Robert Varga [Thu, 13 Dec 2018 15:55:53 +0000 (16:55 +0100)]
Add LLGR openconfig extension
This adds the configuration extension, so that LLGR can be
configured through OpenConfig.
JIRA: BGPCEP-495
Change-Id: Iadcd3915e094a750418480df7b237da88fca00cd
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 15:19:04 +0000 (16:19 +0100)]
Undo damage done by https://git.opendaylight.org/gerrit/#/c/69403/
This patch restores the ability to lookup RIBSupportContext
through NodeIdentifierWithPredicates.
Change-Id: I2b9399620f4935a50e8458042a767bc3e1aa5b59
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 16:00:07 +0000 (17:00 +0100)]
Fix extension's groupId
Rather than inheriting odlparent, make sure we use bgpcep.
Change-Id: I64c696d67d178c243fc6cf76ef3762c329b7b9ef
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 09:33:56 +0000 (10:33 +0100)]
Switch from findbugs to spotbugs
SpotBugs is the replacement for FindBugs, switch to using it.
Change-Id: I2e5b21e87f85d4bb46c343943255364e2231e766
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 12:43:56 +0000 (13:43 +0100)]
Translate GR RPC
Rather than blocking current thread, bridge futures so that we
return without blocking.
Change-Id: I8fec49c96b104b952de7a87b83b1bc4daf17f0d6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 11:58:47 +0000 (12:58 +0100)]
Fix remaining checkstyle in bgp-parser-impl
This fixes up the last straggler and flips enforcement to on.
Change-Id: I5fad39b6105230cb0e4474b6414525aab6736b00
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 11:59:10 +0000 (12:59 +0100)]
Fix most bgp-parser-impl checkstyle
This fixes all but one violation.
Change-Id: I4b34e87cc43946f9bddd485f57573355d076bebf
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 10:53:15 +0000 (11:53 +0100)]
Enforce bgp-parser-spi checkstyle
This patch removes the final violations and flips enforcement on.
Change-Id: Ia02483620694df804ed16c97db1815c066a79be5
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 10:53:37 +0000 (11:53 +0100)]
Fix most bgp-parser-spi checkstyle violations
These violations have been piling up, clean most of them up.
Change-Id: I3f42f7d49028117b81b9963ab3efa750082cc7db
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 10:00:56 +0000 (11:00 +0100)]
Enforce checkstyle in labeled-unicast
This fixes checkstyle violations and flips enforcement on.
Change-Id: I92e080e8a4950727019ce7351afe8311aaac36bf
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 08:17:47 +0000 (09:17 +0100)]
Turn bgp/extensions into an aggregator
This is a pure aggregator, do not also make it a parent, as that
welds project layout to published artifacts.
Change-Id: Ic5190d94d14c2b56e59812e79ce1064e6a866730
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Matej Perina [Sun, 25 Nov 2018 09:26:04 +0000 (10:26 +0100)]
documentation for graceful restart
- documentation for two new capabilities,
graceful restart and long-lived graceful restart
- update documentation of operational state, which
now shows state of graceful restart procedures
Change-Id: Ic698b9e70b2242e53469add7e960af575d76e228
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 07:22:19 +0000 (08:22 +0100)]
Do not install/deploy aggregator poms
THese are just build helpers, there is no point in installing
and deploying them.
Change-Id: Ieda0493d2fb0eb5072c60a234f502fee1a5ad7be
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 13 Dec 2018 07:20:18 +0000 (08:20 +0100)]
Add support for aggregate sonar reports
This patch teaches the top-level aggregator to combine jacoco
reports for all artifacts, so that sonar can pick it up.
Change-Id: Ie116fcb069b6c384783f0a3f751eb15186dd42ff
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 12 Dec 2018 12:52:46 +0000 (13:52 +0100)]
Extract routeKey/pathId from the identifier
Rather than mucking with old/new route content, extract the needed
information from the identifier, which is always available.
Change-Id: I7fedb4a7b4d670bc837388ef492d6d32ce3a244c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 12 Dec 2018 23:41:52 +0000 (00:41 +0100)]
Teach BGPSessionImpl about LLGR
This refactors BGPSessionImpl to unify support for both regular
and LL graceful restart.
JIRA: BGPCEP-495
Change-Id: I95e863fd24fa9118d7068c489856f35df0cb49df
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 12 Dec 2018 21:11:26 +0000 (22:11 +0100)]
Teach BGPSession about LLGR
This adds the API to inquire about LLGR capability, as advertized
by the peer. Also refactors normal GR to take the same default-based
approach.
