Commit 87c9c16317 ("kunit: tool: add support for QEMU") on the 'next'
tree adds 'from __future__ import annotations' in 'kunit_kernel.py'.
Because it is supported on only >=3.7 Python, people using older Python
will get below error:
Traceback (most recent call last):
File "./tools/testing/kunit/kunit.py", line 20, in <module>
import kunit_kernel
File "/home/sjpark/linux/tools/testing/kunit/kunit_kernel.py", line 9
from __future__ import annotations
^
SyntaxError: future feature annotations is not defined
This commit adds a version assertion in 'kunit.py', so that people get
more explicit error message like below:
Traceback (most recent call last):
File "./tools/testing/kunit/kunit.py", line 15, in <module>
assert sys.version_info >= (3, 7), "Python version is too old"
AssertionError: Python version is too old
Signed-off-by: SeongJae Park <sjpark@amazon.de>
Acked-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
This patch addresses misleading error messages reported by kunit_tool in
two cases. First, in the case of TAP output having an incorrect header
format or missing a header, the parser used to output an error message of
'no tests run!'. Now the parser outputs an error message of 'could not
parse test results!'.
As an example:
Before:
$ ./tools/testing/kunit/kunit.py parse /dev/null
[ERROR] no tests run!
...
After:
$ ./tools/testing/kunit/kunit.py parse /dev/null
[ERROR] could not parse test results!
...
Second, in the case of TAP output with the correct header but no
tests, the parser used to output an error message of 'could not parse
test results!'. Now the parser outputs an error message of 'no tests
run!'.
As an example:
Before:
$ echo -e 'TAP version 14\n1..0' | ./tools/testing/kunit/kunit.py parse
[ERROR] could not parse test results!
After:
$ echo -e 'TAP version 14\n1..0' | ./tools/testing/kunit/kunit.py parse
[ERROR] no tests run!
Additionally, this patch also corrects the tests in kunit_tool_test.py
and adds a test to check the error in the case of TAP output with the
correct header but no tests.
Signed-off-by: Rae Moar <rmoar@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Reviewed-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Add support for the SKIP directive to kunit_tool's TAP parser.
Skipped tests now show up as such in the printed summary. The number of
skipped tests is counted, and if all tests in a suite are skipped, the
suite is also marked as skipped. Otherwise, skipped tests do affect the
suite result.
Example output:
[00:22:34] ======== [SKIPPED] example_skip ========
[00:22:34] [SKIPPED] example_skip_test # SKIP this test should be skipped
[00:22:34] [SKIPPED] example_mark_skipped_test # SKIP this test should be skipped
[00:22:34] ============================================================
[00:22:34] Testing complete. 2 tests run. 0 failed. 0 crashed. 2 skipped.
Signed-off-by: David Gow <davidgow@google.com>
Reviewed-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Note: this does not change the parser behavior at all (except for making
one error message more useful). This is just an internal refactor.
The TAP output parser currently operates over a List[str].
This works, but we only ever need to be able to "peek" at the current
line and the ability to "pop" it off.
Also, using a List means we need to wait for all the output before we
can start parsing. While this is not an issue for most tests which are
really lightweight, we do have some longer (~5 minutes) tests.
This patch introduces an LineStream wrapper class that
* Exposes a peek()/pop() interface instead of manipulating an array
* this allows us to more easily add debugging code [1]
* Can consume an input from a generator
* we can now parse results as tests are running (the parser code
currently doesn't print until the end, so no impact yet).
* Tracks the current line number to print better error messages
* Would allow us to add additional features more easily, e.g. storing
N previous lines so we can print out invalid lines in context, etc.
[1] The parsing logic is currently quite fragile.
E.g. it'll often say the kernel "CRASHED" if there's something slightly
wrong with the output format. When debugging a test that had some memory
corruption issues, it resulted in very misleading errors from the parser.
Now we could easily add this to trace all the lines consumed and why
+import inspect
...
def pop(self) -> str:
n = self._next
+ print(f'popping {n[0]}: {n[1].ljust(40, " ")}| caller={inspect.stack()[1].function}')
Example output:
popping 77: TAP version 14 | caller=parse_tap_header
popping 78: 1..1 | caller=parse_test_plan
popping 79: # Subtest: kunit_executor_test | caller=parse_subtest_header
popping 80: 1..2 | caller=parse_subtest_plan
popping 81: ok 1 - parse_filter_test | caller=parse_ok_not_ok_test_case
popping 82: ok 2 - filter_subsuite_test | caller=parse_ok_not_ok_test_case
popping 83: ok 1 - kunit_executor_test | caller=parse_ok_not_ok_test_suite
If we introduce an invalid line, we can see the parser go down the wrong path:
popping 77: TAP version 14 | caller=parse_tap_header
popping 78: 1..1 | caller=parse_test_plan
popping 79: # Subtest: kunit_executor_test | caller=parse_subtest_header
popping 80: 1..2 | caller=parse_subtest_plan
popping 81: 1..2 # this is invalid! | caller=parse_ok_not_ok_test_case
popping 82: ok 1 - parse_filter_test | caller=parse_ok_not_ok_test_case
popping 83: ok 2 - filter_subsuite_test | caller=parse_ok_not_ok_test_case
popping 84: ok 1 - kunit_executor_test | caller=parse_ok_not_ok_test_case
[ERROR] ran out of lines before end token
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Acked-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
This isn't used anywhere. While it's possible that people were manually
referencing it, the new default config (in default.config in the same
path) provides equivalent functionality.
