GitLab

wires table lookup).

object lookups via Interface library.

1. We now allow lans to be implemented by a path. We did not allow this
   before, cause some of the sanity checking code was a pain to
   implement for lans. Well, no more sanity checking, the user is
   responsible for doing things correctly (after all, they are doing
   experiments with their own switches!).

2. We now allow topologies with more then one switch to be wired
   together. The wires between switches are marked as "trunk" wires,
   which informs the configuration generation code in libosload_switch
   to create the trunks and do the little tagged/untagged magic that is
   required on procurve switches. The same information is used to mark
   the the logical wires between switches as trunks.

   Aside: this stuff needs some work; we have spanning tree on by
   default, which causes the trunks to not work correctly. When I turn
   that off, things start working. So need some help from others who now
   about spanning tree stuff.

3. Serious kludging in the Interface and Port libraries related to
   choice of primary keys in the wires table. In order to insert a
   logical wire (or interface) that represents a connection setup by the
   apcon, we have to overload the primary key since the node_id1 side of
   the logical wires is the same as the physical wire to the apcon. We
   have to have overload the node_id2 side too, but that is really just
   a problem when wiring two switches together. Anyway, the kludge just
   maps card1 to a different id, and the Port library unmaps it. It will
   do for now, but really need logical wires to be done better then
   this.

The stuff in db/Port.pm (Port::LookupByIface) wasn't handling the MLE
case where there are two wires table entries for physical wires with a
logical wire implemented by them; for my experiment, when snmpit was
trying to create the physical vlans on the apcon, it would grab the
logical wire instead of the physical one as it populated the Port
objects, and that screwed vlan creation up (the member port for the
physical layer 1 vlan gets labeled with a switch WIRE_END instead of
node -- which is correct for the logical wire, but not the physical).
So, I changed the query to check that wires.logical ==
interfaces.logical ; see the code.  This is a limited fix, maybe --
maybe that method should really do more verification to make sure it's
grabbed the correct wires table entry -- but since we only have 2 layers
it's plenty good for now.

It's useful to know what state an interface is in currently versus what
is mandated in the 'interfaces' table.  E.g., when deciding what to do
in setPortVlan() based on whether or not a port is in 'trunk' mode.

This commit pushes the changes located in the temporary snmpit_portfix
subdirectory into the mainline snmpit_test directory, and updates
db/Ports.pm with the 'portfix' changes.  These changes ensure better
consistency with port lookup/resolution.  In particular, interswitch
trunk lookup will return the switch-local side of the link.

This commit also adds an installer script to cleanup the temporary files
introduced for burning in the snmpit_portfix changes.

In order to test the portfix changes to snmpit, a new version of
the tool and its modules has been split off here.  This new version
will be selected via a site variable / emulab feature.  This also
reverts changes to the mainline Ports module and snmpit modules
in the snmpit_test subdirectory.  The modified Port module will
go by the name of Ports_portfix.pm, and will ultimately disappear
once these changes have been vetted at the Utah Emulab site.

The lookup functions now look at the wire type and which location the
requested node is in (node_id1 or node_id2) to decide which side of the
link the port represents (switch vs. endpoint).  We don't (yet) query the
nodes table for the role since we consistently use the node_id1 (and
related) columns to hold the endpoint (node) information for wires of
type "Node".

For inter-switch
trunks, we always mark the port object as being the "switch" side.  Both
sides are the switch side...  Functions like getPCPort() and getSwitchPort()
are ambiguous when invoked on a switch port object, and will always return
a reference to the object the method was invoked on.

Also update the HP snmpit module to explicitly check the wire type for a port
before deciding whether or not to get the port at the other end in the
listVlans() function.

Run as follows:

boss> wap perl snmpit_test --prunestalevlans --impotent

which will tell you about them. Remove the --impotent option to
actually remove them.

Only numbered lans are considered; ones that derive from entries in
the lans table. Named vlans are skipped since those are generally
created by hand (often via the switch CLI).

Caveat; vlans left on trunk links are still a bit of problem since
listvlans returns the other side of the trunk. Needs to be fixed.

This commit is intended to makes the license status of Emulab and
ProtoGENI source files more clear.  It replaces license symbols like
"EMULAB-COPYRIGHT" and "GENIPUBLIC-COPYRIGHT" with {{{ }}}-delimited
blocks that contain actual license statements.

This change was driven by the fact that today, most people acquire and
track Emulab and ProtoGENI sources via git.

Before the Emulab source code was kept in git, the Flux Research Group
at the University of Utah would roll distributions by making tar
files.  As part of that process, the Flux Group would replace the
license symbols in the source files with actual license statements.

When the Flux Group moved to git, people outside of the group started
to see the source files with the "unexpanded" symbols.  This meant
that people acquired source files without actual license statements in
them.  All the relevant files had Utah *copyright* statements in them,
but without the expanded *license* state...

Add "Unused" to type enum in the wires table.
Change a few critical wires/interfaces queries to avoid logical=1.

Port refactoring for snmpit apcon/hp/cisco done, all run very well, trunking still not tested but be expected to be OK.

commit refactored code before merging old ones, the new_snmpit frontend, lib, lan, stack, hp, apcon already merged