JIRA: BGPCEP-495
Change-Id: I53599053fd3b206e93b470a04511244c2edb74c4
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 12 Dec 2018 20:45:15 +0000 (21:45 +0100)]
Make LocRibWrite.localAs a long
Peer selection is taking a simple long, there is no point in us
retaining the boxed value.
Change-Id: I370382c6225e62e8df0902204551d657e5c02f22
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 12 Dec 2018 20:34:47 +0000 (21:34 +0100)]
Teach AbstractPeer about LLGR_STALE routes
When filtering routes towards a peer, we need to make sure we do not
send out LLGR_STALE routes if the peer has not advertized the
capability.
JIRA: BGPCEP-495
Change-Id: I3f8bd3b13d632d572efa8626f0b29ece06477ffa
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 12 Dec 2018 12:00:28 +0000 (13:00 +0100)]
Fix LocRibWriter.create()
Create method was relying on raw types without properly capturing
types involved. Fix that by propagating types and clean up some
formatting issues.
Change-Id: Ieadb6f42e8529c1e6b40f058a5e5e17ac65e09ff
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 12 Dec 2018 11:03:25 +0000 (12:03 +0100)]
Cache afi/safi type
Rather than looking up afi/safi type for each individual route,
look it up once when we know we'll need it and reuse it afterwards.
Change-Id: Ifac6b5df4ef402fa0a0a12026c4fdf96cf7c885c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 15:13:19 +0000 (16:13 +0100)]
Remove ThreadPool in connection release path
This is a simple read operation, there is no need to allocate
a threadpool to handle the read requests.
Change-Id: I2d482e052a8f321d69f621065a8e11948234a1d8
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Fri, 7 Dec 2018 04:27:26 +0000 (05:27 +0100)]
Remove duplicate OffsetMap code
Having two copies of the same class is confusing and a problem
with maintenance. Create an abstract base template to hold most
of the code.
Change-Id: I3078c2af94d440701d48e43da31c106fdbac78d2
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 12:28:39 +0000 (13:28 +0100)]
Introduce a dedicated RouterId class
We perform quite a few conversion from/to IP addresses and PeerIds,
for which UnsignedInteger-based lookups are not efficient.
Create an equivalent RouterId class, which understands its
relationship with PeerId, hence we can side-step most of the lookups.
Change-Id: If703c620192496962cde9dc68cdf228fbd2fb6ff
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 23:30:33 +0000 (00:30 +0100)]
Fix addpath selection performance
As it turns out, the two selection modes operate under completely
different assuptions, rendering
AddPathAbstractRouteEntry.selectBest() utterly useless.
Specialize N-path and All-path selectors to operate in the offset
order, which makes it clear OffsetMap.getRouteKeys() is unneeded
and that a simple offset-based helper,
AddPathAbstractRouteEntry.processOffset() is all that we need to
efficiently select paths.
Also switch N-path to use int, which is more than enough to cover
the uint8 range we really need.
The end result is that we make minimal memory allocations to
fulfill the needed functionality while also improving access
locality, while side-stepping offsetOf() searches.
For bonus points, we get to completely hide RouteKey, OffsetMap,
the AddPathBestPath constructor and all AddPathAbstractRouteEntry
fields.
Change-Id: Ia68cd3f88aef04a04506f3a2cc96d04485c19ad9
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Matej Perina [Mon, 10 Dec 2018 11:12:29 +0000 (12:12 +0100)]
add P2MP to supported capabilities
Extensions to the Path Computation ElementCommunication Protocol (PCEP)
for Point-to-MultipointTraffic Engineering Label Switched Paths
added to list of supported capabilities
JIRA: BGPCEP-406
Change-Id: I899c272f2aaed0899e3946d220a01c4db7c4c31e
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 18:18:57 +0000 (19:18 +0100)]
Prepare EffRibInWriter for LLGR_STALE handling
When a peer configured with Long-Lived Graceful Restart goes down,
we need to process all tables which it advertised as LLGR-enabled
and mark all routes as being tagged with LLGR_STALE.
LLGR_STALE community must be added to routes behind import
policy processing since policies can also add this community.
This patch should provide procedures necessary to add LLGR_STALE
community to routes.
