arp.c revision cbe25aab
1/*
2 * ethernet/arp.c: IP v4 ARP node
3 *
4 * Copyright (c) 2010 Cisco and/or its affiliates.
5 * Licensed under the Apache License, Version 2.0 (the "License");
6 * you may not use this file except in compliance with the License.
7 * You may obtain a copy of the License at:
8 *
9 *     http://www.apache.org/licenses/LICENSE-2.0
10 *
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
16 */
17
18#include <vnet/arp/arp.h>
19#include <vnet/arp/arp_packet.h>
20
21#include <vnet/fib/ip4_fib.h>
22#include <vnet/fib/fib_entry_src.h>
23#include <vnet/adj/adj_nbr.h>
24#include <vnet/adj/adj_mcast.h>
25
26#include <vnet/ip-neighbor/ip_neighbor.h>
27#include <vnet/ip-neighbor/ip_neighbor_dp.h>
28
29#include <vlibmemory/api.h>
30
31/**
32 * @file
33 * @brief IPv4 ARP.
34 *
35 * This file contains code to manage the IPv4 ARP tables (IP Address
36 * to MAC Address lookup).
37 */
38
39/**
40 * @brief Per-interface ARP configuration and state
41 */
42typedef struct ethernet_arp_interface_t_
43{
44  /**
45   * Is ARP enabled on this interface
46   */
47  u32 enabled;
48} ethernet_arp_interface_t;
49
50typedef struct
51{
52  /* Hash tables mapping name to opcode. */
53  uword *opcode_by_name;
54
55  /** Per interface state */
56  ethernet_arp_interface_t *ethernet_arp_by_sw_if_index;
57
58  /* ARP feature arc index */
59  u8 feature_arc_index;
60} ethernet_arp_main_t;
61
62static ethernet_arp_main_t ethernet_arp_main;
63
64static const u8 vrrp_prefix[] = { 0x00, 0x00, 0x5E, 0x00, 0x01 };
65
66static uword
67unformat_ethernet_arp_opcode_host_byte_order (unformat_input_t * input,
68					      va_list * args)
69{
70  int *result = va_arg (*args, int *);
71  ethernet_arp_main_t *am = &ethernet_arp_main;
72  int x, i;
73
74  /* Numeric opcode. */
75  if (unformat (input, "0x%x", &x) || unformat (input, "%d", &x))
76    {
77      if (x >= (1 << 16))
78	return 0;
79      *result = x;
80      return 1;
81    }
82
83  /* Named type. */
84  if (unformat_user (input, unformat_vlib_number_by_name,
85		     am->opcode_by_name, &i))
86    {
87      *result = i;
88      return 1;
89    }
90
91  return 0;
92}
93
94static uword
95unformat_ethernet_arp_opcode_net_byte_order (unformat_input_t * input,
96					     va_list * args)
97{
98  int *result = va_arg (*args, int *);
99  if (!unformat_user
100      (input, unformat_ethernet_arp_opcode_host_byte_order, result))
101    return 0;
102
103  *result = clib_host_to_net_u16 ((u16) * result);
104  return 1;
105}
106
107typedef struct
108{
109  u8 packet_data[64];
110} ethernet_arp_input_trace_t;
111
112static u8 *
113format_ethernet_arp_input_trace (u8 * s, va_list * va)
114{
115  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*va, vlib_main_t *);
116  CLIB_UNUSED (vlib_node_t * node) = va_arg (*va, vlib_node_t *);
117  ethernet_arp_input_trace_t *t = va_arg (*va, ethernet_arp_input_trace_t *);
118
119  s = format (s, "%U",
120	      format_ethernet_arp_header,
121	      t->packet_data, sizeof (t->packet_data));
122
123  return s;
124}
125
126static int
127arp_is_enabled (ethernet_arp_main_t * am, u32 sw_if_index)
128{
129  if (vec_len (am->ethernet_arp_by_sw_if_index) <= sw_if_index)
130    return 0;
131
132  return (am->ethernet_arp_by_sw_if_index[sw_if_index].enabled);
133}
134
135static void
136arp_enable (ethernet_arp_main_t * am, u32 sw_if_index)
137{
