Commit | Line | Data |
---|---|---|
e453fa60 SH |
1 | % UBIFS Authentication |
2 | % sigma star gmbh | |
3 | % 2018 | |
4 | ||
5 | # Introduction | |
6 | ||
7 | UBIFS utilizes the fscrypt framework to provide confidentiality for file | |
8 | contents and file names. This prevents attacks where an attacker is able to | |
9 | read contents of the filesystem on a single point in time. A classic example | |
10 | is a lost smartphone where the attacker is unable to read personal data stored | |
11 | on the device without the filesystem decryption key. | |
12 | ||
13 | At the current state, UBIFS encryption however does not prevent attacks where | |
14 | the attacker is able to modify the filesystem contents and the user uses the | |
15 | device afterwards. In such a scenario an attacker can modify filesystem | |
16 | contents arbitrarily without the user noticing. One example is to modify a | |
17 | binary to perform a malicious action when executed [DMC-CBC-ATTACK]. Since | |
18 | most of the filesystem metadata of UBIFS is stored in plain, this makes it | |
19 | fairly easy to swap files and replace their contents. | |
20 | ||
21 | Other full disk encryption systems like dm-crypt cover all filesystem metadata, | |
22 | which makes such kinds of attacks more complicated, but not impossible. | |
23 | Especially, if the attacker is given access to the device multiple points in | |
24 | time. For dm-crypt and other filesystems that build upon the Linux block IO | |
25 | layer, the dm-integrity or dm-verity subsystems [DM-INTEGRITY, DM-VERITY] | |
26 | can be used to get full data authentication at the block layer. | |
27 | These can also be combined with dm-crypt [CRYPTSETUP2]. | |
28 | ||
29 | This document describes an approach to get file contents _and_ full metadata | |
30 | authentication for UBIFS. Since UBIFS uses fscrypt for file contents and file | |
31 | name encryption, the authentication system could be tied into fscrypt such that | |
32 | existing features like key derivation can be utilized. It should however also | |
33 | be possible to use UBIFS authentication without using encryption. | |
34 | ||
35 | ||
36 | ## MTD, UBI & UBIFS | |
37 | ||
38 | On Linux, the MTD (Memory Technology Devices) subsystem provides a uniform | |
39 | interface to access raw flash devices. One of the more prominent subsystems that | |
40 | work on top of MTD is UBI (Unsorted Block Images). It provides volume management | |
41 | for flash devices and is thus somewhat similar to LVM for block devices. In | |
42 | addition, it deals with flash-specific wear-leveling and transparent I/O error | |
43 | handling. UBI offers logical erase blocks (LEBs) to the layers on top of it | |
44 | and maps them transparently to physical erase blocks (PEBs) on the flash. | |
45 | ||
46 | UBIFS is a filesystem for raw flash which operates on top of UBI. Thus, wear | |
47 | leveling and some flash specifics are left to UBI, while UBIFS focuses on | |
48 | scalability, performance and recoverability. | |
49 | ||
50 | ||
51 | ||
52 | +------------+ +*******+ +-----------+ +-----+ | |
53 | | | * UBIFS * | UBI-BLOCK | | ... | | |
54 | | JFFS/JFFS2 | +*******+ +-----------+ +-----+ | |
55 | | | +-----------------------------+ +-----------+ +-----+ | |
56 | | | | UBI | | MTD-BLOCK | | ... | | |
57 | +------------+ +-----------------------------+ +-----------+ +-----+ | |
58 | +------------------------------------------------------------------+ | |
59 | | MEMORY TECHNOLOGY DEVICES (MTD) | | |
60 | +------------------------------------------------------------------+ | |
61 | +-----------------------------+ +--------------------------+ +-----+ | |
62 | | NAND DRIVERS | | NOR DRIVERS | | ... | | |
63 | +-----------------------------+ +--------------------------+ +-----+ | |
64 | ||
65 | Figure 1: Linux kernel subsystems for dealing with raw flash | |
66 | ||
67 | ||
68 | ||
69 | Internally, UBIFS maintains multiple data structures which are persisted on | |
70 | the flash: | |
71 | ||
