pimd: Fix size_t zlog_err format string warning

* fc1c114aa / "pimd: Fix warning", the size_t arg should have a %zu format.

pimd: Cleanup interface startup

This patch cleans up some interface startup, removes duplicate
debug messages and protects against some always being displayed.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Notice when we receive a packet type we can't handle yet

There are PIM packet types that have not been implemented yet.
Notice when we get one of those and safely do nothing.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add knowledge of different packet types

There are several additional packet types that pimd is unaware of
Add code to allow pim to understand them in the future.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup zebra debugs to be protected by debug commands

pimd is very chatty without any pim debugs turned on.
This commit fixes a bunch of the debugs to be protected
by appropriate pim debug statement.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Do not display some default values as part of a show run

There is no need to display igmp default values for the query-interval
and the query-max-response-time-dsec

Before change:

!
interface swp4
 description swp4 -> host-212's swp1
 ip igmp
 ip igmp query-interval 125
 ip igmp query-max-response-time-dsec 100
 ip pim ssm
 link-detect
!

After change:
!
interface br1
 ip igmp
 ip pim ssm
 link-detect
!

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix warning

Fix long unsigned / unsigned mixup

Signed-off-by: Martin Winter <mwinter@opensourcerouting.org>

pimd: use IPPROTO_IP (not SOL_IP) for IP_PKTINFO

Solaris uses the same socket API for IP_PKTINFO as Linux, but doesn't
have a SOL_IP define.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

pimd: Fix compile warning (error with Werror) on BSD

* ioctl commands can vary in type between systems, cast to an unsigned long
  before passing to format command.

pimd: Remove stdout zlog changes

Quagga start has unnecessary pim messages to stdout.  Causing some confusion.
Logging is already saving this information in pimd.log

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix leaked fd

When caling pim_sock_open if the failure cause happens, however
unlikely, don't leak the fd on failure.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

build: Enable vtysh and pimd as part of default build

The default build needs vtysh and pimd as a default build so
that when a change is made we can catch build issues before they
become a problem.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

bgpd/pimd: fix zAPI parsing

Commit c99f3481a598 has changed the API. Now, the vrfid has been added in
the header, thus we must read it before parsing the rest of the message.

To ease code maintenance, let's add a new function to read a zAPI header.

Fixes: c99f3481a598 ("*: add VRF ID in the API message header")
Reported-by: Martin Winter <mwinter@opensourcerouting.org>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Donald Sharp <sharpd@cumulusnetworks.com>
Tested-by: Martin Winter <mwinter@opensourcerouting.org>

pimd: Create ability to modify hell and hold timers per interface

Create new per interface command:
'ip pim hello <1-180> {<1-180>}'

The first number is the Hello Timer for this interface
The second number is the Hold Timer to pass to neighbors on this interface.
The second number is optional.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Change ioctl call failure from vty_out to zlog_warn

The command 'show ip multicast' when run, iterates over all interfaces
on the router, if you do not have pim configured on that interface
it would generate an error message:

Interface Address         ifi Vif  PktsIn PktsOut    BytesIn   BytesOut
br1       20.0.15.1         7   7       0       0          0          0
swp1      169.254.0.10      3   3       0       0          0          0
swp2      169.254.0.26      4   4       0       0          0          0
ioctl(SIOCGETVIFCNT=35296) failure for interface swp3 vif_index=-1: errno=22: Invalid argument
ioctl(SIOCGETVIFCNT=35296) failure for interface swp4 vif_index=-1: errno=22: Invalid argument

This fixes the issue to display this instead:

Interface Address         ifi Vif  PktsIn PktsOut    BytesIn   BytesOut
br1       20.0.15.1         7   7       0       0          0          0
swp1      169.254.0.10      3   3       0       0          0          0
swp2      169.254.0.26      4   4       0       0          0          0
swp3      0.0.0.0           5  -1       0       0          0          0
swp4      0.0.0.0           6  -1       0       0          0          0

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Ensure new generation_id is different from previous

The RFC states that an interfaces generation_id must be changed
if it experiences an if down.  From 4.3.1:

The GenID option contains a randomly generated
32-bit value that is regenerated each time PIM forwarding is started
or restarted on the interface, including when the router itself
restarts.

