Re: [PATCH] kunit: Add gnu_printf specifiers

From: David Gow
Date: Thu May 13 2021 - 16:48:37 EST


On Fri, May 14, 2021 at 4:25 AM Daniel Latypov <dlatypov@xxxxxxxxxx> wrote:
>
> On Thu, May 13, 2021 at 1:03 PM 'David Gow' via KUnit Development
> <kunit-dev@xxxxxxxxxxxxxxxx> wrote:
> >
> > Some KUnit functions use variable arguments to implement a printf-like
> > format string. Use the __printf() attribute to let the compiler warn if
> > invalid format strings are passed in.
> >
> > If the kernel is build with W=1, it complained about the lack of these
> > specifiers, e.g.:
> > ../lib/kunit/test.c:72:2: warning: function ‘kunit_log_append’ might be a candidate for ‘gnu_printf’ format attribute [-Wsuggest-attribute=format]
> >
> > Signed-off-by: David Gow <davidgow@xxxxxxxxxx>
>
> Reviewed-by: Daniel Latypov <dlatypov@xxxxxxxxxx>
>
> As noted below, these additions don't really do anything.
> Unfortunately, they just make compiler warnings noisier in the case of
> kunit_log_append().
>
> But if this silences a W=1 warning, then we should probably add them in.
> I guess it also serves as documentation that we're using the same
> standard format specifiers and not something custom, which is nice.
>

Yeah: I did this to get rid of the W=1 warnings. I don't know if
there's a way of doing this which would be less verbose: I do think
that the format checking is worthwhile in general, even if we're
hitting a few nasty cases where things are nested in macros.


> > ---
> > include/kunit/test.h | 2 +-
> > lib/kunit/string-stream.h | 6 +++---
> > 2 files changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/include/kunit/test.h b/include/kunit/test.h
> > index 49601c4b98b8..af2e386b867c 100644
> > --- a/include/kunit/test.h
> > +++ b/include/kunit/test.h
> > @@ -610,7 +610,7 @@ static inline void *kunit_kzalloc(struct kunit *test, size_t size, gfp_t gfp)
> >
> > void kunit_cleanup(struct kunit *test);
> >
> > -void kunit_log_append(char *log, const char *fmt, ...);
> > +void __printf(2, 3) kunit_log_append(char *log, const char *fmt, ...);
>
> Before this patch:
> ../lib/kunit/kunit-example-test.c: In function ‘example_simple_test’:
> ../include/linux/kern_levels.h:5:18: warning: format ‘%s’ expects
> argument of type ‘char *’, but argument 3 has type ‘int’ [-Wformat=]
> 5 | #define KERN_SOH "\001" /* ASCII Start Of Header */
> | ^~~~~~
> ../include/kunit/test.h:622:10: note: in definition of macro ‘kunit_log’
> 622 | printk(lvl fmt, ##__VA_ARGS__); \
> | ^~~
> ../include/kunit/test.h:641:2: note: in expansion of macro ‘kunit_printk’
> 641 | kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
> | ^~~~~~~~~~~~
> ../include/linux/kern_levels.h:14:19: note: in expansion of macro ‘KERN_SOH’
> 14 | #define KERN_INFO KERN_SOH "6" /* informational */
> | ^~~~~~~~
> ../include/kunit/test.h:641:15: note: in expansion of macro ‘KERN_INFO’
> 641 | kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
> | ^~~~~~~~~
> ../lib/kunit/kunit-example-test.c:23:2: note: in expansion of macro ‘kunit_info’
> 23 | kunit_info(test, "invalid: %s", 42);
>
> After this patch, it gets noisier:
> In file included from ../lib/kunit/kunit-example-test.c:9:
> ../lib/kunit/kunit-example-test.c: In function ‘example_simple_test’:
> ../include/linux/kern_levels.h:5:18: warning: format ‘%s’ expects
> argument of type ‘char *’, but argument 3 has type ‘int’ [-Wformat=]
> 5 | #define KERN_SOH "\001" /* ASCII Start Of Header */
> | ^~~~~~
> ../include/kunit/test.h:622:10: note: in definition of macro ‘kunit_log’
> 622 | printk(lvl fmt, ##__VA_ARGS__); \
> | ^~~
> ../include/kunit/test.h:641:2: note: in expansion of macro ‘kunit_printk’
> 641 | kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
> | ^~~~~~~~~~~~
> ../include/linux/kern_levels.h:14:19: note: in expansion of macro ‘KERN_SOH’
> 14 | #define KERN_INFO KERN_SOH "6" /* informational */
> | ^~~~~~~~
> ../include/kunit/test.h:641:15: note: in expansion of macro ‘KERN_INFO’
> 641 | kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
> | ^~~~~~~~~
> ../lib/kunit/kunit-example-test.c:23:2: note: in expansion of macro ‘kunit_info’
> 23 | kunit_info(test, "invalid: %s", 42);
> | ^~~~~~~~~~
> ../include/kunit/test.h:105:31: warning: format ‘%s’ expects argument
> of type ‘char *’, but argument 4 has type ‘int’ [-Wformat=]
> 105 | #define KUNIT_SUBTEST_INDENT " "
> | ^~~~~~
> ../include/kunit/test.h:623:42: note: in definition of macro ‘kunit_log’
> 623 | kunit_log_append((test_or_suite)->log, fmt "\n", \
> | ^~~
> ../include/kunit/test.h:628:23: note: in expansion of macro
> ‘KUNIT_SUBTEST_INDENT’
> 628 | kunit_log(lvl, test, KUNIT_SUBTEST_INDENT "# %s: " fmt, \
> | ^~~~~~~~~~~~~~~~~~~~
> ../include/kunit/test.h:641:2: note: in expansion of macro ‘kunit_printk’
> 641 | kunit_printk(KERN_INFO, test, fmt, ##__VA_ARGS__)
> | ^~~~~~~~~~~~
> ../lib/kunit/kunit-example-test.c:23:2: note: in expansion of macro ‘kunit_info’
> 23 | kunit_info(test, "invalid: %s", 42);
> | ^~~~~~~~~~
>
>

