| blob | 16efd729bf7c97fd875ab3fc9ce30d22faed2bb8 |
1 .. SPDX-License-Identifier: GPL-2.0
3 :orphan:
5 .. UBIFS Authentication
6 .. sigma star gmbh
7 .. 2018
9 Introduction
10 ============
12 UBIFS utilizes the fscrypt framework to provide confidentiality for file
13 contents and file names. This prevents attacks where an attacker is able to
14 read contents of the filesystem on a single point in time. A classic example
15 is a lost smartphone where the attacker is unable to read personal data stored
16 on the device without the filesystem decryption key.
18 At the current state, UBIFS encryption however does not prevent attacks where
19 the attacker is able to modify the filesystem contents and the user uses the
20 device afterwards. In such a scenario an attacker can modify filesystem
21 contents arbitrarily without the user noticing. One example is to modify a
22 binary to perform a malicious action when executed [DMC-CBC-ATTACK]. Since
23 most of the filesystem metadata of UBIFS is stored in plain, this makes it
24 fairly easy to swap files and replace their contents.
26 Other full disk encryption systems like dm-crypt cover all filesystem metadata,
27 which makes such kinds of attacks more complicated, but not impossible.
28 Especially, if the attacker is given access to the device multiple points in
29 time. For dm-crypt and other filesystems that build upon the Linux block IO
30 layer, the dm-integrity or dm-verity subsystems [DM-INTEGRITY, DM-VERITY]
31 can be used to get full data authentication at the block layer.
32 These can also be combined with dm-crypt [CRYPTSETUP2].
34 This document describes an approach to get file contents _and_ full metadata
35 authentication for UBIFS. Since UBIFS uses fscrypt for file contents and file
36 name encryption, the authentication system could be tied into fscrypt such that
37 existing features like key derivation can be utilized. It should however also
38 be possible to use UBIFS authentication without using encryption.
41 MTD, UBI & UBIFS
42 ----------------
44 On Linux, the MTD (Memory Technology Devices) subsystem provides a uniform
45 interface to access raw flash devices. One of the more prominent subsystems that
46 work on top of MTD is UBI (Unsorted Block Images). It provides volume management
47 for flash devices and is thus somewhat similar to LVM for block devices. In
48 addition, it deals with flash-specific wear-leveling and transparent I/O error
49 handling. UBI offers logical erase blocks (LEBs) to the layers on top of it
50 and maps them transparently to physical erase blocks (PEBs) on the flash.
52 UBIFS is a filesystem for raw flash which operates on top of UBI. Thus, wear
53 leveling and some flash specifics are left to UBI, while UBIFS focuses on
54 scalability, performance and recoverability.
56 ::
58 +------------+ +*******+ +-----------+ +-----+
59 | | * UBIFS * | UBI-BLOCK | | ... |
60 | JFFS/JFFS2 | +*******+ +-----------+ +-----+
61 | | +-----------------------------+ +-----------+ +-----+
62 | | | UBI | | MTD-BLOCK | | ... |
63 +------------+ +-----------------------------+ +-----------+ +-----+
64 +------------------------------------------------------------------+
65 | MEMORY TECHNOLOGY DEVICES (MTD) |
66 +------------------------------------------------------------------+
67 +-----------------------------+ +--------------------------+ +-----+
68 | NAND DRIVERS | | NOR DRIVERS | | ... |
69 +-----------------------------+ +--------------------------+ +-----+
71 Figure 1: Linux kernel subsystems for dealing with raw flash
75 Internally, UBIFS maintains multiple data structures which are persisted on
76 the flash:
78 - *Index*: an on-flash B+ tree where the leaf nodes contain filesystem data
79 - *Journal*: an additional data structure to collect FS changes before updating
80 the on-flash index and reduce flash wear.
81 - *Tree Node Cache (TNC)*: an in-memory B+ tree that reflects the current FS
82 state to avoid frequent flash reads. It is basically the in-memory
83 representation of the index, but contains additional attributes.
