- dev - lists.frrouting.org

[cmaster-next] [PATCH 1/3] zebra: fix FIB route updates on *BSD
by Renato Westphal 16 Dec '16

16 Dec '16

On *BSD, we update a route in the FIB by removing the old one and then installing the new version. With that said, on kernel_route_rib() we need to provide a pointer to both the old version and the new version of the route. We were, however, passing a pointer to the new version to both the 'old' and 'new' parameters. This is not a problem on Linux, which uses NLM_F_REPLACE to update routes, but it breaks route updates on *BSD because the 'old' parameter points to a route that is not installed in the kernel. The kernel_route_rib() function then fails to uninstall the supposedly 'old' route and can fail to install the new version as well if the kernel doesn't support ECMP (e.g. FreeBSD with default configuration). Signed-off-by: Renato Westphal <renato(a)opensourcerouting.org> --- zebra/rib.h | 2 +- zebra/zebra_rib.c | 12 ++++++------ zebra/zebra_static.c | 2 +- 3 files changed, 8 insertions(+), 8 deletions(-) diff --git a/zebra/rib.h b/zebra/rib.h index 30929f1..a30c093 100644 --- a/zebra/rib.h +++ b/zebra/rib.h @@ -335,7 +335,7 @@ extern int zebra_check_addr (struct prefix *p); extern void rib_addnode (struct route_node *rn, struct rib *rib, int process); extern void rib_delnode (struct route_node *rn, struct rib *rib); -extern int rib_install_kernel (struct route_node *rn, struct rib *rib, int update); +extern int rib_install_kernel (struct route_node *rn, struct rib *rib, struct rib *old); extern int rib_uninstall_kernel (struct route_node *rn, struct rib *rib); /* NOTE: diff --git a/zebra/zebra_rib.c b/zebra/zebra_rib.c index c573f6b..a792bcc 100644 --- a/zebra/zebra_rib.c +++ b/zebra/zebra_rib.c @@ -1212,7 +1212,7 @@ nexthop_active_update (struct route_node *rn, struct rib *rib, int set) * is only used for IPv4. */ int -rib_install_kernel (struct route_node *rn, struct rib *rib, int update) +rib_install_kernel (struct route_node *rn, struct rib *rib, struct rib *old) { int ret = 0; struct nexthop *nexthop, *tnexthop; @@ -1231,7 +1231,7 @@ rib_install_kernel (struct route_node *rn, struct rib *rib, int update) * the kernel. */ zfpm_trigger_update (rn, "installing in kernel"); - ret = kernel_route_rib (&rn->p, update ? rib : NULL, rib); + ret = kernel_route_rib (&rn->p, old, rib); /* If install succeeds, update FIB flag for nexthops. */ if (!ret) @@ -1388,7 +1388,7 @@ rib_process_add_fib(struct zebra_vrf *zvrf, struct route_node *rn, if (!RIB_SYSTEM_ROUTE (new)) { - if (rib_install_kernel (rn, new, 0)) + if (rib_install_kernel (rn, new, NULL)) { inet_ntop (rn->p.family, &rn->p.u.prefix, buf, INET6_ADDRSTRLEN); zlog_warn ("%u:%s/%d: Route install failed", @@ -1470,7 +1470,7 @@ rib_process_update_fib (struct zebra_vrf *zvrf, struct route_node *rn, /* Non-system route should be installed. */ if (!RIB_SYSTEM_ROUTE (new)) { - if (rib_install_kernel (rn, new, 1)) + if (rib_install_kernel (rn, new, old)) { installed = 0; inet_ntop (rn->p.family, &rn->p.u.prefix, buf, INET6_ADDRSTRLEN); @@ -1542,7 +1542,7 @@ rib_process_update_fib (struct zebra_vrf *zvrf, struct route_node *rn, break; } if (!in_fib) - rib_install_kernel (rn, new, 0); + rib_install_kernel (rn, new, NULL); } } @@ -2779,7 +2779,7 @@ rib_delete (afi_t afi, safi_t safi, vrf_id_t vrf_id, int type, u_short instance, { /* This means someone else, other than Zebra, has deleted * a Zebra router from the kernel. We will add it back */ - rib_install_kernel(rn, fib, 0); + rib_install_kernel(rn, fib, NULL); } } else diff --git a/zebra/zebra_static.c b/zebra/zebra_static.c index e6ae9c7..7c37d6b 100644 --- a/zebra/zebra_static.c +++ b/zebra/zebra_static.c @@ -323,7 +323,7 @@ static_uninstall_route (afi_t afi, safi_t safi, struct prefix *p, struct static_ { /* Update route in kernel if it's in fib */ if (CHECK_FLAG(rib->status, RIB_ENTRY_SELECTED_FIB)) - rib_install_kernel (rn, rib, 1); + rib_install_kernel (rn, rib, rib); /* Update redistribution if it's selected */ if (CHECK_FLAG(rib->flags, ZEBRA_FLAG_SELECTED)) redistribute_update (&rn->p, rib, NULL); -- 1.9.1