72 | - *Index*: an on-flash B+ tree where the leaf nodes contain filesystem data | |
73 | - *Journal*: an additional data structure to collect FS changes before updating | |
74 | the on-flash index and reduce flash wear. | |
75 | - *Tree Node Cache (TNC)*: an in-memory B+ tree that reflects the current FS | |
76 | state to avoid frequent flash reads. It is basically the in-memory | |
77 | representation of the index, but contains additional attributes. | |
78 | - *LEB property tree (LPT)*: an on-flash B+ tree for free space accounting per | |
79 | UBI LEB. | |
80 | ||
81 | In the remainder of this section we will cover the on-flash UBIFS data | |
82 | structures in more detail. The TNC is of less importance here since it is never | |
83 | persisted onto the flash directly. More details on UBIFS can also be found in | |
84 | [UBIFS-WP]. | |
85 | ||
86 | ||
87 | ### UBIFS Index & Tree Node Cache | |
88 | ||
89 | Basic on-flash UBIFS entities are called *nodes*. UBIFS knows different types | |
90 | of nodes. Eg. data nodes (`struct ubifs_data_node`) which store chunks of file | |
91 | contents or inode nodes (`struct ubifs_ino_node`) which represent VFS inodes. | |
92 | Almost all types of nodes share a common header (`ubifs_ch`) containing basic | |
93 | information like node type, node length, a sequence number, etc. (see | |
94 | `fs/ubifs/ubifs-media.h`in kernel source). Exceptions are entries of the LPT | |
95 | and some less important node types like padding nodes which are used to pad | |
96 | unusable content at the end of LEBs. | |
97 | ||
98 | To avoid re-writing the whole B+ tree on every single change, it is implemented | |
99 | as *wandering tree*, where only the changed nodes are re-written and previous | |
100 | versions of them are obsoleted without erasing them right away. As a result, | |
101 | the index is not stored in a single place on the flash, but *wanders* around | |
102 | and there are obsolete parts on the flash as long as the LEB containing them is | |
103 | not reused by UBIFS. To find the most recent version of the index, UBIFS stores | |
104 | a special node called *master node* into UBI LEB 1 which always points to the | |
105 | most recent root node of the UBIFS index. For recoverability, the master node | |
106 | is additionally duplicated to LEB 2. Mounting UBIFS is thus a simple read of | |
107 | LEB 1 and 2 to get the current master node and from there get the location of | |
108 | the most recent on-flash index. | |
109 | ||
110 | The TNC is the in-memory representation of the on-flash index. It contains some | |
111 | additional runtime attributes per node which are not persisted. One of these is | |
112 | a dirty-flag which marks nodes that have to be persisted the next time the | |
113 | index is written onto the flash. The TNC acts as a write-back cache and all | |
114 | modifications of the on-flash index are done through the TNC. Like other caches, | |
115 | the TNC does not have to mirror the full index into memory, but reads parts of | |
116 | it from flash whenever needed. A *commit* is the UBIFS operation of updating the | |
117 | on-flash filesystem structures like the index. On every commit, the TNC nodes | |
118 | marked as dirty are written to the flash to update the persisted index. | |
119 | ||
120 | ||
121 | ### Journal | |
122 | ||
123 | To avoid wearing out the flash, the index is only persisted (*commited*) when | |
124 | certain conditions are met (eg. `fsync(2)`). The journal is used to record | |
125 | any changes (in form of inode nodes, data nodes etc.) between commits | |
126 | of the index. During mount, the journal is read from the flash and replayed | |
127 | onto the TNC (which will be created on-demand from the on-flash index). | |
128 | ||
129 | UBIFS reserves a bunch of LEBs just for the journal called *log area*. The | |
130 | amount of log area LEBs is configured on filesystem creation (using | |
131 | `mkfs.ubifs`) and stored in the superblock node. The log area contains only | |
