xref: /linux/Documentation/filesystems/fscrypt.rst (revision 71dfa617ea9f18e4585fe78364217cd32b1fc382)
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, UBIFS,
35and CephFS.  This allows encrypted files to be read and written
36without caching both the decrypted and encrypted pages in the
37pagecache, thereby nearly halving the memory used and bringing it in
38line with unencrypted files.  Similarly, half as many dentries and
39inodes are needed.  eCryptfs also limits encrypted filenames to 143
40bytes, causing application compatibility issues; fscrypt allows the
41full 255 bytes (NAME_MAX).  Finally, unlike eCryptfs, the fscrypt API
42can be used 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 or inline encryption hardware are.  If a
81vulnerable algorithm is used, such as a table-based implementation of
82AES, it may be possible for an attacker to mount a side channel attack
83against the online system.  Side channel attacks may also be mounted
84against applications consuming 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 security strength of the contents and filenames
180encryption modes being used.  For example, if any AES-256 mode is
181used, the master key must be at least 256 bits, i.e. 32 bytes.  A
182stricter requirement applies if the key is used by a v1 encryption
183policy and AES-256-XTS is used; such keys must be 64 bytes.
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 encryption 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
259DIRECT_KEY policies
260-------------------
261
262The Adiantum encryption mode (see `Encryption modes and usage`_) is
263suitable for both contents and filenames encryption, and it accepts
264long IVs --- long enough to hold both an 8-byte data unit index and a
26516-byte per-file nonce.  Also, the overhead of each Adiantum key is
266greater than that of an AES-256-XTS key.
267
268Therefore, to improve performance and save memory, for Adiantum a
269"direct key" configuration is supported.  When the user has enabled
270this by setting FSCRYPT_POLICY_FLAG_DIRECT_KEY in the fscrypt policy,
271per-file encryption keys are not used.  Instead, whenever any data
272(contents or filenames) is encrypted, the file's 16-byte nonce is
273included in the IV.  Moreover:
274
275- For v1 encryption policies, the encryption is done directly with the
276  master key.  Because of this, users **must not** use the same master
277  key for any other purpose, even for other v1 policies.
278
279- For v2 encryption policies, the encryption is done with a per-mode
280  key derived using the KDF.  Users may use the same master key for
281  other v2 encryption policies.
282
283IV_INO_LBLK_64 policies
284-----------------------
285
286When FSCRYPT_POLICY_FLAG_IV_INO_LBLK_64 is set in the fscrypt policy,
287the encryption keys are derived from the master key, encryption mode
288number, and filesystem UUID.  This normally results in all files
289protected by the same master key sharing a single contents encryption
290key and a single filenames encryption key.  To still encrypt different
291files' data differently, inode numbers are included in the IVs.
292Consequently, shrinking the filesystem may not be allowed.
293
294This format is optimized for use with inline encryption hardware
295compliant with the UFS standard, which supports only 64 IV bits per
296I/O request and may have only a small number of keyslots.
297
298IV_INO_LBLK_32 policies
299-----------------------
300
301IV_INO_LBLK_32 policies work like IV_INO_LBLK_64, except that for
302IV_INO_LBLK_32, the inode number is hashed with SipHash-2-4 (where the
303SipHash key is derived from the master key) and added to the file data
304unit index mod 2^32 to produce a 32-bit IV.
305
306This format is optimized for use with inline encryption hardware
307compliant with the eMMC v5.2 standard, which supports only 32 IV bits
308per I/O request and may have only a small number of keyslots.  This
309format results in some level of IV reuse, so it should only be used
310when necessary due to hardware limitations.
311
312Key identifiers
313---------------
314
315For master keys used for v2 encryption policies, a unique 16-byte "key
316identifier" is also derived using the KDF.  This value is stored in
317the clear, since it is needed to reliably identify the key itself.
318
319Dirhash keys
320------------
321
322For directories that are indexed using a secret-keyed dirhash over the
323plaintext filenames, the KDF is also used to derive a 128-bit
324SipHash-2-4 key per directory in order to hash filenames.  This works
325just like deriving a per-file encryption key, except that a different
326KDF context is used.  Currently, only casefolded ("case-insensitive")
327encrypted directories use this style of hashing.
328
329Encryption modes and usage
330==========================
331
332fscrypt allows one encryption mode to be specified for file contents
333and one encryption mode to be specified for filenames.  Different
334directory trees are permitted to use different encryption modes.
335
336Supported modes
337---------------
338
339Currently, the following pairs of encryption modes are supported:
340
341- AES-256-XTS for contents and AES-256-CBC-CTS for filenames
342- AES-256-XTS for contents and AES-256-HCTR2 for filenames
343- Adiantum for both contents and filenames
344- AES-128-CBC-ESSIV for contents and AES-128-CBC-CTS for filenames
345- SM4-XTS for contents and SM4-CBC-CTS for filenames
346
347Note: in the API, "CBC" means CBC-ESSIV, and "CTS" means CBC-CTS.
348So, for example, FSCRYPT_MODE_AES_256_CTS means AES-256-CBC-CTS.
349
350Authenticated encryption modes are not currently supported because of
351the difficulty of dealing with ciphertext expansion.  Therefore,
352contents encryption uses a block cipher in `XTS mode
353<https://en.wikipedia.org/wiki/Disk_encryption_theory#XTS>`_ or
354`CBC-ESSIV mode
355<https://en.wikipedia.org/wiki/Disk_encryption_theory#Encrypted_salt-sector_initialization_vector_(ESSIV)>`_,
356or a wide-block cipher.  Filenames encryption uses a
357block cipher in `CBC-CTS mode
358<https://en.wikipedia.org/wiki/Ciphertext_stealing>`_ or a wide-block
359cipher.
360
361The (AES-256-XTS, AES-256-CBC-CTS) pair is the recommended default.
362It is also the only option that is *guaranteed* to always be supported
363if the kernel supports fscrypt at all; see `Kernel config options`_.
364
365The (AES-256-XTS, AES-256-HCTR2) pair is also a good choice that
366upgrades the filenames encryption to use a wide-block cipher.  (A
367*wide-block cipher*, also called a tweakable super-pseudorandom
368permutation, has the property that changing one bit scrambles the
369entire result.)  As described in `Filenames encryption`_, a wide-block
370cipher is the ideal mode for the problem domain, though CBC-CTS is the
371"least bad" choice among the alternatives.  For more information about
372HCTR2, see `the HCTR2 paper <https://eprint.iacr.org/2021/1441.pdf>`_.
373
374Adiantum is recommended on systems where AES is too slow due to lack
375of hardware acceleration for AES.  Adiantum is a wide-block cipher
376that uses XChaCha12 and AES-256 as its underlying components.  Most of
377the work is done by XChaCha12, which is much faster than AES when AES
378acceleration is unavailable.  For more information about Adiantum, see
379`the Adiantum paper <https://eprint.iacr.org/2018/720.pdf>`_.
380
381The (AES-128-CBC-ESSIV, AES-128-CBC-CTS) pair exists only to support
382systems whose only form of AES acceleration is an off-CPU crypto
383accelerator such as CAAM or CESA that does not support XTS.
384
385The remaining mode pairs are the "national pride ciphers":
386
387- (SM4-XTS, SM4-CBC-CTS)
388
389Generally speaking, these ciphers aren't "bad" per se, but they
390receive limited security review compared to the usual choices such as
391AES and ChaCha.  They also don't bring much new to the table.  It is
392suggested to only use these ciphers where their use is mandated.
