1 |
W dniu czw, 16.11.2017 o godzinie 11∶19 +0100, użytkownik Michał Górny |
2 |
napisał: |
3 |
> Hi, everyone. |
4 |
> |
5 |
> Here's the updated version of GLEP 74 taking into consideration |
6 |
> the points made during the Council pre-review. |
7 |
> |
8 |
> ReST: https://dev.gentoo.org/~mgorny/tmp/glep-0074.rst |
9 |
> HTML: https://dev.gentoo.org/~mgorny/tmp/glep-0074.html |
10 |
> |
11 |
|
12 |
New changes: |
13 |
|
14 |
9d819c9 glep-0074: Disallow filenames containing whitespace |
15 |
4124b2f glep-0074: Explicitly specify UTF-8 encoding |
16 |
7f9bd9f glep-0074: Include suggestions from Daniel Campbell |
17 |
|
18 |
|
19 |
--- |
20 |
GLEP: 74 |
21 |
Title: Full-tree verification using Manifest files |
22 |
Author: Michał Górny <mgorny@g.o>, |
23 |
Robin Hugh Johnson <robbat2@g.o>, |
24 |
Ulrich Müller <ulm@g.o> |
25 |
Type: Standards Track |
26 |
Status: Draft |
27 |
Version: 1 |
28 |
Created: 2017-10-21 |
29 |
Last-Modified: 2017-11-16 |
30 |
Post-History: 2017-10-26, 2017-11-16 |
31 |
Content-Type: text/x-rst |
32 |
Requires: 59, 61 |
33 |
Replaces: 44, 58, 60 |
34 |
--- |
35 |
|
36 |
Abstract |
37 |
======== |
38 |
|
39 |
This GLEP extends the Manifest file format to cover full-tree file |
40 |
integrity and authenticity checks. The format aims to be future-proof, |
41 |
efficient and provide means of backwards compatibility. |
42 |
|
43 |
|
44 |
Motivation |
45 |
========== |
46 |
|
47 |
The Manifest files as defined by GLEP 44 [#GLEP44]_ provide the current |
48 |
means of verifying the integrity of distfiles and package files |
49 |
in Gentoo. Combined with OpenPGP signatures, they provide means to |
50 |
ensure the authenticity of the covered files. However, as noted |
51 |
in GLEP 57 [#GLEP57]_ they lack the ability to provide full-tree |
52 |
authenticity verification as they do not cover any files outside |
53 |
the package directory. In particular, they provide multiple ways |
54 |
for a third party to inject malicious code into the ebuild environment. |
55 |
|
56 |
Historically, the topic of providing authenticity coverage for the whole |
57 |
repository has been mentioned multiple times. The most noteworthy effort |
58 |
are GLEPs 58 [#GLEP58]_ and 60 [#GLEP60]_ by Robin H. Johnson from 2008. |
59 |
They were accepted by the Council in 2010 but have never been |
60 |
implemented. When potential implementation work started in 2017, a new |
61 |
discussion about the specification arose. It prompted the creation |
62 |
of a competing GLEP that would provide a redesigned alternative to |
63 |
the old GLEPs. |
64 |
|
65 |
This specification is designed with the following goals in mind: |
66 |
|
67 |
1. It should provide means to ensure the authenticity of the complete |
68 |
repository, including preventing the injection of additional files. |
69 |
|
70 |
2. The format should be universal enough to work both for the Gentoo |
71 |
repository and third-party repositories of different characteristics. |
72 |
|
73 |
3. The Manifest files should be verifiable stand-alone, that is without |
74 |
knowing any details about the underlying repository format. |
75 |
|
76 |
|
77 |
Specification |
78 |
============= |
79 |
|
80 |
Manifest file format |
81 |
-------------------- |
82 |
|
83 |
This specification reuses and extends the Manifest file format defined |
84 |
in GLEP 44 [#GLEP44]_. For the purpose of it, the *file type* field is |
85 |
repurposed as a generic *tag* that could also indicate additional |
86 |
(non-checksum) metadata. Appropriately, those tags can be followed by |
87 |
other space-separated values. |
88 |
|
89 |
Unless specified otherwise, the paths used in the Manifest files |
90 |
are relative to the directory containing the Manifest file. The paths |
91 |
must not reference the parent directory (``..``). |
92 |
|
93 |
The Manifest files use UTF-8 encoding. |
94 |
|
95 |
|
96 |
Manifest file locations and nesting |
97 |
----------------------------------- |
98 |
|
99 |
The ``Manifest`` file located in the root directory of the repository |
100 |
is called top-level Manifest, and it is used to perform the full-tree |
101 |
verification. In order to verify the authenticity, it must be signed |
102 |
using OpenPGP, using the armored cleartext format. |
103 |
|
104 |
The top-level Manifest may reference sub-Manifests contained |
105 |
in subdirectories of the repository. The sub-Manifests are traditionally |
106 |
named ``Manifest``; however, the implementation must support arbitrary |
107 |
names, including the possibility of multiple (split) Manifests |
108 |
for a single directory. The sub-Manifest can only cover the files inside |
109 |
the directory tree where it resides. |
110 |
|
111 |
The sub-Manifest can also be signed using OpenPGP armored cleartext |
112 |
format. However, the signature verification can be omitted since it |
113 |
already is covered by the signed top-level Manifest. |
114 |
|
115 |
|
116 |
Directory tree coverage |
117 |
----------------------- |
118 |
|
119 |
The specification provides three ways of skipping Manifest verification |
120 |
of specific files and directories (recursively): |
121 |
|
122 |
1. explicit ``IGNORE`` entries in Manifest files, |
123 |
|
124 |
2. injected ignore paths via package manager configuration, |
125 |
|
126 |
3. using names starting with a dot (``.``) which are always skipped. |
127 |
|
128 |
All files that are not ignored must be covered by at least one |
129 |
of the Manifests. |
130 |
|
131 |
A single file may be matched by multiple identical or equivalent |
132 |
Manifest entries, if and only if the entries have the same semantics, |
133 |
specify the same size and the checksums common to both entries match. |
134 |
It is an error for a single file to be matched by multiple entries |
135 |
