When the last SID in the TI-LFA repair list is an Adj-SID from the
penultimate hop router towards the final hop, the No-PHP flag of the
original Prefix-SID must be honored in the repair list itself since
the penultimate hop router won't have a chance to process that SID
and pop it if necessary.
Reported-by: Fredi Raspall <fredi@voltanet.io>
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
In some cases it's possible that the TI-LFA algorithms will try to
compute a SID repair list more than once for the same backup nexthop
[1]. This of course shouldn't be allowed, as a backup nexthop can't
have multiple label stacks. When that happens, we should just ignore
the new repair list if one is already applied, instead of asserting
and crashing the daemon.
[1] One scenario this can happen is when there's ECMP involving
different P-nodes in the PQ-space intersection.
Reported-by: Fredi Raspall <fredi@voltanet.io>
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
* Clarify which commands are applicable to which flavors of LFA;
* Explain the default prefix priority for different prefix types;
* Rearrange some command descriptions so that they appear in this
order: local LFA, remote LFA and then TI-LFA.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
We can proactively check whether this mroute will be nacked by loopfree
MFC checks so let's do it in the apply phase and emit a useful error
message.
Signed-off-by: Quentin Young <qlyoung@nvidia.com>
This was caused because of uninitialized netlint attrs in the bond-member
netlink parse API.
PS: It was caught by the upstream topotests on ARM8 (passed everywhere
else).
Signed-off-by: Anuradha Karuppiah <anuradhak@nvidia.com>
There are two fixes to handle slow convergence on ARM -
1. Ping on every re-try attempt to account for initial packet loss
2. Handle incomplete show outputs gracefully
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
When a local ES is in LACP bypass state BGP doesn't advertise
reachability to it i.e. the Type-1/EAD-per-ES routes and Type-4
route for the ES is not advertised. This is the equivalent of
oper-down handling.
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
This is needed as kernel currently doesn't allow a mac replace if the dst
changes from a L2NHG to a single-VTEP and viceversa.
Ticket: CM-31561
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
When a ES-bond is in bypass state MACs learnt on it are linked to the
access port instead of the ES. When LACP converges on the bond it moves
out of bypass and the MACs previously learnt on it are flushed to force
a re-learn on new traffic.
Ticket: CM-31326
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
When an ES-bond comes out of bypass FRR needs to flush the local MACs learnt
while the bond was in bypass. To do that efficiently local MACs are linked
to the dest-access port. This only happens if the access-port is in
LACP-bypass or if it is non-ES.
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
Feature overview:
=================
A 802.3ad bond can be setup to allow lacp-bypass. This is done to enable
servers to pxe boot without a LACP license i.e. allows the bond to go oper
up (with a single link) without LACP converging.
If an ES-bond is oper-up in an "LACP-bypass" state MH treats it as a non-ES
bond. This involves the following special handling -
1. If the bond is in a bypass-state the associated ES is placed in a
bypass state.
2. If an ES is in a bypass state -
a. DF election is disabled (i.e. assumed DF)
b. SPH filter is not installed.
3. MACs learnt via the host bond are advertised with a zero ESI.
When the ES moves out of "bypass" the MACs are moved from a zero-ESI to
the correct non-zero id. This is treated as a local station move.
Implementation:
===============
When (a) an ES is detached from a hostbond or (b) an ES-bond goes into
LACP bypass zebra deletes all the local macs (with that ES as destination)
in the kernel and its local db. BGP re-sends any imported MAC-IP routes
that may exist with this ES destination as remote routes i.e. zebra can
end up programming a MAC that was perviously local as remote pointing
to a VTEP-ECMP group.
When an ES is attached to a hostbond or an ES-bond goes
LACP-up (out of bypss) zebra again deletes all the local macs in the
kernel and its local db. At this point BGP resends any imported MAC-IP
routes that may exist with this ES destination as sync routes i.e.
zebra can end up programming a MAC that was perviously remote
as local pointing to an access port.
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
Store instance index at startup and use it when processing vty commands.
The instance itself may be created and deleted by the user in runtime
using `[no] router ospf X` command.
Fixes#7908
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
- Don't document 'no' commands
- Don't use .. index:: for clicmds
- Don't document all possible variants
Signed-off-by: Quentin Young <qlyoung@nvidia.com>
When an unknown LSA is in the database and the user issues the
"show ipv6 ospf6 database" command there is a crash. The code currently
doesn't properly handle display of unknown LSAs.
Signed-off-by: Lynne Morrison <lynne@voltaio.net>
These don't need to be documented, most of the time they are obvious,
when they aren't the behavior can just be described in the command
description.
Signed-off-by: Quentin Young <qlyoung@nvidia.com>
- Generate index entries automatically
- Remove manual command index entries
- Clean up a few other manual index entries
Signed-off-by: Quentin Young <qlyoung@nvidia.com>
When the control plane protocol is created, the vrf structure is
allocated, and its address is stored in the northbound node.
The vrf structure may later be deleted by the user, which will lead to
a stale pointer stored in this node.
Instead of this, allow daemons that use the vrf pointer to register the
dependency between the control plane protocol and vrf nodes. This will
guarantee that the nodes will always be created and deleted together, and
there won't be any stale pointers.
Add such registration to staticd and pimd.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
When handling a large number of events at one time
FRR will call monotime and getrusage 2 times for each
event. With this change modify the code to change
this to (X events / 2) + 1 calls of getrusage and monotime
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
