Loading...
Note: File does not exist in v3.5.6.
1=====================================
2Filesystem-level encryption (fscrypt)
3=====================================
4
5Introduction
6============
7
8fscrypt is a library which filesystems can hook into to support
9transparent encryption of files and directories.
10
11Note: "fscrypt" in this document refers to the kernel-level portion,
12implemented in ``fs/crypto/``, as opposed to the userspace tool
13`fscrypt <https://github.com/google/fscrypt>`_. This document only
14covers the kernel-level portion. For command-line examples of how to
15use encryption, see the documentation for the userspace tool `fscrypt
16<https://github.com/google/fscrypt>`_. Also, it is recommended to use
17the fscrypt userspace tool, or other existing userspace tools such as
18`fscryptctl <https://github.com/google/fscryptctl>`_ or `Android's key
19management system
20<https://source.android.com/security/encryption/file-based>`_, over
21using the kernel's API directly. Using existing tools reduces the
22chance of introducing your own security bugs. (Nevertheless, for
23completeness this documentation covers the kernel's API anyway.)
24
25Unlike dm-crypt, fscrypt operates at the filesystem level rather than
26at the block device level. This allows it to encrypt different files
27with different keys and to have unencrypted files on the same
28filesystem. This is useful for multi-user systems where each user's
29data-at-rest needs to be cryptographically isolated from the others.
30However, except for filenames, fscrypt does not encrypt filesystem
31metadata.
32
33Unlike eCryptfs, which is a stacked filesystem, fscrypt is integrated
34directly into supported filesystems --- currently ext4, F2FS, and
35UBIFS. This allows encrypted files to be read and written without
36caching both the decrypted and encrypted pages in the pagecache,
37thereby nearly halving the memory used and bringing it in line with
38unencrypted files. Similarly, half as many dentries and inodes are
39needed. eCryptfs also limits encrypted filenames to 143 bytes,
40causing application compatibility issues; fscrypt allows the full 255
41bytes (NAME_MAX). Finally, unlike eCryptfs, the fscrypt API can be
42used by unprivileged users, with no need to mount anything.
43
44fscrypt does not support encrypting files in-place. Instead, it
45supports marking an empty directory as encrypted. Then, after
46userspace provides the key, all regular files, directories, and
47symbolic links created in that directory tree are transparently
48encrypted.
49
50Threat model
51============
52
53Offline attacks
54---------------
55
56Provided that userspace chooses a strong encryption key, fscrypt
57protects the confidentiality of file contents and filenames in the
58event of a single point-in-time permanent offline compromise of the
59block device content. fscrypt does not protect the confidentiality of
60non-filename metadata, e.g. file sizes, file permissions, file
61timestamps, and extended attributes. Also, the existence and location
62of holes (unallocated blocks which logically contain all zeroes) in
63files is not protected.
64
65fscrypt is not guaranteed to protect confidentiality or authenticity
66if an attacker is able to manipulate the filesystem offline prior to
67an authorized user later accessing the filesystem.
68
69Online attacks
70--------------
71
72fscrypt (and storage encryption in general) can only provide limited
73protection, if any at all, against online attacks. In detail:
74
75Side-channel attacks
76~~~~~~~~~~~~~~~~~~~~
77
78fscrypt is only resistant to side-channel attacks, such as timing or
79electromagnetic attacks, to the extent that the underlying Linux
80Cryptographic API algorithms are. If a vulnerable algorithm is used,
81such as a table-based implementation of AES, it may be possible for an
82attacker to mount a side channel attack against the online system.
83Side channel attacks may also be mounted against applications
84consuming decrypted data.
85
86Unauthorized file access
87~~~~~~~~~~~~~~~~~~~~~~~~
88
89After an encryption key has been added, fscrypt does not hide the
90plaintext file contents or filenames from other users on the same
91system. Instead, existing access control mechanisms such as file mode
92bits, POSIX ACLs, LSMs, or namespaces should be used for this purpose.
93
94(For the reasoning behind this, understand that while the key is
95added, the confidentiality of the data, from the perspective of the
96system itself, is *not* protected by the mathematical properties of
97encryption but rather only by the correctness of the kernel.
98Therefore, any encryption-specific access control checks would merely
99be enforced by kernel *code* and therefore would be largely redundant
100with the wide variety of access control mechanisms already available.)
101
102Kernel memory compromise
103~~~~~~~~~~~~~~~~~~~~~~~~
104
105An attacker who compromises the system enough to read from arbitrary
106memory, e.g. by mounting a physical attack or by exploiting a kernel
107security vulnerability, can compromise all encryption keys that are
108currently in use.
109
110However, fscrypt allows encryption keys to be removed from the kernel,
111which may protect them from later compromise.
112
113In more detail, the FS_IOC_REMOVE_ENCRYPTION_KEY ioctl (or the
114FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS ioctl) can wipe a master
115encryption key from kernel memory. If it does so, it will also try to
116evict all cached inodes which had been "unlocked" using the key,
117thereby wiping their per-file keys and making them once again appear
118"locked", i.e. in ciphertext or encrypted form.
119
120However, these ioctls have some limitations:
121
122- Per-file keys for in-use files will *not* be removed or wiped.
123 Therefore, for maximum effect, userspace should close the relevant
124 encrypted files and directories before removing a master key, as
125 well as kill any processes whose working directory is in an affected
126 encrypted directory.
127
128- The kernel cannot magically wipe copies of the master key(s) that
129 userspace might have as well. Therefore, userspace must wipe all
130 copies of the master key(s) it makes as well; normally this should
131 be done immediately after FS_IOC_ADD_ENCRYPTION_KEY, without waiting
132 for FS_IOC_REMOVE_ENCRYPTION_KEY. Naturally, the same also applies
133 to all higher levels in the key hierarchy. Userspace should also
134 follow other security precautions such as mlock()ing memory
135 containing keys to prevent it from being swapped out.
136
137- In general, decrypted contents and filenames in the kernel VFS
138 caches are freed but not wiped. Therefore, portions thereof may be
139 recoverable from freed memory, even after the corresponding key(s)
140 were wiped. To partially solve this, you can set
141 CONFIG_PAGE_POISONING=y in your kernel config and add page_poison=1
142 to your kernel command line. However, this has a performance cost.
143
144- Secret keys might still exist in CPU registers, in crypto
145 accelerator hardware (if used by the crypto API to implement any of
146 the algorithms), or in other places not explicitly considered here.
147
148Limitations of v1 policies
149~~~~~~~~~~~~~~~~~~~~~~~~~~
150
151v1 encryption policies have some weaknesses with respect to online
152attacks:
153
154- There is no verification that the provided master key is correct.
