Donald Sharp [Fri, 29 Oct 2021 12:16:13 +0000 (08:16 -0400)]
zebra: Add a timer to nexthop group deletion
Before deleting nexthop groups, that are installed,
from the system, start a timer and hold the nexthop
group for that time.
Suppose you have this scenario
a) create a static route with 1 x ecmp
creates a nhg with 1 x ecmp
b) create a static route with 2 x ecmp
creates a nhg with 2 x ecmp
deletes a's nhg
c) create a static route with 3 x ecmp
creates a nhg with 3 x ecmp
deletes b's nhg
d) create a different route with 1 x ecmp
creates another 1 x ecmp ( since a's ecmp was deleted )
e) create a different route with 2 x ecmp
creates another 2 x ecmp ( since b's ecmp was deleted )
If you don't delete the nhg, start a timer, the nhg's used
in steps a and b can be reused for steps d and e. This reduces
overhead work with zebra <-> kernel interactions and improves
the speed of the system.
So modify the code to note that an installed nexthop group should
be kept around a bit and hopefully reused.
Donald Sharp [Wed, 15 Jun 2022 20:27:07 +0000 (16:27 -0400)]
zebra: Move where zebra marks a nhg as uninstalled in fib
Currently the code is marking the nhg as uninstalled but not
causing that to flood up to the dependent nhgs:
nhg 3 is a group of 1/2
1 -> interface A
2 -> interface B
Suppose A goes down, old code would mark nhg 1 as !VALID and !INSTALLED.
Suppose B then goes down, old code would mark nhg 2 as !VALID and !INSTALLED
But would not mark nhg 3 as !VALID and !INSTALLED (sort of assuming that
it would just be cleaned up by NHG refcounts ). I would prefer that
the code is pedantic about nhg 3 actually being removed from the system.
This code moves the setting of !INSTALLED into zebra_nhg.c where it
really belongs.
Renato Westphal [Wed, 15 Jun 2022 14:39:48 +0000 (11:39 -0300)]
tests: fix ldp_vpls_topo1 to work as expected
In the last step of this test, r1's link to r2 is shut down but
both routers stay connected through a multi-hop LDP session. That
happens because r1 and r2 have a targeted adjacency created by
the pseudowire. The test then checks whether the pseudowire is
still up, using an alternate path for nexthop resolution.
Everything's fine except for the fact that LDP GTSM (aka
ttl-security) is enabled by default. This means that messages sent
over a multi-hop session are not delivered. In the case of this
test, it can prevent PW-Status notifications from being delivered,
which in turn can prevent the pseudowire from coming back up.
Fix the test by disabling GTSM so that LDP multi-hop sessions can
work normally. This is in accordance with RFC6720 which mentions
that GTSM should be disabled (statically or dynamically) for
multi-hop sessions.
Donald Sharp [Wed, 15 Jun 2022 12:27:32 +0000 (08:27 -0400)]
zebra: On linux let interface data come in through netlink messaging
Consolidate on linux to using the netlink api for gathering all data
about a interface. Leave this interface alone in the meantime for
other OS's.
This also has the side effect of reducing the amount of work
being done on linux in that FRR was handling shut/no shut
events 2 times. Once for the ioctl question asked and
once for the netlink message received.
Donald Sharp [Wed, 15 Jun 2022 11:31:53 +0000 (07:31 -0400)]
zebra: Attempt to make ioctl.c have a bit more useful log messges
While examining the code, it was noticed that there was a chance
to improve the log output in some cases to give a fuller understanding
of what went wrong where.
Donald Sharp [Wed, 15 Jun 2022 14:32:53 +0000 (10:32 -0400)]
bgpd, ospfd: Remove extra newline for `show debugging`
This extra newline was adding a weird output to `show debugging`
display where there would be extra newlines sometims and not
others. Make it consistent.
When pim_upstream_inherited_olist_decide calls the api
pim_channel_add_oif, it can pass PIM_OIF_FLAG_PROTO_GM,
PIM_OIF_FLAG_PROTO_PIM and/or PIM_OIF_FLAG_PROTO_STAR.
Now a consider a case where PIM flag was already set
but STAR flag was not set and this api tries to set
both STAR + PIM and passes the same. The api pim_channel_add_oif
returns since it sees that PIM is already set without
setting the STAR flag.
