Ensure that the route-entry QUEUED flag is cleared in the async
notification path, as it is in the normal results processing
code path.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
When we are sending a redistribute_update, pass the old_re in
so that if we still have it around we can update the calling protocol.
Test:
router ospf
redistribute sharp
!
sharp install route 4.5.6.7 nexthop 192.168.201.1 1
Now add a `ip route 4.5.6.7/32 192.168.201.1`.
This causes zebra to replace the sharp route with the static route.
No update is sent to ospf and debug:
2019/08/01 19:02:38.271998 ZEBRA: 0:4.5.6.7/32: Redist update re 0x12fdbda0 (static), old 0x0 (None)
With fix:
2019/08/01 19:15:09.644499 ZEBRA: 0:4.5.6.7/32: Redist update re 0x1ba5bce0 (static), old 0x1beea4e0 (sharp)
2019/08/01 19:15:09.645462 OSPF: ospf_zebra_read_route: from client sharp: vrf_id 0, p 4.5.6.7/32
Ticket: CM-25847
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The flag ROUTE_ENTRY_NEXTHOPS_CHANGED is only ever set or unset.
Since this flag is not used for anything useful, remove from system.
By changing this flag we have re-ordered `internalStatus' of json
output of zebra rib routes. Go through and fix up tetsts to
use the new values.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The code as written before this code change point would enqueue
every system route type to be refigured when we have an
interface event. I believe this was to originally handle bugs
in the way nexthop tracking was handled, mainly that if you keep
asking the question you'll eventually get the right answer.
Modify the code to not do this, we have fixed nexthop tracking
to not be so brain dead and to know when it needs to refigure
a route that it is tracking.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Problem reported where certain routes were not being passed on to
clients if they were operated on while still queued for kernel
installation. Changed it to defer working on entries that were
queued to dplane so we could operate on them after getting an
answer back from kernel installatino.
Ticket: CM-25480
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
Add a file that exposes functions which modify nexthop groups.
Nexthop groups are techincally immutable but there are a
few special cases where we need direct access to add/remove
nexthops after the group has been made. This file provides a
way to expose those functions in a way that makes it clear
this is a private/hidden api.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
The import table code assumes that they will only work
in the default vrf. This is ok, but we should push the
vrf_id and zvrf to be passed in instead of just using
VRF_DEFAULT.
This will allow us to fix a couple of things:
1) A bug in import where we are not creating the
route entry with the appropriate table so the imported
entry is showing up in the wrong spot.
2) In the future allow `ip import-table X` to become
vrf aware very easily.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Improve debugging when we cannot find a route to delete
that we have been told to delete.
New output:
2019/06/25 17:43:49 ZEBRA: default[0]:4.5.6.7/32 doesn't exist in rib
2019/06/25 17:43:49 ZEBRA: default[0]:4.5.6.8/32 doesn't exist in rib
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When dumping rib data about a route for `debug rib detail`
modify the dump command to display the prefix as part
of every line so that we can use a grep on the log
file.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Problem discovered in testing that occasionally when an interface
address was flushed, the corresponding route would be removed from
the kernel and zebra but remain in the bgp table and be advertised
to peers. Discovered that when zebra_rib_evaluate_nexthops spun
thru the tree list of rns, if the timing and circumstances were
right, it would move elements and miss evaluating some. Changed
from frr_each to frr_each_safe and the problem is now gone.
Ticket: CM-25301
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
Add a expected count for the route node we will be processing
as part of nexthop resolution and modify the type to display
a useful string of what the type is instead of a number.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The multicast mode enum was a global static in zebra_rib.c
it does not belong there, it belongs in zebra_router, moving.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
BGP always sends down the correct distance to use. We do
not need rib_add_multipath to double check the code.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Since these functions are not really rib processing problems
let's move them to zebra_nhg.c which is meant for processing of
nexthop groups.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Allow route notifications to trigger route state changes,
such as installed -> not installed.
Clean up the fib-specific nexthop-group in a couple of
un-install paths.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Use some common handling for both route update results
processing and dataplane notification processing. Use the
fib-specific nexthop-group if the update to a route results
in different nexthop status than the default rib-provided
nexthop-group.
