only a small portion may
be accessed.  This is often the case for Android application package
(APK) files, for example.  These typically contain many translations,
classes, and other resources that are infrequently or even never
accessed on a particular device.  It would be slow and wasteful to
read and hash the entire file before starting the application.

Unlike an ahead-of-time hash, fs-verity also re-verifies data each
time it's paged in.  This ensures that malicious disk firmware can't
undetectably change the contents of the file at runtime.

fs-verity does not replace or obsolete dm-verity.  dm-verity should
still be used on read-only filesystems.  fs-verity is for files that
must live on a read-write filesystem because they are independently
updated and potentially user-installed, so dm-verity cannot be used.

fs-verity does not mandate a particular scheme for authenticating its
file hashes.  (Similarly, dm-verity does not mandate a particular
scheme for authenticating its block device root hashes.)  Options for
authenticating fs-verity file hashes include:

- Trusted userspace code.  Often, the userspace code that accesses
  files can be trusted to authenticate them.  Consider e.g. an
  application that wants to authenticate data files before using them,
  or an application loader that is part of the operating system (which
  is already authenticated in a different way, such as by being loaded
  from a read-only partition that uses dm-verity) and that wants to
  authenticate applications before loading them.  In these cases, this
  trusted userspace code can authenticate a file's contents by
  retrieving its fs-verity digest using `FS_IOC_MEASURE_VERITY`_, then
  verifying a signature of it using any userspace cryptographic
  library that supports digital signatures.

- Integrity Measurement Architecture (IMA).  IMA supports fs-verity
  file digests as an alternative to its traditional full file digests.
  "IMA appraisal" enforces that files contain a valid, matching
  signature in their "security.ima" extended attribute, as controlled
  by the IMA policy.  For more information, see the IMA documentation.

- Integrity Policy Enforcement (IPE).  IPE supports enforcing access
  control decisions based on immutable security properties of files,
  including those protected by fs-verity's built-in signatures.
  "IPE policy" specifically allows for the authorization of fs-verity
  files using properties ``fsverity_digest`` for identifying
  files by their verity digest, and ``fsverity_signature`` to authorize
  files with a verified fs-verity's built-in signature. For
  details on configuring IPE policies and understanding its operational
  modes, please refer to :doc:`IPE admin guide </admin-guide/LSM/ipe>`.

- Trusted userspace code in combination with `Built-in signature
  verification`_.  This approach should be used only with great care.

User API
========

FS_IOC_ENABLE_VERITY와 digest 측정

105-243

`FS_IOC_ENABLE_VERITY`는 파일에서 fs-verity를 활성화하며 `struct fsverity_enable_arg` 포인터를 인수로 받습니다.

struct fsverity_enable_arg {
        __u32 version;
        __u32 hash_algorithm;
        __u32 block_size;
        __u32 salt_size;
        __u64 salt_ptr;
        __u32 sig_size;
        __u32 __reserved1;
        __u64 sig_ptr;
        __u64 __reserved2[11];
};

`version`은 1이어야 합니다. `hash_algorithm`은 `FS_VERITY_HASH_ALG_SHA256` 같은 Merkle tree hash 식별자이며 가능한 값은 `include/uapi/linux/fsverity.h`에 정의됩니다.

`block_size`는 Merkle tree block 크기입니다. Linux 6.3 이상에서는 1024부터 system page size와 filesystem block size 중 작은 값까지의 2의 거듭제곱을 허용합니다. 이전 커널은 page size만 허용했습니다.

`salt_size`는 salt 바이트 수이며 없으면 0입니다. salt는 hash하는 모든 block 앞에 붙여 특정 파일이나 장치에 hash를 개인화할 수 있고 현재 최대 32바이트입니다. `salt_ptr`은 salt 주소이며 없으면 NULL입니다.

`sig_size`는 builtin signature 길이이며 없으면 0, 현재 상한은 다소 임의로 정한 16,128바이트입니다. `sig_ptr`은 서명 주소이며 없으면 NULL입니다. 내장 서명 검증을 쓸 때만 필요하고 IMA appraisal이나 사용자 공간에서 전부 처리하는 서명에는 필요하지 않습니다. 모든 reserved field는 0이어야 합니다.

ioctl은 Merkle tree를 만들고 파일에 연결된 파일시스템별 위치에 저장한 뒤 파일을 verity file로 표시합니다. 큰 파일에서는 오래 걸릴 수 있고 fatal signal로 중단할 수 있습니다.

inode에 대한 write access를 검사하지만 fd 자체는 `O_RDONLY`로 열어야 하며 어떤 프로세스도 파일을 쓰기 위해 열고 있으면 안 됩니다. 실행 중 새 write open은 `ETXTBSY`입니다. 활성화 뒤 writable fd가 남지 않고 tree 생성 중 내용이 안정적임을 보장하기 위한 조건입니다.

성공하면 0을 반환하고 파일이 verity file이 됩니다. fatal signal 중단을 포함해 실패하면 파일에는 아무 변경도 하지 않습니다. 즉 파일시스템 transaction과 결합된 all-or-nothing 동작입니다.

ENABLE 오류는 write access 없음 `EACCES`, 잘못된 builtin signature `EBADMSG`, 같은 ioctl 실행 중 `EBUSY`, 이미 활성화 `EEXIST`, 접근 불가 사용자 메모리 `EFAULT`, 지나치게 큰 파일 `EFBIG`, fatal signal `EINTR`, 버전·algorithm·block size·reserved·fd 형식 오류 `EINVAL`, 디렉터리 `EISDIR`입니다.

서명이 file digest와 불일치하면 `EKEYREJECTED`, salt 또는 signature가 너무 길면 `EMSGSIZE`, `.fs-verity` keyring에 필요한 certificate가 없으면 `ENOKEY`, algorithm은 알지만 커널 구성에 없으면 `ENOPKG`입니다.

파일시스템이 ioctl을 구현하지 않으면 `ENOTTY`, 커널 구성·superblock verity feature·해당 파일의 지원 조건이 맞지 않으면 `EOPNOTSUPP`, append-only 또는 필수 서명 누락이면 `EPERM`, read-only filesystem이면 `EROFS`, fd·다른 fd·writable mmap이 쓰기 참조를 보유하면 `ETXTBSY`입니다.

`FS_IOC_MEASURE_VERITY`는 읽을 때 강제되는 파일 내용을 암호학적으로 식별하는 fs-verity file digest를 가져옵니다. Merkle tree로 계산되므로 전통적인 전체 파일 digest와 다릅니다.

struct fsverity_digest {
        __u16 digest_algorithm;
        __u16 digest_size; /* input/output */
        __u8 digest[];
};

호출자는 가변 `digest`에 할당한 바이트 수로 입력 `digest_size`를 초기화합니다. 성공하면 0과 함께 tree에 사용한 `digest_algorithm`, 실제 길이 `digest_size`(SHA-256은 32), digest 바이트를 반환합니다. algorithm은 `fsverity_enable_arg::hash_algorithm`과 일치합니다.

MEASURE는 파일 크기와 관계없이 상수 시간입니다. 오류는 접근 불가 메모리 `EFAULT`, verity file 아님 `ENODATA`, 파일시스템 미구현 `ENOTTY`, 커널 또는 superblock 미지원 `EOPNOTSUPP`, 버퍼가 작음 `EOVERFLOW`이며 마지막 경우 더 큰 버퍼로 재시도합니다.

ENABLE_VERITY 인수
필드의미규칙
`version`UAPI 버전1
`hash_algorithm`tree hashuapi의 지원 식별자
`block_size`data·tree block 크기6.3+: 1024..min(page, fs block)의 2의 거듭제곱
`salt_size`, `salt_ptr`block 앞에 붙일 salt최대 32바이트 또는 없음
`sig_size`, `sig_ptr`DER PKCS#7 내장 서명최대 16,128바이트 또는 없음
reserved미래 확장모두 0

Merkle tree 생성에 들어가는 값과 제약입니다.

ENABLE_VERITY 오류 분류
분류errno
파일 상태·동시성`EBUSY`, `EEXIST`, `EFBIG`, `EISDIR`, `ETXTBSY`
인수·메모리`EFAULT`, `EINVAL`, `EMSGSIZE`
서명·키`EBADMSG`, `EKEYREJECTED`, `ENOKEY`, `EPERM`
기능·구성`ENOPKG`, `ENOTTY`, `EOPNOTSUPP`
접근·중단·마운트`EACCES`, `EINTR`, `EROFS`

운영자가 실패 원인을 빠르게 구분하도록 묶었습니다.

