Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Generate getters in the Builder class #432

Open
tetv opened this issue Aug 21, 2016 · 9 comments
Open

Generate getters in the Builder class #432

tetv opened this issue Aug 21, 2016 · 9 comments

Comments

@tetv
Copy link

tetv commented Aug 21, 2016

Sometimes there is a need to access to a Builder getter to do extra logic.

@Value.Immutable
public abstract class Foo {
    public abstract double price();
    public abstract Instant pricedModifiedOn();

    public static abstract class Builder {
        abstract double price(); // Builder getter access

        abstract Builder setPrice(double price);
        abstract Builder setPriceModifiedOn(Instant priceOn);
        public abstract Foo build();

        public Builder setThePrice(double price) {
            if(price() != price) { setPrice(price); setPriceModifiedOn(Instant.now()); }
        }
    }
}
@elucash
Copy link
Member

elucash commented Aug 21, 2016

There's pre-canned answer to this: If your logic requires complex multistep construction with getters or "is set" queries, please, resort to using Modifiable companion class, generated by @Value.Modifiable annotation. When using builder, if you set something, then you usually know what you've set, so why need to query?

@tetv
Copy link
Author

tetv commented Aug 21, 2016

If the Builder.from() is used, you probably don't know what is the original value in the Builder.

I would like to create an Immutable entity not a Modifiable entity, however during the build phase I may need to query some fields. I think it's a valid request. @AutoValue has this functionality already, and I think @Value.Immutable should have it.

The builder getter should have always package visibility in order the logic be internal to the original definition class (like provided in the above example).

@elucash
Copy link
Member

elucash commented Aug 22, 2016

You use Modifiable object as a builder, you can even use styles so it can be named *Builder and not Modifiable*. Then, there's toImmutable() method (also, rename-able using styles). Please, take a look at Guava library, and compare the API of ImmutableList.Builder with ArrayList. Basically modifiable class is your arraylist which you can manipulate until you create an immutable copy.

@tetv
Copy link
Author

tetv commented Aug 22, 2016

Thank you @elucash.

Can I use the @Value.Modifiable as a static class inside the @Value.Immutable class as we do for Builders? In my understanding to be able to use toImmutable we need to register both @Value.Immutable and @Value.Modifiable for an entity, but then the main Immutable entity will be different from that one.

By chance, do you have any example/documentation that uses Modifiable as a builder for Immutable entities? I would be able to control the setters of the Modifiable entity as shown in the above example.

Many thanks in advance.

@elucash
Copy link
Member

elucash commented Aug 22, 2016

Can I use the @Value.Modifiable as a static class inside the @Value.Immutable class as we do for Builders?

It's a standalone class, cannot be generated as nested to Immutable. This should be only cosmetic problem in practice.

By chance, do you have any example/documentation that uses Modifiable as a builder for Immutable entities?

http://immutables.github.io/immutable.html#modifiable

Here's an example of how I would create builder-in-disguice using modifiable companion class.

@Value.Style(
        create = "new",
        toImmutable = "build",
        typeModifiable = "*Builder")
@Value.Immutable
@Value.Modifiable
public interface MyValue {

    int getAttr();
    String getButtr();

    // just as an example, not required
    static MyValueBuilder builder() {
        return new MyValueBuilder();
    }

    // sample usage
    public static void main(String... args) {
        MyValueBuilder builder = MyValue.builder()
                .setAttr(1)
                .setButtr("2");

        if (builder.attrIsSet()) {
            int attr = builder.getAttr();
        }

        MyValue v = builder.build();
    }
}

@elucash
Copy link
Member

elucash commented Aug 22, 2016

Here's additional example which mimicks initial example

@Value.Immutable
public abstract class Foo {
    public abstract double price();
    public abstract Instant pricedModifiedOn();

    public static abstract class Builder extends FooBuilder  {
        public FooBuilder setThePrice(double price) {
            if (price() != price) { setPrice(price); setPriceModifiedOn(Instant.now()); }
            return this;
        }
    }
}

The problem with this approach is that setThePrice should be called first, as return type of inherited methods would be FooBuilder, so setThePrice would not be visible after first setter invokation returning FooBuilder.

@elucash
Copy link
Member

elucash commented Aug 22, 2016

Another solution might be using "normalization" functionality.

@Value.Immutable
public abstract class Foo {
    public abstract double price();
    @Value.Default
    public double newPrice() {
        return price();
    }
    public abstract Instant pricedModifiedOn();
    @Value.Check
    Foo normalize() {
          if (Double.doubleToLongBits(price()) != Double.doubleToLongBits(newPrice())) {
               return ImmutableFoo.builder()
                     .from(this)
                     .price(newPrice())
                     .pricedModifiedOn(Instant.now())
                     .builder();
          }
          return this;
    }
}

// ...
Foo oldFoo = ...

Foo newFoo =  ImmutableFoo.builder()
      .from(oldFoo)               
      .newPrice(computeNewPrice())
      .build();
// price/timestamp will be auto-updated if differs

@tetv
Copy link
Author

tetv commented Aug 23, 2016

Thank you @elucash for you priceless support on this.
Much appreciated.

The @Value.Modifiable solution is really not very clean.
The normalisation solution it seems a bit better, but has also issues:

  1. Two public methods for updating the price (setPrice and setNewPrice). If users use setPrice the priceModifiedOn is not updated, violating an important requirement.
    Initially I was thinking that I could hide (package visibility) the setPrice method and force users to use always the other, but then I realise that when I am loading from the database I need to use setPrice because I need to maintain the original priceModifiedOn field.

  2. Json representation of Foo entity can't have another unnecessary property "newPrice".

I still think that the cleanest (and also simplest and easy to use) solution to solve this particular problem is having package level getters in the ImmutableFoo.Builder. It would be great if we could activate then using the @Value.Style settings.

@vorburger
Copy link
Contributor

I would also find having getters on generated Builders useful in a particular use case (unrelated to this one; re. "if you set something, then you usually know what you've set, so why need to query?" in my case, I'm initializing imagine like nested structures, and want to get a value I've previously set using a literal, without having to copy/paste it... that can be worked around by assigning the literal to a var and then using it in both setters, of course (that's what I'm currently doing).

Thank you for the tip with Modifiable - I've tried that, and understood that this then creates 3 classes, a @modifiable *Builder (because of typeModifiable = "*Builder"), and an *Immutable, with a nested public static final class Builder ... that's a bit too confusing, and too expensive a price to pay, for just wanting getters on the Builder. Also the @modifiable *Builder won't extend an inner Builder class from the hand written @Value.Immutable @Value.Modifiable - I understand why (because it extends that type), just saying this to clarify why the use of @modifiable is not a straight forward solution to get a (single) Builder with getters..

odl-github pushed a commit to opendaylight/netvirt that referenced this issue Apr 27, 2017
This is dependent on the genius change
https://git.opendaylight.org/gerrit/#/c/53763/, which must be merged
simultaneously.

The change in the Xtend re. flowId is because the FlowEntityBuilder does
not have a getFlowId() like FlowEntity, only a setter; see
immutables/immutables#432

Change-Id: Iecbe1ccd2c1c6cb9c914b70c6e44590d748ad739
Signed-off-by: Michael Vorburger <vorburger@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>
odl-github pushed a commit to opendaylight/releng-autorelease that referenced this issue Apr 27, 2017
Project: netvirt stable/carbon af508b111095b4b7a947d48593ed4c252815acb0

@immutable FlowEntity with FlowEntityBuilder

This is dependent on the genius change
https://git.opendaylight.org/gerrit/#/c/53763/, which must be merged
simultaneously.

The change in the Xtend re. flowId is because the FlowEntityBuilder does
not have a getFlowId() like FlowEntity, only a setter; see
immutables/immutables#432

Change-Id: Iecbe1ccd2c1c6cb9c914b70c6e44590d748ad739
Signed-off-by: Michael Vorburger <vorburger@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>
odl-github pushed a commit to opendaylight/netvirt that referenced this issue Apr 27, 2017
This is dependent on the genius change
https://git.opendaylight.org/gerrit/#/c/53763/, which must be merged
simultaneously.

The change in the Xtend re. flowId is because the FlowEntityBuilder does
not have a getFlowId() like FlowEntity, only a setter; see
immutables/immutables#432

Change-Id: Iecbe1ccd2c1c6cb9c914b70c6e44590d748ad739
Signed-off-by: Michael Vorburger <vorburger@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>
odl-github pushed a commit to opendaylight/releng-autorelease that referenced this issue Apr 27, 2017
Project: netvirt master 483dd2febc9824cda99030346902eef40f0a5566

@immutable FlowEntity with FlowEntityBuilder

This is dependent on the genius change
https://git.opendaylight.org/gerrit/#/c/53763/, which must be merged
simultaneously.

The change in the Xtend re. flowId is because the FlowEntityBuilder does
not have a getFlowId() like FlowEntity, only a setter; see
immutables/immutables#432

Change-Id: Iecbe1ccd2c1c6cb9c914b70c6e44590d748ad739
Signed-off-by: Michael Vorburger <vorburger@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>
odl-github pushed a commit to opendaylight/docs that referenced this issue Apr 27, 2017
Project: netvirt master 483dd2febc9824cda99030346902eef40f0a5566

@immutable FlowEntity with FlowEntityBuilder

This is dependent on the genius change
https://git.opendaylight.org/gerrit/#/c/53763/, which must be merged
simultaneously.

