Re: [PATCH 1/1] kallsyms: add kallsyms_show_value definition in all cases

[Petr Mladek]

On Mon 2022-04-18 10:04:42, Maninder Singh wrote:
> > It is really weird that the function is declared in kallsyms.h
> > and implemented in vsprintf.c.
> >
> 
> Yes it does not look good.
> Initially we thought to make it as static inline function in kallsyms.h only.
> But as function is little big and it will increase code size, so we added
> definition in vsprintf.c, because its alwyas compilable code and also it has
> some wrapper APIs for kallsyms.
>  
> But as you suggested to make a new file, it will be good.
> 
> > What about splitting kallsyms.c into two source files where one
> > would be always compiled? It would be usable also for the
> > spring function introduced by
> > https://lore.kernel.org/r/20220323164742.2984281-1-maninder1.s@xxxxxxxxxxx
> > 
> > It might be something like kallsyms_full.c and/or kallsyms_tiny.c.
> >
> 
> This patch is not taken by Luis yet in module-tetsing branch.
> So what will be the best approach to make new version of this patch ?

I am not sure about Luis' plans and if the patch applies on top of
the current module.c split patchset (linux-next).


> A) to make new file kallsyms_tiny.c and add only one definition in that file
> and when above patch of spring functions is merged in next we can move definitions
> to new file
>
> or
> 
> B) we send patch to Luis's branch of module-testing with moving definition(of earlier patch)
> to new file and current patch also.

I wonder how many conflicts it might cause. The patchset splitting
module.c actually does not touch kernel/kallsyms.c.

Anyway, I would probably prepare it on top of linux-next.

Best Regards,
Petr