393
394Kernel config options
395---------------------
396
397Enabling fscrypt support (CONFIG_FS_ENCRYPTION) automatically pulls in
398only the basic support from the crypto API needed to use AES-256-XTS
399and AES-256-CBC-CTS encryption.  For optimal performance, it is
400strongly recommended to also enable any available platform-specific
401kconfig options that provide acceleration for the algorithm(s) you
402wish to use.  Support for any "non-default" encryption modes typically
403requires extra kconfig options as well.
404
405Below, some relevant options are listed by encryption mode.  Note,
406acceleration options not listed below may be available for your
407platform; refer to the kconfig menus.  File contents encryption can
408also be configured to use inline encryption hardware instead of the
409kernel crypto API (see `Inline encryption support`_); in that case,
410the file contents mode doesn't need to supported in the kernel crypto
411API, but the filenames mode still does.
412
413- AES-256-XTS and AES-256-CBC-CTS
414    - Recommended:
415        - arm64: CONFIG_CRYPTO_AES_ARM64_CE_BLK
416        - x86: CONFIG_CRYPTO_AES_NI_INTEL
417
418- AES-256-HCTR2
419    - Mandatory:
420        - CONFIG_CRYPTO_HCTR2
421    - Recommended:
422        - arm64: CONFIG_CRYPTO_AES_ARM64_CE_BLK
423        - arm64: CONFIG_CRYPTO_POLYVAL_ARM64_CE
424        - x86: CONFIG_CRYPTO_AES_NI_INTEL
425        - x86: CONFIG_CRYPTO_POLYVAL_CLMUL_NI
426
427- Adiantum
428    - Mandatory:
429        - CONFIG_CRYPTO_ADIANTUM
430    - Recommended:
431        - arm32: CONFIG_CRYPTO_CHACHA20_NEON
432        - arm32: CONFIG_CRYPTO_NHPOLY1305_NEON
433        - arm64: CONFIG_CRYPTO_CHACHA20_NEON
434        - arm64: CONFIG_CRYPTO_NHPOLY1305_NEON
435        - x86: CONFIG_CRYPTO_CHACHA20_X86_64
436        - x86: CONFIG_CRYPTO_NHPOLY1305_SSE2
437        - x86: CONFIG_CRYPTO_NHPOLY1305_AVX2
438
439- AES-128-CBC-ESSIV and AES-128-CBC-CTS:
440    - Mandatory:
441        - CONFIG_CRYPTO_ESSIV
442        - CONFIG_CRYPTO_SHA256 or another SHA-256 implementation
443    - Recommended:
444        - AES-CBC acceleration
445
446fscrypt also uses HMAC-SHA512 for key derivation, so enabling SHA-512
447acceleration is recommended:
448
449- SHA-512
450    - Recommended:
451        - arm64: CONFIG_CRYPTO_SHA512_ARM64_CE
452        - x86: CONFIG_CRYPTO_SHA512_SSSE3
453
454Contents encryption
455-------------------
456
457For contents encryption, each file's contents is divided into "data
458units".  Each data unit is encrypted independently.  The IV for each
459data unit incorporates the zero-based index of the data unit within
460the file.  This ensures that each data unit within a file is encrypted
461differently, which is essential to prevent leaking information.
462
463Note: the encryption depending on the offset into the file means that
464operations like "collapse range" and "insert range" that rearrange the
465extent mapping of files are not supported on encrypted files.
466
467There are two cases for the sizes of the data units:
468
469* Fixed-size data units.  This is how all filesystems other than UBIFS
470  work.  A file's data units are all the same size; the last data unit
471  is zero-padded if needed.  By default, the data unit size is equal
472  to the filesystem block size.  On some filesystems, users can select
473  a sub-block data unit size via the ``log2_data_unit_size`` field of
474  the encryption policy; see `FS_IOC_SET_ENCRYPTION_POLICY`_.
475
476* Variable-size data units.  This is what UBIFS does.  Each "UBIFS
477  data node" is treated as a crypto data unit.  Each contains variable
478  length, possibly compressed data, zero-padded to the next 16-byte
479  boundary.  Users cannot select a sub-block data unit size on UBIFS.
480
481In the case of compression + encryption, the compressed data is
482encrypted.  UBIFS compression works as described above.  f2fs
483compression works a bit differently; it compresses a number of
484filesystem blocks into a smaller number of filesystem blocks.
485Therefore a f2fs-compressed file still uses fixed-size data units, and
486it is encrypted in a similar way to a file containing holes.
487
488As mentioned in `Key hierarchy`_, the default encryption setting uses
489per-file keys.  In this case, the IV for each data unit is simply the
490index of the data unit in the file.  However, users can select an
491encryption setting that does not use per-file keys.  For these, some
492kind of file identifier is incorporated into the IVs as follows:
493
494- With `DIRECT_KEY policies`_, the data unit index is placed in bits
495  0-63 of the IV, and the file's nonce is placed in bits 64-191.
496
497- With `IV_INO_LBLK_64 policies`_, the data unit index is placed in
498  bits 0-31 of the IV, and the file's inode number is placed in bits
499  32-63.  This setting is only allowed when data unit indices and
500  inode numbers fit in 32 bits.
501
502- With `IV_INO_LBLK_32 policies`_, the file's inode number is hashed
503  and added to the data unit index.  The resulting value is truncated
504  to 32 bits and placed in bits 0-31 of the IV.  This setting is only
505  allowed when data unit indices and inode numbers fit in 32 bits.
506
507The byte order of the IV is always little endian.
508
509If the user selects FSCRYPT_MODE_AES_128_CBC for the contents mode, an
510ESSIV layer is automatically included.  In this case, before the IV is
511passed to AES-128-CBC, it is encrypted with AES-256 where the AES-256
512key is the SHA-256 hash of the file's contents encryption key.
513
514Filenames encryption
515--------------------
516
517For filenames, each full filename is encrypted at once.  Because of
518the requirements to retain support for efficient directory lookups and
519filenames of up to 255 bytes, the same IV is used for every filename
520in a directory.
521
522However, each encrypted directory still uses a unique key, or
523alternatively has the file's nonce (for `DIRECT_KEY policies`_) or
524inode number (for `IV_INO_LBLK_64 policies`_) included in the IVs.
525Thus, IV reuse is limited to within a single directory.
526
527With CBC-CTS, the IV reuse means that when the plaintext filenames share a
528common prefix at least as long as the cipher block size (16 bytes for AES), the
529corresponding encrypted filenames will also share a common prefix.  This is
530undesirable.  Adiantum and HCTR2 do not have this weakness, as they are
531wide-block encryption modes.
532
533All supported filenames encryption modes accept any plaintext length
534>= 16 bytes; cipher block alignment is not required.  However,
535filenames shorter than 16 bytes are NUL-padded to 16 bytes before
536being encrypted.  In addition, to reduce leakage of filename lengths
537via their ciphertexts, all filenames are NUL-padded to the next 4, 8,
53816, or 32-byte boundary (configurable).  32 is recommended since this
539provides the best confidentiality, at the cost of making directory
540entries consume slightly more space.  Note that since NUL (``\0``) is
541not otherwise a valid character in filenames, the padding will never
542produce duplicate plaintexts.
543
544Symbolic link targets are considered a type of filename and are
545encrypted in the same way as filenames in directory entries, except
546that IV reuse is not a problem as each symlink has its own inode.