The change in the Xtend re. flowId is because the FlowEntityBuilder does
not have a getFlowId() like FlowEntity, only a setter; see
immutables/immutables#432

Change-Id: Iecbe1ccd2c1c6cb9c914b70c6e44590d748ad739
Signed-off-by: Michael Vorburger <vorburger@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>
odl-github pushed a commit to opendaylight/docs that referenced this issue Jun 8, 2017
Project: netvirt stable/carbon 990c2dbdc69783ebf73f81779ea3689e614addd8

Simplify some streaming constructs

Change-Id: If48c31a56e1667597dc540ac5068fe731058a6e1
Signed-off-by: Stephen Kitt <skitt@redhat.com>


Bug 8549 - Inter OVS traffic over EVPN does not work with
openstack-vni-semantics-enforce set to true

Change-Id: I401107b046695ec9af9feebbc47ecce0461b0b07
Signed-off-by: eupakir <kiran.n.upadhyaya@ericsson.com>


Bug7830:Conflict modification exception for NAT

Problem Statement:
===================
Conflicting modification for path exceptions was throwing from NAT module
for the flow creation of table21->26 and table47->21

Solution:
============
Table21->26 flow creation was happening in 2 more classes. To avoid
race-condition have introduced DJC to install and remove the flow.

Change-Id: Ib6c47667c9245d50eb51a7baa3d7969aa05b325b
Signed-off-by: karthikeyan <karthikeyan.k@altencalsoftlabs.com>


Bug 8614: NPE observed in cloud-sc's VrfListener

Change-Id: Ief4b9b90c96e1e426b9399389538cd14b39fc942
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>


Bug 8595 - DNAT traffic from DC gateway to FIP fails

Description :

In PDNAT_TABLE, we have a FIP mac match for performing FIP to internal IP
translation. But in L3_LFIB_TABLE, the destination mac has not been
updated and as a result, the reverse traffic for FIP were dropped.

Changes done to update packet with destination FIP mac as a result,
PDNAT_TABLE has specifc match and FIP to internal IP translation been done
properly.

Change-Id: I3fec733c7a8c049808a09828b8090daed12c83fc
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>


Bug 8585: Exception with invalid QoS Alert params

Root cause - When invalid config parameter is set through karaf shell,
in-memory configuration is updated before data validation. Invalid configuration
later gets rejected but this makes configuration inconsistent. Any subsequent, configuration
is rejected until problematic parameter is set to valid value.

Fix - Update the QoS Alert configuration parameters only after validation.
Same as done for configuration through REST.

Change-Id: Ia46664850f3c7abe0ce74f76cf7e35bdbd79c7c6
Signed-off-by: Arun Sharma <arun.e.sharma@ericsson.com>


Fix checkstyle problems not detected by the current version

This change is required for overall move to new Checkstyle version, see
https://git.opendaylight.org/gerrit/#/q/topic:bumpCheckstyle

Change-Id: I8f9c20211a2936cf964f2cf51f8e41ffa180bb07
Signed-off-by: David Suarez <david.suarez.fuentes@ericsson.com>
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 7451 - guarding NPE

added few null checks to avoid nullpointerexceptions

Change-Id: I8b077f1d87d79bfe3804d8daabdf391a5358f301
Signed-off-by: Periyasamy Palanisamy <periyasamy.palanisamy@ericsson.com>


Bug 8581: DNAT failure with openstack/ocata

Problem Statement:
===================
DNAT traffic is getting failed because of missing DNAT flows in switch
(DPN). This issue was observed in openstack stable/ocata only.

Solution:
==========
Due to invalid check in NeutronPortListener.update(), certain FIP DS entry
where not populated and as a result, required DNAT flows where not installed.
Changes done to create required FIB DS entries properly.

Change-Id: Ia11a7e1d28288d762d26c3897d1612401d4ee8fa
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>


elanName is null

Check for elanName. If null, return.

Change-Id: I72c931477c6d8ae464a654095383f8537ef6748e
Signed-off-by: Janki Chhatbar <jchhatba@redhat.com>


Bug 8586: Alarm raised due to stale stats polling

Root cause - When port stats polling is enabled, first poll returns
old data (Rx bytes and Dropped bytes) when QoS alert was not enabled.
Alert may get logged if stale stats has dropped % > threshold value

Fix - Alert should not be logged based on first poll data. Use first poll data
to set the baseline counts and raise alert on subsequent polls based on the delta
of counts between the polls.

Change-Id: Ic12f3d41e7c0277109acc51ab548e1a7d0b4d13b
Signed-off-by: Arun Sharma <arun.e.sharma@ericsson.com>


Replace LOGGER by LOG

Replace LOGGER by LOG to follow the 
OpenDaylight recommendations [1].

[1]
https://wiki.opendaylight.org/view/BestPractices/Logging_Best_Practices

Change-Id: I038b8fc9a16b7443fcdefcad1bc12283151ba503
Signed-off-by: David Suarez <david.suarez.fuentes@ericsson.com>


Clean up Optional uses

This patch reverts Optional and related streaming uses where they
result in more convoluted code than necessary (or, in the case of
Optional, where Optional instances are created only for conditions).

Change-Id: I8d0b2c456ff0a477e440fb029b8495476aabed27
Signed-off-by: Stephen Kitt <skitt@redhat.com>


Replace logger and log by LOG

Replace logger and log by LOG to follow the
OpenDaylight recommendations [1].

[1]
https://wiki.opendaylight.org/view/BestPractices/Logging_Best_Practices

Change-Id: Ia4f5a1c1692133efa0e49873ba954a360572989b
Signed-off-by: David Suarez <david.suarez.fuentes@ericsson.com>
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 8589: Fix regression in Ipv6PktHandler

Following this change [1], Ipv6PktHandler is not getting
invoked for RS/NS meant for router interfaces. This patch
fixes this issue.

[1] https://git.opendaylight.org/gerrit/#/c/54389/

Change-Id: I5b98b742623df2552205f8bbb24d70ae17ba1325
Signed-off-by: Sridhar Gaddam <sgaddam@redhat.com>


Use named constants in ElanUtils

... instead of commented true or false.

Change-Id: I7f0738f11666f9433ae0942b2be538514d08311b
Signed-off-by: Stephen Kitt <skitt@redhat.com>


Use direct comparisons instead of Objects.isNull

There is no reason to ever use Objects.isNull() directly, it’s
equivalent to “== null”. Its documentation explains its purpose: “This
method exists to be used as a Predicate, filter(Objects::isNull)”.

Change-Id: I86123fec65c09a1d26906935cad6df79750fa2d8
Signed-off-by: Stephen Kitt <skitt@redhat.com>


Bug 7451 - VPN service cleanup of Table 17

Whenever a VM is deleted (or) a VPN where a VM
resides in is removed, the dispatcher-table
entries for that VMs VPN Interface were not
getting removed.

The fix here makes sure that dispatcher Table
(Table 17) gets cleaned up by VPN Service.

The related ELAN patch to clean up dispatcher
table for ELAN service entries is here;
https://git.opendaylight.org/gerrit/#/c/57767

Change-Id: I6460b8a12a37236275e72606bc58c4b59db440cb
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>


SNAT performance improvement for Controller-Based SNAT

Description :
Following changes were done to address SNAT performance improvement
in following area.

1) Seperate Queue for first SNAT packet, retry packet and Flow removed
event on Idle Timeout.
2) Multithreading: using 25 threads for processing first SNAT packets, 25
thread for processing retry packets and 1 thread for processing Flow
Removed event.
3) Updated checkIpPortMap API to have direct query based on routerid,protocol,
internalip+port criteria thus preventing unnecessary looping.
4) Moved the code of getting externalMac Address and external address to
get executed only for first packet, thus preventing the retry packet from
executing it which is unnecessary.

Change-Id: Ib34aa8726295381d0e4ffd1f3cec74292d422c4f
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>


Bug 8417 - [l3vpn_ecmp] Deleting MPLS GRE Tunnel port doesn't delete the

bucket entry in the group

Change-Id: Ic7287a9611bf27453713e47f6dde11373fea3172
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>


Minor Checkstyle fix Custom Import Order error

This shows up as red in Eclipse, but the CLI mvn build missed this.
This is due to a more recent version of Checkstyle.

required for overall move to new Checkstyle version, see
https://git.opendaylight.org/gerrit/#/q/topic:bumpCheckstyle

also includes minor removal of (1) extra parenthesis, (2) generic <>
clean-up and (3) missing @Override from automatic clean-up which happens
on auto-save (technically unrelated to Checkstyle).

Conflicts:
	vpnservice/qosservice/impl/src/main/java/org/opendaylight/netvirt/qosservice/QosAlertManager.java

Change-Id: Ic650ffe9ac42cfbc7cad4c1c31c7553faf5fe603
Signed-off-by: Michael Vorburger <vorburger@redhat.com>
Signed-off-by: David Suarez <david.suarez.fuentes@ericsson.com>


Bug 7451 - Leftovers in dispatcher table when unbind and ietf-state delete

           occurs simultaneously

Change-Id: Id3c44e94da203ddda9dbc3e6fb8b6885523c7fa3
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>
Signed-off-by: Periyasamy Palanisamy <periyasamy.palanisamy@ericsson.com>
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 8498 - ICMP traffic from DHCP NS not blocked when ICMP rule to allow
VM-VM traffic is added in learn mode. 

The  wrong filter table to be used for learn ingress/egress service is
fixed.

Change-Id: I15e8841314c24b10556128bac354ebab1d80a698
Signed-off-by: Aswin Suryanarayanan <asuryana@redhat.com>


Add JUnits for InterVpnLinkLocator

 + Only 2 so far:
    -  testFindInterVpnLinksSameGroup
    -  testSelectSuitableDpns_moreDpnsThanIvpnLinks

 + It adds an L3vpnCatalog that contains several VPN test data
   to play with

 + Idem with InterVpnLinkTestCatalog

Change-Id: I1c6866ec5fe2570a469dd9ea9e5a6591d11dffa4
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>


Fix a few non-null collections

In general, methods returning collections should never return null.
This patch fixes L2GatewayConnectionUtils and propagates the non-null
results.