JIRA: BGPCEP-495
Change-Id: If921e378f998109b073c628f9913f395496b3139
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 21:02:28 +0000 (22:02 +0100)]
Use boxed Path IDs for AddPath
AddPath entries are usually used to instantiate PathId objects,
which are forcing boxing. Since we are already keeping path IDs
in boxed format, this patch makes sure we take advantage of this.
Change-Id: Ia2566e48ef44313f5448bf4e23f4ff0f41c8a69b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 19:26:03 +0000 (20:26 +0100)]
Centralize PathId allocation
Using long for pathId selection is highly inefficient, as it forces
each AFI/SAFI to deal with it separately. Concenrate allocation in
a single place, so we can actually have a proper interface.
Change-Id: I515d7f0c6fce80b45561939999322f273e8a4197
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Matej Perina [Wed, 5 Dec 2018 17:28:30 +0000 (18:28 +0100)]
JUnit tests for attribute parsers
poviders coverage for parsers of attributes:
- AGGREGATOR
- ATOMIC-AGGREGATE
- AS-PATH
- LOCAL-PREFERENCE
- MULTI-EXIT-DISCRIMINATOR
- NEXT-HOP
- ORIGIN
- ORIGINATOR-ID
JIRA: BGPCEP-359
Change-Id: I901f6d3474dd8431d52440c25aa1b0158c951cbb
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 23:29:44 +0000 (00:29 +0100)]
provide way to extract routes from routesContainer
EffectiveRibInWriter operates with binding databroker and while
we have tools to construct InstanceIdentifier for routesContainer
we are missing tool to extract routes from it. This patch should
provide just that.
JIRA:BGPCEP-495
Change-Id: I1aea06ec4f082d11643566aed761fcc4092b6fcc
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Matej Perina [Wed, 28 Nov 2018 17:15:58 +0000 (18:15 +0100)]
P2MP Request message
change to PCReq message according to
https://tools.ietf.org/html/rfc5440#section-6.4
JIRA: BGPCEP-406
Change-Id: Ief3361b5617318b4cb812bed2e8c405c1a2442c5
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Robert Varga [Sun, 9 Dec 2018 20:53:22 +0000 (21:53 +0100)]
Teach RIBSupport about non-addpath routes
This patch exposes an explicit method to create a pathId=0 route
key, which allows us to centralize things a bit.
Change-Id: If2ae95ab3298dca400c14ecbb4e77c94712c06a2
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 21:30:03 +0000 (22:30 +0100)]
Fix RouteEntry.{add,remove}Route()
This API is interfacing with MD-SAL, where values are boxed,
hence these should be propagate as they are, not forcibly unboxed
in LocRibWriter.
This improves non-addpath performance, as we are simply skipping
the unboxing operation, since we do not care at all about pathIds.
Also hide RouteKey constructor and its accessor, as everything
aside of it being an identifier (hence immutable) is a purely
package-private matter.
Change-Id: Ica7612ec5e6df8d07eb16294c1efa5fb0428f884
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Claudio D. Gasparini [Tue, 13 Nov 2018 15:48:07 +0000 (16:48 +0100)]
UNREACH-DESTINATION Object
https://tools.ietf.org/html/rfc8306#section-3.14
JIRA: BGPCEP-406
Change-Id: Ie7ef63e7f166059655ff558623c686f087b6302f
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Robert Varga [Mon, 3 Dec 2018 05:31:26 +0000 (06:31 +0100)]
Update error handling of PE Distinguisher Labels
RFC6514 actually specifies behavior consistent with RFC7606, make
sure we use the same mechanism to correct handling.
Change-Id: If9c5f313bfdf4b54f9f3ae09f91c08e2a7ae8d65
JIRA: BGPCEP-359
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Mon, 3 Dec 2018 05:30:35 +0000 (06:30 +0100)]
Migrate LinkstateAttributeParser
While the error handling of linkstate attribute is not quite clear
yet, migrate it to inherit from AbstractAttributeParser.
Change-Id: I7900d3ec816451daf248c3bacef902d607809ae2
JIRA: BGPCEP-159
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Mon, 3 Dec 2018 01:27:51 +0000 (02:27 +0100)]
Add AbstractAttributeParser
While AttributeParser makes revised error handling optional, this
class makes it mandatory, providing a common baseline for parsers
dealing with attributes which have revised error handling specified.
