ofproto-dpif: Validate ct action support.

Disallow installing rules that execute ct() if conntrack is unsupported
in the datapath.

Reported-by: Ravindra Kenchappa <ravindra.kenchappa@hpe.com>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Jarno Rajahalme <jarno@ovn.org>

ofp-actions: Refactor ofpact_get_mf_dst().

This function finds the mf destination field for any ofpact, returning
NULL if not applicable. It will be used by the next patch to properly
reject OpenFlow flows with conntrack actions when conntrack is
unsupported by the datapath.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>

ipfix: Skip BFD packets.

The patch is to skip BFD packets in ipfix.
Bidirectional Forwarding Detection (BFD) packets are for monitoring
the tunnel link status and consumed by ovs itself, no need to sample
them.
Refer to IETF RFC 5881, BFD control packets are the UDP packets with
destination port 3784 and BFD echo packets are the UDP packets with dst
destination port 3785.
Ipfix will skip both BFD control packets and BFD echo packets.

Signed-off-by: Wenyu Zhang <wenyuz@vmware.com>
[blp@ovn.org added check for IP]
Signed-off-by: Ben Pfaff <blp@ovn.org>

debian: Skip systemctl redirect.

After some experimentation on Ubuntu15.04, I see the
following behavior.

1. If you install openvswitch-switch with 'apt-get install',
then you automatically get a upstart and systemd config files
for openvswitch. The integration with 'interfaces' fails
because both the upstart and systemd jobs do not have logic
to handle it.

The above behavior will likely get fixed soon in upstream
Ubuntu.

2. If you install openvswitch-switch via the packages
created from the openvswitch repo, there is no systemd or
upstart conf files installed. But systemd notices this
and creates a runtime openvswitch conf file which does
nothing but call back the sysv startup script.

In the above case when you call
"/etc/init.d/openvswitch-switch start", it inturn calls
"/bin/systemctl start openvswitch-switch.service" and
that inturn again calls "/etc/init.d/openvswitch-switch start".
But the above for some reason simply hangs. It looks like a call
to ifup when invoked in this manner does not return.
I am not sure why this is happening.

We can avoid the above behavior completely by skipping the
systemctl redirect as done in this commit. This should fix
both 1. and 2. above.

Signed-off-by: Gurucharan Shetty <guru@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

NEWS: Add OVN.

Add OVN to NEWS as a post-2.4.0 feature.

Signed-off-by: Russell Bryant <russell@ovn.org>
Acked-by: Justin Pettit <jpettit@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

ovn.at: Fix races

These tests are racy as nothing prevents packet re-ordering.
Fix them by sorting outputs before comparing.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ovn.at: Use {} to make this less ambiguous

While (surprisingly to me) bash interprets $10 as ${1}0,
many other shells, including NetBSD's /bin/sh, interpret it as ${10}.

Also use already assigned named variables rather than positional
parameters to make things a little more readable.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ovn: Dedicated connection handler for packet-ins

This patch opens and maintains a new connection that is dedicated
to monitor the packet-ins for br-int.

Signed-off-by: Babu Shanmugam <bschanmu@redhat.com>
[blp@nicira.com removed support for OpenFlow 1.0]
Signed-off-by: Ben Pfaff <blp@ovn.org>

ofproto-dpif: Shortcut common case in rule_check().

Typically the datapath will support all available features, so check
that first before attempting to retrieve various values out of a
minimask as the latter doesn't need to be checked if all fields are
supported.

ct_state is an exception, because support for the bits in this field is
not binary; only some bits are defined so far, so they must still be
checked against the current known supported bits.

Suggested-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Jarno Rajahalme <jarno@ovn.org>

ofproto-dpif: Validate ct_* field masks.

When inserting rules that match on connection tracking fields, datapath
support must be checked before allowing or denying the rule insertion.
Previously we only disallowed flows that had non-zero values for the
ct_* field, but allowed non-zero masks. This meant that, eg:

ct_state=-trk,...

Would be allowed, while

ct_state=+trk,...

Would be disallowed, due to lack of datapath support.

Fix this by performing the check on masks instead of the flows.

Reported-by: Ravindra Kenchappa <ravindra.kenchappa@hpe.com>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Jarno Rajahalme <jarno@ovn.org>

ofproto-dpif: Reject partial ct_labels if unsupported.