Change-Id: If249aec9b8494de736c362f0ea04139151670d68
Signed-off-by: Stephen Kitt <skitt@redhat.com>


Bug 8376: Fix DHCP for external tunnels

DHCP for external tunnels does not work with neutron
DHCP server.

ODL is installing rules to punt DHCP packets received
on external tunnels to controller, even when ODL
DHCP is disabled. But, since ODL DHCP is disabled,
no DHCP response is sent by ODL. And, with these
rules installed, neutron dhcp server never sees the
DHCP packet. As a result, VM will not get IP
address.

Do not start listeners that are responsible for
installing DHCP rules to punt DHCP packets to
ODL controller when ODL DHCP is disabled.

Change-Id: Ib0e243a9bca5866e50a5d1cb4c70cd4441dd01e6
Signed-off-by: Victor Pickard <vpickard@redhat.com>


BUG:8341   IAE seen in CSIT logs when port/vlan name is not matching UUID

Change-Id: If90482540cc5cc7c28d16397f84a39f3774eab7b
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>


bug 7599  hwvtep ucast mac add performance improv

Changes for performance improvement.
Added new Cmd TunnelIpCommand and also changed function getKey to return
actual key Identifier in place of String value.

Change-Id: I0fd84c87db6feefcba4c119fe0c05193751ab8cb
Signed-off-by: Akash <a.k.sahu@ericsson.com>


elanmanager-impl <dependency> interfacemanager-impl <scope>test

It is wrong to directly depend on interfacemanager-impl from
elanmanager-impl.  This may only be done with <scope>test, for component
tests.

Remove the sole usage of a util from ITM - genius has the same util.

Change-Id: I0dda82184d547cefa5a447361111438ed3fe40c1
Signed-off-by: Michael Vorburger <vorburger@redhat.com>


Use named constants for PolicyAceFlowWrapper

Instead of commenting each use of true or false, use constants.

Change-Id: Ibf8b19b67dc54e99e6caffe8056a0e04b62601c2
Signed-off-by: Stephen Kitt <skitt@redhat.com>


Adjust tunnel state listeners for logical tunnel

If the logical tunnel exists between the source and destination DPNs it will
be set as the lport-tag of register6 in the remote nexthop actions.

Spec: https://git.opendaylight.org/gerrit/50779/

Change-Id: I98ff89efd047e056529e4bbdf06493e548c561d1
Signed-off-by: Olga Schukin <olga.schukin@hpe.com>
Signed-off-by: Sam Hague <shague@redhat.com>


BUG:8232 updating BGPVPN with the List of RDs

       *Problem: When a new Extra-route is added and if there are not enough RDs
        for the VPN then the new extra-routes are not handled.
        These unprocessed extra-routes will be handled during vpninstance update
        (ie when new RDs are added(Update) to the existing VPN)

Change-Id: Ic380b6aaabdd0d44484c2ec1a33de7cde9c01366
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>


Fix Unused log and private final

Delete unused log and their imports.
Change initailized private final var to
private static final in order to use less memory.

Change-Id: Ieecaf3cf75ac67b55e2bfd87c1c78eb1bbd77782
Signed-off-by: mcauffiez <mcauffiez@inocybe.com>


EVPN RT2 : Silent host changes

When the flow created on packet received from silent host expires,
prefix is removed from macEntry container and prefix is withdrawn,
There is no necessity to withdraw prefix from here, it will be
duplication.

Change-Id: Iab970736a182efaddf4dfc44a5a11377ca22b111
Signed-off-by: Yugandhar Reddy Kaku <yugandhar.reddy.kaku@ericsson.com>


Bug 8508 : Id-manager exception during releasing id for router

Description : Even though opendaylight-semantics is disabled,
the following EVPN NAT use-cases require VNI to be allocated from
OpenDaylight VNI Pool:

a. All traffic on external network will be using a VNI allocated for the
external-VPN aka Internet VPN.

b. Similarly all traffic from Non-NAPT switch to NAPTSwitch will be using
the VNI allocated for the router from the OpenDaylight VNI Pool.
For VNI-L3-L3 Fowarding feature also, a VNI must be allocated for every
router from the OpenDaylight VNI Pool as that VNI will be used for
Non-NAPT to NAPTSwitch traffic.

This fix makes that allocation as only release alone was done earlier.

Changes done to allocate a vni for router(when router-gatway-set done)
and been released during ext-router entry is deleted. In CSIT(having
non-EVPN use-case) were executed, an attempt to release the non-allocated
id was done during when ext-router entry is deleted resulting in exception.
Change done to make sure an id is allocated when router-gateway-set is done to
prevent such occurance of exception.

Change-Id: Ic1e696d6b4c8d0c56eac8b16ed0731d072b5b135
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>


Bug 8539: IPv6 L3 Forwarding broken.

The merge of the following review;
https://git.opendaylight.org/gerrit/57145
broke L3 Forwarding for IPv6 addressed ports.

The fix here addresses the breakage by making
sure that L3_GW_MAC_TABLE entries are filled
correctly for both IPv4 and IPv6 router-interface
ports, before embarking on writing PING
Responder flows for IPv4 based router-interfaces.

Change-Id: Ieb190edda470264006609c62e4af9fc0d94d4a2c
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>


Bug 8485 : EVPN was also coming as Layer_3 VRF

Problem:
Even after configuring EVPN Layer 2 VRFs in quagga its coming as EVPN Layer 3 VRFs.

Solution:
Changes to modify the layer type was missing in VPN engine

Change-Id: I8f45b06c02f3920512f16563d77b13b8ba076a9d
Signed-off-by: Vyshakh Krishnan CH <vyshakh.krishnan.c.h@ericsson.com>


Bug 8358 -  Local next hop group not deleted after VM migration

Issue:

When a VM is migrated from one DPN to another, the FIB flows are not
programmed for the VM in the DPN to which it is migrated.

Analysis and Fix:

The l3nexthop container is used to check whether the local next hop group
is already present for a prefix. If it exits, no new group is created and just
the reference count is increased. When this prefix is removed, this
reference is decreased & the group is removed when the reference count is
0. It was observed that for the first VM creation, the reference count is
2(it should be 1) because of redundant calls to createlocalnexthop. A
check has been added to prevent this.

Also some other minor fixes have been done related to ECMP

1. The addition/deletion of extra route also changes the above mentioned
flow reference count. This caused problems while just updating the
extraroutes(the flow reference count keeps increasing and so the local
next hop group would never be deleted). Now, the addition/deletion of
extra routes doesnt change the flow reference count in the l3nexthop DS.

2. When there is no change in routepath of the extraroute(updated extra
route also present in the same DPN), there is no change in odl-fib DS and
so handling has been added to implicitly call the refresvrfEntry which
would update the FIB flows.

Change-Id: I4d7b435aed73268383b36e1c635c499c71e54c60
Signed-off-by: gobinath <gobinath@ericsson.com>


Bug 8412 - NPE while adding and removing elanmacentry

1. While advertising RT2 route prefix is Optional, hence added a safer
null check. if prefix is not present, still go ahead and advertise RT2
route
2. During withdraw if prefix is not present, return with error log.
3. Remove the elanName null check as its never null.

Change-Id: I3009bb9d19f3413d448031ff47771a986ef10746
Signed-off-by: Janki <jchhatba@redhat.com>


BUG 8537: Get destination from interface remote ip

Genius will not set the tunnel destination IP address until the very
last action. Classifier egress cannot expect to have this IP address
available on it's own pipeline.

Solve this by setting the IP address on REG0 on the egress service
binding actions. This will be the remote IP of the bound interface in
case of tunnel or the local IP otherwise.

To avoid overuse of registries, a REG is no longer used to set the
SFF IP address on table 101. This IP will be set directly on the path
specific flows of table 223 if needed. Previously this REG was being
used to know if the SFF was remote by comparing with the local IP in
these flows but that is not needed as it can be known at configuration
time.

Change-Id: If7a474391ed2d643cc3b3f34743d60c5f418b2e5
Signed-off-by: Jaime Caamaño Ruiz <jaime.caamano.ruiz@ericsson.com>


Replace toUpperCase().equals by equalsIgnoreCase()

Change-Id: I54ac6903362c15c6caab05810861de852d2619a5
Signed-off-by: David Suarez <david.suarez.fuentes@ericsson.com>


Bump versions by x.y.(z+1)

Change-Id: I54e2105583481ca281691c735d81d027350d4711
Signed-off-by: jenkins-releng <jenkins-releng@opendaylight.org>


Bug 8475 - Non-external vlan provider networks do not work

Patch - 2 for handling of review comments on master

Spec link:
http://docs.opendaylight.org/en/latest/submodules/netvirt/docs/specs/vlan-provider-enhancement.html

Change-Id: I4fc70ba20927257046e45ffe5d577106fcc8d09c
Signed-off-by: Ravindra Thakur <ravindra.nath.thakur@ericsson.com>


Bug 7939,8082: Table 19 flow nightmares

Table 19 flows rules aren't getting recycled appropriately
whenever VMs are moved into a routed subnet and then again
backed off.

The fix here attempts to make sure that Table 19 flow rules
are handled consistently when a subnet moves in/out of a router
and also when a router moves in/out of a bgpvpn.

Change-Id: Ief94b1b856a69c631052f11644db6a0b091f2aa7
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>
Signed-off-by: Sam Hague <shague@redhat.com>
Signed-off-by: Jamo Luhrsen <jluhrsen@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 7939 : SubnetRoute functionality for bgpvpn broken.

IPs behind Neutron Interfaces are not discoverable for
router-associated-to-bgpvpn use-cases.