132 | two types of nodes: *reference nodes* and *commit start nodes*. A commit start | |
133 | node is written whenever an index commit is performed. Reference nodes are | |
134 | written on every journal update. Each reference node points to the position of | |
135 | other nodes (inode nodes, data nodes etc.) on the flash that are part of this | |
136 | journal entry. These nodes are called *buds* and describe the actual filesystem | |
137 | changes including their data. | |
138 | ||
139 | The log area is maintained as a ring. Whenever the journal is almost full, | |
140 | a commit is initiated. This also writes a commit start node so that during | |
141 | mount, UBIFS will seek for the most recent commit start node and just replay | |
142 | every reference node after that. Every reference node before the commit start | |
143 | node will be ignored as they are already part of the on-flash index. | |
144 | ||
145 | When writing a journal entry, UBIFS first ensures that enough space is | |
146 | available to write the reference node and buds part of this entry. Then, the | |
147 | reference node is written and afterwards the buds describing the file changes. | |
148 | On replay, UBIFS will record every reference node and inspect the location of | |
149 | the referenced LEBs to discover the buds. If these are corrupt or missing, | |
150 | UBIFS will attempt to recover them by re-reading the LEB. This is however only | |
151 | done for the last referenced LEB of the journal. Only this can become corrupt | |
152 | because of a power cut. If the recovery fails, UBIFS will not mount. An error | |
153 | for every other LEB will directly cause UBIFS to fail the mount operation. | |
154 | ||
155 | ||
156 | | ---- LOG AREA ---- | ---------- MAIN AREA ------------ | | |
157 | ||
158 | -----+------+-----+--------+---- ------+-----+-----+--------------- | |
159 | \ | | | | / / | | | \ | |
160 | / CS | REF | REF | | \ \ DENT | INO | INO | / | |
161 | \ | | | | / / | | | \ | |
162 | ----+------+-----+--------+--- -------+-----+-----+---------------- | |
163 | | | ^ ^ | |
164 | | | | | | |
165 | +------------------------+ | | |
166 | | | | |
167 | +-------------------------------+ | |
168 | ||
169 | ||
170 | Figure 2: UBIFS flash layout of log area with commit start nodes | |
171 | (CS) and reference nodes (REF) pointing to main area | |
172 | containing their buds | |
173 | ||
174 | ||
175 | ### LEB Property Tree/Table | |
176 | ||
177 | The LEB property tree is used to store per-LEB information. This includes the | |
178 | LEB type and amount of free and *dirty* (old, obsolete content) space [1] on | |
179 | the LEB. The type is important, because UBIFS never mixes index nodes with data | |
180 | nodes on a single LEB and thus each LEB has a specific purpose. This again is | |
181 | useful for free space calculations. See [UBIFS-WP] for more details. | |
182 | ||
183 | The LEB property tree again is a B+ tree, but it is much smaller than the | |
184 | index. Due to its smaller size it is always written as one chunk on every | |
185 | commit. Thus, saving the LPT is an atomic operation. | |
186 | ||
187 | ||
188 | [1] Since LEBs can only be appended and never overwritten, there is a | |
189 | difference between free space ie. the remaining space left on the LEB to be | |
190 | written to without erasing it and previously written content that is obsolete | |
191 | but can't be overwritten without erasing the full LEB. | |
192 | ||
193 | ||
194 | # UBIFS Authentication | |
195 | ||
196 | This chapter introduces UBIFS authentication which enables UBIFS to verify | |
197 | the authenticity and integrity of metadata and file contents stored on flash. | |
198 | ||
199 | ||
200 | ## Threat Model | |
201 | ||
202 | UBIFS authentication enables detection of offline data modification. While it | |
203 | does not prevent it, it enables (trusted) code to check the integrity and | |
204 | authenticity of on-flash file contents and filesystem metadata. This covers | |