Since we are only grabbing a new generation_id without comparing
it to the previous generation_id, it is possible that random
can generate the exact same number.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Stop DR election on every hello

The DR election is occurring on every hello received.
This is because the hello receive packet returns a 0
for any successfully received packet.  PIMD then looked
at the 0 returned and did a DR election.

Code was inspected for the cases where DR should happen:
(A) Interface ip address change
(B) DR priority in hello packet changes
(C) Received a new neighbor on an interface
(D) Neighbor timer pops.

Each of these initiate a DR election in the code currently.

Testing was initiated on a pim network:
tor-11# show ip pim designated-router
NonPri: Number of neighbors missing DR Priority hello option

Interface Address         DR              Uptime   Elections Changes NonPri
br1       20.0.15.1       20.0.15.1       00:08:16         1       1      0
swp1      169.254.0.10    169.254.0.10    00:08:16         2       1      0
swp2      169.254.0.26    169.254.0.26    00:08:16         2       1      0
tor-11#

As you can see Elections == 2.  This is because pimd performs
an election on (A) and (C) above.  I see no need to modify
(A) to check if we have any knowledge of the interface before
this call.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix first DR Election

The function pim_neighbor_new caused the dr election before
the addition of the neighbor to the pim_ifp->pim_neighbor_list,
in pim_neighbor_add, which is needed to properly elect the
correct DR.

Move the dr_election and hello trigger till after the addition
to the list.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add ability to set DR Priority for an interface

From RFC 4601, Section 4.3.1:

The DR_Priority Option allows a network administrator to give
preference to a particular router in the DR election process by
giving it a numerically larger DR Priority.  The DR_Priority Option
SHOULD be included in every Hello message, even if no DR Priority is
explicitly configured on that interface.  This is necessary because
priority-based DR election is only enabled when all neighbors on an
interface advertise that they are capable of using the DR_Priority
Option.  The default priority is 1.

This modification allows the setting of the DR Priority on a per
interface basis.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: add support for configuring multicast static routes

Hi,

This patch adds the ability to configure multicast static routes
directly into pimd. Two source files are introduced to implement the new
feature in addition to changes to existing files.

Here is how it can be used the CLI:

interface <incoming interface>
ip mroute <outgoing interface> <group addr>                          #
for asm
or ip mroute <outgoing interface> <group addr> <source>    # for ssm

Please let me know if you have any questions or concerns,

Regards,
Jafar

Acked-by: Donald Sharp <sharpd@cumulusnetworks.com>

Removal of 'show ip mroute'

The 'show ip mroute' in zebra_vty.c collided with the
'show ip mroute' command in pim_cmd.c.  The 'show ip rpf'
command is functionally equivalent to the zebra_vty.c
'show ip mroute'.  Therefore remove the 'show ip mroute' command
in zebra_vty.c.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pim: 'show debugging' collision

The 'show debugging' cli as setup by pim collided with 'show debugging isis'.
Fix that and clean up cli help commands to actually display correctly.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

Fixup of redhat control files to properly start pimd

pimd assert when no route to source from a new igmp join

When pim_upstream_new is called the code looks up the nexthop.
If there is no route to the source, the code silently ignored
the error returned.  When the nexthop lookup fails don't create
the 'struct pim_stream *' to return.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pim_mroute.h has a different version of code than linux/mroute.h provides

linux/mroutes.h and pim_mroute.h both have copies of the same structures.
This is causing failures in setsockopt(..., MRT_ADD_MFC,...) because
of data structure incompatibilities between the kernel and what
pim_mroute.h was providing.  Modify the code to check for mroute.h
and include it if necessary.  I did not modify the non linux/mroute.h
path because I do not have other systems to test on easily.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

PIMD: Fix code to use srandom/random

pimd rolled it's own solution to random #'s, that was not
terribly random.  Rely on the underlying system to generate
random #'s for us

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

*: assorted warning fixes

A few warnings slipped through the cracks...

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

pimd: fix out of tree build

pimd/Makefile.am was missing srcdir/lib from its include paths, breaking
out-of-tree build regarding route_types.h

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

pimd: cast ioctl values when printing them

ioctl values might be int or long, cast them to unsigned long for
consistent printing.  (They're long on FreeBSD, but were printed with
%d.)

