Age | Commit message (Collapse) | Author |
|
The capabilities' DecodeFromBytes() was eating as much data as given,
while these functions would normally be called with all the data
remaining to parse - e.g. including following capabilities.
Fixes osrg/gobgp#2399
|
|
|
|
Currently, just ignored.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
|
|
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
It wasn't configurable at all.
Signed-off-by: Donatas Abraitis <donatas.abraitis@gmail.com>
|
|
SegmentTypeB, others minor fix
|
|
|
|
segment type b
|
|
|
|
|
|
|
|
In general the writer should close the channel, thus we can close the channel here directly.
If zebra has disconnected the `closeChannel` method will in most cases not even close the channel
because no more messages can be received on it.
|
|
|
|
in gobgp policy.
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
https://staticcheck.io/docs/checks#S1025
|
|
so we can safely remove the MP_REACH_NLRI from ipv4_uc with ipv4 nexthop
|
|
It's not kinda RFC (draft), but it's implemented and used in various
other open-source software like FRRouting, Bird, ExaBGP.
It's very handy when dealing with lots of peers.
Exampe between GoBGP and FRRouting:
```
% ./cmd/gobgp/gobgp neighbor 192.168.10.123 | grep -A4 fqdn:
fqdn: advertised and received
Local:
name: donatas-pc, domain:
Remote:
name: exit1-debian-9, domain:
```
```
% vtysh -c 'show bgp neighbors 192.168.10.17 json' | jq .'"192.168.10.17".neighborCapabilities.hostName'
{
"advHostName": "exit1-debian-9",
"advDomainName": "n/a",
"rcvHostName": "donatas-pc",
"rcvDomainName": "n/a"
}
```
Signed-off-by: Donatas Abraitis <donatas.abraitis@gmail.com>
|
|
Signed-off-by: zhouzijiang <zhouzijiang@jd.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
The issue with the initial update was that sometimes a connection had
not yet been established. Trying to get remote and local connection
data caused a segmentation violation because `fsm.conn` was nil.
Also adds a flag to MonitorPeerRequest in the GRPC API to enable the
new behavior to avoid backward incompatibility.
fixes #2047
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
Fixes #2230
|
|
As specified in draft-ietf-idr-tunnel-encaps-22.
Signed-off-by: Mikael Magnusson <mikma@users.sourceforge.net>
|
|
They are not expected to be executed. This seems a silly path, but
some people bother filing bug reports about it, so if you are OK with
it, it could be "fixed".
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Support config set-next-hop = "unchanged" and also cli command.
Signed-off-by: Faicker Mo <faicker.mo@ucloud.cn>
|
|
|
|
non-MP-capable peer
whereby MP-capable peer uses MP_REACH_NLRI to send updates to gobgp
gobgp accepts it but without setting NEXT_HOP pathattr
then gobgp propagates the paths to the non-MP-capable peer with update.NLRI set but without NEXT_HOP
the non-MP-capable peer will treat the update as withdraw message cause routes cannot propagate properly
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
According to https://tools.ietf.org/html/rfc7752#section-3.2.1.4, IGPRouterID could be 4-octet for OSPFv2 or OSPFv3 non-pseudonode.
|
|
Cover additional cases when only one part of the link is presented.
|
|
- Use tags=netgo to ensure net package won't use cgo
- Rebuild stdlib (go build -a) to really disable cgo as it was
tried in https://github.com/osrg/gobgp/pull/2300
- Remove static ldflags as it would statically compile cgo
|
|
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
release action should be triggered only when a tag is pushed.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: Milan Lenco <milan.lenco@pantheon.tech>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
|
|
|