cascardo/ovs.git
9 years agodocs: Add titles to documentation files that lacked them.
Ben Pfaff [Thu, 27 Nov 2014 18:22:40 +0000 (10:22 -0800)]
docs: Add titles to documentation files that lacked them.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Thomas Graf <tgraf@noironetworks.com>
9 years agoovs-testcontroller: Fix manpage .so command.
Ben Pfaff [Thu, 27 Nov 2014 18:22:13 +0000 (10:22 -0800)]
ovs-testcontroller: Fix manpage .so command.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Thomas Graf <tgraf@noironetworks.com>
9 years agoroute-table: Remove Unregister.
Pravin B Shelar [Wed, 26 Nov 2014 19:27:05 +0000 (11:27 -0800)]
route-table: Remove Unregister.

Since dpif registering for routing table at initialization
there is no need to unregister it. Following patch removes
support for turning routing table notifications on and off.
Due to this change OVS always listens for these
notifications.

Reported-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
9 years agodatapath: Use ccflags-y instead of deprecated EXTRA_CFLAGS
Thomas Graf [Wed, 26 Nov 2014 14:52:31 +0000 (15:52 +0100)]
datapath: Use ccflags-y instead of deprecated EXTRA_CFLAGS

This allows users to pass in additional compiler flags through the
environment variable EXTRA_CFLAGS, e.g.

   make EXTRA_CFLAGS=-Wno-error=foo V=1

Reported-by: Alexandru Ardelean <ardeleanalex@gmail.com>
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
9 years agodpif-netdev: Fix a race.
Alex Wang [Tue, 25 Nov 2014 23:51:12 +0000 (15:51 -0800)]
dpif-netdev: Fix a race.

On current master, the 'struct dp_netdev_port' is destroyed
immediately when the ref count reaches 0.  However, non-pmd
threads calling the dpif_netdev_execute() for sending packets
could hold pointer to 'port' that is not ref-counted.  Thusly
those threads could possibly access freed memory when the port
is deleted.

To fix this bug, this commit makes non-pmd threads acquiring
the 'port_mutex' before doing the actual execution in
dpif_netdev_execute().

Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
9 years agovtep: Use unicast MAC as an example for a unicast destination in README.
Ben Pfaff [Mon, 1 Dec 2014 17:43:50 +0000 (09:43 -0800)]
vtep: Use unicast MAC as an example for a unicast destination in README.

Until now, the example for a unicast MAC has been a multicast MAC.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Justin Pettit <jpettit@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agolib: Use MAP_FOR_EACH_INDEX to improve readability.
Jarno Rajahalme [Wed, 26 Nov 2014 23:30:33 +0000 (15:30 -0800)]
lib: Use MAP_FOR_EACH_INDEX to improve readability.

Use MAP_FOR_EACH_INDEX to improve readability when there is no
apparent cost for it.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoflow.c: Improve minimask_equal() and minimask_has_extra().
Jarno Rajahalme [Wed, 26 Nov 2014 23:17:26 +0000 (15:17 -0800)]
flow.c: Improve minimask_equal() and minimask_has_extra().

minimask_equal() and minimask_has_extra() can take benefit from the
fact that minimasks have no zero data.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoflow: Define miniflow_get__() to simplify code.
Jarno Rajahalme [Wed, 26 Nov 2014 23:17:26 +0000 (15:17 -0800)]
flow: Define miniflow_get__() to simplify code.

miniflow_get__() can be used when it is known that the miniflow stores
a value at the given index.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoflow.h: Simplify MAP_FOR_EACH_INDEX()
Jarno Rajahalme [Fri, 21 Nov 2014 19:23:25 +0000 (11:23 -0800)]
flow.h: Simplify MAP_FOR_EACH_INDEX()

Avoid comparing against specific values.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agolib/route-table-*: Fix non-Linux builds
YAMAMOTO Takashi [Tue, 25 Nov 2014 07:56:06 +0000 (16:56 +0900)]
lib/route-table-*: Fix non-Linux builds

Fix build failures introduced by
commit 1bc50ef389d40be9ee215e66b49cf66fcbdb9eeb
("dpctl: Fix crash.")

Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agodpif-netlink: Fix windows build.
Joe Stringer [Wed, 26 Nov 2014 00:35:42 +0000 (16:35 -0800)]
dpif-netlink: Fix windows build.

Commit 7af12bd "dpif: Generate flow_hash for revalidators in dpif."
broke compilation on windows. Fix it.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoAdd Netronome vendor Id: NMX_VENDOR_ID = 0x00001540.
Simon Horman [Wed, 19 Nov 2014 00:44:55 +0000 (09:44 +0900)]
Add Netronome vendor Id: NMX_VENDOR_ID = 0x00001540.

This is based on the Netronome IEEE OUI, 00154D.
And it has been registered with the ONF:

https://rs.opennetworking.org/wiki/display/PUBLIC/ONF+Registry

Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoofproto: Fix checking of packet_in_mask in async config.
Shu Shen [Sat, 22 Nov 2014 00:27:25 +0000 (16:27 -0800)]
ofproto: Fix checking of packet_in_mask in async config.

The check shall use wire protool reasons, which could be different from
the internal packet-in reason.

Signed-off-by: Shu Shen <shu.shen@radisys.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agolib: Add API to set program name and version
Thomas Graf [Mon, 24 Nov 2014 11:49:01 +0000 (12:49 +0100)]
lib: Add API to set program name and version

Required to have reasonable logging messages.

Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agolib: Add new header <openvswitch/version.h> to versioning info
Thomas Graf [Mon, 24 Nov 2014 11:49:00 +0000 (12:49 +0100)]
lib: Add new header <openvswitch/version.h> to versioning info

Exposes the package version as string and the library versioning
as numeric macro.

OVS_LIB_VERSION may be used to check for a particular version of
the interface.

OVS_LIB_REVISION and OVS_LIB_AGE exist to map a set of installed
headers to a particular shared library instance.

Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agodpif: Generate flow_hash for revalidators in dpif.
Joe Stringer [Wed, 24 Sep 2014 03:24:39 +0000 (15:24 +1200)]
dpif: Generate flow_hash for revalidators in dpif.