547
548User API
549========
550
551Setting an encryption policy
552----------------------------
553
554FS_IOC_SET_ENCRYPTION_POLICY
555~~~~~~~~~~~~~~~~~~~~~~~~~~~~
556
557The FS_IOC_SET_ENCRYPTION_POLICY ioctl sets an encryption policy on an
558empty directory or verifies that a directory or regular file already
559has the specified encryption policy.  It takes in a pointer to
560struct fscrypt_policy_v1 or struct fscrypt_policy_v2, defined as
561follows::
562
563    #define FSCRYPT_POLICY_V1               0
564    #define FSCRYPT_KEY_DESCRIPTOR_SIZE     8
565    struct fscrypt_policy_v1 {
566            __u8 version;
567            __u8 contents_encryption_mode;
568            __u8 filenames_encryption_mode;
569            __u8 flags;
570            __u8 master_key_descriptor[FSCRYPT_KEY_DESCRIPTOR_SIZE];
571    };
572    #define fscrypt_policy  fscrypt_policy_v1
573
574    #define FSCRYPT_POLICY_V2               2
575    #define FSCRYPT_KEY_IDENTIFIER_SIZE     16
576    struct fscrypt_policy_v2 {
577            __u8 version;
578            __u8 contents_encryption_mode;
579            __u8 filenames_encryption_mode;
580            __u8 flags;
581            __u8 log2_data_unit_size;
582            __u8 __reserved[3];
583            __u8 master_key_identifier[FSCRYPT_KEY_IDENTIFIER_SIZE];
584    };
585
586This structure must be initialized as follows:
587
588- ``version`` must be FSCRYPT_POLICY_V1 (0) if
589  struct fscrypt_policy_v1 is used or FSCRYPT_POLICY_V2 (2) if
590  struct fscrypt_policy_v2 is used. (Note: we refer to the original
591  policy version as "v1", though its version code is really 0.)
592  For new encrypted directories, use v2 policies.
593
594- ``contents_encryption_mode`` and ``filenames_encryption_mode`` must
595  be set to constants from ``<linux/fscrypt.h>`` which identify the
596  encryption modes to use.  If unsure, use FSCRYPT_MODE_AES_256_XTS
597  (1) for ``contents_encryption_mode`` and FSCRYPT_MODE_AES_256_CTS
598  (4) for ``filenames_encryption_mode``.  For details, see `Encryption
599  modes and usage`_.
600
601  v1 encryption policies only support three combinations of modes:
602  (FSCRYPT_MODE_AES_256_XTS, FSCRYPT_MODE_AES_256_CTS),
603  (FSCRYPT_MODE_AES_128_CBC, FSCRYPT_MODE_AES_128_CTS), and
604  (FSCRYPT_MODE_ADIANTUM, FSCRYPT_MODE_ADIANTUM).  v2 policies support
605  all combinations documented in `Supported modes`_.
606
607- ``flags`` contains optional flags from ``<linux/fscrypt.h>``:
608
609  - FSCRYPT_POLICY_FLAGS_PAD_*: The amount of NUL padding to use when
610    encrypting filenames.  If unsure, use FSCRYPT_POLICY_FLAGS_PAD_32
611    (0x3).
612  - FSCRYPT_POLICY_FLAG_DIRECT_KEY: See `DIRECT_KEY policies`_.
613  - FSCRYPT_POLICY_FLAG_IV_INO_LBLK_64: See `IV_INO_LBLK_64
614    policies`_.
615  - FSCRYPT_POLICY_FLAG_IV_INO_LBLK_32: See `IV_INO_LBLK_32
616    policies`_.
617
618  v1 encryption policies only support the PAD_* and DIRECT_KEY flags.
619  The other flags are only supported by v2 encryption policies.
620
621  The DIRECT_KEY, IV_INO_LBLK_64, and IV_INO_LBLK_32 flags are
622  mutually exclusive.
623
624- ``log2_data_unit_size`` is the log2 of the data unit size in bytes,
625  or 0 to select the default data unit size.  The data unit size is
626  the granularity of file contents encryption.  For example, setting
627  ``log2_data_unit_size`` to 12 causes file contents be passed to the
628  underlying encryption algorithm (such as AES-256-XTS) in 4096-byte
629  data units, each with its own IV.
630
631  Not all filesystems support setting ``log2_data_unit_size``.  ext4
632  and f2fs support it since Linux v6.7.  On filesystems that support
633  it, the supported nonzero values are 9 through the log2 of the
634  filesystem block size, inclusively.  The default value of 0 selects
635  the filesystem block size.
636
637  The main use case for ``log2_data_unit_size`` is for selecting a
638  data unit size smaller than the filesystem block size for
639  compatibility with inline encryption hardware that only supports
640  smaller data unit sizes.  ``/sys/block/$disk/queue/crypto/`` may be
641  useful for checking which data unit sizes are supported by a
642  particular system's inline encryption hardware.
643
644  Leave this field zeroed unless you are certain you need it.  Using
645  an unnecessarily small data unit size reduces performance.
646
647- For v2 encryption policies, ``__reserved`` must be zeroed.
648
649- For v1 encryption policies, ``master_key_descriptor`` specifies how
650  to find the master key in a keyring; see `Adding keys`_.  It is up
651  to userspace to choose a unique ``master_key_descriptor`` for each
652  master key.  The e4crypt and fscrypt tools use the first 8 bytes of
653  ``SHA-512(SHA-512(master_key))``, but this particular scheme is not
654  required.  Also, the master key need not be in the keyring yet when
655  FS_IOC_SET_ENCRYPTION_POLICY is executed.  However, it must be added
656  before any files can be created in the encrypted directory.
657
658  For v2 encryption policies, ``master_key_descriptor`` has been
659  replaced with ``master_key_identifier``, which is longer and cannot
660  be arbitrarily chosen.  Instead, the key must first be added using
661  `FS_IOC_ADD_ENCRYPTION_KEY`_.  Then, the ``key_spec.u.identifier``
662  the kernel returned in the struct fscrypt_add_key_arg must
663  be used as the ``master_key_identifier`` in
664  struct fscrypt_policy_v2.
665
666If the file is not yet encrypted, then FS_IOC_SET_ENCRYPTION_POLICY
667verifies that the file is an empty directory.  If so, the specified
668encryption policy is assigned to the directory, turning it into an
669encrypted directory.  After that, and after providing the
670corresponding master key as described in `Adding keys`_, all regular
671files, directories (recursively), and symlinks created in the
672directory will be encrypted, inheriting the same encryption policy.
673The filenames in the directory's entries will be encrypted as well.
674
675Alternatively, if the file is already encrypted, then
676FS_IOC_SET_ENCRYPTION_POLICY validates that the specified encryption
677policy exactly matches the actual one.  If they match, then the ioctl
678returns 0.  Otherwise, it fails with EEXIST.  This works on both
679regular files and directories, including nonempty directories.
680
681When a v2 encryption policy is assigned to a directory, it is also
682required that either the specified key has been added by the current
683user or that the caller has CAP_FOWNER in the initial user namespace.
684(This is needed to prevent a user from encrypting their data with
685another user's key.)  The key must remain added while
686FS_IOC_SET_ENCRYPTION_POLICY is executing.  However, if the new
687encrypted directory does not need to be accessed immediately, then the
688key can be removed right away afterwards.
689
690Note that the ext4 filesystem does not allow the root directory to be
691encrypted, even if it is empty.  Users who want to encrypt an entire
692filesystem with one key should consider using dm-crypt instead.
693
694FS_IOC_SET_ENCRYPTION_POLICY can fail with the following errors:
695
696- ``EACCES``: the file is not owned by the process's uid, nor does the
697  process have the CAP_FOWNER capability in a namespace with the file
698  owner's uid mapped
699- ``EEXIST``: the file is already encrypted with an encryption policy
700  different from the one specified
701- ``EINVAL``: an invalid encryption policy was specified (invalid
702  version, mode(s), or flags; or reserved bits were set); or a v1
703  encryption policy was specified but the directory has the casefold
704  flag enabled (casefolding is incompatible with v1 policies).
