On Thu, 2007-08-23 at 13:40 -0700, Erik Nordmark wrote:
> Thanks for your review.
>
> Garrett D'Amore wrote:
>
> > ip.c:28759 This should probably be bugged up in bugster. Are there
> > different levels of policy for configuring different stack instances?
>
> It isn't a separate bug. Today it is only possible to enable CGTP from
> the global zone, hence we register PRIV_SYS_NET_CONFIG. But with these
> changes each IP Instance can have CGTP enabled independently. A result
> of that is that PRIV_SYS_IP_CONFIG is sufficient. (The latter privilege
> is given to each exclusive-IP zone, while the former is only in the
> global zone.)
Okay, this makes sense. Thanks.
>
> > ip.c: 28766 You removed the versioning checks on the ops. Why? (Ah,
> > now I see it, you have a redundant check in ip_cgtp_filter_register...)
>
> It was needed before because it was possible to "register" without
> calling the register function; could just set the global
> ip_cgtp_filter_ops variable. Since that is no longer possible we don't
> need the second version check.
Ah, ok. Thanks for the explanation.
>
> > ip_if.c:7281 I'd prefer to see to each side of a boolean && enclosed
> > with parens when it the test is compound (i.e. use " if ((x != NULL)
> > && ... " instead of " if (x != NULL &&"
> >
> > (Yes, I know its stylistic nic, but it avoids having to think about
> > operator predence. :-)
>
> I don't think Sun's cstyle recommends adding such parens.
> And C != LISP ;-)
I don't think the stle guide says anything about it. Its a personal
preference. Hence my statement that its a "nit". Maybe even that is
too strong a word for it...
Its just something I like to see to help delineate logical blocks and
prevent confusion about operator precedence. (Especially later if
someone adds some other logic to that statement...)
-- Garrett
>
> Erik
_______________________________________________
networking-discuss mailing list
[email protected]
