zebra: EVPN DAD trigger was causing zebra to crash
Duplicate address detection and recovery was relying on the l2-vni backptr
in the neighbor entry which was simply not initialized resulting in
a NULL pointer access in a setup with dup-addressed VMs -
VM1:{IP1,M1} and VM2:{IP1,M2}
Call stack:
(gdb) bt 6
at lib/sigevent.c:249
nbr=nbr@entry=0x559347f901d0, vtep_ip=..., vtep_ip@entry=..., do_dad=do_dad@entry=true,
is_dup_detect=is_dup_detect@entry=0x7ffc7f6be59f, is_local=is_local@entry=true)
at ./lib/ipaddr.h:86
ip=0x7ffc7f6be6f0, ifp=0x559347f901d0, zvni=0x559347f86800) at zebra/zebra_vxlan.c:3152
(More stack frames follow...)
(gdb) p nbr->zvni
$8 = (zebra_vni_t *) 0x0 <<<<<<<<<<<<<<<<<<<<
(gdb)
Donald Sharp [Mon, 11 Mar 2019 13:39:19 +0000 (09:39 -0400)]
zebra: System routes sometimes can not be properly selected
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>
Don Slice [Thu, 14 Mar 2019 15:17:47 +0000 (15:17 +0000)]
bgpd: fix redistribution into vrf when networking is restarted
Found that previous fix for this issue caused collatoral damage and
reverted that fix. This fix clears the vrf_bitmaps when the vrf is
disabled/deleted and then re-applies the redist config when the vrf
is re-enabled.
Ticket: CM-24231 Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
Rafael Zalamena [Mon, 11 Mar 2019 18:09:15 +0000 (15:09 -0300)]
bfdd: change session lookup data structure
Use simplier data structure key to avoid having to do complex and
error-prone key building (e.g. avoid expecting caller to know IPv6
scope id, interface index, vrf index etc...).
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Donald Sharp [Fri, 8 Mar 2019 15:46:55 +0000 (10:46 -0500)]
zebra: Add some debugs to neighbor entry processing
When we get a neighbor entry in zebra we start processing it.
Let's add some additional debugs to the processing so that when
it bails out and we don't use the data, we know the reason.
This should help in debugging the problems from why bgp does
not appear to have data associated with a neighbor entry
in the kernel.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Chirag Shah [Thu, 7 Mar 2019 20:36:47 +0000 (12:36 -0800)]
bgpd: vrl route-leak show with all vrfs option
VRF Route Leak's
show bgp vrf all ipv4 unicast route-leak
is not supported with `all` keyword.
Testing Done:
bl1# show bgp vrf all ipv4 unicast route-leak
Instance default:
This VRF is not importing IPv4 Unicast routes from any other VRF
This VRF is not exporting IPv4 Unicast routes to any other VRF
Instance vrf3:
This VRF is importing IPv4 Unicast routes from the following VRFs:
vrf1
Import RT(s): 144.1.1.2:10
This VRF is exporting IPv4 Unicast routes to the following VRFs:
vrf1
RD: 144.1.3.2:9
Export RT: 144.1.3.2:9
Instance vrf1:
This VRF is importing IPv4 Unicast routes from the following VRFs:
vrf3
Import RT(s): 144.1.3.2:9
This VRF is exporting IPv4 Unicast routes to the following VRFs:
vrf3
RD: 144.1.1.2:10
Export RT: 144.1.1.2:10
Signed-off-by: Chirag Shah <chirag@cumulusnetworks.com>
Tuetuopay [Tue, 5 Mar 2019 18:19:28 +0000 (19:19 +0100)]
zebra: Treat ifaces withouth IPv4 as unnumbered
The current definition of an unnumberd interface as an interface with a
/32 IPv4 is too restrictive, especially for EVPN symmetric routing since
commit 2b83602b2 "*: Explicitly mark nexthop of EVPN-sourced routes as
onlink".
It removes the bypass check wether the nexthop is an EVPN VTEP, and
relies on the SVI to be unnumberd to bypass the gateway lookup. While
this works great if the SVI has an IP, it might not, and the test falls
flat and EVPN type 5 routes are not installed into the RIB.
Sample interface setup, where vxlan-blue is the L3VNI and br-blue the
SVI:
bgpd: Incorrect number of peers count in "show bgp ipv6 summary output
The "show bgp ipv6 summary" output displays incorrect number of peers count.
sonic# show bgp ipv6 summary
IPv6 Unicast Summary:
BGP router identifier 10.1.0.1, local AS number 65100 vrf-id 0
BGP table version 0
RIB entries 0, using 0 bytes of memory
Peers 5, using 103 KiB of memory
Peer groups 1, using 64 bytes of memory
Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd
2003::1 4 65099 0 0 0 0 0 never Active
2088::1 4 65100 0 0 0 0 0 never Active
3021::2 4 65100 0 0 0 0 0 never Active
Total number of neighbors 3
sonic#
In the above output, the peers count displays as 5 but the actual peer count is 3, i.e.. 3 neighbors are activated in ipv6 unicast address family.