705- ``ENOKEY``: a v2 encryption policy was specified, but the key with
706  the specified ``master_key_identifier`` has not been added, nor does
707  the process have the CAP_FOWNER capability in the initial user
708  namespace
709- ``ENOTDIR``: the file is unencrypted and is a regular file, not a
710  directory
711- ``ENOTEMPTY``: the file is unencrypted and is a nonempty directory
712- ``ENOTTY``: this type of filesystem does not implement encryption
713- ``EOPNOTSUPP``: the kernel was not configured with encryption
714  support for filesystems, or the filesystem superblock has not
715  had encryption enabled on it.  (For example, to use encryption on an
716  ext4 filesystem, CONFIG_FS_ENCRYPTION must be enabled in the
717  kernel config, and the superblock must have had the "encrypt"
718  feature flag enabled using ``tune2fs -O encrypt`` or ``mkfs.ext4 -O
719  encrypt``.)
720- ``EPERM``: this directory may not be encrypted, e.g. because it is
721  the root directory of an ext4 filesystem
722- ``EROFS``: the filesystem is readonly
723
724Getting an encryption policy
725----------------------------
726
727Two ioctls are available to get a file's encryption policy:
728
729- `FS_IOC_GET_ENCRYPTION_POLICY_EX`_
730- `FS_IOC_GET_ENCRYPTION_POLICY`_
731
732The extended (_EX) version of the ioctl is more general and is
733recommended to use when possible.  However, on older kernels only the
734original ioctl is available.  Applications should try the extended
735version, and if it fails with ENOTTY fall back to the original
736version.
737
738FS_IOC_GET_ENCRYPTION_POLICY_EX
739~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
740
741The FS_IOC_GET_ENCRYPTION_POLICY_EX ioctl retrieves the encryption
742policy, if any, for a directory or regular file.  No additional
743permissions are required beyond the ability to open the file.  It
744takes in a pointer to struct fscrypt_get_policy_ex_arg,
745defined as follows::
746
747    struct fscrypt_get_policy_ex_arg {
748            __u64 policy_size; /* input/output */
749            union {
750                    __u8 version;
751                    struct fscrypt_policy_v1 v1;
752                    struct fscrypt_policy_v2 v2;
753            } policy; /* output */
754    };
755
756The caller must initialize ``policy_size`` to the size available for
757the policy struct, i.e. ``sizeof(arg.policy)``.
758
759On success, the policy struct is returned in ``policy``, and its
760actual size is returned in ``policy_size``.  ``policy.version`` should
761be checked to determine the version of policy returned.  Note that the
762version code for the "v1" policy is actually 0 (FSCRYPT_POLICY_V1).
763
764FS_IOC_GET_ENCRYPTION_POLICY_EX can fail with the following errors:
765
766- ``EINVAL``: the file is encrypted, but it uses an unrecognized
767  encryption policy version
768- ``ENODATA``: the file is not encrypted
769- ``ENOTTY``: this type of filesystem does not implement encryption,
770  or this kernel is too old to support FS_IOC_GET_ENCRYPTION_POLICY_EX
771  (try FS_IOC_GET_ENCRYPTION_POLICY instead)
772- ``EOPNOTSUPP``: the kernel was not configured with encryption
773  support for this filesystem, or the filesystem superblock has not
774  had encryption enabled on it
775- ``EOVERFLOW``: the file is encrypted and uses a recognized
776  encryption policy version, but the policy struct does not fit into
777  the provided buffer
778
779Note: if you only need to know whether a file is encrypted or not, on
780most filesystems it is also possible to use the FS_IOC_GETFLAGS ioctl
781and check for FS_ENCRYPT_FL, or to use the statx() system call and
782check for STATX_ATTR_ENCRYPTED in stx_attributes.
783
784FS_IOC_GET_ENCRYPTION_POLICY
785~~~~~~~~~~~~~~~~~~~~~~~~~~~~
786
787The FS_IOC_GET_ENCRYPTION_POLICY ioctl can also retrieve the
788encryption policy, if any, for a directory or regular file.  However,
789unlike `FS_IOC_GET_ENCRYPTION_POLICY_EX`_,
790FS_IOC_GET_ENCRYPTION_POLICY only supports the original policy
791version.  It takes in a pointer directly to struct fscrypt_policy_v1
792rather than struct fscrypt_get_policy_ex_arg.
793
794The error codes for FS_IOC_GET_ENCRYPTION_POLICY are the same as those
795for FS_IOC_GET_ENCRYPTION_POLICY_EX, except that
796FS_IOC_GET_ENCRYPTION_POLICY also returns ``EINVAL`` if the file is
797encrypted using a newer encryption policy version.
798
799Getting the per-filesystem salt
800-------------------------------
801
802Some filesystems, such as ext4 and F2FS, also support the deprecated
803ioctl FS_IOC_GET_ENCRYPTION_PWSALT.  This ioctl retrieves a randomly
804generated 16-byte value stored in the filesystem superblock.  This
805value is intended to used as a salt when deriving an encryption key
806from a passphrase or other low-entropy user credential.
807
808FS_IOC_GET_ENCRYPTION_PWSALT is deprecated.  Instead, prefer to
809generate and manage any needed salt(s) in userspace.
810
811Getting a file's encryption nonce
812---------------------------------
813
814Since Linux v5.7, the ioctl FS_IOC_GET_ENCRYPTION_NONCE is supported.
815On encrypted files and directories it gets the inode's 16-byte nonce.
816On unencrypted files and directories, it fails with ENODATA.
817
818This ioctl can be useful for automated tests which verify that the
819encryption is being done correctly.  It is not needed for normal use
820of fscrypt.
821
822Adding keys
823-----------
824
825FS_IOC_ADD_ENCRYPTION_KEY
826~~~~~~~~~~~~~~~~~~~~~~~~~
827
828The FS_IOC_ADD_ENCRYPTION_KEY ioctl adds a master encryption key to
829the filesystem, making all files on the filesystem which were
830encrypted using that key appear "unlocked", i.e. in plaintext form.
831It can be executed on any file or directory on the target filesystem,
832but using the filesystem's root directory is recommended.  It takes in
833a pointer to struct fscrypt_add_key_arg, defined as follows::
834
835    struct fscrypt_add_key_arg {
836            struct fscrypt_key_specifier key_spec;
837            __u32 raw_size;
838            __u32 key_id;
839            __u32 __reserved[8];
840            __u8 raw[];
841    };
842
843    #define FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR        1
844    #define FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER        2
845
846    struct fscrypt_key_specifier {
847            __u32 type;     /* one of FSCRYPT_KEY_SPEC_TYPE_* */
848            __u32 __reserved;
849            union {
850                    __u8 __reserved[32]; /* reserve some extra space */
851                    __u8 descriptor[FSCRYPT_KEY_DESCRIPTOR_SIZE];
852                    __u8 identifier[FSCRYPT_KEY_IDENTIFIER_SIZE];
853            } u;
854    };
855
856    struct fscrypt_provisioning_key_payload {
857            __u32 type;
858            __u32 __reserved;
859            __u8 raw[];
860    };
861
862struct fscrypt_add_key_arg must be zeroed, then initialized
863as follows:
864
865- If the key is being added for use by v1 encryption policies, then
866  ``key_spec.type`` must contain FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR, and
867  ``key_spec.u.descriptor`` must contain the descriptor of the key
868  being added, corresponding to the value in the
869  ``master_key_descriptor`` field of struct fscrypt_policy_v1.