FS_IOC_ENABLE_VERITY
--------------------

The FS_IOC_ENABLE_VERITY ioctl enables fs-verity on a file.  It takes
in a pointer to a struct fsverity_enable_arg, defined as
follows::

    struct fsverity_enable_arg {
            __u32 version;
            __u32 hash_algorithm;
            __u32 block_size;
            __u32 salt_size;
            __u64 salt_ptr;
            __u32 sig_size;
            __u32 __reserved1;
            __u64 sig_ptr;
            __u64 __reserved2[11];
    };

This structure contains the parameters of the Merkle tree to build for
the file.  It must be initialized as follows:

- ``version`` must be 1.
- ``hash_algorithm`` must be the identifier for the hash algorithm to
  use for the Merkle tree, such as FS_VERITY_HASH_ALG_SHA256.  See
  ``include/uapi/linux/fsverity.h`` for the list of possible values.
- ``block_size`` is the Merkle tree block size, in bytes.  In Linux
  v6.3 and later, this can be any power of 2 between (inclusively)
  1024 and the minimum of the system page size and the filesystem
  block size.  In earlier versions, the page size was the only allowed
  value.
- ``salt_size`` is the size of the salt in bytes, or 0 if no salt is
  provided.  The salt is a value that is prepended to every hashed
  block; it can be used to personalize the hashing for a particular
  file or device.  Currently the maximum salt size is 32 bytes.
- ``salt_ptr`` is the pointer to the salt, or NULL if no salt is
  provided.
- ``sig_size`` is the size of the builtin signature in bytes, or 0 if no
  builtin signature is provided.  Currently the builtin signature is
  (somewhat arbitrarily) limited to 16128 bytes.
- ``sig_ptr``  is the pointer to the builtin signature, or NULL if no
  builtin signature is provided.  A builtin signature is only needed
  if the `Built-in signature verification`_ feature is being used.  It
  is not needed for IMA appraisal, and it is not needed if the file
  signature is being handled entirely in userspace.
- All reserved fields must be zeroed.

FS_IOC_ENABLE_VERITY causes the filesystem to build a Merkle tree for
the file and persist it to a filesystem-specific location associated
with the file, then mark the file as a verity file.  This ioctl may
take a long time to execute on large files, and it is interruptible by
fatal signals.

FS_IOC_ENABLE_VERITY checks for write access to the inode.  However,
it must be executed on an O_RDONLY file descriptor and no processes
can have the file open for writing.  Attempts to open the file for
writing while this ioctl is executing will fail with ETXTBSY.  (This
is necessary to guarantee that no writable file descriptors will exist
after verity is enabled, and to guarantee that the file's contents are
stable while the Merkle tree is being built over it.)

On success, FS_IOC_ENABLE_VERITY returns 0, and the file becomes a
verity file.  On failure (including the case of interruption by a
fatal signal), no changes are made to the file.

FS_IOC_ENABLE_VERITY can fail with the following errors:

- ``EACCES``: the process does not have write access to the file
- ``EBADMSG``: the builtin signature is malformed
- ``EBUSY``: this ioctl is already running on the file
- ``EEXIST``: the file already has verity enabled
- ``EFAULT``: the caller provided inaccessible memory
- ``EFBIG``: the file is too large to enable verity on
- ``EINTR``: the operation was interrupted by a fatal signal
- ``EINVAL``: unsupported version, hash algorithm, or block size; or
  reserved bits are set; or the file descriptor refers to neither a
  regular file nor a directory.
- ``EISDIR``: the file descriptor refers to a directory
- ``EKEYREJECTED``: the builtin signature doesn't match the file
- ``EMSGSIZE``: the salt or builtin signature is too long
- ``ENOKEY``: the ".fs-verity" keyring doesn't contain the certificate
  needed to verify the builtin signature
- ``ENOPKG``: fs-verity recognizes the hash algorithm, but it's not
  available in the kernel as currently configured
- ``ENOTTY``: this type of filesystem does not implement fs-verity
- ``EOPNOTSUPP``: the kernel was not configured with fs-verity
  support; or the filesystem superblock has not had the 'verity'
  feature enabled on it; or the filesystem does not support fs-verity
  on this file.  (See `Filesystem support`_.)
- ``EPERM``: the file is append-only; or, a builtin signature is
  required and one was not provided.
- ``EROFS``: the filesystem is read-only
- ``ETXTBSY``: someone has the file open for writing.  This can be the
  caller's file descriptor, another open file descriptor, or the file
  reference held by a writable memory map.

FS_IOC_MEASURE_VERITY
---------------------

The FS_IOC_MEASURE_VERITY ioctl retrieves the digest of a verity file.
The fs-verity file digest is a cryptographic digest that identifies
the file contents that are being enforced on reads; it is computed via
a Merkle tree and is different from a traditional full-file digest.

This ioctl takes in a pointer to a variable-length structure::

    struct fsverity_digest {
            __u16 digest_algorithm;
            __u16 digest_size; /* input/output */
            __u8 digest[];
    };

``digest_size`` is an input/output field.  On input, it must be
initialized to the number of bytes allocated for the variable-length
``digest`` field.

On success, 0 is returned and the kernel fills in the structure as
follows:

- ``digest_algorithm`` will be the hash algorithm used for the file
  digest.  It will match ``fsverity_enable_arg::hash_algorithm``.
- ``digest_size`` will be the size of the digest in bytes, e.g. 32
  for SHA-256.  (This can be redundant with ``digest_algorithm``.)
- ``digest`` will be the actual bytes of the digest.

FS_IOC_MEASURE_VERITY is guaranteed to execute in constant time,
regardless of the size of the file.

FS_IOC_MEASURE_VERITY can fail with the following errors:

- ``EFAULT``: the caller provided inaccessible memory
- ``ENODATA``: the file is not a verity file
- ``ENOTTY``: this type of filesystem does not implement fs-verity
- ``EOPNOTSUPP``: the kernel was not configured with fs-verity
  support, or the filesystem superblock has not had the 'verity'
  feature enabled on it.  (See `Filesystem support`_.)
- ``EOVERFLOW``: the digest is longer than the specified
  ``digest_size`` bytes.  Try providing a larger buffer.

Verity metadata 조회와 상태 확인

244-344

Linux 5.12부터 `FS_IOC_READ_VERITY_METADATA`는 verity file의 metadata를 읽습니다. 현재 커널이 아닌 다른 곳에서 fs-verity 호환 검증을 수행해야 하는 전문 용도입니다.

예를 들어 신뢰하지 않는 server가 storage를 제공하고 client가 자체 검증하도록 verity file을 전송하거나, `mkfs.ext4 -d`처럼 사용자 공간에서 filesystem image를 만들 때 verity metadata를 복사하는 데 사용할 수 있습니다. 일반 fs-verity 사용자는 대부분 필요하지 않습니다.

#define FS_VERITY_METADATA_TYPE_MERKLE_TREE 1
#define FS_VERITY_METADATA_TYPE_DESCRIPTOR  2
#define FS_VERITY_METADATA_TYPE_SIGNATURE   3

struct fsverity_read_metadata_arg {
        __u64 metadata_type;
        __u64 offset;
        __u64 length;
        __u64 buf_ptr;
        __u64 __reserved;
};

`FS_VERITY_METADATA_TYPE_MERKLE_TREE`는 root level부터 leaf level 순으로 tree block을 반환합니다. 각 level 안에서는 그 hash들이 다시 hash되는 순서와 같은 순서입니다. `DESCRIPTOR`는 fs-verity descriptor, `SIGNATURE`는 ENABLE에 전달한 builtin signature가 있을 때 이를 읽습니다.

동작은 `pread()`와 유사합니다. `offset`은 metadata item 안의 byte offset, `length`는 최대 읽기 길이, `buf_ptr`은 64비트 정수로 cast한 buffer 주소이며 `__reserved`는 0입니다. 성공하면 실제 읽은 byte 수, item 끝에서는 0을 반환합니다. 중단 등으로 요청보다 적게 반환할 수 있습니다.

READ_METADATA가 반환한 정보는 `FS_IOC_MEASURE_VERITY` digest에 대해 인증되었다고 보장되지 않습니다. 이 ioctl의 목적 자체가 별도 fs-verity 호환 검증 구현이므로 파일시스템은 root 경로를 검증하지 않고 disk의 tree block을 그대로 읽을 수 있습니다. 악의적 disk가 없다면 실제로는 일치합니다.

오류는 접근 불가 메모리 `EFAULT`, 한 byte도 읽기 전 중단 `EINTR`, reserved 설정 또는 `offset + length` overflow `EINVAL`, verity file이 아니거나 요청한 signature가 없음 `ENODATA`, 파일시스템 또는 해당 ioctl 미구현 `ENOTTY`, 커널·superblock 미지원 `EOPNOTSUPP`입니다.

기존 `FS_IOC_GETFLAGS`의 반환 flag에 `FS_VERITY_FL`(0x00100000)이 있는지 검사해 활성화 여부를 알 수 있습니다. 이 flag는 `FS_IOC_SETFLAGS`로 설정할 수 없으며 parameter가 필요한 `FS_IOC_ENABLE_VERITY`만 사용해야 합니다.