With this fix we are disallowing creation of subnetRoutes for internal
subnets off neutron routers.
More specifically, for just neutron routers, subnetRoutes won't be published.
However, if such subnets are part of a router that is within a bgpvpn, the
subnetroutes feature will continue to work.

For the PNF review that requires subnetRoutes to be available,
the fix here needs to be further tweaked to make PNF discovery on internal
subnets of a VLAN-network to work.

PNF Review referredin commit message here is:
https://git.opendaylight.org/gerrit/53984

Change-Id: I507aae2e3d1cbb72457f32b48e598ef0f6cda703
Signed-off-by: HANAMANTAGOUD V Kandagal <hanamantagoud.v.kandagal@ericsson.com>
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>
Signed-off-by: HANAMANTAGOUD V Kandagal <hanamantagoud.v.kandagal@ericsson.com>


BUG-8388: On egress, filter out SF bound packets

Classifier should not handle packets that are already classified and
bound towards the SFF. For this, we set C1 to 0xFFFFFF on the ACL flow.
Then on beginning of egress pipeline, if C1 does not have that value
we assumed that's a packet bound for the SF and already classified so
it is resubmitteed to dispatcher. Otherwise, C1 is reset to 0 and
handling continues.

Change-Id: I1427a9bbec7fb0c8e36c241b727d39de450bc1ed
Signed-off-by: Jaime Caamaño Ruiz <jaime.caamano.ruiz@ericsson.com>


Bug 8451: afi parameter is not the good one used in pushRoute and withdrawRoute

The afi parameter used as parameter is calculated based on the prefix nature,
that is to say either IPv4 or IPv6.

Change-Id: I6d34488e43d4dc589ce842fcd1675d24033c9f28
Signed-off-by: Noel de Prandieres <prandieres@6wind.com>


Bug 8471 - Traffic between a VM without FIP and a VM with FIP is not
working

1)Hairpinning changes are done for conntrack based SNAT, now it matches
on external subnet id that external n/w id.
2) The packets are submitted back to table 21 from 47.
3)L3_GW_MAC_TABLE (Table 19)  replace the external n/w id with external
subnet id for the packet address to router external interface
4)L3_FIB_TABLE (Table21) flow now matched external subnet id  instead of
n/w is along with external ip address.

Change-Id: Ib4de831a58638ed9d4aed9c78450e3d7bebf6222
Signed-off-by: Aswin Suryanarayanan <asuryana@redhat.com>


Bug 8475 - Non-external vlan provider networks do not work

Spec link:
http://docs.opendaylight.org/en/latest/submodules/netvirt/docs/specs/vlan-provider-enhancement.html

Change-Id: Iad7479835a7bd97a4f827c8365601d2a20efe477
Signed-off-by: Ravindra Thakur <ravindra.nath.thakur@ericsson.com>
Signed-off-by: Sam Hague <shague@redhat.com>


Bug8484: Non-NAPT Group action is drop for router associated with BGP-VPN

Problem Description:
======================
Non-NAPT to NAPT VM Traffic is getting dropped at Non-NAPT Group level
when router is being associated with private BGP-VPN.

Solution:
===========
When ever Vxlan tunnel got "down" followed by "up" event between Non-NAPT to
Napt Switch happened, Non-Napt group got re-created through tunnel
interface state listener. In this use case is not been properly having the
code logic for if the router with private bgp-vpn configured.

Added the proper code for creating tunnel bucket information for Non-NAPT
groups got re-created during tunnel state changed.

Change-Id: I509eeda9041d5e1b170e45554b3d9f661e6091fd
Signed-off-by: karthikeyan <karthikeyan.k@altencalsoftlabs.com>


Bug 8473: Reboot fixes for ECMP with extra-routes

* It includes changes in the extraroute-rds-map container. Nexthop is
  added as key which is used to retrieve the allocated rd for an extra
  route. This is useful during reboot to know which rd is allocated for
  an extra route and this change avoids re-allocation of rds during reboot.

* Some minor fixes for reboot is also added in this review.

Change-Id: I31e8d153be947810a66ed3e2ccc686eee876bea8
Signed-off-by: eceghkl <manu.b@ericsson.com>
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>


Bug 8474: Enable openstack-vni-enforcement on data plane

Change-Id: I5f87df4b45e35a3b857f616437b6ede4c0b5f029
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>


Bug 8474: Local FIB entries for VxLAN VPN interfaces not programmed

Local FIB entries for VxLAN networks on router-based VPNs are not getting
programmed when VNI flag is turned to true, as makeConnectedRoute() wasn't
being called.
This commit fixes the same for create and delete cycles.

Change-Id: Ib81ac9aa1270921c11fc35cfb3352afdd5aff577
Signed-off-by: Abhinav Gupta <abhinav.gupta@ericsson.com>
Signed-off-by: Sam Hague <shague@redhat.com>


Bug8471 : L3_GW_MAC_TABLE entry for FIP not removed.

Description : L3_GW_MAC_TABLE entries for Floating-ips where not removed
after FIP disassociation. The flow-key generated during removal was
different then during creation of this flow.

Change-Id: I6dc9d7dbb365d9e1a815241b01139658ebb59d12
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>


Bug 8451 : Problem with EVPN RT5 exchange between ODL and QBGP

Problem:
ODL was always advertising encap type as MPLS and protocol type as ANY
irrespective of L3VPN or EVPN RT5

Fix:
Added the missing code to handle the encap type and protocol type

Change-Id: Ia4e27b62c33977ae5f5ce155695851f97bc47788
Signed-off-by: Vyshakh Krishnan CH <vyshakh.krishnan.c.h@ericsson.com>


Bug 8025: Use RPC with barrier when sending SNAT packet out

When configuring SNAT flows and then sending a packet-out to those
flows, use RPCs with barriers before the packet-out to ensure the
flows are configured once the packet-out is sent.

Change-Id: I330df570857317e50bf5dc404e9691a0053f3f54
Signed-off-by: Alon Kochba <alonko@hpe.com>
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>


Bug 8294 : fib-show cmd throws an exception on usage

When using fib-show, an exception was triggered when parsing FibEntries
from a VRF context.

Correcting the existing fix. Instead of break , changing to continue.

Change-Id: I4637ac14b7fd5bdb9ccb7978d1a2b40538098dfd
Signed-off-by: HANAMANTAGOUD V Kandagal <hanamantagoud.v.kandagal@ericsson.com>


BUG-3874: BGP Manager / fib-show shell cmd triggers an exception on usage

When using fib-show, an exception was triggered when parsing FibEntries
from a VRF context. A sanity check is done in order to prevent accessing
null entries.

Change-Id: If296e3b4770518f53c789130565db8960ca63ef6
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>


EVPN RT2 Installing flows for RT2 received in prior

1. When EVPN is created, vpn instance will be created and bgpmanager will
register with quagga bgp. hence BGPmanager will start receiving RT2 routes
from peer. Bgpmanager will write into macvrfentry(odl-fib.yang).
macvrfentrylistener is listening to macvrfentry and which installs dmac
flow for the received RT2 route only if EVPN is attached to network.

2. Now if EVPN is created and NOT attached to network. BgpManager will
still receive RT2 routes from peer, which will be written into
macvrfentry, but since evpn is not attached to network yet, hence DMAC
flows are not installed. Once Evpn is attached to network we need to go
through all the macvrfentry with the matching RD and install dmac flows.

3. when EVPN is detached from network, we go through all the macvrfentry
for that RD and uninstall DMAC flows.

4. async read is done so that even if it fails to read first time it tries
multiple times till either read is success or failure.

Change-Id: If4cb1b0fb924e7defa4e8c119dcf249b3f9b24f6
Signed-off-by: Riyazahmed D Talikoti <riyazahmed.d.talikoti@ericsson.com>


EVPN RT2 : Receive RT2 BGPManager changes

On receiving RT2, bgpmanager has to updated the macVrfEntry in fib.yang.
Also when the RT2 is from a new TEP, same has to be updated in elan
external tep DS.
A sample cli is added to simulate the RT2 from QBGP as well.

Change-Id: If193af109a3a08d08d6c6f32e715c71f045e5af8
Signed-off-by: Vyshakh Krishnan CH <vyshakh.krishnan.c.h@ericsson.com>


EVPN RT2 : Listener for evpn to network associatn

When EVPN is created, even if no network is associated to it, BGP routes
will come from quagga which will be stored in FIB as MacVRF.
Now when the network association happens, all these routes should be
examined and corresponding elan DS has to be updated.

Change-Id: I53b75890005c67469cab512a3bf5824b6265f06d
Signed-off-by: Vyshakh Krishnan CH <vyshakh.krishnan.c.h@ericsson.com>


bgpmanager VPNv6 shell command update as well as OAM changes

bgp-network command is modified so as to support the ability to
configure prefixes with afi parameter, to be 1 ( IPv4) or 2 ( IPv6).
Moreover, this extension permits retrieving the number of prefixes VPNv6 stored
on quagga, like it has been done for VPNv4

Change-Id: I90a043d5d2a77789b0a5fe376994b201830ad5fa
Signed-off-by: Noel De Prandieres <prandieres@6wind.com>
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>


bug 7599  hwvtep ucast mac add performance improv

Adding batching to improve performance

Change-Id: I2844f7418a8f088064baafe4804a9c1a6b3a8f73
Signed-off-by: Akash Kumar Sahu <a.k.sahu@ericsson.com>


Bug 8014 l2gw connection broken

listener registration is broken in previous async datachange listener
refactoring change.

Adding the remote ucast macs immediately instead of waiting for mcast
to appear in operational datastore.

creating logical switch in config datastore , irrespective of if it is
found in operational datastore or not.