This patch shifts the responsibility for determining the hash for a flow
from the revalidation logic down to the dpif layer. This assists in
handling backward-compatibility for revalidation with the upcoming
unique flow identifier "UFID" patches.

A 128-bit UFID was selected to minimize the likelihood of hash conflicts.
Handler threads will not install a flow that has an identical UFID as
another flow, to prevent misattribution of stats and to ensure that the
correct flow key cache is used for revalidation.

For datapaths that do not support UFID, which is currently all
datapaths, the dpif will generate the UFID and pass it up during upcall
and flow_dump. This is generated based on the datapath flow key.

Later patches will add support for datapaths to store and interpret this
UFID, in which case the dpif has a responsibility to pass it through
transparently.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agohash: Add 128-bit murmurhash.
Joe Stringer [Mon, 11 Aug 2014 23:12:12 +0000 (11:12 +1200)]
hash: Add 128-bit murmurhash.

Add the 128-bit murmurhash by Austin Appleby, r150 from:
http://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoupcall: Revalidate using cache of mask, actions.
Joe Stringer [Wed, 6 Aug 2014 04:40:37 +0000 (16:40 +1200)]
upcall: Revalidate using cache of mask, actions.

This allows us to ignore most fields of a flow_dump, requiring only the
flow key for looking up the ukey. Fetching flows can also be avoided in
the corner case where a flow is missed from a dump but revalidation is
required.

A future patch will modify the datapath interface to allow datapaths to
skip dumping these fields, so this cache will be used instead.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoupcall: Create ukeys in handler threads.
Joe Stringer [Fri, 25 Jul 2014 01:54:24 +0000 (13:54 +1200)]
upcall: Create ukeys in handler threads.

Currently, when a revalidator thread first dumps a flow, it creates a
'udpif_key' object and caches a copy of a kernel flow key. This allows
us to perform lookups in the classifier to attribute stats and validate
the correctness of the datapath flow.

This patch sets up this cache from the handler threads, during flow
setup. While this patch alone causes a decrease in revalidation
performance, it allows future patches increase performance by reducing
the cost of flow dumping.

Revalidators will continue to create ukeys if a flow is dumped that has
no corresponding ukey. This may happen in corner cases such as when
ovs-vswitchd is restarted (and flows remain in the datapath) or a user
installs a flow using ovs-dpctl.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoupcall: Rename dump_op -> ukey_op.
Joe Stringer [Wed, 20 Aug 2014 12:21:03 +0000 (00:21 +1200)]
upcall: Rename dump_op -> ukey_op.

Future patches will make use of the 'struct dump_op' in a broader sense,
so this patch renames it to make things a bit clearer.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoudpif: Separate udpif_key maps from revalidators.
Joe Stringer [Thu, 5 Jun 2014 05:28:46 +0000 (17:28 +1200)]
udpif: Separate udpif_key maps from revalidators.

An upcoming patch will change the access patterns for ukey maps to
increase the number of writers, and shift write-access from revalidator
threads to upcall handler threads. As such, it no longer makes sense to
tie these maps to revalidators in a 1:1 relationship.

This patch separates the ukey maps from the revalidators, and increases
the number of maps used to store ukeys, to reduce contention.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agorevalidator: Protect ukeys with a mutex.
Joe Stringer [Mon, 5 May 2014 03:44:04 +0000 (15:44 +1200)]
revalidator: Protect ukeys with a mutex.

Currently, udpif_keys are protected during revalidator_sweep__() as only
one thread accesses the ukey at a time. This is ensured using barriers:
all revalidators will be in the GC phase, so they will only access their
own ukey collection.

A future patch will change the access patterns to allow these ukey
collections to be read or modified while a revalidator is garbage
collecting it. To protect the ukeys, this patch adds locking on the ukey
collection.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agorevalidator: Use 'cmap' for storing ukeys.
Joe Stringer [Mon, 5 May 2014 03:44:40 +0000 (15:44 +1200)]
revalidator: Use 'cmap' for storing ukeys.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agodatapath: Don't validate IPv6 label masks.
Joe Stringer [Tue, 25 Nov 2014 21:02:42 +0000 (13:02 -0800)]
datapath: Don't validate IPv6 label masks.

When userspace doesn't provide a mask, OVS datapath generates a fully
unwildcarded mask for the flow by copying the flow and setting all bits
in all fields. For IPv6 label, this creates a mask that matches on the
upper 12 bits, causing the following error:

openvswitch: netlink: Invalid IPv6 flow label value (value=ffffffff, max=fffff)

This patch ignores the label validation check for masks, avoiding this
error.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovswitch.xml: Describe NetFlow passive timeout.
Ben Pfaff [Wed, 5 Nov 2014 16:34:51 +0000 (08:34 -0800)]
vswitch.xml: Describe NetFlow passive timeout.

Reported-by: Martin Vizvary <vizvary@ics.muni.cz>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agotests/ovs_client: Remove broken debug code
Stephen Finucane [Wed, 19 Nov 2014 12:03:22 +0000 (12:03 +0000)]
tests/ovs_client: Remove broken debug code

There is some debug "logging" code included in the 'ovs_client'
application. This code appears to be broken as it spams stdout with
duplicated "log" messages. Remove this code.
This patch also renames ovs_client to ring_client.

In addition, add the 'test-dpdkr' executable to the 'tests/.gitignore'
file.

Signed-off-by: Stephen Finucane <stephen.finucane@intel.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
9 years agodatapath-windows: Update DESIGN document.
Nithin Raju [Tue, 25 Nov 2014 17:06:43 +0000 (09:06 -0800)]
datapath-windows: Update DESIGN document.

In this patch, we update the design document to reflect the netlink
based kernel-userspace interface implementation and a few other changes.
I have covered at a high level.

Please feel free to extend the document with more details that you think
got missed out.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoovs-vsctl: Prevent creating duplicate VLAN bridges.
Ben Pfaff [Thu, 6 Nov 2014 22:59:40 +0000 (14:59 -0800)]
ovs-vsctl: Prevent creating duplicate VLAN bridges.

ovs-vsctl has the concept of a VLAN (or "fake") bridge, which is a
sort of a sub-bridge that receives only packets on a particular VLAN.
There is no way to distinguish two VLAN bridges with the same parent on the
same VLAN, but until now ovs-vsctl did not prevent creating duplicates or
report them.  This commit fixes the problem.

