1.. SPDX-License-Identifier: GPL-2.0 2 3.. _fsverity: 4 5======================================================= 6fs-verity: read-only file-based authenticity protection 7======================================================= 8 9Introduction 10============ 11 12fs-verity (``fs/verity/``) is a support layer that filesystems can 13hook into to support transparent integrity and authenticity protection 14of read-only files. Currently, it is supported by the ext4 and f2fs 15filesystems. Like fscrypt, not too much filesystem-specific code is 16needed to support fs-verity. 17 18fs-verity is similar to `dm-verity 19<https://www.kernel.org/doc/Documentation/device-mapper/verity.txt>`_ 20but works on files rather than block devices. On regular files on 21filesystems supporting fs-verity, userspace can execute an ioctl that 22causes the filesystem to build a Merkle tree for the file and persist 23it to a filesystem-specific location associated with the file. 24 25After this, the file is made readonly, and all reads from the file are 26automatically verified against the file's Merkle tree. Reads of any 27corrupted data, including mmap reads, will fail. 28 29Userspace can use another ioctl to retrieve the root hash (actually 30the "fs-verity file digest", which is a hash that includes the Merkle 31tree root hash) that fs-verity is enforcing for the file. This ioctl 32executes in constant time, regardless of the file size. 33 34fs-verity is essentially a way to hash a file in constant time, 35subject to the caveat that reads which would violate the hash will 36fail at runtime. 37 38Use cases 39========= 40 41By itself, the base fs-verity feature only provides integrity 42protection, i.e. detection of accidental (non-malicious) corruption. 43 44However, because fs-verity makes retrieving the file hash extremely 45efficient, it's primarily meant to be used as a tool to support 46authentication (detection of malicious modifications) or auditing 47(logging file hashes before use). 48 49Trusted userspace code (e.g. operating system code running on a 50read-only partition that is itself authenticated by dm-verity) can 51authenticate the contents of an fs-verity file by using the 52`FS_IOC_MEASURE_VERITY`_ ioctl to retrieve its hash, then verifying a 53digital signature of it. 54 55A standard file hash could be used instead of fs-verity. However, 56this is inefficient if the file is large and only a small portion may 57be accessed. This is often the case for Android application package 58(APK) files, for example. These typically contain many translations, 59classes, and other resources that are infrequently or even never 60accessed on a particular device. It would be slow and wasteful to 61read and hash the entire file before starting the application. 62 63Unlike an ahead-of-time hash, fs-verity also re-verifies data each 64time it's paged in. This ensures that malicious disk firmware can't 65undetectably change the contents of the file at runtime. 66 67fs-verity does not replace or obsolete dm-verity. dm-verity should 68still be used on read-only filesystems. fs-verity is for files that 69must live on a read-write filesystem because they are independently 70updated and potentially user-installed, so dm-verity cannot be used. 71 72The base fs-verity feature is a hashing mechanism only; actually 73authenticating the files is up to userspace. However, to meet some 74users' needs, fs-verity optionally supports a simple signature 75verification mechanism where users can configure the kernel to require 76that all fs-verity files be signed by a key loaded into a keyring; see 77`Built-in signature verification`_. Support for fs-verity file hashes 78in IMA (Integrity Measurement Architecture) policies is also planned. 79 80User API 81======== 82 83FS_IOC_ENABLE_VERITY 84-------------------- 85 86The FS_IOC_ENABLE_VERITY ioctl enables fs-verity on a file. It takes 87in a pointer to a struct fsverity_enable_arg, defined as 88follows:: 89 90 struct fsverity_enable_arg { 91 __u32 version; 92 __u32 hash_algorithm; 93 __u32 block_size; 94 __u32 salt_size; 95 __u64 salt_ptr; 96 __u32 sig_size; 97 __u32 __reserved1; 98 __u64 sig_ptr; 99 __u64 __reserved2[11]; 100 }; 101 102This structure contains the parameters of the Merkle tree to build for 103the file, and optionally contains a signature. It must be initialized 104as follows: 105 106- ``version`` must be 1. 107- ``hash_algorithm`` must be the identifier for the hash algorithm to 108 use for the Merkle tree, such as FS_VERITY_HASH_ALG_SHA256. See 109 ``include/uapi/linux/fsverity.h`` for the list of possible values. 110- ``block_size`` must be the Merkle tree block size. Currently, this 111 must be equal to the system page size, which is usually 4096 bytes. 