Change-Id: I7916ec4d457e189d28bf48eb3c11836dd6c311f6
Signed-off-by: Riyazahmed D Talikoti <riyazahmed.d.talikoti@ericsson.com>


EVPN RT2 DMAC flow install

1. When EVPN RT2 route is recieved from peer, BGPManager will write into
macvrfentery to which MAcVrfEntryListener is listening to.
and based on add or delete DMC flow entry is added/deleted.

2. ElanEvpnUtils contains utility methods which access elan related info

3. EvpnFlowUtils contains utility methods which helps
in installing flows in DMAC.

Change-Id: I6c6f1b407d4d209985be964011323019b6965a20
Signed-off-by: Riyazahmed D Talikoti <riyazahmed.d.talikoti@ericsson.com>
(cherry picked from commit cf7e6b3e276e6ea62886945b15c9c20e97b7d954)


Thrift changes to support IPv6 calls over Quagga BGP stack

Quagga BGP stack enhancement to support IPv6 leads to Thrift interface
modifications, which specifically include the following Type changes:
a) new AFI parameter - IPv6
b) pushRoute() includes new AFI parameter
c) withdrawRoute() includes new AFI parameter
d) getRoutes() includes new AFI parameter
e) onUpdatePushRoute() includes new AFI parameter
f) onUpdateWithdrawRoute() includes new AFI parameter
Thrift interface changes to include information on afi type.

Change-Id: Id4aea3809e58910c57e588fa6018119911b57834
Signed-off-by: Noel De Prandieres <prandieres@6wind.com>
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>


BUG-8387: Set tunnel id to 0 on egress towards SFF

When classifier is forwarding packets to the first SFF through a
tunnel, set the tunnel key to 0 as that what the SFF expects.

Change-Id: I02830f46cfe1507463aaabc562ee522646f778d3
Signed-off-by: Jaime Caamaño Ruiz <jaime.caamano.ruiz@ericsson.com>
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 8370 - Fib flows are not deleted after extra route deletion

Issue:

When the extra routes associated with a router are updated, the FIB flows
corresponding to the old extra routes are not deleted.

Analysis:

While there exists handling to update the loadblancing group corresponding
to the extra route, the handling to remove the FIB flows for the extra
route when the extra route is completely removed is missing.

Fix:

Changes have been added to the deletelocalfibentry to remove the FIB flows
of the removed extra route.

Change-Id: I77b5b9d46173b1c2613c28a495bc6a85bff52d9a
Signed-off-by: gobinath <gobinath@ericsson.com>


Fix ivpnlink bgp route leaking

 + After splitting the InterVpnLink refactor commit into 3 smaller
   ones, this line was missed.

Change-Id: Ic31fde68e18a0cb86948301c7e549fab42561168
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>


Bug 8340 - Rules with IP prefix not getting programmed when there
are multiple SG associated with a port

1)Remote filter table programming during port update is fixed.
2)Ingress and Egress filter  table no is fixed in the  respective
classes.
3)AclInterface Map logic is updated to prevent duplicate entries.
4)Test added to cover interface with intially one acl and later two.

Change-Id: If5af5ced20ed81ec27201bd4ec0def91715ebda9
Signed-off-by: Aswin Suryanarayanan <asuryana@redhat.com>


Refactor InterVpnLink. Part THREE

 + This one implements the final purpose of this refactoring:
   taking the route leaking out of the fibmanager to a higher level,
   in this case it's the BgpManager who now takes care of the route
   leaking in case of BGP routes and the VpnManager for static and
   connected routes.

Change-Id: Ie73a7061a26cfa3b95fad633533d44fff87ad6fb
Signed-off-by: Miguel Perez <francisco.miguel.perez@ericsson.com>
Signed-off-by: Sam Hague <shague@redhat.com>


BUG-8331: BGP Manager labels values are passedin correct order to QBGP

L3 Label Parameter is swapped with L2 Label when calling QBGP

Change-Id: Ib5d3260a696078e6b3d1de9cd0b84eaf0ebfef7d
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>


BUG 8375 - Malformed NSH packets get dropped

- The NSH NextProtocol (NP) field must be set to 3 for ethernet
- Since the field isnt set, its set to 0 by default, which causes
  the packets to get dropped by the SFF OVS.

Change-Id: I4d13b86ff563b76a992c5c88791015053f5688e8
Signed-off-by: Brady Johnson <brady.allen.johnson@ericsson.com>


UT - Adding Netvirt classifier GeniusProvider tests

- added GeniusProviderTest.java
- added TestInterfaceManager.java to stub the Genius InterfaceMgr
- added TestOdlInterfaceRpcService.java to stub the Genius RPC calls
- added GeniusProviderTestParams.java with common test params for above 3 files
- changes to GeniusProvider for errors that the UT flushed out

Change-Id: I60e85811959319bf6e163641506aa2f26d9975d5
Signed-off-by: Brady Johnson <brady.allen.johnson@ericsson.com>


BUG:8345 & Bug:8347

   * The code-review contains below Fixes
   BUG:8347 -> NPE at org.opendaylight.netvirt.elan.internal.ElanSmacFlowEventListener.onFlowRemoved
   BUG:8345 -> OptimisticLockFailedException for elanDpnInterface
   Operational Datastore while deleting elan-instance and elan-interfaces

Change-Id: Id930e511f1f0814e17ca1f8cbbda43e9fd3d7e6f
Signed-off-by: epgoraj <p.govinda.rajulu@ericsson.com>


Revert "Thrift changes to support IPv6 calls over Quagga BGP stack"

This reverts commit cae4adebcf7f6f79736859cd2dbb04599b435091.

Change-Id: I09f01ec493bf4dc1023f5fe22f7d8aa9768d0e53
Signed-off-by: harikrishna <hari.i.krishna@ericsson.com>


Revert "Thrift interface changes to support BGP VPNv6"

This reverts commit 21f5d03ca7cc0e0dfd53c3868196b6e93ec9ae19.

Change-Id: I1df917f4385ace0691b6fbf3cec408fd496e8180
Signed-off-by: harikrishna <hari.i.krishna@ericsson.com>


Revert "bgpmanager VPNv6 shell command update"

This reverts commit be855643f1ceadc9077882b81eb35533e5717148.

Change-Id: I9516ee1258f2fd3395a677a7b70ba0e5ad83e884
Signed-off-by: harikrishna <hari.i.krishna@ericsson.com>


Revert "bgpmanager: IPv6 information is collected vty shell"

This reverts commit b918dc13b79a565fb6d16194690c7d6c7eacb1e7.

Change-Id: I4a8ffda31d769ada55481ad708cd619c7c3a3a3b
Signed-off-by: harikrishna <hari.i.krishna@ericsson.com>


Rollback - ADDED INFO logs for tempest issue DEBUG

Change-Id: Ic695dd7bd011313e3f615c1c2f70bd7b7d212771
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>


Bug 7939: Interface management on cluster reboot

Interface state changes are missed on a cluster
reboot resulting in some flows and DS not being
properly re-written.

Change-Id: Ie91cc1226f4d6494e4ce0e82f4dd0d6e00840396
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>


Bug 8349 - Creation of BGPVPN with the same RD is not restricted

ODL does not support handling of multiple bgpvpns with the same RD. Hence,
a bgpvpn create request with an already used RD should be denied.

Change-Id: I60d4cc00e73ca8089a15f5b6f4298d74ab7c7b0c
Signed-off-by: eupakir <kiran.n.upadhyaya@ericsson.com>


bgpmanager: IPv6 information is collected using vty shell

This extension permits retrieving the number of prefixes VPNv6 stored on
quagga, like it has been done for VPNv4.

Change-Id: I413c5b2ed1dc6d004f09a582f8510bccf1f1986a
Signed-off-by: Noel De Prandieres <prandieres@6wind.com>
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Signed-off-by: Hari Krishna <hari.i.krishna@ericsson.com>


BUG 8340: Handle case when aceList is also empty

Change-Id: If76c63ec6c5dd00811f593ac8d6058341b8e1cf1
Signed-off-by: Jamo Luhrsen <jluhrsen@redhat.com>


Netvirt: Qos Alert patch7 - last

*** patch #7 - Added EntityOwnershipListener  ***

Only qosalert entity owner or master only should poll for port stats.
Once a master is elected, it shall continue polling until a new master is elected.

Spec Review Link:
https://git.opendaylight.org/gerrit/50689

Change-Id: I67ad2b3a016204b8e19e2971a0329b52aa8bd4fb
Signed-off-by: Arun Sharma <arun.e.sharma@ericsson.com>


Netvirt: Qos Alert patch6

*** patch #6 - Added log4j configuraton support ***

1. Added a new log4j appender with default logging properties in qosalert.cfg
2. Added listener method update in QosAlertGenerator class to handle any chage in qosalert.cfg

Spec Review Link:
https://git.opendaylight.org/gerrit/50689

Change-Id: I2008a0a2dd22661e0f45251f0aac1088270d3d96
Signed-off-by: Arun Sharma <arun.e.sharma@ericsson.com>


EVPN RT2 advertise and withdraw prefix changes

1) Added elan external tep listener.
When elan external tep is added update the elan remote broadcast
group for all elan dpns.

2) when an elan interface is added , advertise its prefix
when elan interface is deleted, delete its prefix.

3) Advertise prefix when we receive packet in message from silent
host.

4) Withdraw prefix when the smac flow entry of silent host expires

Change-Id: I870174f9fa0741a6508f2edec20dae7489b67d26
Signed-off-by: Yugandhar Reddy Kaku <yugandhar.reddy.kaku@ericsson.com>


Set feature capability for port status update

See: https://bugs.opendaylight.org/show_bug.cgi?id=7718
DEPENDS ON: https://git.opendaylight.org/gerrit/#/c/55854/

Change-Id: Ia83d31c348176947e31e11d383f770ac97390145
Signed-off-by: Josh <jhershbe@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>


bgpmanager VPNv6 shell command update

bgp-network command is modified so as to support the ability to
configure prefixes with afi parameter, to be 1 ( IPv4) or 2 ( IPv6).