Reported-by: rwxybh
Reported-at: https://github.com/openvswitch/ovs/issues/21
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Thomas Graf <tgraf@noironetworks.com>
9 years agoMakefile: Add static check for misuse of LITTLE_ENDIAN and BIG_ENDIAN.
Ben Pfaff [Tue, 25 Nov 2014 17:00:26 +0000 (09:00 -0800)]
Makefile: Add static check for misuse of LITTLE_ENDIAN and BIG_ENDIAN.

I've pointed out two misuses of these macros in review in the last week,
by different authors.  It's time to make it difficult to screw this up.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Thomas Graf <tgraf@noironetworks.com>
9 years agodatapath-windows: Fix broken build on MSVC.
Alin Serdean [Tue, 25 Nov 2014 15:52:28 +0000 (15:52 +0000)]
datapath-windows: Fix broken build on MSVC.

Removing unused local variable.

Signed-off-by: Alin Gabriel Serdean <aserdean@cloudbasesolutions.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agodatapath-windows: Update OVSEXT VS project to support 6.40
Sorin Vinturis [Mon, 24 Nov 2014 16:17:34 +0000 (16:17 +0000)]
datapath-windows: Update OVSEXT VS project to support 6.40

Added support for creating OVS extension driver NDIS 6.40 compliant.

Currently the OVSEXT Visual Studio project has four build configurations,
'Win8 Release', 'Win8 Debug', 'Win8.1 Release' and 'Win8.1 Debug'. All of
them are creating a binary that is NDIS 6.30 compliant. I have changed the
Win8.1 build configurations in order to create a binary that is NDIS 6.40
compliant.

In this way, the OVSEXT project is able to create a release/debug binary
that is NDIS 6.30 compliant, using the 'Win8 Release' and 'Win8 Debug'
build configurations, as well as a release/debug binary that is NDIS 6.40
compliant, using the 'Win8.1 Release' and 'Win8.1 Debug' build
configurations.

Signed-off-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Acked-by: Eitan Eliahu <eliahue@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agodatapath-windows: Fixed Release broken build.
Sorin Vinturis [Mon, 24 Nov 2014 16:26:00 +0000 (16:26 +0000)]
datapath-windows: Fixed Release broken build.

The release configurations of the OVSEXT project were not compiling.
This was due to a warning that was treated as error. Fixed that.

I did not want to remove that variable, because it is used in an
ASSERT in that function.

Signed-off-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Acked-by: Alin Gabriel Serdean <aserdean@cloudbasesolutions.com>
9 years agotests: Fix test broken by introduction of ONFOXM_ET_TCP_FLAGS.
Ben Pfaff [Tue, 25 Nov 2014 16:37:24 +0000 (08:37 -0800)]
tests: Fix test broken by introduction of ONFOXM_ET_TCP_FLAGS.

Commit 847ddeab372220 (meta-flow: Add support for ONF OpenFlow 1.3
extension ONFOXM_ET_TCP_FLAGS.) failed to update a test to match the new
OXM extension.  This fixes the problem.

Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoofproto-dpif: Add test case for OF1.4 packet-in.
Shu Shen [Fri, 21 Nov 2014 19:44:23 +0000 (11:44 -0800)]
ofproto-dpif: Add test case for OF1.4 packet-in.

The test case current fails and shows a bug when OF1.4 is used between
the controller and the switch, the packet-in message still uses OF1.3
wire protocol version.

Signed-off-by: Shu Shen <shu.shen@radisys.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoopenflow: Use *_array_len names in struct ofp15_bucket and ofp15_group_mod
Simon Horman [Fri, 21 Nov 2014 01:16:23 +0000 (10:16 +0900)]
openflow: Use *_array_len names in struct ofp15_bucket and ofp15_group_mod

The spec has been clarified to use _list_len in palce of _list_len
terminology to make it clearer that the data is not an ordered list
(it is a set). The code present in Open vSwitch already avoided
the _list_len terminology. This change brings the code into
line with the updated spec.

ONF-JIRA: EXT-350
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoopenflow: Value 4 of enum ofp15_group_mod_command is reserved.
Simon Horman [Fri, 21 Nov 2014 01:15:32 +0000 (10:15 +0900)]
openflow: Value 4 of enum ofp15_group_mod_command is reserved.

The spec has been clarified to note that element 4 of
enum ofp_group_mod_command is reserved. This patch reflects
that change.

ONF-JIRA: EXT-350
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agometa-flow: Add support for ONF OpenFlow 1.3 extension ONFOXM_ET_TCP_FLAGS.
Ben Pfaff [Mon, 24 Nov 2014 23:06:34 +0000 (15:06 -0800)]
meta-flow: Add support for ONF OpenFlow 1.3 extension ONFOXM_ET_TCP_FLAGS.

Actually the draft extension says "TCP_FLAG", without an "S", but I think
that's just a typo.

ONF-JIRA: EXT-109
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agodatapath-windows: Move vport code from Datapath.c to Vport.c.
Nithin Raju [Fri, 21 Nov 2014 00:51:37 +0000 (16:51 -0800)]
datapath-windows: Move vport code from Datapath.c to Vport.c.

Plain code motion. No additional cleanup or refactoring peformed.

I'll be doing some code refactoring in subsequent patches.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agodpif: Fix initialization order.
Pravin B Shelar [Mon, 24 Nov 2014 23:07:23 +0000 (15:07 -0800)]
dpif: Fix initialization order.

OVS router depends on tnl_conf_seq and all tunnel related
components should be initialized before registering dpif
implementations.

Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agodatapath-windows: Support for hybrid forwarding
Sorin Vinturis [Mon, 24 Nov 2014 22:32:06 +0000 (22:32 +0000)]
datapath-windows: Support for hybrid forwarding

In a hybrid forwarding environment, there are two types of packets that
enter and leave the Hyper-V extensible switch: NVGRE packets and non-NVGRE
packets. Hybrid forwarding involves filtering the incoming traffic based
on packet type.