205 | attacks where file contents are swapped. | |
206 | ||
207 | UBIFS authentication will not protect against rollback of full flash contents. | |
208 | Ie. an attacker can still dump the flash and restore it at a later time without | |
209 | detection. It will also not protect against partial rollback of individual | |
210 | index commits. That means that an attacker is able to partially undo changes. | |
211 | This is possible because UBIFS does not immediately overwrites obsolete | |
212 | versions of the index tree or the journal, but instead marks them as obsolete | |
213 | and garbage collection erases them at a later time. An attacker can use this by | |
214 | erasing parts of the current tree and restoring old versions that are still on | |
215 | the flash and have not yet been erased. This is possible, because every commit | |
216 | will always write a new version of the index root node and the master node | |
217 | without overwriting the previous version. This is further helped by the | |
218 | wear-leveling operations of UBI which copies contents from one physical | |
219 | eraseblock to another and does not atomically erase the first eraseblock. | |
220 | ||
221 | UBIFS authentication does not cover attacks where an attacker is able to | |
222 | execute code on the device after the authentication key was provided. | |
223 | Additional measures like secure boot and trusted boot have to be taken to | |
224 | ensure that only trusted code is executed on a device. | |
225 | ||
226 | ||
227 | ## Authentication | |
228 | ||
229 | To be able to fully trust data read from flash, all UBIFS data structures | |
230 | stored on flash are authenticated. That is: | |
231 | ||
232 | - The index which includes file contents, file metadata like extended | |
233 | attributes, file length etc. | |
234 | - The journal which also contains file contents and metadata by recording changes | |
235 | to the filesystem | |
236 | - The LPT which stores UBI LEB metadata which UBIFS uses for free space accounting | |
237 | ||
238 | ||
239 | ### Index Authentication | |
240 | ||
241 | Through UBIFS' concept of a wandering tree, it already takes care of only | |
242 | updating and persisting changed parts from leaf node up to the root node | |
243 | of the full B+ tree. This enables us to augment the index nodes of the tree | |
244 | with a hash over each node's child nodes. As a result, the index basically also | |
245 | a Merkle tree. Since the leaf nodes of the index contain the actual filesystem | |
246 | data, the hashes of their parent index nodes thus cover all the file contents | |
247 | and file metadata. When a file changes, the UBIFS index is updated accordingly | |
248 | from the leaf nodes up to the root node including the master node. This process | |
249 | can be hooked to recompute the hash only for each changed node at the same time. | |
250 | Whenever a file is read, UBIFS can verify the hashes from each leaf node up to | |
251 | the root node to ensure the node's integrity. | |
252 | ||
253 | To ensure the authenticity of the whole index, the UBIFS master node stores a | |
254 | keyed hash (HMAC) over its own contents and a hash of the root node of the index | |
255 | tree. As mentioned above, the master node is always written to the flash whenever | |
256 | the index is persisted (ie. on index commit). | |
257 | ||
258 | Using this approach only UBIFS index nodes and the master node are changed to | |
259 | include a hash. All other types of nodes will remain unchanged. This reduces | |
260 | the storage overhead which is precious for users of UBIFS (ie. embedded | |
261 | devices). | |
262 | ||
263 | ||
264 | +---------------+ | |
265 | | Master Node | | |
266 | | (hash) | | |
267 | +---------------+ | |
268 | | | |
269 | v | |
270 | +-------------------+ | |
271 | | Index Node #1 | | |
272 | | | | |
273 | | branch0 branchn | | |
274 | | (hash) (hash) | | |
275 | +-------------------+ | |
276 | | ... | (fanout: 8) | |