Also migrate parsers for following attributes to use it:
- ORIGIN (Section 7.1)
- AS_PATH (Section 7.2)
- NEXT_HOP (Section 7.3)
- MULTI_EXIT_DISC (Section 7.4)
- LOCAL_PREF (Section 7.5)
- ATOMIC_AGGREGATE (Section 7.6)
- AGGREGATOR (Section 7.7)
- Community (Section 7.8)
- ORIGINATOR_ID (Section 7.9)
- CLUSTER_LIST (Section 7.10)
- MP_REACH (Section 7.11)
- MP_UNREACH (Section 7.12)
- Extended Community (Section 7.14)
JIRA: BGPCEP-359
Change-Id: I47de7426f811962f6771a5925486b133986f0a44
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Claudio D. Gasparini [Tue, 13 Nov 2018 07:23:45 +0000 (08:23 +0100)]
SVEC object flags
https://tools.ietf.org/html/rfc8306#section-3.12
JIRA: BGPCEP-406
Change-Id: I4a37a79c415e845d984c81707a5a5393423080d9
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Mon, 12 Nov 2018 14:49:20 +0000 (15:49 +0100)]
Branch Node Object
https://tools.ietf.org/html/rfc8306#section-3.11.1
JIRA: BGPCEP-406
Change-Id: I3ae8105f1ec4ec5afbbf94fc34b86f16a8b8f5f6
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Thu, 8 Nov 2018 09:23:05 +0000 (10:23 +0100)]
P2MP end points
JIRA: BGPCEP-406
Change-Id: Ia582bf5e93aef41340bc3f7f19cf9bdfe6003655
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Thu, 8 Nov 2018 09:21:16 +0000 (10:21 +0100)]
P2MP path computation capability
https://tools.ietf.org/html/rfc8306#section-3.1.2
JIRA: BGPCEP-406
Change-Id: Idee90c040d1cf242c3b3146722097f8c68a8b9f8
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Matej Perina [Fri, 7 Dec 2018 09:32:57 +0000 (10:32 +0100)]
fix graceful restart test race-condition
There is possibility that we close session faster
than EOR is send to peer. In order to verify no notification
was sent after OPEN receival we need to wait for EOR to be
sent before we send OPEN.
JIRA:BGPCEP-495
Change-Id: Ifc2b4fa5d74f18bb1d460161559a1cdec10a6282
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Claudio D. Gasparini [Fri, 9 Nov 2018 16:21:53 +0000 (17:21 +0100)]
Add new revision for pcep types model
With P2MP extension model changes we
also need new revision to mark this addition.
JIRA: BGPCEP-406
Change-Id: I90de459addfe451013f2a39e1674e9f4eb3cf933
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Robert Varga [Mon, 3 Dec 2018 10:34:19 +0000 (11:34 +0100)]
Speed up PeerSpecificParserConstraintImpl
This improves design by taking advantage of ClassToInstanceMap,
which provides a type-safe getInstance() method. Furthermore we
recognize that the this class is used mostly for lookups of a few
classes, we use an ImmutableClassToInstanceMap updated concurrently
via a compare-and-set.
This yields significantly faster lookups for the common cases, where
the map is empty or has a single entry. In other cases the speed up
is still there, although not as pronounced.
Change-Id: I4da36e592e88cf22d5ad1654f126eec81660c98c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Matej Perina [Thu, 6 Dec 2018 13:55:43 +0000 (14:55 +0100)]
close effRibInWriter properly
We cannot close effRibInWriter before we clear tables
without graceful restart support.
Change-Id: Ife2f76aa4b2a9db991b570c9930084c17aeb2cc0
JIRA:BGPCEP-495
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 17:30:14 +0000 (18:30 +0100)]
Do not rebox pathId
Path IDs are stored in boxed array, using a long temporary forces
it to be unboxed, only to be boxed again when we put it into the
list.
Change-Id: I591f1259c61a0bc2c4a410af3810222b4fe7dc87
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Ajay Lele [Tue, 27 Nov 2018 19:31:24 +0000 (11:31 -0800)]
Fix unit of BGP operational neighbor uptime in doc
Change-Id: I3873b68f23b57219d1f7b10170ae414c898bcb50
Signed-off-by: Ajay Lele <ajayslele@gmail.com>
Robert Varga [Thu, 6 Dec 2018 14:34:59 +0000 (15:34 +0100)]
Split out EffectiveRibInWriter.processModifications()
If we have non-empty modifications, we know we will need a transaction,
separate that handling into a separate method, so we can side-step
null checks.
Change-Id: Ic391790c5ab55a8259849ce330d3af94db3cf6f6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 14:41:01 +0000 (15:41 +0100)]
RIBImpl.vpnTableRefresher should be final
This field not updated from anywhere, make it final.