Displayed peer count (5) is the number of the neighbors activated in a BGP instance.
Fix : Now the peers count displays the number of neighbors activated per afi/safi.
After Fix:
sonic# show bgp ipv6 summary
IPv6 Unicast Summary:
BGP router identifier 10.1.0.1, local AS number 65100 vrf-id 0
BGP table version 0
RIB entries 0, using 0 bytes of memory
Peers 3, using 62 KiB of memory
Peer groups 1, using 64 bytes of memory
Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd
2003::1 4 65099 0 0 0 0 0 never Active
2088::1 4 65100 0 0 0 0 0 never Active
3021::2 4 65100 0 0 0 0 0 never Active
Donald Sharp [Wed, 6 Mar 2019 15:44:34 +0000 (10:44 -0500)]
configure: Default to 16 way ecmp on compilation
If a person who is compiling FRR does not specify the
multipath number on configure we are defaulting to a ecmp of 1.
Let's change this to 16. In this day and age most everything
supports actual ecmp.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
bgpd: prevent type-5 route creation if bgp_vrf->l3_vni is 0
After a router reboot the L3 network via it converges before the L2
network. This is because MLAG intentionally holds down bridge-access
and vxlan-network ports for some time (MLAG init-delay) to prevent traffic
from switching to a router that is not fully ready. This also means that
routes (from vrf-peering sessions) that qualify for evpn type-5
advertisments are available long before the L3-VNI is available for that
tenant VRF. In these windows bgpd was adding these evpn-type-5 routes with
a L3-VNI of 0 (which was not fixed up after the L3-VNI became available) -
BGP routing table entry for 100.0.0.1:2:[5]:[0]:[0]:[32]:[200.1.1.1]
Paths: (1 available, best #1)
Advertised to non peer-group peers:
MSP1(uplink-1) MSP2(uplink-2)
Route [5]:[0]:[0]:[32]:[200.1.1.1] VNI 0 >>>>>>>>
65001 65535
36.0.0.9 from 0.0.0.0 (27.0.0.9)
Origin incomplete, metric 0, valid, sourced, local, bestpath-from-AS 65001, best
Extended Community: ET:8 RT:5544:4001 Rmac:44:38:39:ff:ff:01
AddPath ID: RX 0, TX 327
Last update: Wed Feb 27 18:37:10 2019
Fix is to defer creating type-5 routes till the L3-VNI is available for
that tenant VRF (this was already being done for most cases; fixup takes
care of some that missed the check).
Mark Stapp [Tue, 5 Mar 2019 20:28:26 +0000 (15:28 -0500)]
libs: make privilege escalation thread-safe
Privs escalation is process-wide, and a multi-threaded process
can deadlock. This adds a mutex and a counter to the privs
object, preventing multiple threads from making the privs
escalation system call.
Donald Sharp [Tue, 5 Mar 2019 15:29:35 +0000 (10:29 -0500)]
pimd: Ensure DR election happens when both sides change prio
Suppose we have 2 routers A and B. Both Router A and B have
the same priority of 1000. Router A is the elected DR.
Now suppose B lowers his priority to 1. He still looses the
DR election and we are not sending a hello with the new priority.
Immediately after this A's priority is also lowered to 1, it
looses the election and sends the hello. B receives this hello
and elects A as the DR( since it has the better ip address)
At this point A believes B is the DR, and B believes A is the
DR until such time that the normal hello from B is sent to A,
which if timed correctly can be a significant amount of time).
This code just causes a hello to be sent if the priority is
changed. Now both sides will be able to converge quickly
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Chirag Shah [Thu, 28 Feb 2019 00:36:47 +0000 (16:36 -0800)]
bgpd: router mac same as self supress bgp update
bgp update can contain router mac address same as one of SVIs
mac address, during processing of evpn route in bpg_update()
check for the flag is set and filter the route from installing.
This check is done prior to attribute lookup or storing in database.
Parse check and set is done once during attribute parse
because all the NLRIs containing evpn prefix
(type-2/type-5) will have same exntended community applicable.
Chirag Shah [Wed, 20 Feb 2019 00:02:00 +0000 (16:02 -0800)]
bgpd: parse and comapre rmac attr against self mac
Any evpn bgp update message comes with router mac extended
community, which can potentially contain the madd adddress
same as any of the local SVIs (L3VNI) MAC address.