If only half of a ct_label is present in a miniflow/minimask (eg, only
matching on one specific bit), then rule_check() would allow the flow
even if ct_label was unsupported, because it required both 64-bit fields
that comprise the ct_label to be present in the miniflow before
performing the check.

Fix this by populating the stack copy of the label directly from the
miniflow fields if available (or zero each 64-bit word if unavailable).

Suggested-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Jarno Rajahalme <jarno@ovn.org>

dp-packet: use dp_packet_reset_offsets during initialisation

Use dp_packet_reset_offsets() in dp_packet_init__() to make it
clearer that initialisation includes a reset of packets
packets the same way as callers to dp_packet_reset_offsets() reset packets.

Signed-off-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

odp-util: Limit scope of vlan in format_odp_action().

Limit the scope of the local vlan variable in format_odp_action()
to where it is used. This is consistent with the treatment of mpls
in the same function.

Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

types: Correct typo in comment above eth_addr.

Correct spelling of struct in comment above eth_addr.

Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

ovsdb-client: Add optional table and columns to dump command

Adding a table to the dump command's arguments retrieves only that table.
One or more columns after the table retrieve only those columns.

Default behavior is unchanged.

Signed-off-by: Bert Vermeulen <bert@biot.com>
[blp@ovn.org updated documentation and usage]
Signed-off-by: Ben Pfaff <blp@ovn.org>

csum: Inline csum_add16(), csum_add32().

These two functions are really simple, to the point that inlining them
yields less code than calling them: with GCC 4.9 on i386, this change
reduces the code generated for packet_csum_pseudoheader() from 112 bytes
to 100 bytes.

I didn't measure whether there is a performance improvement.

Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>

docs: Document that --detach disables logging to the console.

Reported-by: "P.R.Dinesh" <pr.dinesh@gmail.com>
Reported-at: http://openvswitch.org/pipermail/discuss/2015-December/019567.html
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Gurucharan Shetty <guru@ovn.org>

ovs-ofctl: Support multiple tables in replace-flows and diff-flows.

Currently ovs-ofctl replace-flows and diff-flows commands only support
flows in table 0.  Extend this to cover all possible tables.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

AUTHORS: Update email address.

Update AUTHORS to use the email address I'm now using as the primary
address for ovs work.

Signed-off-by: Russell Bryant <russell@ovn.org>
Acked-by: Justin Pettit <jpettit@ovn.org>

ofproto: Implement OF1.4 error code for set-async-config

This patch adds support for Openflow1.4 error codes for set-async-config.
In this patch, a new error type, OFPET_ASYNC_CONFIG_FAILED is introduced
that enables the switch to properly inform the controller when controller
tries to set invalid mask or unsupported configuration.

Signed-off-by: Ambika Arora <ambika.arora@tcs.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

ovsdb-idl: Improve ovsdb_idl_add_table() comment.

The new comment reflects with more clarity what ovsdb_idl_add_table() does.
Previous comment could be misunderstood, leading to believe that this function
replicates all columns on IDL. Hopefully this fix clarifies that columns are
not replicated, just minimal data for reference integrity is replicated.
A comment in ovsdb_idl_table_class is also modified to better reflect this
behaviour.

Signed-off-by: Edward Aymerich <edward.aymerich@hpe.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

ofproto-dpif-xlate: use IPv6 for tnl_neigh_cache

Use IPv4-mapped address and replace tnl_arp_lookup with tnl_neigh_lookup.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

tunneling: extend flow_tnl with ipv6 addresses

Note that because there's been no prerequisite on the outer protocol,
we cannot add it now. Instead, treat the ipv4 and ipv6 dst fields in the way
that either both are null, or at most one of them is non-null.

[cascardo: abstract testing either dst with flow_tnl_dst_is_set]
cascardo: using IPv4-mapped address is an exercise for the future, since this
would require special handling of MFF_TUN_SRC and MFF_TUN_DST and OpenFlow
messages.

Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Co-authored-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

tnl-arp-cache: Rename module and functions to tnl-neigh-cache.

Since we don't distinguish between IPv4 and IPv6 lookups, consolidate ARP
and ND cache into neighbor cache. Other references to ARP related to the
ARP cache but that are not really about ARP have been renamed as well.
tnl_arp_lookup is kept for lookups using IPv4 instead of IPv4-mapped
addresses, but that is going to be removed in a later patch.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

lib: add format_in6_addr and scan_in6_addr

Add in6_addr counterparts to the existing format and scan functions.
Otherwise we'd need to recast all the time.

Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

ofproto-dpif-xlate: use IPv6 for orig_tunnel_ip_dst

orig_tunnel_ip_dst is replaced by orig_tunnel_ipv6_dst and represents IPv4
addresses as IPv4-mapped IPv6 addresses. Since this is only used in
ofproto-dpif-xlate, this change is well contained.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

tunneling: add IPv6 support to netdev_tunnel_config

Allow configuration of IPv6 tunnel endpoints.

Signed-off-by: Jiri Benc <jbenc@redhat.com>
Co-authored-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

smap: smap_add_ipv6 adds IPv4-mapped addresses as IPv4

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

packets: Add ipv6_string_mapped.

ipv6_string_mapped stores an IPv6 or IPv4 representation of an IPv6 address
into a string. If the address is IPv4-mapped, it's represented in IPv4
dotted-decimal format.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

tnl-ports: remove unused tnl-arp-cache.h and include other necessary headers

tnl-ports.c did not use anything from tnl-arp-cache.h besides some other headers
that the latter included, like system string.h and OVS netdev.h.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

AUTHORS: Add Hiteshi Kalra, Saloni Jain, Shashwat Srivastava, Sandeep Kumar.

Signed-off-by: Ben Pfaff <blp@ovn.org>

Vagrantfile: Update to Fedora 22 base box

Fedora 20 base box is no longer available at chef/fedora20.

Signed-off-by: Mark D. Gray <mark.d.gray@intel.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

Implement Vacancy Events for OFPMP_TABLE_DESC.

This patch adds support for vacancy events in table-desc.

ovs-ofctl -O OpenFlow14 dump-tables-desc <switch>
-This command is enhanced to display the Vacancy Event configuration
 of the tables on a <switch>, which is set using the mod-table command.

Signed-off-by: Saloni Jain <saloni.jain@tcs.com>
Co-authored-by: Hiteshi Kalra <hiteshi.kalra@tcs.com>
Signed-off-by: Hiteshi Kalra <hiteshi.kalra@tcs.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

Implement Openflow 1.4 Vacancy Events for OFPT_TABLE_MOD.

OpenFlow 1.4 introduces the ability to turn on vacancy events with an
OFPT_TABLE_MOD message specifying OFPTC_VACANCY_EVENTS. This commit adds
support for the new feature in ovs-ofctl mod-table.
As per the openflow specification-1.4, vacancy event adds a mechanism
enabling the controller to get an early warning based on capacity
threshold chosen by the controller.

With this commit, vacancy events can be configured as:
ovs-ofctl -O OpenFlow14 mod-table <bridge> <table> vacancy:<low,high>
<low,high> specify vacancy threshold values in percentage for vacancy_down
and vacancy_up respectively.

To disable vacancy events, following command should be given:
ovs-ofctl -O OpenFlow14 mod-table <bridge> <table> novacancy

Signed-off-by: Saloni Jain <saloni.jain@tcs.com>
Co-authored-by: Shashwat Srivastava <shashwat.srivastava@tcs.com>
Signed-off-by: Shashwat Srivastava <shashwat.srivastava@tcs.com>
Co-authored-by: Sandeep Kumar <sandeep.kumar16@tcs.com>
Signed-off-by: Sandeep Kumar <sandeep.kumar16@tcs.com>
[blp@ovn.org fixed a few typos]
Signed-off-by: Ben Pfaff <blp@ovn.org>

ovn-northd: Refactor main loop to use ovsdb_idl_loop_* functions

This patch also addresses the issue reported at
http://openvswitch.org/pipermail/discuss/2015-November/019445.html

Suggested-by: Russell Bryant <rbryant@redhat.com>
Signed-off-by: Numan Siddique <nusiddiq@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

AUTHORS: Add Han Zhou.

Signed-off-by: Ben Pfaff <blp@ovn.org>

ovn: use name instead of uuid as key of logical router ports

Before this patch, logical switch ports were using name as key but
logical router ports using uuid, which confuses user when trouble-
shooting, and names of patch ports unreadable. This patch unifies
it by using "name" field also for logical router ports.

Signed-off-by: Han Zhou <zhouhan@gmail.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

vtep/vtep.xml: Improve XML source formatting.

This patch contains only cosmetic changes.