of different semantics, file size or checksum values. It is an error |
136 |
to specify another entry for a file matching ``IGNORE``, or one of its |
137 |
subdirectories. |
138 |
|
139 |
The file entries (except for ``IGNORE``) can be specified for regular |
140 |
files only. Symbolic links are followed when opening files |
141 |
and traversing directories. It is an error to specify an entry for |
142 |
a different file type. If the tree contain files of other types |
143 |
that are not otherwise ignored, they need to be covered by an explicit |
144 |
``IGNORE``. |
145 |
|
146 |
All the local (non-``DIST``) files covered by a Manifest tree must |
147 |
reside on the same filesystem. It is an error to specify entries |
148 |
applying to files on another filesystem. If files or directories that |
149 |
are not otherwise ignored reside on a different filesystem, or symbolic |
150 |
links point to targets on a different filesystem, they must |
151 |
be explicitly excluded via ``IGNORE``. |
152 |
|
153 |
All paths specified in the Manifest file must consist of characters |
154 |
corresponding to valid UTF-8 code points excluding the NULL character |
155 |
(``U+0000``) and characters classified as whitespace in the current |
156 |
version of the Unicode standard [#UNICODE]_. It is an error to use |
157 |
Manifest files in directories containing files whose names contain |
158 |
the disallowed characters. |
159 |
|
160 |
|
161 |
File verification |
162 |
----------------- |
163 |
|
164 |
When verifying a file against the Manifest, the following rules are |
165 |
used: |
166 |
|
167 |
1. If the file is covered directly or indirectly by an entry |
168 |
of the ``IGNORE`` type, the verification always succeeds. |
169 |
|
170 |
2. If the file is covered by an entry of the ``MANIFEST``, ``DATA``, |
171 |
``MISC``, ``EBUILD`` or ``AUX`` type: |
172 |
|
173 |
a. if the file is not present, then the verification fails, |
174 |
|
175 |
b. if the file is present but has a different size or one |
176 |
of the checksums does not match, the verification fails, |
177 |
|
178 |
c. otherwise, the verification succeeds. |
179 |
|
180 |
3. If the file is present but not listed in Manifest, the verification |
181 |
fails. |
182 |
|
183 |
Unless specified otherwise, the package manager must not allow using |
184 |
any files for which the verification failed. The package manager may |
185 |
reject any package or even the whole repository if it may refer to files |
186 |
for which the verification failed. |
187 |
|
188 |
|
189 |
Timestamp verification |
190 |
---------------------- |
191 |
|
192 |
The top-level Manifest file can contain a ``TIMESTAMP`` entry to account |
193 |
for attacks against tree update distribution. If such an entry |
194 |
is present, it should be updated every time at least one |
195 |
of the Manifests changes. Every unique timestamp value must correspond |
196 |
to a single tree state. |
197 |
|
198 |
During the verification process, the client should compare the timestamp |
199 |
against the update time obtained from a local clock or a trusted time |
200 |
source. If the comparison result indicates that the Manifest at the time |
201 |
of receiving was already significantly outdated, the client should |
202 |
either fail the verification or require manual confirmation from user. |
203 |
|
204 |
Furthermore, the Manifest provider may employ additional methods |
205 |
of distributing the timestamps of recently generated Manifests |
206 |
using a secure channel from a trusted source for exact comparison. |
207 |
The exact details of such a solution are outside the scope of this |
208 |
specification. |
209 |
|
210 |
``TIMESTAMP`` entries may also be present in sub-Manifests. Those |
211 |
timestamps must not be newer than the timestamp of the top-level |
212 |
Manifest (if present). This specification does not define any specific |
213 |
use for them. |
214 |
|
215 |
|
216 |
Modern Manifest tags |
217 |
-------------------- |
218 |
|
219 |
The Manifest files can specify the following tags: |
220 |
|
221 |
``TIMESTAMP <iso8601>`` |
222 |
Specifies a timestamp of when the Manifest file was last updated. |
223 |
The timestamp must be a valid second-precision ISO8601 extended format |
224 |
combined date and time in UTC timezone, i.e. using the following |
225 |
``strftime()`` format string: ``%Y-%m-%dT%H:%M:%SZ``. Optional. |
226 |
The package manager can use it to detect an outdated repository |
227 |
checkout as described in `Timestamp verification`_. |
228 |
|
229 |
``MANIFEST <path> <size> <checksums>...`` |
230 |
Specifies a sub-Manifest. The sub-Manifest must be verified like |
231 |
a regular file. If the verification succeeds, the entries from |
232 |
the sub-Manifest are included for verification as described |
233 |
in `Manifest file locations and nesting`_. |
234 |
|
235 |
``IGNORE <path>`` |
236 |
Ignores a subdirectory or file from Manifest checks. If the specified |
237 |
path is present, it and its contents are omitted from the Manifest |
238 |
verification (always pass). *Path* must be a plain file or directory |
239 |
path without a trailing slash, and must not contain wildcards. |
240 |
|
241 |
``DATA <path> <size> <checksums>...`` |
242 |
Specifies a regular file subject to Manifest verification. The file |
243 |
is required to pass verification. Used for all files that do not match |
244 |
any other type. |
245 |
|
246 |
``DIST <filename> <size> <checksums>...`` |
247 |
Specifies a distfile entry used to verify files fetched as part |
248 |
of ``SRC_URI``. The filename must match the filename used to store |
249 |
the fetched file as specified in the PMS [#PMS-FETCH]_. The package |
250 |
manager must reject the fetched file if it fails verification. |
251 |
``DIST`` entries apply to all packages below the Manifest file |
