On Wed, May 12, 2010 at 07:23:13PM -0700, [email protected] wrote:
> > - this bit seems odd to me:
> >
> > ---8<---
> > ... Note that IPv4
> > and IPv6 are considered as independent resources, so that
> > specification of an IPv4 address via zonecfg(1m) does not place any
> > constraints on permissible IPv6 addresses, and vice-versa.
> > ---8<---
> >
> >   ipv4 and ipv6 addresses are considered independent resources from a
> >   zonecfg(1m) perspective, but from a systems management perspective
> >   they don't seem that different to me.  to me, it seems that once an
> >   admin decides to restrict what addresses can be used on a given
> >   interface, that restriction should apply to both ipv4 and ipv6
>
> I could go either way on this.. one interpretation of specifying
> only ipv4 addresses in zonecfg would be "ipv6-addrs are dont-cares".
> Another is that "ipv6-addrs are forbidden".  I've somewhat arbitrarily
> selected the former interpretation, but could just as easily go with
> the latter.
>
> >   addresses.  for example, with the current mechanism, if an admin want
> >   to prevent the zone from configuring any ipv6 address, how do they go
> >   about doing this?
>
> Currently, none, though the  "only ipv4 specified implies ipv6-addrs
> are forbidden" approach solves that.  In retrospect,  that choices
> seems simpler and cleaner. Is that preferable?
>

i think so.

> > - does the existence of a zonecfg(1m) defrouter property on an interface
> >   prevent a zone from changing the default router for that given
> >   interface?
>
> No.
>
> > - does the existence of a zonecfg(1m) defrouter property on an interface
> >   prevent a zone from adding additional routes on a given interface?
>
> No.
>
> Note that, even though a shared-IP zone cannot add routes today, there's
> really nothing that constrains it to using the routes specified in the
> routing table- it could easily circumvent the routing table by using
> IP source-routes, or by using IP*NEXTHOP or IP*PKTINFO socket options.
> So preventing the addition of routes doesn't actually achieve any
> limitations.
>

ok.  so this really isn't much of an issue.

> > - can exclusive stack zones manipulate arp tables?
> yes.
>
>
> > if so, does use of
> >   the zonecfg(1m) defrouter property restrict arp table manipulation in
> >   any way?  (i'm wondering if arp table manipulation could be used to
> >   step around any potential routing limitations placed on an interface.)
>
> there are many ways to step around routing limitations even with shared-IP
> today, so not sure what limitations you have in mind.
> Basically, regardless of what the routing table has, a node can always
> send a packet to any onlink router.
>
> If what we want, is to prevent the NGZ from DOSing some router(s), then
> bandwidth controls are a more reliable way to achieve that.
>

this last question only really made since in a context where we might
try to restrict the routing configuration in a zone.  given that shared
stack zones have always had the ability to circumvent any predefined
routing configuration i see no reason to need to limit arp table
manipulation.

> > - can exclusive stack zones manipulate mac addresses on network
> >   interfaces?
>
>  yes- they can use 'ifconfig .. ether <..>'.
>
> >  if so, does the use of the zonecfg(1m) address property
> >  restrict this in any way?
>
> no, it does not - the address property only clamps dow the IP address,
> and makes no promises about the mac address associated with the IP address.
>

given that one of the motivation for this work is to prevent zones from
using addresses they shouldn't (and there by being capable of DOS-ing
hosts using those addresses) it seems like we should have a zonecfg
mechanism that prevents mac address manipulation.  i don't know if that
should be bundled in with this proposed IP limiting mechanism (ie. if a
user specifies an IP address the mac would automatically be locked down)
or if there should be a seperate knob to control this.  thoughts?

ed
_______________________________________________
opensolaris-arc mailing list
[email protected]

Reply via email to