When we receive an arbitrary table over the netlink bus
save it for later perusal and sweep any routes that
we may have created from an earlier run.
The current redistribute code is limited to
ZEBRA_KERNEL_TABLE_MAX. I left this alone for the
moment because I believe it needs to be converted
to a RB tree instead of a flat array. Which is more
work for the future. Additionally this proposed
change might necessitate some cli changes or rethinks.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
It is possible for clients to install routes into tables
that they desire. Modify the code to delete these routes
from these tables as well.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
There were several places where when I am attempting
to debug zebra functionality that I would really
like to have the ability to know what vrf I think
I am operating on.
Add the vrf_id to a bunch of zlog_debug messages
to help figure out issues when they happen.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Also modify `struct route_entry` to use nexthop_groups.
Move ALL_NEXTHOPS loop to nexthop_group.h
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
If a interested party removes one of it's routes let
it know that it has happened as asked for.
Add a ZAPI_ROUTE_REMOVED to the send of the route_notify_owner
Add a ZAPI_ROUTE_REMOVE_FAIL to the send of the route_notify_owner
Add code in sharpd to notice this and to allow it to keep
track of routes removed for that invocation and give timing
results.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Implement support for EVPN symmetric routing for IPv6 routes. The next hop
for EVPN routes is the IP address of the remote VTEP which is only an IPv4
address. This means that for IPv6 symmetric routing, there will be IPv6
destinations with IPv4 next hops. To make this work, the IPv4 next hops are
converted into IPv4-mapped IPv6 addresses.
As part of support, ensure that "L3" route-targets are not announced with
IPv6 link-local addresses so that they won't be installed in the routing
table.
Signed-off-by: Vivek Venkatraman vivek@cumulusnetworks.com
Reviewed-by: Mitesh Kanjariya mitesh@cumulusnetworks.com
Reviewed-by: Donald Sharp sharpd@cumulusnetworks.com
The ZEBRA_FLAG_INTERNAL flag is used to signal to zebra that
the route being added, the nexthops for it can be recursively
resolved. This name keeps throwing me off when I read it
so let's rename to something that allows the developer to
understand what is going on.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The 'struct route_entry *old' and 'struct route_entry *new' can sometimes
be the same route type( for a route replace ), so when we are checking
to see if a new owner has taken over, don't tell the owner it is
replacing it self.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com.
Some of the tables are no longer stored in the zvrf
and in the zns now. On shutdown zns is cleaned up
after vrf( and rightly so!) As such we should not
attempt to count the information if we don't have
a zvrf.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Problem seen when a prefix was learned with nexthops from multiple
route sources (static and ospf in this case) and the link to that
nexthop flaps. The nht entry was incorrectly deleted so when the
link came back up the static was not re-installed correctly.
Ticket: CM-19675
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
The nh_resolve_via_default function is an accessor function
for NHT in zebra. Let's move this function to it's proper
place.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When a rib_unlink() event is directly called for a
route_entry we need to see if the dest->selected_fib
is the same and just unset the dest->selected_fib.
This was happening for redistributed table 10 routes
into BGP.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The dest->selected_fib assignment needs to happen
after the install and should be controlled by
the southbound api return of success or failure.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The route_node that we are working on is going to be interesting
to the kernel_route_rib_pass_fail. So I am setting up the
code to allow me to pass it. This will be done in a subsuquent
commit.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When a nexthop is resolved via a label based nexthop, copy
the labels into the newly created recursive nexthop.
Please note that this does not fix the case where we
have a label based nexthop that is recursively resolved
through *another* nexthop that is also label based.
In this case we need to create a new label stack
for those routes.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Allow this to work:
vrf DONNA
ip route 4.3.2.1/32 192.168.1.5 nexthop-vrf EVA
The static route code was not properly telling the
nexthop resolution code what vrf to use.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When we are handling nexthops in zebra, use the appropriate
vrf to figure out if the nexthops are active or not.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Add to the rib_add function the ability to pass in the nexthops
vrf.
Additionally when we decode the netlink message from the linux
kernel, properly figure out the nexthops vrf_id.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
With VRF route-leaking we need to know what vrf
the nexthops are in compared to this vrf. This
code adds the nh_vrf_id to the route entry and
sets it up correctly for the non-route-leaking
case.
The assumption here is that future commits
will make the nh_vrf_id *different* than
the vrf_id.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The rn can not have an rn->info pointer and as
such the dest may be NULL. Don't assign
the old_fib pointer if so. This is ok
because we know RNODE_FOREACH... will not
iterate if dest is NULL.