138  if (arp_is_enabled (am, sw_if_index))
139    return;
140
141  vec_validate (am->ethernet_arp_by_sw_if_index, sw_if_index);
142
143  am->ethernet_arp_by_sw_if_index[sw_if_index].enabled = 1;
144
145  vnet_feature_enable_disable ("arp", "arp-reply", sw_if_index, 1, NULL, 0);
146  vnet_feature_enable_disable ("arp", "arp-disabled", sw_if_index, 0, NULL,
147			       0);
148}
149
150static void
151arp_disable (ethernet_arp_main_t * am, u32 sw_if_index)
152{
153  if (!arp_is_enabled (am, sw_if_index))
154    return;
155
156  vnet_feature_enable_disable ("arp", "arp-disabled", sw_if_index, 1, NULL,
157			       0);
158  vnet_feature_enable_disable ("arp", "arp-reply", sw_if_index, 0, NULL, 0);
159
160  am->ethernet_arp_by_sw_if_index[sw_if_index].enabled = 0;
161}
162
163static int
164arp_unnumbered (vlib_buffer_t * p0,
165		u32 input_sw_if_index, u32 conn_sw_if_index)
166{
167  vnet_main_t *vnm = vnet_get_main ();
168  vnet_interface_main_t *vim = &vnm->interface_main;
169  vnet_sw_interface_t *si;
170
171  /* verify that the input interface is unnumbered to the connected.
172   * the connected interface is the interface on which the subnet is
173   * configured */
174  si = &vim->sw_interfaces[input_sw_if_index];
175
176  if (!(si->flags & VNET_SW_INTERFACE_FLAG_UNNUMBERED &&
177	(si->unnumbered_sw_if_index == conn_sw_if_index)))
178    {
179      /* the input interface is not unnumbered to the interface on which
180       * the sub-net is configured that covers the ARP request.
181       * So this is not the case for unnumbered.. */
182      return 0;
183    }
184
185  return !0;
186}
187
188always_inline u32
189arp_learn (u32 sw_if_index,
190	   const ethernet_arp_ip4_over_ethernet_address_t * addr)
191{
192  ip_neighbor_learn_t l = {
193    .ip.ip4 = addr->ip4,
194    .type = IP46_TYPE_IP4,
195    .mac = addr->mac,
196    .sw_if_index = sw_if_index,
197  };
198
199  ip_neighbor_learn_dp (&l);
200
201  return (ETHERNET_ARP_ERROR_l3_src_address_learned);
202}
203
204typedef enum arp_input_next_t_
205{
206  ARP_INPUT_NEXT_DROP,
207  ARP_INPUT_NEXT_DISABLED,
208  ARP_INPUT_N_NEXT,
209} arp_input_next_t;
210
211static uword
212arp_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
213{
214  u32 n_left_from, next_index, *from, *to_next, n_left_to_next;
215  ethernet_arp_main_t *am = &ethernet_arp_main;
216
217  from = vlib_frame_vector_args (frame);
218  n_left_from = frame->n_vectors;
219  next_index = node->cached_next_index;
220
221  if (node->flags & VLIB_NODE_FLAG_TRACE)
222    vlib_trace_frame_buffers_only (vm, node, from, frame->n_vectors,
223				   /* stride */ 1,
224				   sizeof (ethernet_arp_input_trace_t));
225
226  while (n_left_from > 0)
227    {
228      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
229
230      while (n_left_from > 0 && n_left_to_next > 0)
231	{
232	  const ethernet_arp_header_t *arp0;
233	  arp_input_next_t next0;
234	  vlib_buffer_t *p0;
235	  u32 pi0, error0;
236
237	  pi0 = to_next[0] = from[0];
238	  from += 1;
239	  to_next += 1;
240	  n_left_from -= 1;
241	  n_left_to_next -= 1;
242
243	  p0 = vlib_get_buffer (vm, pi0);
244	  arp0 = vlib_buffer_get_current (p0);
245
246	  error0 = ETHERNET_ARP_ERROR_replies_sent;
247	  next0 = ARP_INPUT_NEXT_DROP;
248
249	  error0 =
250	    (arp0->l2_type !=
251	     clib_net_to_host_u16 (ETHERNET_ARP_HARDWARE_TYPE_ethernet) ?
