RFC crypto: tcrypt: small changes to run under KUnit
Get series with b4
b4 am [email protected]Versions newest first
- v1 RFC Latest2 patches 4 repliesIn review
Patches in v1 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/2 | crypto: tcrypt: minimal conversion to run under KUnitcrypto/ crypto/ | +143 | 4 replies | |
| 2/2 | crypto: tcrypt: call KUNIT_FAIL() instead of pr_err()crypto/ | +381 | 0 replies |
Reviews trailers per patch
No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.
Cover letter [RFC v1 0/2] crypto: tcrypt: small changes to run under KUnit
tcrypt.c calls itself a "[q]uick & dirty crypto testing module." One that "will only exist until we have a better testing mechanism."
This RFC seeks to start a discussion if KUnit can fill the role of that "better testing mechanism."
As-is, these example changes don't make the test code much cleaner. But they do provide a new way of running the test that's hopefully more accessible, namely
$ ./tools/testing/kunit/kunit.py run --kunitconfig=crypto ... [16:53:16] Starting KUnit Kernel ... [16:53:19] ============================================================ [16:53:19] ======== [PASSED] tcrypt ======== [16:53:19] [PASSED] tcrypt [16:53:19] ============================================================ [16:53:19] Testing complete. 1 tests run. 0 failed. 0 crashed. 0 skipped. [16:53:19] Elapsed time: 8.806s total, 0.001s configuring, 5.764s building, 0.000s running
This series contains
* an initial patch with the boilerplate needed to run
under KUnit and track pass/fail status in the test context object instead of passing around an int.
* another change to plumb the test context object to other test cases that could previously fail w/o affecting the overall pass/fail status.
I haven't reformatted the code for now just to make the changes a bit easier to read and skim over. checkpatch.pl isn't happy about the spacing in the second patch.
== Other pros ==
If we want, we can go down the path of trying to structure the tests in more idiomatic KUnit fashion to get some benefits like
* automatically freeing memory at the end of tests when allocated using kunit_kmalloc() and friends instead of having to maintain labels
* can be made to call arbitrary cleanup functions as well
* this hopefully makes the tests more readable, at the expense of a bit more runtime overhead dynamically tracking these allocations
* doing this just for the kmalloc's and __get_free_page() directly in
* tcrypt.c saves 100+ lines of code.
* flagging test cases with KASAN issues (and eventually UBSAN)
* a bit easier way to dynamically run subsets of tests via glob
* e.g. kunit.py run 'crypto.*sha512*'
* KUnit currently only supports filtering by the suite ("crypto") right now, but we should have test-level filtering support soonish.
== Cons ==
The main cons are we'd be slightly changing how these tests are built and run with these example patches
These changes are mainly
* building the test now requires CONFIG_KUNIT
* Running insmod on the module will always return 0, even if tests failed
* the test instead prints out (K)TAP output to denote pass/fail
* this is the format kselftest uses, but it's not fully standardized
And if we eventually try to restructure the test as mentioned above:
* more disruptive changes to how the tests are run
* we'd have to move away from using the "mode" parameter
* a decent amount of code churn
Daniel Latypov (2):
crypto: tcrypt: minimal conversion to run under KUnit
crypto: tcrypt: call KUNIT_FAIL() instead of pr_err()
crypto/Kconfig | 5 +- crypto/tcrypt.c | 1063 +++++++++++++++++++++++------------------------ 2 files changed, 524 insertions(+), 544 deletions(-)
base-commit: e9338abf0e186336022293d2e454c106761f262b
Signature · 1 line
2.32.0.402.g57bb445576-goog