ip: Path MTU

Type: feature

Support setting the MTU for a peer on an interface. The minimum value of
the path and interface MTU is used at forwarding time.

the path MTU is specified for a given peer, by address and table-ID.
In the forwarding plane the MTU is enfored either:
 1 - if the peer is attached, then the MTU is set on the peer's
adjacency
 2 - if the peer is not attached, it is remote, then a DPO is added to
the peer's FIB entry to perform the necessary fragmentation.

Signed-off-by: Neale Ranns <neale@graphiant.com>
Change-Id: I8b9ea6a07868b50e97e2561f18d9335407dea7ae

4 files changed, 79 insertions, 0 deletions

diff --git a/src/vnet/dpo/dpo.c b/src/vnet/dpo/dpo.c
index 1331b5501bc..d8342ff17ae 100644
--- a/src/vnet/dpo/dpo.c
+++ b/src/vnet/dpo/dpo.c
@@ -23,6 +23,8 @@
  * The VLIB graph nodes are graph of types, the DPO graph is a graph of instances.
  */
 
+// clang-format off
+
 #include <vnet/dpo/dpo.h>
 #include <vnet/ip/lookup.h>
 #include <vnet/ip/format.h>
@@ -395,6 +397,18 @@ dpo_get_urpf(const dpo_id_t *dpo)
     return (~0);
 }
 
+u16
+dpo_get_mtu(const dpo_id_t *dpo)
+{
+    if (dpo_id_is_valid(dpo) &&
+        (NULL != dpo_vfts[dpo->dpoi_type].dv_get_mtu))
+    {
+        return (dpo_vfts[dpo->dpoi_type].dv_get_mtu(dpo));
+    }
+
+    return (0xffff);
+}
+
 static u32
 dpo_get_next_node (dpo_type_t child_type,
                    dpo_proto_t child_proto,
@@ -649,3 +663,5 @@ VLIB_CLI_COMMAND (show_fib_memory, static) = {
     .short_help = "show dpo memory",
 };
 /* *INDENT-ON* */
+
+// clang-format on
diff --git a/src/vnet/dpo/dpo.h b/src/vnet/dpo/dpo.h
index ee4990d0058..e9976c2dd87 100644
--- a/src/vnet/dpo/dpo.h
+++ b/src/vnet/dpo/dpo.h
@@ -24,6 +24,8 @@
  * instances.
  */
 
+// clang-format off
+
 #ifndef __DPO_H__
 #define __DPO_H__
 
@@ -362,6 +364,16 @@ extern void dpo_stack_from_node(u32 child_node,
 extern u32 dpo_get_urpf(const dpo_id_t *dpo);
 
 /**
+ * Get the MTU DPO
+ *
+ * @param dpo
+ *  The DPO from which to get the MTU
+ *
+ * @return MTU (0xffff if something more usefull was unavailable)
+ */
+extern u16 dpo_get_mtu(const dpo_id_t *dpo);
+
+/**
  * @brief  A lock function registered for a DPO type
  */
 typedef void (*dpo_lock_fn_t)(dpo_id_t *dpo);
@@ -389,6 +401,11 @@ typedef u32* (*dpo_get_next_node_t)(const dpo_id_t *dpo);
 typedef u32 (*dpo_get_urpf_t)(const dpo_id_t *dpo);
 
 /**
+ * @brief Given a DPO instance return the MTU
+ */
+typedef u16 (*dpo_get_mtu_t)(const dpo_id_t *dpo);
+
+/**
  * @brief Called during FIB interposition when the originally
  * registered DPO is used to 'clone' an instance for interposition
  * at a particular location in the FIB graph.
@@ -433,6 +450,10 @@ typedef struct dpo_vft_t_
      */
     dpo_get_urpf_t dv_get_urpf;
     /**
+     * Get MTU
+     */
+    dpo_get_mtu_t dv_get_mtu;
+    /**
      * Signal on an interposed child that the parent has changed
      */
     dpo_mk_interpose_t dv_mk_interpose;
@@ -548,3 +569,5 @@ do {                                                                    \
     if ((YESNO)) vlib_worker_thread_barrier_release((VM));
 