1 2

[cmaster-next] [PATCH 00/10] Issue #12
by Renato Westphal 15 Dec '16

15 Dec '16

Hi all, This patchset makes ldpd display everything in order on CLI output (interfaces, label bindings, etc). This is important to keep consistent with the other routing daemons and to make automated testing easier (CLI output should be predictable and not dependant on the order that things were created). To do this, I converted several linked lists to red-black trees, which store elements in order. I could also use ordered lists, but they have worse time-complexities than red-black trees for basically everything. Fixes issue #12 and addresses a long standing request from Martin. Renato Westphal (10): ldpd: fix segfault when configuring multiple pseudowires ldpd: allow multiple link adjacencies with unnumbered interfaces ldpd: use red-black trees to store 'lde_map' elements ldpd: use red-black trees to store 'iface' elements ldpd: use red-black trees to store 'tnbr' elements ldpd: use red-black trees to store 'nbr_params' elements ldpd: use red-black trees to store 'l2vpn' elements ldpd: use red-black trees to store 'l2vpn_if' elements ldpd: use red-black trees to store 'l2vpn_pw' elements ldpd: use red-black trees to store 'adj' elements ldpd/adjacency.c | 116 +++++++++++++++++----------- ldpd/hello.c | 6 +- ldpd/interface.c | 45 ++++++----- ldpd/l2vpn.c | 108 ++++++++++++++------------ ldpd/lde.c | 49 +++++++----- ldpd/lde.h | 9 ++- ldpd/lde_lib.c | 18 ++--- ldpd/ldp_debug.c | 4 +- ldpd/ldp_vty_conf.c | 104 +++++++++++++------------- ldpd/ldpd.c | 212 ++++++++++++++++++++++++++-------------------------- ldpd/ldpd.h | 72 +++++++++++------- ldpd/ldpe.c | 56 +++++++------- ldpd/ldpe.h | 14 ++-- ldpd/neighbor.c | 29 ++++--- 14 files changed, 468 insertions(+), 374 deletions(-) -- 1.9.1

1 10

[cmaster-next] Name Survey Results (just for reference)
by Martin Winter 15 Dec '16

15 Dec '16

Just for everyones reference, here is the result from the survey as a spreadsheet https://docs.google.com/spreadsheets/d/1QlplLz0t7I4xbGuYa24WB-DwLLqYDCwXG6E… - Martin

1 0

[cmaster-next] Slack - upgrade to standard edition?
by Martin Winter 15 Dec '16

15 Dec '16

Donald, (and everyone else working on it) Thanks for setting up Slack. We (NetDEF) can get a free upgrade to the standard edition as part of bing a 501c3 non-profit. See https://freerangerouting.slack.com/pricing for difference between Free and Standard model. Happy to do this if this community thinks Slack is useful. Please be aware that it’s free “only” up to 250 users - and 85% discount above it. We can only do this for one Slack - so we may want to wait until everyone confirms the new selected name (i.e. no legal issues) - Martin

