Re: [PATCH 00/14] Add support for suppressing warning backtraces

[Guenter Roeck <linux@roeck-us.net>]

On 3/14/24 08:02, Maxime Ripard wrote:
> On Thu, Mar 14, 2024 at 07:37:13AM -0700, Guenter Roeck wrote:
>> On 3/14/24 06:36, Geert Uytterhoeven wrote:
>>> Hi Günter,
>>>
>>> On Tue, Mar 12, 2024 at 6:03 PM Guenter Roeck <linux@roeck-us.net> wrote:
>>>> Some unit tests intentionally trigger warning backtraces by passing bad
>>>> parameters to kernel API functions. Such unit tests typically check the
>>>> return value from such calls, not the existence of the warning backtrace.
>>>>
>>>> Such intentionally generated warning backtraces are neither desirable
>>>> nor useful for a number of reasons.
>>>> - They can result in overlooked real problems.
>>>> - A warning that suddenly starts to show up in unit tests needs to be
>>>>     investigated and has to be marked to be ignored, for example by
>>>>     adjusting filter scripts. Such filters are ad-hoc because there is
>>>>     no real standard format for warnings. On top of that, such filter
>>>>     scripts would require constant maintenance.
>>>>
>>>> One option to address problem would be to add messages such as "expected
>>>> warning backtraces start / end here" to the kernel log.  However, that
>>>> would again require filter scripts, it might result in missing real
>>>> problematic warning backtraces triggered while the test is running, and
>>>> the irrelevant backtrace(s) would still clog the kernel log.
>>>>
>>>> Solve the problem by providing a means to identify and suppress specific
>>>> warning backtraces while executing test code. Support suppressing multiple
>>>> backtraces while at the same time limiting changes to generic code to the
>>>> absolute minimum. Architecture specific changes are kept at minimum by
>>>> retaining function names only if both CONFIG_DEBUG_BUGVERBOSE and
>>>> CONFIG_KUNIT are enabled.
>>>>
>>>> The first patch of the series introduces the necessary infrastructure.
>>>> The second patch introduces support for counting suppressed backtraces.
>>>> This capability is used in patch three to implement unit tests.
>>>> Patch four documents the new API.
>>>> The next two patches add support for suppressing backtraces in drm_rect
>>>> and dev_addr_lists unit tests. These patches are intended to serve as
>>>> examples for the use of the functionality introduced with this series.
>>>> The remaining patches implement the necessary changes for all
>>>> architectures with GENERIC_BUG support.
>>>
>>> Thanks for your series!
>>>
>>> I gave it a try on m68k, just running backtrace-suppression-test,
>>> and that seems to work fine.
>>>
>>>> Design note:
>>>>     Function pointers are only added to the __bug_table section if both
>>>>     CONFIG_KUNIT and CONFIG_DEBUG_BUGVERBOSE are enabled to avoid image
>>>>     size increases if CONFIG_KUNIT=n. There would be some benefits to
>>>>     adding those pointers all the time (reduced complexity, ability to
>>>>     display function names in BUG/WARNING messages). That change, if
>>>>     desired, can be made later.
>>>
>>> Unfortunately this also increases kernel size in the CONFIG_KUNIT=m
>>> case (ca. 80 KiB for atari_defconfig), making it less attractive to have
>>> kunit and all tests enabled as modules in my standard kernel.
>>>
>>
>> Good point. Indeed, it does. I wanted to avoid adding a configuration option,
>> but maybe I should add it after all. How about something like this ?
>>
>> +config KUNIT_SUPPRESS_BACKTRACE
>> +       bool "KUnit - Enable backtrace suppression"
>> +       default y
>> +       help
>> +         Enable backtrace suppression for KUnit. If enabled, backtraces
>> +         generated intentionally by KUnit tests can be suppressed. Disable
>> +         to reduce kernel image size if image size is more important than
>> +         suppression of backtraces generated by KUnit tests.
>> +
> 
> How are tests using that API supposed to handle it then?
> 
> Select the config option or put an ifdef?
> 
> If the former, we end up in the same situation than without the symbol.
> If the latter, we end up in a similar situation than disabling KUNIT
> entirely, with some tests not being run which is just terrible.
> 

The API definitions are themselves within #ifdef and dummies if
KUNIT_SUPPRESS_BACKTRACE (currently CONFIG_KUNIT) is disabled.
In include/kunit/bug.h:

#ifdef CONFIG_KUNIT_SUPPRESS_BACKTRACE
...
#else
#define DEFINE_SUPPRESSED_WARNING(func)
#define START_SUPPRESSED_WARNING(func)
#define END_SUPPRESSED_WARNING(func)
#define IS_SUPPRESSED_WARNING(func) (false)
#define SUPPRESSED_WARNING_COUNT(func) (0)
#endif

Only difference to the current patch series would be

- #if IS_ENABLED(CONFIG_KUNIT)
+ #ifdef CONFIG_KUNIT_SUPPRESS_BACKTRACE

in that file and elsewhere.

With KUNIT_SUPPRESS_BACKTRACE=n you'd still get warning backtraces
triggered by the tests, but the number of tests executed as well
as the test results would still be the same.

Thanks,
Guenter