Ethan Jackson [Wed, 31 Aug 2011 00:37:15 +0000 (17:37 -0700)]
cfm: Allow accurate transmission intervals in extended mode.
The standard CFM protocol only allows a handful of transmission
rates. This is particularly problematic if you want to support a
transmission rate slower than 100 ms and faster than 1000 ms.
This patch allows arbitrary transmission rates (between 1 ms and
65535 ms). It does this by commandeering parts of a reserved
"zero" field in the ccm message. This breaks wire compatibility
with standard 802.1ag implementations, and thus is only supported
in extended mode.
Ethan Jackson [Fri, 19 Aug 2011 20:58:56 +0000 (13:58 -0700)]
cfm: Remove cfm_remote_mpid configuration.
According to the 802.1ag specification, users should be able to
configure the CFM module with a list of remote endpoints with which
the local endpoint should have connectivity. Commit 93b8df3853
"cfm: Remove Maintenance_Point and Monitor tables." changed the
behavior so that only one remote endpoint could be specified. This
commit takes it further, by disallowing specification of any
remote endpoints.
Due to this change, the semantics of the fault flag are slightly
different. Before, a fault was triggered if any of the configured
remote endpoints were unreachable (or with RDI), or if any
unconfigured remote endpoints were reachable. Now a fault is
triggered if no remote endpoints are reachable at all, or if
reachable endpoints have set their RDI.
Ethan Jackson [Thu, 8 Sep 2011 23:15:36 +0000 (16:15 -0700)]
bridge: Write CFM changes more aggressively.
This patch no longer rate limits database updates due to CFM
changes. Due to recent changes, the fault status of CFM only
changes once per 3.5 tx_interval seconds. There doesn't seem to be
a good reason to add an additional rate limit on top of this.
Ben Pfaff [Thu, 8 Sep 2011 23:22:38 +0000 (16:22 -0700)]
ofproto: Fix documentation for calls to ->rule_destruct().
This documented what I intended to implement and what I thought I had
implemented, but not what the code actually did. It is a little easier to
fix the documentation to match the implementation than vice versa, so this
commit does so.
Ethan Jackson [Wed, 31 Aug 2011 01:19:48 +0000 (18:19 -0700)]
ofproto-dpif: Drop special packets more judiciously.
Before this patch, ofproto-dpif would not forward any LACP or CFM
packets regardless of whether or not the relevant modules were
configured on the in_port.
Ethan Jackson [Wed, 31 Aug 2011 01:17:27 +0000 (18:17 -0700)]
lacp: Clean up LACP module interface.
There's no particular reason to force users of the LACP module to
be aware of the lacp_pdu structure. This patch hides that
information in the LACP module implementation. This results in
slightly cleaner code which is more consistent with the CFM
module.
Ben Pfaff [Thu, 8 Sep 2011 23:30:20 +0000 (16:30 -0700)]
datapath: Allow a packet with no input port to omit OVS_KEY_ATTR_IN_PORT.
When ovs-vswitchd executes actions on a synthesized packet, that is, on a
packet that is not being forwarded from any particular port but is being
generated by ovs-vswitchd itself or by an OpenFlow controller (using a
OFPT_PACKET_OUT message with an in_port of OFPP_NONE), there is no good
choice for the in_port to pass to the kernel in the flow in the
OVS_PACKET_CMD_EXECUTE message. This commit allows ovs-vswitchd to omit
the in_port entirely in this case.
This fixes a bug in OFPT_PACKET_OUT: using an in_port of OFPP_NONE would
cause the packet to be dropped by the kernel, since that's an invalid
input port.
datapath: Calculate flow hash after extracting metadata.
When we execute a packet from userspace we first extract the header
fields from the packet and then add supplied metadata. However, we
compute the hash of the packet in between these two steps despite
the fact that the metadata can affect the hash. This can lead to
two separate hashes for packets of the same flow.
Found by code inspection, not an actual real-world problem.
Signed-off-by: Jesse Gross <jesse@nicira.com> Acked-by: Ben Pfaff <blp@nicira.com>
Simon Horman [Wed, 7 Sep 2011 01:10:30 +0000 (10:10 +0900)]
ovs-ctl: Add load-kmod command
On Debian there is a need for the init scripts to die gracefully
if module insertion fails. In such a case it is desirable to print
some sort of informative message.
By adding the load-kmod sub-command to ovs-ctl init scripts
may try to load modules and take appropriate action on failure
or then try to start the daemons.
