Re: [PATCH] Celleb: improve htab lock

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Thu, 2007-02-08 at 07:51 +0100, Arnd Bergmann wrote:
> On Thursday 08 February 2007 07:18, Ishizaki Kou wrote:
> > s);
> > +       local_irq_save(flags);
> > +       spin_lock(&beat_htab_lock);
> >         dummy1 = beat_lpar_hpte_getword0(slot);
> >  
> >         if ((dummy1 & ~0x7FUL) != (want_v & ~0x7FUL)) {
> >                 DBG_LOW("not found !\n");
> > -               spin_unlock_irqrestore(&beat_htab_lock, flags);
> > +               spin_unlock(&beat_htab_lock);
> > +               local_irq_restore(flags);
> >                 return;
> >         }
> >  
> >         lpar_rc = beat_write_htab_entry(0, slot, 0, 0, HPTE_V_VALID, 0,
> >                 &dummy1, &dummy2);
> > -       spin_unlock_irqrestore(&beat_htab_lock, flags);
> > +       spin_unlock(&beat_htab_lock);
> > +       local_irq_restore(flags);
> 
> The function normally used here would be spin_lock_irq()/spin_unlock_irq(),
> which does spin_{,un}lock along with local_irq_{save,restore}.

Not lock_irq/unlock_irq, that doesn't save/restore but disables/enables
(that is doesn't save the previous state).

I'd rather have him use _irqsave/irqrestore, that is, keep the current
code in that function.

Ben.