Use the fib-specific nexthop-group, if present, to provide
the output of 'show ip fib'.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
When setting route nexthops' installation state based on a
dataplane context struct, unset the installed state if a
nexthop was not present in the dataplane context.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Create a helper api that locates a zebra route-node from info
in a dplane context struct. Moved code from the results handler
to make a more-general api that could be used in other paths.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Add an api to update the status of a route based on info
from a dplane context object. Use the api when processing
route update results from the dataplane.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
<Initial Code from Praveen Chaudhary>
Add the a `--graceful_restart X` flag to zebra start that
now creates a timer that pops in X seconds and will go
through and remove all routes that are older than startup.
If graceful_restart is not specified then we will just pop
a timer that cleans everything up immediately.
Signed-off-by: Praveen Chaudhary <pchaudhary@linkedin.com>
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Allow label ignoring when comparing nexthops. Specifically,
add another functon nexthop_same_no_labels() that shares
a path with nexthop_same() but doesn't check labels.
rib_delete() needs to ignore labels in this case.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
The functions nexthop_same() does not check the resolved
nexthops so I don't think this function is even needed
anymore.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
This is necessary to avoid a name collision with std::for_each
from C++.
Fixes the compilation of the gRPC northbound module.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
VRRP doesn't install any routes, but should still have an array entry.
Also add a help string for VRRP to route_types.txt
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
The re->uptime usage of time(NULL) leaves it open to
timing changes from outside influence. Switching
to monotime allows us to ensure that we have a timestamp
that is always increasing.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
For each table created by a vrf, keep track of it and
allow for proper cleanup on shutdown of that particular
table. Cleanup client shutdown to only cleanup data
that the particular vrf owns. Before we were cleaning
the same table 2 times.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
This command is broken and has been broken since the introduction
of vrf's. Since no-one has complained it is safe to assume that
there is no call for this specialized linux command. Remove
from the system with extreme prejudice.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The route_info[X].meta_q_map *must* be less than MQ_SIZE
or we will do some strange stuff, so assert on it at startup.
The distance in route_info is a uint8_t so let's keep the data
structure the same.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The ifp pointer must be pointing at a real location
in memory since right above us in this loop we
return if it is.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The route entry code was using a custom linked list to handle
route entries. Remove and replace with the new lib link list
code. This reduces the size of the route entry by a further
8 bytes.
Observant people will notice that the current linked list
implementation is singly linked, while the Route Entry
is doubly linked. I am not terribly concerned about this
change as that 1) we do not see a large number of route
entries per prefix( say 2 maybe 3 items ) and route entries
do not come and go that often.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The `struct rib_dest_t` was being used to store the linked
list of rnh's associated with the node. This was taking up
a bunch of memory. Replace with new data structure supplied
by David and see the memory reductions associated with 1 million
routes in the zebra rib:
Old:
Memory statistics for zebra:
System allocator statistics:
Total heap allocated: 675 MiB
Holding block headers: 0 bytes
Used small blocks: 0 bytes
Used ordinary blocks: 567 MiB
Free small blocks: 39 MiB
Free ordinary blocks: 69 MiB
Ordinary blocks: 0
Small blocks: 0
Holding blocks: 0
New:
Memory statistics for zebra:
System allocator statistics:
Total heap allocated: 574 MiB
Holding block headers: 0 bytes
Used small blocks: 0 bytes
Used ordinary blocks: 536 MiB
Free small blocks: 33 MiB
Free ordinary blocks: 4600 KiB
Ordinary blocks: 0
Small blocks: 0
Holding blocks: 0
`struct rnh` was moved to rib.h because of the tangled web
of structure dependancies. This data structure is used
in numerous places so it should be ok for the moment.
Future work might be needed to do a better job of splitting
up data structures and function definitions.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The `struct rib_dest_t` was being used to store the linked
list of rnh's associated with the node. This was taking up
a bunch of memory. Replace with new data structure supplied
by David and see the memory reductions associated with 1 million
routes in the zebra rib:
Old:
Memory statistics for zebra:
System allocator statistics:
Total heap allocated: 675 MiB
Holding block headers: 0 bytes
Used small blocks: 0 bytes
Used ordinary blocks: 567 MiB
Free small blocks: 39 MiB
Free ordinary blocks: 69 MiB
Ordinary blocks: 0
Small blocks: 0
Holding blocks: 0
New:
Memory statistics for zebra:
System allocator statistics:
Total heap allocated: 574 MiB
Holding block headers: 0 bytes
Used small blocks: 0 bytes
Used ordinary blocks: 536 MiB
Free small blocks: 33 MiB
Free ordinary blocks: 4600 KiB
Ordinary blocks: 0
Small blocks: 0
Holding blocks: 0
`struct rnh` was moved to rib.h because of the tangled web
of structure dependancies. This data structure is used
in numerous places so it should be ok for the moment.