252	     ETHERNET_ARP_ERROR_l2_type_not_ethernet : error0);
253	  error0 =
254	    (arp0->l3_type !=
255	     clib_net_to_host_u16 (ETHERNET_TYPE_IP4) ?
256	     ETHERNET_ARP_ERROR_l3_type_not_ip4 : error0);
257	  error0 =
258	    (0 == arp0->ip4_over_ethernet[0].ip4.as_u32 ?
259	     ETHERNET_ARP_ERROR_l3_dst_address_unset : error0);
260
261	  if (ETHERNET_ARP_ERROR_replies_sent == error0)
262	    {
263	      next0 = ARP_INPUT_NEXT_DISABLED;
264	      vnet_feature_arc_start (am->feature_arc_index,
265				      vnet_buffer (p0)->sw_if_index[VLIB_RX],
266				      &next0, p0);
267	    }
268	  else
269	    p0->error = node->errors[error0];
270
271	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
272					   n_left_to_next, pi0, next0);
273	}
274
275      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
276    }
277
278  return frame->n_vectors;
279}
280
281typedef enum arp_disabled_next_t_
282{
283  ARP_DISABLED_NEXT_DROP,
284  ARP_DISABLED_N_NEXT,
285} arp_disabled_next_t;
286
287#define foreach_arp_disabled_error					\
288  _ (DISABLED, "ARP Disabled on this interface")                    \
289
290typedef enum
291{
292#define _(sym,string) ARP_DISABLED_ERROR_##sym,
293  foreach_arp_disabled_error
294#undef _
295    ARP_DISABLED_N_ERROR,
296} arp_disabled_error_t;
297
298static char *arp_disabled_error_strings[] = {
299#define _(sym,string) string,
300  foreach_arp_disabled_error
301#undef _
302};
303
304static uword
305arp_disabled (vlib_main_t * vm,
306	      vlib_node_runtime_t * node, vlib_frame_t * frame)
307{
308  u32 n_left_from, next_index, *from, *to_next, n_left_to_next;
309
310  from = vlib_frame_vector_args (frame);
311  n_left_from = frame->n_vectors;
312  next_index = node->cached_next_index;
313
314  if (node->flags & VLIB_NODE_FLAG_TRACE)
315    vlib_trace_frame_buffers_only (vm, node, from, frame->n_vectors,
316				   /* stride */ 1,
317				   sizeof (ethernet_arp_input_trace_t));
318
319  while (n_left_from > 0)
320    {
321      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
322
323      while (n_left_from > 0 && n_left_to_next > 0)
324	{
325	  arp_disabled_next_t next0 = ARP_DISABLED_NEXT_DROP;
326	  vlib_buffer_t *p0;
327	  u32 pi0, error0;
328
329	  next0 = ARP_DISABLED_NEXT_DROP;
330	  error0 = ARP_DISABLED_ERROR_DISABLED;
331
332	  pi0 = to_next[0] = from[0];
333	  from += 1;
334	  to_next += 1;
335	  n_left_from -= 1;
336	  n_left_to_next -= 1;
337
338	  p0 = vlib_get_buffer (vm, pi0);
339	  p0->error = node->errors[error0];
340
341	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
342					   n_left_to_next, pi0, next0);
343	}
344
345      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
346    }
347
348  return frame->n_vectors;
349}
350
351enum arp_dst_fib_type
352{
353  ARP_DST_FIB_NONE,
354  ARP_DST_FIB_ADJ,
355  ARP_DST_FIB_CONN
356};
357
358/*
359 * we're looking for FIB sources that indicate the destination
360 * is attached. There may be interposed DPO prior to the one
361 * we are looking for
362 */
363static enum arp_dst_fib_type
364arp_dst_fib_check (const fib_node_index_t fei, fib_entry_flag_t * flags)
365{
366  const fib_entry_t *entry = fib_entry_get (fei);
367  const fib_entry_src_t *entry_src;
368  fib_source_t src;
369  /* *INDENT-OFF* */
370  FOR_EACH_SRC_ADDED(entry, entry_src, src,
371  ({
372    *flags = fib_entry_get_flags_for_source (fei, src);
373    if (fib_entry_is_sourced (fei, FIB_SOURCE_ADJ))
374        return ARP_DST_FIB_ADJ;
375      else if (FIB_ENTRY_FLAG_CONNECTED & *flags)
376        return ARP_DST_FIB_CONN;
377  }))
378  /* *INDENT-ON* */
379
380  return ARP_DST_FIB_NONE;
381}
382
383static uword