Fixes: #1575
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Linux has the ability to support a concept of 'realms'.
This concept allows you to mark routes with a realm id
value of 1-255. If you have marked the realm
of a route then you can use the tc program to
apply policy to the routes.
This commit adds the ability of FRR to interpret
a tag from (1-255) as a realm when installing into
the kernel. Please note that at this point in time
there is no way to set policy from within FRR. This
must be done outside of it.
The normal methodology for setting tags is valid here
via a route-map.
Finally this is only applied if the --enable-realms configure
option is applied.
Signed-off-by: Kaloyan Kovachev <kkovachev@varna.net>
The SELECTED_FIB flag was placed upon the entry that we
have inserted into the kernel. Remove this flag and replace
with a `rib_dest_t` *selected_fib. Just keep track of the
selected_fib as we modify it. This removes allot of
FOREACH_RE loops as that we do not need to find the
entry anymore.
At this point in time I think this is a very minor performance
boost. Most `rib_dest_t` structures do not typically carry
more than 1 route_entry, but the minute you start having more
than one entry you can and will start having significant processing
time spent finding the selected_fib.
A future commit may re-order the route entries and possibly
keep more pointers on `rib_dest_t` to avoid lookup. This
is a bit tricky because of the FIB_OVERRIDE code.
Signed-off-by Donald Sharp <sharpd@cumulusnetworks.com>
When a route is installed or deleted into the kernel allow a
callback mechanism to handle the success/failure of
the kernel call.
This separation is to allow us to do these things:
1) In the future create a true pthread to handle route
install/deletes. This way we can schedule these
events in a smarter fashion
2) Allow us to use a common southbound api for route
install and deletion.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
zebra_find_client needs to match on instance as well so
protocols like ospfd will work correctly for notification.
Modify the zebra_find_client code to accept the instance
number and to pass it in appropriately.
Signed-off-by: Doanld Sharp <sharpd@cumulusnetworks.com>
When we are installing into the kernel, not the
change points for notification to a higher level
protocol and make it happen
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The rib_uninstall_kernel for non-UNICAST routes when
it is marking a route as no-longer installed should
actually mark it as uninstalled.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Problems reported with zebra nht oscillating when a nexthop is resolved
using the same address to reach the nexthop (for example, 10.0.0.8 is
resolved via 10.0.0.8/32.) This fix removes this attempt to resolve
thru itself unless the route being resolved is also a host route.
This fix also walks up the tree looking for a less specific route to
reach the nexthop if needed. Smoke testing completed successfully.
Ticket: CM-8192
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
Reviewed-by: CCR-6583
Testing done: Manual testing successful, bgp-min completed successfully
l3-smoke completed with two test changes required.
Allow the user to modify the work-queue processing hold time
from 10ms to a value from (0-10000). Make the command hidden
as that it's a semi-dangerous command and it could cause
issues.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Convert the list_delete(struct list *) function to use
struct list **. This is to allow the list pointer to be nulled.
I keep running into uses of this list_delete function where we
forget to set the returned pointer to NULL and attempt to use
it and then experience a crash, usually after the developer
has long since left the building.
Let's make the api explicit in it setting the list pointer
to null.
Cynical Prediction: This code will expose a attempt
to use the NULL'ed list pointer in some obscure bit
of code.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
This is a continuation of 915902cb82. Basically the netlink
read of messages up from the kernel is now noticing the proper
owner of the route. As such when rib_delete was being called
as part of the upcall from the kernel we were not noticing that
we were the originator and not diss-allowing the rib_delete
from happening. This restores this behavior that we were getting
pre-915902cb82cfd
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
While u_char is technically a uint8_t in size I would
like to treat and think about the admin distance
as an actual integer value from 0-255, instead
of a char.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
If we have already scheduled a node to be on the meta_queue, there is no
need to schedule it up again.
On startup we are calling rib_update() multiple times per connected route.
Due to the multiple ways we can get callbacks for adding a connected route
I decided it was best to just improve meta_queue performance as opposed
to trying to figure out all the different ways across all the platforms
that we can decide that a connected route has changed. This appears
to solve the issue with a very large # of interfaces coming up
at the same time on startup.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Both function were very similar, and as we know code duplication is not
good. As an example, in the past couple of weeks some fixes were made
on rib_add() but not on rib_add_multipath(), causing known bugs to still
exist in a different form.