Future work might be needed to do a better job of splitting
up data structures and function definitions.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Add a function to check if the route_info array
has all types specified with data in it. Specifically,
test the 'key' attribute for non-zero data. Ignore
ZEBRA_ROUTE_SYSTEM as it should be zero key anyway.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
Add a comment to indicate that route types added to
Zebra, should also be present in the route_info array.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
Add OpenFabric to the route_info array for handling processing
of the OpenFabric route type.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
Problem reported that route-maps applied to "ip protocol table bgp"
would not be invoked if the ip protocol table command was issued
after the bgp prefixes were installed. Found that a recent change
improving how often nexthop_active_update runs missed causing this
filtering to be applied. This fix resolves that issue as well as
a couple of other places that were problematic with the recent
change.
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
Update the nexthop flag output for the route entry dump to
include all possible flag states be output.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
We currently run nexthop_active_check multiple times. Make the
code run once and figure out state from that.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The nexthop_active_update command looks at each individual
nexthop and decides if it has changed. If any nexthop
has changed we will set the re->status to ROUTE_ENTRY_CHANGED
and ROUTE_ENTRY_NEXTHOPS_CHANGED.
Additionally the test for old_nh_num != curr_active
makes no sense because suppose we have several events
we are processing at the same time and a total ecmp
of 16 but 14 are active at the start and 14 are active
at the end but different interfaces are up or down.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The NEXTHOP_FLAG_FILTERED went away when we started treating
static routes like every other route in the system. This was
a special case for handling static route code that just didn't
get finished cleaning up.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
We are effectively calling nexthop_active_update() on every
route entry being processed for installation at least 2 times.
This is a bit ridiculous. We need to resolve the nexthops
when we know a route has changed in some manner, so do so.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
zlog() should be part of the public logging API as it's useful in
the cases where the logging priority isn't known at compile time
(i.e. it depends on a variable).
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
We don't use th vrf-level VRF_RIB_SCHEDULED flag any longer;
remove it and collapse the zebra_vrf flags' values.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
LSP processing was a zvrf flag based upon a connected route
coming or going. But this did not allow us to know
that we should do lsp processing other than after the meta-queue
processing was finished.
Eventually we moved meta-queue processing of do_nht_processing
to after the dataplane sent the main pthread some results.
This of course left us with a timing hole where if a connected
route came in and we received a data plane response *before*
the meta queue was processed we would not do the work as necessary.
Move the lsp processing to a flag off of the rib_dest_t. If it
is marked then we need to process lsps.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Add a detailed debugging command for NHT tracking and add
the detailed output to the log about why we make some decisions
that we are. I tried to model this like the rib processing
detailed debugs that we added a few months back.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Currently nexthop tracking is performed for all nexthops that
are being tracked after a group of contexts are passed back
from the data plane for post install processing.
This is inefficient and leaves us sending nexthop tracking
changes at an accelerated pace, when we think we've changed
a route. Additionally every route change will cause us
to relook at all nexthops we are tracking irrelevant if
they are possibly related to the route change or not.
Let's modify the code base to track the rnh's off of the rib
table's rn, `rib_dest_t`. So after we process a node, install
it into the data plane, in rib_process_result we can
look at the `rib_dest_t` associated with the rn and see that
a nexthop depended on this route node. If so, refigure it.
Additionally we will store rnh's that are not resolved on the
0.0.0.0/0 nexthop tracking list. As such when a route node
changes we can quickly walk up the rib tree and notice that
it needs to be reprocessed as well.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Add a default route_node for our routing tables. This will allow us
to know that we can hang data off the default route for processing.
We will be hanging the nexthop tracking data structures off the rib_dest_t
so that we can know which nexthops we need to handle. Effectively
nexthops that we are tracking that are unresolved will be stored on the
default route. When something changes in the rib tree we can
work up the rn->parent pointer checking for nexthops we need to re-evaluate.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
We have several route types KERNEL and CONNECT that are handled via special
case in the code. This was causing a lot of work keeping the two different
classes of route types as special(SYSTEM OR NOT). Put the dplane
in charge of the code that sets the bits for signalling route install/failure.
This greatly simplifies the code calling path and makes all route types
be handled exactly the same. Additionaly code that we want to run
post data plane install can just work as per normal then, instead
of having to know we need to run it when we have a special type
of route.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com.