84 - *LEB property tree (LPT)*: an on-flash B+ tree for free space accounting per
85 UBI LEB.
87 In the remainder of this section we will cover the on-flash UBIFS data
88 structures in more detail. The TNC is of less importance here since it is never
89 persisted onto the flash directly. More details on UBIFS can also be found in
90 [UBIFS-WP].
93 UBIFS Index & Tree Node Cache
94 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
96 Basic on-flash UBIFS entities are called *nodes*. UBIFS knows different types
97 of nodes. Eg. data nodes (``struct ubifs_data_node``) which store chunks of file
98 contents or inode nodes (``struct ubifs_ino_node``) which represent VFS inodes.
99 Almost all types of nodes share a common header (``ubifs_ch``) containing basic
100 information like node type, node length, a sequence number, etc. (see
101 ``fs/ubifs/ubifs-media.h`` in kernel source). Exceptions are entries of the LPT
102 and some less important node types like padding nodes which are used to pad
103 unusable content at the end of LEBs.
105 To avoid re-writing the whole B+ tree on every single change, it is implemented
106 as *wandering tree*, where only the changed nodes are re-written and previous
107 versions of them are obsoleted without erasing them right away. As a result,
108 the index is not stored in a single place on the flash, but *wanders* around
109 and there are obsolete parts on the flash as long as the LEB containing them is
110 not reused by UBIFS. To find the most recent version of the index, UBIFS stores
111 a special node called *master node* into UBI LEB 1 which always points to the
112 most recent root node of the UBIFS index. For recoverability, the master node
113 is additionally duplicated to LEB 2. Mounting UBIFS is thus a simple read of
114 LEB 1 and 2 to get the current master node and from there get the location of
115 the most recent on-flash index.
117 The TNC is the in-memory representation of the on-flash index. It contains some
118 additional runtime attributes per node which are not persisted. One of these is
119 a dirty-flag which marks nodes that have to be persisted the next time the
120 index is written onto the flash. The TNC acts as a write-back cache and all
121 modifications of the on-flash index are done through the TNC. Like other caches,
122 the TNC does not have to mirror the full index into memory, but reads parts of
123 it from flash whenever needed. A *commit* is the UBIFS operation of updating the
124 on-flash filesystem structures like the index. On every commit, the TNC nodes
125 marked as dirty are written to the flash to update the persisted index.
128 Journal
129 ~~~~~~~
131 To avoid wearing out the flash, the index is only persisted (*commited*) when
132 certain conditions are met (eg. ``fsync(2)``). The journal is used to record
133 any changes (in form of inode nodes, data nodes etc.) between commits
134 of the index. During mount, the journal is read from the flash and replayed
135 onto the TNC (which will be created on-demand from the on-flash index).
137 UBIFS reserves a bunch of LEBs just for the journal called *log area*. The
138 amount of log area LEBs is configured on filesystem creation (using
139 ``mkfs.ubifs``) and stored in the superblock node. The log area contains only
140 two types of nodes: *reference nodes* and *commit start nodes*. A commit start
141 node is written whenever an index commit is performed. Reference nodes are
142 written on every journal update. Each reference node points to the position of
143 other nodes (inode nodes, data nodes etc.) on the flash that are part of this
144 journal entry. These nodes are called *buds* and describe the actual filesystem
145 changes including their data.
147 The log area is maintained as a ring. Whenever the journal is almost full,
148 a commit is initiated. This also writes a commit start node so that during
149 mount, UBIFS will seek for the most recent commit start node and just replay
150 every reference node after that. Every reference node before the commit start
151 node will be ignored as they are already part of the on-flash index.
153 When writing a journal entry, UBIFS first ensures that enough space is
154 available to write the reference node and buds part of this entry. Then, the
155 reference node is written and afterwards the buds describing the file changes.