Linux 5.5부터 `statx()`는 활성화 파일에 `STATX_ATTR_VERITY`를 설정합니다. 파일을 열 필요가 없어 `FS_IOC_GETFLAGS`나 `FS_IOC_MEASURE_VERITY`보다 빠를 수 있으며 verity file open 자체가 비쌀 수 있는 상황에 적합합니다.

READ_METADATA 형식
type반환대표 용도
`MERKLE_TREE`root에서 leaf 순의 block외부 verifier·image 생성
`DESCRIPTOR``fsverity_descriptor`digest 재계산
`SIGNATURE`ENABLE 시 저장한 builtin signature서명 전달·검사

metadata_type별 반환 단위와 사용 목적입니다.

활성화 상태 확인
API반환 정보비용·주의
`statx()``STATX_ATTR_VERITY`파일 open 불필요
`FS_IOC_GETFLAGS``FS_VERITY_FL`상태만 확인
`FS_IOC_MEASURE_VERITY`실제 file digest상수 시간, file open 필요

필요한 정보의 깊이에 따라 API를 선택합니다.

FS_IOC_READ_VERITY_METADATA
---------------------------

The FS_IOC_READ_VERITY_METADATA ioctl reads verity metadata from a
verity file.  This ioctl is available since Linux v5.12.

This ioctl is useful for cases where the verity verification should be
performed somewhere other than the currently running kernel.

One example is a server program that takes a verity file and serves it
to a client program, such that the client can do its own fs-verity
compatible verification of the file.  This only makes sense if the
client doesn't trust the server and if the server needs to provide the
storage for the client.

Another example is copying verity metadata when creating filesystem
images in userspace (such as with ``mkfs.ext4 -d``).

This is a fairly specialized use case, and most fs-verity users won't
need this ioctl.

This ioctl takes in a pointer to the following structure::

   #define FS_VERITY_METADATA_TYPE_MERKLE_TREE     1
   #define FS_VERITY_METADATA_TYPE_DESCRIPTOR      2
   #define FS_VERITY_METADATA_TYPE_SIGNATURE       3

   struct fsverity_read_metadata_arg {
           __u64 metadata_type;
           __u64 offset;
           __u64 length;
           __u64 buf_ptr;
           __u64 __reserved;
   };

``metadata_type`` specifies the type of metadata to read:

- ``FS_VERITY_METADATA_TYPE_MERKLE_TREE`` reads the blocks of the
  Merkle tree.  The blocks are returned in order from the root level
  to the leaf level.  Within each level, the blocks are returned in
  the same order that their hashes are themselves hashed.
  See `Merkle tree`_ for more information.

- ``FS_VERITY_METADATA_TYPE_DESCRIPTOR`` reads the fs-verity
  descriptor.  See `fs-verity descriptor`_.

- ``FS_VERITY_METADATA_TYPE_SIGNATURE`` reads the builtin signature
  which was passed to FS_IOC_ENABLE_VERITY, if any.  See `Built-in
  signature verification`_.

The semantics are similar to those of ``pread()``.  ``offset``
specifies the offset in bytes into the metadata item to read from, and
``length`` specifies the maximum number of bytes to read from the
metadata item.  ``buf_ptr`` is the pointer to the buffer to read into,
cast to a 64-bit integer.  ``__reserved`` must be 0.  On success, the
number of bytes read is returned.  0 is returned at the end of the
metadata item.  The returned length may be less than ``length``, for
example if the ioctl is interrupted.

The metadata returned by FS_IOC_READ_VERITY_METADATA isn't guaranteed
to be authenticated against the file digest that would be returned by
`FS_IOC_MEASURE_VERITY`_, as the metadata is expected to be used to
implement fs-verity compatible verification anyway (though absent a
malicious disk, the metadata will indeed match).  E.g. to implement
this ioctl, the filesystem is allowed to just read the Merkle tree
blocks from disk without actually verifying the path to the root node.

FS_IOC_READ_VERITY_METADATA can fail with the following errors:

- ``EFAULT``: the caller provided inaccessible memory
- ``EINTR``: the ioctl was interrupted before any data was read
- ``EINVAL``: reserved fields were set, or ``offset + length``
  overflowed
- ``ENODATA``: the file is not a verity file, or
  FS_VERITY_METADATA_TYPE_SIGNATURE was requested but the file doesn't
  have a builtin signature
- ``ENOTTY``: this type of filesystem does not implement fs-verity, or
  this ioctl is not yet implemented on it
- ``EOPNOTSUPP``: the kernel was not configured with fs-verity
  support, or the filesystem superblock has not had the 'verity'
  feature enabled on it.  (See `Filesystem support`_.)

FS_IOC_GETFLAGS
---------------

The existing ioctl FS_IOC_GETFLAGS (which isn't specific to fs-verity)
can also be used to check whether a file has fs-verity enabled or not.
To do so, check for FS_VERITY_FL (0x00100000) in the returned flags.

The verity flag is not settable via FS_IOC_SETFLAGS.  You must use
FS_IOC_ENABLE_VERITY instead, since parameters must be provided.

statx
-----

Since Linux v5.5, the statx() system call sets STATX_ATTR_VERITY if
the file has fs-verity enabled.  This can perform better than
FS_IOC_GETFLAGS and FS_IOC_MEASURE_VERITY because it doesn't require
opening the file, and opening verity files can be expensive.

.. _accessing_verity_files:

접근 의미, Merkle tree와 descriptor

345-451

애플리케이션은 몇 가지 예외를 제외하면 verity file을 일반 파일처럼 투명하게 읽습니다. 파일 내용은 읽기 전용이므로 mode bit가 허용해도 write open이나 `truncate()`는 `EPERM`입니다. owner, mode, timestamp, xattr 같은 metadata는 fs-verity 측정 대상이 아니므로 변경할 수 있고 rename, delete, link도 가능합니다.

direct I/O는 지원하지 않아 buffered I/O로 폴백합니다. DAX는 data verification을 우회하므로 지원하지 않습니다. Merkle tree와 맞지 않는 데이터는 `read()`에서 `EIO`, `mmap()` 읽기에서 `SIGBUS`가 됩니다.

`fs.verity.require_signatures=1`이고 `.fs-verity` keyring의 키로 서명되지 않았다면 file open이 실패합니다. Merkle tree에 직접 접근할 수 없으므로 파일을 일반 복사하거나 backup·restore하면 verity 속성을 잃습니다. package manager가 관리하는 executable 같은 파일이 주된 대상입니다.

지원 파일시스템 모두 같은 알고리즘으로 file contents를 Merkle tree로 hash해 내용을 암호학적으로 식별하는 digest를 만듭니다. 사용자 공간이 직접 digest를 계산해 서명해야 할 때만 이 알고리즘을 구현할 필요가 있습니다.

파일은 보통 4096바이트인 구성 가능한 block으로 나뉘고 마지막 block은 필요하면 0으로 채웁니다. 각 data block hash가 첫 level이 되고, hash들을 다시 block_size 단위 block으로 묶어 끝을 0으로 채운 뒤 hash해 다음 level을 만듭니다. block 하나만 남을 때까지 반복하고 그 block의 hash가 Merkle tree root hash입니다.

0이 아닌 파일이 한 block에 맞으면 root hash는 그 단일 data block hash입니다. 빈 파일의 root hash는 모두 0입니다. 여기서 block은 filesystem block과 반드시 같지 않습니다.

salt가 있으면 hash compression function input size의 가장 가까운 배수로 0 padding합니다. SHA-256은 64바이트, SHA-512는 128바이트 단위입니다. padded salt를 hash하는 모든 data block과 tree block 앞에 붙입니다.

block padding은 모든 hash가 같은 양의 data를 처리하게 해 구현을 단순화하고 hardware acceleration 가능성을 넓힙니다. salt padding은 salted hash state를 미리 계산해 각 hash에 import할 때 salting 비용을 사실상 없앱니다.

권장 SHA-256과 4K block에서는 block마다 128개 hash가 들어갑니다. 각 level은 아래 level의 약 1/128이고 큰 파일에서 tree 크기는 원본의 약 1/127에 수렴합니다. 작은 파일은 padding 비중이 커서 상대 overhead가 큽니다.

root hash만으로는 큰 파일과 그 큰 파일의 top-level hash block을 data로 가진 작은 파일을 구분하지 못하고 block boundary padding에서도 모호성이 생깁니다. 그래서 실제 file digest는 root hash와 file size 등 문맥을 담은 `struct fsverity_descriptor` 전체의 hash입니다.

struct fsverity_descriptor {
        __u8 version;           /* must be 1 */
        __u8 hash_algorithm;
        __u8 log_blocksize;
        __u8 salt_size;
        __le32 __reserved_0x04; /* must be 0 */
        __le64 data_size;
        __u8 root_hash[64];
        __u8 salt[32];
        __u8 __reserved[144];   /* must be 0's */
};
Merkle tree 구성
파일을 block_size 단위로 나누고 마지막 block 0 padding각 data block 앞에 padded salt를 붙여 hashhash 128개씩 4K tree block에 묶고 0 padding상위 level을 반복해 단일 top-level block 생성top-level block hash를 root hash로 계산root hash·data_size·algorithm·salt를 descriptor에 넣고 전체 hash

data block에서 상수 크기 file digest까지 올라가는 구조입니다.