Signed-off-by: Ariel Tubaltsev <tubaltzev@gmail.com>
Acked-by: Bruce Davie <bdavie@vmware.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

netdev-linux: Handle flags for 10G and 40G speeds

Handle advertised and supported flags for the following speeds:

* 1G base KX
* 10G base KX4, KR, R
* 40G base KR4, CR4, SR4, LR4

Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

netdev-linux: correctly detect port speed bits beyond 16bit

This includes bits for:
* Backplane
* 1000 baseKX (full duplex)
* All speeds of 10Gbit and above other than 10000 baseT (full duplex).

Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

ofp-util: Add "check_overlap" and "reset_counts" to stateful flags.

The OpenFlow specification implies that every flag is part of the flow
state, even though that isn't really meaningful for OFPFF_CHECK_OVERLAP
or OFPFF_RESET_COUNTS.  This commit adds them to the flow state (reported
in flow stats replies).

Found by OFTest.

Signed-off-by: Ben Pfaff <blp@ovn.org>
Co-authored-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>

INSTALL.SSL.md: Fix arguments to "ovs-pki req" command.

Reported-by: Aaron Rosen <aaronorosen@gmail.com>
Reported-at: http://openvswitch.org/pipermail/dev/2015-November/061778.html
Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>

ofproto: Fix inserting buckets at the end of an empty group.

This caused a segfault.

Reported-by: Ray Li <rayli1107@gmail.com>
Reported-at: http://openvswitch.org/pipermail/discuss/2015-September/018746.html
Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>

ofp-util: Reject bad group type and command with error instead of abort.

Reported-by: Manpreet Singh <er.manpreet25@gmail.com>
Reported-at: http://openvswitch.org/pipermail/discuss/2015-October/019048.html
Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>

tests: Add tests for out_port and out_group.

Open vSwitch has supported these features on flow_mod for a long time, but
it has never included a test.  This commit adds tests for both.

Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>

ovs-ofctl: Add "out_group" keyword for OF1.1+ matching on output group.

It seems that OVS has never supported this OpenFlow feature in ovs-ofctl,
which makes it hard to test.  This commit adds support.

(Open vSwitch has supported this in OpenFlow for a long time; it's only
ovs-ofctl that didn't.)

Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>

ofp-actions: Look inside write_actions for output ports and groups.

The out_port and out_group matches only looked at apply_actions
instructions, but my interpretation of the OpenFlow spec is that they
should also look inside write_actions.

This affected the output of (and in one case the correctness of) some
tests, so this updates them.

Reported-by: Gavin Remaley <gavin_remaley@selinc.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>

tests: Make parse_listening_port entirely reliable.

In test runs, I've occasionally seen mysterious failures in which
parse_listening_port failed to find the listening port even though an
examination of the log file showed that it was there.  I spent some time
trying to figure out what was going wrong.  It seemed like everything was
lined up properly to ensure that a command like "ovs-vsctl set-controller
br0 tcp:127.0.0.1:0" would only return to the command prompt after the
new listener was ready and the proper log message was written.  It was
obviously a very small race because I could only reproduce it with a large
test parallelism (e.g. -j10 on my quad-core laptop).

The problem turned out to be asynchronous logging in ovs-vswitchd.  If I
turned that off, by commenting out the call to vlog_enable_async() in
bridge.c, parse_listening_port became reliable.

This commit works around the problem by making parse_listening_port retry
for a while if necessary.  It also transforms the shell function into an
m4 macro (so that it can use OVS_WAIT_UNTIL) and renames it to
all-uppercase to follow the convention for macros.

Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>

tests: Make OVS_WAIT_WHILE, OVS_WAIT_UNTIL run even faster in some cases.

Sometimes OVS_WAIT_WHILE and OVS_WAIT_UNTIL are used for conditions that
tend to be true immediately but sometimes take a moment or two.  This
change makes such cases 0.1 seconds faster (which could add up over
hundreds of tests).

Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>

tests: Run OVS_WAIT_WHILE, OVS_WAIT_UNTIL in main shell environment.

AT_CHECK runs its commands in a subshell.  That means that (among other
effects), any variable assignments within its commands will disappear after
the commands' completion.  That doesn't matter for any of the existing
users, which don't do the sorts of things that affect an outer shell
environment anyhow, but an upcoming user wants to make a shell assignment
that persists.  This commit makes that possible, by using AT_CHECK
(actually AT_FAIL_IF but it's moot) only upon failure instead of bracketing
the entire test.

Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>

ofproto-dpif-xlate: Fix byte-order error in comparison.

Found by sparse.