155 Therefore, a malicious user can temporarily associate the wrong key
156 with another user's encrypted files to which they have read-only
157 access. Because of filesystem caching, the wrong key will then be
158 used by the other user's accesses to those files, even if the other
159 user has the correct key in their own keyring. This violates the
160 meaning of "read-only access".
161
162- A compromise of a per-file key also compromises the master key from
163 which it was derived.
164
165- Non-root users cannot securely remove encryption keys.
166
167All the above problems are fixed with v2 encryption policies. For
168this reason among others, it is recommended to use v2 encryption
169policies on all new encrypted directories.
170
171Key hierarchy
172=============
173
174Master Keys
175-----------
176
177Each encrypted directory tree is protected by a *master key*. Master
178keys can be up to 64 bytes long, and must be at least as long as the
179greater of the key length needed by the contents and filenames
180encryption modes being used. For example, if AES-256-XTS is used for
181contents encryption, the master key must be 64 bytes (512 bits). Note
182that the XTS mode is defined to require a key twice as long as that
183required by the underlying block cipher.
184
185To "unlock" an encrypted directory tree, userspace must provide the
186appropriate master key. There can be any number of master keys, each
187of which protects any number of directory trees on any number of
188filesystems.
189
190Master keys must be real cryptographic keys, i.e. indistinguishable
191from random bytestrings of the same length. This implies that users
192**must not** directly use a password as a master key, zero-pad a
193shorter key, or repeat a shorter key. Security cannot be guaranteed
194if userspace makes any such error, as the cryptographic proofs and
195analysis would no longer apply.
196
197Instead, users should generate master keys either using a
198cryptographically secure random number generator, or by using a KDF
199(Key Derivation Function). The kernel does not do any key stretching;
200therefore, if userspace derives the key from a low-entropy secret such
201as a passphrase, it is critical that a KDF designed for this purpose
202be used, such as scrypt, PBKDF2, or Argon2.
203
204Key derivation function
205-----------------------
206
207With one exception, fscrypt never uses the master key(s) for
208encryption directly. Instead, they are only used as input to a KDF
209(Key Derivation Function) to derive the actual keys.
210
211The KDF used for a particular master key differs depending on whether
212the key is used for v1 encryption policies or for v2 encryption
213policies. Users **must not** use the same key for both v1 and v2
214encryption policies. (No real-world attack is currently known on this
215specific case of key reuse, but its security cannot be guaranteed
216since the cryptographic proofs and analysis would no longer apply.)
217
218For v1 encryption policies, the KDF only supports deriving per-file
219encryption keys. It works by encrypting the master key with
220AES-128-ECB, using the file's 16-byte nonce as the AES key. The
221resulting ciphertext is used as the derived key. If the ciphertext is
222longer than needed, then it is truncated to the needed length.
223
224For v2 encryption policies, the KDF is HKDF-SHA512. The master key is
225passed as the "input keying material", no salt is used, and a distinct
226"application-specific information string" is used for each distinct
227key to be derived. For example, when a per-file encryption key is
228derived, the application-specific information string is the file's
229nonce prefixed with "fscrypt\\0" and a context byte. Different
230context bytes are used for other types of derived keys.
231
232HKDF-SHA512 is preferred to the original AES-128-ECB based KDF because
233HKDF is more flexible, is nonreversible, and evenly distributes
234entropy from the master key. HKDF is also standardized and widely
235used by other software, whereas the AES-128-ECB based KDF is ad-hoc.
236
237Per-file keys
238-------------
239
240Since each master key can protect many files, it is necessary to
241"tweak" the encryption of each file so that the same plaintext in two
242files doesn't map to the same ciphertext, or vice versa. In most
243cases, fscrypt does this by deriving per-file keys. When a new
244encrypted inode (regular file, directory, or symlink) is created,
245fscrypt randomly generates a 16-byte nonce and stores it in the
246inode's encryption xattr. Then, it uses a KDF (as described in `Key
247derivation function`_) to derive the file's key from the master key
248and nonce.
249
250Key derivation was chosen over key wrapping because wrapped keys would
251require larger xattrs which would be less likely to fit in-line in the
252filesystem's inode table, and there didn't appear to be any
253significant advantages to key wrapping. In particular, currently
254there is no requirement to support unlocking a file with multiple
255alternative master keys or to support rotating master keys. Instead,
256the master keys may be wrapped in userspace, e.g. as is done by the
257`fscrypt <https://github.com/google/fscrypt>`_ tool.
258
259Including the inode number in the IVs was considered. However, it was
260rejected as it would have prevented ext4 filesystems from being
261resized, and by itself still wouldn't have been sufficient to prevent
262the same key from being directly reused for both XTS and CTS-CBC.
263
264DIRECT_KEY and per-mode keys
265----------------------------
266
267The Adiantum encryption mode (see `Encryption modes and usage`_) is
268suitable for both contents and filenames encryption, and it accepts
269long IVs --- long enough to hold both an 8-byte logical block number
270and a 16-byte per-file nonce. Also, the overhead of each Adiantum key
271is greater than that of an AES-256-XTS key.
272
273Therefore, to improve performance and save memory, for Adiantum a
274"direct key" configuration is supported. When the user has enabled
275this by setting FSCRYPT_POLICY_FLAG_DIRECT_KEY in the fscrypt policy,
276per-file keys are not used. Instead, whenever any data (contents or
277filenames) is encrypted, the file's 16-byte nonce is included in the
278IV. Moreover:
279
280- For v1 encryption policies, the encryption is done directly with the
281 master key. Because of this, users **must not** use the same master
282 key for any other purpose, even for other v1 policies.
283
284- For v2 encryption policies, the encryption is done with a per-mode
285 key derived using the KDF. Users may use the same master key for
286 other v2 encryption policies.
287
288Key identifiers
289---------------
290
291For master keys used for v2 encryption policies, a unique 16-byte "key
292identifier" is also derived using the KDF. This value is stored in
293the clear, since it is needed to reliably identify the key itself.
294
295Encryption modes and usage
296==========================
297
298fscrypt allows one encryption mode to be specified for file contents
299and one encryption mode to be specified for filenames. Different
300directory trees are permitted to use different encryption modes.
301Currently, the following pairs of encryption modes are supported:
302
303- AES-256-XTS for contents and AES-256-CTS-CBC for filenames
304- AES-128-CBC for contents and AES-128-CTS-CBC for filenames
305- Adiantum for both contents and filenames
306
307If unsure, you should use the (AES-256-XTS, AES-256-CTS-CBC) pair.
308
309AES-128-CBC was added only for low-powered embedded devices with
310crypto accelerators such as CAAM or CESA that do not support XTS. To
311use AES-128-CBC, CONFIG_CRYPTO_SHA256 (or another SHA-256
312implementation) must be enabled so that ESSIV can be used.