156 On replay, UBIFS will record every reference node and inspect the location of
157 the referenced LEBs to discover the buds. If these are corrupt or missing,
158 UBIFS will attempt to recover them by re-reading the LEB. This is however only
159 done for the last referenced LEB of the journal. Only this can become corrupt
160 because of a power cut. If the recovery fails, UBIFS will not mount. An error
161 for every other LEB will directly cause UBIFS to fail the mount operation.
163 ::
165 | ---- LOG AREA ---- | ---------- MAIN AREA ------------ |
167 -----+------+-----+--------+---- ------+-----+-----+---------------
168 \ | | | | / / | | | \
169 / CS | REF | REF | | \ \ DENT | INO | INO | /
170 \ | | | | / / | | | \
171 ----+------+-----+--------+--- -------+-----+-----+----------------
172 | | ^ ^
173 | | | |
174 +------------------------+ |
175 | |
176 +-------------------------------+
179 Figure 2: UBIFS flash layout of log area with commit start nodes
180 (CS) and reference nodes (REF) pointing to main area
181 containing their buds
184 LEB Property Tree/Table
185 ~~~~~~~~~~~~~~~~~~~~~~~
187 The LEB property tree is used to store per-LEB information. This includes the
188 LEB type and amount of free and *dirty* (old, obsolete content) space [1]_ on
189 the LEB. The type is important, because UBIFS never mixes index nodes with data
190 nodes on a single LEB and thus each LEB has a specific purpose. This again is
191 useful for free space calculations. See [UBIFS-WP] for more details.
193 The LEB property tree again is a B+ tree, but it is much smaller than the
194 index. Due to its smaller size it is always written as one chunk on every
195 commit. Thus, saving the LPT is an atomic operation.
198 .. [1] Since LEBs can only be appended and never overwritten, there is a
199 difference between free space ie. the remaining space left on the LEB to be
200 written to without erasing it and previously written content that is obsolete
201 but can't be overwritten without erasing the full LEB.
204 UBIFS Authentication
205 ====================
207 This chapter introduces UBIFS authentication which enables UBIFS to verify
208 the authenticity and integrity of metadata and file contents stored on flash.
211 Threat Model
212 ------------
214 UBIFS authentication enables detection of offline data modification. While it
215 does not prevent it, it enables (trusted) code to check the integrity and
216 authenticity of on-flash file contents and filesystem metadata. This covers
217 attacks where file contents are swapped.
219 UBIFS authentication will not protect against rollback of full flash contents.
220 Ie. an attacker can still dump the flash and restore it at a later time without
221 detection. It will also not protect against partial rollback of individual
222 index commits. That means that an attacker is able to partially undo changes.
223 This is possible because UBIFS does not immediately overwrites obsolete
224 versions of the index tree or the journal, but instead marks them as obsolete
225 and garbage collection erases them at a later time. An attacker can use this by
226 erasing parts of the current tree and restoring old versions that are still on
227 the flash and have not yet been erased. This is possible, because every commit
228 will always write a new version of the index root node and the master node
229 without overwriting the previous version. This is further helped by the
230 wear-leveling operations of UBI which copies contents from one physical
231 eraseblock to another and does not atomically erase the first eraseblock.
233 UBIFS authentication does not cover attacks where an attacker is able to
234 execute code on the device after the authentication key was provided.
235 Additional measures like secure boot and trusted boot have to be taken to
236 ensure that only trusted code is executed on a device.
239 Authentication
240 --------------
242 To be able to fully trust data read from flash, all UBIFS data structures
243 stored on flash are authenticated. That is:
245 - The index which includes file contents, file metadata like extended
246 attributes, file length etc.
247 - The journal which also contains file contents and metadata by recording changes
248 to the filesystem
249 - The LPT which stores UBI LEB metadata which UBIFS uses for free space accounting
252 Index Authentication
253 ~~~~~~~~~~~~~~~~~~~~
255 Through UBIFS' concept of a wandering tree, it already takes care of only
256 updating and persisting changed parts from leaf node up to the root node
257 of the full B+ tree. This enables us to augment the index nodes of the tree
258 with a hash over each node's child nodes. As a result, the index basically also
259 a Merkle tree. Since the leaf nodes of the index contain the actual filesystem
260 data, the hashes of their parent index nodes thus cover all the file contents
261 and file metadata. When a file changes, the UBIFS index is updated accordingly
262 from the leaf nodes up to the root node including the master node. This process
263 can be hooked to recompute the hash only for each changed node at the same time.