2 2

[cmaster-next] [PATCH 3/3] bgp: garbage "neighbor (af 54816) remote-as internal" peer in show run
by Donald Sharp 15 Dec '16

15 Dec '16

From: Daniel Walton <dwalton(a)cumulusnetworks.com> Signed-off-by: Daniel Walton <dwalton(a)cumulusnetworks.com> Reviewed-by: vivek(a)cumulusnetworks.com Ticket: CM-14065 (cherry picked from commit 82e528daec27ca87323dab9e60d564b2a3cc0c40) --- bgpd/bgp_vty.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/bgpd/bgp_vty.c b/bgpd/bgp_vty.c index 7def8eb..abfe9b8 100644 --- a/bgpd/bgp_vty.c +++ b/bgpd/bgp_vty.c @@ -2870,7 +2870,7 @@ peer_conf_interface_get (struct vty *vty, const char *conf_if, afi_t afi, if (peer) { if (as_str) - ret = peer_remote_as (bgp, &su, NULL, &as, as_type, afi, safi); + ret = peer_remote_as (bgp, &su, conf_if, &as, as_type, afi, safi); } else { -- 2.5.5

1 0

[cmaster-next] [PATCH 2/3] bgpd: "neighbor swpX interface remote-as XYZ" is ignored
by Donald Sharp 15 Dec '16

15 Dec '16

From: Daniel Walton <dwalton(a)cumulusnetworks.com> Signed-off-by: Daniel Walton <dwalton(a)cumulusnetworks.com> Reviewed-by: Don Slice <dslice(a)cumulusnetworks.com> Ticket: CM-13979 (cherry picked from commit 7d308ec5b45bf66d241535b23c6413e9cb2c9435) --- bgpd/bgp_vty.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/bgpd/bgp_vty.c b/bgpd/bgp_vty.c index 5dafac1..7def8eb 100644 --- a/bgpd/bgp_vty.c +++ b/bgpd/bgp_vty.c @@ -2867,7 +2867,12 @@ peer_conf_interface_get (struct vty *vty, const char *conf_if, afi_t afi, } peer = peer_lookup_by_conf_if (bgp, conf_if); - if (!peer) + if (peer) + { + if (as_str) + ret = peer_remote_as (bgp, &su, NULL, &as, as_type, afi, safi); + } + else { if (bgp_flag_check (bgp, BGP_FLAG_NO_DEFAULT_IPV4) && afi == AFI_IP && safi == SAFI_UNICAST) @@ -2891,8 +2896,9 @@ peer_conf_interface_get (struct vty *vty, const char *conf_if, afi_t afi, } peer_flag_set (peer, PEER_FLAG_CAPABILITY_ENHE); } - else if ((v6only && !CHECK_FLAG(peer->flags, PEER_FLAG_IFPEER_V6ONLY)) || - (!v6only && CHECK_FLAG(peer->flags, PEER_FLAG_IFPEER_V6ONLY))) + + if ((v6only && !CHECK_FLAG(peer->flags, PEER_FLAG_IFPEER_V6ONLY)) || + (!v6only && CHECK_FLAG(peer->flags, PEER_FLAG_IFPEER_V6ONLY))) { if (v6only) SET_FLAG(peer->flags, PEER_FLAG_IFPEER_V6ONLY); -- 2.5.5

1 0

[cmaster-next] [PATCH 1/3] zebrad: Correct problem removing dead routes from rib due to NHT issue
by Donald Sharp 15 Dec '16