313
314Adiantum is a (primarily) stream cipher-based mode that is fast even
315on CPUs without dedicated crypto instructions. It's also a true
316wide-block mode, unlike XTS. It can also eliminate the need to derive
317per-file keys. However, it depends on the security of two primitives,
318XChaCha12 and AES-256, rather than just one. See the paper
319"Adiantum: length-preserving encryption for entry-level processors"
320(https://eprint.iacr.org/2018/720.pdf) for more details. To use
321Adiantum, CONFIG_CRYPTO_ADIANTUM must be enabled. Also, fast
322implementations of ChaCha and NHPoly1305 should be enabled, e.g.
323CONFIG_CRYPTO_CHACHA20_NEON and CONFIG_CRYPTO_NHPOLY1305_NEON for ARM.
324
325New encryption modes can be added relatively easily, without changes
326to individual filesystems. However, authenticated encryption (AE)
327modes are not currently supported because of the difficulty of dealing
328with ciphertext expansion.
329
330Contents encryption
331-------------------
332
333For file contents, each filesystem block is encrypted independently.
334Currently, only the case where the filesystem block size is equal to
335the system's page size (usually 4096 bytes) is supported.
336
337Each block's IV is set to the logical block number within the file as
338a little endian number, except that:
339
340- With CBC mode encryption, ESSIV is also used. Specifically, each IV
341 is encrypted with AES-256 where the AES-256 key is the SHA-256 hash
342 of the file's data encryption key.
343
344- In the "direct key" configuration (FSCRYPT_POLICY_FLAG_DIRECT_KEY
345 set in the fscrypt_policy), the file's nonce is also appended to the
346 IV. Currently this is only allowed with the Adiantum encryption
347 mode.
348
349Filenames encryption
350--------------------
351
352For filenames, each full filename is encrypted at once. Because of
353the requirements to retain support for efficient directory lookups and
354filenames of up to 255 bytes, the same IV is used for every filename
355in a directory.
356
357However, each encrypted directory still uses a unique key; or
358alternatively (for the "direct key" configuration) has the file's
359nonce included in the IVs. Thus, IV reuse is limited to within a
360single directory.
361
362With CTS-CBC, the IV reuse means that when the plaintext filenames
363share a common prefix at least as long as the cipher block size (16
364bytes for AES), the corresponding encrypted filenames will also share
365a common prefix. This is undesirable. Adiantum does not have this
366weakness, as it is a wide-block encryption mode.
367
368All supported filenames encryption modes accept any plaintext length
369>= 16 bytes; cipher block alignment is not required. However,
370filenames shorter than 16 bytes are NUL-padded to 16 bytes before
371being encrypted. In addition, to reduce leakage of filename lengths
372via their ciphertexts, all filenames are NUL-padded to the next 4, 8,
37316, or 32-byte boundary (configurable). 32 is recommended since this
374provides the best confidentiality, at the cost of making directory
375entries consume slightly more space. Note that since NUL (``\0``) is
376not otherwise a valid character in filenames, the padding will never
377produce duplicate plaintexts.
378
379Symbolic link targets are considered a type of filename and are
380encrypted in the same way as filenames in directory entries, except
381that IV reuse is not a problem as each symlink has its own inode.
382
383User API
384========
385
386Setting an encryption policy
387----------------------------
388
389FS_IOC_SET_ENCRYPTION_POLICY
390~~~~~~~~~~~~~~~~~~~~~~~~~~~~
391
392The FS_IOC_SET_ENCRYPTION_POLICY ioctl sets an encryption policy on an
393empty directory or verifies that a directory or regular file already
394has the specified encryption policy. It takes in a pointer to a
395:c:type:`struct fscrypt_policy_v1` or a :c:type:`struct
396fscrypt_policy_v2`, defined as follows::
397
398 #define FSCRYPT_POLICY_V1 0
399 #define FSCRYPT_KEY_DESCRIPTOR_SIZE 8
400 struct fscrypt_policy_v1 {
401 __u8 version;
402 __u8 contents_encryption_mode;
403 __u8 filenames_encryption_mode;
404 __u8 flags;
405 __u8 master_key_descriptor[FSCRYPT_KEY_DESCRIPTOR_SIZE];
406 };
407 #define fscrypt_policy fscrypt_policy_v1
408
409 #define FSCRYPT_POLICY_V2 2
410 #define FSCRYPT_KEY_IDENTIFIER_SIZE 16
411 struct fscrypt_policy_v2 {
412 __u8 version;
413 __u8 contents_encryption_mode;
414 __u8 filenames_encryption_mode;
415 __u8 flags;
416 __u8 __reserved[4];
417 __u8 master_key_identifier[FSCRYPT_KEY_IDENTIFIER_SIZE];
418 };
419
420This structure must be initialized as follows:
421
422- ``version`` must be FSCRYPT_POLICY_V1 (0) if the struct is
423 :c:type:`fscrypt_policy_v1` or FSCRYPT_POLICY_V2 (2) if the struct
424 is :c:type:`fscrypt_policy_v2`. (Note: we refer to the original
425 policy version as "v1", though its version code is really 0.) For
426 new encrypted directories, use v2 policies.
427
428- ``contents_encryption_mode`` and ``filenames_encryption_mode`` must
429 be set to constants from ``<linux/fscrypt.h>`` which identify the
430 encryption modes to use. If unsure, use FSCRYPT_MODE_AES_256_XTS
431 (1) for ``contents_encryption_mode`` and FSCRYPT_MODE_AES_256_CTS
432 (4) for ``filenames_encryption_mode``.
433
434- ``flags`` must contain a value from ``<linux/fscrypt.h>`` which
435 identifies the amount of NUL-padding to use when encrypting
436 filenames. If unsure, use FSCRYPT_POLICY_FLAGS_PAD_32 (0x3).
437 Additionally, if the encryption modes are both
438 FSCRYPT_MODE_ADIANTUM, this can contain
439 FSCRYPT_POLICY_FLAG_DIRECT_KEY; see `DIRECT_KEY and per-mode keys`_.
440
441- For v2 encryption policies, ``__reserved`` must be zeroed.
442
443- For v1 encryption policies, ``master_key_descriptor`` specifies how
444 to find the master key in a keyring; see `Adding keys`_. It is up
445 to userspace to choose a unique ``master_key_descriptor`` for each
446 master key. The e4crypt and fscrypt tools use the first 8 bytes of
447 ``SHA-512(SHA-512(master_key))``, but this particular scheme is not
448 required. Also, the master key need not be in the keyring yet when
449 FS_IOC_SET_ENCRYPTION_POLICY is executed. However, it must be added
450 before any files can be created in the encrypted directory.
