Port libkeyutils tests #34

mathstuf · 2019-07-08T02:39:33Z

Continuted work towards #32.

TODO:

fix formatting in each commit
add KDF API/tests

mathstuf · 2019-07-08T02:42:06Z

Seems that the CI containers do not have a new enough libkeyutils for keyctl_dh_compute (another reason to move off the C library and return a Rust-y version of ENOSYS instead). See #24.

mathstuf · 2019-07-08T11:52:05Z

Marking it as a weak symbol seems to be a perma-unstable solution: rust-lang/rust#29603. For now, I'll comment out the dh_compute tests and once we finish #24, they can be added back again with the appropriate ENOSYS detection logic.

mathstuf · 2019-07-30T11:50:10Z

Seems that the EPERM error occurs more often on the CI machines than locally. Looking into the kernel as to what codepath makes that happen.

mathstuf · 2019-08-02T13:34:46Z

src/api.rs

+        // Avoid a panic in the code below be ensuring that we actually have a keyring. Parsing
+        // a key's payload as a keyring payload.
+        let desc = self.description()?;
+        if desc.type_ != keytypes::Keyring::name() {


See #35 for a better way of doing this.

mathstuf · 2019-08-13T13:10:29Z

There are some tests which end up with EDQUOT (quota reached) when run in parallel. Tracking down what tests are overflowing should probably be guarded with #[serial]. Or we just commit to running "rerun tests" when it happens. It's not common, but not rare either.

The documentation states that a keyring may be given or a special keyring used. The special IDs are looked up in the requesting process' namespace. An ID of `0` means that no link is requested.

The parser for the key payload is blindly trusting that the key ID is actually a keyring and not transformed or replaced in some way. Guard against this case be ensuring that we actually have a keyring ID.

The destination keyring does not need to be the same as the one that is searched. Fix the API and remove the now-broken tests.

The API did not match what is actually possible with the kernel API. Instead, just offer two simple methods which do what is needed. Also remove the old test which doesn't match what the API looks like anymore.

This is more ergonomic and allows for just passing a plain string around.

Some tests handle it manually to test behavior once the keyring is deallocated, so leave the old functionality there.

Probably related to not having a user `1` available on the machine.

codecov · 2020-01-10T01:47:23Z

Codecov Report

Merging #34 into master will increase coverage by 22.61%.
The diff coverage is 98.91%.

@@             Coverage Diff             @@
##           master      #34       +/-   ##
===========================================
+ Coverage    67.1%   89.71%   +22.61%     
===========================================
  Files          19       33       +14     
  Lines         766     2032     +1266     
===========================================
+ Hits          514     1823     +1309     
+ Misses        252      209       -43

Impacted Files	Coverage Δ
src/constants.rs	`100% <ø> (ø)`	⬆️
src/keytypes/encrypted.rs	`0% <0%> (ø)`	⬆️
src/tests/update.rs	`100% <100%> (ø)`
src/tests/revoke.rs	`100% <100%> (ø)`
src/tests/utils/mod.rs	`100% <100%> (ø)`
src/tests/invalidate.rs	`100% <100%> (ø)`
src/tests/describe.rs	`100% <100%> (ø)`
src/tests/link.rs	`100% <100%> (ø)`
src/tests/search.rs	`100% <100%> (ø)`
src/tests/add.rs	`98.93% <100%> (+0.78%)`	⬆️
... and 24 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a5d26a4...7728eda. Read the comment docs.

mathstuf · 2020-01-10T01:47:43Z

Well, this is far better.

mathstuf force-pushed the port-libkeyutils-tests branch from 4a60aa3 to 3291db8 Compare July 9, 2019 00:59

mathstuf force-pushed the port-libkeyutils-tests branch 2 times, most recently from 99b6361 to 318a35a Compare July 30, 2019 11:45

mathstuf force-pushed the port-libkeyutils-tests branch from 421f739 to 82d2878 Compare August 2, 2019 13:14

mathstuf commented Aug 2, 2019

View reviewed changes

mathstuf force-pushed the port-libkeyutils-tests branch 2 times, most recently from a4384cb to 6fb0df5 Compare August 14, 2019 13:48

mathstuf force-pushed the port-libkeyutils-tests branch from 6fb0df5 to 2bc9508 Compare October 3, 2019 00:29

