zebra: only show vrf matched to a l3-vni in show vrf vni

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: pass ERR_STR_SZ in zebra_vxlan_process_vrf_vni_cmd

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: NUD_STALE is only valid in GNU linux

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: resolve memory leak in bm_master_init

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra/bgpd: cleanup l3vni on no advertise-all-vni

EVPN is only enabled when user configures advertise-all-vni.
All VNIs (L2 and L3) should be cleared upon removal of this config.

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: uninstall type-5 routes from vrf

When we receive an MP_UNREACH,
we try to uninstall routes from the VRF and the VNI.
The route-node in the VRF corresponds
to the ip prefix formed from EVPN prefix.
We should correctly form the prefix based on the EVPN route-type.

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: move vrf rd command under address-family l2vpn evpn

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: advertise/withdraw new added/deleted type-5 routes

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: process evpn type-5 routes received from peers

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: distinguish between frr prefixlen and packet prefixlen for EVPN type-5 routes

for EVPN routes prefixlen filed in struct prefix
represents the sizeof of the struct rather than the actual prefix len.
This is later used in looking up route node in RIB.

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: only advertise valid subnet routes as evpn type-5 routes

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: write advertise <ipv4|ipv6> unicast under bgp vrf config

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: follow AFI/SAFI style for advertising/withdrawing type-5 routes

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: update type-5 routes upon vrf export-rt change

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: update/withdraw type-5 routes upon l3-vni add/del

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: evpn enabled should only be checked for default instance

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: update type-5 routes when RD changes

when router-id/RD changes for a bgp vrf instance,
we need to update all type-5 routes.

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: advertise/withdraw type-5 routes upon user config

CLI config for enabling/disabling type-5 routes

router bgp <as> vrf <vrf>
  address-family l2vpn evpn
    [no] advertise <ipv4|ipv6|both>

loop through all the routes in VRF instance and advertise/withdraw
all ip routes as type-5 routes in default instance.

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: set vrf originator ip to kernels local-ip

For EVPN type-5 route the NH in the NLRI is set to the local tunnel ip.
This information has to be obtained from kernel notification.
We need to pass this info from zebra to bgp in l3vni call flow.
This patch doesn't handle the tunnel-ip change.

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: RD derivation for VRF

1. VRF RD can be auto-derived (simillar to RD for a VNI)
2. VRF RD can be configured manually through a config

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: move rd id bitfield to bgp_master

currently, we have a rd_id bitfield
to assign an unique index for auto RD.
This bitfield currently resides under struct bgp which seems wrong.
We need to shift this to a global space
as this ID space is really global per box.
One more reason to keep it at a global data structure is,
the ID space could be used by both VNIs and VRFs.

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: vrf to vni mapping command is only valid under vrf submode

Ticket: CM-18821
Review: Trivial
Testing: Manual

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: handle different sequence of bgp vrf create/delete

BGP VRF can be created/deleted either via config or via l3vni add/del.
We need to handle various sequences.

1. If user config is presented, an l3vni del should not delete the vrf instance
2. do not write bgp config in show running for auto created vrf
2. If l3vni present, disallow the cli for deleting bgp vrf instance
3. If l3vni is added and vrf config is present set the flags properly
4. if bgp vrf is configured unset the AUTO flag

Ticket: CM-18630
Review: CCR-6906
Testing: Manual

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: properly initialize ret variable

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: move vrf RT command under address-family l2vpn evpn

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: do not advertise ipv6 host routes with l3-vni related ext-comm

Currently, kernel doesn't support ipv6 routes with ipv4 nexthops.
To avoid the crash,
we will only attach l3-vni related
RTs/ecommunities only to ipv4 host routes.

Ticket: CM-18743
Review: ccr-6912
Testing: Manaul

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: Reinstall remote VTEP next hop when it becomes stale

When a remote VTEP next hop entry (for symmetric routing) becomes
stale, reinstall it. This makes the behavior the same as what is
done for remote host next hops (for asymmetric routing and ARP
suppression).

Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>

zebra: use list_delete_and_null instead of list_delete

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: adjust show bgp l2vpn evpn vni command to avoid sanity breakages

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: remove l3vni/rmac from gw macip APIs

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: only write vni config under vrf if the vni is valid

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: re-install remote next-hop neigh entries upon age out

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: write vrf import/export RT config to frr.conf

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: update all routes when vrf changes on a VNI

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: uninstall all nh/rmac when l3vni is deleted

Upon a l3vni delete (no vni under a vrf) is executed,
we should uninstall all the RMACs and NHs associated with the l3vni.
This is because by the time we get a route delete in zebra
l3vni is already deleted and we dont have refernce to RMACs and NHs

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show evpn rmac specific mac command

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show evpn rmac vni command

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show evpn rmac command

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show evpn specific next-hop command

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show evpn next-hop l3vni <vni>

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show evpn next-hop vni all

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show evpn l3vni vni command

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show evpn l3vni

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: json support for show vrf vni command

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: json support for show bgp vrf <> l3vni info

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: json support for show bgp l2vpn evpn vrf-import-rt

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: free host_list while deleting rmac/nh entry

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: set type-2 route flag if necessary in bgp_zebra_witgdraw

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: show commands for specific rmac/nh entries

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: proper refcounting for rmac/nh entries

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.comy>

bgpd: do not send label to zebra route add for evpn routes

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: only install mac_ip routes in vrf