451
452 For v2 encryption policies, ``master_key_descriptor`` has been
453 replaced with ``master_key_identifier``, which is longer and cannot
454 be arbitrarily chosen. Instead, the key must first be added using
455 `FS_IOC_ADD_ENCRYPTION_KEY`_. Then, the ``key_spec.u.identifier``
456 the kernel returned in the :c:type:`struct fscrypt_add_key_arg` must
457 be used as the ``master_key_identifier`` in the :c:type:`struct
458 fscrypt_policy_v2`.
459
460If the file is not yet encrypted, then FS_IOC_SET_ENCRYPTION_POLICY
461verifies that the file is an empty directory. If so, the specified
462encryption policy is assigned to the directory, turning it into an
463encrypted directory. After that, and after providing the
464corresponding master key as described in `Adding keys`_, all regular
465files, directories (recursively), and symlinks created in the
466directory will be encrypted, inheriting the same encryption policy.
467The filenames in the directory's entries will be encrypted as well.
468
469Alternatively, if the file is already encrypted, then
470FS_IOC_SET_ENCRYPTION_POLICY validates that the specified encryption
471policy exactly matches the actual one. If they match, then the ioctl
472returns 0. Otherwise, it fails with EEXIST. This works on both
473regular files and directories, including nonempty directories.
474
475When a v2 encryption policy is assigned to a directory, it is also
476required that either the specified key has been added by the current
477user or that the caller has CAP_FOWNER in the initial user namespace.
478(This is needed to prevent a user from encrypting their data with
479another user's key.) The key must remain added while
480FS_IOC_SET_ENCRYPTION_POLICY is executing. However, if the new
481encrypted directory does not need to be accessed immediately, then the
482key can be removed right away afterwards.
483
484Note that the ext4 filesystem does not allow the root directory to be
485encrypted, even if it is empty. Users who want to encrypt an entire
486filesystem with one key should consider using dm-crypt instead.
487
488FS_IOC_SET_ENCRYPTION_POLICY can fail with the following errors:
489
490- ``EACCES``: the file is not owned by the process's uid, nor does the
491 process have the CAP_FOWNER capability in a namespace with the file
492 owner's uid mapped
493- ``EEXIST``: the file is already encrypted with an encryption policy
494 different from the one specified
495- ``EINVAL``: an invalid encryption policy was specified (invalid
496 version, mode(s), or flags; or reserved bits were set)
497- ``ENOKEY``: a v2 encryption policy was specified, but the key with
498 the specified ``master_key_identifier`` has not been added, nor does
499 the process have the CAP_FOWNER capability in the initial user
500 namespace
501- ``ENOTDIR``: the file is unencrypted and is a regular file, not a
502 directory
503- ``ENOTEMPTY``: the file is unencrypted and is a nonempty directory
504- ``ENOTTY``: this type of filesystem does not implement encryption
505- ``EOPNOTSUPP``: the kernel was not configured with encryption
506 support for filesystems, or the filesystem superblock has not
507 had encryption enabled on it. (For example, to use encryption on an
508 ext4 filesystem, CONFIG_FS_ENCRYPTION must be enabled in the
509 kernel config, and the superblock must have had the "encrypt"
510 feature flag enabled using ``tune2fs -O encrypt`` or ``mkfs.ext4 -O
511 encrypt``.)
512- ``EPERM``: this directory may not be encrypted, e.g. because it is
513 the root directory of an ext4 filesystem
514- ``EROFS``: the filesystem is readonly
515
516Getting an encryption policy
517----------------------------
518
519Two ioctls are available to get a file's encryption policy:
520
521- `FS_IOC_GET_ENCRYPTION_POLICY_EX`_
522- `FS_IOC_GET_ENCRYPTION_POLICY`_
523
524The extended (_EX) version of the ioctl is more general and is
525recommended to use when possible. However, on older kernels only the
526original ioctl is available. Applications should try the extended
527version, and if it fails with ENOTTY fall back to the original
528version.
529
530FS_IOC_GET_ENCRYPTION_POLICY_EX
531~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
532
533The FS_IOC_GET_ENCRYPTION_POLICY_EX ioctl retrieves the encryption
534policy, if any, for a directory or regular file. No additional
535permissions are required beyond the ability to open the file. It
536takes in a pointer to a :c:type:`struct fscrypt_get_policy_ex_arg`,
537defined as follows::
538
539 struct fscrypt_get_policy_ex_arg {
540 __u64 policy_size; /* input/output */
541 union {
542 __u8 version;
543 struct fscrypt_policy_v1 v1;
544 struct fscrypt_policy_v2 v2;
545 } policy; /* output */
546 };
547
548The caller must initialize ``policy_size`` to the size available for
549the policy struct, i.e. ``sizeof(arg.policy)``.
550
551On success, the policy struct is returned in ``policy``, and its
552actual size is returned in ``policy_size``. ``policy.version`` should
553be checked to determine the version of policy returned. Note that the
554version code for the "v1" policy is actually 0 (FSCRYPT_POLICY_V1).
555
556FS_IOC_GET_ENCRYPTION_POLICY_EX can fail with the following errors:
557
558- ``EINVAL``: the file is encrypted, but it uses an unrecognized
559 encryption policy version
560- ``ENODATA``: the file is not encrypted
561- ``ENOTTY``: this type of filesystem does not implement encryption,
562 or this kernel is too old to support FS_IOC_GET_ENCRYPTION_POLICY_EX
563 (try FS_IOC_GET_ENCRYPTION_POLICY instead)
564- ``EOPNOTSUPP``: the kernel was not configured with encryption
565 support for this filesystem, or the filesystem superblock has not
566 had encryption enabled on it
567- ``EOVERFLOW``: the file is encrypted and uses a recognized
568 encryption policy version, but the policy struct does not fit into
569 the provided buffer
570
571Note: if you only need to know whether a file is encrypted or not, on
572most filesystems it is also possible to use the FS_IOC_GETFLAGS ioctl
573and check for FS_ENCRYPT_FL, or to use the statx() system call and
574check for STATX_ATTR_ENCRYPTED in stx_attributes.
575
576FS_IOC_GET_ENCRYPTION_POLICY
577~~~~~~~~~~~~~~~~~~~~~~~~~~~~
578
579The FS_IOC_GET_ENCRYPTION_POLICY ioctl can also retrieve the
580encryption policy, if any, for a directory or regular file. However,
581unlike `FS_IOC_GET_ENCRYPTION_POLICY_EX`_,
582FS_IOC_GET_ENCRYPTION_POLICY only supports the original policy
583version. It takes in a pointer directly to a :c:type:`struct
584fscrypt_policy_v1` rather than a :c:type:`struct
585fscrypt_get_policy_ex_arg`.
586
587The error codes for FS_IOC_GET_ENCRYPTION_POLICY are the same as those
588for FS_IOC_GET_ENCRYPTION_POLICY_EX, except that
589FS_IOC_GET_ENCRYPTION_POLICY also returns ``EINVAL`` if the file is
590encrypted using a newer encryption policy version.