CC: Jarno Rajahalme <jarno@ovn.org>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Flavio Leitner <fbl@sysclose.org>

classifier-private.h: Whitespace in a comment

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ovs-ofctl.at: Fix a comment typo

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ofproto.at: Workaround a race

While a barrier serializes requests from the same connection,
it doesn't wait for requests from other connections to the switch.
Replace the barrier with infamous "sleep 1" to workaround the problem.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ofproto.at: Portability improvement

== is a bash dialect.  Use = instead.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ofproto-dpif: Remove a debug log

The message seems stale and it happens on port-mod.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ofproto-dpif.at: Fix a race

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

vtep-ctl.at: Fix a regex

} is an ordinary character in "basic" regex.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

dpif_dummy_override: Allow overriding a non-existing provider

This allows --enable-dummy=system with a userland-only build.
It's useful for testsuite.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ovn-controller-vtep.at: Fix a race

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

use "/usr/bin/env python" for shebangs for python scripts

On NetBSD, a typical path of python interpreter is /usr/pkg/bin/python.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ofproto-macros.at: Rename vconn_windows_sub to vconn_sub

This is a preparation to add non-windows stuff.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

test-atomic: Bump test duration for multi-thread tests

This makes tests pass on on my single-cpu NetBSD-7 VM.

How frequently preemption happens depends on the cpu scheduler.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

completion.at: Improve portability

NetBSD's /bin/sh complains on the syntax of bash array.

While the use of eval might seem overkill, it's tricky to avoid
the error because the generated code will be a part of the surrounding
subshell and the syntax check is done a bit earlier than the execution
of these conditionals.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

daemon_switch_user: Improve portablility

NetBSD doesn't have [gs]etres[ug]id.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Andy Zhou <azhou@nicira.com>

openvswitch.m4: Portability improvement

${parameter/pattern/string} substituion is not portable.
Replace them with $SED equivalent.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

netdev-bsd: Update after eth_addr changes

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

tnl-arp-cache: Fix NetBSD build

On NetBSD, netinet/icmp6.h has a few pre-requisite headers.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ofp-actions: Define IPPORT_FTP by ourselves

IPPORT_FTP is not always provided by system headers.  (eg. NetBSD, OS X)

This hides the enum on Linux but I don't think it causes a problem.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

extract-odp-netlink-h: Portablitiy improvement

\t is GNU sed extension.  Use [[:space:]] instead.

Signed-off-by: YAMAMOTO Takashi <yamamoto@midokura.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ovs-router: Report ovs/route/add errors as errors.

The _error version should be used to report errors.

Also, add missing return in one error case.

Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>

netdev-linux: Remove unreachable code in netdev_linux_rx_recv_tap().

While splitting netdev_linux_rx_recv() into netdev_linux_rx_recv_sock()
and netdev_linux_rx_recv_tap() in commit
b73c85181df9 ("netdev-linux: Read packet auxdata to obtain vlan_tid")
error handling part was copied 'as is' to both functions.
But in case of netdev_linux_rx_recv_tap(), according to POSIX, the
number of bytes read shall never be greater than 'size'.

Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

datapath-windows: fix MIN() macro

A quick implementation of MIN() didn't take into account operator
precedence as shown in the following example:

 #include <stdio.h>
 #define MYMIN(_a, _b) (_a) > (_b) ? (_b) : (_a)
 int main() {
     if (MYMIN(512, 256) < 14) {
         printf("buggy MYMIN\n");
     }
     return 0;
 }

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Russell Bryant <russell@ovn.org>
Signed-off-by: Ben Pfaff <blp@ovn.org>

AUTHORS: Add Panu Matilainen.

Signed-off-by: Ben Pfaff <blp@ovn.org>

rpms: %exclude has side-effects, remove the files from buildroot instead

Due to the way debuginfo package creation is done in existing rpm
versions, its better to physically remove any non-packaged files
from the buildroot, using %exclude for the purposes causes artifacts
like broken symlinks  in the -debuginfo packages.

Reported-at: https://bugzilla.redhat.com/show_bug.cgi?id=1281913
Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>
Signed-off-by: Ben Pfaff <blp@ovn.org>

conntrack: Add support for NAT.

Extend OVS conntrack interface to cover NAT.  New nested NAT action
may be included with a CT action.  A bare NAT action only mangles
existing connections.  If a NAT action with src or dst range attribute
is included, new (non-committed) connections are mangled according to
the NAT attributes.