252 |
specifying them. |
253 |
|
254 |
|
255 |
Deprecated Manifest tags |
256 |
------------------------ |
257 |
|
258 |
For backwards compatibility, the following tags are additionally |
259 |
allowed at the package directory level: |
260 |
|
261 |
``EBUILD <filename> <size> <checksums>...`` |
262 |
Equivalent to the ``DATA`` type. |
263 |
|
264 |
``MISC <path> <size> <checksums>...`` |
265 |
Equivalent to the ``DATA`` type. Historically indicated that |
266 |
the package manager may ignore a verification failure if operating |
267 |
in non-strict mode. However, that behavior is deprecated. |
268 |
|
269 |
``AUX <filename> <size> <checksums>...`` |
270 |
Equivalent to the ``DATA`` type, except that the filename is relative |
271 |
to ``files/`` subdirectory. |
272 |
|
273 |
|
274 |
Algorithm for full-tree verification |
275 |
------------------------------------ |
276 |
|
277 |
In order to perform full-tree verification, the following algorithm |
278 |
can be used: |
279 |
|
280 |
1. Collect all files present in the repository into *present* set. |
281 |
|
282 |
2. Start at the top-level Manifest file. Verify its OpenPGP signature. |
283 |
Optionally verify the ``TIMESTAMP`` entry if present as specified |
284 |
in `timestamp verification`. Remove the top-level Manifest |
285 |
from the *present* set. |
286 |
|
287 |
3. Process all ``MANIFEST`` entries, recursively. Verify the Manifest |
288 |
files according to `file verification`_ section, and include their |
289 |
entries in the current Manifest entry list (using paths relative |
290 |
to directories containing the Manifests). |
291 |
|
292 |
4. Process all ``IGNORE`` entries. Remove any paths matching them |
293 |
from the *present* set. |
294 |
|
295 |
5. Collect all files covered by ``DATA``, ``MISC``, ``EBUILD`` |
296 |
and ``AUX`` entries into the *covered* set. |
297 |
|
298 |
6. Verify the entries in *covered* set for incompatible duplicates |
299 |
and collisions with ignored files as explained in `Manifest file |
300 |
locations and nesting`_. |
301 |
|
302 |
7. Verify all the files in the union of the *present* and *covered* |
303 |
sets, according to `file verification`_ section. |
304 |
|
305 |
|
306 |
Algorithm for finding parent Manifests |
307 |
-------------------------------------- |
308 |
|
309 |
In order to find the top-level Manifest from the current directory |
310 |
the following algorithm can be used: |
311 |
|
312 |
1. Store the current directory as *original* and the device ID |
313 |
of the containing filesystem (``st_dev``) as *startdev*, |
314 |
|
315 |
2. If the device ID of the containing filesystem (``st_dev``) |
316 |
of the current directory is different than *startdev*, stop. |
317 |
|
318 |
3. If the current directory contains a ``Manifest`` file: |
319 |
|
320 |
a. If a ``IGNORE`` entry in the ``Manifest`` file covers |
321 |
the *original* directory (or one of the parent directories), stop. |
322 |
|
323 |
b. Otherwise, store the current directory as *last_found*. |
324 |
|
325 |
4. If the current directory is the root system directory (``/``), stop. |
326 |
|
327 |
5. Otherwise, enter the parent directory and jump to step 2. |
328 |
|
329 |
Once the algorithm stops, *last_found* will contain the relevant |
330 |
top-level Manifest. If *last_found* is null, then the directory tree |
331 |
does not contain any valid top-level Manifest candidates and one should |
332 |
be created in the *original* directory. |
333 |
|
334 |
Once the top-level Manifest is found, its ``MANIFEST`` entries should |
335 |
be used to find any sub-Manifests below the top-level Manifest, |
336 |
up to and including the *original* directory. Note that those |
337 |
sub-Manifests can use different filenames than ``Manifest``. |
338 |
|
339 |
|
340 |
Checksum algorithms |
341 |
------------------- |
342 |
|
343 |
This section is informational only. Specifying the exact set |
344 |
of supported algorithms is outside the scope of this specification. |
345 |
|
346 |
The algorithm names reserved at the time of writing are: |
347 |
|
348 |
- ``MD5`` [#MD5]_, |
349 |
- ``RMD160`` -- RIPEMD-160 [#RIPEMD160]_, |
350 |
- ``SHA1`` [#SHS]_, |
351 |
- ``SHA256`` and ``SHA512`` -- SHA-2 family of hashes [#SHS]_, |
352 |
- ``WHIRLPOOL`` [#WHIRLPOOL]_, |
353 |
- ``BLAKE2B`` and ``BLAKE2S`` -- BLAKE2 family of hashes [#BLAKE2]_, |
354 |
- ``SHA3_256`` and ``SHA3_512`` -- SHA-3 family of hashes [#SHA3]_, |
355 |
- ``STREEBOG256`` and ``STREEBOG512`` -- Streebog family of hashes |
356 |
[#STREEBOG]_. |
357 |
|
358 |
The method of introducing new hashes is defined by GLEP 59 [#GLEP59]_. |
359 |
It is recommended that any new hashes are named after the Python |
360 |
``hashlib`` module algorithm names, transformed into uppercase. |
361 |
|
362 |
|
363 |
Manifest compression |
364 |
-------------------- |
365 |
|
366 |
The topic of Manifest file compression is covered by GLEP 61 [#GLEP61]_. |
367 |
This section merely addresses interoperability issues between Manifest |
368 |
compression and this specification. |
369 |
|
370 |
The compressed Manifest files are required to be suffixed for their |
371 |
compression algorithm. This suffix should be used to recognize |
372 |
the compression and decompress Manifests transparently. The exact list |
373 |
of algorithms and their corresponding suffixes are outside the scope |
374 |
of this specification. |
375 |
|
376 |
The top-level Manifest file must not be compressed. Since the OpenPGP |
377 |
signature covers the uncompressed text and is compressed itself, |
378 |
the data would have to be decompressed without any prior verification. |
379 |
This could expose users e.g. to zip bombs or exploits on decompressor |
380 |
vulnerabilities. |
381 |
|
382 |
Whenever this specification refers to sub-Manifests, they can use any |
383 |
