Re: 2.6.9-rc3-mm2

[Andrew Morton]

Rusty Russell <rusty@xxxxxxxxxxxxxxx> wrote:
>
> On Wed, 2004-10-06 at 00:07, Dominik Karall wrote:
> > On Monday 04 October 2004 11:02, Andrew Morton wrote:
> > > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.9-rc3/2.6
> > >.9-rc3-mm2/
> > 
> > some more scheduling/preempt problems. following patches were applied:
> > --- 
> > 25/include/linux/netfilter_ipv4/ip_conntrack.h~conntrack-preempt-safety-fix 
> > Mon Oct  4 14:36:19 2004
> > +++ 25-akpm/include/linux/netfilter_ipv4/ip_conntrack.h Mon Oct  4 14:37:02 
> > 2004
> > @@ -311,10 +311,11 @@ struct ip_conntrack_stat
> >         unsigned int expect_delete;
> >  };
> >  
> > -#define CONNTRACK_STAT_INC(count)                              \
> > -       do {                                                    \
> > -               per_cpu(ip_conntrack_stat, get_cpu()).count++;  \
> > -               put_cpu();                                      \
> > +#define CONNTRACK_STAT_INC(count)                                      \
> > +       do {                                                            \
> > +               preempt_disable();                                      \
> > +               per_cpu(ip_conntrack_stat, smp_processor_id()).count++; \
> > +               preempt_disable();                                      \
> >         } while (0)
> 
> Please, please please!  Never use per_cpu(XXX, smp_processor_id())!

Why?

>  In
> this case, it's "get_cpu_var(ip_conntrack_stat).count++; put_cpu(),

Actually, it's put_cpu_var().

> although I think this code should only be called in a softirq, so a
> simple "__get_cpu_var(ip_conntrack_stat).count++;" is sufficient.

We were getting warnings from somewhere or other due to smp_processor_id()
within preemptible code - I don't recall the callsite.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/