On Thu, Feb 21, 2019 at 01:12:40PM +0100, Cornelia Huck wrote:
> On Thu, 21 Feb 2019 11:42:25 +0100
> Harald Freudenberger <fre...@linux.ibm.com> wrote:
>
> > On 30.01.19 19:32, Sebastian Ott wrote:
> > > On Wed, 30 Jan 2019, Tony Krowiak wrote:
>
> > >> /*
> > >> +* A config change has happened, Force an ap bus rescan.
> > >> +*/
> > >> +void ap_bus_cfg_chg(void)
> > >> +{
> > >> + AP_DBF(DBF_INFO, "%s config change, forcing bus rescan\n",
> > >> __func__);
> > >> +
> > >> + ap_bus_force_rescan();
> > >> +}
> > >> +EXPORT_SYMBOL(ap_bus_cfg_chg);
> > > There is no need for the export symbol - you don't call that function
> > > from module code.
> > That's what I have learned now: You don't need to export a symbol
> > as long as the symbol is only called in static code parts of the kernel.
> > But you need to export it when it is intended to be used by code
> > which sits in a kernel module. So now the big question:
> > How does a provider of a function in the kernel know, if the caller is in
> > static
> > code or in module code ? And ... maybe this may even change over
> > the time. So my recommendation is to always export the symbol with
> > the EXPORT_SYMBOL macro. This way you don't need to change the
> > code providing a function when the caller code changes or additional
> > code uses the symbol.
> >
> > Other opinions ?
>
> Well, if you know it will be called from module code in upcoming
> patches, export it. If not, I consider it the choice of the maintainer.
> You can easily add the export later on, if needed, anyway, and I don't
> consider changing the code a problem.
>
> In this particular case, both exporting and not exporting looked like
> reasonable choices to me.
The number of exported symbols should be as small as possible. Each
exported symbol eats up extra memory for meta information, plus I
really do not want to deal with semi-automated bots sending patches to
remove not needed exported symbols.
You can check usage of exported symbols with "make export_report".