591
592Getting the per-filesystem salt
593-------------------------------
594
595Some filesystems, such as ext4 and F2FS, also support the deprecated
596ioctl FS_IOC_GET_ENCRYPTION_PWSALT. This ioctl retrieves a randomly
597generated 16-byte value stored in the filesystem superblock. This
598value is intended to used as a salt when deriving an encryption key
599from a passphrase or other low-entropy user credential.
600
601FS_IOC_GET_ENCRYPTION_PWSALT is deprecated. Instead, prefer to
602generate and manage any needed salt(s) in userspace.
603
604Adding keys
605-----------
606
607FS_IOC_ADD_ENCRYPTION_KEY
608~~~~~~~~~~~~~~~~~~~~~~~~~
609
610The FS_IOC_ADD_ENCRYPTION_KEY ioctl adds a master encryption key to
611the filesystem, making all files on the filesystem which were
612encrypted using that key appear "unlocked", i.e. in plaintext form.
613It can be executed on any file or directory on the target filesystem,
614but using the filesystem's root directory is recommended. It takes in
615a pointer to a :c:type:`struct fscrypt_add_key_arg`, defined as
616follows::
617
618 struct fscrypt_add_key_arg {
619 struct fscrypt_key_specifier key_spec;
620 __u32 raw_size;
621 __u32 __reserved[9];
622 __u8 raw[];
623 };
624
625 #define FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR 1
626 #define FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER 2
627
628 struct fscrypt_key_specifier {
629 __u32 type; /* one of FSCRYPT_KEY_SPEC_TYPE_* */
630 __u32 __reserved;
631 union {
632 __u8 __reserved[32]; /* reserve some extra space */
633 __u8 descriptor[FSCRYPT_KEY_DESCRIPTOR_SIZE];
634 __u8 identifier[FSCRYPT_KEY_IDENTIFIER_SIZE];
635 } u;
636 };
637
638:c:type:`struct fscrypt_add_key_arg` must be zeroed, then initialized
639as follows:
640
641- If the key is being added for use by v1 encryption policies, then
642 ``key_spec.type`` must contain FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR, and
643 ``key_spec.u.descriptor`` must contain the descriptor of the key
644 being added, corresponding to the value in the
645 ``master_key_descriptor`` field of :c:type:`struct
646 fscrypt_policy_v1`. To add this type of key, the calling process
647 must have the CAP_SYS_ADMIN capability in the initial user
648 namespace.
649
650 Alternatively, if the key is being added for use by v2 encryption
651 policies, then ``key_spec.type`` must contain
652 FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER, and ``key_spec.u.identifier`` is
653 an *output* field which the kernel fills in with a cryptographic
654 hash of the key. To add this type of key, the calling process does
655 not need any privileges. However, the number of keys that can be
656 added is limited by the user's quota for the keyrings service (see
657 ``Documentation/security/keys/core.rst``).
658
659- ``raw_size`` must be the size of the ``raw`` key provided, in bytes.
660
661- ``raw`` is a variable-length field which must contain the actual
662 key, ``raw_size`` bytes long.
663
664For v2 policy keys, the kernel keeps track of which user (identified
665by effective user ID) added the key, and only allows the key to be
666removed by that user --- or by "root", if they use
667`FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS`_.
668
669However, if another user has added the key, it may be desirable to
670prevent that other user from unexpectedly removing it. Therefore,
671FS_IOC_ADD_ENCRYPTION_KEY may also be used to add a v2 policy key
672*again*, even if it's already added by other user(s). In this case,
673FS_IOC_ADD_ENCRYPTION_KEY will just install a claim to the key for the
674current user, rather than actually add the key again (but the raw key
675must still be provided, as a proof of knowledge).
676
677FS_IOC_ADD_ENCRYPTION_KEY returns 0 if either the key or a claim to
678the key was either added or already exists.
679
680FS_IOC_ADD_ENCRYPTION_KEY can fail with the following errors:
681
682- ``EACCES``: FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR was specified, but the
683 caller does not have the CAP_SYS_ADMIN capability in the initial
684 user namespace
685- ``EDQUOT``: the key quota for this user would be exceeded by adding
686 the key
687- ``EINVAL``: invalid key size or key specifier type, or reserved bits
688 were set
689- ``ENOTTY``: this type of filesystem does not implement encryption
690- ``EOPNOTSUPP``: the kernel was not configured with encryption
691 support for this filesystem, or the filesystem superblock has not
692 had encryption enabled on it
693
694Legacy method
695~~~~~~~~~~~~~
696
697For v1 encryption policies, a master encryption key can also be
698provided by adding it to a process-subscribed keyring, e.g. to a
699session keyring, or to a user keyring if the user keyring is linked
700into the session keyring.
701
702This method is deprecated (and not supported for v2 encryption
703policies) for several reasons. First, it cannot be used in
704combination with FS_IOC_REMOVE_ENCRYPTION_KEY (see `Removing keys`_),
705so for removing a key a workaround such as keyctl_unlink() in
706combination with ``sync; echo 2 > /proc/sys/vm/drop_caches`` would
707have to be used. Second, it doesn't match the fact that the
708locked/unlocked status of encrypted files (i.e. whether they appear to
709be in plaintext form or in ciphertext form) is global. This mismatch
710has caused much confusion as well as real problems when processes
711running under different UIDs, such as a ``sudo`` command, need to
712access encrypted files.
713
714Nevertheless, to add a key to one of the process-subscribed keyrings,
715the add_key() system call can be used (see:
716``Documentation/security/keys/core.rst``). The key type must be
717"logon"; keys of this type are kept in kernel memory and cannot be
718read back by userspace. The key description must be "fscrypt:"
719followed by the 16-character lower case hex representation of the
720``master_key_descriptor`` that was set in the encryption policy. The
721key payload must conform to the following structure::
722
723 #define FSCRYPT_MAX_KEY_SIZE 64
724
725 struct fscrypt_key {
726 __u32 mode;
727 __u8 raw[FSCRYPT_MAX_KEY_SIZE];
728 __u32 size;
729 };
730
731``mode`` is ignored; just set it to 0. The actual key is provided in
732``raw`` with ``size`` indicating its size in bytes. That is, the
733bytes ``raw[0..size-1]`` (inclusive) are the actual key.
734
735The key description prefix "fscrypt:" may alternatively be replaced
736with a filesystem-specific prefix such as "ext4:". However, the
737filesystem-specific prefixes are deprecated and should not be used in
738new programs.
739
740Removing keys
741-------------
742
743Two ioctls are available for removing a key that was added by
744`FS_IOC_ADD_ENCRYPTION_KEY`_:
745
746- `FS_IOC_REMOVE_ENCRYPTION_KEY`_
747- `FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS`_
748
749These two ioctls differ only in cases where v2 policy keys are added
750or removed by non-root users.