15 Dec '16

From: Don Slice <dslice(a)cumulusnetworks.com> Fix for another issue with next-hop tracking for overlapping static routes created a problem removing the dead routes. This fix corrects that error. Ticket: CM-13710 Signed-off-by: Don Slice Reviewed By: Donald Sharp Testing Done: ospf-smoke, bgp-smoke --- zebra/zebra_rnh.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/zebra/zebra_rnh.c b/zebra/zebra_rnh.c index c4c11f9..f5a9571 100644 --- a/zebra/zebra_rnh.c +++ b/zebra/zebra_rnh.c @@ -546,7 +546,7 @@ zebra_rnh_process_static_routes (vrf_id_t vrfid, int family, { RNODE_FOREACH_RIB(static_rn, srib) { - if (srib->type == ZEBRA_ROUTE_STATIC) + if (srib->type != ZEBRA_ROUTE_STATIC) continue; /* Set the filter flag for the correct nexthop - static route may -- 2.5.5

1 0

[cmaster-next] [PATCH 0/3] stable/2.0 - Some bug fixes for stable/2.0
by Donald Sharp 15 Dec '16

15 Dec '16

During a final testing cycle we have found these 3 bugs FRR. Daniel Walton (2): bgpd: "neighbor swpX interface remote-as XYZ" is ignored bgp: garbage "neighbor (af 54816) remote-as internal" peer in show run Don Slice (1): zebrad: Correct problem removing dead routes from rib due to NHT issue bgpd/bgp_vty.c | 12 +++++++++--- zebra/zebra_rnh.c | 2 +- 2 files changed, 10 insertions(+), 4 deletions(-) -- 2.5.5

1 0

[cmaster-next] [PATCH 9/9] lib: Allow json to work across different versions.
by Donald Sharp 15 Dec '16

15 Dec '16

The json_object_object_get_ex function is not fully available across all versions of json. Write a wrapper to allow it to work. Ticket: CM-13872 Signed-off-by: Donald Sharp <sharpd(a)cumulusnetworks.com> --- lib/json.c | 15 +++++++++++++++ lib/json.h | 4 ++++ 2 files changed, 19 insertions(+) diff --git a/lib/json.c b/lib/json.c index 3edd146..ccbecb7 100644 --- a/lib/json.c +++ b/lib/json.c @@ -87,3 +87,18 @@ json_object_free(struct json_object *obj) { json_object_put(obj); } + +#if !defined(HAVE_JSON_C_JSON_H) +int +json_object_object_get_ex(struct json_object *obj, + const char *key, + struct json_object **value) +{ + *value = json_object_object_get(obj, key); + + if (*value) + return 1; + + return 0; +} +#endif diff --git a/lib/json.h b/lib/json.h index bacc799..7e98614 100644 --- a/lib/json.h +++ b/lib/json.h @@ -32,6 +32,10 @@ * so let's just turn it back to the original usage. */ #define json_object_to_json_string_ext(A, B) json_object_to_json_string (A) + +extern int json_object_object_get_ex(struct json_object *obj, + const char *key, + struct json_object **value); #endif #include "command.h" -- 2.5.5

1 0

[cmaster-next] [PATCH 8/9] lib: Fix clang SA warnings.
by Donald Sharp 15 Dec '16

15 Dec '16

Signed-off-by: Donald Sharp <sharpd(a)cumulusnetworks.com> (cherry picked from commit 0894e4b1d07eaef801c300eb9550dbd3fc32ddd3) --- lib/imsg.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/lib/imsg.c b/lib/imsg.c index 246430c..fc62c13 100644 --- a/lib/imsg.c +++ b/lib/imsg.c @@ -182,7 +182,8 @@ imsg_get(struct imsgbuf *ibuf, struct imsg *imsg) else imsg->fd = -1; - memcpy(imsg->data, ibuf->r.rptr, datalen); + if (imsg->data) + memcpy(imsg->data, ibuf->r.rptr, datalen); if (imsg->hdr.len < av) { left = av - imsg->hdr.len; -- 2.5.5

1 0