Set route mac exist and during route processing in
bgp_update() filter the route.
Ticket:CM-23674
Reviewed By:CCR-8336
Testing Done:
Configure L3vni mac on TORS1 which is similar to TORC11
L3vni MAC. When TORC11 received the EVPN update with
Router mac extended community, this check rejected the
BGP update message.
Signed-off-by: Chirag Shah <chirag@cumulusnetworks.com>
Renato Westphal [Thu, 28 Feb 2019 22:54:47 +0000 (19:54 -0300)]
lib: fix removal of yang non-presence containers
Non-presence containers don't have "destroy" callbacks. So, once
a np-container is deleted, we need to call the "destroy" callbacks
of its child nodes instead.
This commit doesn't fix any real problem as of now since all
np-containers from the FRR YANG modules contain or one more mandatory
child nodes, so they can't be deleted (libyang will add missing
np-containers when validating data). Nevertheless, upcoming YANG
modules should benefit from this change.
Renato Westphal [Sat, 2 Mar 2019 20:45:14 +0000 (17:45 -0300)]
ripd: fix removal of configured passive interfaces
libyang-0.16-rc3 fixed a bug [1] in which data would be auto-deleted
when it shouldn't. The problem is that the "no passive-interface"
command was relying on that wrong behavior, so the command was
affected when the libyang bug was fixed. Adapt the command to do
the right thing in order to get rid of the problem (regardless of
the libyang version being used).
"passive-interface default" still has problems though, but that
will be addressed separetely in the future.
Don Slice [Sat, 2 Mar 2019 19:40:17 +0000 (19:40 +0000)]
bpgd: resolve more neighbor peer-group issues
Found in testing that in a certain sequence, a neighbor's peer-group
membership would be lost. This fix resolves that issue. Additionally
found that "no neighbor swp1 remote-as 2" would sometimes leave the
config with "neighbor swp1 remote-as 0" rather than removing from the
config. That one is also resolved.
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
Donald Sharp [Thu, 28 Feb 2019 14:11:41 +0000 (09:11 -0500)]
zebra: Upon vrf deletion, actually release this data.
When a vrf is deleted we need to tell the zebra_router that we have
finished using the tables we are keeping track of. This will allow
us to properly cleanup the data structures associated with them.
This fixes this valgrind error found:
==8579== Invalid read of size 8
==8579== at 0x430034: zvrf_id (zebra_vrf.h:167)
==8579== by 0x432366: rib_process (zebra_rib.c:1580)
==8579== by 0x432366: process_subq (zebra_rib.c:2092)
==8579== by 0x432366: meta_queue_process (zebra_rib.c:2188)
==8579== by 0x48C99FE: work_queue_run (workqueue.c:291)
==8579== by 0x48C3788: thread_call (thread.c:1607)
==8579== by 0x48A2E9E: frr_run (libfrr.c:1011)
==8579== by 0x41316A: main (main.c:473)
==8579== Address 0x5aeb750 is 0 bytes inside a block of size 4,424 free'd
==8579== at 0x4839A0C: free (vg_replace_malloc.c:540)
==8579== by 0x438914: zebra_vrf_delete (zebra_vrf.c:279)
==8579== by 0x48C4225: vrf_delete (vrf.c:243)
==8579== by 0x48C4225: vrf_delete (vrf.c:217)
==8579== by 0x4151CE: netlink_vrf_change (if_netlink.c:364)
==8579== by 0x416810: netlink_link_change (if_netlink.c:1189)
==8579== by 0x41C1FC: netlink_parse_info (kernel_netlink.c:904)
==8579== by 0x41C2D3: kernel_read (kernel_netlink.c:389)
==8579== by 0x48C3788: thread_call (thread.c:1607)
==8579== by 0x48A2E9E: frr_run (libfrr.c:1011)
==8579== by 0x41316A: main (main.c:473)
==8579== Block was alloc'd at
==8579== at 0x483AB1A: calloc (vg_replace_malloc.c:762)
==8579== by 0x48A6030: qcalloc (memory.c:110)
==8579== by 0x4389EF: zebra_vrf_alloc (zebra_vrf.c:382)
==8579== by 0x438A42: zebra_vrf_new (zebra_vrf.c:93)
==8579== by 0x48C40AD: vrf_get (vrf.c:209)
==8579== by 0x415144: netlink_vrf_change (if_netlink.c:319)
==8579== by 0x415E90: netlink_interface (if_netlink.c:653)
==8579== by 0x41C1FC: netlink_parse_info (kernel_netlink.c:904)
==8579== by 0x4163E8: interface_lookup_netlink (if_netlink.c:760)
==8579== by 0x42BB37: zebra_ns_enable (zebra_ns.c:130)
==8579== by 0x42BC5E: zebra_ns_init (zebra_ns.c:208)
==8579== by 0x4130F4: main (main.c:401)
This can be found by: `ip link del <VRF DEVICE NAME>` then `ip link add <NAME> type vrf table X` again and
then attempting to use the vrf.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Donald Sharp [Fri, 1 Mar 2019 18:56:12 +0000 (13:56 -0500)]
zebra: When installing a new route always use REPLACE
When we install a new route into the kernel always use
REPLACE. Else if the route is already there it can
be translated into an append with the flags we are
using.
