On Wed, Jan 09, 2013 at 10:13:51AM +0100, Franck Jullien wrote:
> 2013/1/9 Sascha Hauer <[email protected]>:
> > On Wed, Jan 09, 2013 at 10:02:32AM +0100, Franck Jullien wrote:
> >> Hi,
> >>
> >> I have a question not directly related to Barebox but I think I can
> >> find some answer here: )
> >>
> >> I would like to use initcalls in a Linux user's land program on a x86 
> >> target.
> >>
> >> I'm doing something like this:
> >>
> >> #ifndef _INIT_H
> >> #define _INIT_H
> >>
> >> typedef int (*initcall_t)(void);
> >>
> >> extern initcall_t __start_target, __stop_target;
> >>
> >> #define target_initcall(fn)   static initcall_t _##fn \
> >>                               __attribute__((used)) \
> >>                               __attribute__ ((section("target"))) = fn
> >>
> >> #endif
> >>
> >> then:
> >>
> >>       initcall_t *initcall;
> >>
> >>       for (initcall = &__start_target;
> >>                       initcall < &__stop_target; initcall++) {
> >>               printf("initcall-> %p\n", *initcall);
> >>               ret = (*initcall)();
> >>               if (ret)
> >>                       printf("initcall %p failed: %d\n", *initcall, ret);
> >>       }
> >>
> >> Everything looks fine except the linker removes the function
> >> "initcalled" because it is not
> >> referenced anywhere and this is normal.
> >>
> >> I have not modified the linker script (I'm using the default one). I'm
> >> using auto generated
> >> __start_target and __stop_target symbols generated by the linker.
> >>
> >> My question is: why does it work in barebox ? For example, in
> >> nios2/generic.c we have only
> >> static function and initcalls. So why the linker does optimize out
> >> those functions ? Is it
> >> because we have initcall corresponding sections in the linker script ?
> >
> > Yes. In nios2 this is:
> >
> >         __barebox_initcalls_start = .;
> >         .barebox_initcalls : { INITCALLS }
> >         __barebox_initcalls_end = .;
> >
> > With INITCALLS being defined as:
> >
> > #define INITCALLS                       \
> >         KEEP(*(.initcall.0))                    \
> >         KEEP(*(.initcall.1))                    \
> >         KEEP(*(.initcall.2))                    \
> >         KEEP(*(.initcall.3))                    \
> >         KEEP(*(.initcall.4))                    \
> >         KEEP(*(.initcall.5))                    \
> >         KEEP(*(.initcall.6))                    \
> >         KEEP(*(.initcall.7))                    \
> >         KEEP(*(.initcall.8))                    \
> >         KEEP(*(.initcall.9))                    \
> >         KEEP(*(.initcall.10))                   \
> >         KEEP(*(.initcall.11))
> >
> > The 'KEEP' keyword keeps the linker from throwing away these.
> >
> > Sascha
> >
> > --
> > Pengutronix e.K.                           |                             |
> > Industrial Linux Solutions                 | http://www.pengutronix.de/  |
> > Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
> > Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
> 
> Thanks Sascha, I missed this one.....
> 
> So is there no way I can do what I want without modification to the
> default linker script.
> I searched for a gcc attibute equivalent to KEEP but it doesn't seems to 
> exist.

No, and this wouldn't be enough anyway. You also have to tell the linker
that it shall put all functions matching .initcall.* in a separate
section. Otherwise they would end up in the regular sections.

Sascha


-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

Reply via email to