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

pimd: cast to sockaddr_in to sockaddr

While glibc seems to have something in the system headers that prevents
this from triggering a warning, FreeBSD doesn't.  Fix the warning.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

pimd: Rework pim to handle changes between us and upstream

The code in pimd needs to be updated to handle the changes
we've made to api's that it depends on.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: mask unused zclient_broken()

this function is used by the currently not present zclient reconnect
code.  It'll be unmasked again when that code hits master.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

build: enable pimd in test script

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

pimd: merge pimd as of 2015-01-19

Welcome pimd to the Quagga daemon zoo!

This is a merge of commit 77ae369 ("pimd: Log ifindex found for an
interface when zebra lib reports a new connected address."), with
the intermediate "reconnect" changes removed (c9adf00...d274381).
d274381 is replaced with b162ab7, which includes some changes.  In
addition, 4 reconnect-related changes and 1 cosmetic one have been
bumped out.

The rebase command used to produce the branch that is merged here is:
  git rebase --onto b162ab7 c9adf00 77ae369

Note that 3 patches had their author rewritten from
    "Anonymous SR#108542 <>" (which is not a valid git author ID)
to: "Savannah SR#108542 <nbahr@atcorp.com>" (which is the e-mail address
                               listed in the associated Savannah ticket)

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

doc: explain rpf lookup default mode

Reported-by: Alexis Rosen <quagga-users@alexis.users.panix.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

doc: zebra multicast RIB commands

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: mark multicast commands experimental

depending on feedback from actually having these commands in a released
version, we may want to adjust them.  Thus, mark them as experimental so
users are aware of this.

Cc: Everton Marques <everton.marques@gmail.com>
Cc: Balaji G <balajig81@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: add "show ip rpf" to get result of RPF lookup

Checking what route exactly a RPF lookup for a given source uses is
essential for an administrator to debug multicast routing issues.  This
command provides exactly that, using the multicst RPF lookup function
and printing out its result to the CLI.

Cc: Everton Marques <everton.marques@gmail.com>
Cc: Balaji G <balajig81@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: make MRIB lookup behaviour switchable

depending on the usage scenario (and availability of multitopology IGP
protocols, which is currently zero in Quagga), different approaches of
Multicast RPF lookups are useful.

Reference behaviours from commercial vendors are urib-only/mrib-only
(Juniper, depending on inet.2 availability) and lowest-distance (Cisco).
As we are currently without MT IGP support, mrib-first seems the most
useful default for Quagga.

Cc: Everton Marques <everton.marques@gmail.com>
Cc: Balaji G <balajig81@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: return route_node from rib_match_ipv4_safi

The multicast code needs to know the route_node in addition to the rib
entry in order to perform distance or prefix-length comparisons.  Add it
as optional "out" pointer parameter.

Cc: Everton Marques <everton.marques@gmail.com>
Cc: Balaji G <balajig81@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: dummy kernel "install" multicast routes

This is a followup to 9511633 ("zebra: MBGP routes should not be
installed in the kernel"), which was correct in disabling MRIB routes
being installed in the kernel, yet broke the MRIB since now routes were
never marked as active.

Hence, push down the check into the kernel install functions, so that
the routes are still marked active.  At the same time, the FPM calls get
a check each since otherwise we'd bump the FPM interface on MRIB
updates.

Fixes: 9511633 ("zebra: MBGP routes should not be installed in the kernel")
Cc: Everton Marques <everton.marques@gmail.com>
Cc: Balaji G <balajig81@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: fix optional distance on static mrib route

Unfortunately, the quagga CLI parser doesn't support [<1-255>].  Fix by
working around with an alias.

Replaces the following commits:
- zebra: mrib: [no] ip mroute - require distance.
- zebra: mrib: [no] ip mroute - make distance optional.
(Rewritten as alias)

Cc: Everton Marques <everton.marques@gmail.com>
Cc: Balaji G <balajig81@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: mrib: static route support

With the MRIB being independent from the Unicast RIB, there's currently
now way to add static routes to the MRIB.  Address that by adding a
separate set of commands for MRIB static routes.

