If we are running with a delayed timer to handle route-map changes
in zebra, if another route-map change is made to the cli, push
out the timer instead of not modifying the timer. This will
allow a large set of route-maps to be possibly be read in by
the system and we don't have a state where new route-map
changes are being read in and having the timer pop in
the middle of it.
Additionally convert to use THREAD_OFF, preventing a possible
use after free as well as aligning the thread api usage
with what we consider correct.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Current code when a route map changes schedules a rerun of all routes in the
particular table. So if you modify the `ip protocol XX route-map FOO`
route-map `FOO` all routes will be rechecked. This is extremely expensive.
Modify zebra to only update the routes associated with the route-map. So
if we have 800k bgp routes and 50 ospf routes and we are route-map'ing
the ospf routes we'll only look at 50 routes.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
When we need to cause a reprocessing of data the code currently
marks all routes as needing to be looked at. Modify the
rib_update_table code to allow us to specify a specific route
type we only want to reprocess. At this point none
of the code is behaving differently this is just setup
for a future code change.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
When the routemap code was rewritten for performance the
code to track the number of times a particular section of
a route-map was applied was not correctly updated. In
this case I found another sequence of events where the
number of times a section was invoked was not being correctly
kept.
Effectively in this case when route_map_get_index is called
and returns an index the route map has been applied( see that
skip_match_clause is set to true and then in the for loop
below the skip_match_clause is tested and index->applied is
incremented.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Commit 220e848cc5 introduced an optimization that would prevent ldpd
from sending redundant label mappings when it receives notifications
from zebra about routes that didn't effectively change (such
notifications can happen under certain circumstances).
The problem is that that commit didn't take into account the metric
of the received routes, so it would dismiss a notification of a
route with a better metric taking the place of another route in the
RIB, preventing the newly selected route from receiving the label
mappings it needs.
Revert 220e848cc5 temporarily to fix sporadic failures in the CI
system until we have a better solution.
Debugged-by: Lynne Morrison lynne@voltanet.io
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Use nl_pid from the netlink socket used for programming the kernel
(netlink_dplane) in netlink route messages sent by the 'fpm' module.
This makes 'fpm' consistent with 'dplane_fpm_nl' which already
behaves this way, and allows FPM server implementations to determine
route origin via nlmsg_pid.
Signed-off-by: Duncan Eastoe <duncan.eastoe@att.com>
The command `neighbor PGROUP ttl-security hops X` was being
accepted but ignored. Allow it to be stored. I am still
not sure that this is applied correctly, but that is another
problem.
Fixes: #7848
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Add the ability for our topotests to take advantage of pytest `mark`ing.
This effectively allows you to tell pytest to run against certain sets
of tests. For a demonstration purpose I've added in marks for:
babel
eigrp
ldp
ospf
pim
rip
And setup tests to run against those tests that only test those protocols.
You can run against eigrp tests by running `pytest -k eigrp`
Other combinations are also available based upon simple boolean logic.
Just read the pytest.mark documentation.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Talking w/ Chirag and he indicated that we can just backout the command
to the original and things would `work` and they do( at least a quick test does )
Put this in place until a proper fix can be done.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
A bgp neighbor remains in Idle state in the event that the number
of received prefixes exceeds the configured maximum prefix for the
neighbor. The neighbor remains in idle state even after de-configuring
the maximum prefix limit for the neighbor.
The fix is to clear the neighbor overflow state if set, after
de-configuring the neighbor maximum-prefix commnd.
This allows the neighbor to establish without having to perform a
clear operation. It also avoids the misleading neigbor summary
indicating that the neighbor is in prefix overflow state (PfxCt)
when no limit is configured for the neighbor.
Signed-off-by: Dewi Morgan <dewi.morgan@intl.att.com>
Create a function that can dump the mac->flags in human readable
output and convert all debugs to use it.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
At present, libyang validate api takes longer time to complete
for a transaction to complete if the same config is re-applied.
For instance if set of static routes are reapplied the config
completion takes longer than it took initial time.
One of the solution is to remove when statement from staticd nexthop yang OM.
When condition adds peformance toll on libyang's validate api.
The same when condition checks are done in frr northbound
validation phase (which are must faster).
With this change, if the same static routes are configured
agian and again, the time to completion does not go up and
perfomance does not degrade.
Ticket:CM-32530
Testing Done:
Configure 400 static routes across two vrfs and keep re-applying them.
The time to complete the config remains in few seconds.
Before:
root@bharat:~/stash/frr4# time vtysh -f static_route_cfg
real 0m19.877s
user 0m0.263s
sys 0m0.014s
After:
root@bharat:~/stash/frr4# time vtysh -f static_route_cfg
real 0m3.857s
user 0m0.239s
sys 0m0.034s
Co-developed-by: VishalDhingra <vdhingra@vmware.com>
Signed-off-by: Chirag Shah <chirag@nvidia.com>
Remove when conditions from the yang OM as it degrades
the performance in libyang.
Instead do the same when conditional check in frr northbound
validate phase. Reject the config if condiion do not meet.
Ticket:CM-32530
Testing Done:
Co-developed-by: VishalDhingra <vdhingra@vmware.com>
Signed-off-by: Chirag Shah <chirag@nvidia.com>
The re->flags and re->status in debugs were being dumped as hex values.
I can never quickly decode this. Here is an idea. Let's let FRR do
it for me.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
If frr.conf contains a prefix-list or access-list without a seq number,
frr-reload needs to be aware that it should not delete/add if the running
config contains a seq number.
Ticket: CM-32623
Signed-off-by: Don Slice <dslice@nvidia.com>
Since new workflow instructions state to run black against
python change and it found formatting changes required that
were not part of my change set, committing those changes
separately.
Signed-off-by: Don Slice <dslice@nvidia.com>