112 Other sizes may be supported in the future. This value is not 113 necessarily the same as the filesystem block size. 114- ``salt_size`` is the size of the salt in bytes, or 0 if no salt is 115 provided. The salt is a value that is prepended to every hashed 116 block; it can be used to personalize the hashing for a particular 117 file or device. Currently the maximum salt size is 32 bytes. 118- ``salt_ptr`` is the pointer to the salt, or NULL if no salt is 119 provided. 120- ``sig_size`` is the size of the signature in bytes, or 0 if no 121 signature is provided. Currently the signature is (somewhat 122 arbitrarily) limited to 16128 bytes. See `Built-in signature 123 verification`_ for more information. 124- ``sig_ptr`` is the pointer to the signature, or NULL if no 125 signature is provided. 126- All reserved fields must be zeroed. 127 128FS_IOC_ENABLE_VERITY causes the filesystem to build a Merkle tree for 129the file and persist it to a filesystem-specific location associated 130with the file, then mark the file as a verity file. This ioctl may 131take a long time to execute on large files, and it is interruptible by 132fatal signals. 133 134FS_IOC_ENABLE_VERITY checks for write access to the inode. However, 135it must be executed on an O_RDONLY file descriptor and no processes 136can have the file open for writing. Attempts to open the file for 137writing while this ioctl is executing will fail with ETXTBSY. (This 138is necessary to guarantee that no writable file descriptors will exist 139after verity is enabled, and to guarantee that the file's contents are 140stable while the Merkle tree is being built over it.) 141 142On success, FS_IOC_ENABLE_VERITY returns 0, and the file becomes a 143verity file. On failure (including the case of interruption by a 144fatal signal), no changes are made to the file. 145 146FS_IOC_ENABLE_VERITY can fail with the following errors: 147 148- ``EACCES``: the process does not have write access to the file 149- ``EBADMSG``: the signature is malformed 150- ``EBUSY``: this ioctl is already running on the file 151- ``EEXIST``: the file already has verity enabled 152- ``EFAULT``: the caller provided inaccessible memory 153- ``EINTR``: the operation was interrupted by a fatal signal 154- ``EINVAL``: unsupported version, hash algorithm, or block size; or 155 reserved bits are set; or the file descriptor refers to neither a 156 regular file nor a directory. 157- ``EISDIR``: the file descriptor refers to a directory 158- ``EKEYREJECTED``: the signature doesn't match the file 159- ``EMSGSIZE``: the salt or signature is too long 160- ``ENOKEY``: the fs-verity keyring doesn't contain the certificate 161 needed to verify the signature 162- ``ENOPKG``: fs-verity recognizes the hash algorithm, but it's not 163 available in the kernel's crypto API as currently configured (e.g. 164 for SHA-512, missing CONFIG_CRYPTO_SHA512). 165- ``ENOTTY``: this type of filesystem does not implement fs-verity 166- ``EOPNOTSUPP``: the kernel was not configured with fs-verity 167 support; or the filesystem superblock has not had the 'verity' 168 feature enabled on it; or the filesystem does not support fs-verity 169 on this file. (See `Filesystem support`_.) 170- ``EPERM``: the file is append-only; or, a signature is required and 171 one was not provided. 172- ``EROFS``: the filesystem is read-only 173- ``ETXTBSY``: someone has the file open for writing. This can be the 174 caller's file descriptor, another open file descriptor, or the file 175 reference held by a writable memory map. 176 177FS_IOC_MEASURE_VERITY 178--------------------- 179 180The FS_IOC_MEASURE_VERITY ioctl retrieves the digest of a verity file. 181The fs-verity file digest is a cryptographic digest that identifies 182the file contents that are being enforced on reads; it is computed via 183a Merkle tree and is different from a traditional full-file digest. 184 185This ioctl takes in a pointer to a variable-length structure:: 186 187 struct fsverity_digest { 188 __u16 digest_algorithm; 189 __u16 digest_size; /* input/output */ 190 __u8 digest[]; 191 }; 192 193``digest_size`` is an input/output field. On input, it must be 194initialized to the number of bytes allocated for the variable-length 195``digest`` field. 196 197On success, 0 is returned and the kernel fills in the structure as 198follows: 199 200- ``digest_algorithm`` will be the hash algorithm used for the file 201 digest. It will match ``fsverity_enable_arg::hash_algorithm``. 202- ``digest_size`` will be the size of the digest in bytes, e.g. 32 203 for SHA-256. (This can be redundant with ``digest_algorithm``.) 204- ``digest`` will be the actual bytes of the digest. 