870  To add this type of key, the calling process must have the
871  CAP_SYS_ADMIN capability in the initial user namespace.
872
873  Alternatively, if the key is being added for use by v2 encryption
874  policies, then ``key_spec.type`` must contain
875  FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER, and ``key_spec.u.identifier`` is
876  an *output* field which the kernel fills in with a cryptographic
877  hash of the key.  To add this type of key, the calling process does
878  not need any privileges.  However, the number of keys that can be
879  added is limited by the user's quota for the keyrings service (see
880  ``Documentation/security/keys/core.rst``).
881
882- ``raw_size`` must be the size of the ``raw`` key provided, in bytes.
883  Alternatively, if ``key_id`` is nonzero, this field must be 0, since
884  in that case the size is implied by the specified Linux keyring key.
885
886- ``key_id`` is 0 if the raw key is given directly in the ``raw``
887  field.  Otherwise ``key_id`` is the ID of a Linux keyring key of
888  type "fscrypt-provisioning" whose payload is
889  struct fscrypt_provisioning_key_payload whose ``raw`` field contains
890  the raw key and whose ``type`` field matches ``key_spec.type``.
891  Since ``raw`` is variable-length, the total size of this key's
892  payload must be ``sizeof(struct fscrypt_provisioning_key_payload)``
893  plus the raw key size.  The process must have Search permission on
894  this key.
895
896  Most users should leave this 0 and specify the raw key directly.
897  The support for specifying a Linux keyring key is intended mainly to
898  allow re-adding keys after a filesystem is unmounted and re-mounted,
899  without having to store the raw keys in userspace memory.
900
901- ``raw`` is a variable-length field which must contain the actual
902  key, ``raw_size`` bytes long.  Alternatively, if ``key_id`` is
903  nonzero, then this field is unused.
904
905For v2 policy keys, the kernel keeps track of which user (identified
906by effective user ID) added the key, and only allows the key to be
907removed by that user --- or by "root", if they use
908`FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS`_.
909
910However, if another user has added the key, it may be desirable to
911prevent that other user from unexpectedly removing it.  Therefore,
912FS_IOC_ADD_ENCRYPTION_KEY may also be used to add a v2 policy key
913*again*, even if it's already added by other user(s).  In this case,
914FS_IOC_ADD_ENCRYPTION_KEY will just install a claim to the key for the
915current user, rather than actually add the key again (but the raw key
916must still be provided, as a proof of knowledge).
917
918FS_IOC_ADD_ENCRYPTION_KEY returns 0 if either the key or a claim to
919the key was either added or already exists.
920
921FS_IOC_ADD_ENCRYPTION_KEY can fail with the following errors:
922
923- ``EACCES``: FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR was specified, but the
924  caller does not have the CAP_SYS_ADMIN capability in the initial
925  user namespace; or the raw key was specified by Linux key ID but the
926  process lacks Search permission on the key.
927- ``EDQUOT``: the key quota for this user would be exceeded by adding
928  the key
929- ``EINVAL``: invalid key size or key specifier type, or reserved bits
930  were set
931- ``EKEYREJECTED``: the raw key was specified by Linux key ID, but the
932  key has the wrong type
933- ``ENOKEY``: the raw key was specified by Linux key ID, but no key
934  exists with that ID
935- ``ENOTTY``: this type of filesystem does not implement encryption
936- ``EOPNOTSUPP``: the kernel was not configured with encryption
937  support for this filesystem, or the filesystem superblock has not
938  had encryption enabled on it
939
940Legacy method
941~~~~~~~~~~~~~
942
943For v1 encryption policies, a master encryption key can also be
944provided by adding it to a process-subscribed keyring, e.g. to a
945session keyring, or to a user keyring if the user keyring is linked
946into the session keyring.
947
948This method is deprecated (and not supported for v2 encryption
949policies) for several reasons.  First, it cannot be used in
950combination with FS_IOC_REMOVE_ENCRYPTION_KEY (see `Removing keys`_),
951so for removing a key a workaround such as keyctl_unlink() in
952combination with ``sync; echo 2 > /proc/sys/vm/drop_caches`` would
953have to be used.  Second, it doesn't match the fact that the
954locked/unlocked status of encrypted files (i.e. whether they appear to
955be in plaintext form or in ciphertext form) is global.  This mismatch
956has caused much confusion as well as real problems when processes
957running under different UIDs, such as a ``sudo`` command, need to
958access encrypted files.
959
960Nevertheless, to add a key to one of the process-subscribed keyrings,
961the add_key() system call can be used (see:
962``Documentation/security/keys/core.rst``).  The key type must be
963"logon"; keys of this type are kept in kernel memory and cannot be
964read back by userspace.  The key description must be "fscrypt:"
965followed by the 16-character lower case hex representation of the
966``master_key_descriptor`` that was set in the encryption policy.  The
967key payload must conform to the following structure::
968
969    #define FSCRYPT_MAX_KEY_SIZE            64
970
971    struct fscrypt_key {
972            __u32 mode;
973            __u8 raw[FSCRYPT_MAX_KEY_SIZE];
974            __u32 size;
975    };
976
977``mode`` is ignored; just set it to 0.  The actual key is provided in
978``raw`` with ``size`` indicating its size in bytes.  That is, the
979bytes ``raw[0..size-1]`` (inclusive) are the actual key.
980
981The key description prefix "fscrypt:" may alternatively be replaced
982with a filesystem-specific prefix such as "ext4:".  However, the
983filesystem-specific prefixes are deprecated and should not be used in
984new programs.
985
986Removing keys
987-------------
988
989Two ioctls are available for removing a key that was added by
990`FS_IOC_ADD_ENCRYPTION_KEY`_:
991
992- `FS_IOC_REMOVE_ENCRYPTION_KEY`_
993- `FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS`_
994
995These two ioctls differ only in cases where v2 policy keys are added
996or removed by non-root users.
997
998These ioctls don't work on keys that were added via the legacy
999process-subscribed keyrings mechanism.
1000
1001Before using these ioctls, read the `Kernel memory compromise`_
1002section for a discussion of the security goals and limitations of
1003these ioctls.
1004
1005FS_IOC_REMOVE_ENCRYPTION_KEY
1006~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1007
1008The FS_IOC_REMOVE_ENCRYPTION_KEY ioctl removes a claim to a master
1009encryption key from the filesystem, and possibly removes the key
1010itself.  It can be executed on any file or directory on the target
1011filesystem, but using the filesystem's root directory is recommended.
1012It takes in a pointer to struct fscrypt_remove_key_arg, defined
1013as follows::
1014
1015    struct fscrypt_remove_key_arg {
1016            struct fscrypt_key_specifier key_spec;
1017    #define FSCRYPT_KEY_REMOVAL_STATUS_FLAG_FILES_BUSY      0x00000001
1018    #define FSCRYPT_KEY_REMOVAL_STATUS_FLAG_OTHER_USERS     0x00000002
1019            __u32 removal_status_flags;     /* output */
1020            __u32 __reserved[5];
1021    };
1022
1023This structure must be zeroed, then initialized as follows:
1024
1025- The key to remove is specified by ``key_spec``:
1026
1027    - To remove a key used by v1 encryption policies, set
1028      ``key_spec.type`` to FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR and fill
1029      in ``key_spec.u.descriptor``.  To remove this type of key, the
1030      calling process must have the CAP_SYS_ADMIN capability in the
1031      initial user namespace.
1032
1033    - To remove a key used by v2 encryption policies, set
1034      ``key_spec.type`` to FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER and fill
1035      in ``key_spec.u.identifier``.