Combines these original patches:
- zebra: mrib: ip mroute command to add unicast route to MRIB for multicast RPF.
- zebra: mrib: no ip mroute: Fix removal of static multicast RPF route.
- zebra: mrib: remove unused static_add/delete_ipv4
- zebra: Cleanups to zebra_rib.
- pimd: Merge pim-only branch.

Cc: Everton Marques <everton.marques@gmail.com>
Cc: Balaji G <balajig81@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: add ZEBRA_IPV4_NEXTHOP_LOOKUP_MRIB

This adds a new zapi call "ZEBRA_IPV4_NEXTHOP_LOOKUP_MRIB" performing a
Multicast RPF lookup for a given source.  Details of the lookup
behaviour are left to the zebra side of things.

Note: this is non-reactive, as in, only delivers a snapshot of the state
at a particular point in time.  There's no push notification of changes
happening to the RIB.

This combines the following 3 original patches:
- zebra: add zsend_ipv4_nexthop_lookup_mrib()
- zserv: Query mrib (SAFI_MULTICAST).
- zebra: Cleanups to zebra_rib.

Cc: Everton Marques <everton.marques@gmail.com>
Cc: Balaji G <balajig81@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: add rib_match_ipv4_safi()

This is the same as rib_lookup_ipv4(), without the SAFI hardcoded.

Cc: Balaji G <balajig81@gmail.com>
Cc: Everton Marques <everton.marques@gmail.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: identify MRIB on debug messages

since the same code handles both URIB and MRIB, the debug messages can
get rather confusing if the RIB isn't identified.  Mark the MRIB in
debug messages so we can distinguish that.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: factor out rib debug logs

Introduces a logging function that takes a struct route_node * argument,
and prefixes log output with that node's prefix.  While this removes
some duplication, it will also later be useful for srcdest route nodes.

Behaviour before and after the patch should be exactly identical.

Signed-off-by: David Lamparter <equinox@opensourcerouting.org>

zebra: Ensure IPv6 RA configuration is displayed correctly

IPv6 RAs on an interface can be enabled either by the operator or internally
due to the configuration of a BGP unnumbered neighbor. Ensure that this is
displayed in the configuration correctly.

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-11076
Reviewed By: CCR-4770
Testing Done: Manual; also verified by defect submitter.

cumulus: Increase netlink socket buffer size

Ticket: CM-10950
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>

Quagga: Make sure order of route-maps in list and hash table matches

Quick create/delete actions on a route-map can result in the same route-map
entity having multiple entries created for it — because BGP hasn't run the
update processing to complete prior delete action. The route-map is present
in both a hash table as well as a linked list and the order in each is
different. This can lead to problems when the BGP route-map update processing
runs and finds the same route-map entity present for deletion multiple times.
For example, while processing instance-2 of rmap-A, the code may end up
freeing the hash bucket corresponding to instance-1 of rmap-A.

The fix works by ensuring the list is ordered the same way as the hash
buckets.

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Shrijeet Mukherjee <shm@cumulusnetworks.com>
Ticket: CM-10023
Reviewed By: CCR-4747
Testing Done: manual, bgp-smoke

lib: Compiler warning fix

There exists a code path where fld would
be used uninitialized.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

vtysh: fix vtysh warnings from another compiler

Remove some dead code and fix char `\0` being
used as NULL.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

ospfd: Fix compiler warnings

Two issues:

1) nbr->oi->ifp->name is an array it would
always evaluate to true.

2) There exist a code path where addr
would be used without initialization.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

isisd: Remove dead case

It's impossible to ever get to the else statement.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

lib: Fix some more compiler warnings

Three different changes:
a) in nexthop.c we have an actual bug, we've rearranged
the enumerated types and the list of enums has changed, we
were displaying the wrong nexthop information.

b) Remove some dead code from sockunion.h

c) '\0' evaluates to 0 which is NULL.  vector_set takes
a pointer.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

bgpd: Fix more clang compiler warnings

Remove some dead code and fix initialization of the
sockunion.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

bgpd: Fix warnings -> errors issues

Switched to !another! new compiler and a few more
warnings creeped in.  Also noticed a json issue
that caused an error in output.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

bgpd: Fix possible api.tag uninitialized usage

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

lib: Fix possible uninitialized pointer usage.