This work extends on a branch by Thomas Graf at
https://github.com/tgraf/ovs/tree/nat.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

system-tests: Add IPv6 FTP system test.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Joe Stringer <joe@ovn.org>

system-tests: Workaround for pyftpdlib bug handling IPv6 addresses.

Hack around a bug in pyftpdlib that rejects EPRT connection due to
mismatching textual representation of the IPv6 address when the
address is not in the normalized format.  This happens when the
control connection is mangled by Linux NAT.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Joe Stringer <joe@ovn.org>

packets: Reorder CS_* flags to remove gap.

This changes the conntrack state flags used in the OpenFlow interface
to match the ones we currently use in the datapath.  While these do
not need to be synced, it is nice to get rid of the gap.

This should be merged before the first OVS release with connection
tracking, or not at all.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

ofproto: Enable in-place modification for recirc actions.

When modifying an existing datapath flow with recirculation actions,
the references to old (if any) recirculation actions need to be freed,
and references to new recirculation actions need to be stored.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@ovn.org>

ofproto: Allow xlate_actions() to fail.

Sometimes xlate_actions() fails due to too deep recursion, too many
MPLS labels, or missing recirculation context.  Make xlate_actions()
clear out the produced odp actions in these cases to make it easy for
the caller to install a drop flow (instead or installing a flow with
partially translated actions).  Also, return a specific error code, so
that the error can be properly propagated where meaningful.

There are may cases in which the NORMAL action decides to drop the
packet.  Most of these are not, however, traslation errors, but just
reactions to malformed input. In these cases it is correct to make the
NORMAL action do nothing, but allow other actions in the pipeline (if
any) to take effect.

Before this patch it was possible that the revalidation installed a
flow with a recirculation ID with an invalid recirc ID (== 0), due to
the introduction of in-place modification in commit 43b2f131a229
(ofproto: Allow in-place modifications of datapath flows).

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Joe Stringer <joe@ovn.org>

datapath-windows: Fix HvUpdateNic() to handle name changes

If the name of an internal or external NIC changes, we need to
disconnect the NIC from OVS since the name is the key. In this
change, we generate a link down event. It is as though we got a
call to HvDisconnectNic() for the old name and got a HvConnectNic()
for the new name.

Also, HvCreateNic() has been cleaned up to remove the code to look
for existing vport. We won't have a vport now since we'd have deleted
the vport in HvDeleteNic().

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sairam Venugopal <vsairam@vmware.com>
Acked-by: Alin Gabriel Serdean <aserdean@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <guru@ovn.org>

datapath-windows: cleanup AssignNicNameSpecial()

AssignNicNameSpecial() needed to be called outside of a lock and was
moved out in a previous change. But, it was accessing vport structure
outside of the lock which isn't safe. In this change, we take care of
that.

I tried to trigger a call to HvUpdateNic() by renaming the interface
from the GUI and didn't see any callback. Other changes are tested.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sairam Venugopal <vsairam@vmware.com>
Acked-by: Alin Gabriel Serdean <aserdean@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <guru@ovn.org>

datapath-windows: nuke port # argument in IP helper

Port # doesn't make much sense since it has not been computed yet.

Also, get rid of OVS_DEFAULT_PORT_NO and use OVS_DPPORT_NUMBER_INVALID
instead.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sairam Venugopal <vsairam@vmware.com>
Acked-by: Alin Gabriel Serdean <aserdean@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <guru@ovn.org>

datapath-windows: cleanup events code

Turns out that we don't need to generate an event is practically
useful only in case of a port disconnect to let userspace know.
Hence, this event is being posted from HvDisconnectNic().

In case of a new port appearing, it seems that userspace is not
interested in a new port unless it was added by userspace itself.
In my tests, userspce would end up deleting the port when it got
a new port notification, despite the port existing in OVSDB.

The reasoning seems simple enough:
- On Linux, OVS is integrated with the hypervisor (libvirt for eg)
and a port (ie. netdev) gets created in the Linux kernel and then
get added to OVSDB. When vswitchd picks up the port addition in OVSDB,
it adds the port in the OVS kernel DP.
- If the kernel netdev does not exist while OVS userspace tries to
create the port in OVS kernel DP, port addition fails. Moreover, the
only way to re-add the port is to trigger userspace to re-add the port
by deleting the port in OVSDB and re-adding it.