names but are also required to use a specific compression suffix. |
384 |
The ``MANIFEST`` entries are required to specify the full name including |
385 |
compression suffix, and the verification is performed on the compressed |
386 |
file. |
387 |
|
388 |
The specification permits uncompressed Manifests to exist alongside |
389 |
their compressed counterparts, and multiple compressed formats |
390 |
to coexist. If that is the case, the files must have the same |
391 |
uncompressed content and the specification is free to choose either |
392 |
of the files using the same base name. |
393 |
|
394 |
|
395 |
Combining multiple Manifest trees (informational) |
396 |
------------------------------------------------- |
397 |
|
398 |
This specification permits nesting multiple hierarchical Manifest trees. |
399 |
In this layout, the specific directories of the Manifest tree can |
400 |
be verified both as a part of another top-level Manifest, |
401 |
and as an independent Manifest tree (when obtained without the parent |
402 |
directory). |
403 |
|
404 |
For this to work, the sub-Manifest file in the directory must also |
405 |
satisfy the requirements for the top-level Manifest file. That is: |
406 |
|
407 |
- it must be named ``Manifest`` and not compressed, |
408 |
|
409 |
- it must cover all the files in this directory and its subdirectories |
410 |
(i.e. no files from the directory tree can be covered by parent |
411 |
Manifest), |
412 |
|
413 |
- if authenticity verification is desired, it must be OpenPGP-signed. |
414 |
|
415 |
It should be noted that if such a directory is a subdirectory of a valid |
416 |
Manifest tree, the sub-Manifest needs to be valid according |
417 |
to the top-level Manifest and the OpenPGP signature is disregarded |
418 |
as detailed in `Manifest file locations and nesting`_. The top-level |
419 |
behavior is exhibited only when the directory is obtained without parent |
420 |
directories. |
421 |
|
422 |
|
423 |
An example Manifest file (informational) |
424 |
---------------------------------------- |
425 |
|
426 |
An example top-level Manifest file for the Gentoo repository would have |
427 |
the following content:: |
428 |
|
429 |
TIMESTAMP 2017-10-30T10:11:12Z |
430 |
IGNORE distfiles |
431 |
IGNORE local |
432 |
IGNORE lost+found |
433 |
IGNORE packages |
434 |
MANIFEST app-accessibility/Manifest 14821 SHA256 1b5f.. SHA512 f7eb.. |
435 |
... |
436 |
MANIFEST eclass/Manifest.gz 50812 SHA256 8c55.. SHA512 2915.. |
437 |
... |
438 |
|
439 |
An example modern Manifest (disregarding backwards compatibility) |
440 |
for a package directory would have the following content:: |
441 |
|
442 |
DATA SphinxTrain-0.9.1-r1.ebuild 932 SHA256 3d3b.. SHA512 be4d.. |
443 |
DATA SphinxTrain-1.0.8.ebuild 912 SHA256 f681.. SHA512 0749.. |
444 |
DATA metadata.xml 664 SHA256 97c6.. SHA512 1175.. |
445 |
DATA files/gcc.patch 816 SHA256 b56e.. SHA512 2468.. |
446 |
DATA files/gcc34.patch 333 SHA256 c107.. SHA512 9919.. |
447 |
DIST SphinxTrain-0.9.1-beta.tar.gz 469617 SHA256 c1a4.. SHA512 1b33.. |
448 |
DIST sphinxtrain-1.0.8.tar.gz 8925803 SHA256 548e.. SHA512 465d.. |
449 |
|
450 |
|
451 |
Rationale |
452 |
========= |
453 |
|
454 |
Stand-alone format |
455 |
------------------ |
456 |
|
457 |
The first question that needed to be asked before proceeding with |
458 |
the design was whether the Manifest file format was supposed to be |
459 |
stand-alone, or tightly bound to the repository format. |
460 |
|
461 |
The stand-alone format has been selected because of its three |
462 |
advantages: |
463 |
|
464 |
1. It is more future-proof. If an incompatible change to the repository |
465 |
format is introduced, only developers need to upgrade the tools |
466 |
they use to generate the Manifests. The tools used to verify |
467 |
the updated Manifests will continue to work. |
468 |
|
469 |
2. It is more flexible and universal. With a dedicated tool, |
470 |
the Manifest files can be used to sign and verify arbitrary file |
471 |
sets. |
472 |
|
473 |
3. It keeps the verification tool simpler. In particular, we can easily |
474 |
write an independent verification tool that could work on any |
475 |
distribution without needing to depend on a package manager |
476 |
implementation or rewrite parts of it. |
477 |
|
478 |
Designing a stand-alone format requires that the Manifest carries enough |
479 |
information to perform the verification following all the rules specific |
480 |
to the Gentoo repository. |
481 |
|
482 |
|
483 |
Tree design |
484 |
----------- |
485 |
|
486 |
The second important point of the design was determining whether |
487 |
the Manifest files should be structured hierarchically, or independent. |
488 |
Both options have their advantages. |
489 |
|
490 |
In the hierarchical model, each sub-Manifest file is covered by a higher |
491 |
level Manifest. As a result, only the top-level Manifest has to be |
492 |
OpenPGP-signed, and subsequent Manifests need to be only verified by |
493 |
checksum stored in the parent Manifest. This has the following |
494 |
implications: |
495 |
|
496 |
- Verifying any set of files in the repository requires using checksums |
497 |
from the most relevant Manifests and the parent Manifests. |
498 |
|
499 |
- The OpenPGP signature of the top-level Manifest needs to be verified |
500 |
only once per process. |
501 |
|
502 |
- Altering any set of files requires updating the relevant Manifests, |
503 |
and their parent Manifests up to the top-level Manifest, and signing |
504 |
the last one. |
505 |
|
506 |
- As a result, the top-level Manifest changes on every commit, |
507 |
and various middle-level Manifests change (and need to be transferred) |
508 |
frequently. |
509 |
|
510 |
In the independent model, each sub-Manifest file is independent |
511 |
