Re: kernel/trace/ftrace.c:7157:20: error: unused function 'ftrace_startup_enable'

From: Nathan Chancellor
Date: Mon Feb 14 2022 - 12:00:54 EST


On Tue, Feb 15, 2022 at 01:28:53AM +0900, Masahiro Yamada wrote:
> On Tue, Feb 15, 2022 at 12:53 AM Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
> >
> > Hi Steve,
> >
> > On Mon, Feb 14, 2022 at 10:20:00AM -0500, Steven Rostedt wrote:
> > > On Sun, 13 Feb 2022 21:03:29 +0800
> > > kernel test robot <lkp@xxxxxxxxx> wrote:
> > >
> > > > All errors (new ones prefixed by >>):
> > > >
> > > > >> kernel/trace/ftrace.c:7157:20: error: unused function 'ftrace_startup_enable' [-Werror,-Wunused-function]
> > > > static inline void ftrace_startup_enable(int command) { }
> > > > ^
> > > > 1 error generated.
> > >
> > > Strange. I always thought that static inline functions do not cause
> > > warnings when not used? Especially, since they are often in headers when
> > > things are turned off. Or is it because this is in a C file?
> >
> > With -Wunused-function, clang will warn about unused static inline
> > functions within a .c file (but not .h), whereas GCC will not warn for
> > either. The unused attribute was added to the definition of inline to
> > make clang's behavior match GCC's.
> >
> > > Is this a new warning caused by a commit, or is it a new warning because
> > > the compiler now complains about it?
> >
> > However, in commit 6863f5643dd7 ("kbuild: allow Clang to find unused
> > static inline functions for W=1 build"), Masahiro made it so that the
> > unused attribute does not get added at W=1 so that instances of unused
> > static inline functions can be caught and eliminated (or put into use,
> > if the function should have been used), hence this report.
>
> BTW, my hope was to move this warning to W=0 someday.
> (that is, remove __inline_maybe_unused entirely).
>
> I do not know how many warnings are still remaining, though.
> Is it now more difficult due to CONFIG_WERROR?

I have not done an audit in quite a while and I do not remember the
results of the last one. I don't think -Werror makes it more difficult,
it is more so just finding the time to sit down and work on it :) I
agree that it would be nice to make this the default behavior for
cleaner code, perhaps we can use this as a "good first issue" for
working with clang.

Cheers,
Nathan