zebra: allow multiple connected routes for the same prefix

With unnumbered interfaces on Linux we have the same IP address configured
on several different interfaces and hence multiple connected routes for
the same prefix.

With that said, add an exception in rib_add() to allow zebra to keep
track of all connected routes. We don't need to worry about the bugs
reported in a3d18ce because connected routes are always added from the
connected_up() function, and connected_update() already takes care of
handling duplicate addresses per interface.

Fixes #1112.

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

Merge pull request #1122 from opensourcerouting/non-recursive-3

non-recursive pimd & tools & watchfrr

build: non-recursive pimd

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

build: non-recursive watchfrr & tools

(also forgot to remove AC_SUBST(ISISD) earlier)

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

Merge pull request #1099 from donaldsharp/hashing

Hashing

Merge pull request #1105 from qlyoung/fixup-hashtable-stats-aaa

vtysh: complete 40818cec

Merge pull request #1113 from qlyoung/vtysh-pathspace-fix

vtysh: fix pathspace

vtysh: fix pathspace

It didn't work

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

Merge pull request #1092 from donaldsharp/unreachable

Grab bag of issues

bgpd: Cleanup rmap name memory leak

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

eigrpd: Clean up some new coverity issues

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

pimd: Add usage of vty_frame for vrf configuration

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

zebra: Allow zebra to delete self originated routes

With the change to make zebra pass routes to the kernel
with the 'correct' proto name, it caused zebra to
not properly recognize them on startup again
the next time such that the route would not
be deleted.

Modify rt_netlink.c to notice that we have a
self originated route and to properly mark
the type of route it was.

Modify rib_table_sweep to mark the nexthops
as active so that when we go to delete the
self originated routes it would properly
delete from the kernel.

Fixes: #1061
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

nhrpd: Add hash table names

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

bgpd: Add various hash optimizations

1) Add hash names to all hash_create calls

2) Fix community_hash, ecommunity_hash and lcommunity_hash key
creation

3) Fix output of community and lcommunity iterators( why would
we want to see the memory location of the backet? ).

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

zebra: Add hash name to mpls hash

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

lib: Hash creation cleanup

1) Some hash key functions where converting pointers
directly to a 32 bit value via downcasting.  Pointers
are 64 bit on a majority of our platforms.

2) Some hashes were being created with 256 entries,
downsize the hash creation size to more appropriate
values.

3) Add hash names to hash creation so we can watch
the hash via 'show debugging hashtable'

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

Merge pull request #1108 from donaldsharp/rip

ripd: Fix warning about metric value less than 0

ripd: Fix warning about metric value less than 0

RIP is testing to ensure that the metric returned
isn't negative.  We should be looking at the returned
value from the cli strtol.

If we get a metric value that is less than zero that
means that we shouldn't use this value in RIP currently.
So signify that by returning mod here.

Fixes: #1107
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

Merge pull request #1093 from opensourcerouting/static_blackhole_display

zebra: display static blackhole routes consistently

Merge pull request #1087 from qlyoung/fix-weirdness

lib: fix thread scheduling weirdness

Merge pull request #1075 from donaldsharp/rip_metric

lib, ospfd, ripd, ripngd:  Fix 'set metric'

Merge pull request #1058 from chiragshah6/mdev

ospfd: fix route_node_get

vtysh: complete 40818cec

I missed this in review; the old vtysh command should have been removed
when adding the new one.

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

zebra: accept Null0 again in addition to null0

Restore behavior before commit a830942.

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

zebra: display static blackhole routes consistently

If we configure a static route pointing to null0 and zebra displays
it with the 'blackhole' keyword in the running configuration, the
frr-reload.py script will have issues. Fix this by introducing the
'static_blackhole_type' enum.

Fixes #1091.

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

Merge pull request #1102 from donaldsharp/eigrp_compile

eigrp: Fix missing neighbor->nexthop changes?

Merge pull request #1047 from dwalton76/bgpd-draft-ietf-grow-bgp-gshut-10

bgpd: implement draft-ietf-grow-bgp-gshut-10

Merge pull request #1020 from opensourcerouting/ferr

"ferr" error reporting extensions

eigrp: Fix missing neighbor->nexthop changes?

I'm not sure how this was missed or why master
is not failing to compile.

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

Merge pull request #1051 from donaldsharp/plists

Refactor Access and Prefix Lists application

Merge pull request #1086 from donaldsharp/zebra_import

Zebra import

Merge pull request #1095 from LabNConsulting/working/master/patch/grammer

bgpd rfapi: fixup vnc l2_group config

Merge pull request #1096 from opensourcerouting/isisd-fix-solaris-tests

tests: skip isis tlv fuzztest on SunOS

Merge pull request #1090 from opensourcerouting/rib_issues

RIB issues

Merge pull request #1098 from nkukard/lcommunity-cleanup

Large community critical fixes and misc cleanups

bgpd: Fixed some comments in bgp_lcommunity.h

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Fixed incorrect MTYPE being used for large communites

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: rfapi - fix vestige from gratuitous grammar change.

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

bgpd: Fixed pointer calculation issue in lcommunity

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

tests: skip isis tlv fuzztest on SunOS

zebra: add an exception for link-local connected routes

Link-local routes are unique in the sense that they all have the same
prefix but have different nexthops (local interfaces). Add an exception
in rib_add() to allows us to keep track of all of them.

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

Merge pull request #1088 from opensourcerouting/isis-show-topo-format

isisd: cleanup output of show isis topology

bgpd: Use lcommunity_new() to allocate new lcommunity

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

zebra: allow blackhole routes from the kernel to be deleted

Fixes the following bug:
% ip -6 route add unreachable 4000::/64
% ip -6 route del unreachable 4000::/64
% vtysh -c "show ipv6 route"
[snip]
K>* 4000::/64 [0/1024] unreachable (ICMP unreachable)

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

zebra: fix detection of duplicate system routes

Fixes the following bugs:

1)
% ip -6 route add 5000::/64 via 3000::2
% ip -6 route replace 5000::/64 via 3000::2
% ip -6 route replace 5000::/64 via 3000::2
% ip -6 route replace 5000::/64 via 3000::2
% ip -6 route replace 5000::/64 via 3000::2
%
% vtysh -c "show ipv6 route"
[snip]
K * 5000::/64 [0/1024] via 3000::2, rt1-eth0
K * 5000::/64 [0/1024] via 3000::2, rt1-eth0
K * 5000::/64 [0/1024] via 3000::2, rt1-eth0
K * 5000::/64 [0/1024] via 3000::2, rt1-eth0
K>* 5000::/64 [0/1024] via 3000::2, rt1-eth0

2)
% ip -6 route add 7000::/64 via 3000::2
% ip -6 route replace 7000::/64 via 3000::3
% ip -6 ro | grep 7000
7000::/64 via 3000::3 dev rt1-eth0 metric 1024  pref medium
%
% vtysh -c "show ipv6 route"
[snip]
K * 7000::/64 [0/1024] via 3000::3, rt1-eth0
K>* 7000::/64 [0/1024] via 3000::2, rt1-eth0

NOTE: the check for ROUTE_ENTRY_REMOVED was redundant as it was already
performed at the beginning of the loop.

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

zebra: remove refcnt from the route_entry structure

This reference counter was introduced in 2001, apparently to solve a
problem with connected routes being added/removed multiple times. The RIB
code changed a lot since then, and giving the current callers of rib_add()
and rib_delete() it's safe to assume that we don't need this anymore.

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

Merge pull request #1089 from opensourcerouting/zebra-fix-build

zebra: fix compiler error introduced by previous commit

zebra: fix compiler error introduced by previous commit

Lesson learned: always wait for the CI tests to finish before merging :)

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

Merge pull request #1083 from donaldsharp/unreachable

zebra: Allow bh_type to override nexthop selection

bgpd: Fixed pointer arithmatic miscalculation

If we increment PTR by i * size each time, we end up doing 1, 3, 6
etc.

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Fixed incorrect large community pointer calculation

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Fixed incorrect calculation of pointer location

com_index_to_delete[i] holds the location in lcom->val of where the
value is stored, we cannot just increment it by this value on each
iteration as we'll overflow the size of lcom->val.

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Fixed potential memory leak if additive is false

The else portion of the IF dups the lc, but doesn't free old

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Fixed potential memory leak if additive is false in normal communities

The else portion of the IF dups the community, but doesn't free old

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Cleaned up some comments

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Fixed typo in comment

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Fixed typo in the lc list type being matched on

This should be EXPANDED, not STANDARD.

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: Cleaned up the lcom->size * LCOMMUNITY_SIZE with lcom_length()

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

bgpd: The large community size is easier to read if its using the constant

Signed-off-by: Nigel Kukard <nkukard@lbsd.net>

zebra: Allow bh_type to override nexthop selection

If we've set the bh_type to something besides BLACKHOLE_UNSPEC
due to the received route type being RTN_BLACKHOLE,
RTN_UNREACHABLE or RTN_PROHIBIT then just trust that
the nexthop is just what it is and set accordingly.

Fixes: #1082
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

lib: fix thread scheduling weirdness

Restores 3.0 behavior in terms of thread scheduling & prioritization.

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

isisd: cleanup output of show isis topology

Signed-off-by: Christian Franke <chris@opensourcerouting.org>

zebra: Add v6 support for import-table

Now that all rib functions are afi/safi independent,
allow import-table to work with v6 routes too.

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

*: Cleanup multiple is_default... code

There are 3 different implementations of is_prefix.
Standardize on is_prefix_default and fix it's implementation.

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

zebra: Clean up indentation

Code was hard to look at.  Clean up a bit
since I'm going to be in there.

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

Merge pull request #1076 from donaldsharp/connected

Connected

zebra: Cleanup some typos

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

zebra: Abstract what we can of connected_delete_ipv[4|6]

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

zebra: Refactor connected_down_ipv[4|6]

The connected_down_ipv[4|6] functions are basically identical.
Refactor into one common interface.

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

zebra: Refactor connected_up_ipv[4|6]

The connected_up_ipv[4|6] functions were almost identical.
Combine the forces for the goodness of mankind

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

lib: Add PREFIX_COPY

Add a generic PREFIX_COPY to complement PREFIX_COPY_V4 and
PREFIX_COPY_V6

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

Merge pull request #1078 from dwalton76/ospfd-network-cmd-warning

*: return CMD_WARNING if command was already configured

Merge pull request #1079 from qlyoung/fix-style-a

*: fix style

Merge pull request #1044 from donaldsharp/combination

Coverity Cleanup of Stuff

*: return CMD_WARNING if command was already configured

Signed-off-by: Daniel Walton <dwalton@cumulusnetworks.com>
If the user configures some command that is already in the config we
should return CMD_WARNING instead of CMD_WARNING_CONFIG_FAILED

Merge pull request #1077 from donaldsharp/ws

Cleanup some issues in pim

Merge pull request #1080 from mgsmith1000/systemd_service_perms

Set permissions on systemd service file

redhat: Set permissions on systemd service file

Systemd on CentOS 7.3 logs a warning about the execute bit being
set every time the frr service file is invoked by systemctl.
Modify the spec file to explicitly set the permissions to 644.

Signed-off-by: Matthew Smith <mgsmith@netgate.com>

*: fix style

Fixes style nits introduced by recent pull requests.

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

pimd: Fix pim->keep_alive_time missed patch

The commit '19b807c pimd: Allow the keepalive time to be per vrf.'
is missing some data.  Probably as a result of the indentation
and I accidently dropped it.

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

pimd: Add missed frame for vrf based interface

pim is vrf aware and the interface based line for vrf's was
not being started with a vrf_frame as the non interface
based line was.

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

pimd: Fix indentation issue that slipped in?

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

Merge pull request #1059 from opensourcerouting/oldbits-1

zebra: PtP address configuration support

Merge pull request #1056 from opensourcerouting/oldbits-0

"pathspace" options, vtysh-suid-cleanups, "vty_frame()"

Merge pull request #1074 from opensourcerouting/ospf6-plist-segv

ospf6d: fix prefix list name SEGV

Merge pull request #1073 from LabNConsulting/working/master/patch/fix-pr939-bgp-race

Fix bgpd race

lib: apply_mask to route_node_get

remove address of prefix while typecasting to apply_mask

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

bgp: add asserts to catch unsupported double enqueue of an RN on a bgp_process_queue

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

bgp: fix handling of bgp_process_queue to allow for RNs added
     during callback.  Issue introduced in PR #939.

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

bgp: refactor bgp_process_queue to be consistent and cleanup related event queuing

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

ospf6d: fix prefix list name SEGV

Oops, I tested this with prefix lists configured and failed to notice it
blows up when nothing is set...

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

lib, ospfd, ripd, ripngd:  Fix 'set metric'

There are a variety of cli's associated with the
'set metric ...' command.  The problem that we
are experiencing is that not all the daemons
support all the varieties of the set metric
and the returned of NULL during the XXX_compile
phase for these unsupported commands is causing
issues.  Modify the code base to only return
NULL if we encounter a true parsing issue.
Else we need to keep track if this metric
applies to us or not.

In the case of rip or ripngd if the metric
passed to us is greater than 16 just turn
it internally into a MAX_METRIC.

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

Revert "doc: work around weird texinfo bug"

This reverts commit 50ffbc124bca15b2ca09c1b8b2fd57de6c4d3da4.

The workaround stopped working :(

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

doc: add "ip address LOCAL peer PEER/SUBNET"

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

Merge pull request #998 from mkanjari/dev-master

Dev master

Merge pull request #1072 from opensourcerouting/isisd-fix-confusion-purge

isisd: keep original TLVs when purging in response to confusion

Merge pull request #1071 from opensourcerouting/isis-spf-debug-guard

isisd: guard SPF warning by debug flag

Merge pull request #1060 from opensourcerouting/oldbits-2

zebra: clean up blackhole support

doc: new frr-args(8) man page with general options

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