of the parent Manifests. As a result, each of them needs to be signed |
512 |
and verified independently. However, the parent Manifests still need |
513 |
to list sub-Manifests (albeit without verification data) in order |
514 |
to detect removal or replacement of subdirectories. This has |
515 |
the following implications: |
516 |
|
517 |
- Verifying any set of files in the repository requires using checksums |
518 |
and verifying signatures of the most relevant Manifest files. |
519 |
|
520 |
- Altering any set of files requires updating the relevant Manifests |
521 |
and signing them again. |
522 |
|
523 |
- Parent Manifests are updated only when Manifests are added or removed |
524 |
from subdirectories. As a result, they change infrequently. |
525 |
|
526 |
While both models have their advantages, the hierarchical model was |
527 |
selected because it reduces the number of OpenPGP operations |
528 |
(which are comparatively costly) to the minimum. |
529 |
|
530 |
|
531 |
Tree layout restrictions |
532 |
------------------------ |
533 |
|
534 |
The algorithm is meant to work primarily with ebuild repositories which |
535 |
normally contain only files and directories. Directories provide |
536 |
no useful metadata for verification, and specifying special entries |
537 |
for additional file types is purposeless. Therefore, the specification |
538 |
is restricted to dealing with regular files. |
539 |
|
540 |
The Gentoo repository does not use symbolic links. Some Gentoo |
541 |
repositories do, however. To provide a simple solution for dealing with |
542 |
symlinks without having to take care to implement special handling for |
543 |
them, the common behavior of implicitly resolving them is used. |
544 |
Therefore, symbolic links to files are stored as if they were regular |
545 |
files, and symbolic links to directories are followed as if they were |
546 |
regular directories. |
547 |
|
548 |
Dotfiles are implicitly ignored as that is a common notion used |
549 |
in software written for POSIX systems. All other filenames require |
550 |
explicit ``IGNORE`` lines. |
551 |
|
552 |
An ability to inject additional ignore entries is provided to account |
553 |
for site configuration affecting the repository tree -- placing |
554 |
additional files in it, skipping some of the categories from syncing. |
555 |
This configuration can extend beyond the limits of this GLEP, |
556 |
e.g. by allowing wildcards or regular expressions. |
557 |
|
558 |
The algorithm is restricted to work on a single filesystem. This is |
559 |
mostly relevant when scanning for top-level Manifest -- we do not want |
560 |
to cross filesystem boundaries then. However, to ensure consistent |
561 |
bidirectional behavior we need to also ban them when operating downwards |
562 |
the tree. |
563 |
|
564 |
The directories and files on different filesystems need to be ignored |
565 |
explicitly as implicitly skipping them would cause confusion. |
566 |
In particular, tools might then claim that a file does not exist when |
567 |
it clearly does because it was skipped due to filesystem boundaries. |
568 |
|
569 |
|
570 |
Filename character set restriction |
571 |
---------------------------------- |
572 |
|
573 |
The valid set of filename characters for the Gentoo repository |
574 |
is restricted by the devmanual 'File Naming Rules' section |
575 |
[#FILE-NAMING-RULES]_, and enforced via a git hook. The valid distfile |
576 |
names are not restricted explicitly -- however, the PMS dependency |
577 |
specification syntax [#PMS-FETCH]_ implicitly makes it impossible to use |
578 |
filenames containing whitespace. |
579 |
|
580 |
This specification aims to avoid arbitrary restrictions. For this |
581 |
reason, the filename characters are only restricted by excluding two |
582 |
technically problematic groups: |
583 |
|
584 |
1. The NULL character (``U+0000``) is normally used to indicate the end |
585 |
of a null-terminated string. Its use could therefore break programs |
586 |
written using C. Furthermore, it is not allowed in any known |
587 |
filesystem. |
588 |
|
589 |
2. The whitespace characters are used to separate Manifest fields. While |
590 |
technically it would be enough to restrict space (``U+0020``) |
591 |
character that is normally used as the separator, all whitespace |
592 |
characters are forbidden to avoid confusion and implementation |
593 |
errors. |
594 |
|
595 |
While the specification could be extended to allow such filenames |
596 |
by using some form of escaping, there is currently no apparent need |
597 |
for such a feature. |
598 |
|
599 |
Historically, Portage attempted to overcome the whitespace limitation |
600 |
by attempting to locate the size field and take everything before it |
601 |
as filename. This was terribly fragile and even if it worked, it would |
602 |
solve the problem only partially. |
603 |
|
604 |
Since the same restrictions apply to ``IGNORE`` rules, it is currently |
605 |
not possible to either list or ignore the file using whitespace |
606 |
characters. Therefore, the presence of such files is forbidden entirely. |
607 |
|
608 |
|
609 |
File verification model |
610 |
----------------------- |
611 |
|
612 |
The verification model aims to provide full coverage against different |
613 |
forms of attack. In particular, three different kinds of manipulation |
614 |
are considered: |
615 |
|
616 |
1. Alteration of the file content. |
617 |
|
618 |
2. Removal of a file. |
619 |
|
620 |
3. Addition of a new file. |
621 |
|
622 |
In order to prevent against all three, the system requires that all |
623 |
files in the repository are listed in Manifests and verified against |
624 |
them. |
625 |
|
626 |
As a special case, ignores are allowed to account for directories |
627 |
that are not part of the repository but were traditionally placed inside |