When we get a route install failure from the kernel, actually
indicate in the rib the status of the routes.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When switching routes from one route type to another actually
unset the old route as enqueued.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
System Routes if received over the netlink bus in a
specific pattern that causes an update operation for that
route in zebra can leave the dest->selected_fib pointer NULL,
while having the ZEBRA_FLAG_SELECTED flag set. Specifically
one way to achieve this is to do this:
`ip addr del 4.5.6.7/32 dev swp1 ; ip addr add 4.5.6.7/32 dev swp1 metric 9`
Why is this a big deal?
Because nexthop tracking is looking at ZEBRA_FLAG_SELECTED to
know if we can use a route, while nexthop active checking uses
dest->selected_fib.
So imagine we have bgp registering a nexthop. nexthop tracking in
the above case will be able to choose the 4.5.6.7/32 route
if that is what the nexthop is, due to the ZEBRA_FLAG_SELECTED being
properly set. BGP then allows the peers connection to come up and we
install routes with a 4.5.6.7 nexthop. The rib processing for route
installation will then look at the 4.5.6.7 route see no
dest->selected_fib and then start walking up the tree to resolve
the route. In our case we could easily hit the default route and be
unable to resolve the route. Which then becomes inactive in the
rib so we never attempt to install it.
This commit fixes this problem because when the rib_process decides
that we need to update the fib( ie replace old w/ new ), the
replacement with new was not setting the `dest->selected_fib` pointer
to the new route_entry, when the route was a system route.
Ticket: CM-24203
Signed-off-by: Donald Sharp <sharpd@cumulusnetworkscom>
In the case of EVPN symmetric routing, the tenant VRF is associated with
a VNI that is used for routing and commonly referred to as the L3 VNI or
VRF VNI. Corresponding to this VNI is a VLAN and its associated L3 (IP)
interface (SVI). Overlay next hops (i.e., next hops for routes in the
tenant VRF) are reachable over this interface. Howver, in the model that
is supported in the implementation and commonly deployed, there is no
explicit Overlay IP address associated with the next hop in the tenant
VRF; the underlay IP is used if (since) the forwarding plane requires
a next hop IP. Therefore, the next hop has to be explicit flagged as
onlink to cause any next hop reachability checks in the forwarding plane
to be skipped.
https://tools.ietf.org/html/draft-ietf-bess-evpn-prefix-advertisement
section 4.4 provides additional description of the above constructs.
Use existing mechanism to specify the nexthops as onlink when installing
these routes from bgpd to zebra and get rid of a special flag that was
introduced for EVPN-sourced routes. Also, use the onlink flag during next
hop validation in zebra and eliminate other special checks.
Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Reviewed-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Running zebra after commit 888756b208
in valgrind produces this item:
==17102== Invalid read of size 8
==17102== at 0x44D84C: rib_dest_from_rnode (rib.h:375)
==17102== by 0x4546ED: rib_process_result (zebra_rib.c:1904)
==17102== by 0x45436D: rib_process_dplane_results (zebra_rib.c:3295)
==17102== by 0x4D0902B: thread_call (thread.c:1607)
==17102== by 0x4CC3983: frr_run (libfrr.c:1011)
==17102== by 0x4266F6: main (main.c:473)
==17102== Address 0x83bd468 is 88 bytes inside a block of size 96 free'd
==17102== at 0x4A35F54: free (vg_replace_malloc.c:530)
==17102== by 0x4CCAC00: qfree (memory.c:129)
==17102== by 0x4D03DC6: route_node_destroy (table.c:501)
==17102== by 0x4D039EE: route_node_free (table.c:90)
==17102== by 0x4D03971: route_node_delete (table.c:382)
==17102== by 0x44D82A: route_unlock_node (table.h:256)
==17102== by 0x454617: rib_process_result (zebra_rib.c:1882)
==17102== by 0x45436D: rib_process_dplane_results (zebra_rib.c:3295)
==17102== by 0x4D0902B: thread_call (thread.c:1607)
==17102== by 0x4CC3983: frr_run (libfrr.c:1011)
==17102== by 0x4266F6: main (main.c:473)
==17102== Block was alloc'd at
==17102== at 0x4A36FF6: calloc (vg_replace_malloc.c:752)
==17102== by 0x4CCAA2D: qcalloc (memory.c:110)
==17102== by 0x4D03D88: route_node_create (table.c:489)
==17102== by 0x4D0360F: route_node_new (table.c:65)
==17102== by 0x4D034F8: route_node_set (table.c:74)
==17102== by 0x4D03486: route_node_get (table.c:327)
==17102== by 0x4CFB700: srcdest_rnode_get (srcdest_table.c:243)
==17102== by 0x4545C1: rib_process_result (zebra_rib.c:1872)
==17102== by 0x45436D: rib_process_dplane_results (zebra_rib.c:3295)
==17102== by 0x4D0902B: thread_call (thread.c:1607)
==17102== by 0x4CC3983: frr_run (libfrr.c:1011)
==17102== by 0x4266F6: main (main.c:473)
==17102==
This is happening because of this order of events:
1) Route is deleted in the main thread and scheduled for rib processing.
