Fix rib generation count for fib algo.

Currently, PCB caching mechanism relies on the rib generation
 counter (rnh_gen) to invalidate cached nhops/LLE entries.

With certain fib algorithms, it is now possible that the
 datapath lookup state applies RIB changes with some delay.
In that scenario, PCB cache will invalidate on the RIB change,
 but the new lookup may result in the same nexthop being returned.
When fib algo finally gets in sync with the RIB changes, PCB cache
 will not receive any notification and will end up caching the stale data.

To fix this, introduce additional counter, rnh_gen_rib, which is used
 only when FIB_ALGO is enabled.
This counter is incremented by the control plane. Each time when fib algo
 synchronises with the RIB, it updates rnh_gen to the current rnh_gen_rib value.

Differential Revision: https://reviews.freebsd.org/D29812
Reviewed by:	donner
MFC after:	2 weeks

3 files changed, 29 insertions, 5 deletions

diff --git a/sys/net/route/fib_algo.c b/sys/net/route/fib_algo.c
index 30e715e1e1ef..91565d727a9c 100644
--- a/sys/net/route/fib_algo.c
+++ b/sys/net/route/fib_algo.c
@@ -525,6 +525,13 @@ schedule_fd_rebuild(struct fib_data *fd, const char *reason)
 	}
 }
 
+static void
+sync_rib_gen(struct fib_data *fd)
+{
+	FD_PRINTF(LOG_DEBUG, fd, "Sync gen %u -> %u", fd->fd_rh->rnh_gen, fd->fd_rh->rnh_gen_rib);
+	fd->fd_rh->rnh_gen = fd->fd_rh->rnh_gen_rib;
+}
+
 static int64_t
 get_tv_diff_ms(const struct timeval *old_tv, const struct timeval *new_tv)
 {
@@ -680,6 +687,7 @@ apply_rtable_changes(struct fib_data *fd)
 	result = fd->fd_flm->flm_change_rib_items_cb(fd->fd_rh, q, fd->fd_algo_data);
 
 	if (result == FLM_SUCCESS) {
+		sync_rib_gen(fd);
 		for (int i = 0; i < q->count; i++)
 			if (q->entries[i].nh_old)
 				fib_unref_nhop(fd, q->entries[i].nh_old);
@@ -811,6 +819,7 @@ handle_rtable_change_cb(struct rib_head *rnh, struct rib_cmd_info *rc,
 
 	switch (result) {
 	case FLM_SUCCESS:
+		sync_rib_gen(fd);
 		/* Unref old nexthop on success */
 		if (rc->rc_nh_old != NULL)
 			fib_unref_nhop(fd, rc->rc_nh_old);
@@ -1236,7 +1245,9 @@ setup_fd_instance(struct fib_lookup_module *flm, struct rib_head *rh,
 		result = try_setup_fd_instance(flm, rh, prev_fd, &new_fd);
 
 		if ((result == FLM_SUCCESS) && attach) {
-			if (!fib_set_datapath_ptr(new_fd, &new_fd->fd_dp))
+			if (fib_set_datapath_ptr(new_fd, &new_fd->fd_dp))
+				sync_rib_gen(new_fd);
+			else
 				result = FLM_REBUILD;
 		}
 
diff --git a/sys/net/route/route_ctl.c b/sys/net/route/route_ctl.c
index fb3dfff87019..a0c4a2283a00 100644
--- a/sys/net/route/route_ctl.c
+++ b/sys/net/route/route_ctl.c
@@ -808,7 +808,7 @@ rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, struct rib_cmd_info
 	rt->rte_flags &= ~RTF_UP;
 
 	/* Finalize notification */
-	rnh->rnh_gen++;
+	rib_bump_gen(rnh);
 	rnh->rnh_prefixes--;
 
 	rc->rc_cmd = RTM_DELETE;
@@ -1060,7 +1060,7 @@ add_route_nhop(struct rib_head *rnh, struct rtentry *rt,
 			tmproutes_update(rnh, rt);
 
 		/* Finalize notification */
-		rnh->rnh_gen++;
+		rib_bump_gen(rnh);
 		rnh->rnh_prefixes++;
 
 		rc->rc_cmd = RTM_ADD;
@@ -1116,7 +1116,7 @@ change_route_nhop(struct rib_head *rnh, struct rtentry *rt,
 	}
 
 	/* Finalize notification */
-	rnh->rnh_gen++;
+	rib_bump_gen(rnh);
 	if (rnd->rnd_nhop == NULL)
 		rnh->rnh_prefixes--;
 
diff --git a/sys/net/route/route_var.h b/sys/net/route/route_var.h
index 427c286a5090..f12931476fd3 100644
--- a/sys/net/route/route_var.h
+++ b/sys/net/route/route_var.h
@@ -60,7 +60,7 @@ struct rib_head {
 	rn_walktree_t		*rnh_walktree;	/* traverse tree */
 	rn_walktree_from_t	*rnh_walktree_from; /* traverse tree below a */
 	rnh_preadd_entry_f_t	*rnh_preadd;	/* hook to alter record prior to insertion */
-	rt_gen_t		rnh_gen;	/* generation counter */
+	rt_gen_t		rnh_gen;	/* datapath generation counter */
 	int			rnh_multipath;	/* multipath capable ? */
 	struct radix_node	rnh_nodes[3];	/* empty tree for common case */
 	struct rmlock		rib_lock;	/* config/data path lock */
@@ -71,6 +71,9 @@ struct rib_head {
 	struct callout		expire_callout;	/* Callout for expiring dynamic routes */
 	time_t			next_expire;	/* Next expire run ts */
 	uint32_t		rnh_prefixes;	/* Number of prefixes */
+#ifdef FIB_ALGO
+	rt_gen_t		rnh_gen_rib;	/* rib generation counter */
+#endif
 	uint32_t		rib_dying:1;	/* rib is detaching */
 	uint32_t		rib_algo_fixed:1;/* fixed algorithm */
 	struct nh_control	*nh_control;	/* nexthop subsystem data */
@@ -116,6 +119,16 @@ CHK_STRUCT_ROUTE_FIELDS(_ro_new);						\
 _Static_assert(__offsetof(struct route, ro_dst) == __offsetof(_ro_new, _dst_new),\
 		"ro_dst and " #_dst_new " are at different offset")
 
+static inline void
+rib_bump_gen(struct rib_head *rnh)
+{
+#ifdef FIB_ALGO
+	rnh->rnh_gen_rib++;
+#else
+	rnh->rnh_gen++;
+#endif
+}
+
 struct rib_head *rt_tables_get_rnh(uint32_t table, sa_family_t family);
 int rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum);
 struct rib_cmd_info;