Signed-off-by: Donald Sharp <sharp@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

bgpd: Clean up some static analysis warnings

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

zebra: Clean up some small static analysis warnings

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

lib: Clean up some small static analysis warnings

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

ospf6d: Fix printf % code usage

Use correct % code in printf statement.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

bgpd: Fix function return

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

isisd: Use correct boolean operator type.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

quagga: Remove WEAK_ALIAS pragma usage

The code was using pragma's to override function calls
while there was code for platforms that don't support pragma's
Just remove the pragma usage from the program in it's entirety

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

bgpd: peer_uptime overflows after 1 year

* bgpd.c: (peer_uptime) Wraps after 1 year, and doesn't indicate years.
  Fix. Assume a year is 365 days, for an easy life.

Fixes: Bug #836
Reported-by: Rolf Hanßen
Acked-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

bgpd: Refactor peer_uptime

peer_uptime was doing the same work unnecessarily.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

OSPF LSA may not be refreshed within 1800s, ANVL OSPF test 12.4 fails

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-10948

The previous formula for calculating delay was:

    delay = OSPF_LS_REFRESH_TIME - LS_AGE (lsa) - OSPF_LS_REFRESH_JITTER
              + (random () % (2*OSPF_LS_REFRESH_JITTER));

If LS_AGE() is 0 we get:

  delay = 1800 - 0 - 60 + (random() % 2*60)

So delay will be 1740 + 0..120, so we could easily pick a delay that is
greater than OSPF_LS_REFRESH_TIME (1800).  For the normal case where
we hit this else the LS_AGE will be > 0 so that reduces the chances of
delay being greater than 1800.

BGP: "show ip bgp neighbor json" keys are stepping all over each other

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-11007

bgp_show_neighbor() was allocating a json_neigh struct and using that
for every single peer...so the resulting json was a munge of state from
all peers

We need to allocate a new json_neigh for each neighbor.

quagga reload failure (Tested on VX)

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-10974

ctx_keys needs to be a tuple and not a list but there was one spot
where I created a tmp copy of the ctx_keys but failed to convert it to
tuple

BGP: Unlink BGP instance from VRF only at the end of deletion

Additional change missed out in earlier commit. Updates
d3f5a0d3f51a8a7f3f02c38bd4c1c6870b4b2a5b.

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Ticket: CM-10930
Reviewed By: CCR-4717
Testing Done: Manual, bgp-smoke

Merge branch 'cmaster' of ssh://stash.cumulusnetworks.com:7999/quag/quagga into cmaster

BGP: Unlink BGP instance from VRF only at the end of deletion

When a BGP instance including the default instance is deleted, it needs to be
unlinked from the corresponding VRF structure. However, instance deletion does
not happen in one shot but needs a lot of threads to run - peer event handling,
route processing etc. - before it can complete. Premature unlinking of the
instance from underlying VRF would result in BGP routes not being deleted from
the zebra RIB.

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-10930
Reviewed By: CCR-4717
Testing Done: Manual, bgp-smoke

OSPF ANVL failures in 3.0

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-10088

When we pulled this patch from upstream there was one spot where we
incorrectly used min_ls_interval (5s) instead of min_ls_arrival (1s).
commit ac7424f9dec0d49d891e0b457afa8e31c2b854fe
Author: Michael Rossberg <michael.rossberg@tu-ilmenau.de>
Date:   Mon Jul 27 21:05:44 2015 +0200

    configuration options for faster OSPF convergence

    Allow configuration of faster OSPF convergence via the
    min_ls_interval and min_ls_arrival timer lengths.

    This patch was originated by Michael, and cross-ported
    to Cumulus's Quagga.

zebra: Fix show ip route for ibgp only routes

When 'show ip route summ' is entered
and there are only ibgp routes they
are not being displayed.  This commit
fixes this issue.

Ticket: CM-10931
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

OSPF refreshTimerMsecs in json output is off

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-10960

The timer is in seconds and was being divided by 1000 instead of
multiplying by 1000 when converting to ms.