205 206FS_IOC_MEASURE_VERITY is guaranteed to execute in constant time, 207regardless of the size of the file. 208 209FS_IOC_MEASURE_VERITY can fail with the following errors: 210 211- ``EFAULT``: the caller provided inaccessible memory 212- ``ENODATA``: the file is not a verity file 213- ``ENOTTY``: this type of filesystem does not implement fs-verity 214- ``EOPNOTSUPP``: the kernel was not configured with fs-verity 215 support, or the filesystem superblock has not had the 'verity' 216 feature enabled on it. (See `Filesystem support`_.) 217- ``EOVERFLOW``: the digest is longer than the specified 218 ``digest_size`` bytes. Try providing a larger buffer. 219 220FS_IOC_READ_VERITY_METADATA 221--------------------------- 222 223The FS_IOC_READ_VERITY_METADATA ioctl reads verity metadata from a 224verity file. This ioctl is available since Linux v5.12. 225 226This ioctl allows writing a server program that takes a verity file 227and serves it to a client program, such that the client can do its own 228fs-verity compatible verification of the file. This only makes sense 229if the client doesn't trust the server and if the server needs to 230provide the storage for the client. 231 232This is a fairly specialized use case, and most fs-verity users won't 233need this ioctl. 234 235This ioctl takes in a pointer to the following structure:: 236 237 #define FS_VERITY_METADATA_TYPE_MERKLE_TREE 1 238 239 struct fsverity_read_metadata_arg { 240 __u64 metadata_type; 241 __u64 offset; 242 __u64 length; 243 __u64 buf_ptr; 244 __u64 __reserved; 245 }; 246 247``metadata_type`` specifies the type of metadata to read: 248 249- ``FS_VERITY_METADATA_TYPE_MERKLE_TREE`` reads the blocks of the 250 Merkle tree. The blocks are returned in order from the root level 251 to the leaf level. Within each level, the blocks are returned in 252 the same order that their hashes are themselves hashed. 253 See `Merkle tree`_ for more information. 254 255The semantics are similar to those of ``pread()``. ``offset`` 256specifies the offset in bytes into the metadata item to read from, and 257``length`` specifies the maximum number of bytes to read from the 258metadata item. ``buf_ptr`` is the pointer to the buffer to read into, 259cast to a 64-bit integer. ``__reserved`` must be 0. On success, the 260number of bytes read is returned. 0 is returned at the end of the 261metadata item. The returned length may be less than ``length``, for 262example if the ioctl is interrupted. 263 264The metadata returned by FS_IOC_READ_VERITY_METADATA isn't guaranteed 265to be authenticated against the file digest that would be returned by 266`FS_IOC_MEASURE_VERITY`_, as the metadata is expected to be used to 267implement fs-verity compatible verification anyway (though absent a 268malicious disk, the metadata will indeed match). E.g. to implement 269this ioctl, the filesystem is allowed to just read the Merkle tree 270blocks from disk without actually verifying the path to the root node. 271 272FS_IOC_READ_VERITY_METADATA can fail with the following errors: 273 274- ``EFAULT``: the caller provided inaccessible memory 275- ``EINTR``: the ioctl was interrupted before any data was read 276- ``EINVAL``: reserved fields were set, or ``offset + length`` 277 overflowed 278- ``ENODATA``: the file is not a verity file 279- ``ENOTTY``: this type of filesystem does not implement fs-verity, or 280 this ioctl is not yet implemented on it 281- ``EOPNOTSUPP``: the kernel was not configured with fs-verity 282 support, or the filesystem superblock has not had the 'verity' 283 feature enabled on it. (See `Filesystem support`_.) 284 285FS_IOC_GETFLAGS 286--------------- 287 288The existing ioctl FS_IOC_GETFLAGS (which isn't specific to fs-verity) 289can also be used to check whether a file has fs-verity enabled or not. 290To do so, check for FS_VERITY_FL (0x00100000) in the returned flags. 291 292The verity flag is not settable via FS_IOC_SETFLAGS. You must use 293FS_IOC_ENABLE_VERITY instead, since parameters must be provided. 294 295statx 296----- 297 298Since Linux v5.5, the statx() system call sets STATX_ATTR_VERITY if 299the file has fs-verity enabled. This can perform better than 300FS_IOC_GETFLAGS and FS_IOC_MEASURE_VERITY because it doesn't require 301opening the file, and opening verity files can be expensive. 302 303Accessing verity files 304====================== 305 306Applications can transparently access a verity file just like a 307non-verity one, with the following exceptions: 308 309- Verity files are readonly. They cannot be opened for writing or 310 truncate()d, even if the file mode bits allow it. Attempts to do 311 one of these things will fail with EPERM. However, changes to 312 metadata such as owner, mode, timestamps, and xattrs are still 313 allowed, since these are not measured by fs-verity. Verity files 314 can also still be renamed, deleted, and linked to. 315 316- Direct I/O is not supported on verity files. Attempts to use direct 317 I/O on such files will fall back to buffered I/O. 318 319- DAX (Direct Access) is not supported on verity files, because this 320 would circumvent the data verification. 