Thus, we must split the incoming traffic into NVGRE and non-NVGRE packets.
All non-NVGRE traffic is forwarded by the OVS extension and processed as
usual, and the NVGRE traffic is passed to NDIS to be handled by the HNV
module.

Necessary VS project changes regarding the compiler settings will be made
in a different patch.

More details about hybrid forwarding and the necessary NDIS 6.40 support
is provided in issue #52.

Signed-off-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Reported-by: Nithin Raju <nithin@vmware.com>
Reported-at: https://github.com/openvswitch/ovs-issues/issues/52
Acked-by: Nithin Raju <nithin@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoOPENFLOW-1.1+.md: Add OpenFlow 1.5 features.
Jean Tourrilhes [Mon, 24 Nov 2014 23:35:53 +0000 (15:35 -0800)]
OPENFLOW-1.1+.md: Add OpenFlow 1.5 features.

Signed-off-by: Jean Tourrilhes <jt@hpl.hp.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoOPENFLOW-1.1+.md: Update and correct OpenFlow 1.4 features.
Jean Tourrilhes [Sat, 22 Nov 2014 02:05:10 +0000 (18:05 -0800)]
OPENFLOW-1.1+.md: Update and correct OpenFlow 1.4 features.

Signed-off-by: Jean Tourrilhes <jt@hpl.hp.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agodatapath-windows: Removed duplicate checking of previous dump state
Sorin Vinturis [Tue, 18 Nov 2014 14:20:39 +0000 (14:20 +0000)]
datapath-windows: Removed duplicate checking of previous dump state

Removed the check for previous dump state before calling FreeUserDumpState
as the check is also performed in the latter function.

Signed-off-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Acked-by: Nithin Raju <nithin@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agodatapath-windows: BSOD for a transactional NL cmd without dump state.
Sorin Vinturis [Tue, 18 Nov 2014 14:20:39 +0000 (14:20 +0000)]
datapath-windows: BSOD for a transactional NL cmd without dump state.

The BSOD is triggered by the incorect extraction of the OVS message
from the received input buffer from userspace, in _FlowNlGetCmdHandler
function.

Signed-off-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Reported-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Reported-at: https://github.com/openvswitch/ovs-issues/issues/54
Acked-by: Nithin Raju <nithin@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoovs-ofctl: Document that mod-group requires the type field
Simon Horman [Tue, 18 Nov 2014 02:31:51 +0000 (11:31 +0900)]
ovs-ofctl: Document that mod-group requires the type field

Document that mod-group rather than del-group requires the type field.

Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoofp-errors: Use official extension code for OFPBAC_BAD_SET_MASK.
Ben Pfaff [Mon, 24 Nov 2014 18:27:01 +0000 (10:27 -0800)]
ofp-errors: Use official extension code for OFPBAC_BAD_SET_MASK.

ONF-JIRA: EXT-425
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agometa-flow: OXM_OF_PKT_REG<n> now available as (draft) OF1.3 extension.
Ben Pfaff [Mon, 24 Nov 2014 18:41:54 +0000 (10:41 -0800)]
meta-flow: OXM_OF_PKT_REG<n> now available as (draft) OF1.3 extension.

ONF-JIRA: EXT-244
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agoofp-util.c: Enable packet-in messages for OpenFlow 1.4 and 1.5.
Jean Tourrilhes [Thu, 13 Nov 2014 19:35:21 +0000 (11:35 -0800)]
ofp-util.c: Enable packet-in messages for OpenFlow 1.4 and 1.5.

Signed-off-by: Jean Tourrilhes <jt@hpl.hp.com>
[blp@nicira.com added simple test]
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agotests: Add command to purge revalidators of flows.
Joe Stringer [Thu, 13 Nov 2014 18:42:47 +0000 (10:42 -0800)]
tests: Add command to purge revalidators of flows.

This patch adds a new 'ovs-appctl revalidator/purge' command which
flushes all flows from all datapaths, and updates the revalidator
udpif_key cache at the same time.

Update the ofproto-dpif fragment tests which may fail when ukeys are
created from handler threads.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agodpif-netdev: Remove redundant hash action handling.
Pravin B Shelar [Thu, 13 Nov 2014 23:03:39 +0000 (15:03 -0800)]
dpif-netdev: Remove redundant hash action handling.

odp_execute_actions() already handles hash execution part.

Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agodpctl: Fix crash.
Pravin B Shelar [Thu, 20 Nov 2014 06:12:21 +0000 (22:12 -0800)]
dpctl: Fix crash.

ovs-dpctl crashed due to uninitialized router classifier. To
fix this issue move ovs router initialization to route table
module.

Reported-by: Madhu Challa <challa@noironetworks.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agotunnel-push-pop.at: Make test more deterministic.
Jarno Rajahalme [Fri, 21 Nov 2014 23:42:42 +0000 (15:42 -0800)]
tunnel-push-pop.at: Make test more deterministic.

Using the '-march=native' broke this test in a i7 laptop, most likely
due to a different hash implementation, which resulted in different
datapath port number assignment accross different compilations.

Break up the ports setup so that the datapath port number assignment
is more deterministic.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
9 years agonetlink-socket: Set socket pid number in NL message on Windows.
Eitan Eliahu [Sat, 22 Nov 2014 01:34:56 +0000 (17:34 -0800)]
netlink-socket: Set socket pid number in NL message on Windows.

The pid must be set in the NL header as the driver checks it against the pid in
the instance paired with the socket.

Signed-off-by: Eitan Eliahu <eliahue@vmware.com>
Acked-by: Nithin Raju <nithin@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agodatapath-windows: do not complete an IRP when pending status returned
Eitan Eliahu [Sat, 22 Nov 2014 00:58:54 +0000 (16:58 -0800)]
datapath-windows: do not complete an IRP when pending status returned

The driver completes a pending IRP, the IRP is being marked free by the
I/O manager which causes a system crash when the IRP is completed at
the second time.
The driver should not complete a pending IRP in the context of the
Device I/O control thread.