Merge branch 'cmaster' of ssh://stash.cumulusnetworks.com:7999/quag/quagga into cmaster

zebra: Accept and process RAs with lifetime of zero

Fix code to not discard received RAs with a lifetime of 0. The router lifetime
is only applicable for default router processing which is not relevant here.
For the purposes of BGP unnumbered, the neighbor should be learnt without
consideration of the value of router lifetime in received RA.

Note: This patch brings in a portion of the earlier commit
690baa53592320dddee5c729f959150cc9a72699 - this included some additional
changes which have been reverted.

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>
Ticket: CM-10943
Reviewed By: CCR-4611
Testing Done: bgp-smoke

lib: Fix crash in 'no vrf' command

If a routing protocol does not have a vrf
configed and the vrf happens to be down
a 'no vrf X' line will cause the system
to crash.

This fixes the issue, I do believe though
that we need to revisit the issue and re-think
start/stop/config/unconfig of vrf's a bit more.

Ticket: CM-10952
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Signed-off-by: Dinesh Dutt <ddutt@cumulusnetworks.com>

Revert "Zebra: Update/fix router_lifetime in IPv6 RAs"

This reverts commit 690baa53592320dddee5c729f959150cc9a72699.

Making the router lifetime in the IPv6 RAs as 0 by default would break BGP
unnumbered when this version of Quagga goes up against a 2.5.x Quagga. This
is because of a defect in the Quagga code that ignores any received RAs with
a lifetime of 0.

zebra: Add route-map support to ip import-table

Added the ability to supply a route-map to the ip import-table command,
which greatly improves filtering between the kernel prefixes in a
non-default table that are imported into the zebra rib.

Ticket:CM-8168
Signed-off-by: Donald Slice
Reviewed By: Donald Sharp

BGP: Set advertisement interval when triggering IPv6 RAs

This change extends the earlier change which added the ability in BGP to
trigger IPv6 Router Advertisements when an unnumbered neighbor is configured.
In addition to triggering the RAs, the advertisement interval is also set to
10 seconds. This is needed to handle the scenario where the peer may start
later.

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-10896
Reviewed By: CCR-4693
Testing Done: Manual, bgp-min, bgp-smoke

Fix unprotected debugs to warns and fix

In the case of a route replace failing we
saw two issues with the logging:

1) The route replace was a debug instead of a warn
  -> In this case change code to zlog_warn
2) The buf in the route replace was not being initialized
because buf initialization was protected by a debug check.
  -> In this case move the buf initialization to inside
the failure case.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>

Fix non initialized usage of data in zebra_rnh.c

In zebra_deregister_rnh_static_nexthops the nh_p
structure was not being properly initialized for
all the cases that we could be storing a nexthop
for.  This was causing code later to retrieve
the table from an nh_p->family which was garbage.

In the case of BLACKHOLE and Ifindex based routes
do nothing because they shouldn't be a nexthop
considered for NHT.

==2239== Conditional jump or move depends on uninitialised value(s)
==2239== at 0x4E5F6CE: family2afi (prefix.c:217)
==2239== by 0x155F7C: get_rnh_table (zebra_rnh.c:83)
==2239== by 0x156194: zebra_lookup_rnh (zebra_rnh.c:148)
==2239== by 0x15655E: zebra_deregister_rnh_static_nh (zebra_rnh.c:242)
==2239== by 0x156681: zebra_deregister_rnh_static_nexthops
(zebra_rnh.c:280)
==2239== by 0x12F3DF: rib_unlink (zebra_rib.c:2210)
==2239== by 0x12E9CE: rib_process (zebra_rib.c:1843)
==2239== by 0x12EA8A: process_subq (zebra_rib.c:1873)
==2239== by 0x12ECAF: meta_queue_process (zebra_rib.c:1936)
==2239== by 0x4E89625: work_queue_run (workqueue.c:298)
==2239== by 0x4E63230: thread_call (thread.c:1577)
==2239== by 0x125830: main (main.c:432)
==2239==
==2239== Conditional jump or move depends on uninitialised value(s)
==2239== at 0x4E5F6DB: family2afi (prefix.c:220)
==2239== by 0x155F7C: get_rnh_table (zebra_rnh.c:83)
==2239== by 0x156194: zebra_lookup_rnh (zebra_rnh.c:148)
==2239== by 0x15655E: zebra_deregister_rnh_static_nh (zebra_rnh.c:242)
==2239== by 0x156681: zebra_deregister_rnh_static_nexthops
(zebra_rnh.c:280)
==2239== by 0x12F3DF: rib_unlink (zebra_rib.c:2210)
==2239== by 0x12E9CE: rib_process (zebra_rib.c:1843)
==2239== by 0x12EA8A: process_subq (zebra_rib.c:1873)
==2239== by 0x12ECAF: meta_queue_process (zebra_rib.c:1936)
==2239== by 0x4E89625: work_queue_run (workqueue.c:298)
==2239== by 0x4E63230: thread_call (thread.c:1577)
==2239== by 0x125830: main (main.c:432)