628 |
it. Those directories were ``distfiles``, ``local`` and ``packages``. It |
629 |
could be also used to ignore VCS directories such as ``CVS``. |
630 |
|
631 |
|
632 |
Non-strict Manifest verification |
633 |
-------------------------------- |
634 |
|
635 |
Originally the Manifest2 format provided a special ``MISC`` tag that |
636 |
was used for ``metadata.xml`` and ``ChangeLog`` files. This tag |
637 |
indicated that the Manifest verification failures could be ignored for |
638 |
those files unless the package manager was working in strict mode. |
639 |
|
640 |
The first versions of this specification continued the use of this tag. |
641 |
However, after a long debate it was decided to deprecate it along with |
642 |
the non-strict behavior, and require all files to strictly match. |
643 |
|
644 |
Two arguments were mentioned for the usefulness of a ``MISC`` type: |
645 |
|
646 |
1. being able to reduce the checkout size by stripping unnecessary |
647 |
files out, and |
648 |
|
649 |
2. being able to run update automatically generated files locally |
650 |
without causing unnecessary verification failures. |
651 |
|
652 |
However, the usefulness of ``MISC`` in both cases is doubtful. |
653 |
|
654 |
The cases for stripping unnecessary files mostly focused around space |
655 |
savings. For this purpose, stripping ``metadata.xml`` and similar files |
656 |
has little value. It is much more common for users to strip whole |
657 |
packages or categories. The ``MISC`` type is not suitable for that, |
658 |
and so a dedicated package manager mechanism needs to be developed |
659 |
instead. The same mechanism can also handle files that historically used |
660 |
the ``MISC`` type. As an example, the package manager may choose |
661 |
to generate both the rsync exclusion list and Manifest ignore list |
662 |
using a single source list. |
663 |
|
664 |
The cases for autogenerated files involve such cache files |
665 |
as ``use.local.desc``. However, we can not include ``md5-cache`` there |
666 |
due to security concerns which results in inconsistent cache handling. |
667 |
Furthermore, the tools were historically modified to provide stable |
668 |
output which means that their content can not change without |
669 |
a non-``MISC`` content being changed first. This practically defeats |
670 |
the purpose of using ``MISC``. |
671 |
|
672 |
Finally, the non-strict mode could be used as means to an attack. |
673 |
The allowance of missing or modified documentation file could be used |
674 |
to spread misinformation, resulting in bad decisions made by the user. |
675 |
A modified file could also be used, e.g. to exploit vulnerabilities |
676 |
of an XML parser. |
677 |
|
678 |
|
679 |
Timestamp field |
680 |
--------------- |
681 |
|
682 |
The top-level Manifest optionally allows using a ``TIMESTAMP`` tag |
683 |
to include a generation timestamp in the Manifest. A similar feature |
684 |
was originally proposed in GLEP 58 [#GLEP58]_. |
685 |
|
686 |
A malicious third-party may use the principles of exclusion or replay |
687 |
[#C08]_ to deny an update to clients, while at the same time recording |
688 |
the identity of clients to attack. The timestamp field can be used to |
689 |
detect that. |
690 |
|
691 |
In order to provide more complete protection, the Gentoo Infrastructure |
692 |
should provide an ability to obtain the timestamps of all Manifests |
693 |
from a recent timeframe over a secure channel from a trusted source |
694 |
for comparison. |
695 |
|
696 |
Strictly speaking, this information is already provided by the various |
697 |
``metadata/timestamp*`` files that are already present. However, |
698 |
including the value in the Manifest itself has a little cost |
699 |
and provides the ability to perform the verification stand-alone. |
700 |
|
701 |
Furthermore, some of the timestamp files are added very late |
702 |
in the distribution process, past the Manifest generation phase. Those |
703 |
files will most likely receive ``IGNORE`` entries and therefore |
704 |
be unsafe to use. |
705 |
|
706 |
The specification permits additional timestamps in sub-Manifest files |
707 |
for local use. A generic testing tool should ignore them. |
708 |
|
709 |
|
710 |
New vs deprecated tags |
711 |
---------------------- |
712 |
|
713 |
Out of the four types defined by Manifest2, only one is reused |
714 |
and the remaining three are replaced by a single, universal ``DATA`` |
715 |
type. |
716 |
|
717 |
The ``DIST`` tag is reused since the specification does not change |
718 |
anything with regard to distfile handling. |
719 |
|
720 |
The ``EBUILD`` tag could potentially be reused for generic file |
721 |
verification data. However, it would be confusing if all the different |
722 |
data files were marked as ``EBUILD``. Therefore, an equivalent ``DATA`` |
723 |
type was introduced as a replacement. |
724 |
|
725 |
The ``MISC`` tag and the relevant non-strict mode has been removed |
726 |
as being of little value, as detailed in the `Non-strict Manifest |
727 |
verification`_ section. |
728 |
|
729 |
The ``AUX`` tag is deprecated as it is redundant to ``DATA``, and has |
730 |
the limiting property of implicit ``files/`` path prefix. |
731 |
|
732 |
|
733 |
Finding top-level Manifest |
734 |
-------------------------- |
735 |
|
736 |
The development of a reference implementation for this GLEP has brought |
737 |
the following problem: how to find all the relevant Manifests when |
738 |
the Manifest tool is run inside a subdirectory of the repository? |
739 |
|
740 |
One of the options would be to provide a bi-directional linking |
741 |
of Manifests via a ``PARENT`` tag. However, that would not solve |
742 |
the problem when a new Manifest file is being created. |
743 |
|
744 |