Change-Id: Ie0a5faf7160a049e930d2eb8f89af9d87693ddcd
Signed-off-by: Noel De Prandieres <prandieres@6wind.com>
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Signed-off-by: Hari Krishna <hari.i.krishna@ericsson.com>


BUG 8296 - Erroneous Egress Classifier flows

- Egress classifier filter flow now just matches on NSH
  MD Type as set in the Ingress classifier ACL flow
- OpenflowRenderer.renderPath() is now passed the nodeIP
  instead of the destination SFF IP
- Patch Set 2 : fixed minor merge issue
- Patch Set 3 : fixes found by ClassifierEntryTest UT

Change-Id: Ide70b7222b08b2b363acc36126dc08e8a9add556
Signed-off-by: Brady Johnson <brady.allen.johnson@ericsson.com>


Factor out executor from ClassifierService & UT

Depends on [1]

[1] https://git.opendaylight.org/gerrit/#/c/55251/

Change-Id: I4cf5bffcd33e2f4190a699a9982f75ca2d3ad8fd
Signed-off-by: Jaime Caamaño Ruiz <jaime.caamano.ruiz@ericsson.com>


Thrift interface changes to support BGP VPNv6

Thrift interface changes to include information on afi type.

Change-Id: Ie82e9a086621a551f19a0faaf9107d630277ae43
Signed-off-by: Noel De Prandieres <prandieres@6wind.com>
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Signed-off-by: Hari Krishna <hari.i.krishna@ericsson.com>
Signed-off-by: Vivekanandan Narasimhan <n.vivekanandan@ericsson.com>


Thrift changes to support IPv6 calls over Quagga BGP stack

Quagga BGP stack enhancement to support IPv6 leads to Thrift interface
modifications, which specifically include the following Type changes:
a) new AFI parameter - IPv6
b) pushRoute() includes new AFI parameter
c) withdrawRoute() includes new AFI parameter
d) getRoutes() includes new AFI parameter
e) onUpdatePushRoute() includes new AFI parameter
f) onUpdateWithdrawRoute() includes new AFI parameter

Change-Id: Ibd4842624d55fd07ae88b0d2ca8094a641f8749d
Signed-off-by: Noel De Prandieres <prandieres@6wind.com>
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Signed-off-by: Hari Krishna <hari.i.krishna@ericsson.com>


Bug 7866 adding retries for remote dmac programming during tunnel up event

* currently there is no retries in case of OptimisticLockFailedException
while programming remote dmac during tunnel up event. This leads to remote
dmac is programmed with drop action which leads to packet drop. Now added
retries for jc job.
* Also adding retries in case remote BC group fails (during tunnel up event) due to datastore exceptions.

Change-Id: I2bc024fc7133170c321c4a5071d6b58e00a2666b
Signed-off-by: Periyasamy Palanisamy <periyasamy.palanisamy@ericsson.com>


EVPN RT2 NeutonVpn changes

1. EVPN RT2 RPC for create, delete, and get EVPN
creation and deletion of EVPN via REST is handled by NeutronEvpnManager.
When create EVPN RPC is received, NeutronEvpnManager will check if
vpninstance already exist with same name if not then it will create
vpninstance of type L2 for EVPN. similarly vpninstance will be deleted for
delete event.

2. Changes for attach/detach network to EVPN
NeutronEvpnManager will add/remove/update evpn name in Elan augmentation so that
EvpnElanInstanceManager will advertised and withdraw routes accordingly.

3. Few previous comments mandatory false from yang removal fix

4. Added ElanHelper file to move commonly used utility methods to avoid
cyclic dependency

Change-Id: I259734fad1ca0e19f38093b4e22a326fb6969459
Signed-off-by: Riyazahmed D Talikoti <riyazahmed.d.talikoti@ericsson.com>


Bug 8244  In Conntrack SNAT implementation G/w is not reachable when no
port in the router subnet is present in the node

1)Added a the pseudo port to vpn-dpn list.
2)Moved the router cache util to a separate class.

Change-Id: I6f5834ab3af1c3478274191bd7eafebc162dc25e
Signed-off-by: Aswin Suryanarayanan <asuryana@redhat.com>


stale entry in flow rules, after deleting port associated to qos policy with dscp

When we are delete the VM which has an associated qos policy with it,
the stale entry of DSCP marking use to persist for the table number 90.

The reason for this was that when QosNeutronPortChangeListner is called
rpc called to fetch DPN id use to fail. By moving the code to
QosInterfaceStateChangeLister we can use Interface to find out DPN id.

Change-Id: I485d84806eae2bcc8341518f8186973efb171658
Signed-off-by: Naveen Kumar Verma <naveen.kumar.verma@ericsson.com>


Bug 8310 - SNAT flows not added when the first subnet is added to the
external n/w after router g/w set

1)Check for the presence of external IP during add.
2)Update now considers whether a subnet is added after router g/w set.

Change-Id: Ibb6f2a5cf031db18a624e68391415e6574242434
Signed-off-by: Aswin Suryanarayanan <asuryana@redhat.com>


ClassifierUpdate unit tests

Change-Id: I4baa75c75c590e35a44da8b61e86eb27e625f79c
Signed-off-by: Jaime Caamaño Ruiz <jaime.caamano.ruiz@ericsson.com>


Fix remote acl bugs

1. Merging metadatas of lPort and aclId.
2. Handling reg6Match for ingress
3. Fixing dupplicate AclInterfaces in AclDataUtil
4. Adding all ips to remote acl filter table, not only those with remote acl rules
5. Fixing the flows instalation logic for the case of port in more than one SG.
6. Fixing handling of SG that uses itself as remote SG.

Change-Id: I64b0e294317e0ac1a3444192a92f8058f2604f1d
Signed-off-by: Slava Radune <slava.radune@hpe.com>


Netvirt: Qos Alert patch5

*** patch #5 - Port direct statistics polling  ***

Added support of  retrieval of port statistics data using OpenflowPlugin
direct-statistics RPC and log the alert message if packet drop ratio is
greater than the configured threshold value.

Spec Review Link:
https://git.opendaylight.org/gerrit/50689

Change-Id: I0a708ccd7eeb10c3d71fd8c08d5a2efeacdd3e25
Signed-off-by: Arun Sharma <arun.e.sharma@ericsson.com>


BUG 8284 - Unmatched packets get dropped by Classifier

- Packets that enter the Netvirt classifier that dont get
  matched are dropped.
- These packets should instead be resubmit back to the
  Ingress Dispatcher table.
- The solution is to add a lower priority MatchAny flow
  to the INGRESS_SFC_CLASSIFIER_ACL_TABLE table that
  resubmits back to the Ingress Dispatcher.
- Adding UT for the newly created flow.

Change-Id: I2d8b598a1146aa72a2c64c9b12835a7612e16c11
Signed-off-by: Brady Johnson <brady.allen.johnson@ericsson.com>


Fix the error in the egress classifier spec

The table numbers were wrong and could confuse newcomers

Change-Id: Idd57ab06f4c17801fc40dc90b00a410f5f4cfb74
Signed-off-by: Manuel Buil <mbuil@suse.com>
(cherry picked from commit ba22f7cf19d8a827d77a3391a7f654344ade43d8)


@Immutable FlowEntity with FlowEntityBuilder

This is dependent on the genius change
https://git.opendaylight.org/gerrit/#/c/53763/, which must be merged
simultaneously.

The change in the Xtend re. flowId is because the FlowEntityBuilder does
not have a getFlowId() like FlowEntity, only a setter; see
https://github.com/immutables/immutables/issues/432

Change-Id: Iecbe1ccd2c1c6cb9c914b70c6e44590d748ad739
Signed-off-by: Michael Vorburger <vorburger@redhat.com>
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 7758: Use Trunk instead of Transparent port for Flat networks

Netvirt should have used Trunk ports for Flat networks, instead of
Transparent ports - otherwise there are issues of mixing Trunk and
Transparent ports together, since Trunks are automatically created
once there are Trunk Members (VLAN networks) on the same port.

Change-Id: If666b672b1158919623db11b5324321f2f7b15b0
Signed-off-by: Alon Kochba <alonko@hpe.com>


Netvirt: Qos Alert patch4

*** patch #4 - Building Qos Alert port cache ***

This patch builds a cache of port stats data having QoS rate-limit rule applied.

A neutron port is added into the cache if -

1. Port QoS policy has bandwidth rate limit rule.
2. If port does not have QoS policy applied, then if network QoS policy
   has bandwidth rate limit rule.

Spec Review Link:
https://git.opendaylight.org/gerrit/50689

Change-Id: Iafb22bd98009b891a2ef4911ef0edf9f2ab9c767
Signed-off-by: Arun Sharma <arun.e.sharma@ericsson.com>


Bug 8244 - In Conntrack SNAT implementation G/w is not reachable
when no
port in the router subnet is present in the node

Creates an elan interface when a router port is created amd is deleted
on port delete.

Change-Id: I18fa39310f22c7ad633037c7bda54f51d193374b
Signed-off-by: Aswin Suryanarayanan <asuryana@redhat.com>


ClassifierEntry unit tests + javadoc

Change-Id: I1317a8e7cc0bc08bd027e8a91bf29111572feab8
Signed-off-by: Jaime Caamaño Ruiz <jaime.caamano.ruiz@ericsson.com>


EVPN RT2 elan changes to adv/withdraw RT2 routes..

1. Elan instance changes when network is attached and detached to EVPN
Neutronvpnmanager will add/delete/update evpn augmentaion in elan instance
EvpnElanInstanceManager is listening to the augmentation and based on type
of event EvpnElanInstanceManager will advertise or withdraw RT2 routes.