*** Fatal System Error: 0x00000044
                       (0xFFFFFA800CF897D0,0x0000000000000F7A,0x0000000000000000,0x0000000000000000)

Break instruction exception - code 80000003 (first chance)

BugCheck 44, {fffffa800cf897d0, f7a, 0, 0}

Probably caused by : OVSExt.sys ( OVSExt!OvsCompleteIrpRequest+3a )

Followup: MachineOwner
---------

nt!DbgBreakPointWithStatus:
fffff800`5a100930 cc              int     3
11: kd> !analyze -v
*******************************************************************************
*                                                                             *
*                        Bugcheck Analysis                                    *
*                                                                             *
*******************************************************************************

MULTIPLE_IRP_COMPLETE_REQUESTS (44)
A driver has requested that an IRP be completed (IoCompleteRequest()), but
the packet has already been completed.  This is a tough bug to find because
the easiest case, a driver actually attempted to complete its own packet
twice, is generally not what happened.  Rather, two separate drivers each
believe that they own the packet, and each attempts to complete it.  The
first actually works, and the second fails.  Tracking down which drivers
in the system actually did this is difficult, generally because the trails
of the first driver have been covered by the second.  However, the driver
stack for the current request can be found by examining the DeviceObject
fields in each of the stack locations.
Arguments:
Arg1: fffffa800cf897d0, Address of the IRP
Arg2: 0000000000000f7a
Arg3: 0000000000000000
Arg4: 0000000000000000

Signed-off-by: Eitan Eliahu <eliahue@vmware.com>
Acked-by: Nithin Raju <nithin@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agoxen.spec: Fix xenserver build failure.
Alex Wang [Fri, 21 Nov 2014 18:55:03 +0000 (10:55 -0800)]
xen.spec: Fix xenserver build failure.

This commit conducts similar operations as commit 2fd84d5e5
(rhel: Include pkgconfig and update header install in
Fedora/RHEL spec.) to xenserver spec file to prevent the
following rpmbuild failure.

  RPM build errors:
       Installed (but unpackaged) file(s) found:
           /usr/include/openflow/nicira-ext.h
           /usr/include/openflow/openflow-1.0.h
           /usr/include/openflow/openflow-1.1.h
           /usr/include/openflow/openflow-1.2.h
           /usr/include/openflow/openflow-1.3.h
           /usr/include/openflow/openflow-1.4.h
           /usr/include/openflow/openflow-1.5.h
           /usr/include/openflow/openflow-common.h
           /usr/include/openflow/openflow.h
           /usr/include/openvswitch/types.h
           /usr/lib/pkgconfig/libofproto.pc
           /usr/lib/pkgconfig/libopenvswitch.pc
           /usr/lib/pkgconfig/libovsdb.pc
           /usr/lib/pkgconfig/libsflow.pc

VMware-BZ: #1362092

Reported-by: Edwin Chiu <echiu@nicira.com>
Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agocmap: Shrink cmap when load factor is below 20%.
Alex Wang [Thu, 20 Nov 2014 18:41:52 +0000 (10:41 -0800)]
cmap: Shrink cmap when load factor is below 20%.

This commit adds check in cmap_remove() and shrinks the cmap by half
if the load factor is below 20%.  This is to reduce the memory
utilization of cmap and to avoid the allocated cmap memory occupying
the top of heap memory, preventing the trim of heap.

Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agodpif-netdev: Garbage collect the exact match cache periodically.
Alex Wang [Fri, 14 Nov 2014 23:54:56 +0000 (15:54 -0800)]
dpif-netdev: Garbage collect the exact match cache periodically.

On current master, the exact match cache entry can keep reference to
'struct dp_netdev_flow' even after the flow is removed from the flow
table.  This means the free of allocated memory of the flow is delayed
until the exact match cache entry is cleared or replaced.

If the allocated memory is ahead of chunks of freed memory on heap,
the delay will prevent the reclaim of those freed chunks, causing
falsely high memory utilization.

To fix the issue, this commit makes the owning thread conduct periodic
garbage collection on the exact match cache and clear dead entries.

Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
---
PATCH -> V2:
- Adopt Jarno's suggestion and conduct slow sweep to avoid introducing
  jitter.

9 years agoovs-vsctl: Update RSTP documentation.
Daniele Venturino [Thu, 20 Nov 2014 17:49:10 +0000 (18:49 +0100)]
ovs-vsctl: Update RSTP documentation.

Mention that the AUTO detection of admin point to point MAC parameter
is not implemented.

Signed-off-by: Martino Fornasa <mf@fornasa.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agoNEWS: Information about Ixia ANVL testing.
Daniele Venturino [Thu, 20 Nov 2014 17:49:09 +0000 (18:49 +0100)]
NEWS: Information about Ixia ANVL testing.

Update the NEWS section with some information about testing.

Signed-off-by: Martino Fornasa <mf@fornasa.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agodatapath-windows: rename 'hvDeleted' to 'isPresentOnHv'
Nithin Raju [Thu, 20 Nov 2014 03:41:22 +0000 (19:41 -0800)]
datapath-windows: rename 'hvDeleted' to 'isPresentOnHv'

'hvDeleted' was originally added to track vports that got deleted from
the Hyper-V switch, but which still exists in OVS. Since then, it has
taken a new meaning: eg. it is set to TRUE for VXLAN ports which don't
exist on the Hyper-V switch at all.

Hence renaming it to something more appropriate.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatpaath-windows: consolidate check for output buffer
Nithin Raju [Thu, 20 Nov 2014 17:06:34 +0000 (09:06 -0800)]
datpaath-windows: consolidate check for output buffer

Earlier, output buffer was optional in a transaction command. Thus each
command handler, had to check if the output buffer had indeed been
specified by userspace or not.

Now that output buffer is mandatory in transaction command, let's
consolidate the check in one place, and also convert the previous check
to ASSERTs.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: cleanup DpInternal.h
Nithin Raju [Thu, 20 Nov 2014 17:06:33 +0000 (09:06 -0800)]
datapath-windows: cleanup DpInternal.h

Getting rid of unused definitions in DpInternal.h.