321 322- Reads of data that doesn't match the verity Merkle tree will fail 323 with EIO (for read()) or SIGBUS (for mmap() reads). 324 325- If the sysctl "fs.verity.require_signatures" is set to 1 and the 326 file is not signed by a key in the fs-verity keyring, then opening 327 the file will fail. See `Built-in signature verification`_. 328 329Direct access to the Merkle tree is not supported. Therefore, if a 330verity file is copied, or is backed up and restored, then it will lose 331its "verity"-ness. fs-verity is primarily meant for files like 332executables that are managed by a package manager. 333 334File digest computation 335======================= 336 337This section describes how fs-verity hashes the file contents using a 338Merkle tree to produce the digest which cryptographically identifies 339the file contents. This algorithm is the same for all filesystems 340that support fs-verity. 341 342Userspace only needs to be aware of this algorithm if it needs to 343compute fs-verity file digests itself, e.g. in order to sign files. 344 345.. _fsverity_merkle_tree: 346 347Merkle tree 348----------- 349 350The file contents is divided into blocks, where the block size is 351configurable but is usually 4096 bytes. The end of the last block is 352zero-padded if needed. Each block is then hashed, producing the first 353level of hashes. Then, the hashes in this first level are grouped 354into 'blocksize'-byte blocks (zero-padding the ends as needed) and 355these blocks are hashed, producing the second level of hashes. This 356proceeds up the tree until only a single block remains. The hash of 357this block is the "Merkle tree root hash". 358 359If the file fits in one block and is nonempty, then the "Merkle tree 360root hash" is simply the hash of the single data block. If the file 361is empty, then the "Merkle tree root hash" is all zeroes. 362 363The "blocks" here are not necessarily the same as "filesystem blocks". 364 365If a salt was specified, then it's zero-padded to the closest multiple 366of the input size of the hash algorithm's compression function, e.g. 36764 bytes for SHA-256 or 128 bytes for SHA-512. The padded salt is 368prepended to every data or Merkle tree block that is hashed. 369 370The purpose of the block padding is to cause every hash to be taken 371over the same amount of data, which simplifies the implementation and 372keeps open more possibilities for hardware acceleration. The purpose 373of the salt padding is to make the salting "free" when the salted hash 374state is precomputed, then imported for each hash. 375 376Example: in the recommended configuration of SHA-256 and 4K blocks, 377128 hash values fit in each block. Thus, each level of the Merkle 378tree is approximately 128 times smaller than the previous, and for 379large files the Merkle tree's size converges to approximately 1/127 of 380the original file size. However, for small files, the padding is 381significant, making the space overhead proportionally more. 382 383.. _fsverity_descriptor: 384 385fs-verity descriptor 386-------------------- 387 388By itself, the Merkle tree root hash is ambiguous. For example, it 389can't a distinguish a large file from a small second file whose data 390is exactly the top-level hash block of the first file. Ambiguities 391also arise from the convention of padding to the next block boundary. 392 393To solve this problem, the fs-verity file digest is actually computed 394as a hash of the following structure, which contains the Merkle tree 395root hash as well as other fields such as the file size:: 396 397 struct fsverity_descriptor { 398 __u8 version; /* must be 1 */ 399 __u8 hash_algorithm; /* Merkle tree hash algorithm */ 400 __u8 log_blocksize; /* log2 of size of data and tree blocks */ 401 __u8 salt_size; /* size of salt in bytes; 0 if none */ 402 __le32 __reserved_0x04; /* must be 0 */ 403 __le64 data_size; /* size of file the Merkle tree is built over */ 404 __u8 root_hash[64]; /* Merkle tree root hash */ 405 __u8 salt[32]; /* salt prepended to each hashed block */ 406 __u8 __reserved[144]; /* must be 0's */ 407 }; 408 409Built-in signature verification 410=============================== 411 412With CONFIG_FS_VERITY_BUILTIN_SIGNATURES=y, fs-verity supports putting 413a portion of an authentication policy (see `Use cases`_) in the 414kernel. Specifically, it adds support for: 415 4161. At fs-verity module initialization time, a keyring ".fs-verity" is 417 created. The root user can add trusted X.509 certificates to this 418 keyring using the add_key() system call, then (when done) 419 optionally use keyctl_restrict_keyring() to prevent additional 420 certificates from being added. 