fix(dhcp): let Kea start, let ISC refuse, and ask for what is asserted

Three things the wire cases turned up, only the first of which is xCAT
answering a client wrongly -- the other two never got as far as an answer.

kea-dhcp4 would not start at all. The ONIE class named option "www-server",
which xCAT's dhcpd.conf declares as code 114 = string but which means the
standard option 72 to Kea -- a list of IPv4 addresses, so an installer URL
in one is a configuration error and the whole Kea pass of the wire cases
never ran. Naming the code instead says the same thing to both.

ISC answered a request for an address on a network it has never heard of
with silence, where the spec says DHCPNAK, because "authoritative" was
written into each subnet declaration and read from the subnet the
requested address belongs to -- which, for this case, is none of them. It
is now global as well, which is what Kea's authoritative:true already
covered.

The remaining two were the cases asking the wrong question. ISC sends an
option when the client asks for it, so an ONIE case that never put 114 in
its parameter request list could not see it however the class was
written; and the common-option case announced a PXEClient vendor class
while asserting the cluster's lease time, which is the ten-minute
firmware lease of S-51 rather than S-50. S-50 now says which clients it
is about.

perl-xCAT/xCAT/DHCP/BootPolicy.pm

@@ -254,16 +254,30 @@ sub kea_onie_network_classes {
             test            => onie_vendor_class_test(),
             additional_only => 1,
             'option-data'   => [
-                {
-                    name          => 'www-server',
-                    data          => "http://$opts{next_server}$portsuffix/install/onie/onie-installer",
-                    'always-send' => 1,
-                },
+                kea_onie_url_option( undef, "http://$opts{next_server}$portsuffix/install/onie/onie-installer" ),
             ],
         },
     ];
 }
 
+# The installer URL goes in option 114, which is where ONIE looks for it.
+#
+# xCAT's dhcpd.conf says so by declaring "option www-server code 114 = string"
+# -- a local redefinition, because ISC's own www-server is the standard option
+# 72. Kea has no such redefinition: to it, www-server means option 72, a list
+# of IPv4 addresses, and a URL in one is a configuration error that stops the
+# server from starting at all. Naming the code says the same thing to both,
+# and puts the same bytes on the wire.
+sub kea_onie_url_option {
+    my ( $class, $url ) = @_;
+
+    return {
+        code          => 114,
+        data          => $url,
+        'always-send' => 1,
+    };
+}
+
 # The user class a chainloaded second stage announces itself with, as an ISC
 # dhcpd condition.
 #

xCAT-server/lib/xcat/plugins/dhcp.pm

@@ -4155,7 +4155,7 @@ sub kea_boot_for_node
         }
     } elsif ($netboot and $netboot eq 'onie') {
         my $onie_url = kea_onie_url_for_node($node, $ntent, $nxtsrv, $httpport);
-        push @{ $boot{'option-data'} }, { name => 'www-server', data => $onie_url } if $onie_url;
+        push @{ $boot{'option-data'} }, xCAT::DHCP::BootPolicy->kea_onie_url_option($onie_url) if $onie_url;
     }
 
     push @{ $boot{'option-data'} }, { name => 'host-name', data => $node };
@@ -5117,6 +5117,17 @@ sub newconfig
 
     push @dhcpconf, "option iscsi-initiator-iqn code 203 = string;\n"; #Only via gPXE, not a standard
     push @dhcpconf, "ddns-update-style interim;\n";
+
+    # A node that was moved to another rack comes up asking for the address it
+    # held there. Answering nothing leaves it retrying an address it can never
+    # have until its own timers give up; a DHCPNAK tells it to start over now.
+    #
+    # Each subnet declaration says "authoritative" as well, but that flag is
+    # read from the subnet the requested address belongs to -- and the whole
+    # point of this case is that it belongs to none of them. Only the global
+    # statement covers an address this server has never heard of, which is
+    # what Kea's global authoritative:true covers.
+    push @dhcpconf, "authoritative;\n";
     push @dhcpconf, "ignore client-updates;\n"; #Windows clients like to do all caps, very un xCAT-like
 
