| blob | 6a9584f6ff4677e8a974c1f27f7b223b405ffb25 |
1 :orphan:
3 .. UBIFS Authentication
4 .. sigma star gmbh
5 .. 2018
7 Introduction
8 ============
10 UBIFS utilizes the fscrypt framework to provide confidentiality for file
11 contents and file names. This prevents attacks where an attacker is able to
12 read contents of the filesystem on a single point in time. A classic example
13 is a lost smartphone where the attacker is unable to read personal data stored
14 on the device without the filesystem decryption key.
16 At the current state, UBIFS encryption however does not prevent attacks where
17 the attacker is able to modify the filesystem contents and the user uses the
18 device afterwards. In such a scenario an attacker can modify filesystem
19 contents arbitrarily without the user noticing. One example is to modify a
20 binary to perform a malicious action when executed [DMC-CBC-ATTACK]. Since
21 most of the filesystem metadata of UBIFS is stored in plain, this makes it
22 fairly easy to swap files and replace their contents.
24 Other full disk encryption systems like dm-crypt cover all filesystem metadata,
25 which makes such kinds of attacks more complicated, but not impossible.
26 Especially, if the attacker is given access to the device multiple points in
27 time. For dm-crypt and other filesystems that build upon the Linux block IO
28 layer, the dm-integrity or dm-verity subsystems [DM-INTEGRITY, DM-VERITY]
29 can be used to get full data authentication at the block layer.
30 These can also be combined with dm-crypt [CRYPTSETUP2].
32 This document describes an approach to get file contents _and_ full metadata
33 authentication for UBIFS. Since UBIFS uses fscrypt for file contents and file
34 name encryption, the authentication system could be tied into fscrypt such that
35 existing features like key derivation can be utilized. It should however also
36 be possible to use UBIFS authentication without using encryption.
39 MTD, UBI & UBIFS
40 ----------------
42 On Linux, the MTD (Memory Technology Devices) subsystem provides a uniform
43 interface to access raw flash devices. One of the more prominent subsystems that
44 work on top of MTD is UBI (Unsorted Block Images). It provides volume management
45 for flash devices and is thus somewhat similar to LVM for block devices. In
46 addition, it deals with flash-specific wear-leveling and transparent I/O error
47 handling. UBI offers logical erase blocks (LEBs) to the layers on top of it
48 and maps them transparently to physical erase blocks (PEBs) on the flash.
50 UBIFS is a filesystem for raw flash which operates on top of UBI. Thus, wear
51 leveling and some flash specifics are left to UBI, while UBIFS focuses on
52 scalability, performance and recoverability.
54 ::
56 +------------+ +*******+ +-----------+ +-----+
57 | | * UBIFS * | UBI-BLOCK | | ... |
58 | JFFS/JFFS2 | +*******+ +-----------+ +-----+
59 | | +-----------------------------+ +-----------+ +-----+
60 | | | UBI | | MTD-BLOCK | | ... |
61 +------------+ +-----------------------------+ +-----------+ +-----+
62 +------------------------------------------------------------------+
63 | MEMORY TECHNOLOGY DEVICES (MTD) |
64 +------------------------------------------------------------------+
65 +-----------------------------+ +--------------------------+ +-----+
66 | NAND DRIVERS | | NOR DRIVERS | | ... |
67 +-----------------------------+ +--------------------------+ +-----+
69 Figure 1: Linux kernel subsystems for dealing with raw flash
73 Internally, UBIFS maintains multiple data structures which are persisted on
74 the flash:
76 - *Index*: an on-flash B+ tree where the leaf nodes contain filesystem data
77 - *Journal*: an additional data structure to collect FS changes before updating
78 the on-flash index and reduce flash wear.
79 - *Tree Node Cache (TNC)*: an in-memory B+ tree that reflects the current FS
80 state to avoid frequent flash reads. It is basically the in-memory
81 representation of the index, but contains additional attributes.