384arp_reply (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame)
385{
386  vnet_main_t *vnm = vnet_get_main ();
387  u32 n_left_from, next_index, *from, *to_next;
388  u32 n_replies_sent = 0;
389
390  from = vlib_frame_vector_args (frame);
391  n_left_from = frame->n_vectors;
392  next_index = node->cached_next_index;
393
394  if (node->flags & VLIB_NODE_FLAG_TRACE)
395    vlib_trace_frame_buffers_only (vm, node, from, frame->n_vectors,
396				   /* stride */ 1,
397				   sizeof (ethernet_arp_input_trace_t));
398
399  while (n_left_from > 0)
400    {
401      u32 n_left_to_next;
402
403      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
404
405      while (n_left_from > 0 && n_left_to_next > 0)
406	{
407	  vlib_buffer_t *p0;
408	  ethernet_arp_header_t *arp0;
409	  ethernet_header_t *eth_rx;
410	  const ip4_address_t *if_addr0;
411	  u32 pi0, error0, next0, sw_if_index0, conn_sw_if_index0, fib_index0;
412	  u8 dst_is_local0, is_vrrp_reply0;
413	  fib_node_index_t dst_fei, src_fei;
414	  const fib_prefix_t *pfx0;
415	  fib_entry_flag_t src_flags, dst_flags;
416
417	  pi0 = from[0];
418	  to_next[0] = pi0;
419	  from += 1;
420	  to_next += 1;
421	  n_left_from -= 1;
422	  n_left_to_next -= 1;
423
424	  p0 = vlib_get_buffer (vm, pi0);
425	  arp0 = vlib_buffer_get_current (p0);
426	  /* Fill in ethernet header. */
427	  eth_rx = ethernet_buffer_get_header (p0);
428
429	  next0 = ARP_REPLY_NEXT_DROP;
430	  error0 = ETHERNET_ARP_ERROR_replies_sent;
431	  sw_if_index0 = vnet_buffer (p0)->sw_if_index[VLIB_RX];
432
433	  /* Check that IP address is local and matches incoming interface. */
434	  fib_index0 = ip4_fib_table_get_index_for_sw_if_index (sw_if_index0);
435	  if (~0 == fib_index0)
436	    {
437	      error0 = ETHERNET_ARP_ERROR_interface_no_table;
438	      goto drop;
439
440	    }
441
442	  {
443	    /*
444	     * we're looking for FIB entries that indicate the source
445	     * is attached. There may be more specific non-attached
446	     * routes that match the source, but these do not influence
447	     * whether we respond to an ARP request, i.e. they do not
448	     * influence whether we are the correct way for the sender
449	     * to reach us, they only affect how we reach the sender.
450	     */
451	    fib_entry_t *src_fib_entry;
452	    const fib_prefix_t *pfx;
453	    fib_entry_src_t *src;
454	    fib_source_t source;
455	    int attached;
456	    int mask;
457
458	    mask = 32;
459	    attached = 0;
460
461	    do
462	      {
463		src_fei = ip4_fib_table_lookup (ip4_fib_get (fib_index0),
464						&arp0->
465						ip4_over_ethernet[0].ip4,
466						mask);
467		src_fib_entry = fib_entry_get (src_fei);
468
469		/*
470		 * It's possible that the source that provides the
471		 * flags we need, or the flags we must not have,
472		 * is not the best source, so check then all.
473		 */
474                /* *INDENT-OFF* */
475                FOR_EACH_SRC_ADDED(src_fib_entry, src, source,
476                ({
477                  src_flags = fib_entry_get_flags_for_source (src_fei, source);
478
479                  /* Reject requests/replies with our local interface
480                     address. */
481                  if (FIB_ENTRY_FLAG_LOCAL & src_flags)
482                    {
483                      error0 = ETHERNET_ARP_ERROR_l3_src_address_is_local;
484                      /*
485                       * When VPP has an interface whose address is also
486                       * applied to a TAP interface on the host, then VPP's
487                       * TAP interface will be unnumbered  to the 'real'
488                       * interface and do proxy ARP from the host.