Signed-off-by: David Gow <davidgow@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
The default .kunitconfig file is currently kept in
arch/um/configs/kunit_defconfig, but -- with the impending QEMU patch
-- will no-longer be exclusively used for UML-based kernels.
Move it alongside the other KUnit configs in
tools/testing/kunit/configs, and give it a name which matches the
existing all_tests.config and broken_on_uml.config files.
Also update the Getting Started documentation to point to the new file.
Signed-off-by: David Gow <davidgow@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Add a new kernel command-line option, 'kunit_shutdown', which allows the
user to specify that the kernel poweroff, halt, or reboot after
completing all KUnit tests; this is very handy for running KUnit tests
on UML or a VM so that the UML/VM process exits cleanly immediately
after running all tests without needing a special initramfs.
Signed-off-by: David Gow <davidgow@google.com>
Signed-off-by: Brendan Higgins <brendanhiggins@google.com>
Reviewed-by: Stephen Boyd <sboyd@kernel.org>
Tested-By: Daniel Latypov <dlatypov@google.com>
Reviewed-by: Daniel Latypov <dlatypov@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Pull KUnit updates from Shuah Khan:
"Several fixes and a new feature to support failure from dynamic
analysis tools such as UBSAN and fake ops for testing.
- a fake ops struct for testing a "free" function to complain if it
was called with an invalid argument, or caught a double-free. Most
return void and have no normal means of signalling failure (e.g.
super_operations, iommu_ops, etc.)"
* tag 'linux-kselftest-kunit-5.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
Documentation: kunit: add tips for using current->kunit_test
kunit: fix -Wunused-function warning for __kunit_fail_current_test
kunit: support failure from dynamic analysis tools
kunit: tool: make --kunitconfig accept dirs, add lib/kunit fragment
kunit: make KUNIT_EXPECT_STREQ() quote values, don't print literals
kunit: Match parenthesis alignment to improve code readability
TL;DR
$ ./tools/testing/kunit/kunit.py run --kunitconfig=lib/kunit
Per suggestion from Ted [1], we can reduce the amount of typing by
assuming a convention that these files are named '.kunitconfig'.
In the case of [1], we now have
$ ./tools/testing/kunit/kunit.py run --kunitconfig=fs/ext4
Also add in such a fragment for kunit itself so we can give that as an
example more close to home (and thus less likely to be accidentally
broken).
[1] https://lore.kernel.org/linux-ext4/YCNF4yP1dB97zzwD@mit.edu/
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
kunit_tool maintains a list of config options which are broken under
UML, which we exclude from an otherwise 'make ARCH=um allyesconfig'
build used to run all tests with the --alltests option.
Something in UML allyesconfig is causing segfaults when page poisining
is enabled (and is poisoning with a non-zero value). Previously, this
didn't occur, as allyesconfig enabled the CONFIG_PAGE_POISONING_ZERO
option, which worked around the problem by zeroing memory. This option
has since been removed, and memory is now poisoned with 0xAA, which
triggers segfaults in many different codepaths, preventing UML from
booting.
Note that we have to disable both CONFIG_PAGE_POISONING and
CONFIG_DEBUG_PAGEALLOC, as the latter will 'select' the former on
architectures (such as UML) which don't implement __kernel_map_pages().
Ideally, we'd fix this properly by tracking down the real root cause,
but since this is breaking KUnit's --alltests feature, it's worth
disabling there in the meantime so the kernel can boot to the point
where tests can actually run.
Fixes: f289041ed4 ("mm, page_poison: remove CONFIG_PAGE_POISONING_ZERO")
Signed-off-by: David Gow <davidgow@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
The first argument to namedtuple() should match the name of the type,
which wasn't the case for KconfigEntryBase.
Fixing this is enough to make mypy show no python typing errors again.
Fixes 97752c39bd ("kunit: kunit_tool: Allow .kunitconfig to disable config items")
Signed-off-by: David Gow <davidgow@google.com>
Reviewed-by: Daniel Latypov <dlatypov@google.com>
Acked-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
This is a bug that has been present since the first version of this
code.
Using [] as a default parameter is dangerous, since it's mutable.
Example using the REPL:
>>> def bad(param = []):
... param.append(len(param))
... print(param)
...
>>> bad()
[0]
>>> bad()
[0, 1]
This wasn't a concern in the past since it would just keep appending the
same values to it.