82 - *LEB property tree (LPT)*: an on-flash B+ tree for free space accounting per
83 UBI LEB.
85 In the remainder of this section we will cover the on-flash UBIFS data
86 structures in more detail. The TNC is of less importance here since it is never
87 persisted onto the flash directly. More details on UBIFS can also be found in
88 [UBIFS-WP].
91 UBIFS Index & Tree Node Cache
92 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
94 Basic on-flash UBIFS entities are called *nodes*. UBIFS knows different types
95 of nodes. Eg. data nodes (`struct ubifs_data_node`) which store chunks of file
96 contents or inode nodes (`struct ubifs_ino_node`) which represent VFS inodes.
97 Almost all types of nodes share a common header (`ubifs_ch`) containing basic
98 information like node type, node length, a sequence number, etc. (see
99 `fs/ubifs/ubifs-media.h`in kernel source). Exceptions are entries of the LPT
100 and some less important node types like padding nodes which are used to pad
101 unusable content at the end of LEBs.
103 To avoid re-writing the whole B+ tree on every single change, it is implemented
104 as *wandering tree*, where only the changed nodes are re-written and previous
105 versions of them are obsoleted without erasing them right away. As a result,
106 the index is not stored in a single place on the flash, but *wanders* around
107 and there are obsolete parts on the flash as long as the LEB containing them is
108 not reused by UBIFS. To find the most recent version of the index, UBIFS stores
109 a special node called *master node* into UBI LEB 1 which always points to the
110 most recent root node of the UBIFS index. For recoverability, the master node
111 is additionally duplicated to LEB 2. Mounting UBIFS is thus a simple read of
112 LEB 1 and 2 to get the current master node and from there get the location of
113 the most recent on-flash index.
115 The TNC is the in-memory representation of the on-flash index. It contains some
116 additional runtime attributes per node which are not persisted. One of these is
117 a dirty-flag which marks nodes that have to be persisted the next time the
118 index is written onto the flash. The TNC acts as a write-back cache and all
119 modifications of the on-flash index are done through the TNC. Like other caches,
120 the TNC does not have to mirror the full index into memory, but reads parts of
121 it from flash whenever needed. A *commit* is the UBIFS operation of updating the
122 on-flash filesystem structures like the index. On every commit, the TNC nodes
123 marked as dirty are written to the flash to update the persisted index.
126 Journal
127 ~~~~~~~
129 To avoid wearing out the flash, the index is only persisted (*commited*) when
130 certain conditions are met (eg. ``fsync(2)``). The journal is used to record
131 any changes (in form of inode nodes, data nodes etc.) between commits
132 of the index. During mount, the journal is read from the flash and replayed
133 onto the TNC (which will be created on-demand from the on-flash index).
135 UBIFS reserves a bunch of LEBs just for the journal called *log area*. The
136 amount of log area LEBs is configured on filesystem creation (using
137 ``mkfs.ubifs``) and stored in the superblock node. The log area contains only
138 two types of nodes: *reference nodes* and *commit start nodes*. A commit start
139 node is written whenever an index commit is performed. Reference nodes are
140 written on every journal update. Each reference node points to the position of
141 other nodes (inode nodes, data nodes etc.) on the flash that are part of this
142 journal entry. These nodes are called *buds* and describe the actual filesystem
143 changes including their data.
145 The log area is maintained as a ring. Whenever the journal is almost full,
146 a commit is initiated. This also writes a commit start node so that during
147 mount, UBIFS will seek for the most recent commit start node and just replay
148 every reference node after that. Every reference node before the commit start
149 node will be ignored as they are already part of the on-flash index.
151 When writing a journal entry, UBIFS first ensures that enough space is
152 available to write the reference node and buds part of this entry. Then, the
153 reference node is written and afterwards the buds describing the file changes.