421 4222. `FS_IOC_ENABLE_VERITY`_ accepts a pointer to a PKCS#7 formatted 423 detached signature in DER format of the file's fs-verity digest. 424 On success, this signature is persisted alongside the Merkle tree. 425 Then, any time the file is opened, the kernel will verify the 426 file's actual digest against this signature, using the certificates 427 in the ".fs-verity" keyring. 428 4293. A new sysctl "fs.verity.require_signatures" is made available. 430 When set to 1, the kernel requires that all verity files have a 431 correctly signed digest as described in (2). 432 433fs-verity file digests must be signed in the following format, which 434is similar to the structure used by `FS_IOC_MEASURE_VERITY`_:: 435 436 struct fsverity_formatted_digest { 437 char magic[8]; /* must be "FSVerity" */ 438 __le16 digest_algorithm; 439 __le16 digest_size; 440 __u8 digest[]; 441 }; 442 443fs-verity's built-in signature verification support is meant as a 444relatively simple mechanism that can be used to provide some level of 445authenticity protection for verity files, as an alternative to doing 446the signature verification in userspace or using IMA-appraisal. 447However, with this mechanism, userspace programs still need to check 448that the verity bit is set, and there is no protection against verity 449files being swapped around. 450 451Filesystem support 452================== 453 454fs-verity is currently supported by the ext4 and f2fs filesystems. 455The CONFIG_FS_VERITY kconfig option must be enabled to use fs-verity 456on either filesystem. 457 458``include/linux/fsverity.h`` declares the interface between the 459``fs/verity/`` support layer and filesystems. Briefly, filesystems 460must provide an ``fsverity_operations`` structure that provides 461methods to read and write the verity metadata to a filesystem-specific 462location, including the Merkle tree blocks and 463``fsverity_descriptor``. Filesystems must also call functions in 464``fs/verity/`` at certain times, such as when a file is opened or when 465pages have been read into the pagecache. (See `Verifying data`_.) 466 467ext4 468---- 469 470ext4 supports fs-verity since Linux v5.4 and e2fsprogs v1.45.2. 471 472To create verity files on an ext4 filesystem, the filesystem must have 473been formatted with ``-O verity`` or had ``tune2fs -O verity`` run on 474it. "verity" is an RO_COMPAT filesystem feature, so once set, old 475kernels will only be able to mount the filesystem readonly, and old 476versions of e2fsck will be unable to check the filesystem. Moreover, 477currently ext4 only supports mounting a filesystem with the "verity" 478feature when its block size is equal to PAGE_SIZE (often 4096 bytes). 479 480ext4 sets the EXT4_VERITY_FL on-disk inode flag on verity files. It 481can only be set by `FS_IOC_ENABLE_VERITY`_, and it cannot be cleared. 482 483ext4 also supports encryption, which can be used simultaneously with 484fs-verity. In this case, the plaintext data is verified rather than 485the ciphertext. This is necessary in order to make the fs-verity file 486digest meaningful, since every file is encrypted differently. 487 488ext4 stores the verity metadata (Merkle tree and fsverity_descriptor) 489past the end of the file, starting at the first 64K boundary beyond 490i_size. This approach works because (a) verity files are readonly, 491and (b) pages fully beyond i_size aren't visible to userspace but can 492be read/written internally by ext4 with only some relatively small 493changes to ext4. This approach avoids having to depend on the 494EA_INODE feature and on rearchitecturing ext4's xattr support to 495support paging multi-gigabyte xattrs into memory, and to support 496encrypting xattrs. Note that the verity metadata *must* be encrypted 497when the file is, since it contains hashes of the plaintext data. 498 499Currently, ext4 verity only supports the case where the Merkle tree 500block size, filesystem block size, and page size are all the same. It 501also only supports extent-based files. 502 503f2fs 504---- 505 506f2fs supports fs-verity since Linux v5.4 and f2fs-tools v1.11.0. 507 508To create verity files on an f2fs filesystem, the filesystem must have 509been formatted with ``-O verity``. 510 511f2fs sets the FADVISE_VERITY_BIT on-disk inode flag on verity files. 512It can only be set by `FS_IOC_ENABLE_VERITY`_, and it cannot be 513cleared. 514 515Like ext4, f2fs stores the verity metadata (Merkle tree and 516fsverity_descriptor) past the end of the file, starting at the first 51764K boundary beyond i_size. See explanation for ext4 above. 