1036
1037For v2 policy keys, this ioctl is usable by non-root users.  However,
1038to make this possible, it actually just removes the current user's
1039claim to the key, undoing a single call to FS_IOC_ADD_ENCRYPTION_KEY.
1040Only after all claims are removed is the key really removed.
1041
1042For example, if FS_IOC_ADD_ENCRYPTION_KEY was called with uid 1000,
1043then the key will be "claimed" by uid 1000, and
1044FS_IOC_REMOVE_ENCRYPTION_KEY will only succeed as uid 1000.  Or, if
1045both uids 1000 and 2000 added the key, then for each uid
1046FS_IOC_REMOVE_ENCRYPTION_KEY will only remove their own claim.  Only
1047once *both* are removed is the key really removed.  (Think of it like
1048unlinking a file that may have hard links.)
1049
1050If FS_IOC_REMOVE_ENCRYPTION_KEY really removes the key, it will also
1051try to "lock" all files that had been unlocked with the key.  It won't
1052lock files that are still in-use, so this ioctl is expected to be used
1053in cooperation with userspace ensuring that none of the files are
1054still open.  However, if necessary, this ioctl can be executed again
1055later to retry locking any remaining files.
1056
1057FS_IOC_REMOVE_ENCRYPTION_KEY returns 0 if either the key was removed
1058(but may still have files remaining to be locked), the user's claim to
1059the key was removed, or the key was already removed but had files
1060remaining to be the locked so the ioctl retried locking them.  In any
1061of these cases, ``removal_status_flags`` is filled in with the
1062following informational status flags:
1063
1064- ``FSCRYPT_KEY_REMOVAL_STATUS_FLAG_FILES_BUSY``: set if some file(s)
1065  are still in-use.  Not guaranteed to be set in the case where only
1066  the user's claim to the key was removed.
1067- ``FSCRYPT_KEY_REMOVAL_STATUS_FLAG_OTHER_USERS``: set if only the
1068  user's claim to the key was removed, not the key itself
1069
1070FS_IOC_REMOVE_ENCRYPTION_KEY can fail with the following errors:
1071
1072- ``EACCES``: The FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR key specifier type
1073  was specified, but the caller does not have the CAP_SYS_ADMIN
1074  capability in the initial user namespace
1075- ``EINVAL``: invalid key specifier type, or reserved bits were set
1076- ``ENOKEY``: the key object was not found at all, i.e. it was never
1077  added in the first place or was already fully removed including all
1078  files locked; or, the user does not have a claim to the key (but
1079  someone else does).
1080- ``ENOTTY``: this type of filesystem does not implement encryption
1081- ``EOPNOTSUPP``: the kernel was not configured with encryption
1082  support for this filesystem, or the filesystem superblock has not
1083  had encryption enabled on it
1084
1085FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS
1086~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1087
1088FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS is exactly the same as
1089`FS_IOC_REMOVE_ENCRYPTION_KEY`_, except that for v2 policy keys, the
1090ALL_USERS version of the ioctl will remove all users' claims to the
1091key, not just the current user's.  I.e., the key itself will always be
1092removed, no matter how many users have added it.  This difference is
1093only meaningful if non-root users are adding and removing keys.
1094
1095Because of this, FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS also requires
1096"root", namely the CAP_SYS_ADMIN capability in the initial user
1097namespace.  Otherwise it will fail with EACCES.
1098
1099Getting key status
1100------------------
1101
1102FS_IOC_GET_ENCRYPTION_KEY_STATUS
1103~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1104
1105The FS_IOC_GET_ENCRYPTION_KEY_STATUS ioctl retrieves the status of a
1106master encryption key.  It can be executed on any file or directory on
1107the target filesystem, but using the filesystem's root directory is
1108recommended.  It takes in a pointer to
1109struct fscrypt_get_key_status_arg, defined as follows::
1110
1111    struct fscrypt_get_key_status_arg {
1112            /* input */
1113            struct fscrypt_key_specifier key_spec;
1114            __u32 __reserved[6];
1115
1116            /* output */
1117    #define FSCRYPT_KEY_STATUS_ABSENT               1
1118    #define FSCRYPT_KEY_STATUS_PRESENT              2
1119    #define FSCRYPT_KEY_STATUS_INCOMPLETELY_REMOVED 3
1120            __u32 status;
1121    #define FSCRYPT_KEY_STATUS_FLAG_ADDED_BY_SELF   0x00000001
1122            __u32 status_flags;
1123            __u32 user_count;
1124            __u32 __out_reserved[13];
1125    };
1126
1127The caller must zero all input fields, then fill in ``key_spec``:
1128
1129    - To get the status of a key for v1 encryption policies, set
1130      ``key_spec.type`` to FSCRYPT_KEY_SPEC_TYPE_DESCRIPTOR and fill
1131      in ``key_spec.u.descriptor``.
1132
1133    - To get the status of a key for v2 encryption policies, set
1134      ``key_spec.type`` to FSCRYPT_KEY_SPEC_TYPE_IDENTIFIER and fill
1135      in ``key_spec.u.identifier``.
1136
1137On success, 0 is returned and the kernel fills in the output fields:
1138
1139- ``status`` indicates whether the key is absent, present, or
1140  incompletely removed.  Incompletely removed means that removal has
1141  been initiated, but some files are still in use; i.e.,
1142  `FS_IOC_REMOVE_ENCRYPTION_KEY`_ returned 0 but set the informational
1143  status flag FSCRYPT_KEY_REMOVAL_STATUS_FLAG_FILES_BUSY.
1144
1145- ``status_flags`` can contain the following flags:
1146
1147    - ``FSCRYPT_KEY_STATUS_FLAG_ADDED_BY_SELF`` indicates that the key
1148      has added by the current user.  This is only set for keys
1149      identified by ``identifier`` rather than by ``descriptor``.
1150
1151- ``user_count`` specifies the number of users who have added the key.
1152  This is only set for keys identified by ``identifier`` rather than
1153  by ``descriptor``.
1154
1155FS_IOC_GET_ENCRYPTION_KEY_STATUS can fail with the following errors:
1156
1157- ``EINVAL``: invalid key specifier type, or reserved bits were set
1158- ``ENOTTY``: this type of filesystem does not implement encryption
1159- ``EOPNOTSUPP``: the kernel was not configured with encryption
1160  support for this filesystem, or the filesystem superblock has not
1161  had encryption enabled on it
1162
1163Among other use cases, FS_IOC_GET_ENCRYPTION_KEY_STATUS can be useful
1164for determining whether the key for a given encrypted directory needs
1165to be added before prompting the user for the passphrase needed to
1166derive the key.
1167
1168FS_IOC_GET_ENCRYPTION_KEY_STATUS can only get the status of keys in
1169the filesystem-level keyring, i.e. the keyring managed by
1170`FS_IOC_ADD_ENCRYPTION_KEY`_ and `FS_IOC_REMOVE_ENCRYPTION_KEY`_.  It
1171cannot get the status of a key that has only been added for use by v1
1172encryption policies using the legacy mechanism involving
1173process-subscribed keyrings.
1174
1175Access semantics
1176================
1177
1178With the key
1179------------
1180
1181With the encryption key, encrypted regular files, directories, and
1182symlinks behave very similarly to their unencrypted counterparts ---
1183after all, the encryption is intended to be transparent.  However,
1184astute users may notice some differences in behavior:
1185
1186- Unencrypted files, or files encrypted with a different encryption
1187  policy (i.e. different key, modes, or flags), cannot be renamed or
1188  linked into an encrypted directory; see `Encryption policy
1189  enforcement`_.  Attempts to do so will fail with EXDEV.  However,
1190  encrypted files can be renamed within an encrypted directory, or
1191  into an unencrypted directory.
