Re: [PATCH 3/4] tracing: Remove redundant module.h includes

[Tom Zanussi <tom.zanussi@linux.intel.com>]

On Fri, 2015-04-10 at 12:27 -0400, Steven Rostedt wrote:
> On Fri, 10 Apr 2015 11:05:02 -0500
> Tom Zanussi <tom.zanussi@linux.intel.com> wrote:
> 
> > kallsyms.h now includes module.h, so remove module.h includes that
> > were apparently there only to satisfy kallsyms use of MODULE_NAME_LEN.
> 
> Only remove it for those that need it for MODULE_NAME_LEN because it
> was used in KSYM_SYMBOL_LEN. But not the rest. Lots of these files use
> module code directly. Now you made it require kallsyms.h to get to that
> code. If for some reason we no longer need kallsyms.h, or that
> kallsyms.h no longer needs module.h, these files will not build.
> 
> If code uses module code, it must include module.h.
> 

I had grepped all those files for 'module' and didn't find anything
other than comments.  The only files of those affected that use
KSYM_SYMBOL_LEN are trace_output and trace_syscalls, so I can respin the
patch including just those.

But I don't see that any of the other files really use module code, and
looking at those again, none of them seem to need kallsyms either so
they could remove those includes as well...  But I agree all that that
probably doesn't belong as part of this patchset.

> So NACK for this patch.
> 
> To see what needs module.h, before reverting this patch, remove the
> module.h from kallsyms.h, and also remove the MODULE_NAME_LEN from it as
> well. Compile all the tracing code, and see what does not build. Those
> files will be the ones that you need to add back the module.h to.
> 

Yeah, doing that I didn't see any build errors, but I'll stick with just
the explicit KSYM_SYMBOL_LEN-using files for this patchset to keep
things restricted to the specific case at hand.

Tom

> -- Steve
>