Verity file 접근 차이
연산동작
write open·`truncate()``EPERM`
metadata 변경·rename·delete·link허용, 측정 대상 아님
direct I/Obuffered I/O로 폴백
DAX미지원
손상 `read()``EIO`
손상 `mmap()``SIGBUS`
일반 copy·restoreverity 속성 소실

일반 파일과 달라지는 연산과 실패 형태입니다.


Accessing verity files
======================

Applications can transparently access a verity file just like a
non-verity one, with the following exceptions:

- Verity files are readonly.  They cannot be opened for writing or
  truncate()d, even if the file mode bits allow it.  Attempts to do
  one of these things will fail with EPERM.  However, changes to
  metadata such as owner, mode, timestamps, and xattrs are still
  allowed, since these are not measured by fs-verity.  Verity files
  can also still be renamed, deleted, and linked to.

- Direct I/O is not supported on verity files.  Attempts to use direct
  I/O on such files will fall back to buffered I/O.

- DAX (Direct Access) is not supported on verity files, because this
  would circumvent the data verification.

- Reads of data that doesn't match the verity Merkle tree will fail
  with EIO (for read()) or SIGBUS (for mmap() reads).

- If the sysctl "fs.verity.require_signatures" is set to 1 and the
  file is not signed by a key in the ".fs-verity" keyring, then
  opening the file will fail.  See `Built-in signature verification`_.

Direct access to the Merkle tree is not supported.  Therefore, if a
verity file is copied, or is backed up and restored, then it will lose
its "verity"-ness.  fs-verity is primarily meant for files like
executables that are managed by a package manager.

File digest computation
=======================

This section describes how fs-verity hashes the file contents using a
Merkle tree to produce the digest which cryptographically identifies
the file contents.  This algorithm is the same for all filesystems
that support fs-verity.

Userspace only needs to be aware of this algorithm if it needs to
compute fs-verity file digests itself, e.g. in order to sign files.

.. _fsverity_merkle_tree:

Merkle tree
-----------

The file contents is divided into blocks, where the block size is
configurable but is usually 4096 bytes.  The end of the last block is
zero-padded if needed.  Each block is then hashed, producing the first
level of hashes.  Then, the hashes in this first level are grouped
into 'blocksize'-byte blocks (zero-padding the ends as needed) and
these blocks are hashed, producing the second level of hashes.  This
proceeds up the tree until only a single block remains.  The hash of
this block is the "Merkle tree root hash".

If the file fits in one block and is nonempty, then the "Merkle tree
root hash" is simply the hash of the single data block.  If the file
is empty, then the "Merkle tree root hash" is all zeroes.

The "blocks" here are not necessarily the same as "filesystem blocks".

If a salt was specified, then it's zero-padded to the closest multiple
of the input size of the hash algorithm's compression function, e.g.
64 bytes for SHA-256 or 128 bytes for SHA-512.  The padded salt is
prepended to every data or Merkle tree block that is hashed.

The purpose of the block padding is to cause every hash to be taken
over the same amount of data, which simplifies the implementation and
keeps open more possibilities for hardware acceleration.  The purpose
of the salt padding is to make the salting "free" when the salted hash
state is precomputed, then imported for each hash.

Example: in the recommended configuration of SHA-256 and 4K blocks,
128 hash values fit in each block.  Thus, each level of the Merkle
tree is approximately 128 times smaller than the previous, and for
large files the Merkle tree's size converges to approximately 1/127 of
the original file size.  However, for small files, the padding is
significant, making the space overhead proportionally more.

.. _fsverity_descriptor:

fs-verity descriptor
--------------------

By itself, the Merkle tree root hash is ambiguous.  For example, it
can't a distinguish a large file from a small second file whose data
is exactly the top-level hash block of the first file.  Ambiguities
also arise from the convention of padding to the next block boundary.

To solve this problem, the fs-verity file digest is actually computed
as a hash of the following structure, which contains the Merkle tree
root hash as well as other fields such as the file size::

    struct fsverity_descriptor {
            __u8 version;           /* must be 1 */
            __u8 hash_algorithm;    /* Merkle tree hash algorithm */
            __u8 log_blocksize;     /* log2 of size of data and tree blocks */
            __u8 salt_size;         /* size of salt in bytes; 0 if none */
            __le32 __reserved_0x04; /* must be 0 */
            __le64 data_size;       /* size of file the Merkle tree is built over */
            __u8 root_hash[64];     /* Merkle tree root hash */
            __u8 salt[32];          /* salt prepended to each hashed block */
            __u8 __reserved[144];   /* must be 0's */
    };

내장 서명 검증의 동작과 한계

452-568

`CONFIG_FS_VERITY_BUILTIN_SIGNATURES=y`는 fs-verity builtin signature의 커널 내부 검증을 추가합니다. 그러나 유일한 서명 방법이 아니며 사용자 공간 검증이나 IMA appraisal이 훨씬 나을 수 있습니다. 실제보다 더 많은 문제를 해결한다고 오해하기 쉬우므로 사용 전 각 한계를 검토해야 합니다.

부팅 때 커널은 `.fs-verity` keyring을 만들고 root는 `add_key()`로 신뢰할 X.509 certificate를 추가할 수 있습니다.

`FS_IOC_ENABLE_VERITY`는 file digest에 대한 DER 형식 PKCS#7 detached signature를 받을 수 있습니다. 성공하면 Merkle tree와 함께 저장합니다. 그 뒤 파일을 열 때마다 `.fs-verity` certificate로 실제 digest와 signature를 검증합니다. signature가 존재하는 동안은 `fs.verity.require_signatures` 값과 무관하게 검증하며 IPE LSM이 검증된 builtin signature 파일을 인식·표시하는 데 이 동작을 사용합니다.

새 sysctl `fs.verity.require_signatures`를 1로 설정하면 모든 verity file이 올바르게 서명된 digest를 가져야 합니다.

서명 대상은 단순 digest byte가 아니라 magic `FSVerity`, little-endian algorithm과 size, digest를 연결한 `struct fsverity_formatted_digest`입니다.

struct fsverity_formatted_digest {
        char magic[8];          /* must be "FSVerity" */
        __le16 digest_algorithm;
        __le16 digest_size;
        __u8 digest[];
};

내장 검증은 어떤 파일이 fs-verity를 실제로 사용하도록 강제하지 않으므로 완전한 인증 정책이 아닙니다. 신뢰 사용자 공간이 접근 전에 서명과 활성화를 명시적으로 확인해야 합니다. `require_signatures=1`이면 활성화 여부만 확인해도 되지만, 그 정도 사용자 공간 신뢰가 있다면 signature를 파일 옆에 저장하고 암호 라이브러리로 직접 검증할 수도 있습니다.

대안으로 IPE LSM과 결합해 검증된 builtin signature 파일만 실행 같은 특정 연산을 허용하는 커널 강제 시스템 전체 정책을 만들 수 있습니다. IPE는 `fs.verity.require_signatures=1`을 요구하지 않습니다.

builtin signature는 fs-verity 활성화와 동시에만 설정할 수 있습니다. 나중에 변경하거나 삭제하려면 파일을 다시 만들어야 합니다.

시스템의 모든 fs-verity 파일이 같은 공개키 집합을 공유합니다. 파일마다 다른 신뢰 키 범위를 지정할 수 없고 각 키는 전체 파일에 대해 all-or-nothing입니다. `fs.verity.require_signatures`도 시스템 전체에 적용되어 모든 사용자가 1 설정에 동의해야 하므로 유용한 일부 배포를 막을 수 있습니다.

커널이 지원하는 signature algorithm만 쓸 수 있고 새 암호 설계에 흔히 권장되는 Ed25519는 아직 커널이 지원하지 않습니다.

builtin signature는 PKCS#7, 공개키는 X.509입니다. 널리 쓰이고 기능이 많지만 1990년대 ASN.1 기반 형식의 복잡한 parser·handler는 역사적으로 취약점의 원인이 됐으며 이는 암호 자체에 필연적인 복잡성이 아닙니다.

X.509·PKCS#7 고급 기능이 필요 없다면 단순 Ed25519 key와 signature를 사용자 공간에서 검증하는 방식을 강하게 고려해야 합니다. 두 형식을 계속 쓰더라도 사용자 공간 검증은 더 유연하고 `CONFIG_FS_VERITY_BUILTIN_SIGNATURES`가 늘리는 kernel attack surface를 피합니다.