Ticket: CM-10667
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Shrijeet Mukherjee <shm@cumulusnetworks.com>

lib: Fix connected lookup

When looking up the connected route, the delete was
causing crashes in OSPF due to the oi having copies
of the freshly deleted connected interface.  Fix
code to first lookup the connected route and use that
instead of just deleting it.

Valgrind Findings:

==24112== Invalid read of size 1
==24112== at 0x4E8283F: ospf_intra_add_stub (ospf_route.c:614)
==24112== by 0x4E80B15: ospf_spf_process_stubs (ospf_spf.c:1064)
==24112== by 0x4E80F74: ospf_spf_calculate (ospf_spf.c:1269)
==24112== by 0x4E811C9: ospf_spf_calculate_timer (ospf_spf.c:1339)
==24112== by 0x5126230: thread_call (thread.c:1577)
==24112== by 0x401E00: main (ospf_main.c:377)
==24112== Address 0x7f56a09 is 9 bytes inside a block of size 40 free'd
==24112== at 0x4C29E90: free (in
/usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==24112== by 0x51290B3: zfree (memory.c:132)
==24112== by 0x51287F0: connected_free (if.c:987)
==24112== by 0x514406A: zebra_interface_address_read (zclient.c:1146)
==24112== by 0x4E5A81C: ospf_interface_address_add (ospf_zebra.c:262)
==24112== by 0x5144838: zclient_read (zclient.c:1397)
==24112== by 0x5126230: thread_call (thread.c:1577)
==24112== by 0x401E00: main (ospf_main.c:377)

Ticket: CM-10890
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

lib: refactor connected_lookup_prefix

The connected_lookup_address function should really
be a connected_lookup_prefix function.  Refactor
the code to use it.

Ticket: CM-10890
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>

"ip protocol" is missing ospf as an option

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Ticket: CM-10849

We were incorrectly listing protocols that supported ipv6 for "ip
protocol" instead of listing the protocols that supported ipv4.

With the fix

superm-redxp-05(config)# ip protocol ?
  any        Any of the above protocols
  bgp        Border Gateway Protocol (BGP)
  connected  Connected routes (directly attached subnet or host)
  isis       Intermediate System to Intermediate System (IS-IS)
  kernel     Kernel routes (not installed via the zebra RIB)
  ospf       Open Shortest Path First (OSPFv2)
  rip        Routing Information Protocol (RIP)
  static     Statically configured routes
  table      Non-main Kernel Routing Table
superm-redxp-05(config)#
superm-redxp-05(config)#
superm-redxp-05(config)# ipv6 protocol ?
  any        Any of the above protocols
  bgp        Border Gateway Protocol (BGP)
  connected  Connected routes (directly attached subnet or host)
  isis       Intermediate System to Intermediate System (IS-IS)
  kernel     Kernel routes (not installed via the zebra RIB)
  ospf6      Open Shortest Path First (IPv6) (OSPFv3)
  ripng      Routing Information Protocol next-generation (IPv6) (RIPng)
  static     Statically configured routes
  table      Non-main Kernel Routing Table
superm-redxp-05(config)#
superm-redxp-05(config)#

zebra: Fix interface based static routes

This commit fixes interface based static routes.

static routes are now stored and if an interface
comes up it finds the route and installs it.