So basically this will lead to issues in scenarios where for the
same OIF multiple flags(IGMP, PIM, STAR) needs to be set.
Donald Sharp [Tue, 14 Jun 2022 16:21:19 +0000 (12:21 -0400)]
tests: Fix verify_rib such that it will look at the selected route
When you have a static route with multiple different admin
distances there exists a chance that route will have been
installed multiple times due to system load when inserted
at about the same time. If this is the case then the
verify_rib function can and will select the wrong route
that happens to have a nexthop group that is still installed.
Modify verify_rib to ensure that the route that is going to
be looked at for nexthop correctness is the actual installed
route, not a previous version of it.
Renato Westphal [Mon, 13 Jun 2022 18:49:45 +0000 (15:49 -0300)]
tests: fix sporadic failures in the ldp_topo1 topotest
The sporadic failures were happening because, under heavy load,
the r4 router could form an OSPF adjacency with r3 a few seconds
before doing the same with r2. In that interim, LDP could establish
a neighborship with r2 going through r3 (instead of connecting
directly). That would cause all label mappings received from r3
to be ignored since they can't be mapped to the routes' nexthops
received from zebra, causing all sorts of test failures. None of
this is erroneous behavior as LDP simply follows the IGP.
The fix consists of updating the test to ensure all expected OSPF
adjacencies fully converged before proceeding to the LDP checks.
Donald Sharp [Tue, 14 Jun 2022 13:50:54 +0000 (09:50 -0400)]
pimd: Cleanup rpf lookup debug to help us figure out what is going on
The rpf lookup debug was not taking into account the fact that a prefix-list
might be applied and also we might need to make a choice between the two.
So let's give ourselves a bit more data.
For show ip pim interface traffic cli, doing the below changes
1. Changing DEFUN to DEFPY
2. Move the whole code to a common api and modify it so that can
be reused for pimv6.
lib: Update sysrepo code with the latest API changes
* sr_event_notif_send -> sr_notif_send
* sr_process_events -> sr_subscription_process_events
* sr_oper_get_items_subscribe -> sr_oper_get_subscribe
* Removed SR_SUBSCR_CTX_REUSE flag from the code at all
Donald Sharp [Thu, 9 Jun 2022 14:29:04 +0000 (10:29 -0400)]
pimd: Show interface traffic even if interface is currently `down`
the `show ip pim interface [x] traffic` command was deciding
to skip display of interfaces if they happened to be down at
that moment. This of course does not make a bunch of sense
to limit the output for a interface that may have sent data
in the past.
This fixes this test crash:
rnode = <lib.topogen.TopoRouter object at 0x7fc755be3880>, dut = 'c1', input_dict = {'c1': {'c1-l1-eth2': ['helloTx', 'helloRx']}}, output_dict = {'c1': {}}
Donald Sharp [Wed, 8 Jun 2022 20:51:48 +0000 (16:51 -0400)]
pimd: Fix pim_channel_oil_empty tests
The pim_channel_oil_empty() function was setting
the pimreg if it ever existed for NULL comparison
but of course the pimreg device is never pulled back
out again when it was needed to be when the pimreg
is not present.
Donald Sharp [Wed, 8 Jun 2022 13:01:31 +0000 (09:01 -0400)]
babeld: Ignore sub-tlv's with mandatory bit set
RFC 8966 4.4
If the mandatory bit is set, then the
whole enclosing TLV MUST be silently ignored (except for updating the
parser state by a Router-Id, Next Hop, or Update TLV, as described in
the next section).
Fixes: #11349 Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Donald Sharp [Wed, 8 Jun 2022 12:41:05 +0000 (08:41 -0400)]
babeld: Update to ignore unicast hellos
RFC 8966 requires a bit of work to make FRR forward compabitble with
it for the moment.
tl;dr -> Flags of hellos should be 0 except the main bit and
The Unicast Hello bit if set should cause the packet to be ignored.
Additionally FRR needs to ignore Hello packets with the interval field
set to 0.
Fixes: #11349 Signed-off-by: Donald Sharp <sharpd@nvidia.com>
projects / mirror_frr.git / log