mathstuf added 18 commits October 2, 2019 22:41

tests: factor out creating an invalid keyring

87d6cd2

tests: use test names as key descriptions

e55d2a0

tests/add: add a test for treating a key as a keyring

58b940a

tests/add: whitespace nit

9891f15

tests: move test keyring creation to tests/utils

6a7de94

tests: remove unnecessary cloning

25820b0

tests: clean up test keyrings

4156e62

tests/add: migrate old test_add_key test

7d1c913

tests/clear: add tests for clearing keyrings

19cb0f5

tests/utils/kernel: add statics for the uid and gid

ee811f0

tests/utils: add a function to create an invalid key

a2a04fb

tests/describe: port description tests from libkeyutils

e33d00a

tests/utils/kernel: explicitly use the lazy_static macro

c12e3c1

tests/session: add session-based tests

b32d2e4

tests/utils: refactor out waiting for key garbage collection

d0b3225

utils/kernel: add functions for extracting the kernel version

bac112c

tests/invalidate: test key and keyring invalidation

396a24c

tests/update: port update tests from libkeyutils

7112055

mathstuf added 15 commits October 2, 2019 23:06

KeyManager: fix API to match kernel semantics

375d8c8

The documentation states that a keyring may be given or a special keyring used. The special IDs are looked up in the requesting process' namespace. An ID of `0` means that no link is requested.

tests/instantiate: add tests for key instantiation

30f0e53

Keyring::read: guard against a key-as-keyring panic

ba3637d

The parser for the key payload is blindly trusting that the key ID is actually a keyring and not transformed or replaced in some way. Guard against this case be ensuring that we actually have a keyring ID.

tests/reading: test reading keys and keyrings as each other

b283904

constants: fix a typo in the Permission docstring

317a16c

encrypted: add support for the enc32 format

2e0c1b9

tests/link: add tests for link behavior

8e054fc

tests/unlink: add tests for unlinking keyrings and keys

afa55af

Keyring: fix API for searching

fdbfb5a

The destination keyring does not need to be the same as the one that is searched. Fix the API and remove the now-broken tests.

tests/search: test searching for keys and keyrings

2c2c748

tests/permitting: add tests for permission modification

05b8ddb

api: fix request methods

87003fa

The API did not match what is actually possible with the kernel API. Instead, just offer two simple methods which do what is needed. Also remove the old test which doesn't match what the API looks like anymore.

tests/add: expand on tests for adding keyrings

8a6eeb6

api: fix the request API

bd5dcd7

This is more ergonomic and allows for just passing a plain string around.

tests: make a ScopedKeyring for handling cleanup automatically

1a70ad2

Some tests handle it manually to test behavior once the keyring is deallocated, so leave the old functionality there.

mathstuf force-pushed the port-libkeyutils-tests branch from 736ae3d to 1a70ad2 Compare October 3, 2019 03:07

mathstuf mentioned this pull request Jan 9, 2020

test: add basic encrypted key management tests #38

Open

mathstuf added 7 commits January 9, 2020 20:15

ci: run tests with a single thread

21f2f02

tests: disable session tests (for now)

7004edc

ci: update to tarpaulin 0.10.0

f39e205

clippy: add safety sections to unsafe functions

2f6daae

ci: require 1.35.0 for Option::copied

dbac0f0

tests/permitting: allow root not having chown permission too

da585d1

Probably related to not having a user `1` available on the machine.

ci: download a precompiled cargo-tarpaulin binary

7728eda

mathstuf force-pushed the port-libkeyutils-tests branch from 8b7e6d4 to 7728eda Compare January 10, 2020 01:45

mathstuf merged commit 04e4f30 into master Jan 10, 2020

mathstuf deleted the port-libkeyutils-tests branch January 10, 2020 01:47

mathstuf mentioned this pull request May 13, 2020

Improve keyutils tests #32

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Port libkeyutils tests #34

Port libkeyutils tests #34

mathstuf commented Jul 8, 2019 •

edited

Loading

mathstuf commented Jul 8, 2019

mathstuf commented Jul 8, 2019

mathstuf commented Jul 30, 2019

mathstuf Aug 2, 2019

mathstuf commented Aug 13, 2019

codecov bot commented Jan 10, 2020

mathstuf commented Jan 10, 2020

Port libkeyutils tests #34

Port libkeyutils tests #34

Conversation

mathstuf commented Jul 8, 2019 • edited Loading

mathstuf commented Jul 8, 2019

mathstuf commented Jul 8, 2019

mathstuf commented Jul 30, 2019

mathstuf Aug 2, 2019

Choose a reason for hiding this comment

mathstuf commented Aug 13, 2019

codecov bot commented Jan 10, 2020

Codecov Report

mathstuf commented Jan 10, 2020

mathstuf commented Jul 8, 2019 •

edited

Loading