1192
1193  Note: "moving" an unencrypted file into an encrypted directory, e.g.
1194  with the `mv` program, is implemented in userspace by a copy
1195  followed by a delete.  Be aware that the original unencrypted data
1196  may remain recoverable from free space on the disk; prefer to keep
1197  all files encrypted from the very beginning.  The `shred` program
1198  may be used to overwrite the source files but isn't guaranteed to be
1199  effective on all filesystems and storage devices.
1200
1201- Direct I/O is supported on encrypted files only under some
1202  circumstances.  For details, see `Direct I/O support`_.
1203
1204- The fallocate operations FALLOC_FL_COLLAPSE_RANGE and
1205  FALLOC_FL_INSERT_RANGE are not supported on encrypted files and will
1206  fail with EOPNOTSUPP.
1207
1208- Online defragmentation of encrypted files is not supported.  The
1209  EXT4_IOC_MOVE_EXT and F2FS_IOC_MOVE_RANGE ioctls will fail with
1210  EOPNOTSUPP.
1211
1212- The ext4 filesystem does not support data journaling with encrypted
1213  regular files.  It will fall back to ordered data mode instead.
1214
1215- DAX (Direct Access) is not supported on encrypted files.
1216
1217- The maximum length of an encrypted symlink is 2 bytes shorter than
1218  the maximum length of an unencrypted symlink.  For example, on an
1219  EXT4 filesystem with a 4K block size, unencrypted symlinks can be up
1220  to 4095 bytes long, while encrypted symlinks can only be up to 4093
1221  bytes long (both lengths excluding the terminating null).
1222
1223Note that mmap *is* supported.  This is possible because the pagecache
1224for an encrypted file contains the plaintext, not the ciphertext.
1225
1226Without the key
1227---------------
1228
1229Some filesystem operations may be performed on encrypted regular
1230files, directories, and symlinks even before their encryption key has
1231been added, or after their encryption key has been removed:
1232
1233- File metadata may be read, e.g. using stat().
1234
1235- Directories may be listed, in which case the filenames will be
1236  listed in an encoded form derived from their ciphertext.  The
1237  current encoding algorithm is described in `Filename hashing and
1238  encoding`_.  The algorithm is subject to change, but it is
1239  guaranteed that the presented filenames will be no longer than
1240  NAME_MAX bytes, will not contain the ``/`` or ``\0`` characters, and
1241  will uniquely identify directory entries.
1242
1243  The ``.`` and ``..`` directory entries are special.  They are always
1244  present and are not encrypted or encoded.
1245
1246- Files may be deleted.  That is, nondirectory files may be deleted
1247  with unlink() as usual, and empty directories may be deleted with
1248  rmdir() as usual.  Therefore, ``rm`` and ``rm -r`` will work as
1249  expected.
1250
1251- Symlink targets may be read and followed, but they will be presented
1252  in encrypted form, similar to filenames in directories.  Hence, they
1253  are unlikely to point to anywhere useful.
1254
1255Without the key, regular files cannot be opened or truncated.
1256Attempts to do so will fail with ENOKEY.  This implies that any
1257regular file operations that require a file descriptor, such as
1258read(), write(), mmap(), fallocate(), and ioctl(), are also forbidden.
1259
1260Also without the key, files of any type (including directories) cannot
1261be created or linked into an encrypted directory, nor can a name in an
1262encrypted directory be the source or target of a rename, nor can an
1263O_TMPFILE temporary file be created in an encrypted directory.  All
1264such operations will fail with ENOKEY.
1265
1266It is not currently possible to backup and restore encrypted files
1267without the encryption key.  This would require special APIs which
1268have not yet been implemented.
1269
1270Encryption policy enforcement
1271=============================
1272
1273After an encryption policy has been set on a directory, all regular
1274files, directories, and symbolic links created in that directory
1275(recursively) will inherit that encryption policy.  Special files ---
1276that is, named pipes, device nodes, and UNIX domain sockets --- will
1277not be encrypted.
1278
1279Except for those special files, it is forbidden to have unencrypted
1280files, or files encrypted with a different encryption policy, in an
1281encrypted directory tree.  Attempts to link or rename such a file into
1282an encrypted directory will fail with EXDEV.  This is also enforced
1283during ->lookup() to provide limited protection against offline
1284attacks that try to disable or downgrade encryption in known locations
1285where applications may later write sensitive data.  It is recommended
1286that systems implementing a form of "verified boot" take advantage of
1287this by validating all top-level encryption policies prior to access.
1288
1289Inline encryption support
1290=========================
1291
1292By default, fscrypt uses the kernel crypto API for all cryptographic
1293operations (other than HKDF, which fscrypt partially implements
1294itself).  The kernel crypto API supports hardware crypto accelerators,
1295but only ones that work in the traditional way where all inputs and
1296outputs (e.g. plaintexts and ciphertexts) are in memory.  fscrypt can
1297take advantage of such hardware, but the traditional acceleration
1298model isn't particularly efficient and fscrypt hasn't been optimized
1299for it.
1300
1301Instead, many newer systems (especially mobile SoCs) have *inline
1302encryption hardware* that can encrypt/decrypt data while it is on its
1303way to/from the storage device.  Linux supports inline encryption
1304through a set of extensions to the block layer called *blk-crypto*.
1305blk-crypto allows filesystems to attach encryption contexts to bios
1306(I/O requests) to specify how the data will be encrypted or decrypted
1307in-line.  For more information about blk-crypto, see
1308:ref:`Documentation/block/inline-encryption.rst <inline_encryption>`.
1309
1310On supported filesystems (currently ext4 and f2fs), fscrypt can use
1311blk-crypto instead of the kernel crypto API to encrypt/decrypt file
1312contents.  To enable this, set CONFIG_FS_ENCRYPTION_INLINE_CRYPT=y in
1313the kernel configuration, and specify the "inlinecrypt" mount option
1314when mounting the filesystem.
1315
1316Note that the "inlinecrypt" mount option just specifies to use inline
1317encryption when possible; it doesn't force its use.  fscrypt will
1318still fall back to using the kernel crypto API on files where the
1319inline encryption hardware doesn't have the needed crypto capabilities
1320(e.g. support for the needed encryption algorithm and data unit size)
1321and where blk-crypto-fallback is unusable.  (For blk-crypto-fallback
1322to be usable, it must be enabled in the kernel configuration with
1323CONFIG_BLK_INLINE_ENCRYPTION_FALLBACK=y.)
1324
1325Currently fscrypt always uses the filesystem block size (which is
1326usually 4096 bytes) as the data unit size.  Therefore, it can only use
1327inline encryption hardware that supports that data unit size.
1328
1329Inline encryption doesn't affect the ciphertext or other aspects of
1330the on-disk format, so users may freely switch back and forth between
1331using "inlinecrypt" and not using "inlinecrypt".
1332
1333Direct I/O support
1334==================
1335
1336For direct I/O on an encrypted file to work, the following conditions
1337must be met (in addition to the conditions for direct I/O on an
1338unencrypted file):
1339
1340* The file must be using inline encryption.  Usually this means that
1341  the filesystem must be mounted with ``-o inlinecrypt`` and inline
1342  encryption hardware must be present.  However, a software fallback
1343  is also available.  For details, see `Inline encryption support`_.
1344
1345* The I/O request must be fully aligned to the filesystem block size.
1346  This means that the file position the I/O is targeting, the lengths
1347  of all I/O segments, and the memory addresses of all I/O buffers
1348  must be multiples of this value.  Note that the filesystem block
1349  size may be greater than the logical block size of the block device.