751
752These ioctls don't work on keys that were added via the legacy
753process-subscribed keyrings mechanism.
754
755Before using these ioctls, read the `Kernel memory compromise`_
756section for a discussion of the security goals and limitations of
757these ioctls.
758
759FS_IOC_REMOVE_ENCRYPTION_KEY
760~~~~~~~~~~~~~~~~~~~~~~~~~~~~
761
762The FS_IOC_REMOVE_ENCRYPTION_KEY ioctl removes a claim to a master
763encryption key from the filesystem, and possibly removes the key
764itself. It can be executed on any file or directory on the target
765filesystem, but using the filesystem's root directory is recommended.
766It takes in a pointer to a :c:type:`struct fscrypt_remove_key_arg`,
767defined as follows::
768
769 struct fscrypt_remove_key_arg {
770 struct fscrypt_key_specifier key_spec;
771 #define FSCRYPT_KEY_REMOVAL_STATUS_FLAG_FILES_BUSY 0x00000001
772 #define FSCRYPT_KEY_REMOVAL_STATUS_FLAG_OTHER_USERS 0x00000002
773 __u32 removal_status_flags; /* output */
774 __u32 __reserved[5];
775 };
776
777This structure must be zeroed, then initialized as follows:
778
779- The key to remove is specified by ``key_spec``:
780
781 - To remove a key used by v1 encryption policies, set
782 ``key_spec.type`` to FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR and fill
783 in ``key_spec.u.descriptor``. To remove this type of key, the
784 calling process must have the CAP_SYS_ADMIN capability in the
785 initial user namespace.
786
787 - To remove a key used by v2 encryption policies, set
788 ``key_spec.type`` to FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER and fill
789 in ``key_spec.u.identifier``.
790
791For v2 policy keys, this ioctl is usable by non-root users. However,
792to make this possible, it actually just removes the current user's
793claim to the key, undoing a single call to FS_IOC_ADD_ENCRYPTION_KEY.
794Only after all claims are removed is the key really removed.
795
796For example, if FS_IOC_ADD_ENCRYPTION_KEY was called with uid 1000,
797then the key will be "claimed" by uid 1000, and
798FS_IOC_REMOVE_ENCRYPTION_KEY will only succeed as uid 1000. Or, if
799both uids 1000 and 2000 added the key, then for each uid
800FS_IOC_REMOVE_ENCRYPTION_KEY will only remove their own claim. Only
801once *both* are removed is the key really removed. (Think of it like
802unlinking a file that may have hard links.)
803
804If FS_IOC_REMOVE_ENCRYPTION_KEY really removes the key, it will also
805try to "lock" all files that had been unlocked with the key. It won't
806lock files that are still in-use, so this ioctl is expected to be used
807in cooperation with userspace ensuring that none of the files are
808still open. However, if necessary, this ioctl can be executed again
809later to retry locking any remaining files.
810
811FS_IOC_REMOVE_ENCRYPTION_KEY returns 0 if either the key was removed
812(but may still have files remaining to be locked), the user's claim to
813the key was removed, or the key was already removed but had files
814remaining to be the locked so the ioctl retried locking them. In any
815of these cases, ``removal_status_flags`` is filled in with the
816following informational status flags:
817
818- ``FSCRYPT_KEY_REMOVAL_STATUS_FLAG_FILES_BUSY``: set if some file(s)
819 are still in-use. Not guaranteed to be set in the case where only
820 the user's claim to the key was removed.
821- ``FSCRYPT_KEY_REMOVAL_STATUS_FLAG_OTHER_USERS``: set if only the
822 user's claim to the key was removed, not the key itself
823
824FS_IOC_REMOVE_ENCRYPTION_KEY can fail with the following errors:
825
826- ``EACCES``: The FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR key specifier type
827 was specified, but the caller does not have the CAP_SYS_ADMIN
828 capability in the initial user namespace
829- ``EINVAL``: invalid key specifier type, or reserved bits were set
830- ``ENOKEY``: the key object was not found at all, i.e. it was never
831 added in the first place or was already fully removed including all
832 files locked; or, the user does not have a claim to the key (but
833 someone else does).
834- ``ENOTTY``: this type of filesystem does not implement encryption
835- ``EOPNOTSUPP``: the kernel was not configured with encryption
836 support for this filesystem, or the filesystem superblock has not
837 had encryption enabled on it
838
839FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS
840~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
841
842FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS is exactly the same as
843`FS_IOC_REMOVE_ENCRYPTION_KEY`_, except that for v2 policy keys, the
844ALL_USERS version of the ioctl will remove all users' claims to the
845key, not just the current user's. I.e., the key itself will always be
846removed, no matter how many users have added it. This difference is
847only meaningful if non-root users are adding and removing keys.
848
849Because of this, FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS also requires
850"root", namely the CAP_SYS_ADMIN capability in the initial user
851namespace. Otherwise it will fail with EACCES.
852
853Getting key status
854------------------
855
856FS_IOC_GET_ENCRYPTION_KEY_STATUS
857~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
858
859The FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl retrieves the status of a
860master encryption key. It can be executed on any file or directory on
861the target filesystem, but using the filesystem's root directory is
862recommended. It takes in a pointer to a :c:type:`struct
863fscrypt_get_key_status_arg`, defined as follows::
864
865 struct fscrypt_get_key_status_arg {
866 /* input */
867 struct fscrypt_key_specifier key_spec;
868 __u32 __reserved[6];
869
870 /* output */
871 #define FSCRYPT_KEY_STATUS_ABSENT 1
872 #define FSCRYPT_KEY_STATUS_PRESENT 2
873 #define FSCRYPT_KEY_STATUS_INCOMPLETELY_REMOVED 3
874 __u32 status;
875 #define FSCRYPT_KEY_STATUS_FLAG_ADDED_BY_SELF 0x00000001
876 __u32 status_flags;
877 __u32 user_count;
878 __u32 __out_reserved[13];
879 };
880
881The caller must zero all input fields, then fill in ``key_spec``:
882
883 - To get the status of a key for v1 encryption policies, set
884 ``key_spec.type`` to FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR and fill
885 in ``key_spec.u.descriptor``.
886
887 - To get the status of a key for v2 encryption policies, set
888 ``key_spec.type`` to FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER and fill
889 in ``key_spec.u.identifier``.
890
891On success, 0 is returned and the kernel fills in the output fields:
892
893- ``status`` indicates whether the key is absent, present, or
894 incompletely removed. Incompletely removed means that the master
895 secret has been removed, but some files are still in use; i.e.,
896 `FS_IOC_REMOVE_ENCRYPTION_KEY`_ returned 0 but set the informational
897 status flag FSCRYPT_KEY_REMOVAL_STATUS_FLAG_FILES_BUSY.