This is especially true for the way we handle pbr
routes as that we are re-installing the same route
entry from pbr at the moment.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
vivek [Fri, 1 Mar 2019 07:17:16 +0000 (07:17 +0000)]
bgpd: Recursively determine if route's source is EVPN
With leaking of IPv4 or IPv6 unicast routes whose source is a EVPN
type-2 or type-5 route between VRFs, the determination of whether
the route's source is EVPN has to be made recursively. This is used
during route install to pass along appropriate parameters to zebra.
vivek [Fri, 1 Mar 2019 07:10:53 +0000 (07:10 +0000)]
zebra: Use next hop's VRF for EVPN-based routes
Ensure that the next hop's VRF is used for IPv4 and IPv6 unicast routes
sourced from EVPN routes, for next hop and Router MAC tracking and
install. This way, leaked routes from other instances are handled properly.
vivek [Fri, 1 Mar 2019 06:45:04 +0000 (06:45 +0000)]
bgpd: Correctly identify VPN-imported routes in a VRF
Refine check that looks for VPN routes imported into a VRF because
a VRF can have other imported routes too like IPv4 and IPv6 unicast
routes sourced from EVPN type-2 and type-5 routes.
bgpd: Code to remove the bottleneck in aggregation.
The code that causes the bottleneck has been written generically to
handle the below two cases:
a) When a new aggregate-address is configured.
b) When new routes, that can be aggregated under an existing
aggregate-address, are received.
This change optimizes the code that handles case-(b).
bgpd: Code to handle BGP aggregate's l-communities.
With this commit:
1) The code to manage the large-communities attribute of the routes that are
aggregatable under a configured aggregate-address is introduced.
2) The code to compute the aggregate-route's large-communities attribute is
introduced.
bgpd: Code to handle BGP aggregate's e-communities.
With this commit:
1) The code to manage the extended-communities attribute of the routes that are
aggregatable under a configured aggregate-address is introduced.
2) The code to compute the aggregate-route's extended-communities attribute is
introduced.
With this commit:
1) The code to manage the communities attribute of the routes that are
aggregatable under a configured aggregate-address is introduced.
2) The code to compute the aggregate-route's communities attribute is
introduced.
With this commit:
1) 'struct bgp_aggregate' is moved to bgp_route.h from bgp_route.c
2) Hashes to accommodate the as-path, communities, extended-communities and
large-communities attributes of all the routes aggregated by an
aggregate route is introduced in 'struct bgp_aggregate'.
3) Place-holders for the aggregate route's as-path, communities,
extended-communities and large-communities attributes are introduced in
'struct bgp_aggregate'.
4) The code to manage the as-path of the routes that are aggregatable under
a configured aggregate-address is introduced.
5) The code to compute the aggregate-route's as-path is introduced.
vivek [Thu, 28 Feb 2019 11:18:10 +0000 (11:18 +0000)]
bgpd: Remove route sub-type checks in route-leak withdraw
The check on which routes are exportable is a superset, so remove the
route sub-type checks. Also, this change is needed to handle EVPN-imported
leaked routes correctly.
vivek [Thu, 28 Feb 2019 11:11:01 +0000 (11:11 +0000)]
bgpd: No nexthop tracking for EVPN-imported leaked routes
IPv4 or IPv6 unicast routes which are imported from EVPN routes
(type-2 or type-5) and installed in a BGP instance and then leaked
do not need any nexthop tracking, as any tracking should happen in
the source instance.
vivek [Thu, 28 Feb 2019 08:30:51 +0000 (08:30 +0000)]
bgpd: Refine check for which routes can be exported into VPN
A non-imported route or a non-VPN imported route is a candidate to be
exported into the VPN routing table for leaking to other BGP instances
or advertisement into BGP/MPLS VPN. The former is a local or learnt
IPv4 or IPv6 route. The latter is an IPv4 or IPv6 route that is based
on a received EVPN type-2 or type-5 route.
Implement a function to specify if a route can be exported into VPN
and use in the appropriate places.
projects / mirror / frr.git / log