2) Rib garbage collection is run and we remove the route node since it
is no longer needed.
3) Data plane returns from the deletion in the kernel and we call
the srcdest_rnode_get function to get the prefix that was deleted.
This recreates a new route node. This creates a route_node with
a lock count of 1, which we freed via the route_unlock_node call.
Then we continued to use the rn pointer. Which leaves us with use
after frees.
The solution is, of course, to just move the unlock the node at the
end of the function if we have a route_node.
Fixes: #3854
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
With the data plane changes that were made, we are now running
nexthop tracking 2 times. Once at the end of meta-queue insertion
and once at the end of receiving a bunch of data from the dataplane.
The Addition of the data plane code caused flags to not be set
fully for the resolved routes( since we do not know the answer yet ),
This in turn caused the nexthop tracking run after the meta-queue
to think that the route was not `good`. This would cause it to
tell all interested parties that there was no nexthop.
After the dataplane insertion we are also no running nht code.
This was re-figuring out the nexthop correctly and also
correctly reporting to interested parties that there was a path again.
Example:
donna.cumulusnetworks.com(config)# do show ip route
Codes: K - kernel route, C - connected, S - static, R - RIP,
O - OSPF, I - IS-IS, B - BGP, E - EIGRP, N - NHRP,
T - Table, v - VNC, V - VNC-Direct, A - Babel, D - SHARP,
F - PBR, f - OpenFabric,
> - selected route, * - FIB route, q - queued route, f - failed route
K>* 0.0.0.0/0 [0/103] via 10.50.11.1, enp0s3, 00:06:47
S>* 4.5.6.7/32 [1/0] via 192.168.209.1, enp0s8, 00:04:47
C>* 10.50.11.0/24 is directly connected, enp0s3, 00:06:47
C>* 192.168.209.0/24 is directly connected, enp0s8, 00:06:47
C>* 192.168.210.0/24 is directly connected, enp0s9, 00:06:47
donna.cumulusnetworks.com(config)# ip route 4.5.6.7/32 192.168.210.1
donna.cumulusnetworks.com(config)# do show ip route
Codes: K - kernel route, C - connected, S - static, R - RIP,
O - OSPF, I - IS-IS, B - BGP, E - EIGRP, N - NHRP,
T - Table, v - VNC, V - VNC-Direct, A - Babel, D - SHARP,
F - PBR, f - OpenFabric,
> - selected route, * - FIB route, q - queued route, f - failed route
K>* 0.0.0.0/0 [0/103] via 10.50.11.1, enp0s3, 00:07:06
S>* 4.5.6.7/32 [1/0] via 192.168.209.1, enp0s8, 00:00:04
* via 192.168.210.1, enp0s9, 00:00:04
C>* 10.50.11.0/24 is directly connected, enp0s3, 00:07:06
C>* 192.168.209.0/24 is directly connected, enp0s8, 00:07:06
C>* 192.168.210.0/24 is directly connected, enp0s9, 00:07:06
donna.cumulusnetworks.com(config)#
Log files for sharp, which is watching 4.5.6.7:
2019/02/04 15:20:54.844288 SHARP: Received update for 4.5.6.7/32
2019/02/04 15:20:54.844820 SHARP: Received update for 4.5.6.7/32
2019/02/04 15:20:54.844836 SHARP: Nexthop 192.168.209.1, type: 2, ifindex: 3, vrf: 0, label_num: 0
2019/02/04 15:20:54.844853 SHARP: Nexthop 192.168.210.1, type: 2, ifindex: 4, vrf: 0, label_num: 0
As you can see we have received an update with no nexthops( invalid route )
and a second update immediately after it with 2 nexthops.
What's the big deal you say? Well we have code in other daemons that reacts
to not having a path for a nexthop. In BGP this will cause us to tear
down the peer. In staticd we'll remove the recursively resolved route.
In pim we'll remove all paths to the mroute. This is not desirable.