898
899- ``status_flags`` can contain the following flags:
900
901 - ``FSCRYPT_KEY_STATUS_FLAG_ADDED_BY_SELF`` indicates that the key
902 has added by the current user. This is only set for keys
903 identified by ``identifier`` rather than by ``descriptor``.
904
905- ``user_count`` specifies the number of users who have added the key.
906 This is only set for keys identified by ``identifier`` rather than
907 by ``descriptor``.
908
909FS_IOC_GET_ENCRYPTION_KEY_STATUS can fail with the following errors:
910
911- ``EINVAL``: invalid key specifier type, or reserved bits were set
912- ``ENOTTY``: this type of filesystem does not implement encryption
913- ``EOPNOTSUPP``: the kernel was not configured with encryption
914 support for this filesystem, or the filesystem superblock has not
915 had encryption enabled on it
916
917Among other use cases, FS_IOC_GET_ENCRYPTION_KEY_STATUS can be useful
918for determining whether the key for a given encrypted directory needs
919to be added before prompting the user for the passphrase needed to
920derive the key.
921
922FS_IOC_GET_ENCRYPTION_KEY_STATUS can only get the status of keys in
923the filesystem-level keyring, i.e. the keyring managed by
924`FS_IOC_ADD_ENCRYPTION_KEY`_ and `FS_IOC_REMOVE_ENCRYPTION_KEY`_. It
925cannot get the status of a key that has only been added for use by v1
926encryption policies using the legacy mechanism involving
927process-subscribed keyrings.
928
929Access semantics
930================
931
932With the key
933------------
934
935With the encryption key, encrypted regular files, directories, and
936symlinks behave very similarly to their unencrypted counterparts ---
937after all, the encryption is intended to be transparent. However,
938astute users may notice some differences in behavior:
939
940- Unencrypted files, or files encrypted with a different encryption
941 policy (i.e. different key, modes, or flags), cannot be renamed or
942 linked into an encrypted directory; see `Encryption policy
943 enforcement`_. Attempts to do so will fail with EXDEV. However,
944 encrypted files can be renamed within an encrypted directory, or
945 into an unencrypted directory.
946
947 Note: "moving" an unencrypted file into an encrypted directory, e.g.
948 with the `mv` program, is implemented in userspace by a copy
949 followed by a delete. Be aware that the original unencrypted data
950 may remain recoverable from free space on the disk; prefer to keep
951 all files encrypted from the very beginning. The `shred` program
952 may be used to overwrite the source files but isn't guaranteed to be
953 effective on all filesystems and storage devices.
954
955- Direct I/O is not supported on encrypted files. Attempts to use
956 direct I/O on such files will fall back to buffered I/O.
957
958- The fallocate operations FALLOC_FL_COLLAPSE_RANGE,
959 FALLOC_FL_INSERT_RANGE, and FALLOC_FL_ZERO_RANGE are not supported
960 on encrypted files and will fail with EOPNOTSUPP.
961
962- Online defragmentation of encrypted files is not supported. The
963 EXT4_IOC_MOVE_EXT and F2FS_IOC_MOVE_RANGE ioctls will fail with
964 EOPNOTSUPP.
965
966- The ext4 filesystem does not support data journaling with encrypted
967 regular files. It will fall back to ordered data mode instead.
968
969- DAX (Direct Access) is not supported on encrypted files.
970
971- The st_size of an encrypted symlink will not necessarily give the
972 length of the symlink target as required by POSIX. It will actually
973 give the length of the ciphertext, which will be slightly longer
974 than the plaintext due to NUL-padding and an extra 2-byte overhead.
975
976- The maximum length of an encrypted symlink is 2 bytes shorter than
977 the maximum length of an unencrypted symlink. For example, on an
978 EXT4 filesystem with a 4K block size, unencrypted symlinks can be up
979 to 4095 bytes long, while encrypted symlinks can only be up to 4093
980 bytes long (both lengths excluding the terminating null).
981
982Note that mmap *is* supported. This is possible because the pagecache
983for an encrypted file contains the plaintext, not the ciphertext.
984
985Without the key
986---------------
987
988Some filesystem operations may be performed on encrypted regular
989files, directories, and symlinks even before their encryption key has
990been added, or after their encryption key has been removed:
991
992- File metadata may be read, e.g. using stat().
993
994- Directories may be listed, in which case the filenames will be
995 listed in an encoded form derived from their ciphertext. The
996 current encoding algorithm is described in `Filename hashing and
997 encoding`_. The algorithm is subject to change, but it is
998 guaranteed that the presented filenames will be no longer than
999 NAME_MAX bytes, will not contain the ``/`` or ``\0`` characters, and
1000 will uniquely identify directory entries.
1001
1002 The ``.`` and ``..`` directory entries are special. They are always
1003 present and are not encrypted or encoded.
1004
1005- Files may be deleted. That is, nondirectory files may be deleted
1006 with unlink() as usual, and empty directories may be deleted with
1007 rmdir() as usual. Therefore, ``rm`` and ``rm -r`` will work as
1008 expected.
1009
1010- Symlink targets may be read and followed, but they will be presented
1011 in encrypted form, similar to filenames in directories. Hence, they
1012 are unlikely to point to anywhere useful.
1013
1014Without the key, regular files cannot be opened or truncated.
1015Attempts to do so will fail with ENOKEY. This implies that any
1016regular file operations that require a file descriptor, such as
1017read(), write(), mmap(), fallocate(), and ioctl(), are also forbidden.
1018
1019Also without the key, files of any type (including directories) cannot
1020be created or linked into an encrypted directory, nor can a name in an
1021encrypted directory be the source or target of a rename, nor can an
1022O_TMPFILE temporary file be created in an encrypted directory. All
1023such operations will fail with ENOKEY.
1024
1025It is not currently possible to backup and restore encrypted files
1026without the encryption key. This would require special APIs which
1027have not yet been implemented.
1028
1029Encryption policy enforcement
1030=============================
1031
1032After an encryption policy has been set on a directory, all regular
1033files, directories, and symbolic links created in that directory
1034(recursively) will inherit that encryption policy. Special files ---
1035that is, named pipes, device nodes, and UNIX domain sockets --- will
1036not be encrypted.
1037
1038Except for those special files, it is forbidden to have unencrypted
1039files, or files encrypted with a different encryption policy, in an
1040encrypted directory tree. Attempts to link or rename such a file into
1041an encrypted directory will fail with EXDEV. This is also enforced
1042during ->lookup() to provide limited protection against offline
1043attacks that try to disable or downgrade encryption in known locations
1044where applications may later write sensitive data. It is recommended
1045that systems implementing a form of "verified boot" take advantage of
1046this by validating all top-level encryption policies prior to access.