518Moreover, f2fs supports at most 4096 bytes of xattr entries per inode 519which wouldn't be enough for even a single Merkle tree block. 520 521Currently, f2fs verity only supports a Merkle tree block size of 4096. 522Also, f2fs doesn't support enabling verity on files that currently 523have atomic or volatile writes pending. 524 525Implementation details 526====================== 527 528Verifying data 529-------------- 530 531fs-verity ensures that all reads of a verity file's data are verified, 532regardless of which syscall is used to do the read (e.g. mmap(), 533read(), pread()) and regardless of whether it's the first read or a 534later read (unless the later read can return cached data that was 535already verified). Below, we describe how filesystems implement this. 536 537Pagecache 538~~~~~~~~~ 539 540For filesystems using Linux's pagecache, the ``->readpage()`` and 541``->readpages()`` methods must be modified to verify pages before they 542are marked Uptodate. Merely hooking ``->read_iter()`` would be 543insufficient, since ``->read_iter()`` is not used for memory maps. 544 545Therefore, fs/verity/ provides a function fsverity_verify_page() which 546verifies a page that has been read into the pagecache of a verity 547inode, but is still locked and not Uptodate, so it's not yet readable 548by userspace. As needed to do the verification, 549fsverity_verify_page() will call back into the filesystem to read 550Merkle tree pages via fsverity_operations::read_merkle_tree_page(). 551 552fsverity_verify_page() returns false if verification failed; in this 553case, the filesystem must not set the page Uptodate. Following this, 554as per the usual Linux pagecache behavior, attempts by userspace to 555read() from the part of the file containing the page will fail with 556EIO, and accesses to the page within a memory map will raise SIGBUS. 557 558fsverity_verify_page() currently only supports the case where the 559Merkle tree block size is equal to PAGE_SIZE (often 4096 bytes). 560 561In principle, fsverity_verify_page() verifies the entire path in the 562Merkle tree from the data page to the root hash. However, for 563efficiency the filesystem may cache the hash pages. Therefore, 564fsverity_verify_page() only ascends the tree reading hash pages until 565an already-verified hash page is seen, as indicated by the PageChecked 566bit being set. It then verifies the path to that page. 567 568This optimization, which is also used by dm-verity, results in 569excellent sequential read performance. This is because usually (e.g. 570127 in 128 times for 4K blocks and SHA-256) the hash page from the 571bottom level of the tree will already be cached and checked from 572reading a previous data page. However, random reads perform worse. 573 574Block device based filesystems 575~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 576 577Block device based filesystems (e.g. ext4 and f2fs) in Linux also use 578the pagecache, so the above subsection applies too. However, they 579also usually read many pages from a file at once, grouped into a 580structure called a "bio". To make it easier for these types of 581filesystems to support fs-verity, fs/verity/ also provides a function 582fsverity_verify_bio() which verifies all pages in a bio. 583 584ext4 and f2fs also support encryption. If a verity file is also 585encrypted, the pages must be decrypted before being verified. To 586support this, these filesystems allocate a "post-read context" for 587each bio and store it in ``->bi_private``:: 588 589 struct bio_post_read_ctx { 590 struct bio *bio; 591 struct work_struct work; 592 unsigned int cur_step; 593 unsigned int enabled_steps; 594 }; 595 596``enabled_steps`` is a bitmask that specifies whether decryption, 597verity, or both is enabled. After the bio completes, for each needed 598postprocessing step the filesystem enqueues the bio_post_read_ctx on a 599workqueue, and then the workqueue work does the decryption or 600verification. Finally, pages where no decryption or verity error 601occurred are marked Uptodate, and the pages are unlocked. 602 603Files on ext4 and f2fs may contain holes. Normally, ``->readpages()`` 604simply zeroes holes and sets the corresponding pages Uptodate; no bios 605are issued. To prevent this case from bypassing fs-verity, these 606filesystems use fsverity_verify_page() to verify hole pages. 607 608ext4 and f2fs disable direct I/O on verity files, since otherwise 609direct I/O would bypass fs-verity. (They also do the same for 610encrypted files.) 