고급 X.509·PKCS#7 기능은 커널에서 항상 의도대로 작동하지 않습니다. 예를 들어 커널은 X.509 certificate validity time을 검사하지 않습니다. fs-verity를 지원하는 IMA appraisal은 signature에 PKCS#7을 쓰지 않아 일부 문제를 피하지만 X.509는 사용합니다.

내장 서명 검증 경로
부팅 시 `.fs-verity` keyring 생성root가 X.509 certificate 추가사용자 공간이 formatted digest에 DER PKCS#7 detached signature 생성ENABLE_VERITY가 signature와 Merkle tree 저장매 file open에서 실제 digest와 signature 검증선택적으로 require_signatures 또는 IPE가 정책 강제

keyring과 파일 open 사이의 검증 관계입니다.

내장 서명의 주요 제약
제약영향
활성화 자체를 강제하지 않음별도 trusted userspace 또는 IPE 정책 필요
signature 사후 변경 불가파일 재생성 필요
공개키 집합 시스템 전체 공유파일별 trust domain 분리 불가
sysctl 시스템 전체서로 다른 사용자 요구 충돌
커널 algorithm 한정Ed25519 미지원
PKCS#7·X.509 복잡성parser attack surface와 기능 제약

사용자 공간 검증과 비교할 때 먼저 확인할 항목입니다.

Built-in signature verification
===============================

CONFIG_FS_VERITY_BUILTIN_SIGNATURES=y adds supports for in-kernel
verification of fs-verity builtin signatures.

**IMPORTANT**!  Please take great care before using this feature.
It is not the only way to do signatures with fs-verity, and the
alternatives (such as userspace signature verification, and IMA
appraisal) can be much better.  It's also easy to fall into a trap
of thinking this feature solves more problems than it actually does.

Enabling this option adds the following:

1. At boot time, the kernel creates a keyring named ".fs-verity".  The
   root user can add trusted X.509 certificates to this keyring using
   the add_key() system call.

2. `FS_IOC_ENABLE_VERITY`_ accepts a pointer to a PKCS#7 formatted
   detached signature in DER format of the file's fs-verity digest.
   On success, the ioctl persists the signature alongside the Merkle
   tree.  Then, any time the file is opened, the kernel verifies the
   file's actual digest against this signature, using the certificates
   in the ".fs-verity" keyring. This verification happens as long as the
   file's signature exists, regardless of the state of the sysctl variable
   "fs.verity.require_signatures" described in the next item. The IPE LSM
   relies on this behavior to recognize and label fsverity files
   that contain a verified built-in fsverity signature.

3. A new sysctl "fs.verity.require_signatures" is made available.
   When set to 1, the kernel requires that all verity files have a
   correctly signed digest as described in (2).

The data that the signature as described in (2) must be a signature of
is the fs-verity file digest in the following format::

    struct fsverity_formatted_digest {
            char magic[8];                  /* must be "FSVerity" */
            __le16 digest_algorithm;
            __le16 digest_size;
            __u8 digest[];
    };

That's it.  It should be emphasized again that fs-verity builtin
signatures are not the only way to do signatures with fs-verity.  See
`Use cases`_ for an overview of ways in which fs-verity can be used.
fs-verity builtin signatures have some major limitations that should
be carefully considered before using them:

- Builtin signature verification does *not* make the kernel enforce
  that any files actually have fs-verity enabled.  Thus, it is not a
  complete authentication policy.  Currently, if it is used, one
  way to complete the authentication policy is for trusted userspace
  code to explicitly check whether files have fs-verity enabled with a
  signature before they are accessed.  (With
  fs.verity.require_signatures=1, just checking whether fs-verity is
  enabled suffices.)  But, in this case the trusted userspace code
  could just store the signature alongside the file and verify it
  itself using a cryptographic library, instead of using this feature.

- Another approach is to utilize fs-verity builtin signature
  verification in conjunction with the IPE LSM, which supports defining
  a kernel-enforced, system-wide authentication policy that allows only
  files with a verified fs-verity builtin signature to perform certain
  operations, such as execution. Note that IPE doesn't require
  fs.verity.require_signatures=1.
  Please refer to :doc:`IPE admin guide </admin-guide/LSM/ipe>` for
  more details.

- A file's builtin signature can only be set at the same time that
  fs-verity is being enabled on the file.  Changing or deleting the
  builtin signature later requires re-creating the file.

- Builtin signature verification uses the same set of public keys for
  all fs-verity enabled files on the system.  Different keys cannot be
  trusted for different files; each key is all or nothing.

- The sysctl fs.verity.require_signatures applies system-wide.
  Setting it to 1 only works when all users of fs-verity on the system
  agree that it should be set to 1.  This limitation can prevent
  fs-verity from being used in cases where it would be helpful.

- Builtin signature verification can only use signature algorithms
  that are supported by the kernel.  For example, the kernel does not
  yet support Ed25519, even though this is often the signature
  algorithm that is recommended for new cryptographic designs.

- fs-verity builtin signatures are in PKCS#7 format, and the public
  keys are in X.509 format.  These formats are commonly used,
  including by some other kernel features (which is why the fs-verity
  builtin signatures use them), and are very feature rich.
  Unfortunately, history has shown that code that parses and handles
  these formats (which are from the 1990s and are based on ASN.1)
  often has vulnerabilities as a result of their complexity.  This
  complexity is not inherent to the cryptography itself.

  fs-verity users who do not need advanced features of X.509 and
  PKCS#7 should strongly consider using simpler formats, such as plain
  Ed25519 keys and signatures, and verifying signatures in userspace.

  fs-verity users who choose to use X.509 and PKCS#7 anyway should
  still consider that verifying those signatures in userspace is more
  flexible (for other reasons mentioned earlier in this document) and
  eliminates the need to enable CONFIG_FS_VERITY_BUILTIN_SIGNATURES
  and its associated increase in kernel attack surface.  In some cases
  it can even be necessary, since advanced X.509 and PKCS#7 features
  do not always work as intended with the kernel.  For example, the
  kernel does not check X.509 certificate validity times.

  Note: IMA appraisal, which supports fs-verity, does not use PKCS#7
  for its signatures, so it partially avoids the issues discussed
  here.  IMA appraisal does use X.509.

Filesystem support
==================

fs-verity is supported by several filesystems, described below.  The

ext4, f2fs, btrfs 지원과 metadata 저장

569-647

모든 지원 파일시스템에서 `CONFIG_FS_VERITY`가 필요합니다. `include/linux/fsverity.h`는 `fs/verity/` 계층과 파일시스템 사이 인터페이스를 선언합니다. 파일시스템은 Merkle tree block과 `fsverity_descriptor`를 파일시스템별 위치에서 읽고 쓰는 `fsverity_operations`를 제공하고, file open이나 pagecache read 완료 같은 시점에 `fs/verity/` 함수를 호출해야 합니다.

ext4는 Linux 5.4와 e2fsprogs 1.45.2부터 지원합니다. 포맷할 때 `-O verity`를 주거나 `tune2fs -O verity`를 실행해야 합니다. verity는 RO_COMPAT feature이므로 설정 뒤 구형 커널은 read-only로만 마운트하고 구형 e2fsck는 검사할 수 없습니다.

초기 ext4 구현은 filesystem block size가 보통 4096바이트인 system page size와 같아야 했지만 Linux 6.3에서 제한이 제거됐습니다. verity inode의 on-disk `EXT4_VERITY_FL`은 ENABLE ioctl만 설정할 수 있고 지울 수 없습니다.

ext4 암호화와 fs-verity를 동시에 쓸 때 암호문이 아니라 평문을 검증합니다. 파일마다 암호문이 달라져도 file digest가 의미 있게 유지되기 위한 조건입니다.

ext4는 `i_size`를 넘는 첫 64K boundary부터 파일 끝 뒤에 Merkle tree와 descriptor를 저장합니다. verity file이 읽기 전용이고 `i_size` 완전히 뒤의 page는 사용자 공간에 보이지 않으면서 ext4 내부에서는 작은 변경으로 읽고 쓸 수 있어 가능한 설계입니다.

이 방식은 EA_INODE에 의존하거나 multi-gigabyte xattr를 paging하고 암호화하도록 ext4 xattr를 재설계하는 일을 피합니다. 파일이 암호화됐다면 평문 hash를 담는 verity metadata도 반드시 암호화해야 합니다. ext4는 extent 기반 파일에만 verity를 허용합니다.

f2fs는 Linux 5.4와 f2fs-tools 1.11.0부터 지원하며 포맷할 때 `-O verity`가 필요합니다. `FADVISE_VERITY_BIT`는 ENABLE ioctl만 설정하고 지울 수 없습니다.

f2fs도 `i_size` 뒤 첫 64K boundary부터 metadata를 저장합니다. inode당 xattr entry가 최대 4096바이트라 보통 Merkle tree block 하나도 담기 어려워 이 방식이 필요합니다. atomic write나 volatile write가 pending인 파일에는 verity를 활성화할 수 없습니다.

btrfs는 Linux 5.15부터 지원합니다. verity inode에는 RO_COMPAT inode flag를 표시하고 metadata는 별도 btree item에 저장합니다.

