On Fri, Oct 09, 2009 at 04:27:51PM -0400, Anurag S. Maskey wrote: > Alan Maguire wrote: >> Anurag S. Maskey wrote: >>> code review requested for >>> >>> 11834 IPv6 only NCUs don't go online >>> http://defect.opensolaris.org/bz/show_bug.cgi?id=11834 >>> >>> webrev at >>> >>> >>> http://zhadum.east/export/ws/am223141/checkout-area/nwam1-fixes/webrev/ >> minor nit: it might make sense to rename >> to nwamd_if_belongs_to_this_zone() >> since it takes as argument strings that are >> valid interface names but not valid link names. > accepted and webrev updated.
Wait, I think Alan's observation uncovers a bigger problem (in the existing code, not in Anurag's fix). The reason the function is called nwamd_link_belongs_to_this_zone() is because it's a wrapper around zone_check_datalink(). We should not be giving it interface names at all. I would assert that we don't need to be doing this check at all in this case. If we're getting a routing socket message about something that happened on an interface, it belongs to our zone. Either we're the global zone and own the underlying link, or we're the exclusive stack zone that owns it. We only need to do this check when we're deciding whether or not to pay attention to a given *link*. So rather than making Anurag's suggested fix, I think we should just rip out all the calls to nwamd_link_belongs_to_this_zone() from within the routing event handling code. -renee