611 612Userspace utility 613================= 614 615This document focuses on the kernel, but a userspace utility for 616fs-verity can be found at: 617 618 https://git.kernel.org/pub/scm/linux/kernel/git/ebiggers/fsverity-utils.git 619 620See the README.md file in the fsverity-utils source tree for details, 621including examples of setting up fs-verity protected files. 622 623Tests 624===== 625 626To test fs-verity, use xfstests. For example, using `kvm-xfstests 627<https://github.com/tytso/xfstests-bld/blob/master/Documentation/kvm-quickstart.md>`_:: 628 629 kvm-xfstests -c ext4,f2fs -g verity 630 631FAQ 632=== 633 634This section answers frequently asked questions about fs-verity that 635weren't already directly answered in other parts of this document. 636 637:Q: Why isn't fs-verity part of IMA? 638:A: fs-verity and IMA (Integrity Measurement Architecture) have 639 different focuses. fs-verity is a filesystem-level mechanism for 640 hashing individual files using a Merkle tree. In contrast, IMA 641 specifies a system-wide policy that specifies which files are 642 hashed and what to do with those hashes, such as log them, 643 authenticate them, or add them to a measurement list. 644 645 IMA is planned to support the fs-verity hashing mechanism as an 646 alternative to doing full file hashes, for people who want the 647 performance and security benefits of the Merkle tree based hash. 648 But it doesn't make sense to force all uses of fs-verity to be 649 through IMA. As a standalone filesystem feature, fs-verity 650 already meets many users' needs, and it's testable like other 651 filesystem features e.g. with xfstests. 652 653:Q: Isn't fs-verity useless because the attacker can just modify the 654 hashes in the Merkle tree, which is stored on-disk? 655:A: To verify the authenticity of an fs-verity file you must verify 656 the authenticity of the "fs-verity file digest", which 657 incorporates the root hash of the Merkle tree. See `Use cases`_. 658 659:Q: Isn't fs-verity useless because the attacker can just replace a 660 verity file with a non-verity one? 661:A: See `Use cases`_. In the initial use case, it's really trusted 662 userspace code that authenticates the files; fs-verity is just a 663 tool to do this job efficiently and securely. The trusted 664 userspace code will consider non-verity files to be inauthentic. 665 666:Q: Why does the Merkle tree need to be stored on-disk? Couldn't you 667 store just the root hash? 668:A: If the Merkle tree wasn't stored on-disk, then you'd have to 669 compute the entire tree when the file is first accessed, even if 670 just one byte is being read. This is a fundamental consequence of 671 how Merkle tree hashing works. To verify a leaf node, you need to 672 verify the whole path to the root hash, including the root node 673 (the thing which the root hash is a hash of). But if the root 674 node isn't stored on-disk, you have to compute it by hashing its 675 children, and so on until you've actually hashed the entire file. 676 677 That defeats most of the point of doing a Merkle tree-based hash, 678 since if you have to hash the whole file ahead of time anyway, 679 then you could simply do sha256(file) instead. That would be much 680 simpler, and a bit faster too. 681 682 It's true that an in-memory Merkle tree could still provide the 683 advantage of verification on every read rather than just on the 684 first read. However, it would be inefficient because every time a 685 hash page gets evicted (you can't pin the entire Merkle tree into 686 memory, since it may be very large), in order to restore it you 687 again need to hash everything below it in the tree. This again 688 defeats most of the point of doing a Merkle tree-based hash, since 689 a single block read could trigger re-hashing gigabytes of data. 690 691:Q: But couldn't you store just the leaf nodes and compute the rest? 692:A: See previous answer; this really just moves up one level, since 693 one could alternatively interpret the data blocks as being the 694 leaf nodes of the Merkle tree. It's true that the tree can be 695 computed much faster if the leaf level is stored rather than just 696 the data, but that's only because each level is less than 1% the 697 size of the level below (assuming the recommended settings of 698 SHA-256 and 4K blocks). For the exact same reason, by storing 699 "just the leaf nodes" you'd already be storing over 99% of the 700 tree, so you might as well simply store the whole tree. 701 702:Q: Can the Merkle tree be built ahead of time, e.g. distributed as 703 part of a package that is installed to many computers? 704:A: This isn't currently supported. It was part of the original 705 design, but was removed to simplify the kernel UAPI and because it 706 wasn't a critical use case. Files are usually installed once and 707 used many times, and cryptographic hashing is somewhat fast on 708 most modern processors. 