파일시스템별 fs-verity
파일시스템지원 시작준비·flagmetadata 위치
ext4Linux 5.4 / e2fsprogs 1.45.2`-O verity`, `EXT4_VERITY_FL``i_size` 뒤 첫 64K boundary
f2fsLinux 5.4 / f2fs-tools 1.11.0`-O verity`, `FADVISE_VERITY_BIT``i_size` 뒤 첫 64K boundary
btrfsLinux 5.15RO_COMPAT inode flag별도 btree items

도입 버전과 metadata 저장 방식을 비교합니다.

CONFIG_FS_VERITY kconfig option must be enabled to use fs-verity on
any of these filesystems.

``include/linux/fsverity.h`` declares the interface between the
``fs/verity/`` support layer and filesystems.  Briefly, filesystems
must provide an ``fsverity_operations`` structure that provides
methods to read and write the verity metadata to a filesystem-specific
location, including the Merkle tree blocks and
``fsverity_descriptor``.  Filesystems must also call functions in
``fs/verity/`` at certain times, such as when a file is opened or when
pages have been read into the pagecache.  (See `Verifying data`_.)

ext4
----

ext4 supports fs-verity since Linux v5.4 and e2fsprogs v1.45.2.

To create verity files on an ext4 filesystem, the filesystem must have
been formatted with ``-O verity`` or had ``tune2fs -O verity`` run on
it.  "verity" is an RO_COMPAT filesystem feature, so once set, old
kernels will only be able to mount the filesystem readonly, and old
versions of e2fsck will be unable to check the filesystem.

Originally, an ext4 filesystem with the "verity" feature could only be
mounted when its block size was equal to the system page size
(typically 4096 bytes).  In Linux v6.3, this limitation was removed.

ext4 sets the EXT4_VERITY_FL on-disk inode flag on verity files.  It
can only be set by `FS_IOC_ENABLE_VERITY`_, and it cannot be cleared.

ext4 also supports encryption, which can be used simultaneously with
fs-verity.  In this case, the plaintext data is verified rather than
the ciphertext.  This is necessary in order to make the fs-verity file
digest meaningful, since every file is encrypted differently.

ext4 stores the verity metadata (Merkle tree and fsverity_descriptor)
past the end of the file, starting at the first 64K boundary beyond
i_size.  This approach works because (a) verity files are readonly,
and (b) pages fully beyond i_size aren't visible to userspace but can
be read/written internally by ext4 with only some relatively small
changes to ext4.  This approach avoids having to depend on the
EA_INODE feature and on rearchitecturing ext4's xattr support to
support paging multi-gigabyte xattrs into memory, and to support
encrypting xattrs.  Note that the verity metadata *must* be encrypted
when the file is, since it contains hashes of the plaintext data.

ext4 only allows verity on extent-based files.

f2fs
----

f2fs supports fs-verity since Linux v5.4 and f2fs-tools v1.11.0.

To create verity files on an f2fs filesystem, the filesystem must have
been formatted with ``-O verity``.

f2fs sets the FADVISE_VERITY_BIT on-disk inode flag on verity files.
It can only be set by `FS_IOC_ENABLE_VERITY`_, and it cannot be
cleared.

Like ext4, f2fs stores the verity metadata (Merkle tree and
fsverity_descriptor) past the end of the file, starting at the first
64K boundary beyond i_size.  See explanation for ext4 above.
Moreover, f2fs supports at most 4096 bytes of xattr entries per inode
which usually wouldn't be enough for even a single Merkle tree block.

f2fs doesn't support enabling verity on files that currently have
atomic or volatile writes pending.

btrfs
-----

btrfs supports fs-verity since Linux v5.15.  Verity-enabled inodes are
marked with a RO_COMPAT inode flag, and the verity metadata is stored
in separate btree items.

Implementation details
======================

Pagecache·bio 검증 구현과 사용자 공간 도구

648-742

fs-verity는 `mmap()`, `read()`, `pread()` 등 syscall 종류와 최초·후속 읽기를 가리지 않고 verity file의 모든 data read를 검증합니다. 이미 검증되어 cache된 data를 그대로 반환하는 경우만 재검증이 필요하지 않습니다.

Linux pagecache를 쓰는 파일시스템은 `->read_folio()`와 `->readahead()`가 folio를 Uptodate로 표시하기 전에 검증하도록 수정해야 합니다. memory map은 `->read_iter()`를 쓰지 않으므로 여기에만 hook하는 것은 불충분합니다.

`fsverity_verify_blocks()`는 verity inode의 pagecache로 읽은 data를 검증합니다. 호출할 때 folio는 여전히 lock된 상태이고 Uptodate가 아니어야 하므로 사용자 공간이 읽을 수 없습니다. 필요한 tree hash block은 `fsverity_operations::read_merkle_tree_page()` callback으로 파일시스템에서 읽습니다.

검증 실패 시 `fsverity_verify_blocks()`는 false를 반환하고 파일시스템은 folio를 Uptodate로 표시하면 안 됩니다. 이후 일반 pagecache 동작에 따라 해당 부분의 `read()`는 `EIO`, memory map 접근은 `SIGBUS`가 됩니다.

원칙적으로 data block 하나를 검증하려면 그 block에서 root hash까지 전체 경로를 확인해야 합니다. 효율을 위해 이미 검증된 hash block을 cache할 수 있으므로 함수는 tree를 올라가다 검증된 block을 만나면 그 block까지만 경로를 확인합니다.

dm-verity도 쓰는 이 최적화는 순차 읽기 성능이 매우 좋습니다. 4K block과 SHA-256에서는 보통 128번 중 127번 아래 level hash block이 직전 data block 읽기에서 이미 cache·검증되어 있습니다. 반면 random read는 더 느립니다.

ext4와 f2fs 같은 block device 기반 파일시스템도 pagecache 규칙을 적용하지만 여러 data block을 `bio` 하나로 묶어 읽습니다. `fsverity_verify_bio()`는 bio 안 모든 data block을 검증하도록 제공됩니다.

verity file이 암호화도 됐다면 검증 전에 복호화해야 합니다. ext4와 f2fs는 각 bio에 `bio_post_read_ctx`를 할당해 `->bi_private`에 저장합니다. `enabled_steps` bitmask는 decryption, verity 또는 둘 다 필요한지 표시합니다.

struct bio_post_read_ctx {
        struct bio *bio;
        struct work_struct work;
        unsigned int cur_step;
        unsigned int enabled_steps;
};

bio 완료 뒤 필요한 각 postprocessing step에 대해 context를 workqueue에 넣고 work가 복호화 또는 검증을 수행합니다. 오류가 없는 folio만 Uptodate로 표시한 뒤 unlock합니다.

파일 hole은 일반 `->readahead()`가 bio 없이 0으로 채우고 Uptodate로 처리할 수 있습니다. 이 경로가 검증을 우회하지 않도록 hole block도 `fsverity_verify_blocks()`로 검증합니다. direct I/O 역시 우회 가능성이 있어 verity file에서 비활성화합니다.

커널 외 사용자 공간 도구는 `https://git.kernel.org/pub/scm/fs/fsverity/fsverity-utils.git`에 있습니다. 보호 파일 구성 예시는 그 source tree의 `README.md`를 참고합니다.

암호화된 verity bio 후처리
storage에서 여러 암호문 block을 bio로 읽기`bio_post_read_ctx`가 enabled_steps 확인workqueue에서 먼저 복호화평문 data를 fsverity_verify_bio로 검증hole block도 별도 verify_blocks로 검증오류 없는 folio만 Uptodate 후 unlock

평문 검증 순서를 보장하는 작업 단계입니다.

검증 hook 선택
경로검증 함수·지점핵심 조건
pagecache`->read_folio()`, `->readahead()` + `fsverity_verify_blocks()`locked, not Uptodate
bio`fsverity_verify_bio()`모든 block, 복호화 뒤
hole`fsverity_verify_blocks()`zero-fill만으로 Uptodate 금지
direct I/O비활성화pagecache 검증 우회 방지

I/O 경로별 검증 지점과 우회 방지 조건입니다.

Verifying data
--------------

fs-verity ensures that all reads of a verity file's data are verified,
regardless of which syscall is used to do the read (e.g. mmap(),
read(), pread()) and regardless of whether it's the first read or a
later read (unless the later read can return cached data that was
already verified).  Below, we describe how filesystems implement this.

