> > CR 6730167[1] questions whether dhcp_drop() should be called when one > > or more of attributes of a DHCP-managed interfaces change. What is the > > general consensus on this? Should dhcp_drop() be called without > > canonize_lif() being called? Or, should a different (new) function be > > called that simply removes the IFF_DHCPRUNNING flag from the interface? > > I think you still have to do dhcp_drop() so that the lease is properly > cached, but that the path that goes through deprecate_leases() to > remove_lease() to unplumb_lif() needs to be tweaked so that we > recognize the "abandoning" case, and deal with it by turning off the > running flag and *not* by zeroing out the interface.
Agreed. -- meem _______________________________________________ networking-discuss mailing list [email protected]