264 Whenever a file is read, UBIFS can verify the hashes from each leaf node up to
265 the root node to ensure the node's integrity.
267 To ensure the authenticity of the whole index, the UBIFS master node stores a
268 keyed hash (HMAC) over its own contents and a hash of the root node of the index
269 tree. As mentioned above, the master node is always written to the flash whenever
270 the index is persisted (ie. on index commit).
272 Using this approach only UBIFS index nodes and the master node are changed to
273 include a hash. All other types of nodes will remain unchanged. This reduces
274 the storage overhead which is precious for users of UBIFS (ie. embedded
275 devices).
277 ::
279 +---------------+
280 | Master Node |
281 | (hash) |
282 +---------------+
283 |
284 v
285 +-------------------+
286 | Index Node #1 |
287 | |
288 | branch0 branchn |
289 | (hash) (hash) |
290 +-------------------+
291 | ... | (fanout: 8)
292 | |
293 +-------+ +------+
294 | |
295 v v
296 +-------------------+ +-------------------+
297 | Index Node #2 | | Index Node #3 |
298 | | | |
299 | branch0 branchn | | branch0 branchn |
300 | (hash) (hash) | | (hash) (hash) |
301 +-------------------+ +-------------------+
302 | ... | ... |
303 v v v
304 +-----------+ +----------+ +-----------+
305 | Data Node | | INO Node | | DENT Node |
306 +-----------+ +----------+ +-----------+
309 Figure 3: Coverage areas of index node hash and master node HMAC
313 The most important part for robustness and power-cut safety is to atomically
314 persist the hash and file contents. Here the existing UBIFS logic for how
315 changed nodes are persisted is already designed for this purpose such that
316 UBIFS can safely recover if a power-cut occurs while persisting. Adding
317 hashes to index nodes does not change this since each hash will be persisted
318 atomically together with its respective node.
321 Journal Authentication
322 ~~~~~~~~~~~~~~~~~~~~~~
324 The journal is authenticated too. Since the journal is continuously written
325 it is necessary to also add authentication information frequently to the
326 journal so that in case of a powercut not too much data can't be authenticated.
327 This is done by creating a continuous hash beginning from the commit start node
328 over the previous reference nodes, the current reference node, and the bud
329 nodes. From time to time whenever it is suitable authentication nodes are added
330 between the bud nodes. This new node type contains a HMAC over the current state
331 of the hash chain. That way a journal can be authenticated up to the last
332 authentication node. The tail of the journal which may not have a authentication
333 node cannot be authenticated and is skipped during journal replay.
335 We get this picture for journal authentication::
337 ,,,,,,,,
338 ,......,...........................................
339 ,. CS , hash1.----. hash2.----.
340 ,. | , . |hmac . |hmac
341 ,. v , . v . v
342 ,.REF#0,-> bud -> bud -> bud.-> auth -> bud -> bud.-> auth ...
343 ,..|...,...........................................
344 , | ,
345 , | ,,,,,,,,,,,,,,,
346 . | hash3,----.
347 , | , |hmac
348 , v , v
349 , REF#1 -> bud -> bud,-> auth ...
350 ,,,|,,,,,,,,,,,,,,,,,,
351 v
352 REF#2 -> ...
353 |
354 V
355 ...
357 Since the hash also includes the reference nodes an attacker cannot reorder or
358 skip any journal heads for replay. An attacker can only remove bud nodes or
359 reference nodes from the end of the journal, effectively rewinding the
360 filesystem at maximum back to the last commit.