2. EvpnElanInstanceManager uses vpninstance to get rd to advertise or
withdraw routes.

Change-Id: I147da46e5d869cde560f25452e9e7511206a61dd
Signed-off-by: Riyazahmed D Talikoti <riyazahmed.d.talikoti@ericsson.com>


Bug-7718 Operational neutron port status

networking-odl requires feedback as to when a neutron
port can be marked as "ACTIVE". This commit places an
estimation of that information in the operational data
store. For ports that connect to OVS we wait for them
to connect and for basic L2 flows to be configured.
Note that we do not validate the flows in operational
nor do we check all flows. Ports that are implemented
purely as flows are marked ACTIVE immediately. The plan
is for networking-odl to receive notifications of these
status changes via a websocket.

It was decided at the time being
to keep the port status out of the ODL neutron project
since (a) functionally, this issue is mainly an issue for
netvirt and (b) to do it right in netvirt would require
some re-architecting.

Change-Id: Id719e904b277fe4dbb9c3d118d24c3bedf110a33
Signed-off-by: Josh <jhershbe@redhat.com>


Error messages are displayed when the port updated with qos policy stating
that binding are not allowed to the port

Following Error message was seen  when the same port updated with second
qos policy:

2017-04-05 14:42:54,116 | ERROR | pool-39-thread-7 |
FlowBasedServicesConfigListener  | 338 -
org.opendaylight.genius.interfacemanager-impl - 0.1.4.SNAPSHOT | Service
Binding entry update not allowed for:
b193e270-8053-4650-8059-a8ec57713069, Data:
BoundServices{getServiceName=qos.b193e270-8053-4650-8059-a8ec57713069,
getServicePriority=3, getServiceType=class
org.opendaylight.yang.gen.v1.urn.opendaylight.genius.interfacemanager.servicebinding.rev160406.ServiceTypeFlowBased,
augmentations={interface
org.opendaylight.yang.gen.v1.urn.opendaylight.genius.interfacemanager.servicebinding.rev160406.StypeOpenflow=StypeOpenflow{getFlowCookie=67108865,
getFlowPriority=10,
getInstruction=[Instruction{getInstruction=GoToTableCase{getGoToTable=GoToTable{getTableId=90,
augmentations={}}, augmentations={}}, getOrder=1, augmentations={}}]}}}

Use case:

- create a port, say port1
- create qos policy q1 with dscp and bandwidth
- Associate qos policy to the port
- observe the policy applied properly
- update the port again with qos policy q2
- Though the 2nd qos policy (here q2) seen applied properly , below error
  message was seen in ODL side:

  "Service Binding entry update not allowed"

  RCA:

  Qos service binding is happening even during the 2nd qos policy updation
  on the same port on which a 1st qos policy already associated.

Change-Id: Ia3754833d716f1e387bd7bdce00f82ed51a394ce
Signed-off-by: A Vamsikrishna <a.vamsikrishna@ericsson.com>


Bug 8189 - Policy flows are not updated after ovsdb other-config changes

Change-Id: Ic27b87922dd42aac020fa6f7e65bbab53e8c4489
Signed-off-by: Tali <tali.ben-meir@hpe.com>


Bug 8255 - VPN creation with multiple RD via Rest
is not working

fixed this by removing throwing of an error when
multiple RDs are passed to create VPN.

Change-Id: I28e546ca781d87566a2eb8b70148d980fa7dad3b
Signed-off-by: eswanit <swati.udhavrao.niture@ericsson.com>


Bug 8200:NAPT_PFIB_TABLE(47) with Internet VPN is not programmed for EVPN

This code review will address the following problem for EVPN NAT UCs.

1) NAPT_PFIB_TABLE (47) resubmit to L3_FIB_TABLE (21) {47->21} with
external (internet) VPN Id as match flow is not programmed for if external
network with Provider type VXLAN is configured.

2) When the external router is deleted some of the EVPN stale entry flows are
presented.

3) If more than one floatingIp is available in vpn-to-dpn-list for given
dpn id, do not call for installing INTERNAL_TUNNEL_TABLE (table=36) -> PDNAT_TABLE (table=25)
flow entry with same tunnel_id again and again.

Change-Id: I5a54764da323ed1ac20f91dd40fc36c06ed59ecb
Signed-off-by: karthikeyan <karthikeyan.k@altencalsoftlabs.com>
Signed-off-by: Sam Hague <shague@redhat.com>


ADDED INFO logs for tempest issue DEBUG

Change-Id: I25cb78056fbc6b7cbd372ac7565c21472e2c3060
Signed-off-by: cgowdru <chetan.arakere@altencalsoftlabs.com>
(cherry picked from commit 8a75dc835b6ff248da8762081f8bcc44e432e679)


BUG 8240 - Fix odl-netvirt-sfc dependent features

- Adding the odl-sfc-genius feature.
- Setting the correct version for carbon

Change-Id: Iee6f0a3df0e4dd9b3ae5f57035d0c4a1584dd163
Signed-off-by: Brady Johnson <brady.allen.johnson@ericsson.com>


Fix equals() bug in FederatedNetworkPair found by Checkstyle

This shows up as red in Eclipse, but the CLI mvn build missed this.

required for overall move to new Checkstyle version, see
https://git.opendaylight.org/gerrit/#/q/topic:bumpCheckstyle-stable/carbon

Change-Id: I7a46ac4c96e541d9fec17f3e6add9bcf65999a48
Signed-off-by: Michael Vorburger <vorburger@redhat.com>


Bug 8235 - Bug in group programming for directly connected subnet routes

Issue:

1. The ECMP feature enables the loadbalancing of traffic to the the extra
routes only but the discovered routes are also getting loadbalanced similar
to the extra routes.

2. This also leads to the issue in case of the discovered routes
containing macaddress(directly connected subnet route), the
loadbalancing groupId overrides the local nexthop group Id
as the parameters used for computing the groupId turns out to be same.

Analysis:

1. Since the discovered routes are also present in the vpntoextraroute
container, the discovered routes are also being currently loadbalanced by
ECMP feature.

2. Also, the loadbalancing groupId is computed from the prefix or the nexthop Ip
depending on whether the macadress is present or not(if mac is present,
the nexthopIp is used and prefix for otherwise). In case of the directly
connected subnet route, the macaddress is present and so the nexthopIp
is used for computing the loadbalancing groupId. This overlaps with
local next hop group programmed for that nexthop.

Fix:

Since the discovered routes have only 1 next hop, this
condition is now used to distinguish between discovered routes and the
extra routes. Only the extra routes are loadbalanced now and the
discovered routes are handled as before(creating a new local nexthop
group).

Change-Id: I6b0c5506f438ec1f816dfa0fdb045a8494950f48
Signed-off-by: gobinath <gobinath@ericsson.com>


FIB: VNI support in datapath for VxLAN networks

This commit adds support for VNI based L3 forwarding for VxLAN based
provider networks.
Spec: https://git.opendaylight.org/gerrit/#/c/48640/

Following changes are done:

1.
  i)   FIB_TABLE will set the destination network VNI in the tun_id
       field instead of the MPLS label.

  ii)  On egress OVS, match will happen on this VNI in table 36,
       and packet will be taken to ELAN pipeline in order to reach the
       destination VM.

  iii) Table 36 will no longer be programmed by FIB to match on MPLS
       label for internal router-based VPNs whose subnets are part
       of a VxLAN network

2. Some changes are done to program VNI in tun_id for ECMP LB NH
   groups.

3. The commit also incorporates the enforce-openstack-semantics
   flag to enable easy turning on/off of the feature.

4. Some refactoring and cleanup has also been done, majorly to
   remove the local utils/references for MDSAL sync
   write/delete/update APIs.

Change-Id:Id3244bd9f72028381e1790e96428a8f541f0e6b2
Signed-off-by: Abhinav Gupta <abhinav.gupta@ericsson.com>


ELAN: VNI support in datapath for VxLAN networks

This commit adds support for VNI based datapath forwarding for VxLAN
based provider networks.
Spec: https://git.opendaylight.org/gerrit/#/c/48640/

a. Unicast
  1. Table 51 now sets VNI in tun_id field for packets egressing on
     tunnel ports
  2. On egress, table 36 now matches on VNI in tun_id field, sets the
     ELAN tag in metadata, and forwards the traffic to table 51

b. Broadcast
  1. Remote broadcast group now sets VNI in tun_id field for packets
     egressing on tunnel ports
  2. On egress, table 36 now matches on VNI in tun_id field, sets the
     ELAN tag in metadata, and forwards the traffic to table 51

Also, integrates the openstack-vni-semantics-enforced flag to enable
easier turning on/off of the feature

Change-Id: I7a6b14897344a8d00cb19291602d92cf586f15f7
Signed-off-by: Abhinav Gupta <abhinav.gupta@ericsson.com>


Bug 8201 - ECMP not working with router based VPN

Issue:

In case of router based VPNs, the loadbalancing groups created for extra
routes are not updated with correct buckets when the extra route is
present behind different DPNs.

Analysis:

The issue is caused as the current design requires list of rds to
differentiate the extraroute present behind different DPNs.

Fix:

In case of the router based VPNs, the DpnIds are used as rds so the
vpntoextraroute is now populated correctly.

Change-Id: I18f6d39c67047a31817d442ba79aee6244bf3712
Signed-off-by: gobinath <gobinath@ericsson.com>


Bug 8221 - Remove Wait Time To Install L3VPN Groups on OVS

Currently we have a wait of 1.5s to ensure L3VPN groups are programmed
correctly in OVS, before we start programming the L3VPN Flows. This needs
to be optimized further.
This patch makes use of the OpenFlowPlugin addGroup() RPC call to directly
install the group in the OVS. This will ensure that the subsequent installation
of flows will always find the required groups in the OVS.
Group installation via the FRM is still retained to ensure consistency in the
datastore.