Instead, an algorithm for iterating over parent directories is proposed. |
745 |
Since there is no obligatory explicit indicator for the top-level |
746 |
Manifest, the algorithm assumes that the top-level Manifest |
747 |
is the highest ``Manifest`` in the directory hierarchy that can cover |
748 |
the current directory. This generally makes sense since the Manifest |
749 |
files are required to provide coverage for all subdirectories, so all |
750 |
Manifests starting from that one need to be updated. |
751 |
|
752 |
If independent Manifest trees are nested in the directory structure, |
753 |
then an ``IGNORE`` entry needs to be used to separate them. |
754 |
|
755 |
Since sub-Manifests can use any filenames, the Manifest finding |
756 |
algorithm must not short-cut the procedure by storing all ``Manifest`` |
757 |
files along the parent directories. Instead, it needs to retrace |
758 |
the relevant sub-Manifest files along ``MANIFEST`` entries |
759 |
in the top-level Manifest. |
760 |
|
761 |
|
762 |
Injecting ChangeLogs into the checkout |
763 |
-------------------------------------- |
764 |
|
765 |
One of the problems considered in the new Manifest format was injecting |
766 |
historical and autogenerated ChangeLog into the repository. We normally |
767 |
don't include those files, to reduce the checkout size. However, some |
768 |
users have shown interest in them and Infra is working on providing them |
769 |
via an additional rsync module. |
770 |
|
771 |
If such files were injected into the repository, they would cause |
772 |
verification failures of Manifests. To account for this, Infra could |
773 |
provide ``IGNORE`` entries to allow them to exist. |
774 |
|
775 |
|
776 |
Splitting distfile checksums from file checksums |
777 |
------------------------------------------------ |
778 |
|
779 |
Another problem with the current Manifest format is that the checksums |
780 |
for fetched files are combined with checksums for local files |
781 |
in a single file inside the package directory. It has been specifically |
782 |
pointed out that: |
783 |
|
784 |
- since distfiles are sometimes reused across different packages, |
785 |
the repeating checksums are redundant [#DIST]_. |
786 |
|
787 |
- mirror admins were interested in the possibility of verifying all |
788 |
the distfiles with a single tool. |
789 |
|
790 |
This specification does not provide a clean solution to this problem. |
791 |
It technically permits moving ``DIST`` entries to higher-level Manifests |
792 |
but the usefulness of such a solution is doubtful. |
793 |
|
794 |
However, for the second problem we will probably deliver a dedicated |
795 |
tool working with this Manifest format. |
796 |
|
797 |
|
798 |
Hash algorithms |
799 |
--------------- |
800 |
|
801 |
While maintaining a consistent supported hash set is important |
802 |
for interoperability, it is not a good fit for the generic layout |
803 |
of this GLEP. Furthermore, it would require updating the GLEP |
804 |
in the future every time the used algorithms change. |
805 |
|
806 |
Instead, the specification focuses on listing the currently used |
807 |
algorithm names for interoperability, and sets a recommendation |
808 |
for consistent naming of algorithms in the future. The Python |
809 |
``hashlib`` module is used as a reference since it is used |
810 |
as the provider of hash functions for most of the Python software, |
811 |
including Portage and PkgCore. |
812 |
|
813 |
The basic rules for changing hash algorithms are defined in GLEP 59 |
814 |
[#GLEP59]_. The implementations can focus only on those algorithms |
815 |
that are actually used or planned on being used. It may be feasible |
816 |
to devise a new GLEP that specifies the currently used hashes (or update |
817 |
GLEP 59 accordingly). |
818 |
|
819 |
|
820 |
Manifest compression |
821 |
-------------------- |
822 |
|
823 |
The support for Manifest compression is introduced with minimal changes |
824 |
to the file format. The ``MANIFEST`` entries are required to provide |
825 |
the real (compressed) file path for compatibility with other file |
826 |
entries and to avoid confusion. |
827 |
|
828 |
The compression of top-level Manifest file has been prohibited |
829 |
as the specification currently does not provide any means of verifying |
830 |
the file prior to decompression. If the top-level Manifest is |
831 |
compressed, tooling will have to unpack the file before being able |
832 |
to verify the contents. This makes it possible for a malicious third |
833 |
party to attack the system by providing a compressed Manifest that |
834 |
exposes decompressor vulnerabilities, or a zip bomb. |
835 |
|
836 |
The OpenPGP cleartext signature covers the contents of the Manifest, |
837 |
and is therefore compressed along with them. The possibility of using |
838 |
detached signature has been considered but it was rejected as |
839 |
unnecessary complexity for minor gain. |
840 |
|
841 |
Technically, a similar result could be effected via moving all the data |
842 |
into a compressed sub-Manifest in the top directory (e.g. |
843 |
``Manifest.sub.gz``), and including a ``MANIFEST`` entry for this file |
844 |
in a signed, uncompressed top-level Manifest. |
845 |
|
846 |
The existence of additional entries for uncompressed Manifest checksums |
847 |
was debated. However, plain entries for the uncompressed file would |
848 |
be confusing if only the compressed file existed, and conflicting |
849 |
if both uncompressed and compressed variants existed. Furthermore, |
850 |
it has been pointed out that ``DIST`` entries do not have uncompressed |
851 |
variant either. |
852 |
|
853 |
|
854 |
Performance considerations |
855 |
-------------------------- |
856 |
|
857 |
Performing a full-tree verification on every sync raises some |