The fix is to remove the meta-queue run of nexthop tracking.
While running after data plane notice of routes to handle is not ideal
we will be fixing this in the future with the nexthop group code, which
should know what nexthops are affected by a nexthop group change.
Fixed code debug code:
donna.cumulusnetworks.com(config)# do show ip route
Codes: K - kernel route, C - connected, S - static, R - RIP,
O - OSPF, I - IS-IS, B - BGP, E - EIGRP, N - NHRP,
T - Table, v - VNC, V - VNC-Direct, A - Babel, D - SHARP,
F - PBR, f - OpenFabric,
> - selected route, * - FIB route, q - queued route, f - failed route
K>* 0.0.0.0/0 [0/103] via 10.50.11.1, enp0s3, 00:00:46
S>* 4.5.6.7/32 [1/0] via 192.168.209.1, enp0s8, 00:00:02
C>* 10.50.11.0/24 is directly connected, enp0s3, 00:00:46
C>* 192.168.209.0/24 is directly connected, enp0s8, 00:00:46
C>* 192.168.210.0/24 is directly connected, enp0s9, 00:00:46
donna.cumulusnetworks.com(config)# ip route 4.5.6.7/32 192.168.210.1
donna.cumulusnetworks.com(config)# do show ip route
Codes: K - kernel route, C - connected, S - static, R - RIP,
O - OSPF, I - IS-IS, B - BGP, E - EIGRP, N - NHRP,
T - Table, v - VNC, V - VNC-Direct, A - Babel, D - SHARP,
F - PBR, f - OpenFabric,
> - selected route, * - FIB route, q - queued route, f - failed route
K>* 0.0.0.0/0 [0/103] via 10.50.11.1, enp0s3, 00:00:59
S>* 4.5.6.7/32 [1/0] via 192.168.209.1, enp0s8, 00:00:02
* via 192.168.210.1, enp0s9, 00:00:02
C>* 10.50.11.0/24 is directly connected, enp0s3, 00:00:59
C>* 192.168.209.0/24 is directly connected, enp0s8, 00:00:59
C>* 192.168.210.0/24 is directly connected, enp0s9, 00:00:59
2019/02/04 15:26:20.656395 SHARP: Received update for 4.5.6.7/32
2019/02/04 15:26:20.656440 SHARP: Nexthop 192.168.209.1, type: 2, ifindex: 3, vrf: 0, label_num: 0
2019/02/04 15:26:33.688251 SHARP: Received update for 4.5.6.7/32
2019/02/04 15:26:33.688322 SHARP: Nexthop 192.168.209.1, type: 2, ifindex: 3, vrf: 0, label_num: 0
2019/02/04 15:26:33.688329 SHARP: Nexthop 192.168.210.1, type: 2, ifindex: 4, vrf: 0, label_num: 0
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The master thread handler is really part of the zrouter structure.
So let's move it over to that. Eventually zserv.h will only be
used for zapi messages.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When we get into rib_process_result and the operation we are handling
is DPLANE_OP_ROUTE_UPDATE *and* the route entry being looked at
is a route replace, we currently have no way to decode to the old_re
and the re due to how we have stored context. As such they are the
same pointer.
As such the route replace for the same route type is causing the re
to set the installed flag and then immediately unset the installed
flag, leaving us in a state where the kernel has the route but
the rib thinks we are not installed.
Since the true old_re( the one being replaced by the update operation )
is going away( as that it zebra deletes the old one for us already )
this fix is not optimal but will get us moving forward.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
If we receive a valid message from the kernel that
is either a kernel or system route, we should trust
that the route is legit and just use it.
Old behavior:
K * 172.22.0.0/15 [0/0] via 172.22.2.254, eva_dummy1 inactive, 00:00:16
New Behavior:
K>* 172.22.0.0/15 [0/0] via 172.22.2.254, eva_dummy1, 00:02:35
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The route entry being displayed in debugs was displaying
the originating route type as a number. While numbers
are cool, I for one am not terribly interested in
memorizing them. Modify the (type %d) to a (%s) to
just list the string type of the route.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The onlink attribute was being passed from upper level protocols
as an attribute of the route *not* the individual nexthop. When
we pass this data to the kernel, we treat the onlink as a attribute
of the nexthop. This commit modifies the code base to allow
us to pass the ONLINK attribute as an attribute of the nexthop.