489                       * The curious aspect of this setup is that ARP requests
490                       * from the host will come from the VPP's own address.
491                       * So don't drop immediately here, instead go see if this
492                       * is a proxy ARP case.
493                       */
494                      goto next_feature;
495                    }
496                  /* A Source must also be local to subnet of matching
497                   * interface address. */
498                  if ((FIB_ENTRY_FLAG_ATTACHED & src_flags) ||
499                      (FIB_ENTRY_FLAG_CONNECTED & src_flags))
500                    {
501                      attached = 1;
502                      break;
503                    }
504                  /*
505                   * else
506                   *  The packet was sent from an address that is not
507                   *  connected nor attached i.e. it is not from an
508                   *  address that is covered by a link's sub-net,
509                   *  nor is it a already learned host resp.
510                   */
511                }));
512                /* *INDENT-ON* */
513
514		/*
515		 * shorter mask lookup for the next iteration.
516		 */
517		pfx = fib_entry_get_prefix (src_fei);
518		mask = pfx->fp_len - 1;
519
520		/*
521		 * continue until we hit the default route or we find
522		 * the attached we are looking for. The most likely
523		 * outcome is we find the attached with the first source
524		 * on the first lookup.
525		 */
526	      }
527	    while (!attached &&
528		   !fib_entry_is_sourced (src_fei, FIB_SOURCE_DEFAULT_ROUTE));
529
530	    if (!attached)
531	      {
532		/*
533		 * the matching route is a not attached, i.e. it was
534		 * added as a result of routing, rather than interface/ARP
535		 * configuration. If the matching route is not a host route
536		 * (i.e. a /32)
537		 */
538		error0 = ETHERNET_ARP_ERROR_l3_src_address_not_local;
539		goto drop;
540	      }
541	  }
542
543	  dst_fei = ip4_fib_table_lookup (ip4_fib_get (fib_index0),
544					  &arp0->ip4_over_ethernet[1].ip4,
545					  32);
546	  switch (arp_dst_fib_check (dst_fei, &dst_flags))
547	    {
548	    case ARP_DST_FIB_ADJ:
549	      /*
550	       * We matched an adj-fib on ths source subnet (a /32 previously
551	       * added as a result of ARP). If this request is a gratuitous
552	       * ARP, then learn from it.
553	       * The check for matching an adj-fib, is to prevent hosts
554	       * from spamming us with gratuitous ARPS that might otherwise
555	       * blow our ARP cache
556	       */
557	      if (arp0->ip4_over_ethernet[0].ip4.as_u32 ==
558		  arp0->ip4_over_ethernet[1].ip4.as_u32)
559		error0 =
560		  arp_learn (sw_if_index0, &arp0->ip4_over_ethernet[0]);
561	      goto drop;
562	    case ARP_DST_FIB_CONN:
563	      /* destination is connected, continue to process */
564	      break;
565	    case ARP_DST_FIB_NONE:
566	      /* destination is not connected, stop here */
567	      error0 = ETHERNET_ARP_ERROR_l3_dst_address_not_local;
568	      goto next_feature;
569	    }
570
571	  dst_is_local0 = (FIB_ENTRY_FLAG_LOCAL & dst_flags);
572	  pfx0 = fib_entry_get_prefix (dst_fei);
573	  if_addr0 = &pfx0->fp_addr.ip4;
574
575	  is_vrrp_reply0 =
576	    ((arp0->opcode ==
577	      clib_host_to_net_u16 (ETHERNET_ARP_OPCODE_reply))
578	     &&
579	     (!memcmp
580	      (arp0->ip4_over_ethernet[0].mac.bytes, vrrp_prefix,
581	       sizeof (vrrp_prefix))));
582
583	  /* Trash ARP packets whose ARP-level source addresses do not
584	     match their L2-frame-level source addresses, unless it's
585	     a reply from a VRRP virtual router */
586	  if (!ethernet_mac_address_equal
587	      (eth_rx->src_address,
588	       arp0->ip4_over_ethernet[0].mac.bytes) && !is_vrrp_reply0)
589	    {
590	      error0 = ETHERNET_ARP_ERROR_l2_address_mismatch;
591	      goto drop;
592	    }
593
594	  /* Learn or update sender's mapping only for replies to addresses
595	   * that are local to the subnet */
596	  if (arp0->opcode ==
597	      clib_host_to_net_u16 (ETHERNET_ARP_OPCODE_reply))
598	    {
599	      if (dst_is_local0)
600		error0 =
601		  arp_learn (sw_if_index0, &arp0->ip4_over_ethernet[0]);
602	      else
603		/* a reply for a non-local destination could be a GARP.
