Yes, do_schedule has to be exported. It used to be exported. Looks like the
export definition was lost somewhere.
I am going to check in this fix.
On Friday 12 Mar 2004 12:53 am, Michel Hubert wrote:
> This fix is required when you DON'T use thread analysis.
> I should specify that I based my fix on core.patch v 1.3.
> I guess when you refer to kernel/kgdb.c, you mean kernel/kgdbstub.c.
> In that file what I see is:
> +#ifdef CONFIG_KGDB_THREAD
> There is nothing about do_schedule which is called from
> include/linux/sched.h which is included in several modules:
> static inline void schedule(void)
> #ifdef CONFIG_KGDB_THREAD
> Sorry, I am new to patches and open source development. Please correct
> me if I am wrong.
> I took the latest stable kernel 2.6.3 from http://www.kernel.org
> Applied core.patch v1.3
> Applied i386.patch v1.2
> Applied 8250.patch v1.2
> *I would have applied all the latest patches if I wouldn't have
> encountered an error with kernel/Kconfig.kgdb with core.patch v1.6
> ** I suppose that in order to apply core.patch v1.6, I previously need
> to apply core.patch v1.3, then core.patch v1.4, then v.15 then v1.6.
> Two problems are arising:
> - 1) EXPORT_SYMBOL(kern_schedule); and a load of code appears twice in
> - 2) I do not have the file kernel/Kconfig.kgdb, so the patch fails.
> What am I missing ?
> On Thu, 2004-03-11 at 00:03, Amit S. Kale wrote:
> > On Thursday 11 Mar 2004 3:14 am, Michel Hubert wrote:
> > > - Fix unknown symbol do_schedule() when compiling modules without
> > > thread analysis with kgdb patch for kernel 2.6.3.
> > This shouldn't be required. When thread support is enabled, kernel/kgdb.c
> > exports do_shedule. Where did you get an unknown symbol error?
> > -Amit