277 | | | | |
278 | +-------+ +------+ | |
279 | | | | |
280 | v v | |
281 | +-------------------+ +-------------------+ | |
282 | | Index Node #2 | | Index Node #3 | | |
283 | | | | | | |
284 | | branch0 branchn | | branch0 branchn | | |
285 | | (hash) (hash) | | (hash) (hash) | | |
286 | +-------------------+ +-------------------+ | |
287 | | ... | ... | | |
288 | v v v | |
289 | +-----------+ +----------+ +-----------+ | |
290 | | Data Node | | INO Node | | DENT Node | | |
291 | +-----------+ +----------+ +-----------+ | |
292 | ||
293 | ||
294 | Figure 3: Coverage areas of index node hash and master node HMAC | |
295 | ||
296 | ||
297 | ||
298 | The most important part for robustness and power-cut safety is to atomically | |
299 | persist the hash and file contents. Here the existing UBIFS logic for how | |
300 | changed nodes are persisted is already designed for this purpose such that | |
301 | UBIFS can safely recover if a power-cut occurs while persisting. Adding | |
302 | hashes to index nodes does not change this since each hash will be persisted | |
303 | atomically together with its respective node. | |
304 | ||
305 | ||
306 | ### Journal Authentication | |
307 | ||
308 | The journal is authenticated too. Since the journal is continuously written | |
309 | it is necessary to also add authentication information frequently to the | |
310 | journal so that in case of a powercut not too much data can't be authenticated. | |
311 | This is done by creating a continuous hash beginning from the commit start node | |
312 | over the previous reference nodes, the current reference node, and the bud | |
313 | nodes. From time to time whenever it is suitable authentication nodes are added | |
314 | between the bud nodes. This new node type contains a HMAC over the current state | |
315 | of the hash chain. That way a journal can be authenticated up to the last | |
316 | authentication node. The tail of the journal which may not have a authentication | |
317 | node cannot be authenticated and is skipped during journal replay. | |
318 | ||
319 | We get this picture for journal authentication: | |
320 | ||
321 | ,,,,,,,, | |
322 | ,......,........................................... | |
323 | ,. CS , hash1.----. hash2.----. | |
324 | ,. | , . |hmac . |hmac | |
325 | ,. v , . v . v | |
326 | ,.REF#0,-> bud -> bud -> bud.-> auth -> bud -> bud.-> auth ... | |
327 | ,..|...,........................................... | |
328 | , | , | |
329 | , | ,,,,,,,,,,,,,,, | |
330 | . | hash3,----. | |
331 | , | , |hmac | |
332 | , v , v | |
333 | , REF#1 -> bud -> bud,-> auth ... | |
334 | ,,,|,,,,,,,,,,,,,,,,,, | |
335 | v | |
336 | REF#2 -> ... | |
337 | | | |
338 | V | |
339 | ... | |
340 | ||
341 | Since the hash also includes the reference nodes an attacker cannot reorder or | |
342 | skip any journal heads for replay. An attacker can only remove bud nodes or | |
343 | reference nodes from the end of the journal, effectively rewinding the | |
344 | filesystem at maximum back to the last commit. | |
345 | ||
346 | The location of the log area is stored in the master node. Since the master | |
347 | node is authenticated with a HMAC as described above, it is not possible to | |
348 | tamper with that without detection. The size of the log area is specified when | |
349 | the filesystem is created using `mkfs.ubifs` and stored in the superblock node. | |
350 | To avoid tampering with this and other values stored there, a HMAC is added to | |
351 | the superblock struct. The superblock node is stored in LEB 0 and is only | |
352 | modified on feature flag or similar changes, but never on file changes. | |
353 | ||
354 | ||
355 | ### LPT Authentication | |
356 | ||
357 | The location of the LPT root node on the flash is stored in the UBIFS master | |
358 | node. Since the LPT is written and read atomically on every commit, there is | |
359 | no need to authenticate individual nodes of the tree. It suffices to | |
360 | protect the integrity of the full LPT by a simple hash stored in the master | |
