Re: linux-next: kernel/trace/trace.c:658: error:

From: Geert Uytterhoeven
Date: Thu Oct 23 2008 - 11:30:20 EST


On Thu, 23 Oct 2008, Frédéric Weisbecker wrote:
> 2008/10/23 Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>:
> > If kernel/trace/trace.c calls irqs_disabled_flags(), it should include
> > the include that defines irqs_disabled_flags(). You should not add it to
> > some other random include.
> >
> > It's also happening on m68k:
> > http://kisskb.ellerman.id.au/kisskb/buildresult/50641/
>
> The fact is that other archs include this header into their
> asm/system.h. It seems that's because
> they need some non-traced irq_save/restore .
> I wanted to stay in the same approach because future use of
> raw_local_irq could be used elsewhere
> and seem to work perfectly whithout adding special headers on most
> arch. But there will be some
> bug report for each future use of these functions for alpha.... (and
> even m68k as it seems).
>
> What do you think? Should I let this patch as is or send a new one
> (and one other for m68k).

kernel/trace/trace.c needs to include at least <linux/irqflags.h>, as
that's where irqs_disabled_flags() is defined.

If there are still other issues with the raw_local_irq(), IMHO they should be
fixed separately.

Hmm, what's this doing in <linux/irqflags.h>:
| #ifdef CONFIG_TRACE_IRQFLAGS_SUPPORT
| ...
| #endif /* CONFIG_X86 */
^^^^^^^^^^

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds