Merge remote-tracking branch 'origin/stable/2.0'

Merge pull request #665 from opensourcerouting/rpm-postun-fix

redhat/frr.spec.in: Fix postun script for Fedora.

redhat/frr.spec.in: Fix postun script for Fedora.

Fedora uninstall showed some non-fatal postun script error because of an empty if which only applied to other distro's

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

Merge pull request #656 from qlyoung/fix-varhandler-reachable

lib: free varhandlers on exit

Merge pull request #657 from donaldsharp/rompapotamus

Rompapotamus

Merge pull request #663 from chiragshah6/pim_dev

ospfd: Fix ospf timers commands

Merge pull request #646 from qlyoung/fix-ospf-timers

Fix ospf timers

Merge pull request #664 from opensourcerouting/ldpd-fixes

Ldpd fixes/cleanup

ospfd: fix clear ip ospf interface cmd

clear ip ospf interface <IFNAME> parsing issue,
instead of arguement of interface "IFNAME" parsed.

Signed-off-by: Chirag Shah <chirag@cumulusnetworks.com>

Merge pull request #654 from qlyoung/ignore-compile-commands

frr: ignore 'compile_commands.json'

Merge pull request #647 from qlyoung/zebra-mpls-error-messages

zebra: more informative error messages & help strings for mpls labels

Merge pull request #652 from donaldsharp/miscapotamus

Miscapotamus

Merge pull request #650 from donaldsharp/zebrafication

zebra: Refactor 'struct rib' to be 'struct route_entry'

ldpd: further improvements in the configuration handling

This patch attempts to make the code easier to read by removing a
global variable and changing a few other things. Also, ldpd now calls
merge_config() only after reading the whole initial configuration at
startup, instead of doing that for every command in the configuration
file.

Signed-off-by: Renato Westphal <renato@opensourcerouting.org>

ldpd: remove variable that is not necessary anymore

Signed-off-by: Renato Westphal <renato@opensourcerouting.org>

ldpd: fix bug in pseudowire control-word negotiation

Bingen discovered a bug in the pseudowire control-word negotiation that
might happen when the "control-word exclude" command is used. Under some
very specific conditions, ldpd might ignore a PWID label mapping when
it shouldn't.

This patch removes a wrong optimization that was preventing ldpd to call
l2vpn_pw_reset() every time we change the configuration of a pseudowire.

Signed-off-by: Renato Westphal <renato@opensourcerouting.org>

ospfd: Fix ospf timers command

ospf timers related commands under router ospf were not accepted
due to incorrect argument count check.

timers throttle lsa all (0-5000)
timers throttle spf (0-600000) (0-600000) (0-600000)
timers lsa arrival (0-1000)
timers lsa min-arrival (0-600000)

Signed-off-by: Chirag Shah <chirag@cumulusnetworks.com>

pimd: Remove qpim_inaddr_any

In prep for switching over to having PIM understand VRF
remove the qpim_inaddr_any global variable and just
use INADDR_ANY directly.

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

ospfd: Make OSPF TE warnings be less chatty

If you are not using OSPF-TE than there is no need to
be warned that we cannot find anything about it.

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

vtysh: Cleanup some stuff(tm)

1) vtysh_config_dump was not properly indented.  Fix

2) Add VRF_NODE to be line_uniq when parsing return from sub-daemons

3) Fix 'no log monitor' to be line_uniq as well

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

lib, zebra: Pass up kernel table_id for vrf's

pim controls the vrf table creation for due to the way that
pim must interact with the kernel.  In order to match the
table_id for unicast <-> multicast( not necessary but a
real nice to have ) we need to pass up from zebra the
table_id associated with the vrf.

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

lib: free varhandler on exit

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

frr: ignore 'compile_commands.json'

This file is generated by most tooling for Clang and generally wants to
be in the root directory. Best to ignore it.

https://clang.llvm.org/docs/JSONCompilationDatabase.html

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

zebra: Fix help string ordering

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

pimd: Remove some unneeded debugs from pim

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

nhrpd: Guard a debug that was constantly spewing information

This debug when nhrpd was just running but not configured
was constantly being sent to the log file.  Filling it
with useless information

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

doc: Fix up some missed stuff for Ubuntu 12.04 build doc

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

zebra: Refactor 'struct rib' to be 'struct route_entry'

The 'struct rib' data structure is missnamed.  It really
is a 'struct route_entry' as part of the 'struct route_node'.
We have 1 'struct route_entry' per route src.  As such
1 route node can have multiple route entries if multiple
protocols attempt to install the same route.

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

zebra: define mpls label helpstr

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

zebra: more informative error messages for mpls labels

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

ospfd: shufflin

move virtual-link command down to be next to its no form

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

Merge pull request #634 from dwalton76/bgp-ipv6-nexthop-ll-and-global-takeII

bgpd: resolve IPv6 LL and Global nexthop issue

ospfd: fix virtual-link timers commands

Use {} semantics, correct docstrings, switchup parsing.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

Merge pull request #645 from qlyoung/nonblocking-fixup

poll() fixup

Merge pull request #643 from opensourcerouting/2.0-pcreposix

[2.0] fix --enable-pcreposix

frr: remove --enable-poll

vestigial

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: make writing end of pipe nonblocking

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

build: fix pcreposix check

pcreposix_regexec doesn't exist (anymore?), it's just regexec.  Also, if
the user specifies --enable-pcreposix, not finding it is a fatal error.

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

Merge pull request #378 from opensourcerouting/rpm-fixes-2.0

redhat RPM fixes

bgpd: resolve IPv6 LL and Global nexthop issue

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>

lib: missed AWAKEN in r/w thread scheduler

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: allow pthreads to poke poll()

When scheduling a task onto a thread master owned by another pthread, we
need to lock the thread master's mutex. However, if the pthread which
owns that thread master is in poll(), we could be stuck waiting for a
very long time. To solve this, we copy all data poll() needs and unlock
during poll(). To break the target pthread out of poll(), thread_master
has gained a pipe whose reading end is passed into poll(). After an event
that requires immediate action by the target pthread, a byte is written
into the pipe in order to wake it up.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
[DL: split off from select() removal]

lib: remove select()

poll() is present on every supported platform and does not have an upper
limit on file descriptors.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
[DL: split off from AWAKEN() change]

Merge pull request #617 from donaldsharp/bgp_nexthop

vtysh: Add back some missing bgp commands

Merge pull request #632 from opensourcerouting/bgpd-show-segfaults

bgpd: fix segfault with some show commands

Merge pull request #630 from opensourcerouting/zebra-lbl-unicast-issues

Zebra labeled unicast issues

Merge pull request #563 from qlyoung/enforce-no-pthread-cancellation

lib: enforce thread_cancel() MT-unsafe invariant

Merge pull request #605 from dwalton76/show-ip-route-vrf-main

zebra: one "show ip route" to rule them all

Merge pull request #576 from chiragshah6/pim_dev

pimd: pim bfd support

Merge pull request #623 from dslicenc/cm16512

bgpd: labeled-unicast issue changing from no import-check to using im…

Merge pull request #627 from donaldsharp/ripng

Ripng

Merge pull request #603 from qlyoung/fix-access-packed-member

*: do not take address of packed member

bgpd: labeled-unicast issue changing from no import-check to using import-check

Problem seen in testing import-check using labeled-unicast address-family. When
transitioning from "no bgp network import-check" to "bgp network import-check",
previously installed networks were not removed.  This fix resolves this.

Ticket: CM-16512
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>

bgpd: fix segfault with some show commands

The following commands were causing bgpd to crash when vpnv4/vpnv6 static
routes are configured:

bgpd aborted: vtysh  -c "show ip bgp view all ipv4 vpn json"
bgpd aborted: vtysh  -c "show ip bgp view all ipv4 vpn"
bgpd aborted: vtysh  -c "show ip bgp view all ipv6 vpn json"
bgpd aborted: vtysh  -c "show ip bgp view all ipv6 vpn"
bgpd aborted: vtysh  -c "show ip bgp vrf all ipv4 vpn json"
bgpd aborted: vtysh  -c "show ip bgp vrf all ipv4 vpn"
bgpd aborted: vtysh  -c "show ip bgp vrf all ipv6 vpn json"
bgpd aborted: vtysh  -c "show ip bgp vrf all ipv6 vpn"
bgpd aborted: vtysh  -c "show bgp view all ipv4 vpn json"
bgpd aborted: vtysh  -c "show bgp view all ipv4 vpn"
bgpd aborted: vtysh  -c "show bgp view all ipv6 vpn json"
bgpd aborted: vtysh  -c "show bgp view all ipv6 vpn"
bgpd aborted: vtysh  -c "show bgp vrf all ipv4 vpn json"
bgpd aborted: vtysh  -c "show bgp vrf all ipv4 vpn"
bgpd aborted: vtysh  -c "show bgp vrf all ipv6 vpn json"
bgpd aborted: vtysh  -c "show bgp vrf all ipv6 vpn"

Problem found with the CLI fuzzer.

Signed-off-by: Renato Westphal <renato@opensourcerouting.org>

zebra: fix installation of ipv4 labeled unicast routes

Rearrange the _netlink_route_build*() functions so the labels of the
nexthops are always installed, even for IPv4 routes with IPv6 nexthops.

Fixes Labeled Unicast with BGP Unnumbered.

Signed-off-by: Renato Westphal <renato@opensourcerouting.org>

zebra: fix processing of labeled ipv6 routes

We're incrementing nh_count twice for the same nexthop.

Signed-off-by: Renato Westphal <renato@opensourcerouting.org>

Merge pull request #615 from dslicenc/cm16531

zebra: add bgp unnumbered labeled-unicast to zserv

Merge pull request #614 from qlyoung/fix-ospf6-oa-scope

ospf6d: default to dotted quad instead of ret NULL

Merge pull request #625 from donaldsharp/show_zebra

zebra: Fix 64 bit number printf issues

Merge pull request #629 from donaldsharp/eigrp_network

eigrpd: Fix network statement

Merge pull request #618 from donaldsharp/hopola

Hopola

Merge pull request #621 from dslicenc/cm15545

bgpd: fix issue with ipv6 ecmp with vrfs

eigrpd: Fix network statement

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

ripngd: Let ripng announce interface turned on for usage.

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

ripngd: Fix missing newlines at end of help strings

Found by visual inspection and some squinting

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

nhrpd: Fix missing include for 'access_list_init' function

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

zebra: Fix 64 bit number printf issues

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

bgpd: remove struct capability_orf_entry

virtually unused

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

*: do not take address of packed member

May result in alignment errors on certain platforms

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

bgpd: fix issue with ipv6 ecmp with vrfs

Problem reported by customer that ipv6 wasn't installing ecmp paths
when using vrfs.  Found a vrf-unware call in bgp_zebra_announce that
was the culprit.  Testing of the fix looks good.

Ticket: CM-15545
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>

zebra: Remove string operations from non-debug path

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

zebra: Store commonly used values instead of regenerating

the ipv4_ll address used for 5549 routes does not need
to be figured out every single time that we attempt
to install/remove a route of that type.

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

vtysh: Add back some missing bgp commands

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

zebra: add bgp unnumbered labeled-unicast to zserv

Prior to the fix, labels weren't getting installed in zebra nor were the
ifindex values correctly set if labeled-unicast was used in conjunction
with bgp unnumbered.

Ticket: CM-16531
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: CCR-6276

Merge pull request #610 from donaldsharp/rpprefixlen

pimd: Allow secondary address comparisons to work

ospf6d: default to dotted quad instead of ret NULL

to fix coverity warning

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

Merge remote-tracking branch 'origin/stable/3.0'

Merge pull request #612 from LabNConsulting/working/master/patch-set/deprecateEncap

master:  remove support for config & show of encap safi

Merge pull request #611 from LabNConsulting/working/3.0/patch-set/deprecateEncap

3.0: remove support for config & show of encap safi

bgpd: remove encap safi vty related files bgp_encap.h|c

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd: remove encap_safi rx processing

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd: remove encap safi show commands

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd: remove additional instance of BGP_ENCAP(v6)_NODE

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd rfapi: remove ability to configure use of encap-safi (continue to use encap attribute)

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd: remove support for ipv4|6 encap safi config

Signed-off-by: Lou Berger <lberger@labn.net>

vtysh: remove BGP_ENCAP(V6)_NODE

Signed-off-by: Lou Berger <lberger@labn.net>

lib: remove BGP_ENCAP(V6)_NODE

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd: remove encap safi vty related files bgp_encap.h|c

Signed-off-by: Lou Berger <lberger@labn.net>

pimd: Allow secondary address comparisons to work

The secondary address comparison done to determine if we are
an RP for a specified address was comparing A.B.C.D/32 to A.B.C.D/0
because when we created the rp_info we were not setting the prefixlen

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

bgpd: remove encap_safi rx processing

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd: remove encap safi show commands

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd rfapi: remove ability to configure use of encap-safi (continue to use encap attribute)

Signed-off-by: Lou Berger <lberger@labn.net>

bgpd: remove support for ipv4|6 encap safi config

Signed-off-by: Lou Berger <lberger@labn.net>

vtysh: remove BGP_ENCAP(V6)_NODE

Signed-off-by: Lou Berger <lberger@labn.net>

lib: remove BGP_ENCAP(V6)_NODE

Signed-off-by: Lou Berger <lberger@labn.net>

Merge pull request #598 from donaldsharp/to_3.0

To 3.0

zebra: one "show ip route" to rule them all

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>

Merge pull request #593 from qlyoung/fix-nhrpd-acl-3.0

nhrpd: install access-list commands

Merge pull request #591 from qlyoung/ospf-df-areaid-3.0

ospfd: remember format for ospf area id

Merge pull request #582 from qlyoung/ospf6-df-areaid

ospf6d: remember format for ospf6 area id