361 | node. Since the master node itself is authenticated, the LPTs authenticity can | |
362 | be verified by verifying the authenticity of the master node and comparing the | |
363 | LTP hash stored there with the hash computed from the read on-flash LPT. | |
364 | ||
365 | ||
366 | ## Key Management | |
367 | ||
368 | For simplicity, UBIFS authentication uses a single key to compute the HMACs | |
369 | of superblock, master, commit start and reference nodes. This key has to be | |
370 | available on creation of the filesystem (`mkfs.ubifs`) to authenticate the | |
371 | superblock node. Further, it has to be available on mount of the filesystem | |
372 | to verify authenticated nodes and generate new HMACs for changes. | |
373 | ||
374 | UBIFS authentication is intended to operate side-by-side with UBIFS encryption | |
375 | (fscrypt) to provide confidentiality and authenticity. Since UBIFS encryption | |
376 | has a different approach of encryption policies per directory, there can be | |
377 | multiple fscrypt master keys and there might be folders without encryption. | |
378 | UBIFS authentication on the other hand has an all-or-nothing approach in the | |
379 | sense that it either authenticates everything of the filesystem or nothing. | |
380 | Because of this and because UBIFS authentication should also be usable without | |
381 | encryption, it does not share the same master key with fscrypt, but manages | |
382 | a dedicated authentication key. | |
383 | ||
384 | The API for providing the authentication key has yet to be defined, but the | |
385 | key can eg. be provided by userspace through a keyring similar to the way it | |
386 | is currently done in fscrypt. It should however be noted that the current | |
387 | fscrypt approach has shown its flaws and the userspace API will eventually | |
388 | change [FSCRYPT-POLICY2]. | |
389 | ||
390 | Nevertheless, it will be possible for a user to provide a single passphrase | |
391 | or key in userspace that covers UBIFS authentication and encryption. This can | |
392 | be solved by the corresponding userspace tools which derive a second key for | |
393 | authentication in addition to the derived fscrypt master key used for | |
394 | encryption. | |
395 | ||
396 | To be able to check if the proper key is available on mount, the UBIFS | |
397 | superblock node will additionally store a hash of the authentication key. This | |
398 | approach is similar to the approach proposed for fscrypt encryption policy v2 | |
399 | [FSCRYPT-POLICY2]. | |
400 | ||
401 | ||
402 | # Future Extensions | |
403 | ||
404 | In certain cases where a vendor wants to provide an authenticated filesystem | |
405 | image to customers, it should be possible to do so without sharing the secret | |
406 | UBIFS authentication key. Instead, in addition the each HMAC a digital | |
407 | signature could be stored where the vendor shares the public key alongside the | |
408 | filesystem image. In case this filesystem has to be modified afterwards, | |
409 | UBIFS can exchange all digital signatures with HMACs on first mount similar | |
410 | to the way the IMA/EVM subsystem deals with such situations. The HMAC key | |
411 | will then have to be provided beforehand in the normal way. | |
412 | ||
413 | ||
414 | # References | |
415 | ||
416 | [CRYPTSETUP2] http://www.saout.de/pipermail/dm-crypt/2017-November/005745.html | |
417 | ||
418 | [DMC-CBC-ATTACK] http://www.jakoblell.com/blog/2013/12/22/practical-malleability-attack-against-cbc-encrypted-luks-partitions/ | |
419 | ||
f0ba4377 | 420 | [DM-INTEGRITY] https://www.kernel.org/doc/Documentation/device-mapper/dm-integrity.rst |
e453fa60 | 421 | |
f0ba4377 | 422 | [DM-VERITY] https://www.kernel.org/doc/Documentation/device-mapper/verity.rst |
e453fa60 SH |
423 | |
424 | [FSCRYPT-POLICY2] https://www.spinics.net/lists/linux-ext4/msg58710.html | |
425 | ||
426 | [UBIFS-WP] http://www.linux-mtd.infradead.org/doc/ubifs_whitepaper.pdf |