Yeah: that is pretty ugly. TBH, it was pretty ugly beforehand, and
this does make it worse. I guess that's the price we pay for having so
many nested macros, as well.
Personally, I suspect this is still worth it to get rid of the
compiler warnings, but only just.

> >
> > /*
> > * printk and log to per-test or per-suite log buffer. Logging only done
> > diff --git a/lib/kunit/string-stream.h b/lib/kunit/string-stream.h
> > index fe98a00b75a9..5e94b623454f 100644
> > --- a/lib/kunit/string-stream.h
> > +++ b/lib/kunit/string-stream.h
> > @@ -35,9 +35,9 @@ struct string_stream *alloc_string_stream(struct kunit *test, gfp_t gfp);
> > int __printf(2, 3) string_stream_add(struct string_stream *stream,
> > const char *fmt, ...);
> >
> > -int string_stream_vadd(struct string_stream *stream,
> > - const char *fmt,
> > - va_list args);
> > +int __printf(2, 0) string_stream_vadd(struct string_stream *stream,
> > + const char *fmt,
> > + va_list args);
>
> This is never called with a literal `fmt` string.
> It's currently only ever called through the _add variant, which does
> have __printf(2,3).
>
> So this can't catch any mistakes currently.
> And I think it's hard to imagine we'd ever pass in a literal format
> string w/ a va_list.
>

Yeah: I was tempted to leave this one out, but it was triggering
warnings with the "you should use __printf()" heuristic. In fact, it
had two warnings.
The __printf() specifier documentation does specifically call out
cases where a va_list is passed in as a case to use '0' for the
positional argument, but only the format string is checked for
validity: there's no typechecking.

> >
> > char *string_stream_get_string(struct string_stream *stream);
> >
> > --
> > 2.31.1.751.gd2f1c929bd-goog
> >
> > --
> > You received this message because you are subscribed to the Google Groups "KUnit Development" group.
> > To unsubscribe from this group and stop receiving emails from it, send an email to kunit-dev+unsubscribe@xxxxxxxxxxxxxxxx.
> > To view this discussion on the web visit https://groups.google.com/d/msgid/kunit-dev/20210513200350.854429-1-davidgow%40google.com.