E.g. before, `args` would just grow in size like:
[mem=1G', 'console=tty']
[mem=1G', 'console=tty', mem=1G', 'console=tty']
But with now filter_glob, this is more dangerous, e.g.
run_kernel(filter_glob='my-test*') # default modified here
run_kernel() # filter_glob still applies here!
That earlier `filter_glob` will affect all subsequent calls that don't
specify `args`.
Note: currently the kunit tool only calls run_kernel() at most once, so
it's not possible to trigger any negative side-effects right now.
Fixes: 6ebf5866f2 ("kunit: tool: add Python wrappers for running KUnit tests")
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
This allows running different subsets of tests, e.g.
$ ./tools/testing/kunit/kunit.py build
$ ./tools/testing/kunit/kunit.py exec 'list*'
$ ./tools/testing/kunit/kunit.py exec 'kunit*'
This passes the "kunit_filter.glob" commandline option to the UML
kernel, which currently only supports filtering by suite name.
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Currently running tests via KUnit tool means tweaking a .kunitconfig
file, which you'd keep around locally and never commit.
This changes makes it so users can pass in a path to a kunitconfig.
One of the imagined use cases is having kunitconfig fragments in-tree
to formalize interesting sets of tests for features/subsystems, e.g.
$ ./tools/testing/kunit/kunit.py run --kunticonfig=fs/ext4/kunitconfig
For now, this hypothetical fs/ext4/kunitconfig would contain
CONFIG_KUNIT=y
CONFIG_EXT4_FS=y
CONFIG_EXT4_KUNIT_TESTS=y
At the moment, it's not hard to manually whip up this file, but as more
and more tests get added, this will get tedious.
It also opens the door to documenting how to run all the tests relevant
to a specific subsystem or feature as a simple one-liner.
This can be seen as an analogue to tools/testing/selftests/*/config
But in the case of KUnit, the tests live in the same directory as the
code-under-test, so it feels more natural to allow the kunitconfig
fragments to live anywhere. (Though, people could create a separate
directory if wanted; this patch imposes no restrictions on the path).
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Tested-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Don't use an O(nm) algorithm* and make it more readable by using a dict.
*Most obviously, it does a nested for-loop over the entire other config.
A bit more subtle, it calls .entries(), which constructs a set from the
list for _every_ outer iteration.
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Tested-by: Brendan Higgins <brendanhiggins@google.com>
Acked-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
The use of manual open() and .close() calls seems to be an attempt to
keep the contents in scope.
But Python doesn't restrict variables like that, so we can introduce new
variables inside of a `with` and use them outside.
Do so to make the code more Pythonic.
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Tested-by: Brendan Higgins <brendanhiggins@google.com>
Acked-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Use self.assertEqual/assertNotEqual() instead.
Besides being more appropriate in a unit test, it'll also give a better
error message by show the unexpected values.
Also
* Delete redundant check of exception types. self.assertRaises does this.
* s/kall/call. There's no reason to name it this way.
* This is probably a misunderstanding from the docs which uses it
since `mock.call` is in scope as `call`.
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Tested-by: Brendan Higgins <brendanhiggins@google.com>
Acked-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
LinuxSourceTree will unceremoniously crash if the user doesn't call
read_kunitconfig() first in a number of functions.
And currently every place we create an instance, the caller also calls
create_kunitconfig() and read_kunitconfig().
Move these instead into __init__() so they can't be forgotten and to
reduce copy-paste.
The https://github.com/google/pytype type-checker complained that
_config wasn't initialized. With this, kunit_tool now type checks
under both pytype and mypy.
Add an optional boolean that can be used to disable this for use cases
in the future where we might not need/want to load the config.
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Tested-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
The code to handle aggregating statuses didn't check that the status
actually got set to some non-None value.
Default the value to SUCCESS instead of adding a bunch of `is None`
checks.
This sorta follows the precedent in commit 3fc48259d5 ("kunit: Don't
fail test suites if one of them is empty").
Also slightly simplify the code and add type annotations.
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Reviewed-by: Brendan Higgins <brendanhiggins@google.com>
Tested-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
The authors of this tool were more familiar with a different
type-checker, https://github.com/google/pytype.
That's open source, but mypy seems more prevalent (and runs faster).
And unlike pytype, mypy doesn't try to infer types so it doesn't check
unanotated functions.
So annotate ~all functions in kunit tool to increase type-checking
coverage.
Note: per https://www.python.org/dev/peps/pep-0484/, `__init__()` should
be annotated as `-> None`.
Doing so makes mypy discover a number of new violations.
Exclude main() since we reuse `request` for the different types of
requests, which mypy isn't happy about.
This commit fixes all but one error, where `TestSuite.status` might be
None.
Signed-off-by: Daniel Latypov <dlatypov@google.com>
Reviewed-by: David Gow <davidgow@google.com>
Tested-by: Brendan Higgins <brendanhiggins@google.com>
Acked-by: Brendan Higgins <brendanhiggins@google.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