Ticket: CM-10869
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>

debian: Fix changelog

Fixup Changelog

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

BGP: Fix nexthop setting for default originate

The nexthop setting for IPv6 default origination was incorrect, resulting in
the same (incorrect) IPv6 link-local nexthop being advertised to multiple
peers. The issue was also present for IPv4 default origination when IPv4
routes are advertised with IPv6 nextops (RFC 5549).

For default origination, we should just reset the nexthop when forming the
update for the update-group and let the outbound update/nexthop setting code
set it correctly.

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>
Ticket: CM-10623
Reviewed By: CCR-4677
Testing Done: Manual, bgp-min

Quagga: Check and cleanup prior address when processing address add on interface

IPv4 addresses can be learnt from the kernel even when the interface is down.
Quagga notifis clients of addresses upon learning of them (initial read or
upon NEWADDR) as well as when the interface comes up. The problem is that
while zebra code itself has checks to ensure duplicate addresses aren't added
to an interface, that is not true for the clients.

This patch checks for duplicates when the client receives the address add.
Upon this, the patch does a delete and add as done in zebra - because there
is a possibility that some other parameter could have changed. The fix also
takes care of the extra memory allocation for 'connected' in clients.

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Dinesh Dutt <ddutt@cumulusnetworks.com>
Ticket: CM-10745
Reviewed By: CCR-4660
Testing Done: Manual, bgp-min

zebra - BFD client de-registration support

CM-10680
Issue: When BGP daemon is stopped, all the BGP BFD sessions are not getting deleted from PTM.
Root cause: BGP daemon stop causes BFD de-register message to be sent for every peer on which BFD is enabled. But, all the de-register messages from bgpd to zebra are not processed before the socket close. This results in some stale BGP BFD sessions.
Fix: Support for client de-register message has been added in PTM/BFD.  Changes in Quagga to support BFD client de-registrations:
−   The BFD clients de-registration is sent directly from zebra daemon when zebra client (bgpd, ospfd and ospf6d) socket close is detected.
−   Introduced a BFD flag for the zebra clients to prevent BFD de-registration messages from being sent to zebra daemon when the client is shutting down. This reduces the BFD messaging.

CM-10540
Issue: Invalid ptm status “fail” instead of “n/a” being displayed for VRF interfaces.
Root cause: ptm status is not being initialized to “unknown” status when VRF interface is added or changed. The uninitialized value is ‘0’ which is the value for “fail”
Fix: Initialized the ptm status to the correct value.

Signed-off-by: Radhika Mahankali <radhika@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Kanna Rajagopal <kanna@cumulusnetworks.com>
Ticket: CM-10680, CM-10540
Reviewed By: CCR-4653
Testing Done: PTM smoke, BGP smoke and ptmd_test.py:TestMultipleAddrsIntfOspfBgp

BGP: At exit, remove callbacks before invoking vrf_terminate()

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Daniel Walton <dwalton@cumulusnetworks.com>
Ticket: CM-10831
Reviewed By: CCR-4657
Testing Done: Manual

debian: Update changelog

Fix the changelog to show what we are releasing

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

lib: Vrf disable hook was never being run

The check for the fact a vrf was enabled
was immediately after we disabled it in vrf_disable()

Ticket: CM-10139
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>

zebra: fix show ip route vrf to use zvrf_list

Iterate over the zvrf_list to allow 'show ip route vrf...'
to show vrfs that are inactive -vs- not even configured

Ticket: CM-10139
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>

lib: Swap show run ordering of interfaces and vrf's

When doing a 'show run' display vrf information
first to allow later commands to just
work correctly for vrf cli read in.

Ticket: CM-10139
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>

zebra: Fix show vrf and show run to use the zvrf_list

The show vrf and show run commands were iterating
over the vrf_list.  Use the zvrf_list instead
so that we can understand the differences
between something configed -vs- created

Ticket: CM-10139
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>

zebra: Save the zvrf in a zvrf_list

The zebra vrf needs to be saved in a
zvrf_list so that we can tell when
things start/stop correctly

Ticket: CM-10139
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
Signed-off-by: Donald Sharp <sharpd@cumulustnetworks.com>
Reviewed-by: Vivek Venkatraman <vivek@cumulusnetworks.com>