709 710:Q: Why doesn't fs-verity support writes? 711:A: Write support would be very difficult and would require a 712 completely different design, so it's well outside the scope of 713 fs-verity. Write support would require: 714 715 - A way to maintain consistency between the data and hashes, 716 including all levels of hashes, since corruption after a crash 717 (especially of potentially the entire file!) is unacceptable. 718 The main options for solving this are data journalling, 719 copy-on-write, and log-structured volume. But it's very hard to 720 retrofit existing filesystems with new consistency mechanisms. 721 Data journalling is available on ext4, but is very slow. 722 723 - Rebuilding the Merkle tree after every write, which would be 724 extremely inefficient. Alternatively, a different authenticated 725 dictionary structure such as an "authenticated skiplist" could 726 be used. However, this would be far more complex. 727 728 Compare it to dm-verity vs. dm-integrity. dm-verity is very 729 simple: the kernel just verifies read-only data against a 730 read-only Merkle tree. In contrast, dm-integrity supports writes 731 but is slow, is much more complex, and doesn't actually support 732 full-device authentication since it authenticates each sector 733 independently, i.e. there is no "root hash". It doesn't really 734 make sense for the same device-mapper target to support these two 735 very different cases; the same applies to fs-verity. 736 737:Q: Since verity files are immutable, why isn't the immutable bit set? 738:A: The existing "immutable" bit (FS_IMMUTABLE_FL) already has a 739 specific set of semantics which not only make the file contents 740 read-only, but also prevent the file from being deleted, renamed, 741 linked to, or having its owner or mode changed. These extra 742 properties are unwanted for fs-verity, so reusing the immutable 743 bit isn't appropriate. 744 745:Q: Why does the API use ioctls instead of setxattr() and getxattr()? 746:A: Abusing the xattr interface for basically arbitrary syscalls is 747 heavily frowned upon by most of the Linux filesystem developers. 748 An xattr should really just be an xattr on-disk, not an API to 749 e.g. magically trigger construction of a Merkle tree. 750 751:Q: Does fs-verity support remote filesystems? 752:A: Only ext4 and f2fs support is implemented currently, but in 753 principle any filesystem that can store per-file verity metadata 754 can support fs-verity, regardless of whether it's local or remote. 755 Some filesystems may have fewer options of where to store the 756 verity metadata; one possibility is to store it past the end of 757 the file and "hide" it from userspace by manipulating i_size. The 758 data verification functions provided by ``fs/verity/`` also assume 759 that the filesystem uses the Linux pagecache, but both local and 760 remote filesystems normally do so. 761 762:Q: Why is anything filesystem-specific at all? Shouldn't fs-verity 763 be implemented entirely at the VFS level? 764:A: There are many reasons why this is not possible or would be very 765 difficult, including the following: 766 767 - To prevent bypassing verification, pages must not be marked 768 Uptodate until they've been verified. Currently, each 769 filesystem is responsible for marking pages Uptodate via 770 ``->readpages()``. Therefore, currently it's not possible for 771 the VFS to do the verification on its own. Changing this would 772 require significant changes to the VFS and all filesystems. 773 774 - It would require defining a filesystem-independent way to store 775 the verity metadata. Extended attributes don't work for this 776 because (a) the Merkle tree may be gigabytes, but many 777 filesystems assume that all xattrs fit into a single 4K 778 filesystem block, and (b) ext4 and f2fs encryption doesn't 779 encrypt xattrs, yet the Merkle tree *must* be encrypted when the 780 file contents are, because it stores hashes of the plaintext 781 file contents. 782 783 So the verity metadata would have to be stored in an actual 784 file. Using a separate file would be very ugly, since the 785 metadata is fundamentally part of the file to be protected, and 786 it could cause problems where users could delete the real file 787 but not the metadata file or vice versa. On the other hand, 788 having it be in the same file would break applications unless 789 filesystems' notion of i_size were divorced from the VFS's, 790 which would be complex and require changes to all filesystems. 791 792 - It's desirable that FS_IOC_ENABLE_VERITY uses the filesystem's 793 transaction mechanism so that either the file ends up with 794 verity enabled, or no changes were made. Allowing intermediate 795 states to occur after a crash may cause problems. 796