 #endif
+
+// clang-format on
diff --git a/src/vnet/dpo/load_balance.c b/src/vnet/dpo/load_balance.c
index fb876a09ec2..a212532dffd 100644
--- a/src/vnet/dpo/load_balance.c
+++ b/src/vnet/dpo/load_balance.c
@@ -25,6 +25,8 @@
 #include <vnet/ip/ip4_inlines.h>
 #include <vnet/ip/ip6_inlines.h>
 
+// clang-format off
+
 /*
  * distribution error tolerance for load-balancing
  */
@@ -918,11 +920,30 @@ load_balance_mem_show (void)
     load_balance_map_show_mem();
 }
 
+static u16
+load_balance_dpo_get_mtu (const dpo_id_t *dpo)
+{
+    const dpo_id_t *buckets;
+    load_balance_t *lb;
+    u16 i, mtu = 0xffff;
+
+    lb = load_balance_get(dpo->dpoi_index);
+    buckets = load_balance_get_buckets(lb);
+
+    for (i = 0; i < lb->lb_n_buckets; i++)
+    {
+        mtu = clib_min (mtu, dpo_get_mtu (&buckets[i]));
+    }
+
+    return (mtu);
+}
+
 const static dpo_vft_t lb_vft = {
     .dv_lock = load_balance_lock,
     .dv_unlock = load_balance_unlock,
     .dv_format = format_load_balance_dpo,
     .dv_mem_show = load_balance_mem_show,
+    .dv_get_mtu = load_balance_dpo_get_mtu,
 };
 
 /**
@@ -1323,3 +1344,5 @@ VLIB_REGISTER_NODE (bier_load_balance_node) = {
   .format_trace = format_bier_load_balance_trace,
   .sibling_of = "mpls-load-balance",
 };
+
+// clang-format on
diff --git a/src/vnet/dpo/mpls_label_dpo.c b/src/vnet/dpo/mpls_label_dpo.c
index 683b5449513..b87cb1efcd2 100644
--- a/src/vnet/dpo/mpls_label_dpo.c
+++ b/src/vnet/dpo/mpls_label_dpo.c
@@ -18,6 +18,8 @@
 #include <vnet/mpls/mpls.h>
 #include <vnet/dpo/drop_dpo.h>
 
+// clang-format off
+
 #ifndef CLIB_MARCH_VARIANT
 /*
  * pool of all MPLS Label DPOs
@@ -1213,12 +1215,25 @@ mpls_label_interpose (const dpo_id_t *original,
             mpls_label_dpo_get_index(mld_clone));
 }
 
+static u16
+mpls_label_dpo_get_mtu (const dpo_id_t *dpo)
+{
+    mpls_label_dpo_t *mld;
+
+    mld = mpls_label_dpo_get(dpo->dpoi_index);
+
+    /* return the parent's MTU minus the amount of header
+     * this DPO imposes */
+    return (dpo_get_mtu (&mld->mld_dpo) - sizeof(mpls_label_t) * mld->mld_n_labels);
+}
+
 const static dpo_vft_t mld_vft = {
     .dv_lock = mpls_label_dpo_lock,
     .dv_unlock = mpls_label_dpo_unlock,
     .dv_format = format_mpls_label_dpo,
     .dv_mem_show = mpls_label_dpo_mem_show,
     .dv_mk_interpose = mpls_label_interpose,
+    .dv_get_mtu = mpls_label_dpo_get_mtu,
 };
 
 const static char* const mpls_label_imp_pipe_ip4_nodes[] =
@@ -1337,3 +1352,5 @@ mpls_label_dpo_get_type (mpls_label_dpo_flags_t flags)
     return (mpls_label_dpo_types[flags]);
 }
 #endif /* CLIB_MARCH_VARIANT */
+
+// clang-format on