362 The location of the log area is stored in the master node. Since the master
363 node is authenticated with a HMAC as described above, it is not possible to
364 tamper with that without detection. The size of the log area is specified when
365 the filesystem is created using `mkfs.ubifs` and stored in the superblock node.
366 To avoid tampering with this and other values stored there, a HMAC is added to
367 the superblock struct. The superblock node is stored in LEB 0 and is only
368 modified on feature flag or similar changes, but never on file changes.
371 LPT Authentication
372 ~~~~~~~~~~~~~~~~~~
374 The location of the LPT root node on the flash is stored in the UBIFS master
375 node. Since the LPT is written and read atomically on every commit, there is
376 no need to authenticate individual nodes of the tree. It suffices to
377 protect the integrity of the full LPT by a simple hash stored in the master
378 node. Since the master node itself is authenticated, the LPTs authenticity can
379 be verified by verifying the authenticity of the master node and comparing the
380 LTP hash stored there with the hash computed from the read on-flash LPT.
383 Key Management
384 --------------
386 For simplicity, UBIFS authentication uses a single key to compute the HMACs
387 of superblock, master, commit start and reference nodes. This key has to be
388 available on creation of the filesystem (`mkfs.ubifs`) to authenticate the
389 superblock node. Further, it has to be available on mount of the filesystem
390 to verify authenticated nodes and generate new HMACs for changes.
392 UBIFS authentication is intended to operate side-by-side with UBIFS encryption
393 (fscrypt) to provide confidentiality and authenticity. Since UBIFS encryption
394 has a different approach of encryption policies per directory, there can be
395 multiple fscrypt master keys and there might be folders without encryption.
396 UBIFS authentication on the other hand has an all-or-nothing approach in the
397 sense that it either authenticates everything of the filesystem or nothing.
398 Because of this and because UBIFS authentication should also be usable without
399 encryption, it does not share the same master key with fscrypt, but manages
400 a dedicated authentication key.
402 The API for providing the authentication key has yet to be defined, but the
403 key can eg. be provided by userspace through a keyring similar to the way it
404 is currently done in fscrypt. It should however be noted that the current
405 fscrypt approach has shown its flaws and the userspace API will eventually
406 change [FSCRYPT-POLICY2].
408 Nevertheless, it will be possible for a user to provide a single passphrase
409 or key in userspace that covers UBIFS authentication and encryption. This can
410 be solved by the corresponding userspace tools which derive a second key for
411 authentication in addition to the derived fscrypt master key used for
412 encryption.
414 To be able to check if the proper key is available on mount, the UBIFS
415 superblock node will additionally store a hash of the authentication key. This
416 approach is similar to the approach proposed for fscrypt encryption policy v2
417 [FSCRYPT-POLICY2].
420 Future Extensions
421 =================
423 In certain cases where a vendor wants to provide an authenticated filesystem
424 image to customers, it should be possible to do so without sharing the secret
425 UBIFS authentication key. Instead, in addition the each HMAC a digital
426 signature could be stored where the vendor shares the public key alongside the
427 filesystem image. In case this filesystem has to be modified afterwards,
428 UBIFS can exchange all digital signatures with HMACs on first mount similar
429 to the way the IMA/EVM subsystem deals with such situations. The HMAC key
430 will then have to be provided beforehand in the normal way.
433 References
434 ==========
436 [CRYPTSETUP2] http://www.saout.de/pipermail/dm-crypt/2017-November/005745.html
438 [DMC-CBC-ATTACK] http://www.jakoblell.com/blog/2013/12/22/practical-malleability-attack-against-cbc-encrypted-luks-partitions/
440 [DM-INTEGRITY] https://www.kernel.org/doc/Documentation/device-mapper/dm-integrity.rst
442 [DM-VERITY] https://www.kernel.org/doc/Documentation/device-mapper/verity.rst
444 [FSCRYPT-POLICY2] https://www.spinics.net/lists/linux-ext4/msg58710.html
446 [UBIFS-WP] http://www.linux-mtd.infradead.org/doc/ubifs_whitepaper.pdf