Change-Id: I843bc7c87d5eb69b6c48841013ec3d7d11aa93a6
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 19:08:56 +0000 (20:08 +0100)]
Do not always update eff-rib-in attributes
Check if the attributes changed before propagating the change,
reducing the number of operations executed.
Change-Id: I6435a50aefbaefabec86f8d6d7f220c9a1bfaa20
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Olivier Dugeon [Wed, 28 Nov 2018 13:15:28 +0000 (14:15 +0100)]
Correct infinite loop in base parser
When sending a PCReply message with more than one ERO object, as per RFC540,
the PCEP base parser start infinite loop and stop with a java heap exception.
The problem is located in PCEPReplyMessageParser.java (pcep-base-parser
package) at line 217 in handleEro() method. handleEro() is called with the first
ERO object and remaining object to parse metrics associated with this ERO. The
parsing is done by parserPath() method, but this method expect to find only
metrics object. When a second ERO follow the last metric of the first ERO,
parsePath() method do nothing and handleEro() method start infinite loop at line
220 as the object list is not empty and parsePath() will not empty it.
This patch replace handleEro() method by handleEros() in order to parse
correctly multiple ERO in PCReply message. It also add a new test in
PCEPValidatorTest.java class: testPCRepMsgWithTwoEros() in order to verify
that a PCPep Message with two EROs is correctly parse.
JIRA: BGPCEP-851 PCEP base parser is unable to handle multiple ERO
Change-Id: I194b7cbb347a3b377e7b3000f93ef1d21a86f39a
Signed-off-by: Olivier Dugeon <olivier.dugeon@orange.com>
Robert Varga [Thu, 6 Dec 2018 11:56:18 +0000 (12:56 +0100)]
Prevent unnecessary boxing in AddPathBestPath
The callers are of two varieties: boxed and unboxed. Unboxing is
inherently faster, hence force the single boxed callsite to unbox
allowing us to convert it when needed.
Change-Id: Ie4d80398724b2f98a186e500c73e33ed1775a406
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 11:52:10 +0000 (12:52 +0100)]
Prevent boxing of AS in BasePathSelector
Superclass uses long, just as all callers. Ditch intermediate
boxing.
Change-Id: I67e8678ebbac4084460e69766686b0feb25ae3db
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 11:48:15 +0000 (12:48 +0100)]
Optimize path selectors
Creating multiple object just for the purposes of LOG.trace() is
utterly wasteful. Guard the LOG.trace() with isTraceEnabled() and
also expose a String-only formatter.
Change-Id: If6f3011e89772f82948816a975034a38a82a7922
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 11:39:20 +0000 (12:39 +0100)]
Fix RouteKey and surrounding code
This cleans up RouteKey to be a properly-designed and efficient
DTO:
- prevent useless boxing
- use proper prime in hashCode
- make sure nullness is propagated
- improve compareTo performance
Change-Id: Ibfa1f57fee8dbc4a0a4c1a0f0e47664b48f4d055
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Thu, 6 Dec 2018 10:57:50 +0000 (11:57 +0100)]
Remove unneeded casts
We already are comparing longs, no need to cast.
Change-Id: I96b9bb5e4a85a96236295abb995bb8b421a57039
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 22:18:34 +0000 (23:18 +0100)]
Allow AttributeRegistry to signal treat-as-withdraw
When any attribute indicates we should be treating the resulting
message as withdraw, we need to report this indication upwards,
so that the it can be evaluated once the complete contents of
the message are known.
This can result in either a conversion of reachability to
unreachability, or, if we fail to parse the NLRI-bearing fields,
in a session teardown.
Change-Id: I8c3b7b9d9f9523b42a328820cc6b7c9168976aeb
JIRA: BGPCEP-359
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 17:15:52 +0000 (18:15 +0100)]
Optimize AbstractBestPathSelector
There are inefficiencies in how we compare local preference, let's
refactor the code to have clear comparisons without unnecessary
boxing.
Also remove implicit boxing of AS numbers, as we are already getting
them as primitive longs.
Change-Id: I3cbba14fa2c2bd2add35c0f827da8ba4a06a0b4b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 21:53:20 +0000 (22:53 +0100)]
Propagate depreferenced status through AbstractAdvertizedRoute
AbstractPeer will need this info to differentiate which routes
to advertize when, propagate it out of selection state.