Change-Id: Ie4c2cf2a6199bad2a27d067d444a42ae04d3f191
Signed-off-by: eupakir <kiran.n.upadhyaya@ericsson.com>


Remove unused setFibManager

Change-Id: I627678151987577273bca88f5506a9faaedcd0c1
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 8188: L3VNI without configured UC handled for EVPN

Problem Description:
==================
L3VNI without configured in Internet VPN needs to be handled for EVPN in
NAT feature.
Currently Internet(External) VPN with L3VNI use case only handled for
external VXLAN network (EVPN) in NAT feature.

Internet(External) VPN without L3VNI use case also needs to be handled for
external VXLAN in NAT feature. Since Openstack Operator may configure L3VNI or
may not configure L3VNI.

Solution:
=========
Unconfigured of L3VNI value in Internet(External) VPN, will carve-out
the L3VNI value from OpenDaylight VXLAN VNI Pool to use SNAT and DNAT
flows.
The following UCs are handled based on the L3VNI configured/unconfigured
for GRE and VXLAN provider type.

Ext-Net-Type || Internet-VPN-L3VNI || External and Internal VXLAN uses
===============================================================================
GRE           Configured          Ignore L3VNI, allocate VNI from pool and
                                  use it for Intra-DC only
--------------------------------------------------------------------------------
GRE           Not Configured      A VNI taken from ODL Pool and used for
                                  Intra-DC communication
---------------------------------------------------------------------------------
VXLAN         Configured          Internet-VPN-L3VNI used for both
                                  Intra-Inter-DC communication
---------------------------------------------------------------------------------
VXLAN         Not Configured      A VNI taken from ODL Pool and used for both intra-
                                  inter-DC communication
---------------------------------------------------------------------------------

Note: If VNI value is unable to get from ODL VNI Pool, router-id will be
used as tunnel-id for external network provider type as VXLAN

Change-Id: I3c3f21e026303edf7eb2ad88bfd33622523085f0
Signed-off-by: karthikeyan <karthikeyan.k@altencalsoftlabs.com>
(cherry picked from commit 767d666b4a15f064be1478006328abf720c11c65)


cleanup ovsdb-ui bundle

Change-Id: I60da5608800bb36353e166334262e6f3bf33ac46
Signed-off-by: Sam Hague <shague@redhat.com>


Add ovsdb-ui bundle

Change-Id: If6b04d9ba61bcc32eac5603c13304b933318d184
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 8142 : DHCP timeout issue.

Issue:
stable/boron not usable - DHCP timeout

Analysis:
more than 5 thread were waiting on BGPConfigurationManager when BGP was
trying to send thrift messages to QBGP and QBGP is not up. BGP was
retrying it again and again. Meanwhile, services which call addPrefix is
getting blocked due to synchronized behavior of addPrefix.

Fix:
APIs exposed by BGP to other services - addPrefix, delPrefix, addvrf and
delVrf is made unsynchronized.

Change-Id: I6010b1eeef680b1f8d8908c2b9faf283b3827cfc
Signed-off-by: Vyshakh Krishnan CH <vyshakh.krishnan.c.h@ericsson.com>


Deprecate legacy features

Depends-on: https://git.opendaylight.org/gerrit/55084

Change-Id: Ia0ac9c5d427bd616eaa651df2e85297225afdfe3
Signed-off-by: Sam Hague <shague@redhat.com>


add odl-sfc-genius to odl-netvirt-sfc

Change-Id: I2512b14e15210bda5e5b8339621ea8a7b2b02f13
Signed-off-by: Sam Hague <shague@redhat.com>


BUG 8240 - Fix odl-netvirt-sfc dependent features

- Needed to install odl-sfc-openflow-renderer instead of
  odl-sfc-provider in order to bring in all the needed
  SFC features.

Change-Id: Id83d65c5b67214a826ff903cf48efeeca61aa99a
Signed-off-by: Brady Johnson <brady.allen.johnson@ericsson.com>


Bug 8241: Fix openstack-sfc ACLs IllegalArgumentException

Fix ace port range to be set only if mandatory lower port
is specified, otherwise we hit an exception.

Change-Id: Ic6206394e4255653a837c2b57cefc538555eefa4
Signed-off-by: Alon Kochba <alonko@hpe.com>


Bug 8170 - Loadbalancing nexthop group not updated after VM deletion

Issue:

When the last nexthop present in a DPN is deleted, the loadbalancing next
hop group is not updated. The group still contains the bucket pointing to
the deleted nexthop.

Fix:

There was an error while retrieving the endpointIpaddress for the DPN in
method getEndpointIpAddressForDPN. There was an error in converting the
IpAddress object to string. Appropriate conversion is used now.

Change-Id: I86cc77bd69446881d2ab669471f3142845e4d549
Signed-off-by: gobinath <gobinath@ericsson.com>


Bug 8162 - NPE at org.opendaylight.netvirt.vpnmanager.VpnSubnetRouteHandler.onSubnetAddedToVpn

Change-Id: I2936a473841b995a496c76064249a2bb9bd96354
Signed-off-by: Janki <jchhatba@redhat.com>
(cherry picked from commit f100cfb4e361fb07f91d2ec087839d292e94f1d0)


Set copyright for sfc translator back to Brocade

Change-Id: I576ff1dfe9c97cde540c1a56f77a857fb32e43eb
Signed-off-by: Sam Hague <shague@redhat.com>


Bug 8166: Acl interfaces are not set and cached correctly

The acl interface can exist in the cache before an interface
was assigned one. This can happen via the AclInterfaceStateListener.
Also, an interface ID should be set when a new entry is created.

Change-Id: I3e882582731eb273da533ac15dd2c8cc7f27ffe2
Signed-off-by: Yakir Dorani <yakir.dorani@hpe.com>


Use ArpConstants from arputil

Depend-on: https://git.opendaylight.org/gerrit/54890

This avoids duplicating constants and also helps to
break cyclics.

Change-Id: I76689a97d1131d13cffa382bbf4d07d7a3d19044
Signed-off-by: Sam Hague <shague@redhat.com>


Name all blueprint.xml file consistently

Change-Id: I7497339a96ff0f76f11e3f3c9937f57b6993169c
Signed-off-by: Sam Hague <shague@redhat.com>


Cleanup unused dependencies

Change-Id: I98eb44130f247108589db3aedf4e015b76ce03c1
Signed-off-by: Sam Hague <shague@redhat.com>


add interface for VpnFootprintService

This is another attempt to reduce cyclics.

Change-Id: I29acdc7fa4c681c24b0cafc8430716f189bd7be1
Signed-off-by: Sam Hague <shague@redhat.com>


Remove utils.config

Change-Id: Iaf232a25f055b26e392d337e85a2b15409b9c03d
Signed-off-by: Sam Hague <shague@redhat.com>


Remove legacy utils from new netvirt

Change-Id: I43a183b689846924ac8344117bfa59cae0457a4a
Signed-off-by: Sam Hague <shague@redhat.com>


BUG 8193 - Fix Netvirt classifier egress service port binding

- The Netvirt classifier egress service should bind on egress
  ports, not ingress ports like it does now.
- Since its not possible to know all the possible egress ports
  to bind on before-hand, the egress classifier service will
  bind on all switch ports. It will only process NSH packets
  and return all others to the egress dispatcher.
- Also in this patch, when the SFF is on the same bridge as
  the classifier, the egress classifier will resubmit the
  packets directly to the SFF instead of going through the
  ingress dispatcher, since we dont know the correct ingress
  port to use for the SFF.
- Updated OpenFlow13ProviderTest to reflect the change to the
  egress classifier resubmit.
- Changes from code review comments
- Final changes to make sure the egress binding is working.
  Now using getDpnInterfaceList RPC to get the switch ports
  to bind to.

Change-Id: I97bc38722064738ee22b8ddbc7163bc5dc3dd276
Signed-off-by: Brady Johnson <brady.allen.johnson@ericsson.com>


Bug 8105: IllegalArgumentException in getSrcIp fix

1. reading IP and ARP payload was wrong, which is fixed by getting payload from
deserialised ethernet data based on IPv4 or ARP, getPayload will return
the right payload as the offsets are maintained in Packet based on
type of payload.
2. IllegalArgumentException was for payload type IPv4 where
getSourceAddress method will return of type int, and
NWUtil.toStringIpAddress will accept input as byte.
The same was working fine with payload type ARP as
getSenderProtocolAddress return byte.
3. Tested these changes by creating 2 ports in neutron network, delete 1
port and send ping packets from deleted port to the other port.
First ARP packet hit and was able to retrieve IP address properly, and
then data traffic hit and was able to retrieve IP address.

Change-Id: I16af2e18e2c857a5529f4ffef1caf9da014d521f
Signed-off-by: Riyazahmed D Talikoti <riyazahmed.d.talikoti@ericsson.com>
(cherry picked from commit 0c9a0f34e776cf3620cc4eed8a5f0afe5eec1d23)


Bug 8014 - L2Gw connectivity not working

Tunnel creation issue as cache were not populated properly
Fixed .

Change-Id: I7613afe762f0a865275c3fc492045ef4d035ba84
Signed-off-by: Akash Kumar Sahu <a.k.sahu@ericsson.com>


Cleanup version properties

Change-Id: I15c3131032866d0094b7f34ac7e4c85cfe139665
Signed-off-by: Sam Hague <shague@redhat.com>


Update .gitreview to stable/carbon

Change-Id: Ib6980e6ee2c6ff084075771ff350881fe8024265
Signed-off-by: Anil Belur <abelur@linuxfoundation.org>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants