MPLS VPN networks can either peer with iBGP or eBGP. When
calculating the distance to send to zebra, the imported prefix
is never sent with distance information, even if the vty
command is used under the ipv4 unicast address family:
router bgp 65505 vrf vrf1
address-family ipv4 unicast
distance bgp 26 27 28
[vpn config]
The observation is that the distance sent to zebra for an
imported prefix is still 20:
[..]
VRF vrf1:
B> 192.168.0.0/24 [20/0] via 2.2.2.2 (vrf default) (recursive), label 20, weight 1, 00:00:12
* via 10.125.0.6, ntfp3 (vrf default), label implicit-null/20, weight 1, 00:00:12
The expectation is that the incoming prefix has to follow the
distance that is configured, or the distance derived from the peer
relationship established by the parent prefix.
In the case, an iBGP relationship is done, and no distance
configuration is done, the below show is expected:
[..]
VRF vrf1:
B*> 192.168.0.0/24 [200/0] via 192.168.0.2, r1-gre0 (vrf default), label 20, weight 1, 00:00:12
In the case an iBGP relationship is done, and distance configuration
is performed as below:
[..]
distance bgp 21 201 41
[..]
Then the below show is expected:
[..]
VRF vrf1:
B*> 192.168.0.0/24 [201/0] via 192.168.0.2, r1-gre0 (vrf default), label 20, weight 1, 00:00:12
To get this behaviour, get the peer origin where the prefix is coming
from.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
When setting rule for access-list ( and prefix-list ) without sequence, it
will automatically get a sequence by `acl_get_seq()`, and return
`CMD_SUCCESS` for command even this sequence value is wrong.
In this scene, `CMD_WARNING_CONFIG_FAILED` should be returned with a
warning.
So, add the check in `acl_get_seq()` and move `nb_cli_enqueue_change()`
after the check of wrong sequence.
Both `plist_remove_if_empty()` and `acl_remove_if_empty()` should ignore
this check, there is no change on them.
Before:
```
anlan(config)# access-list aa seq 4294967295 deny 6.6.6.6/32
anlan(config)# access-list aa deny 6.6.6.7/32 <- Return CMD_SUCCESS
YANG error(s):
Value "4294967300" is out of uint32's min/max bounds.
Value "4294967300" is out of uint32's min/max bounds.
Value "4294967300" is out of uint32's min/max bounds.
Value "4294967300" is out of uint32's min/max bounds.
Value "4294967300" is out of uint32's min/max bounds.
YANG path: Schema location /frr-filter:lib/prefix-list/entry/sequence.
% Failed to edit configuration.
```
After:
```
anlan(config)# access-list aa seq 4294967295 deny 6.6.6.6/32
anlan(config)# access-list aa deny 6.6.6.7/32 <- Return CMD_WARNING_CONFIG_FAILED
% Malformed sequence value
```
Additionally, fixed the overflow issue on `acl_get_seq()` on **32bit** platforms.
Just change the returned type of `acl_get_seq()` from `long` to `int64_t`.
Before:
```
anlan(config)# access-list bb seq 4294967295 deny 6.6.6.6/32
anlan(config)# access-list bb deny 6.6.6.7/32
anlan(config)# do show run
...
access-list bb seq 4294967295 deny 6.6.6.6/32
access-list bb seq 4 deny 6.6.6.7/32 <- Overflow
```
After:
```
anlan(config)# access-list bb seq 4294967295 deny 6.6.6.6/32
anlan(config)# access-list bb deny 6.6.6.7/32
% Malformed sequence value
```
Signed-off-by: anlan_cs <vic.lan@pica8.com>
After calling `rib_unlink` the variable `re` will point to `free()`d
memory, so don't attempt to use it after this point.
Found by Coverity Scan (Coverity ID 1519784)
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
[New LWP 2524]
[New LWP 2539]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Core was generated by `/opt/avi/bin/bgpd -f /run/frr/avi_ns3_bgpd.config -i /opt/avi/etc/avi_ns3_bgpd.'.
Program terminated with signal SIGABRT, Aborted.
[Current thread is 1 (Thread 0x7f92ac8f1740 (LWP 2524))]
0 0x00007f92acb3800b in raise () from /lib/x86_64-linux-gnu/libc.so.6
[Current thread is 1 (Thread 0x7f92ac8f1740 (LWP 2524))]
0 0x00007f92acb3800b in raise () from /lib/x86_64-linux-gnu/libc.so.6
1 0x00007f92acb17859 in abort () from /lib/x86_64-linux-gnu/libc.so.6
2 0x00007f92acb17729 in ?? () from /lib/x86_64-linux-gnu/libc.so.6
3 0x00007f92acb28fd6 in __assert_fail () from /lib/x86_64-linux-gnu/libc.so.6
4 0x00007f92accf2164 in pthread_mutex_lock () from /lib/x86_64-linux-gnu/libpthread.so.0
5 0x000055b46be1ef63 in bgp_keepalives_wake () at bgpd/bgp_keepalives.c:311
6 0x000055b46be1f111 in bgp_keepalives_stop (fpt=0x55b46cfacf20, result=<optimized out>) at bgpd/bgp_keepalives.c:323
7 0x00007f92acea9521 in frr_pthread_stop (fpt=0x55b46cfacf20, result=result@entry=0x0) at lib/frr_pthread.c:176
8 0x00007f92acea9586 in frr_pthread_stop_all () at lib/frr_pthread.c:188
9 0x000055b46bdde54a in bgp_pthreads_finish () at bgpd/bgpd.c:8150
10 0x000055b46bd696ca in bgp_exit (status=0) at bgpd/bgp_main.c:210
11 sigint () at bgpd/bgp_main.c:154
12 0x00007f92acecc1e9 in quagga_sigevent_process () at lib/sigevent.c:105
13 0x00007f92aced689a in thread_fetch (m=m@entry=0x55b46cf23540, fetch=fetch@entry=0x7fff95379238) at lib/thread.c:1487
14 0x00007f92aceb2681 in frr_run (master=0x55b46cf23540) at lib/libfrr.c:1010
15 0x000055b46bd676f4 in main (argc=11, argv=0x7fff953795a8) at bgpd/bgp_main.c:482
Root cause:
This is due to race condition between main thread & keepalive thread during clean-up.
This happens when the keepalive thread is processing a wake signal owning the mutex, when meanwhile the main thread tries to stop the keepalives thread.
In main thread, the keepalive thread’s running bit (fpt->running) is set to false, without taking the mutex & then it blocks on mutex.
Meanwhile, keepalive thread which owns the mutex sees that the running bit is false & executes bgp_keepalives_finish() which also frees up mutex.
Main thread that is waiting on mutex with pthread_mutex_lock() will cause core while trying to access mutex.
Fix:
Take the lock in main thread while setting the fpt->running to false.
Signed-off-by: Samanvitha B Bhargav <bsamanvitha@vmware.com>
Add a function to find the VRF or the loopback interface: the loopback
interface for the default VRF and the VRF master interface otherwise.
Signed-off-by: Louis Scalbert <louis.scalbert@6wind.com>
Before this patch we allowed importing routes between VRFs in the same node,
only for external routes, but not for local (e.g.: redistribute).
Relax here a bit, and allow importing local routes between VRFs when the RT
list is modified using route reflectors.
Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>
I'm seeing test failures after in micronet runs in CI
after 7 seconds * 30 attempts at seeing if it succeeds.
Let's see if another 60 seconds of attempts allows
this to work properly.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Got `ERROR: Daemon babeld is not a valid option for 'show running-config'` when using `frr-reload.py --reload --daemon babeld`.
Adds `babeld` and `nhrpd` as valid daemons.
Signed-off-by: Yuxiang Zhu <vfreex@gmail.com>
Before this patch, we needed to explicitly define a neighbor to be SOLO
(= separate update-group). Let's ease this functionality for an operator to
avoid confusions.
Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>
The idea is to drop unwanted attributes from the BGP UPDATE messages and
continue by just ignoring them. This improves the security, flexiblity, etc.
This is the command that Cisco has also.
Signed-off-by: Donatas Abraitis <donatas@opensourcerouting.org>
Single run of this test suite on my machine was 8 minutes.
Breaking this up into 3 test suites halves the run time.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
This change alters the behavior of existing test code. The
default mode (before any call to luSetWaitType()) is now
"strict".
The historical behavior of luCommand(op="wait) is to ignore
failures to match the specified regexp in the specified time.
In those cases, no result was logged and no error was signaled.
This change introduces a new "strict" mode for luCommand(op="wait):
in "strict" wait mode, each invocation of luCommand(op="wait)
generates an explicit, logged failure result when it fails to match
the specified regexp in the specified time. These failures signal
an error for the test.
Calling luSetWaitType("nostrict") restores the historical behavior.
Calling luSetWaitType("strict") (re)enables the new strict behavior.
Individual calls to luCommand() may also specify op="wait-nostrict"
to override any default and use the historical behavior.
Individual calls to luCommand() may also specify op="wait-strict"
to override any default and use the new behavior.
Signed-off-by: G. Paul Ziemba <paulz@labn.net>
This command helps in troubleshooting static bfd feature.
Add traces upon bfd events.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Test that BFD static monitoring works:
When BFD session is up the routes are installed in the RIB and
distributed with routing protocol (in this case BGP). When the session
is down it is removed from RIB and propagated.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Let the user know how to use the static route monitoring commands.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
When configuring a route with multiple next hops on boot and the
interface is missing the route never shows up (even after interface is
properly learned).
Fix: force route clean up by uninstalling it and putting it back
entirely instead of just updating the missing next hop.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Implement all BFD integration northbound callbacks and integrate BFD
with `staticd` route installation procedure.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Define a generic BFD monitoring group template and use it to add support
for static route monitoring.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Implement clean up function to be called on shutdown to make daemon exit
clean for valgrind and other memory sanitizers.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Implement new BFD library issue to allow protocols to configure BFD
sessions with automatic source selection.
The source selection will be based on the Next Hop Tracking feature:
`zebra` will do RIB lookups to determine the output interface and the
primary source address of that interface will be used as source.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