Instead of merging the two functions into one, let's make rib_add()
call rib_add_multipath() with the appropriate parameters. This way we
remove the code duplication but still keep the easy-to-use rib_add()
function for single-path routes.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Fixes the following bug:
% ip route add 50.0.0.0/8 nexthop via 10.0.1.2 nexthop via 10.0.2.2
% ip route replace 50.0.0.0/8 nexthop via 10.0.1.3 nexthop via 10.0.2.3
% ip route replace 50.0.0.0/8 nexthop via 10.0.1.4 nexthop via 10.0.2.4
%
% vtysh -c "show ip route"
[snip]
K * 50.0.0.0/8 [0/0] via 10.0.1.4, rt1-eth1, 00:00:00
* via 10.0.2.4, rt1-eth2, 00:00:00
K * 50.0.0.0/8 [0/0] via 10.0.1.3, rt1-eth1, 00:00:10
* via 10.0.2.3, rt1-eth2, 00:00:10
K>* 50.0.0.0/8 [0/0] via 10.0.1.2, rt1-eth1, 00:00:24
* via 10.0.2.2, rt1-eth2, 00:00:24
Commit a3d18ce6 fixed a similar problem for single-path routes.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Commit f19435a8 fixed rib_add() but didn't fix rib_add_multipath().
While here, remove the unnecessary 'same->table == re->table' check as
it always evaluate to true.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This fixes the broken indentation of several foreach loops throughout
the code.
From clang's documentation[1]:
ForEachMacros: A vector of macros that should be interpreted as foreach
loops instead of as function calls.
[1] http://clang.llvm.org/docs/ClangFormatStyleOptions.html
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
There exists situations where it is possible to have duplicate
nexthops passed from a higher level protocol into zebra.
This code notices this duplication of nexthops and marks
the duplicates as DUPLICATE so we don't attempt to install
it into the kernel.
This is important on *BSD as I understand it because passing
duplicate nexthops will cause the route to be rejected.
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
* Reuse route_distance() on rib_add_multipath() and on rib_add();
* Set the admin distance of LDP and BGP MPLS LSPs.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Set the default admin distance for some route types
more appropriately. The route_distance function
would return 0 for array items not configured, which
is not the right thing to do.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
When we get a route install for a route that needs to be recursively
resolved allow the blackhole to be considered and used if it is
available.
This allows bgp to install a route that will be blackholed.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
We should only be operating RIB_UPDATE_IF_CHANGE on
types that zebra has control of. We assume that
the calling routing protocol is going to take care
of their own route changes based upon the interface
state change.
Also try to re-organize the code a tiny bit to allow
it to fit better within a tabed world.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
There is no need to retrieve the zvrf *unless* we are doing
debugs. So move the retrieval under the debug statement.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
With unnumbered interfaces on Linux we have the same IP address configured
on several different interfaces and hence multiple connected routes for
the same prefix.
With that said, add an exception in rib_add() to allow zebra to keep
track of all connected routes. We don't need to worry about the bugs
reported in a3d18ce because connected routes are always added from the
connected_up() function, and connected_update() already takes care of
handling duplicate addresses per interface.
Fixes#1112.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
With the change to make zebra pass routes to the kernel
with the 'correct' proto name, it caused zebra to
not properly recognize them on startup again
the next time such that the route would not
be deleted.
Modify rt_netlink.c to notice that we have a
self originated route and to properly mark
the type of route it was.
Modify rib_table_sweep to mark the nexthops
as active so that when we go to delete the
self originated routes it would properly
delete from the kernel.