604		 * GARPs for hosts we know were handled above, so this one
605		 * we drop */
606		error0 = ETHERNET_ARP_ERROR_l3_dst_address_not_local;
607
608	      goto next_feature;
609	    }
610	  else if (arp0->opcode ==
611		   clib_host_to_net_u16 (ETHERNET_ARP_OPCODE_request) &&
612		   (dst_is_local0 == 0))
613	    {
614	      goto next_feature;
615	    }
616
617	  /* Honor unnumbered interface, if any */
618	  conn_sw_if_index0 = fib_entry_get_resolving_interface (dst_fei);
619	  if (sw_if_index0 != conn_sw_if_index0 ||
620	      sw_if_index0 != fib_entry_get_resolving_interface (src_fei))
621	    {
622	      /*
623	       * The interface the ARP is sent to or was received on is not the
624	       * interface on which the covering prefix is configured.
625	       * Maybe this is a case for unnumbered.
626	       */
627	      if (!arp_unnumbered (p0, sw_if_index0, conn_sw_if_index0))
628		{
629		  error0 = ETHERNET_ARP_ERROR_unnumbered_mismatch;
630		  goto drop;
631		}
632	    }
633	  if (arp0->ip4_over_ethernet[0].ip4.as_u32 ==
634	      arp0->ip4_over_ethernet[1].ip4.as_u32)
635	    {
636	      error0 = ETHERNET_ARP_ERROR_gratuitous_arp;
637	      goto drop;
638	    }
639
640	  next0 = arp_mk_reply (vnm, p0, sw_if_index0,
641				if_addr0, arp0, eth_rx);
642
643	  /* We are going to reply to this request, so, in the absence of
644	     errors, learn the sender */
645	  if (!error0)
646	    error0 = arp_learn (sw_if_index0, &arp0->ip4_over_ethernet[1]);
647
648	  n_replies_sent += 1;
649	  goto enqueue;
650
651	next_feature:
652	  vnet_feature_next (&next0, p0);
653	  goto enqueue;
654
655	drop:
656	  p0->error = node->errors[error0];
657
658	enqueue:
659	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
660					   n_left_to_next, pi0, next0);
661	}
662
663      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
664    }
665
666  vlib_error_count (vm, node->node_index,
667		    ETHERNET_ARP_ERROR_replies_sent, n_replies_sent);
668
669  return frame->n_vectors;
670}
671
672
673static char *ethernet_arp_error_strings[] = {
674#define _(sym,string) string,
675  foreach_ethernet_arp_error
676#undef _
677};
678
679/* *INDENT-OFF* */
680
681VLIB_REGISTER_NODE (arp_input_node, static) =
682{
683  .function = arp_input,
684  .name = "arp-input",
685  .vector_size = sizeof (u32),
686  .n_errors = ETHERNET_ARP_N_ERROR,
687  .error_strings = ethernet_arp_error_strings,
688  .n_next_nodes = ARP_INPUT_N_NEXT,
689  .next_nodes = {
690    [ARP_INPUT_NEXT_DROP] = "error-drop",
691    [ARP_INPUT_NEXT_DISABLED] = "arp-disabled",
692  },
693  .format_buffer = format_ethernet_arp_header,
694  .format_trace = format_ethernet_arp_input_trace,
695};
696
697VLIB_REGISTER_NODE (arp_disabled_node, static) =
698{
699  .function = arp_disabled,
700  .name = "arp-disabled",
701  .vector_size = sizeof (u32),
702  .n_errors = ARP_DISABLED_N_ERROR,
703  .error_strings = arp_disabled_error_strings,
704  .n_next_nodes = ARP_DISABLED_N_NEXT,
705  .next_nodes = {
706    [ARP_INPUT_NEXT_DROP] = "error-drop",
707  },
708  .format_buffer = format_ethernet_arp_header,
709  .format_trace = format_ethernet_arp_input_trace,