Change-Id: I83a243ffe98e55c719913f7c7b708b0ccab2cee2
JIRA: BGPCEP-495
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 21:22:27 +0000 (22:22 +0100)]
Merge ActualBestPathRoutes and AdvertizedRoute
These two classes are almost identical, extract the common bits
to a common superclass, reducing complexity and allowing them
us to share codepaths.
Change-Id: Ib6bbf1e4285ed0f1376d932367c8cb0d3f6f2662
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 21:26:28 +0000 (22:26 +0100)]
Fix isFirstBestPath brain damage
Using a List.indexOf() followed by equality to 0 is ... inefficient
to say the least.
Kill isFirstBestPath() and use a simple check against the first path,
if the list is non-empty.
Also optimize list allocation and remove useless copying while we're
at it.
Change-Id: Ia1a48736ff363ab15776b9f188207720f23169e2
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Mon, 3 Dec 2018 00:10:55 +0000 (01:10 +0100)]
Introduce BGPTreatAsWithdrawException
Attribute parsers need to be able to report that the attribute
failed to parse and the UPDATE message needs to enter
treat-as-withdraw error recovery.
BGPTreatAsWithdrawException holds this information and can be
reported from individual attribute parsers.
Change-Id: I459eb70692cde1035ca03110f414acb60d302abd
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 17:48:07 +0000 (18:48 +0100)]
Teach AbstractBestPathSelector about depreferenced routes
draft-uttaro-idr-bgp-persistence-04 mandates that routes tagged
with LLGR_STALE are to be used as routes of least preference.
Teach best path selection about this use it as the first step
of discrimination: depreferenced paths will always lose to
non-depreferenced paths.
JIRA: BGPCEP-495
Change-Id: I81a1ba1d09c554dee754329b5a907fbff3d6bf84
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 17:38:44 +0000 (18:38 +0100)]
Teach BestPathState about depreferenced routes
draft-uttaro-idr-bgp-persistence-04 defines the concept of a
"depreferenced" route, i.e. routes which are tagged with LLGR_STALE
community. These routes should be used as routes of last resort.
Change-Id: I19032bf94dce2f073458dfd9a9cd120a496db4f6
JIRA: BGPCEP-495
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 16:12:00 +0000 (17:12 +0100)]
Add Long-lived Graceful restart capabilities
This adds the definition of protocol-level capabilities
for draft-uttaro-idr-bgp-persistence-04.
Change-Id: I9885f65615678d8f01e238b0f2aa367de61bb60c
JIRA: BGPCEP-495
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 16:49:56 +0000 (17:49 +0100)]
Fix eclipse nullness warnings
Eclipse does not understand that Attributes is expected to be
immutable, hence it emits warnings about items being potentially
null.
Since normally we get attributes from the datastore, it is actually
helpful to store them in local variables, preventing roundtrips
to LazyDataObject -- fixing warnings as well.
Change-Id: I82fc82a69b07c739af566a069d225bbf18db98ae
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 14:54:59 +0000 (15:54 +0100)]
Add Long-lived Graceful Restart communities
https://tools.ietf.org/html/draft-uttaro-idr-bgp-persistence-04
introduces two new communities, NO_LLGR and LLGR_STALE.
This patch adds support for them.
JIRA: BGPCEP-495
Change-Id: I21745dffe9d57164f55020cf50bb443fcf6fc533
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 5 Dec 2018 14:05:00 +0000 (15:05 +0100)]
Clean up PeerTrackerInformation
Fix documentation and add a final qualifier.
Change-Id: I19446071c07b1613629501f137b6db90cc9daeeb
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Sun, 2 Dec 2018 23:49:24 +0000 (00:49 +0100)]
Distinguish between ignored and erroring attributes
MP_REACH/MP_UNREACH attributes need to cause a BGPDocumentedException
to be thrown when in RFC7606 mode. This adds that capability and
makes parser report it.
Change-Id: Iabbd16e64332ad9ffd823e942262b7ba324ad328
JIRA: MDSAL-359
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Mon, 26 Nov 2018 09:48:50 +0000 (10:48 +0100)]
Use templates for predicates
This speeds up instantiation slightly.