This is only the first round. There's scope for more cleanup.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatpath-windows: include/OvsPub.h => ovsext/DpInternal.h
Nithin Raju [Wed, 19 Nov 2014 17:14:33 +0000 (09:14 -0800)]
datpath-windows: include/OvsPub.h => ovsext/DpInternal.h

OvsPub.h is not longer the interface file that published the interface
of the kernel datapath to userspace. Nevertheless it is still being used
internal to the kernel datapath.

We rename the file for this reason.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: nuke non-netlink based interface
Nithin Raju [Wed, 19 Nov 2014 17:14:32 +0000 (09:14 -0800)]
datapath-windows: nuke non-netlink based interface

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: Avoid BSOD when switch context is NULL
Sorin Vinturis [Wed, 19 Nov 2014 07:16:32 +0000 (07:16 +0000)]
datapath-windows: Avoid BSOD when switch context is NULL

I came around a BSOD that happened when trying to access pidHashLock
from the gOvsSwitchContext, which was NULL. The stop happened in
OvsAcquirePidHashLock function.

To reproduce this BSOD, make sure the extension is enabled and running,
disable it and, after that, execute 'ovs-dpctl.exe show'. The BSOD is
triggered afterwards.

Signed-off-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Reported-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Reported-at: https://github.com/openvswitch/ovs-issues/issues/53
Acked-by: Eitan Eliahu <eliahue@vmware.com>
Acked-by: Nithin Raju <nithin@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: Fixes in HvCreatePort() to re-add a port
Nithin Raju [Wed, 19 Nov 2014 01:42:48 +0000 (17:42 -0800)]
datapath-windows: Fixes in HvCreatePort() to re-add a port

In this patch, we update HvCreatePort() to be able to re-add a Hyper-V
port. Specifically, we handle the case where the port had also been
added by OVS userspace, so that when the port was previously deleted
from Hyper-V, we did not deallocate the port.

The key to a vport is its name. We lookup the list of vports both in the
'portIdHashArray' as well as 'portNoHashArray' to make sure that we
don't have a port with the same name.

Validation:
- deleted an re-added a port with and without the corresponding OVS port
existing
- deleted, changed the name of a port, and re-added it back with and
without the corresponding OVS port existing.
- uninstall was succcessful. No asserts hit.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Ankur Sharma <ankursharma@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: Change return type of OvsInitVxlanTunnel()
Nithin Raju [Wed, 19 Nov 2014 01:42:47 +0000 (17:42 -0800)]
datapath-windows: Change return type of OvsInitVxlanTunnel()

OvsInitVxlanTunnel() need not return a NL_ERROR. In this patch, we
change it to NTSTATUS, and also update the mapping function that maps a
NTSTATUS to NL_ERROR.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Ankur Sharma <ankursharma@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: Changes to InitOvsVportCommon()
Nithin Raju [Wed, 19 Nov 2014 01:42:46 +0000 (17:42 -0800)]
datapath-windows: Changes to InitOvsVportCommon()

If a Hyper-V port (for which there exists an OVS port), gets deleted and
re-added, we'll call into InitOvsVportCommon() for the port to insert
the port into the 'portIdHashArray' as well as do a few other
initialization in the switch ocntext.

We should not be incrementing 'numHvPorts' at this point since this
vport has been counted before when it was first allocated. To account
for this, we add a new parameter to InitOvsVportCommon(). The arguments
passed by some of the callers are not 100% correct, and will be fixed in
future commit in the series.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Ankur Sharma <ankursharma@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: Changes to OvsRemoveAndDeleteVport()
Nithin Raju [Wed, 19 Nov 2014 01:42:45 +0000 (17:42 -0800)]
datapath-windows: Changes to OvsRemoveAndDeleteVport()

In this patch, we add changes to OvsRemoveAndDeleteVport() to allow the
caller to specify if a vport is being deleted because it got deleted on
Hyper-V or if it got deleted from OVS userspace.

The reason we need to make the distinction is to be able to delete the
vport from the relevant hash tables.

If a port has been deleted from all the hash tables ie. has been deleted
from Hyper-V as well as OVS userspace, it gets deallocated.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Ankur Sharma <ankursharma@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: clarify 'portIdHashArray' in the swithc context
Nithin Raju [Wed, 19 Nov 2014 01:42:44 +0000 (17:42 -0800)]
datapath-windows: clarify 'portIdHashArray' in the swithc context

In this patch, we update the documentation for 'portIdHashArray' to
indicate that a vport would exist in this hash table if and only if it
also exists on the Hyper-V switch.

This functionality to implement this semantic will follow in subsequent
patches.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: add WCHAR version of OvsFindVportByHvName()
Nithin Raju [Wed, 19 Nov 2014 01:42:43 +0000 (17:42 -0800)]
datapath-windows: add WCHAR version of OvsFindVportByHvName()

In this patch, we rename the existing OvsFindVportByHvName() to
OvsFindVportByHvNameA() to indicate that the input string is an ASCII
string. We also define a OvsFindVportByHvNameW() that takes as input a
WCHAR string.

This will be used later in HvCreatPort() to check for ports with
duplicate names.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Ankur Sharma <ankursharma@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath-windows: fixes in OvsGetExtInfoIoctl()
Nithin Raju [Wed, 19 Nov 2014 01:42:42 +0000 (17:42 -0800)]
datapath-windows: fixes in OvsGetExtInfoIoctl()

A few fixes around locking.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Ankur Sharma <ankursharma@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agoFix build break in ofproto/tunnel.c for windows platform.
Saurabh Shah [Wed, 19 Nov 2014 11:28:41 +0000 (03:28 -0800)]
Fix build break in ofproto/tunnel.c for windows platform.

The breakage was introduced by commit: a36de779
("openvswitch: Userspace tunneling.").

Reported-by: Edwin Chiu <echiu@nicira.com>
Signed-off-by: Saurabh Shah <ssaurabh@vmware.com>
Acked-by: Nithin Raju <nithin@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
9 years agodatapath: compat: Fix build on RHEL 6.6
Pravin B Shelar [Wed, 12 Nov 2014 02:56:34 +0000 (18:56 -0800)]
datapath: compat: Fix build on RHEL 6.6

RHEL 6.6 kernel percpu APIs are broken, so following patch is using OVS
backported version.