This commit also fixes static routes that have multiple nexthops
some onlink and some not.
ip route 4.5.6.7/32 192.168.41.1 eveth1 onlink
ip route 4.5.6.7/32 192.168.42.2
S>* 4.5.6.7/32 [1/0] via 192.168.41.1, eveth1 onlink, 00:03:04
* via 192.168.42.2, eveth2, 00:03:04
sharpd@robot ~/frr2> sudo ip netns exec EVA ip route show
4.5.6.7 proto 196 metric 20
nexthop via 192.168.41.1 dev eveth1 weight 1 onlink
nexthop via 192.168.42.2 dev eveth2 weight 1
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When we process the dataplane data, keep track of whether or not a route
is in transit or not.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
zebra is using NEXTHOP_FLAG_FIB as the basis of whether or not
a route_entry is installed. This is problematic in that we plan
to separate out nexthop handling from route installation. So modify
the code to keep track of whether or not a route_entry is installed/failed.
This basically means that every place we set/unset NEXTHOP_FLAG_FIB, we
actually also set/unset ROUTE_ENTRY_INSTALLED on the route_entry.
Additionally where we check for route installed via NEXTHOP_FLAG_FIB
switch over to checking if the route think's it is installed.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Start performing LSP updates through the async dataplane
subsystem. This is plumbed through for linux/netlink.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Favor usage of the afi_t enumeration to identify address-families
over using the classic AF_INET[6] constants for that. The choice to
use either of the two seems to be mostly arbitrary throughout our
code base, which leads to confusion and bugs like the one fixed by
commit 6f95d11a1. To address this problem, favor usage of the afi_t
enumeration whenever possible, since 1) it's an enumeration (helps
the compilers to catch some bugs), 2) has a safi_t sibling and 3)
can be used to index static arrays. AF_INET[6] should then be used
only when interfacing with the kernel or external libraries like
libc. The family2afi() and afi2family() functions can be used to
convert between the two different representations back and forth.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Future commits are going to introduce more rigor in
state setting in the case of received results from
the data plane. So let us move the DPLANE_OP_ROUTE_DELETE
state check to the same spot as the rest of the code that
is handling a particular operation.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Modify the meta_queue insertion such that we only enqueue
the route_node into one meta_queue instead of several.
Suppose we have multiple route_entries associated with
a particular node from rip, bgp, staticd. If we receive a
route update from rip, we would enqueue the route_node into
the 1, 2, 3 meta-nodes. Which means that we would run
the entire process of figuring out a route 3 times, while
nothing would change the second two times.
Modify the code to choose the lowest meta-queue and
install it into that one for processing.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Pass lists of results back to zebra from the dataplane subsystem
(and pthread). This helps reduce the lock/unlock cycles when
zebra is busy. Also remove a couple of typedefs that made their
way into the dataplane header file - those violate the FRR style
guidelines.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
The route_info data structure already had a mapping of route type
to admin distance. Consolidate the meta_queue_map information
into this route_info data structure. This is to reduce the number
of places we need to remember to touch when adding a new routing
protocol.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When a route removal failure happens return to the installing
protocol that the route deletion failed.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
In the zebra rib processing workqueue, set a small timeout
so that we will wait a short time if the queue into the
async dataplane is full. This helps avoid a situation where
the zebra main pthread constantly retries rib work without
giving the dataplane pthread a chance to make progress.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
NEXTHOP_FLAG_ACTIVE currently means that the nexthop is considered
good enough to be installed. With current ecmp restrictions this
translation from multipath_num is enforced in the data plane.
The problem with this is of course that every data plane now
becomes concerned about the multipath num and must enforce it
independently. Currently *bsd does not honor multipath_num at
all and linux marks all nexthops as being installed even when
it honors a multipath_num that is less than the total.
This code change moves the multipath_num enforcement from a dataplane
decision to a zebra nexthop decision. Thus dataplanes now can
just install those nexthops marked as NEXTHOP_FLAG_ACTIVE
without having to worry about multipath_num.