Change-Id: I325ff2e7ffe197553d922abd0169d5b21e15e483
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
(cherry picked from commit
c6c29b64ca7ca58aef3e5e8d6aa08505ce5b75cd)
Claudio D. Gasparini [Mon, 3 Dec 2018 08:17:08 +0000 (09:17 +0100)]
Decouple and create features per each topology bundle
Change-Id: I06e914041d48b4aefb6d234094f6c2af0f94b55b
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Sun, 2 Dec 2018 20:26:45 +0000 (21:26 +0100)]
Use features instead od bundles
under existent features. Avoid duplication
Change-Id: Ia7503d65e7b63d82de114c8d9485028ba6de8984
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Matej Perina [Mon, 5 Nov 2018 11:30:19 +0000 (12:30 +0100)]
send EOR marker after initial route exchange
JIRA: BGPCEP-807
Change-Id: I374826c8faaa94e8ba8394588b7de0a00acf101f
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Robert Varga [Mon, 3 Dec 2018 01:18:21 +0000 (02:18 +0100)]
Migrate Matchers.any()
This method has been moved to ArgumentMatchers, make the switch.
Change-Id: I710791942f2f342e636cc77713e8867b3a75cd32
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Sun, 2 Dec 2018 23:44:54 +0000 (00:44 +0100)]
Add RevisedErrorHandlingSupport ParserConstraint
We need to communicate to the parser whether RFC7606 is enabled
and whether the peer is internal or external. This patch adds
the interface to do that.
JIRA: BGPCEP-359
Change-Id: Id123c4d25019bd1be1ec7b228cd010c81186a4a4
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Matej Perina [Wed, 19 Sep 2018 06:48:18 +0000 (08:48 +0200)]
introduce graceful restart speaker capability
JIRA: BGPCEP-808
Change-Id: I6c4ac1b7697311c31405d0c660fb3e53ca1d8e78
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Robert Varga [Sun, 2 Dec 2018 23:42:20 +0000 (00:42 +0100)]
Add Ipv4Util.prefixForByteBuf(ByteBuf, int)
This utility method is needed to parse the IPv4 prefix in two separate
validation steps.
Change-Id: I43f05987222c2cd953760d0b57c3f73240d75f1b
JIRA: BGPCEP-359
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Mon, 26 Nov 2018 09:20:49 +0000 (10:20 +0100)]
Bump to yangtools-2.1.4 and mdsal-3.0.2
This bumps the dependencies to 2.1.4 and 3.0.2, respectively.
Change-Id: I73c9ae102298ae3c7bae7f3d1a74a43b626509ba
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Matej Perina [Thu, 6 Sep 2018 09:37:46 +0000 (11:37 +0200)]
introduce graceful restart listener capability
JIRA: BGPCEP-807
Change-Id: I0f31bd03d63ec48749d8cf4a2491016e30180c0f
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Matej Perina [Tue, 20 Nov 2018 14:15:57 +0000 (15:15 +0100)]
add option to close peer session without notification
- from RFC4724 comes need to have option to close
session without sending NOTIFICATION in certain
situations
JIRA: BGPCEP-807
Change-Id: Ib560c3a15c1f2227829c151697a6fde98e3a7eed
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Robert Varga [Wed, 21 Nov 2018 09:31:27 +0000 (10:31 +0100)]
Remove AttributeParser.parseAttribute(ByteBuf, int)
We require peer information for parsing, promote the peer-aware
parsing method as the primary contract.
JIRA: BGPCEP-359
Change-Id: I7fe039034563ce416ac6f6d08fbf6de867c35344
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Tue, 20 Nov 2018 13:34:34 +0000 (14:34 +0100)]
Remove MessageParser.parseMessageBody(ByteBuf, int)
This method is no longer the proper entrypoint, remove it and update
all parsers to implement the replacement.
JIRA: BGPCEP-359
Change-Id: I6fff6cf9f0275d9562c9e37ce55d34287251794b
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Tue, 20 Nov 2018 12:10:44 +0000 (13:10 +0100)]
Reduce use of newArrayList()/newHashSet()
These methods are not needed with modern java, do not use them.
Change-Id: If6cb8780f2277c80ef84d72daa89f92833677e84
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Tue, 20 Nov 2018 12:01:51 +0000 (13:01 +0100)]
Require Peer to return a FluentFuture
Rather than specifying ListenableFuture, make sure we get a FluentFuture.
Change-Id: I35999c0c7ae8c0c5fa64fdbf4391b6604c60f91c
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Tue, 20 Nov 2018 11:59:44 +0000 (12:59 +0100)]
Remove duplicate message type check
Rather than performing two instanceof checks, use an if/else
construct to dispatch message.