Ben Pfaff [Tue, 6 Sep 2011 16:33:26 +0000 (09:33 -0700)]
netlink-protocol: Move CTRL_ATTR_MCAST definitions for consistency.
One of the current goals of netlink-protocol.h, for better or for worse, is
to ensure that the same definitions are available whether a Linux kernel is
in use or not. One of the ways it accomplishes this is by putting the
conditional definitions that test for features missing in old kernels at
the very end, after the dummy definitions used on non-Linux platforms.
However, commit b0025c8389f "netlink-protocol: Define missing symbols"
added new conditional definitions only in the Linux platform case, which
means that those definitions won't be available on non-Linux platforms.
This commit moves them to the end, instead.
The symbols that are moved are only used from netlink-socket.c, which is
only built on Linux platforms, so this does not change an actual bug. It
only makes the location of the definitions consistent with prior practice.
Ethan Jackson [Wed, 24 Aug 2011 00:20:00 +0000 (17:20 -0700)]
datapath: genl_notify() on port disappearances.
Before this patch, if a vport detached itself from the datapath
without interaction from userspace, rtnetlink notifications would
be sent, but genl notifications would not.
Feature #6809.
Signed-off-by: Ethan Jackson <ethan@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
Ethan Jackson [Wed, 24 Aug 2011 22:17:32 +0000 (15:17 -0700)]
netlink-notifier: Rename rtnetlink code.
This patch renames the rtnetlink module's code to "nln" for
"netlink notifier". Callers are now required to pass in the
netlink protocol to he newly renamed nln_create() function.
Ethan Jackson [Thu, 25 Aug 2011 21:06:54 +0000 (14:06 -0700)]
lib: Rename rtnetlink.[ch] files.
The only rtnetlink specific functionality contained in the
rtnetlink module is the use of the NETLINK_ROUTE protocol. This
can easily be passed in by callers.
In preparation for generalization, this patch renames
rtnetlink.[ch] to netlink-notifier.[ch]. Future patches will
complete the transition.
Justin Pettit [Sat, 27 Aug 2011 06:34:40 +0000 (23:34 -0700)]
datapath: Disable LRO from userspace instead of the kernel.
Whenever a port is added to the datapath, LRO is automatically disabled.
In the future, we may want to enable LRO in some circumstances, so have
userspace disable LRO through the ethtool ioctls.
As part of this change, the MTU and LRO checks are moved to
netdev-vport's send(), which is where they're actually needed.
Justin Pettit [Fri, 26 Aug 2011 22:05:53 +0000 (15:05 -0700)]
bridge: Make bridge_pick_local_hw_addr() easier to reason.
The use of eth_addr_is_multicast() to see if a reasonable address was
found always caused me momentary confusion. This commit uses a flag
instead, and also saves a bit of unnecessary array reading and writing.
Simon Horman [Fri, 26 Aug 2011 09:03:42 +0000 (18:03 +0900)]
debian: Transition to dh_python2
dh_pysupport that the packaging used until now is deprecated, with
dh_python2 as its successor.
This commit removes the PYTHONPATH setting from
debian/openvswitch-ipsec.init because it is not needed, as the Python
packaging is public. In fact, the Python packaging was public,
unintentionally, before, so the PYTHONPATH could have been removed earlier.
Chuck Short [Tue, 23 Aug 2011 22:37:11 +0000 (15:37 -0700)]
debian: Apply Ubuntu patch to add DKMS support.
I tested that installing openvswitch-datapath-dkms worked OK on my own
Debian machine.
The bulk of this patch is taken from downstream Ubuntu DKMS support written
by Chuck Short <zulcss@ubuntu.com>, version 1.2.0-1ubuntu1. I made the
following changes:
* Update debian/.gitignore.
* Update debian/automake.mk.
* Correct description in debian/control (it was a cut-and-paste from
the openvswitch-datapath-source description without editing).
* Fix up for --with-l26 to --with-linux and datapath/linux-2.6 to
datapath/linux transitions.
CC: Chuck Short <zulcss@ubuntu.com> CC: Dave Walker <DaveWalker@ubuntu.com> Acked-by: Simon Horman <horms@verge.net.au>
Ben Pfaff [Tue, 23 Aug 2011 16:36:39 +0000 (09:36 -0700)]
ovs.db.idl: Use top-level class to represent IDL rows.