858 |
performance concerns for end-user systems. The initial testing has shown |
859 |
that a cold-cache verification on a btrfs file system can take up around |
860 |
4 minutes, with the process being mostly I/O bound. On the other hand, |
861 |
it can be expected that the verification will be performed directly |
862 |
after syncing, taking advantage of a warm filesystem cache. |
863 |
|
864 |
To improve speed on I/O and/or CPU-restrained systems even further, |
865 |
the algorithms can be easily extended to perform incremental |
866 |
verification. Given that rsync does not preserve mtimes by default, |
867 |
the tool can take advantage of mtime and Manifest comparisons to recheck |
868 |
only the parts of the repository that have changed. |
869 |
|
870 |
Furthermore, the package manager implementations can restrict checking |
871 |
only to the parts of the repository that are actually being used. |
872 |
|
873 |
|
874 |
Backwards Compatibility |
875 |
======================= |
876 |
|
877 |
This GLEP provides optional means of preserving backwards compatibility. |
878 |
To preserve the backwards compatibility, the following needs to hold |
879 |
for the ``Manifest`` file in every package directory: |
880 |
|
881 |
- all files must be covered by the single ``Manifest`` file, |
882 |
|
883 |
- all distfiles used by the package must be included, |
884 |
|
885 |
- all files inside the ``files/`` subdirectory need to use |
886 |
the ``AUX`` tag (rather than ``DATA``), |
887 |
|
888 |
- all ``.ebuild`` files need to use the ``EBUILD`` tag, |
889 |
|
890 |
- the ``metadata.xml`` and ``ChangeLog`` files need to use |
891 |
the ``MISC`` tag, |
892 |
|
893 |
- the Manifest can be signed to provide authenticity verification, |
894 |
|
895 |
- an uncompressed Manifest must always exist, and a compressed Manifest |
896 |
of identical content may be present. |
897 |
|
898 |
Once the backwards compatibility is no longer a concern, the above |
899 |
no longer needs to hold and the deprecated tags can be removed. |
900 |
|
901 |
|
902 |
Reference Implementation |
903 |
======================== |
904 |
|
905 |
The reference implementation for this GLEP is being developed |
906 |
as the gemato project [#GEMATO]_. |
907 |
|
908 |
|
909 |
Credits |
910 |
======= |
911 |
|
912 |
Thanks to all the people whose contributions were invaluable |
913 |
to the creation of this GLEP. This includes but is not limited to: |
914 |
|
915 |
- Robin Hugh Johnson, |
916 |
- Ulrich Müller. |
917 |
|
918 |
Additionally, thanks to Robin Hugh Johnson for the original |
919 |
MetaManifest GLEP series which served both as inspiration and source |
920 |
of many concepts used in this GLEP. Recursively, also thanks to all |
921 |
the people who contributed to the original GLEPs. |
922 |
|
923 |
|
924 |
References |
925 |
========== |
926 |
|
927 |
.. [#GLEP44] GLEP 44: Manifest2 format |
928 |
(https://www.gentoo.org/glep/glep-0044.html) |
929 |
|
930 |
.. [#GLEP57] GLEP 57: Security of distribution of Gentoo software |
931 |
- Overview |
932 |
(https://www.gentoo.org/glep/glep-0057.html) |
933 |
|
934 |
.. [#GLEP58] GLEP 58: Security of distribution of Gentoo software |
935 |
- Infrastructure to User distribution - MetaManifest |
936 |
(https://www.gentoo.org/glep/glep-0058.html) |
937 |
|
938 |
.. [#GLEP59] GLEP 59: Manifest2 hash policies and security implications |
939 |
(https://www.gentoo.org/glep/glep-0059.html) |
940 |
|
941 |
.. [#GLEP60] GLEP 60: Manifest2 filetypes |
942 |
(https://www.gentoo.org/glep/glep-0060.html) |
943 |
|
944 |
.. [#GLEP61] GLEP 61: Manifest2 compression |
945 |
(https://www.gentoo.org/glep/glep-0061.html) |
946 |
|
947 |
.. [#UNICODE] The Unicode standard |
948 |
(https://unicode.org/versions/latest/) |
949 |
|
950 |
.. [#PMS-FETCH] Package Manager Specification: Dependency Specification |
951 |
Format - SRC_URI |
952 |
(https://projects.gentoo.org/pms/6/pms.html#x1-940008.2.10) |
953 |
|
954 |
.. [#FILE-NAMING-RULES] Ebuild File Format -- Gentoo Development Guide |
955 |
(https://devmanual.gentoo.org/ebuild-writing/file-format/#file-naming-rules) |
956 |
|
957 |
.. [#MD5] RFC1321: The MD5 Message-Digest Algorithm |
958 |
(https://www.ietf.org/rfc/rfc1321.txt) |
959 |
|
960 |
.. [#RIPEMD160] The hash function RIPEMD-160 |
961 |
(https://homes.esat.kuleuven.be/~bosselae/ripemd160.html) |
962 |
|
963 |
.. [#SHS] FIPS PUB 180-4: Secure Hash Standard (SHS) |
964 |
(http://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.180-4.pdf) |
965 |
|
966 |
.. [#WHIRLPOOL] The WHIRLPOOL Hash Function |
967 |
(http://www.larc.usp.br/~pbarreto/WhirlpoolPage.html) |
968 |
|
969 |
.. [#BLAKE2] BLAKE2 -- fast secure hashing |
970 |
(https://blake2.net/) |
971 |
|
972 |
.. [#SHA3] FIPS PUB 202: SHA-3 Standard: Permutation-Based Hash |
973 |
and Extendable-Output Functions |
974 |
(http://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf) |
975 |
|
976 |
.. [#STREEBOG] GOST R 34.11-2012: Streebog Hash Function |
977 |
(https://www.streebog.net/) |
978 |
|
979 |
.. [#C08] Cappos, J et al. (2008). "Attacks on Package Managers" |
980 |
(https://www2.cs.arizona.edu/stork/packagemanagersecurity/attacks-on-package-managers.html) |
981 |
|
982 |
.. [#DIST] According to Robin H. Johnson, 8.4% of all DIST entries |
983 |
at the time of writing are duplicate, representing a 2 MiB |
984 |
out of 25 MiB of DIST entries altogether. |
985 |
|
986 |
.. [#GEMATO] gemato: Gentoo Manifest Tool |
987 |
(https://github.com/mgorny/gemato/) |
988 |
|
989 |
Copyright |
990 |
========= |
991 |
This work is licensed under the Creative Commons Attribution-ShareAlike 3.0 |
992 |
Unported License. To view a copy of this license, visit |
993 |
http://creativecommons.org/licenses/by-sa/3.0/. |
994 |
|
995 |
-- |
996 |
Best regards, |
997 |
Michał Górny |