710};
711
712VLIB_REGISTER_NODE (arp_reply_node, static) =
713{
714  .function = arp_reply,
715  .name = "arp-reply",
716  .vector_size = sizeof (u32),
717  .n_errors = ETHERNET_ARP_N_ERROR,
718  .error_strings = ethernet_arp_error_strings,
719  .n_next_nodes = ARP_REPLY_N_NEXT,
720  .next_nodes = {
721    [ARP_REPLY_NEXT_DROP] = "error-drop",
722    [ARP_REPLY_NEXT_REPLY_TX] = "interface-output",
723  },
724  .format_buffer = format_ethernet_arp_header,
725  .format_trace = format_ethernet_arp_input_trace,
726};
727
728/* Built-in ARP rx feature path definition */
729VNET_FEATURE_ARC_INIT (arp_feat, static) =
730{
731  .arc_name = "arp",
732  .start_nodes = VNET_FEATURES ("arp-input"),
733  .last_in_arc = "error-drop",
734  .arc_index_ptr = &ethernet_arp_main.feature_arc_index,
735};
736
737VNET_FEATURE_INIT (arp_reply_feat_node, static) =
738{
739  .arc_name = "arp",
740  .node_name = "arp-reply",
741  .runs_before = VNET_FEATURES ("arp-disabled"),
742};
743
744VNET_FEATURE_INIT (arp_proxy_feat_node, static) =
745{
746  .arc_name = "arp",
747  .node_name = "arp-proxy",
748  .runs_after = VNET_FEATURES ("arp-reply"),
749  .runs_before = VNET_FEATURES ("arp-disabled"),
750};
751
752VNET_FEATURE_INIT (arp_disabled_feat_node, static) =
753{
754  .arc_name = "arp",
755  .node_name = "arp-disabled",
756  .runs_before = VNET_FEATURES ("error-drop"),
757};
758
759VNET_FEATURE_INIT (arp_drop_feat_node, static) =
760{
761  .arc_name = "arp",
762  .node_name = "error-drop",
763  .runs_before = 0,	/* last feature */
764};
765
766/* *INDENT-ON* */
767
768typedef struct
769{
770  pg_edit_t l2_type, l3_type;
771  pg_edit_t n_l2_address_bytes, n_l3_address_bytes;
772  pg_edit_t opcode;
773  struct
774  {
775    pg_edit_t mac;
776    pg_edit_t ip4;
777  } ip4_over_ethernet[2];
778} pg_ethernet_arp_header_t;
779
780static inline void
781pg_ethernet_arp_header_init (pg_ethernet_arp_header_t * p)
782{
783  /* Initialize fields that are not bit fields in the IP header. */
784#define _(f) pg_edit_init (&p->f, ethernet_arp_header_t, f);
785  _(l2_type);
786  _(l3_type);
787  _(n_l2_address_bytes);
788  _(n_l3_address_bytes);
789  _(opcode);
790  _(ip4_over_ethernet[0].mac);
791  _(ip4_over_ethernet[0].ip4);
792  _(ip4_over_ethernet[1].mac);
793  _(ip4_over_ethernet[1].ip4);
794#undef _
795}
796
797uword
798unformat_pg_arp_header (unformat_input_t * input, va_list * args)
799{
800  pg_stream_t *s = va_arg (*args, pg_stream_t *);
801  pg_ethernet_arp_header_t *p;
802  u32 group_index;
803
804  p = pg_create_edit_group (s, sizeof (p[0]), sizeof (ethernet_arp_header_t),
805			    &group_index);
806  pg_ethernet_arp_header_init (p);
807
808  /* Defaults. */
809  pg_edit_set_fixed (&p->l2_type, ETHERNET_ARP_HARDWARE_TYPE_ethernet);
810  pg_edit_set_fixed (&p->l3_type, ETHERNET_TYPE_IP4);
811  pg_edit_set_fixed (&p->n_l2_address_bytes, 6);
812  pg_edit_set_fixed (&p->n_l3_address_bytes, 4);
813
814  if (!unformat (input, "%U: %U/%U -> %U/%U",
815		 unformat_pg_edit,
816		 unformat_ethernet_arp_opcode_net_byte_order, &p->opcode,
817		 unformat_pg_edit,
818		 unformat_mac_address_t, &p->ip4_over_ethernet[0].mac,