With this patch, I have verified that if a VIF gets disconnected on the
Hyper-V switch, it disappears from the OVS kernel DP as well.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sairam Venugopal <vsairam@vmware.com>
Acked-by: Alin Gabriel Serdean <aserdean@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <guru@ovn.org>

datapath-windows: cleanup InitHvVportCommon()

The workflow being implemented is that, we need to assign a special
name to the internal and external NICs, and it it not necessary to do
that from InitHvVportCommon(). The purpose of InitHvVportCommon() is to
insert the vport into the hash tables and update the switch context.

We special case the name assignment in HvCreateNIC() for
internal and external NICs. That seems more meaningful.

Also, reused HvCreatePort() to allocate a Vport for each of the external
NICs with NicIndex != 0. Due to this HvCreatePort() now takes 'nicIndex'
as an additional parameter.

Renamed InitHvVportCommon() to UpdateSwitchCtxWithVport().

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sairam Venugopal <vsairam@vmware.com>
Acked-by: Alin Gabriel Serdean <aserdean@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <guru@ovn.org>

datapath-windows: refactor port enumeration code

We already have functions HvCreatePort() and HvCreateNic() to
do the work. Might as well use that during port enumeration.
More refactoring in later patches.

Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Sairam Venugopal <vsairam@vmware.com>
Acked-by: Alin Gabriel Serdean <aserdean@cloudbasesolutions.com>
Signed-off-by: Gurucharan Shetty <guru@ovn.org>

AUTHORS: Update email address.

Signed-off-by: Gurucharan Shetty <guru@ovn.org>

AUTHORS: Update email address.

Signed-off-by: Jesse Gross <jesse@kernel.org>

ofproto: Correctly reject duplicate bucket ID for OFPGC_INSERT_BUCKET.

Otherwise duplicate bucket IDs cause linked list loops and other nastiness
because the ofputil_bucket_find() in the OFPG15_BUCKET_LAST case later in
copy_buckets_for_insert_bucket() will find the new bucket instead of the
old one and the list_splice() call becomes nonsensical.

Reported-by: Ray Li <rayli1107@gmail.com>
Reported-at: http://openvswitch.org/pipermail/discuss/2015-September/018731.html
Signed-off-by: Ben Pfaff <blp@ovn.org>
Reviewed-by: Simon Horman <simon.horman@netronome.com>

datapath-windows: Support for OVS_KEY_ATTR_SCTP attribute

This patch adds OVS_KEY_ATTR_SCTP to the OVS flow mechanism.

Signed-off-by: Sorin Vinturis <svinturis@cloudbasesolutions.com>
Acked-by: Sairam Venugopal <vsairam@vmware.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

stream-ssl: Replace client CA list instead of adding to it.

SSL_CTX_add_client_CA() appends to the client CA list without replacing any
already on the list, and furthermore wastes memory if the certificate in
the file is already on the list.  This commit thus fixes an effective
memory leak.

Signed-off-by: YongQiangLiu <liu.liuyongqiang@huawei.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

system-traffic: Add internal port conntrack tests.

Add an additional test that ensures that when receiving packets from
internal ports that reside in a foreign namespace, the conntrack
information is not populated in the flow.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>

system-traffic: Remove netcat from ICMP test.

Netcat is different on each platform I tried (Debian, Ubuntu, RHEL),
so rather than handling version differences it's better to just do the
same test with some hardcoded packets.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>

system-tests: Use '--bundle'

Use OpenFlow bundles for setting up flow tables.  This has the benefit
that when debugging test failures, no packet gets processed by
partially set-up flow table, which may seem confusing.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

system-tests: Make bridge creation more consistent.

Create all bridges with the same set of supported OpenFlow protocols
and fail-safe-mode secure, so that each test explicitly specifies flow
handling.

Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

nx-match: Add nxm_put_128m() function.

This allows us to remove nxm_put_ct_label() with something more general.
It will also have other callers later.

Signed-off-by: Justin Pettit <jpettit@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

byte-order: Make hton128() and ntoh128() behave like their counterparts.

Instead of taking the source and destination as arguments, make these
functions act like their short and long counterparts.

Signed-off-by: Justin Pettit <jpettit@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

ovs-router: Use standard libraries in scan_ipv6_route().

Signed-off-by: Justin Pettit <jpettit@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>

meta-flow: Use standard libraries in mf_from_ipv6_string().

Signed-off-by: Justin Pettit <jpettit@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>