1350
1351If either of the above conditions is not met, then direct I/O on the
1352encrypted file will fall back to buffered I/O.
1353
1354Implementation details
1355======================
1356
1357Encryption context
1358------------------
1359
1360An encryption policy is represented on-disk by
1361struct fscrypt_context_v1 or struct fscrypt_context_v2.  It is up to
1362individual filesystems to decide where to store it, but normally it
1363would be stored in a hidden extended attribute.  It should *not* be
1364exposed by the xattr-related system calls such as getxattr() and
1365setxattr() because of the special semantics of the encryption xattr.
1366(In particular, there would be much confusion if an encryption policy
1367were to be added to or removed from anything other than an empty
1368directory.)  These structs are defined as follows::
1369
1370    #define FSCRYPT_FILE_NONCE_SIZE 16
1371
1372    #define FSCRYPT_KEY_DESCRIPTOR_SIZE  8
1373    struct fscrypt_context_v1 {
1374            u8 version;
1375            u8 contents_encryption_mode;
1376            u8 filenames_encryption_mode;
1377            u8 flags;
1378            u8 master_key_descriptor[FSCRYPT_KEY_DESCRIPTOR_SIZE];
1379            u8 nonce[FSCRYPT_FILE_NONCE_SIZE];
1380    };
1381
1382    #define FSCRYPT_KEY_IDENTIFIER_SIZE  16
1383    struct fscrypt_context_v2 {
1384            u8 version;
1385            u8 contents_encryption_mode;
1386            u8 filenames_encryption_mode;
1387            u8 flags;
1388            u8 log2_data_unit_size;
1389            u8 __reserved[3];
1390            u8 master_key_identifier[FSCRYPT_KEY_IDENTIFIER_SIZE];
1391            u8 nonce[FSCRYPT_FILE_NONCE_SIZE];
1392    };
1393
1394The context structs contain the same information as the corresponding
1395policy structs (see `Setting an encryption policy`_), except that the
1396context structs also contain a nonce.  The nonce is randomly generated
1397by the kernel and is used as KDF input or as a tweak to cause
1398different files to be encrypted differently; see `Per-file encryption
1399keys`_ and `DIRECT_KEY policies`_.
1400
1401Data path changes
1402-----------------
1403
1404When inline encryption is used, filesystems just need to associate
1405encryption contexts with bios to specify how the block layer or the
1406inline encryption hardware will encrypt/decrypt the file contents.
1407
1408When inline encryption isn't used, filesystems must encrypt/decrypt
1409the file contents themselves, as described below:
1410
1411For the read path (->read_folio()) of regular files, filesystems can
1412read the ciphertext into the page cache and decrypt it in-place.  The
1413folio lock must be held until decryption has finished, to prevent the
1414folio from becoming visible to userspace prematurely.
1415
1416For the write path (->writepage()) of regular files, filesystems
1417cannot encrypt data in-place in the page cache, since the cached
1418plaintext must be preserved.  Instead, filesystems must encrypt into a
1419temporary buffer or "bounce page", then write out the temporary
1420buffer.  Some filesystems, such as UBIFS, already use temporary
1421buffers regardless of encryption.  Other filesystems, such as ext4 and
1422F2FS, have to allocate bounce pages specially for encryption.
1423
1424Filename hashing and encoding
1425-----------------------------
1426
1427Modern filesystems accelerate directory lookups by using indexed
1428directories.  An indexed directory is organized as a tree keyed by
1429filename hashes.  When a ->lookup() is requested, the filesystem
1430normally hashes the filename being looked up so that it can quickly
1431find the corresponding directory entry, if any.
1432
1433With encryption, lookups must be supported and efficient both with and
1434without the encryption key.  Clearly, it would not work to hash the
1435plaintext filenames, since the plaintext filenames are unavailable
1436without the key.  (Hashing the plaintext filenames would also make it
1437impossible for the filesystem's fsck tool to optimize encrypted
1438directories.)  Instead, filesystems hash the ciphertext filenames,
1439i.e. the bytes actually stored on-disk in the directory entries.  When
1440asked to do a ->lookup() with the key, the filesystem just encrypts
1441the user-supplied name to get the ciphertext.
1442
1443Lookups without the key are more complicated.  The raw ciphertext may
1444contain the ``\0`` and ``/`` characters, which are illegal in
1445filenames.  Therefore, readdir() must base64url-encode the ciphertext
1446for presentation.  For most filenames, this works fine; on ->lookup(),
1447the filesystem just base64url-decodes the user-supplied name to get
1448back to the raw ciphertext.
1449
1450However, for very long filenames, base64url encoding would cause the
1451filename length to exceed NAME_MAX.  To prevent this, readdir()
1452actually presents long filenames in an abbreviated form which encodes
1453a strong "hash" of the ciphertext filename, along with the optional
1454filesystem-specific hash(es) needed for directory lookups.  This
1455allows the filesystem to still, with a high degree of confidence, map
1456the filename given in ->lookup() back to a particular directory entry
1457that was previously listed by readdir().  See
1458struct fscrypt_nokey_name in the source for more details.
1459
1460Note that the precise way that filenames are presented to userspace
1461without the key is subject to change in the future.  It is only meant
1462as a way to temporarily present valid filenames so that commands like
1463``rm -r`` work as expected on encrypted directories.
1464
1465Tests
1466=====
1467
1468To test fscrypt, use xfstests, which is Linux's de facto standard
1469filesystem test suite.  First, run all the tests in the "encrypt"
1470group on the relevant filesystem(s).  One can also run the tests
1471with the 'inlinecrypt' mount option to test the implementation for
1472inline encryption support.  For example, to test ext4 and
1473f2fs encryption using `kvm-xfstests
1474<https://github.com/tytso/xfstests-bld/blob/master/Documentation/kvm-quickstart.md>`_::
1475
1476    kvm-xfstests -c ext4,f2fs -g encrypt
1477    kvm-xfstests -c ext4,f2fs -g encrypt -m inlinecrypt
1478
1479UBIFS encryption can also be tested this way, but it should be done in
1480a separate command, and it takes some time for kvm-xfstests to set up
1481emulated UBI volumes::
1482
1483    kvm-xfstests -c ubifs -g encrypt
1484
1485No tests should fail.  However, tests that use non-default encryption
1486modes (e.g. generic/549 and generic/550) will be skipped if the needed
1487algorithms were not built into the kernel's crypto API.  Also, tests
1488that access the raw block device (e.g. generic/399, generic/548,
1489generic/549, generic/550) will be skipped on UBIFS.
1490
1491Besides running the "encrypt" group tests, for ext4 and f2fs it's also
1492possible to run most xfstests with the "test_dummy_encryption" mount
1493option.  This option causes all new files to be automatically
1494encrypted with a dummy key, without having to make any API calls.
1495This tests the encrypted I/O paths more thoroughly.  To do this with
1496kvm-xfstests, use the "encrypt" filesystem configuration::
1497
1498    kvm-xfstests -c ext4/encrypt,f2fs/encrypt -g auto
1499    kvm-xfstests -c ext4/encrypt,f2fs/encrypt -g auto -m inlinecrypt
1500
1501Because this runs many more tests than "-g encrypt" does, it takes
1502much longer to run; so also consider using `gce-xfstests
1503<https://github.com/tytso/xfstests-bld/blob/master/Documentation/gce-xfstests.md>`_
1504instead of kvm-xfstests::
1505
1506    gce-xfstests -c ext4/encrypt,f2fs/encrypt -g auto
1507    gce-xfstests -c ext4/encrypt,f2fs/encrypt -g auto -m inlinecrypt
1508