154 On replay, UBIFS will record every reference node and inspect the location of
155 the referenced LEBs to discover the buds. If these are corrupt or missing,
156 UBIFS will attempt to recover them by re-reading the LEB. This is however only
157 done for the last referenced LEB of the journal. Only this can become corrupt
158 because of a power cut. If the recovery fails, UBIFS will not mount. An error
159 for every other LEB will directly cause UBIFS to fail the mount operation.
161 ::
163 | ---- LOG AREA ---- | ---------- MAIN AREA ------------ |
165 -----+------+-----+--------+---- ------+-----+-----+---------------
166 \ | | | | / / | | | \
167 / CS | REF | REF | | \ \ DENT | INO | INO | /
168 \ | | | | / / | | | \
169 ----+------+-----+--------+--- -------+-----+-----+----------------
170 | | ^ ^
171 | | | |
172 +------------------------+ |
173 | |
174 +-------------------------------+
177 Figure 2: UBIFS flash layout of log area with commit start nodes
178 (CS) and reference nodes (REF) pointing to main area
179 containing their buds
182 LEB Property Tree/Table
183 ~~~~~~~~~~~~~~~~~~~~~~~
185 The LEB property tree is used to store per-LEB information. This includes the
186 LEB type and amount of free and *dirty* (old, obsolete content) space [1]_ on
187 the LEB. The type is important, because UBIFS never mixes index nodes with data
188 nodes on a single LEB and thus each LEB has a specific purpose. This again is
189 useful for free space calculations. See [UBIFS-WP] for more details.
191 The LEB property tree again is a B+ tree, but it is much smaller than the
192 index. Due to its smaller size it is always written as one chunk on every
193 commit. Thus, saving the LPT is an atomic operation.
196 .. [1] Since LEBs can only be appended and never overwritten, there is a
197 difference between free space ie. the remaining space left on the LEB to be
198 written to without erasing it and previously written content that is obsolete
199 but can't be overwritten without erasing the full LEB.
202 UBIFS Authentication
203 ====================
205 This chapter introduces UBIFS authentication which enables UBIFS to verify
206 the authenticity and integrity of metadata and file contents stored on flash.
209 Threat Model
210 ------------
212 UBIFS authentication enables detection of offline data modification. While it
213 does not prevent it, it enables (trusted) code to check the integrity and
214 authenticity of on-flash file contents and filesystem metadata. This covers
215 attacks where file contents are swapped.
217 UBIFS authentication will not protect against rollback of full flash contents.
218 Ie. an attacker can still dump the flash and restore it at a later time without
219 detection. It will also not protect against partial rollback of individual
220 index commits. That means that an attacker is able to partially undo changes.
221 This is possible because UBIFS does not immediately overwrites obsolete
222 versions of the index tree or the journal, but instead marks them as obsolete
223 and garbage collection erases them at a later time. An attacker can use this by
224 erasing parts of the current tree and restoring old versions that are still on
225 the flash and have not yet been erased. This is possible, because every commit
226 will always write a new version of the index root node and the master node
227 without overwriting the previous version. This is further helped by the
228 wear-leveling operations of UBI which copies contents from one physical
229 eraseblock to another and does not atomically erase the first eraseblock.
231 UBIFS authentication does not cover attacks where an attacker is able to
232 execute code on the device after the authentication key was provided.
233 Additional measures like secure boot and trusted boot have to be taken to
234 ensure that only trusted code is executed on a device.
237 Authentication
238 --------------
240 To be able to fully trust data read from flash, all UBIFS data structures
241 stored on flash are authenticated. That is:
243 - The index which includes file contents, file metadata like extended
244 attributes, file length etc.
245 - The journal which also contains file contents and metadata by recording changes
246 to the filesystem
247 - The LPT which stores UBI LEB metadata which UBIFS uses for free space accounting
250 Index Authentication
251 ~~~~~~~~~~~~~~~~~~~~
253 Through UBIFS' concept of a wandering tree, it already takes care of only
254 updating and persisting changed parts from leaf node up to the root node
255 of the full B+ tree. This enables us to augment the index nodes of the tree
256 with a hash over each node's child nodes. As a result, the index basically also
257 a Merkle tree. Since the leaf nodes of the index contain the actual filesystem
258 data, the hashes of their parent index nodes thus cover all the file contents
259 and file metadata. When a file changes, the UBIFS index is updated accordingly
260 from the leaf nodes up to the root node including the master node. This process
261 can be hooked to recompute the hash only for each changed node at the same time.
