Peter Memishian wrote: > I've completed my initial testing and self-review, so it's time to start > code review for the DHCP client sockets changes. The webrev is at: > > http://cr.opensolaris.org/~meem/dhcp-sock
A couple of random thoughts, and one comment: usr/src/cmd/cmd-inet/sbin/dhcpagent/interface.c * 1322: Not a comment for you, but for SIOCSLIFFLAGS in general. There needs to be a better interface for "toggling" a single flag in an atomic way. Otherwise, any other flag change that occurs by another thread between SIOCGLIFFLAGS (as on line 1316) and SIOCSLIFFLAGS (as on line 1322) get overridden. usr/src/cmd/cmd-inet/sbin/dhcpagent/states.c * 704,705: Is closing and re-opening the socket necessary? Could we not re-bind to the appropriate address using the same socket? I'm more curious than anything, as there's nothing wrong with the code you have. usr/src/uts/common/inet/ip/ip.c * 15116: Why check for IS_SIMPLE_IPH() here? Does the DHCP protocol mandate that IP options must not be used by the DHCP server? If not, then the offset to the UDP header can easily be obtained even when the IP header contains options. -Seb