According to Reid, there may be some disadvantages to having this class be
anonymous, for example, cannot do instance/typechecking, might be
allocating a new class for every row as well, which isn't the most memory
efficient.
Ben Pfaff [Tue, 23 Aug 2011 16:26:29 +0000 (09:26 -0700)]
ovs.db.idl: Actually use Idl.__modify_row()'s return value.
Idl.__parse_row_update() assumed that every change that the database server
sent down actually modified the database. This is generally true, but
since Idl.__modify_row() already returns whether there was a change, we
might as well use it.
Ben Pfaff [Mon, 22 Aug 2011 21:31:18 +0000 (14:31 -0700)]
python: Use getattr() and setattr() instead of __dict__.
This leaves one use of __dict__ used for iterating through attributes.
I could use dir() instead, but I was put off by this note in its
documentation in the Python Library Reference:
Because dir() is supplied primarily as a convenience for use at an
interactive prompt, it tries to supply an interesting set of names more
than it tries to supply a rigorously or consistently defined set of names,
and its detailed behavior may change across releases. For example,
metaclass attributes are not in the result list when the argument is a
class.
Ben Pfaff [Tue, 23 Aug 2011 18:09:46 +0000 (11:09 -0700)]
ovs.stream: Fix logic bug in Stream.connect().
The loop here is supposed to run at least once, and to continue looping as
long as the loop body changes the current state, but this bug caused it to
continue looping until the connection completed in success or failure. It
probably didn't cause many problems in practice because only Unix domain
socket connections are currently supported, and those connections normally
complete immediately.
Simon Horman [Wed, 24 Aug 2011 01:40:52 +0000 (10:40 +0900)]
docs: Suppress "warning: macro `DD' not defined" warning
Suppress "warning: macro `DD' not defined" warning for ovs-brcompatd.8.
As per the description by Ben Pfaff for the same problem effecting
other files:
deamon.man allows the file that is including it to include extra
text in the description of --detach by defining a macro named DD.
Only some of the manpages that included it did this (only those
manpages that needed extra text there). But it's better to be
quiet in "man --warnings", so this defines DD to an empty value in
the other manpages that include daemon.man.
Simon Horman [Wed, 24 Aug 2011 01:40:51 +0000 (10:40 +0900)]
Debian: Add dependency on ${misc:Depends}
Add dependency on ${misc:Depends} to openvswitch-brcompat and ovsdbmonitor.
As reported by Lintian:
The source package uses debhelper, but it does not include
${misc:Depends} in the given binary package's debian/control entry.
Any debhelper command may add dependencies to ${misc:Depends} that
are required for the work that it does, so recommended best
practice is always add ${misc:Depends} to the dependencies of each
binary package if debhelper is in use.
Refer to the debhelper(7) manual page for details.
Justin Pettit [Thu, 18 Aug 2011 17:35:40 +0000 (10:35 -0700)]
datapath: Use "OVS_*" as opposed to "ODP_*" for user<->kernel interactions.
The prefix "ODP_*" is not overly descriptive in the context of the
larger Linux tree. This commit changes the prefix to "OVS_*" for the
userpace to kernel interactions. The userspace libraries still use
"ODP_" in many of their interfaces since it is more descriptive in the
OVS oeuvre.
Ben Pfaff [Fri, 19 Aug 2011 21:29:27 +0000 (14:29 -0700)]
ofproto-dpif: Delete MAC learning entries when they expire.
Commit fa066f015f716c7 "bridge: Move packet processing functionality into
ofproto" deleted the call to mac_learning_run() that deletes MAC learning
table entries when they expire. This fixes the problem.
Justin Pettit [Thu, 18 Aug 2011 16:57:23 +0000 (09:57 -0700)]
datapath: Correct comment for vport_add().
The comment describing vport_add() incorrectly stated that the function
added the vport to the datapath. It is the responsibility of the caller
to do that.
Ben Pfaff [Thu, 18 Aug 2011 17:33:32 +0000 (10:33 -0700)]
odp-util: Fix parsing of Ethertypes 0x8000 and above.
An existing comment in the function being updated explains the problem:
* Many of the sscanf calls in this function use oversized destination
* fields because some sscanf() implementations truncate the range of %i
* directives, so that e.g. "%"SCNi16 interprets input of "0xfedc" as a
* value of 0x7fff. The other alternatives are to allow only a single
* radix (e.g. decimal or hexadecimal) or to write more sophisticated
* parsers.