Reported-by: Wang Sheng-Hui <shhuiw@gmail.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
9 years agoofproto-dpif: fix bundle_move() call in rstp_run().
Daniele Venturino [Wed, 19 Nov 2014 21:16:17 +0000 (22:16 +0100)]
ofproto-dpif: fix bundle_move() call in rstp_run().

bundle_move() expects two struct ofbundle pointers as arguments, but
void pointers hide the error.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp: refactor RSTP setters.
Daniele Venturino [Thu, 6 Nov 2014 15:31:15 +0000 (16:31 +0100)]
rstp: refactor RSTP setters.

With this patch setters invoke procedures only if values have changed.
Also rstp_set_bridge_address__() keeps the existing priority in the
bridge_identifier.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp-state-machines: fix proposal reception behaviour.
Jarno Rajahalme [Fri, 14 Nov 2014 21:58:25 +0000 (13:58 -0800)]
rstp-state-machines: fix proposal reception behaviour.

Other ports should stop forwarding and learning when a port receives a
superior BPDU carrying a Proposal flag.
Without this patch this does not happen and other ports keep executing
the learning and forwarding processes.
This patch contains some fixes reported in the 802.1q-2008 standard.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp: Fix global transitions.
Jarno Rajahalme [Wed, 19 Nov 2014 17:27:56 +0000 (09:27 -0800)]
rstp: Fix global transitions.

Global transitions are highest priority transitions.  When the
condition associated with a global transition is met, it supersedes
all other exit conditions including UCT.

Extracted from 802.1D-2004 standard (17.16):

A transition that is global in nature (i.e., a transition that occurs
from any of the possible states if the condition attached to the arrow
is met) is denoted by an open arrow, i.e., no specific state is
identified as the origin of the transition.  When the condition
associated with a global transition is met, it supersedes all other
exit conditions including UCT.  The special global condition BEGIN
supersedes all other global conditions, and once asserted remains
asserted until all state blocks have executed to the point that
variable assignments and other consequences of their execution remain
unchanged.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp: shift learned MAC addresses to new Root port.
Jarno Rajahalme [Wed, 19 Nov 2014 17:24:52 +0000 (09:24 -0800)]
rstp: shift learned MAC addresses to new Root port.

All MAC addresses previously learned on a Root Port can be moved to an
Alternate Port that becomes the new Root Port; i.e., Dynamic Filtering
Entries for those addresses may be modified to show the new Root Port as
their source, reducing the need to flood frames when recovering from
some component failures.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agoclassifier: Defer pvector publication.
Jarno Rajahalme [Thu, 13 Nov 2014 19:54:31 +0000 (11:54 -0800)]
classifier: Defer pvector publication.

This patch adds a new functions classifier_defer() and
classifier_publish(), which control when the classifier modifications
are made available to lookups.  By default, all modifications are made
available to lookups immediately.  Modifications made after a
classifier_defer() call MAY be 'deferred' for later 'publication'.  A
call to classifier_publish() will both publish any deferred
modifications, and cause subsequent changes to to be published
immediately.

Currently any deferring is limited to the visibility of the subtable
vector changes.  pvector now processes modifications mostly in a
working copy, which needs to be explicitly published with
pvector_publish().  pvector_publish() sorts the working copy and
removes gaps before publishing it.

This change helps avoiding O(n**2) memory behavior in corner cases,
where large number of rules with different masks are inserted or
deleted.

VMware-BZ: #1322017
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoclassifier: Make insert and replace take a const rule, too.
Jarno Rajahalme [Thu, 13 Nov 2014 19:54:31 +0000 (11:54 -0800)]
classifier: Make insert and replace take a const rule, too.

classifier_remove() was recently changed to take a const struct
cls_rule *.  Make the corresponding change to classifier_replace() and
classifier_insert().  This simplifies existing calling sites in
ofproto.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoclassifier: Remove internal mutex.
Jarno Rajahalme [Fri, 14 Nov 2014 23:58:09 +0000 (15:58 -0800)]
classifier: Remove internal mutex.

Almost all classifier users already exclude concurrent modifications,
or are single-threaded, hence the classifier internal mutex can be
removed.  Due to this change, ovs-router.c and tnl-ports.c need new
mutexes, which are added.

As noted by Ben in review, ovs_router_flush() should also free the
entries it removes from the classifier.  It now calls
ovsrcu_postpone() to that effect.

Suggested-by: Ben Pfaff <blp@nicira.com>
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoclassifier: Lockless and robust classifier iteration.
Jarno Rajahalme [Thu, 13 Nov 2014 19:54:31 +0000 (11:54 -0800)]
classifier: Lockless and robust classifier iteration.

Previously, accurate iteration required writers to be excluded during
iteration.  This patch adds an rculist to struct cls_subtable, and a
corresponding list node to struct cls_rule, which makes iteration more
straightforward, and allows the iterators to remain ignorant of the
internals of the cls_match.  This new list allows iteration of rules
in the classifier by traversing the RCU-friendly subtables vector, and
the rculist of rules in each subtable.

Classifier modifications may be performed concurrently, but whether or
not the concurrent iterator sees those changes depends on the timing
of change.  More specifically, an concurrent iterator:

- May or may not see a rule that is being inserted or removed.
- Will see either the new or the old version of a rule that is replaced.
- Will see all the other rules (that are not being modified).

Finally, The subtable's rculist also allows to make
classifier_rule_overlaps() lockless, which this patch also does.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agoclassifier: Do not insert duplicate rules in indices.
Jarno Rajahalme [Fri, 14 Nov 2014 22:47:03 +0000 (14:47 -0800)]
classifier: Do not insert duplicate rules in indices.

There is no point in adding duplicate information into prefix tries.

Also, since the lower-priority duplicate rules are not visible to
lookups, they do not need to be in staged lookup indices directly
either (the head rule is).

Finally, now that cmap operations return the number of elements in the
cmap, subtable's 'n_rules' member is not needed any more.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
9 years agodatapath-windows: Remove all duplicate checks for NULL.
Sorin Vinturis [Fri, 14 Nov 2014 11:35:21 +0000 (11:35 +0000)]
datapath-windows: Remove all duplicate checks for NULL.