262 Whenever a file is read, UBIFS can verify the hashes from each leaf node up to
263 the root node to ensure the node's integrity.
265 To ensure the authenticity of the whole index, the UBIFS master node stores a
266 keyed hash (HMAC) over its own contents and a hash of the root node of the index
267 tree. As mentioned above, the master node is always written to the flash whenever
268 the index is persisted (ie. on index commit).
270 Using this approach only UBIFS index nodes and the master node are changed to
271 include a hash. All other types of nodes will remain unchanged. This reduces
272 the storage overhead which is precious for users of UBIFS (ie. embedded
273 devices).
275 ::
277 +---------------+
278 | Master Node |
279 | (hash) |
280 +---------------+
281 |
282 v
283 +-------------------+
284 | Index Node #1 |
285 | |
286 | branch0 branchn |
287 | (hash) (hash) |
288 +-------------------+
289 | ... | (fanout: 8)
290 | |
291 +-------+ +------+
292 | |
293 v v
294 +-------------------+ +-------------------+
295 | Index Node #2 | | Index Node #3 |
296 | | | |
297 | branch0 branchn | | branch0 branchn |
298 | (hash) (hash) | | (hash) (hash) |
299 +-------------------+ +-------------------+
300 | ... | ... |
301 v v v
302 +-----------+ +----------+ +-----------+
303 | Data Node | | INO Node | | DENT Node |
304 +-----------+ +----------+ +-----------+
307 Figure 3: Coverage areas of index node hash and master node HMAC
311 The most important part for robustness and power-cut safety is to atomically
312 persist the hash and file contents. Here the existing UBIFS logic for how
313 changed nodes are persisted is already designed for this purpose such that
314 UBIFS can safely recover if a power-cut occurs while persisting. Adding
315 hashes to index nodes does not change this since each hash will be persisted
316 atomically together with its respective node.
319 Journal Authentication
320 ~~~~~~~~~~~~~~~~~~~~~~
322 The journal is authenticated too. Since the journal is continuously written
323 it is necessary to also add authentication information frequently to the
324 journal so that in case of a powercut not too much data can't be authenticated.
325 This is done by creating a continuous hash beginning from the commit start node
326 over the previous reference nodes, the current reference node, and the bud
327 nodes. From time to time whenever it is suitable authentication nodes are added
328 between the bud nodes. This new node type contains a HMAC over the current state
329 of the hash chain. That way a journal can be authenticated up to the last
330 authentication node. The tail of the journal which may not have a authentication
331 node cannot be authenticated and is skipped during journal replay.
333 We get this picture for journal authentication::
335 ,,,,,,,,
336 ,......,...........................................
337 ,. CS , hash1.----. hash2.----.
338 ,. | , . |hmac . |hmac
339 ,. v , . v . v
340 ,.REF#0,-> bud -> bud -> bud.-> auth -> bud -> bud.-> auth ...
341 ,..|...,...........................................
342 , | ,
343 , | ,,,,,,,,,,,,,,,
344 . | hash3,----.
345 , | , |hmac
346 , v , v
347 , REF#1 -> bud -> bud,-> auth ...
348 ,,,|,,,,,,,,,,,,,,,,,,
349 v
350 REF#2 -> ...
351 |
352 V
353 ...
355 Since the hash also includes the reference nodes an attacker cannot reorder or
356 skip any journal heads for replay. An attacker can only remove bud nodes or
357 reference nodes from the end of the journal, effectively rewinding the
358 filesystem at maximum back to the last commit.
