Currently if the sysctl net.ipv4.raw_l3mdev_accept is 1, packets
destined to a specific vrf also end up being delivered to the default
vrf. We will see logs like this in ospf:
2021/02/10 21:17:05.245727 OSPF: ospf_recv_packet: fd 20(default) on interface 1265(swp1s1.26)
2021/02/10 21:17:05.245740 OSPF: Hello received from [9.9.36.12] via [swp1s1.26:200.254.26.13]
2021/02/10 21:17:05.245741 OSPF: src [200.254.26.14],
2021/02/10 21:17:05.245743 OSPF: dst [224.0.0.5]
2021/02/10 21:17:05.245769 OSPF: ospf_recv_packet: fd 45(vrf1036) on interface 1265(swp1s1.26)
2021/02/10 21:17:05.245774 OSPF: Hello received from [9.9.36.12] via [swp1s1.26:200.254.26.13]
2021/02/10 21:17:05.245775 OSPF: src [200.254.26.14],
2021/02/10 21:17:05.245777 OSPF: dst [224.0.0.5]
This really really makes ospf unhappy in the vrf we are running in.
I am approaching the problem by just dropping the packet if read in the
default vrf because of:
commit 0556fc33c7
Author: Donald Sharp <sharpd@cumulusnetworks.com>
Date: Fri Feb 1 11:54:59 2019 -0500
lib: Allow bgp to always create a listen socket for the vrf
Effectively if we have `router ospf vrf BLUE` but no ospf running
in the default vrf, we will not have a listener and that would
require a fundamental change in our approach to handle the ospf->fd
at a global level. I think this is less than ideal at the moment
but it will get us moving again and allow FRR to work with
a bunch of vrf's and ospf neighbors.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Modify the timers on the bgp_blackhole_community test to
be more aggressive so our test system will recover faster
when we drop packets.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
`same_attr` has been computed and `hook_call(bgp_process)` (calling
BMP module) would not change it. We could reuse the value to filter
same attribute updates, avoiding an extra comparison.
Signed-off-by: zyxwvu Shi <i@shiyc.cn>
Valgrind is still reporting:
466020-==466020== by 0x11B9F4: main (bfdd.c:403)
466020-==466020== Address 0x5a7d544 is 84 bytes inside a block of size 272 free'd
466020:==466020== at 0x48399AB: free (vg_replace_malloc.c:538)
466020-==466020== by 0x490A947: qfree (memory.c:140)
466020-==466020== by 0x48F2AE8: if_delete (if.c:322)
466020-==466020== by 0x48F250D: if_destroy_via_zapi (if.c:195)
466020-==466020== by 0x497071E: zclient_interface_delete (zclient.c:2040)
466020-==466020== by 0x49745F6: zclient_read (zclient.c:3687)
466020-==466020== by 0x4955AEC: thread_call (thread.c:1684)
466020-==466020== by 0x48FF64E: frr_run (libfrr.c:1126)
466020-==466020== by 0x11B9F4: main (bfdd.c:403)
466020-==466020== Block was alloc'd at
466020:==466020== at 0x483AB65: calloc (vg_replace_malloc.c:760)
466020-==466020== by 0x490A805: qcalloc (memory.c:115)
466020-==466020== by 0x48F23D6: if_new (if.c:160)
466020-==466020== by 0x48F257F: if_create_name (if.c:214)
466020-==466020== by 0x48F3493: if_get_by_name (if.c:558)
466020-==466020== by 0x49705F2: zclient_interface_add (zclient.c:1989)
466020-==466020== by 0x49745E0: zclient_read (zclient.c:3684)
466020-==466020== by 0x4955AEC: thread_call (thread.c:1684)
466020-==466020== by 0x48FF64E: frr_run (libfrr.c:1126)
466020-==466020== by 0x11B9F4: main (bfdd.c:403)
Apparently the bs->ifp pointer is being set even in cases when
the bs->key.ifname is not being set. So go through and just
match the interface pointer and cut-to-the-chase.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Valgrind reports:
469901-==469901==
469901-==469901== Conditional jump or move depends on uninitialised value(s)
469901:==469901== at 0x3A090D: bgp_bfd_dest_update (bgp_bfd.c:416)
469901-==469901== by 0x497469E: zclient_read (zclient.c:3701)
469901-==469901== by 0x4955AEC: thread_call (thread.c:1684)
469901-==469901== by 0x48FF64E: frr_run (libfrr.c:1126)
469901-==469901== by 0x213AB3: main (bgp_main.c:540)
469901-==469901== Uninitialised value was created by a stack allocation
469901:==469901== at 0x3A0725: bgp_bfd_dest_update (bgp_bfd.c:376)
469901-==469901==
469901-==469901== Conditional jump or move depends on uninitialised value(s)
469901:==469901== at 0x3A093C: bgp_bfd_dest_update (bgp_bfd.c:421)
469901-==469901== by 0x497469E: zclient_read (zclient.c:3701)
469901-==469901== by 0x4955AEC: thread_call (thread.c:1684)
469901-==469901== by 0x48FF64E: frr_run (libfrr.c:1126)
469901-==469901== by 0x213AB3: main (bgp_main.c:540)
469901-==469901== Uninitialised value was created by a stack allocation
469901:==469901== at 0x3A0725: bgp_bfd_dest_update (bgp_bfd.c:376)
On looking at bgp_bfd_dest_update the function call into bfd_get_peer_info
when it fails to lookup the ifindex ifp pointer just returns leaving
the dest and src prefix pointers pointing to whatever was passed in.