Right now the gOvsSwitchContext pointer is checked against NULL
in a lot of places of the OVS extension code. This check should
be done only once to avoid wasteful checks. Thus I have added the
check in the dispatch routine, before doing any processing, and
removed all other checks from the rest of the code.

Signed-off-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Acked-by: Eitan Eliahu <eliahue@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
9 years agorstp-state-machines: fix rcv_info().
Daniele Venturino [Fri, 14 Nov 2014 22:07:51 +0000 (14:07 -0800)]
rstp-state-machines: fix rcv_info().

A Config BPDU always conveys a Designated Port Role.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp: Refactor rstp_port_set_administrative_bridge_port__().
Jarno Rajahalme [Fri, 14 Nov 2014 22:07:51 +0000 (14:07 -0800)]
rstp: Refactor rstp_port_set_administrative_bridge_port__().

Invoke move_rstp__() in rstp_port_set_administrative_bridge_port__()
if port is not initializing.  This is necessary in a test that checks
that a Port becoming alternate/backup/disabled stops learning and
forwarding.  The move_rstp__() call is necessary to immediatly disable
learning and forwarding on that port.  Without this, the test fails.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp: Refactor rstp_check_and_reset_fdb_flush().
Jarno Rajahalme [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp: Refactor rstp_check_and_reset_fdb_flush().

With this commit, RSTP is able to flush from the MAC learning table
entries pertaining to a single port.  Before this commit the whole
table was flushed every time a port requested flushing actions.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agostp,rstp: disable learning and forwarding in STP/RSTP disabled state.
Daniele Venturino [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
stp,rstp: disable learning and forwarding in STP/RSTP disabled state.

There is a difference between a port with STP/RSTP protocol enabled and a
disabled role and a port which has a disabled role because STP/RSTP is
not active.  This commit ensure to make such distinction.

Standard 802.1D claims that the Topology Change state machine (17.31)
treats a Port as no longer active when it becomes an Alternate, Backup,
or Disabled Port and stops learning from received frames.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp: Invoke updt_roles_tree__() in max_age and forward_delay setters.
Jarno Rajahalme [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp: Invoke updt_roles_tree__() in max_age and forward_delay setters.

This patch was necessary in a test on the acceptable range values of
max_age and forward_delay.  Since rstp_set_bridge_max_age__() and
rstp_set_bridge_forward_delay__() set the rstp->bridge_max_age and
rstp->bridge_forward_delay variables, it was necessary to call
updt_roles_tree__() to immediately update p->designated_times, used in
tx_rstp() (in lib/rstp-state-machines.c).  After this change the
validation software received the expected maxAge and forwardDelay
values.  Otherwise, such test failed.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp-state-machines: fix TCN reception.
Daniele Venturino [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp-state-machines: fix TCN reception.

Set rcdv_tcn and return OTHER_INFO when a
TOPOLOGY_CHANGE_NOTIFICATION_BPDU is received, as required by
802.1Q-2008.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp-state-machines: Rename ALTERNATE_AGREED state as ALTERNATE_AGREED_EXEC.
Daniele Venturino [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp-state-machines: Rename ALTERNATE_AGREED state as ALTERNATE_AGREED_EXEC.

All other similar port_role_transition_sm states are named according
to this pattern.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp-state-machines: Fix updt_roles_tree().
Daniele Venturino [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp-state-machines: Fix updt_roles_tree().

If the DesignatedBridgeID Bridge Address component is equal to that
component of the Bridge's own bridge priority vector skip to the next
port.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp-state-machines: Fix Topology Change State Machine transition.
Daniele Venturino [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp-state-machines: Fix Topology Change State Machine transition.

Accordingly to the 802.1D-2004 standard, the transition from the
TOPOLOGY_CHANGE_SM_NOTIFIED_TCN_EXEC state should go to the
TOPOLOGY_CHANGE_SM_NOTIFIED_TC_EXEC state and not to the
TOPOLOGY_CHANGE_SM_ACTIVE.

See section 17.31 of 802.1D-2004 standard.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp-state-machines: Fix compare_rstp_priority_vector().
Daniele Venturino [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp-state-machines: Fix compare_rstp_priority_vector().

The bridge_port_id is never conveyed in Configuration Messages, but is
used as a tie-breaker within a Bridge.  This patch extends the
comparison to this fifth field in a rstp_priority_vector.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp-state-machines: Fix reception of BPDUs conveying an unknown port role.
Jarno Rajahalme [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp-state-machines: Fix reception of BPDUs conveying an unknown port role.

If the Unknown value of the Port Role parameter is received, the state
machines should treat the RST BPDU as if it were a Configuration BPDU.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp: Add 'rstp-admin-p2p-mac' and 'rstp-admin-port-state' setters.
Jarno Rajahalme [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp: Add 'rstp-admin-p2p-mac' and 'rstp-admin-port-state' setters.

'rstp-admin-port-state' is the Administrative Bridge Port state
variable defined in the 802.1D-2004 standard.  It can be set to
include or exclude a port from the active topology by management
(section 7.4).

operPointToPointMAC and 'rstp-admin-p2p-mac' are a pair of parameters
that permit inspection of, and control over, the administrative and
operational state of the point-to-point status of the MAC entity by
the MAC Relay Entity.  adminPointToPointMAC can be set by management
and its value is reflected on operPointToPointMAC.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agorstp: Show some useful rstp port fields.
Jarno Rajahalme [Fri, 14 Nov 2014 22:07:50 +0000 (14:07 -0800)]
rstp: Show some useful rstp port fields.

designated_bridge_id, designated_port_id and designated_path_cost are
now displayed in rstp_status when using 'ovs-vsctl list port'.

Signed-off-by: Daniele Venturino <daniele.venturino@m3s.it>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
9 years agotests: Skip tunnel-push-pop test on Windows and BSD.
Pravin B Shelar [Thu, 13 Nov 2014 22:52:51 +0000 (14:52 -0800)]
tests: Skip tunnel-push-pop test on Windows and BSD.

Also updates ovs-router README documentation.

Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>