Fixes: #1061
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Link-local routes are unique in the sense that they all have the same
prefix but have different nexthops (local interfaces). Add an exception
in rib_add() to allows us to keep track of all of them.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Fixes the following bugs:
1)
% ip -6 route add 5000::/64 via 3000::2
% ip -6 route replace 5000::/64 via 3000::2
% ip -6 route replace 5000::/64 via 3000::2
% ip -6 route replace 5000::/64 via 3000::2
% ip -6 route replace 5000::/64 via 3000::2
%
% vtysh -c "show ipv6 route"
[snip]
K * 5000::/64 [0/1024] via 3000::2, rt1-eth0
K * 5000::/64 [0/1024] via 3000::2, rt1-eth0
K * 5000::/64 [0/1024] via 3000::2, rt1-eth0
K * 5000::/64 [0/1024] via 3000::2, rt1-eth0
K>* 5000::/64 [0/1024] via 3000::2, rt1-eth0
2)
% ip -6 route add 7000::/64 via 3000::2
% ip -6 route replace 7000::/64 via 3000::3
% ip -6 ro | grep 7000
7000::/64 via 3000::3 dev rt1-eth0 metric 1024 pref medium
%
% vtysh -c "show ipv6 route"
[snip]
K * 7000::/64 [0/1024] via 3000::3, rt1-eth0
K>* 7000::/64 [0/1024] via 3000::2, rt1-eth0
NOTE: the check for ROUTE_ENTRY_REMOVED was redundant as it was already
performed at the beginning of the loop.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This reference counter was introduced in 2001, apparently to solve a
problem with connected routes being added/removed multiple times. The RIB
code changed a lot since then, and giving the current callers of rib_add()
and rib_delete() it's safe to assume that we don't need this anymore.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
blackhole support was horribly broken. cleanup by removing blackhole
stuff from ZEBRA_FLAG_*
introduces support for "prohibit" routes (Linux/netlink only)
also clean up blackhole options on "ip route" vty commands.
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
1) Various socket close issues
2) Ensure afi passed is usable
3) Fix some reads beyond buffer and reads after free
4) Ensure some failure modes are handled properly
5) Memory Leak(s) fix
6) There is no 6.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Change all callers of IPV4_ADDR_SAME() to pass a pointer to a struct in_addr
Use assignment and comparison instead of memcpy() and memcmp(). Avoids function
calls. Faster.
Signed-off-by: Jorge Boncompte <jbonor@gmail.com>
Convert the work queue implementation to not use the generic linked list
to mantain the item list and use instead a simple queue from queue.h that
does not allocate memory for each node.
Signed-off-by: Jorge Boncompte <jbonor@gmail.com>
When the linux kernel adds/deletes routes, the
metric is important, but our routing protocols
add/delete in a slightly different manner,
so allow kernel metrics to match so that our
rib matches the kernel's fib.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Kernel does not send the best route after adding or deleting routes, if
we treat routes for an existing prefix as implicit withdraw the zebra RIB
goes out of sync with FIB and can announce wrong route to protocols.
host:~# vtysh -c 'show ip route'
S>* 0.0.0.0/0 [0/0] via 192.168.1.1, eth0
C>* 192.168.1.0/24 is directly connected, eth0
host:~# ip route add 192.0.2.0/24 via 192.168.1.101 metric 100
host:~# vtysh -c 'show ip route'
S>* 0.0.0.0/0 [0/0] via 192.168.1.1, eth0
K>* 192.0.2.0/24 via 192.168.1.101, eth0
C>* 192.168.1.0/24 is directly connected, eth0
host:~# ip route add 192.0.2.0/24 via 192.168.1.102 metric 50
host:~# vtysh -c 'show ip route'
S>* 0.0.0.0/0 [0/0] via 192.168.1.1, eth0
K>* 192.0.2.0/24 via 192.168.1.102, eth0
C>* 192.168.1.0/24 is directly connected, eth0
host:~# ip route del 192.0.2.0/24 via 192.168.1.102 metric 50
host:~# vtysh -c 'show ip route'
S>* 0.0.0.0/0 [0/0] via 192.168.1.1, eth0
C>* 192.168.1.0/24 is directly connected, eth0
host:~# ip route show 192.0.2.0/24
192.0.2.0/24 via 10.10.1.101 dev eth0 metric 100
Signed-off-by: Jorge Boncompte <jbonor@gmail.com>
Some routing protocols advertise route MTU (e.g. NHRP), with this patch
installed routes in the kernel have the advertised MTU.
Signed-off-by: Jorge Boncompte <jbonor@gmail.com>
This reverts commit c14777c6bf.
clang 5 is not widely available enough for people to indent with. This
is particularly problematic when rebasing/adjusting branches.
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
Switch the RB tree implementation completely to the new dlg@'s version
that uses pre-declared functions instead of macros for tree functions.
Original e-mail/diff:
https://marc.info/?l=openbsd-tech&m=147087487111068&w=2
Pros:
* Reduces the amount of code that the usage of those macros generate
* Allows the compiler to do a better compile-time check job
* Might have better i-cache utilization since the tree code is shared
Con:
* dlg@ benchmarks shows it has 'very slightly slower' insertions
* imported RB_* code must adapt the following calls:
RB_INIT(), RB_GENERATE(), RB_ROOT(), RB_EMPTY(), make compare
functions use 'const' (if not already) and maybe others.