Let's do two things:
a) The src pointer was sometimes assumed to be passed in and sometimes not.
Forget that. Make it always be passed in
b) memset the src and dst pointers to be all zeros. Then when we look
at either of the pointers we are not making decisions based upon random
data in the pointers.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Just for more debug information regarding malformed aggregator_as.
```
bgpd[5589]: [EC 33554434] 192.168.10.25: AGGREGATOR AS number is 0 for aspath: 65030
bgpd[5589]: bgp_attr_aggregator: attributes: nexthop 192.168.10.25, origin i, path 65030
```
Signed-off-by: Donatas Abraitis <donatas.abraitis@gmail.com>
Already not necessary, because if BGP aggregator AS attribute is with
value of 0, then the attribute is already discarded at early processing.
Signed-off-by: Donatas Abraitis <donatas.abraitis@gmail.com>
Avoid mangling packet size which is expected to be the same as received.
Stream pointer advancing is necessary to avoid changing the packet and
reseting BGP sessions.
Signed-off-by: Donatas Abraitis <donatas.abraitis@gmail.com>
People keep asking about the default unreachable route
in the linux vrf table. Add a bit of color about the
design choices and what is going on.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Only one of the four reference files was present; add the missing
three. The test just silently passed if a ref file was missing:
change that to a failure.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Wrote a little guide for cross-compiling FRR, gleaned from notes I took
while compiling for a RPi 3B+ on a Gentoo x86_64 system.
Care was taken to keep this documentation as generic as possible so
these steps could be applied to any cross-compile targeting a supported
architecture.
Signed-off-by: Wesley Coakley <wcoakley@nvidia.com>
An UPDATE message that contains the AS number of zero in the AS_PATH
or AGGREGATOR attribute MUST be considered as malformed and be
handled by the procedures specified in [RFC7606].
An UPDATE message with a malformed AGGREGATOR attribute SHALL be
handled using the approach of "attribute discard".
Attribute discard: In this approach, the malformed attribute MUST
be discarded and the UPDATE message continues to be processed.
This approach MUST NOT be used except in the case of an attribute
that has no effect on route selection or installation.
Signed-off-by: Donatas Abraitis <donatas.abraitis@gmail.com>
the old VXLAN function for local MAC deletion was still in
existence and being called from the VXLAN code whilst the new
generic function was not being called at all. Resolve this so
the generic function matches the old function and is called
exclusively.
Signed-off-by: Pat Ruddy <pat@voltanet.io>
This didn't exist yet when the xref code came around, and since
frrtrace() gets collapsed to nothing by the preprocessor when
tracepoints are disabled, it didn't cause any compiler errors...
Signed-off-by: David Lamparter <equinox@diac24.net>
gcc fucks up global variables with section attributes when they're used
in templated C++ code. The template instantiation "magic" kinda breaks
down (it's implemented through COMDAT in the linker, which clashes with
the section attribute.)
The workaround provides full runtime functionality, but the xref
extraction tool (xrelfo.py) won't work on C++ code compiled by GCC.
FWIW, clang gets this right.
Signed-off-by: David Lamparter <equinox@diac24.net>
