Re: Link modules to DYN type

On Wed, Apr 26, 2023 at 01:38:32PM +0200, Hans Petter Selasky wrote:
> On 4/26/23 13:12, Konstantin Belousov wrote:
> > No, in-kernel linker does not behave this way.
> > Modules need to contain explicit reference to all modules they depend upon,
> > using the MODULE_DEPEND() macro.  Only symbols from the dependencies are
> > resolved.
> > 
> > All modules get an implicit reference to kernel.
> 
> Hi Konstantin,
> 
> Maybe I wasn't so clear. Trying again:
> 
> > diff --git a/sys/tests/ktest.c b/sys/tests/ktest.c
> > index 495fedf95dde..eb42cf062487 100644
> > --- a/sys/tests/ktest.c
> > +++ b/sys/tests/ktest.c
> > @@ -409,6 +409,12 @@ static moduledata_t ktestmod = {
> >          0
> >  };
> > +int
> > +printf(const char *fmt, ...)
> > +{
> > +       return (0);
> > +}
> > +
> >  DECLARE_MODULE(ktestmod, ktestmod, SI_SUB_PSEUDO, SI_ORDER_ANY);
> >  MODULE_VERSION(ktestmod, 1);
> >  MODULE_DEPEND(ktestmod, netlink, 1, 1, 1);
> 
> Then kldload ktest.ko . Which printf() function will be used if ktest.c
> calls printf() ?
My best guess is that printf is resolved locally by the static linker,
even before the module is loaded.  In other words, it would be ktest.c
printf.

> 
> I would expect a warning from the kernel at least ...
This is not how ELF behaves, even in such not fully compliant env as kernel.