Signed-off-by: Mitesh Kanjariya<mitesh@cumulusnetworks.com>

bgpd: use bgp_process while processing evpn routes in vrf

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: set evpn rvtep nexthops as active by default

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: import/unimport vrf routes upon l3vni change

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: program nh/rmac entries

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: program mac-ip routes in matching vrfs

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: import rt to vrf mapping

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: rmac ext comm

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: don't get rmac in remote macip delete

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: handle export rt change for vrf

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: add VRF export RTs to mac-ip routes

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: bgpevpn APIs to get l3vni/rmac and import/export RT list

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: link l2vnis(bgpevpn) to l3vni(vrf)

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra: don't get rmac in remote macip add

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: import/export rt for BGP vrf

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: l3vni/rmac association with bgp vrf

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

bgpd: Bgpevpn tenant vrf association

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

zebra, lib: zebra changes for symmetric routing support

Signed-off-by: Mitesh Kanjariya <mitesh@cumulusnetworks.com>

Merge pull request #1554 from opensourcerouting/cover-fix1

bgpd: fix some coverity scan issues

Merge pull request #1550 from opensourcerouting/missing-lintian-file

debianpkg: Add missing frr-dbg.lintian-overrides to Makefile

bgpd: handle argv_find_and_parse_afi return value

Handle the return value of argv_find_and_parse_afi() to avoid passing
along bad values.

Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>

bgpd: use buffer size instead of hardcoded value

This is a possible buffer overflow.

We should always use the buffer size (whenever possible) to tell
functions what the size of the buffer is, instead of a hardcoded value.

Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>

Merge pull request #1546 from pguibert6WIND/issue__1537

Issue  1537

Merge pull request #1527 from donaldsharp/zserv_thread

Zserv.h modifications

Merge pull request #1521 from donaldsharp/bufsiz

Cleanup Some Code Issues

Merge pull request #1536 from opensourcerouting/isis-l2conv

Fix ISIS L2 formations

debianpkg: Add missing frr-dbg.lintian-overrides to Makefile

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

isisd: fix l2 neighbor formations

Add a timestamp information for level 2 circuits, otherwise if the
circuit is marked as already processed on level 1 we will not process
level 2 areas.

Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>

zebra: remove useless iotcl linux calls

As netlink is available for all linux systems ( old linux distributions
are not considered), this commit removes the ipv6 ioctl support for
linux.

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

zebra: ipv6 addressing uses netlink socket instead of standard ioctl

It is possible to configure IPv6 addresses from interfaces by using
netlink socket, intead of using standard sockets.

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

zebra: Make zserv.h the true api for zapi northbound

zserv.c had a grab bag of function declarations that
did not belong in it.  Move those to where they better
belong.

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

zebra: Try to move non zapi cli commands from zserv.c

zserv.c has become a bit of a dumping ground for zebra cli.
I'd like to focus the zserv.c code into it's core functionality
which is handling the zapi interface.

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

zebra: Move zebrad initialization outside of cli init

The zebrad initialization does not need to be part of cli
initialization and should be done separately.

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

Merge pull request #1478 from bingen/zeromq4

lib: Address ZMQ lib TODOs

Merge pull request #1540 from opensourcerouting/isis-spfperf1

isisd: save a clock_gettime() call

Merge pull request #1539 from LabNConsulting/working/master/community-decisions

COMMUNITY.md: add paragraph on use of development list and discussing…

Merge pull request #1514 from donaldsharp/watchfrr

tools, watchfrr: Modify timeout to 90 seconds

isisd: save a clock_gettime() call

Use the thread cached clock to use as start time. It will save a call to
clock_gettime() and also provide a more 'accurate' time measurement from
the start of the procedure.

COMMUNITY.md: add paragraph on use of development list and discussing/documenting decisions

isisd: fix debug message

Show the actual timer instead of doing the arithmetic operation again.

Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>

Merge pull request #1526 from chiragshah6/ospfv3_dev

ospf6d: Fix multi nexthop route remove

bgpd: Fixup buffer sizes for prefix_rd2str

The prefix_rd2str uses a buffer size of RD_ADDRSTRLEN.
Modify the code to consistently use this for all of BGP.

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

bgpd: Modify prefix_rd2str to return "Unknown" when unknown

Make prefix_rd2str return an "Unknown" string when something
goes wrong.  This will allow for simplification of the
code that uses prefix_rd2str.

Additionally ensure that size is big enough with an assert.

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

Merge pull request #1524 from dslicenc/zebra-ra-display-cm18702

zebra: do not display ipv6 ra commands created by bgpd

bgpd: Cleanup SA error in ignoring return from function

Ignoring the return from argv_find_and_parse_afi
makes the SA system assume that you could pass a AFI_MAX
value to bgp_show_route.  Which in turn would cause
an array out of bounds read.

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

eigrpd: Fixup some SA issues

1) strlen(buf) on an uninted value is the wrong thing to do
we should be getting sizeof(buf)

2) tlv is not freed in this error case, so let's free it up.

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

bgpd: Fix prefix2str using BUFSIZ to PREFIX_STRLEN

PREFIX_STRLEN is the correct length for buffers needed to output
a prefix2str.  Additionally cleanup some setting of the last
value to a `\0` this is handled by prefix2str.

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

Merge pull request #1531 from chiragshah6/ospf_vrf_dev

ospfd: prevent passive interface cmd crash

Merge pull request #1528 from donaldsharp/ldp_macro

ldpd: Switch over to new debug style