Last Call: 'Connecting IPv6 Islands over IPv4 MPLS using IPv6 Provider Edge Routers (6PE)' to Prop

Robert,

> Hey Francois,
> 
> Ok so you proved that the text describing AFI/SAFI is
poorly written and
> does not even reflect implementations reality (for
example as you said:
> rt-constrain).
> 
> I don't think that adding your change will make a huge
difference and
> will make the spec perfect.

While Francois' proposed change may not "make the
spec perfect",
it would certainly improve the match between the MP_BGP spec
and
some of the existing applications that use MP_BGP (e.g.,
rt-constrain,
l2vpn).  Whether making this change "will make a huge
difference"
is a matter of opinion...

Yakov.

_______________________________________________
Softwires mailing list
Softwiresietf.org
http
s://www1.ietf.org/mailman/listinfo/softwires

On 9/15/06, Dmitry Torokhov <dmitry.torokhovgmail.com> wrote:
> On 9/15/06, Jiri Kosina <jikosjikos.cz> wrote:
> > On Fri, 15 Sep 2006, Dmitry Torokhov wrote:
> >
> > > I understand what Ingo is saying about
detecting deadlocks across the
> > > pool of locks of the same class not waiting
till they really clash, it
> > > is really useful. I also want to make my code
as independent of lockdep
> > > as possible. Having a speciall marking on the
locks themselves (done
> > > upon creation) instead of altering call sites
is the cleanest way IMHO.
> > > Can we have a flag in the lock structure that
would tell lockdep that it
> > > is OK for the given lock to be taken several
times (i.e. the locks are
> > > really on the different objects)? This would
still allow to detect
> > > incorrect locking across different classes.
> >
> > Yes, but unfortunately marking the lock as
'can-be-taken-multiple-times'
> > is weaker than using the nested locking provided
by lockdep.
> >
> > i.e. if you mark a lock this way, it opens door
for having deadlock, which
> > won't be detected by lockdep. This will happen if
the code, by mistake,
> > really takes the _very same_ lock twice. lockdep
will not be able to
> > detect this, when the lock is marked in a way you
propose, but is able to
> > detect this when using the nested semantics.
> >
>
> But nested semantics breaks the notion of the locks
belonging to the
> same pool so both solutions have tradeoffs. I could use
either one of
> these as long as details are hidden and callers do not
have to care.
>

One more thing I forgot to add - how will we deal with
lockdep in
cases when we have X-over-Y-over-X protocol, when there is
no tight
coupling between the X parts so it is impossible to know
when to apply
special marking on the lock or callers?

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