*BSD will now respect multipath_num and Linux now properly notes
which routes are actually installed or not:
sharpd@donna ~/f/t/topotests> ps -ef | grep frr
frr 6261 1556 0 09:12 ? 00:00:00 /usr/lib/frr/zebra -e 2 --daemon -A 127.0.0.1
frr 6279 1556 0 09:12 ? 00:00:00 /usr/lib/frr/staticd --daemon -A 127.0.0.1
donna.cumulusnetworks.com(config)# do show ip route
Codes: K - kernel route, C - connected, S - static, R - RIP,
O - OSPF, I - IS-IS, B - BGP, E - EIGRP, N - NHRP,
T - Table, v - VNC, V - VNC-Direct, A - Babel, D - SHARP,
F - PBR, f - OpenFabric,
> - selected route, * - FIB route
K>* 0.0.0.0/0 [0/106] via 10.0.2.2, enp0s3, 00:00:45
S>* 4.4.4.4/32 [1/0] via 10.0.2.1, enp0s3, 00:00:02
* via 192.168.209.1, enp0s8, 00:00:02
via 192.168.210.1, enp0s9 inactive, 00:00:02
C>* 10.0.2.0/24 is directly connected, enp0s3, 00:00:45
C>* 192.168.209.0/24 is directly connected, enp0s8, 00:00:45
C>* 192.168.210.0/24 is directly connected, enp0s9, 00:00:45
donna.cumulusnetworks.com(config)#
sharpd@donna ~/f/t/topotests> ip route show
default via 10.0.2.2 dev enp0s3 proto dhcp metric 106
4.4.4.4 proto 196 metric 20
nexthop via 10.0.2.1 dev enp0s3 weight 1
nexthop via 192.168.209.1 dev enp0s8 weight 1
10.0.2.0/24 dev enp0s3 proto kernel scope link src 10.0.2.15 metric 106
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1 linkdown
192.168.122.0/24 dev virbr0 proto kernel scope link src 192.168.122.1 linkdown
192.168.209.0/24 dev enp0s8 proto kernel scope link src 192.168.209.2 metric 105
192.168.210.0/24 dev enp0s9 proto kernel scope link src 192.168.210.2 metric 103
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The rib_lookup_ipv4_route function is only used in a debug path.
Is only used for v4 and only checks to make sure that the rib
and fib are in sync( which is not needed/used/supported on other
platforms ). So let's just remove it.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Limit the number of updates processed from the incoming queue;
add more stats. Fill out apis for dataplane providers; convert
route update processing to provider model; move dataplane
status enum
Signed-off-by: Mark Stapp <mjs@voltanet.io>
When the rib code is informed that a table is closing/
going away, only try once to uninstall associated routes from
the fib/dataplane. The close path can be called multiple times
in some cases - zebra shutdown, e.g.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Impose a configurable limit on the number of route updates
that can be queued towards the dataplane subsystem.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Correct use of netlink_parse_info() in the netlink fuzzing path.
Also clarify a couple of comments about pthreads.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
We need a bit of special handling for system routes, which need
to be offered for redistribution even though they won't be
passing through the dplane system.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Set SELECTED re immediately in rib_process, without expecting
that fib install has completed. Remove premature redistribute
call also.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Reduce or eliminate use of global zebra_ns structs in
a couple of netlink/kernel code paths, so that those paths
can potentially be made asynch eventually.
Slide netlink_talk_info into place to remove dependency on core
zebra structs; add accessors for dplane context block
Start init of route context from zebra core re and rn structs;
start queueing and event handling for incoming route updates.
Expose netlink apis that don't rely on zebra core structs;
add parallel route-update code path using the dplane ctx;
simplest possible event loop to process queued route'
updates.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
The `struct zebra_ns` data structure is being used
for both router information as well as support for
the vrf backend( as appropriate ). This is a confusing
state. Start the movement of `struct zebra_ns` into
2 things `struct zebra_router` and `struct zebra_ns`.
In this new regime `struct zebra_router` is purely
for handling data about the router. It has no knowledge
of the underlying representation of the Data Plane.
`struct zebra_ns` becomes a linux specific bit of code
that allows us to handle the vrf backend and is allowed
to have knowledge about underlying data plane constructs.
When someone implements a *bsd backend the zebra_vrf data
structure will need to be abstracted to take advantage of this
instead of relying on zebra_ns.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
During a debugging session last night I discovered that I was
still having some `fun` figuring out why zebra was not making
a route's nexthop active. After some debugging I figured out
that I was missing some states that we could end up in that
didn't have debug information about what happened in nexthop_active.
Add the missing breadcrumbs for nexthop resolution. In addition
add a bit of code to notice the ebgp state without recursion turned
on and to let the user know about it.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
on some cases, kernel routes are not selected, because the kernel
suppressed it without informing the netlink layer that the route has
been suppressed ( for instance, when an interface goes down, the route
never goes back when interface goes up again). This commit intends to
suppress that entry from zebra.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Work to handle the route-maps, namely the header changes in zebra_vrf.h
and the mapping of using that everywhere
Signed-off-by: vishaldhingra vdhingra@vmware.com