819		 unformat_pg_edit,
820		 unformat_ip4_address, &p->ip4_over_ethernet[0].ip4,
821		 unformat_pg_edit,
822		 unformat_mac_address_t, &p->ip4_over_ethernet[1].mac,
823		 unformat_pg_edit,
824		 unformat_ip4_address, &p->ip4_over_ethernet[1].ip4))
825    {
826      /* Free up any edits we may have added. */
827      pg_free_edit_group (s);
828      return 0;
829    }
830  return 1;
831}
832
833/*
834 * callback when an interface address is added or deleted
835 */
836static void
837arp_enable_disable_interface (ip4_main_t * im,
838			      uword opaque, u32 sw_if_index, u32 is_enable)
839{
840  ethernet_arp_main_t *am = &ethernet_arp_main;
841
842  if (is_enable)
843    arp_enable (am, sw_if_index);
844  else
845    arp_disable (am, sw_if_index);
846}
847
848/*
849 * Remove any arp entries associated with the specified interface
850 */
851static clib_error_t *
852vnet_arp_add_del_sw_interface (vnet_main_t * vnm, u32 sw_if_index, u32 is_add)
853{
854  ethernet_arp_main_t *am = &ethernet_arp_main;
855
856  if (!is_add && sw_if_index != ~0)
857    {
858      arp_disable (am, sw_if_index);
859    }
860  else if (is_add)
861    {
862      vnet_feature_enable_disable ("arp", "arp-disabled",
863				   sw_if_index, 1, NULL, 0);
864    }
865
866  return (NULL);
867}
868
869VNET_SW_INTERFACE_ADD_DEL_FUNCTION (vnet_arp_add_del_sw_interface);
870
871const static ip_neighbor_vft_t arp_vft = {
872  .inv_proxy4_add = arp_proxy_add,
873  .inv_proxy4_del = arp_proxy_del,
874  .inv_proxy4_enable = arp_proxy_disable,
875  .inv_proxy4_disable = arp_proxy_disable,
876};
877
878static clib_error_t *
879ethernet_arp_init (vlib_main_t * vm)
880{
881  ethernet_arp_main_t *am = &ethernet_arp_main;
882  ip4_main_t *im = &ip4_main;
883  pg_node_t *pn;
884
885  ethernet_register_input_type (vm, ETHERNET_TYPE_ARP, arp_input_node.index);
886
887  pn = pg_get_node (arp_input_node.index);
888  pn->unformat_edit = unformat_pg_arp_header;
889
890  am->opcode_by_name = hash_create_string (0, sizeof (uword));
891#define _(o) hash_set_mem (am->opcode_by_name, #o, ETHERNET_ARP_OPCODE_##o);
892  foreach_ethernet_arp_opcode;
893#undef _
894
895  /* don't trace ARP error packets */
896  {
897    vlib_node_runtime_t *rt =
898      vlib_node_get_runtime (vm, arp_input_node.index);
899
900#define _(a,b)                                  \
901    vnet_pcap_drop_trace_filter_add_del         \
902        (rt->errors[ETHERNET_ARP_ERROR_##a],    \
903         1 /* is_add */);
904    foreach_ethernet_arp_error
905#undef _
906  }
907
908  {
909    ip4_enable_disable_interface_callback_t cb = {
910      .function = arp_enable_disable_interface,
911    };
912    vec_add1 (im->enable_disable_interface_callbacks, cb);
913  }
914
915  ip_neighbor_register (IP46_TYPE_IP4, &arp_vft);
916
917  return 0;
918}
919
920/* *INDENT-OFF* */
921VLIB_INIT_FUNCTION (ethernet_arp_init) =
922{
923  .runs_after = VLIB_INITS("ethernet_init",
924                           "ip_neighbor_init"),
925};
926/* *INDENT-ON* */
927
928/*
929 * fd.io coding-style-patch-verification: ON
930 *
931 * Local Variables:
932 * eval: (c-set-style "gnu")
933 * End:
934 */
935