Pagecache
~~~~~~~~~

For filesystems using Linux's pagecache, the ``->read_folio()`` and
``->readahead()`` methods must be modified to verify folios before
they are marked Uptodate.  Merely hooking ``->read_iter()`` would be
insufficient, since ``->read_iter()`` is not used for memory maps.

Therefore, fs/verity/ provides the function fsverity_verify_blocks()
which verifies data that has been read into the pagecache of a verity
inode.  The containing folio must still be locked and not Uptodate, so
it's not yet readable by userspace.  As needed to do the verification,
fsverity_verify_blocks() will call back into the filesystem to read
hash blocks via fsverity_operations::read_merkle_tree_page().

fsverity_verify_blocks() returns false if verification failed; in this
case, the filesystem must not set the folio Uptodate.  Following this,
as per the usual Linux pagecache behavior, attempts by userspace to
read() from the part of the file containing the folio will fail with
EIO, and accesses to the folio within a memory map will raise SIGBUS.

In principle, verifying a data block requires verifying the entire
path in the Merkle tree from the data block to the root hash.
However, for efficiency the filesystem may cache the hash blocks.
Therefore, fsverity_verify_blocks() only ascends the tree reading hash
blocks until an already-verified hash block is seen.  It then verifies
the path to that block.

This optimization, which is also used by dm-verity, results in
excellent sequential read performance.  This is because usually (e.g.
127 in 128 times for 4K blocks and SHA-256) the hash block from the
bottom level of the tree will already be cached and checked from
reading a previous data block.  However, random reads perform worse.

Block device based filesystems
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Block device based filesystems (e.g. ext4 and f2fs) in Linux also use
the pagecache, so the above subsection applies too.  However, they
also usually read many data blocks from a file at once, grouped into a
structure called a "bio".  To make it easier for these types of
filesystems to support fs-verity, fs/verity/ also provides a function
fsverity_verify_bio() which verifies all data blocks in a bio.

ext4 and f2fs also support encryption.  If a verity file is also
encrypted, the data must be decrypted before being verified.  To
support this, these filesystems allocate a "post-read context" for
each bio and store it in ``->bi_private``::

    struct bio_post_read_ctx {
           struct bio *bio;
           struct work_struct work;
           unsigned int cur_step;
           unsigned int enabled_steps;
    };

``enabled_steps`` is a bitmask that specifies whether decryption,
verity, or both is enabled.  After the bio completes, for each needed
postprocessing step the filesystem enqueues the bio_post_read_ctx on a
workqueue, and then the workqueue work does the decryption or
verification.  Finally, folios where no decryption or verity error
occurred are marked Uptodate, and the folios are unlocked.

On many filesystems, files can contain holes.  Normally,
``->readahead()`` simply zeroes hole blocks and considers the
corresponding data to be up-to-date; no bios are issued.  To prevent
this case from bypassing fs-verity, filesystems use
fsverity_verify_blocks() to verify hole blocks.

Filesystems also disable direct I/O on verity files, since otherwise
direct I/O would bypass fs-verity.

Userspace utility
=================

This document focuses on the kernel, but a userspace utility for
fs-verity can be found at:

        https://git.kernel.org/pub/scm/fs/fsverity/fsverity-utils.git

See the README.md file in the fsverity-utils source tree for details,
including examples of setting up fs-verity protected files.

Tests
=====

xfstests와 설계 FAQ

743-912

fs-verity는 xfstests의 `verity` group으로 검증합니다. kvm-xfstests 예시는 다음과 같으며 ext4, f2fs, btrfs에서 실패가 없어야 합니다.

kvm-xfstests -c ext4,f2fs,btrfs -g verity

fs-verity와 IMA의 초점은 다릅니다. fs-verity는 Merkle tree로 개별 파일을 hash하는 filesystem mechanism이고, IMA는 어떤 파일을 hash하고 결과를 기록·인증·measurement list에 추가할지 정하는 system-wide policy입니다. IMA는 성능과 보안 이점을 위해 fs-verity hash를 전체 file hash 대신 사용할 수 있지만 모든 fs-verity 용도를 IMA로 강제할 이유는 없습니다. 독립 기능으로도 많은 요구를 충족하고 xfstests로 검사할 수 있습니다.

공격자가 disk의 Merkle tree hash를 바꿀 수 있다는 이유로 무용하지 않습니다. 진본성을 확인하려면 tree root hash가 들어간 fs-verity file digest 자체를 인증해야 합니다. verity file을 비verity file로 바꾸는 공격도 신뢰 사용자 공간이 활성화와 digest를 인증하고 비verity 파일을 신뢰하지 않으면 차단됩니다.

root hash만 disk에 저장하고 tree를 저장하지 않으면 한 byte만 처음 읽어도 전체 tree를 계산해야 합니다. leaf를 검증하려면 root node를 포함한 root까지 전체 경로가 필요한데 root node가 없으면 children을 hash하고 다시 그 children을 계산하는 식으로 결국 전체 파일을 hash하게 됩니다. 이 경우 단순 `sha256(file)`이 더 쉽고 조금 더 빠릅니다.

메모리 tree는 매 읽기 검증 이점은 남지만 큰 tree 전체를 pin할 수 없습니다. hash page가 퇴거될 때마다 그 아래 tree 전체를 다시 hash해야 해 block 하나 읽기가 수 GiB 재hash를 유발할 수 있으므로 Merkle tree 목적을 대부분 잃습니다.

leaf node만 저장하는 것도 한 level 위로 문제를 옮길 뿐입니다. data block을 leaf로 볼 수도 있습니다. SHA-256·4K 설정에서 각 level은 아래보다 1% 미만이므로 leaf level만 저장해도 tree의 99% 이상을 저장합니다. 전체 tree를 저장하는 편이 낫습니다.

package에 미리 만든 Merkle tree를 배포하는 기능은 현재 지원하지 않습니다. 초기 설계에는 있었지만 kernel UAPI 단순화와 낮은 우선순위 때문에 제거됐습니다. 파일은 보통 한 번 설치해 여러 번 쓰고 현대 CPU의 암호 hash는 비교적 빠릅니다.

write 지원은 완전히 다른 설계가 필요합니다. crash 뒤 data와 모든 hash level의 일관성을 유지해야 하므로 data journaling, copy-on-write, log-structured volume 같은 mechanism이 필요하지만 기존 파일시스템에 새 일관성 체계를 덧붙이기 어렵고 ext4 data journaling은 매우 느립니다.

매 write마다 Merkle tree를 다시 만들면 극도로 비효율적이고 authenticated skiplist 같은 다른 authenticated dictionary는 훨씬 복잡합니다. dm-verity는 read-only data와 read-only tree를 단순 검증하지만 dm-integrity는 write를 지원하는 대신 느리고 복잡하며 sector를 독립 인증해 root hash 기반 full-device authentication도 제공하지 않습니다. 서로 다른 문제를 한 target이나 fs-verity 하나에 합칠 이유가 없습니다.

verity file이 변경 불가능해도 `FS_IMMUTABLE_FL`을 재사용하지 않습니다. immutable bit는 내용 쓰기뿐 아니라 delete, rename, link, owner·mode 변경까지 막는데 fs-verity에는 이런 추가 제한이 필요하지 않습니다.

API가 `setxattr()`·`getxattr()`가 아니라 ioctl인 이유는 xattr interface를 임의 syscall처럼 사용하는 것이 Linux filesystem 개발자에게 강하게 권장되지 않기 때문입니다. xattr는 on-disk xattr여야지 Merkle tree 생성을 마술처럼 trigger하는 API가 아니어야 합니다.

현재 구현은 모두 local filesystem이지만 per-file verity metadata를 저장할 수 있다면 원칙적으로 remote filesystem도 지원할 수 있습니다. metadata 저장 선택지가 적다면 파일 끝 뒤에 저장하고 `i_size` 조작으로 사용자 공간에서 숨길 수 있습니다. `fs/verity/` 검증 함수는 Linux pagecache를 가정하지만 local과 remote 파일시스템 모두 보통 pagecache를 사용합니다.

모든 것을 VFS에만 구현하기 어려운 첫 이유는 검증 전에 folio를 Uptodate로 표시하면 안 되는데 현재 각 파일시스템의 `->readahead()`가 그 상태를 책임지기 때문입니다. VFS 단독 검증으로 바꾸려면 VFS와 모든 파일시스템을 크게 수정해야 합니다.

두 번째 이유는 파일시스템 독립 metadata 저장 방식이 마땅하지 않기 때문입니다. Merkle tree는 수 GiB일 수 있지만 많은 파일시스템은 모든 xattr가 4K block 하나에 맞는다고 가정합니다. ext4·f2fs 암호화는 xattr를 암호화하지 않지만 tree는 평문 data hash를 담으므로 파일과 함께 반드시 암호화해야 합니다.

별도 metadata file은 보호 대상 파일의 본질적 일부를 분리해 한쪽만 삭제되는 문제를 만들고, 같은 file에 넣으면 파일시스템의 `i_size`와 VFS `i_size`를 분리하지 않는 한 application을 깨뜨립니다. 이는 모든 파일시스템에 복잡한 변경이 필요합니다.

세 번째 이유는 `FS_IOC_ENABLE_VERITY`가 파일시스템 transaction을 사용해 활성화 완료 또는 변경 없음 중 하나만 남기는 것이 바람직하기 때문입니다. crash 뒤 intermediate state가 남으면 문제가 될 수 있습니다.

FAQ 핵심 설계 판단
질문판단
왜 IMA 일부가 아닌가개별 파일 hash mechanism과 시스템 정책은 독립 계층
왜 tree를 disk에 저장하나부분 읽기에서 root 경로만 검증하기 위해 필요
왜 write를 지원하지 않나일관성·재구축 비용이 완전히 다른 설계를 요구
왜 immutable bit가 아닌가delete·rename·metadata 변경은 허용해야 함
왜 ioctl인가xattr를 임의 동작 trigger로 오용하지 않기 위해
왜 파일시스템별 코드가 필요한가Uptodate 시점·metadata 저장·transaction이 파일시스템 책임

대안이 채택되지 않은 이유를 요약합니다.

To test fs-verity, use xfstests.  For example, using `kvm-xfstests
<https://github.com/tytso/xfstests-bld/blob/master/Documentation/kvm-quickstart.md>`_::

    kvm-xfstests -c ext4,f2fs,btrfs -g verity

FAQ
===

This section answers frequently asked questions about fs-verity that
weren't already directly answered in other parts of this document.

:Q: Why isn't fs-verity part of IMA?
:A: fs-verity and IMA (Integrity Measurement Architecture) have
    different focuses.  fs-verity is a filesystem-level mechanism for
    hashing individual files using a Merkle tree.  In contrast, IMA
    specifies a system-wide policy that specifies which files are
    hashed and what to do with those hashes, such as log them,
    authenticate them, or add them to a measurement list.

    IMA supports the fs-verity hashing mechanism as an alternative
    to full file hashes, for those who want the performance and
    security benefits of the Merkle tree based hash.  However, it