     #    push @dhcpconf, "update-static-leases on;\n"; #makedns rendered optional

xCAT-test/dhcptest/conf/provision-vs-discovery.conf

@@ -83,9 +83,17 @@ description = The node is told how to route, resolve and keep time, not just its
 #
 # Asked for as an INIT-REBOOT request -- option 50, no option 54 -- which is
 # what a node that already holds this lease sends when it comes back up.
+#
+# And asked without the PXEClient vendor class this file otherwise sends,
+# because this is the reply an installed system gets: the options here are the
+# ones an operating system needs, and a client that announces itself as PXE
+# firmware is deliberately given a ten-minute lease instead of the cluster's
+# (S-51), which is a different question asked in loader-absent.conf.
 [step known-network-settings]
 type              = request
 mac               = %(node_mac)s
+vendor_class      =
+client_arch       =
 requested_address = %(node_ip)s
 expect            = ack
 assert =

xCAT-test/dhcptest/conf/pxe-arch-matrix.conf

@@ -208,9 +208,16 @@ description = S-19: An ONIE switch is offered an installer URL before it is a no
 # A switch announces onie_vendor on its very first boot, which is necessarily
 # before anyone has defined it as a node. If only a node definition can produce
 # the installer URL, the switch can never reach the state where it gets one.
+#
+# Option 114 is asked for, because a server sends an option when the client
+# asks for it: ISC dhcpd puts nothing in the reply for an option missing from
+# the parameter request list, however carefully the class set it. A real ONIE
+# switch asks for 114, which is the whole reason xCAT's dhcpd.conf redefines
+# www-server to that code.
 [step onie-discover]
-type         = discover
-vendor_class = onie_vendor:x86_64
+type            = discover
+vendor_class    = onie_vendor:x86_64
+request_options = 1, 3, 6, 54, 114
 assert =
     msgtype    == OFFER
     option:114 == %(onie_url)s

xCAT-test/dhcptest/spec.md

@@ -515,10 +515,12 @@ Scenario: The reply carries an MTU where the network defines one
 @S-50
 Scenario: The lease is as long as site.dhcplease says
   Given site.dhcplease is set
-  When a client is acknowledged
+  When a client that is not PXE firmware is acknowledged
   Then option 51 (lease-time) is that value
   And 43200 seconds when site.dhcplease is unset
-  # dhcp.pm:4524
+  # dhcp.pm:4524. "Not PXE firmware" because a client announcing a PXEClient
+  # vendor class is answered by S-51 instead, and the two are the same reply
+  # field.
 
 @S-51
 Scenario: A PXE client gets a short lease

xCAT-test/unit/dhcp_boot_policy.t

@@ -290,7 +290,7 @@ is_deeply(
             additional_only => 1,
             'option-data'   => [
                 {
-                    name          => 'www-server',
+                    code          => 114,
                     data          => 'http://10.0.0.1/install/onie/onie-installer',
                     'always-send' => 1,
                 },
@@ -299,6 +299,15 @@ is_deeply(
     ],
     'an ONIE switch is pointed at the installer over HTTP, as the ISC path does',
 );
+
+# The option is named by code and not by name on purpose. "www-server" means
+# option 114 in xCAT's dhcpd.conf, which declares it that way, and option 72 to
+# Kea, which does not -- and option 72 holds IPv4 addresses, so a URL in it
+# stops kea-dhcp4 from starting.
+is( $onie->[0]{'option-data'}[0]{code}, 114,
+    'the installer URL is option 114, the one ONIE reads' );
+ok( !exists $onie->[0]{'option-data'}[0]{name},
+    'and it is not named www-server, which is a different option to Kea' );
 is(
     xCAT::DHCP::BootPolicy->kea_onie_network_classes(
         net => '10.0.0.0', prefix => 24, next_server => '10.0.0.1', httpport => 8080,