1047
1048Implementation details
1049======================
1050
1051Encryption context
1052------------------
1053
1054An encryption policy is represented on-disk by a :c:type:`struct
1055fscrypt_context_v1` or a :c:type:`struct fscrypt_context_v2`. It is
1056up to individual filesystems to decide where to store it, but normally
1057it would be stored in a hidden extended attribute. It should *not* be
1058exposed by the xattr-related system calls such as getxattr() and
1059setxattr() because of the special semantics of the encryption xattr.
1060(In particular, there would be much confusion if an encryption policy
1061were to be added to or removed from anything other than an empty
1062directory.) These structs are defined as follows::
1063
1064 #define FS_KEY_DERIVATION_NONCE_SIZE 16
1065
1066 #define FSCRYPT_KEY_DESCRIPTOR_SIZE 8
1067 struct fscrypt_context_v1 {
1068 u8 version;
1069 u8 contents_encryption_mode;
1070 u8 filenames_encryption_mode;
1071 u8 flags;
1072 u8 master_key_descriptor[FSCRYPT_KEY_DESCRIPTOR_SIZE];
1073 u8 nonce[FS_KEY_DERIVATION_NONCE_SIZE];
1074 };
1075
1076 #define FSCRYPT_KEY_IDENTIFIER_SIZE 16
1077 struct fscrypt_context_v2 {
1078 u8 version;
1079 u8 contents_encryption_mode;
1080 u8 filenames_encryption_mode;
1081 u8 flags;
1082 u8 __reserved[4];
1083 u8 master_key_identifier[FSCRYPT_KEY_IDENTIFIER_SIZE];
1084 u8 nonce[FS_KEY_DERIVATION_NONCE_SIZE];
1085 };
1086
1087The context structs contain the same information as the corresponding
1088policy structs (see `Setting an encryption policy`_), except that the
1089context structs also contain a nonce. The nonce is randomly generated
1090by the kernel and is used as KDF input or as a tweak to cause
1091different files to be encrypted differently; see `Per-file keys`_ and
1092`DIRECT_KEY and per-mode keys`_.
1093
1094Data path changes
1095-----------------
1096
1097For the read path (->readpage()) of regular files, filesystems can
1098read the ciphertext into the page cache and decrypt it in-place. The
1099page lock must be held until decryption has finished, to prevent the
1100page from becoming visible to userspace prematurely.
1101
1102For the write path (->writepage()) of regular files, filesystems
1103cannot encrypt data in-place in the page cache, since the cached
1104plaintext must be preserved. Instead, filesystems must encrypt into a
1105temporary buffer or "bounce page", then write out the temporary
1106buffer. Some filesystems, such as UBIFS, already use temporary
1107buffers regardless of encryption. Other filesystems, such as ext4 and
1108F2FS, have to allocate bounce pages specially for encryption.
1109
1110Filename hashing and encoding
1111-----------------------------
1112
1113Modern filesystems accelerate directory lookups by using indexed
1114directories. An indexed directory is organized as a tree keyed by
1115filename hashes. When a ->lookup() is requested, the filesystem
1116normally hashes the filename being looked up so that it can quickly
1117find the corresponding directory entry, if any.
1118
1119With encryption, lookups must be supported and efficient both with and
1120without the encryption key. Clearly, it would not work to hash the
1121plaintext filenames, since the plaintext filenames are unavailable
1122without the key. (Hashing the plaintext filenames would also make it
1123impossible for the filesystem's fsck tool to optimize encrypted
1124directories.) Instead, filesystems hash the ciphertext filenames,
1125i.e. the bytes actually stored on-disk in the directory entries. When
1126asked to do a ->lookup() with the key, the filesystem just encrypts
1127the user-supplied name to get the ciphertext.
1128
1129Lookups without the key are more complicated. The raw ciphertext may
1130contain the ``\0`` and ``/`` characters, which are illegal in
1131filenames. Therefore, readdir() must base64-encode the ciphertext for
1132presentation. For most filenames, this works fine; on ->lookup(), the
1133filesystem just base64-decodes the user-supplied name to get back to
1134the raw ciphertext.
1135
1136However, for very long filenames, base64 encoding would cause the
1137filename length to exceed NAME_MAX. To prevent this, readdir()
1138actually presents long filenames in an abbreviated form which encodes
1139a strong "hash" of the ciphertext filename, along with the optional
1140filesystem-specific hash(es) needed for directory lookups. This
1141allows the filesystem to still, with a high degree of confidence, map
1142the filename given in ->lookup() back to a particular directory entry
1143that was previously listed by readdir(). See :c:type:`struct
1144fscrypt_digested_name` in the source for more details.
1145
1146Note that the precise way that filenames are presented to userspace
1147without the key is subject to change in the future. It is only meant
1148as a way to temporarily present valid filenames so that commands like
1149``rm -r`` work as expected on encrypted directories.
1150
1151Tests
1152=====
1153
1154To test fscrypt, use xfstests, which is Linux's de facto standard
1155filesystem test suite. First, run all the tests in the "encrypt"
1156group on the relevant filesystem(s). For example, to test ext4 and
1157f2fs encryption using `kvm-xfstests
1158<https://github.com/tytso/xfstests-bld/blob/master/Documentation/kvm-quickstart.md>`_::
1159
1160 kvm-xfstests -c ext4,f2fs -g encrypt
1161
1162UBIFS encryption can also be tested this way, but it should be done in
1163a separate command, and it takes some time for kvm-xfstests to set up
1164emulated UBI volumes::
1165
1166 kvm-xfstests -c ubifs -g encrypt
1167
1168No tests should fail. However, tests that use non-default encryption
1169modes (e.g. generic/549 and generic/550) will be skipped if the needed
1170algorithms were not built into the kernel's crypto API. Also, tests
1171that access the raw block device (e.g. generic/399, generic/548,
1172generic/549, generic/550) will be skipped on UBIFS.
1173
1174Besides running the "encrypt" group tests, for ext4 and f2fs it's also
1175possible to run most xfstests with the "test_dummy_encryption" mount
1176option. This option causes all new files to be automatically
1177encrypted with a dummy key, without having to make any API calls.
1178This tests the encrypted I/O paths more thoroughly. To do this with
1179kvm-xfstests, use the "encrypt" filesystem configuration::
1180
1181 kvm-xfstests -c ext4/encrypt,f2fs/encrypt -g auto
1182
1183Because this runs many more tests than "-g encrypt" does, it takes
1184much longer to run; so also consider using `gce-xfstests
1185<https://github.com/tytso/xfstests-bld/blob/master/Documentation/gce-xfstests.md>`_
1186instead of kvm-xfstests::
1187
1188 gce-xfstests -c ext4/encrypt,f2fs/encrypt -g auto