    doesn't make sense to force all uses of fs-verity to be through
    IMA.  fs-verity already meets many users' needs even as a
    standalone filesystem feature, and it's testable like other
    filesystem features e.g. with xfstests.

:Q: Isn't fs-verity useless because the attacker can just modify the
    hashes in the Merkle tree, which is stored on-disk?
:A: To verify the authenticity of an fs-verity file you must verify
    the authenticity of the "fs-verity file digest", which
    incorporates the root hash of the Merkle tree.  See `Use cases`_.

:Q: Isn't fs-verity useless because the attacker can just replace a
    verity file with a non-verity one?
:A: See `Use cases`_.  In the initial use case, it's really trusted
    userspace code that authenticates the files; fs-verity is just a
    tool to do this job efficiently and securely.  The trusted
    userspace code will consider non-verity files to be inauthentic.

:Q: Why does the Merkle tree need to be stored on-disk?  Couldn't you
    store just the root hash?
:A: If the Merkle tree wasn't stored on-disk, then you'd have to
    compute the entire tree when the file is first accessed, even if
    just one byte is being read.  This is a fundamental consequence of
    how Merkle tree hashing works.  To verify a leaf node, you need to
    verify the whole path to the root hash, including the root node
    (the thing which the root hash is a hash of).  But if the root
    node isn't stored on-disk, you have to compute it by hashing its
    children, and so on until you've actually hashed the entire file.

    That defeats most of the point of doing a Merkle tree-based hash,
    since if you have to hash the whole file ahead of time anyway,
    then you could simply do sha256(file) instead.  That would be much
    simpler, and a bit faster too.

    It's true that an in-memory Merkle tree could still provide the
    advantage of verification on every read rather than just on the
    first read.  However, it would be inefficient because every time a
    hash page gets evicted (you can't pin the entire Merkle tree into
    memory, since it may be very large), in order to restore it you
    again need to hash everything below it in the tree.  This again
    defeats most of the point of doing a Merkle tree-based hash, since
    a single block read could trigger re-hashing gigabytes of data.

:Q: But couldn't you store just the leaf nodes and compute the rest?
:A: See previous answer; this really just moves up one level, since
    one could alternatively interpret the data blocks as being the
    leaf nodes of the Merkle tree.  It's true that the tree can be
    computed much faster if the leaf level is stored rather than just
    the data, but that's only because each level is less than 1% the
    size of the level below (assuming the recommended settings of
    SHA-256 and 4K blocks).  For the exact same reason, by storing
    "just the leaf nodes" you'd already be storing over 99% of the
    tree, so you might as well simply store the whole tree.

:Q: Can the Merkle tree be built ahead of time, e.g. distributed as
    part of a package that is installed to many computers?
:A: This isn't currently supported.  It was part of the original
    design, but was removed to simplify the kernel UAPI and because it
    wasn't a critical use case.  Files are usually installed once and
    used many times, and cryptographic hashing is somewhat fast on
    most modern processors.

:Q: Why doesn't fs-verity support writes?
:A: Write support would be very difficult and would require a
    completely different design, so it's well outside the scope of
    fs-verity.  Write support would require:

    - A way to maintain consistency between the data and hashes,
      including all levels of hashes, since corruption after a crash
      (especially of potentially the entire file!) is unacceptable.
      The main options for solving this are data journalling,
      copy-on-write, and log-structured volume.  But it's very hard to
      retrofit existing filesystems with new consistency mechanisms.
      Data journalling is available on ext4, but is very slow.

    - Rebuilding the Merkle tree after every write, which would be
      extremely inefficient.  Alternatively, a different authenticated
      dictionary structure such as an "authenticated skiplist" could
      be used.  However, this would be far more complex.

    Compare it to dm-verity vs. dm-integrity.  dm-verity is very
    simple: the kernel just verifies read-only data against a
    read-only Merkle tree.  In contrast, dm-integrity supports writes
    but is slow, is much more complex, and doesn't actually support
    full-device authentication since it authenticates each sector
    independently, i.e. there is no "root hash".  It doesn't really
    make sense for the same device-mapper target to support these two
    very different cases; the same applies to fs-verity.

:Q: Since verity files are immutable, why isn't the immutable bit set?
:A: The existing "immutable" bit (FS_IMMUTABLE_FL) already has a
    specific set of semantics which not only make the file contents
    read-only, but also prevent the file from being deleted, renamed,
    linked to, or having its owner or mode changed.  These extra
    properties are unwanted for fs-verity, so reusing the immutable
    bit isn't appropriate.

:Q: Why does the API use ioctls instead of setxattr() and getxattr()?
:A: Abusing the xattr interface for basically arbitrary syscalls is
    heavily frowned upon by most of the Linux filesystem developers.
    An xattr should really just be an xattr on-disk, not an API to
    e.g. magically trigger construction of a Merkle tree.

:Q: Does fs-verity support remote filesystems?
:A: So far all filesystems that have implemented fs-verity support are
    local filesystems, but in principle any filesystem that can store
    per-file verity metadata can support fs-verity, regardless of
    whether it's local or remote.  Some filesystems may have fewer
    options of where to store the verity metadata; one possibility is
    to store it past the end of the file and "hide" it from userspace
    by manipulating i_size.  The data verification functions provided
    by ``fs/verity/`` also assume that the filesystem uses the Linux
    pagecache, but both local and remote filesystems normally do so.

:Q: Why is anything filesystem-specific at all?  Shouldn't fs-verity
    be implemented entirely at the VFS level?
:A: There are many reasons why this is not possible or would be very
    difficult, including the following:

    - To prevent bypassing verification, folios must not be marked
      Uptodate until they've been verified.  Currently, each
      filesystem is responsible for marking folios Uptodate via
      ``->readahead()``.  Therefore, currently it's not possible for
      the VFS to do the verification on its own.  Changing this would
      require significant changes to the VFS and all filesystems.

    - It would require defining a filesystem-independent way to store
      the verity metadata.  Extended attributes don't work for this
      because (a) the Merkle tree may be gigabytes, but many
      filesystems assume that all xattrs fit into a single 4K
      filesystem block, and (b) ext4 and f2fs encryption doesn't
      encrypt xattrs, yet the Merkle tree *must* be encrypted when the
      file contents are, because it stores hashes of the plaintext
      file contents.

      So the verity metadata would have to be stored in an actual
      file.  Using a separate file would be very ugly, since the
      metadata is fundamentally part of the file to be protected, and
      it could cause problems where users could delete the real file
      but not the metadata file or vice versa.  On the other hand,
      having it be in the same file would break applications unless
      filesystems' notion of i_size were divorced from the VFS's,
      which would be complex and require changes to all filesystems.

    - It's desirable that FS_IOC_ENABLE_VERITY uses the filesystem's
      transaction mechanism so that either the file ends up with
      verity enabled, or no changes were made.  Allowing intermediate
      states to occur after a crash may cause problems.