Change-Id: I4ce5aba7e1c826fe343b279c9767193d0ba83154
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Claudio D. Gasparini [Wed, 14 Nov 2018 17:18:35 +0000 (18:18 +0100)]
SERO & SRRO
https://tools.ietf.org/html/rfc8306#section-3.2
JIRA: BGPCEP-406
Change-Id: I47e496fc821325b30c31368a4f906a2cb66434b2
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Thu, 8 Nov 2018 09:15:01 +0000 (10:15 +0100)]
Decouple sero/srro subojects models from sero/srro objects models
JIRA: BGPCEP-406
Change-Id: I197735ed18278dec010f5e595c7619efb8c5d232
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Robert Varga [Thu, 15 Nov 2018 18:17:00 +0000 (19:17 +0100)]
Assert route change identifier being IdentifiableItem
Rather than doing a straight cast, verify the object beforehand,
giving us a proper data instead of a ClassCastException.
Change-Id: Ia9dafc9d62028fbe66fc11036d6873821c5ef8a7
JIRA: BGPCEP-848
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Robert Varga [Wed, 14 Nov 2018 21:14:43 +0000 (22:14 +0100)]
Fix an eclipse nullness warning
Eclipse (rightfully) flags this access as potentiall null. Since
we are invoking an interface getter, it is not statically known that
the two invocations will return the same value.
Cache the second field access in a local variable, which ensures
we access the property exactly once, which keeps eclipse happy.
In case the implementation LazyDataObjectModification, this has
the benefit of not deserializing twice.
Change-Id: Idfb8c631bc0267a63710c2cb650e6bda1c227df3
Signed-off-by: Robert Varga <robert.varga@pantheon.tech>
Matej Perina [Wed, 14 Nov 2018 11:11:16 +0000 (12:11 +0100)]
use multi thread executor in unit tests
- instantiatie AbstractConcurentDataBroker the way to use
CachedThreadPoolExecutor instead of DirectExecutor
- with new executor we are not in sync with .get() on transaction
future, so timers continuos verification with timeout is used
in tests where needed
JIRA: BGPCEP-847
Change-Id: I86cbe54a17e65d4ac7bdc7ca447ac58f345c473a
Signed-off-by: Matej Perina <matej.perina@pantheon.tech>
Claudio D. Gasparini [Mon, 12 Nov 2018 07:12:03 +0000 (08:12 +0100)]
BGP Stats collector
simplify synchronization of stats
JIRA: BGPCEP-846
Change-Id: I55d30f9cddb33006fb8d2519ddbe9f8b330de71b
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Sat, 10 Nov 2018 17:53:15 +0000 (18:53 +0100)]
PCEP Thread deadlock
Change-Id: I7b1f6fcf8c678c76750c501ce979813a20fa8313
JIRA: BGPCEP-845
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Fri, 9 Nov 2018 12:27:47 +0000 (13:27 +0100)]
Do not create empty route list container
to keep consistency with mdsal changes
Change-Id: Ibd58e7538fbefda96f96d7b3166553c3f8f27181
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Fri, 9 Nov 2018 15:03:11 +0000 (16:03 +0100)]
Encode with Base64 Linkstate route key
Change-Id: Ib9e9803d6c8287dffb97e9c89b5e975db60bc979
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Thu, 8 Nov 2018 14:05:02 +0000 (15:05 +0100)]
Update ObjectParser API
- return Object Class
- return Object Type
This will allow to simplify parser registration
and remove duplicated code.
Change-Id: I1f712f87ad0be190db45e77cc13f55ab6352c2a7
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Thu, 8 Nov 2018 16:36:53 +0000 (17:36 +0100)]
Fix application peer handling of DISAPPEARED / APPEARED case
JIRA: BGPCEP-843
Change-Id: I82368d8cf029ffe83391333c6e11633511688b69
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Tue, 6 Nov 2018 12:26:37 +0000 (13:26 +0100)]
Solve deadlock
JIRA: BGPCEP-841
Change-Id: I9daaa786e9d1b1cf7aa74141d346f1bd86b36452
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Thu, 1 Nov 2018 14:20:43 +0000 (15:20 +0100)]
Migrate bgp-benchmark to MD-SAL APIs
Change-Id: Ie6a1bfcd810524b08485537b6b2e54ceb6c2bf49
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>
Claudio D. Gasparini [Tue, 30 Oct 2018 21:03:38 +0000 (22:03 +0100)]
Migrate pcep to MD-SAL APIs
Change-Id: Id3e35a7636cd6b4b7e0d319a022ba588840de67c
Signed-off-by: Claudio D. Gasparini <claudio.gasparini@pantheon.tech>