360 The location of the log area is stored in the master node. Since the master
361 node is authenticated with a HMAC as described above, it is not possible to
362 tamper with that without detection. The size of the log area is specified when
363 the filesystem is created using `mkfs.ubifs` and stored in the superblock node.
364 To avoid tampering with this and other values stored there, a HMAC is added to
365 the superblock struct. The superblock node is stored in LEB 0 and is only
366 modified on feature flag or similar changes, but never on file changes.
369 LPT Authentication
370 ~~~~~~~~~~~~~~~~~~
372 The location of the LPT root node on the flash is stored in the UBIFS master
373 node. Since the LPT is written and read atomically on every commit, there is
374 no need to authenticate individual nodes of the tree. It suffices to
375 protect the integrity of the full LPT by a simple hash stored in the master
376 node. Since the master node itself is authenticated, the LPTs authenticity can
377 be verified by verifying the authenticity of the master node and comparing the
378 LTP hash stored there with the hash computed from the read on-flash LPT.
381 Key Management
382 --------------
384 For simplicity, UBIFS authentication uses a single key to compute the HMACs
385 of superblock, master, commit start and reference nodes. This key has to be
386 available on creation of the filesystem (`mkfs.ubifs`) to authenticate the
387 superblock node. Further, it has to be available on mount of the filesystem
388 to verify authenticated nodes and generate new HMACs for changes.
390 UBIFS authentication is intended to operate side-by-side with UBIFS encryption
391 (fscrypt) to provide confidentiality and authenticity. Since UBIFS encryption
392 has a different approach of encryption policies per directory, there can be
393 multiple fscrypt master keys and there might be folders without encryption.
394 UBIFS authentication on the other hand has an all-or-nothing approach in the
395 sense that it either authenticates everything of the filesystem or nothing.
396 Because of this and because UBIFS authentication should also be usable without
397 encryption, it does not share the same master key with fscrypt, but manages
398 a dedicated authentication key.
400 The API for providing the authentication key has yet to be defined, but the
401 key can eg. be provided by userspace through a keyring similar to the way it
402 is currently done in fscrypt. It should however be noted that the current
403 fscrypt approach has shown its flaws and the userspace API will eventually
404 change [FSCRYPT-POLICY2].
406 Nevertheless, it will be possible for a user to provide a single passphrase
407 or key in userspace that covers UBIFS authentication and encryption. This can
408 be solved by the corresponding userspace tools which derive a second key for
409 authentication in addition to the derived fscrypt master key used for
410 encryption.
412 To be able to check if the proper key is available on mount, the UBIFS
413 superblock node will additionally store a hash of the authentication key. This
414 approach is similar to the approach proposed for fscrypt encryption policy v2
415 [FSCRYPT-POLICY2].
418 Future Extensions
419 =================
421 In certain cases where a vendor wants to provide an authenticated filesystem
422 image to customers, it should be possible to do so without sharing the secret
423 UBIFS authentication key. Instead, in addition the each HMAC a digital
424 signature could be stored where the vendor shares the public key alongside the
425 filesystem image. In case this filesystem has to be modified afterwards,
426 UBIFS can exchange all digital signatures with HMACs on first mount similar
427 to the way the IMA/EVM subsystem deals with such situations. The HMAC key
428 will then have to be provided beforehand in the normal way.
431 References
432 ==========
434 [CRYPTSETUP2] http://www.saout.de/pipermail/dm-crypt/2017-November/005745.html
436 [DMC-CBC-ATTACK] http://www.jakoblell.com/blog/2013/12/22/practical-malleability-attack-against-cbc-encrypted-luks-partitions/
438 [DM-INTEGRITY] https://www.kernel.org/doc/Documentation/device-mapper/dm-integrity.rst
440 [DM-VERITY] https://www.kernel.org/doc/Documentation/device-mapper/verity.rst
442 [FSCRYPT-POLICY2] https://www.spinics.net/lists/linux-ext4/msg58710.html
444 [UBIFS-WP] http://www.linux-mtd.infradead.org/doc/ubifs_whitepaper.pdf