ipsec_tun_in.c revision 02950406
1/*
2 * ipsec_tun_protect_in.c : IPSec interface input node
3 *
4 * Copyright (c) 2015 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/vnet.h>
19#include <vnet/api_errno.h>
20#include <vnet/ip/ip.h>
21
22#include <vnet/ipsec/ipsec.h>
23#include <vnet/ipsec/esp.h>
24#include <vnet/ipsec/ipsec_io.h>
25#include <vnet/ipsec/ipsec_punt.h>
26#include <vnet/ipsec/ipsec_tun.h>
27#include <vnet/ip/ip4_input.h>
28
29/* Statistics (not really errors) */
30#define foreach_ipsec_tun_protect_input_error                     \
31  _(RX, "good packets received")                                  \
32  _(DISABLED, "ipsec packets received on disabled interface")     \
33  _(NO_TUNNEL, "no matching tunnel")                              \
34  _(TUNNEL_MISMATCH, "SPI-tunnel mismatch")                       \
35  _(NAT_KEEPALIVE, "NAT Keepalive")                               \
36  _(TOO_SHORT, "Too Short")                                       \
37  _(SPI_0, "SPI 0")
38
39static char *ipsec_tun_protect_input_error_strings[] = {
40#define _(sym,string) string,
41  foreach_ipsec_tun_protect_input_error
42#undef _
43};
44
45typedef enum
46{
47#define _(sym,str) IPSEC_TUN_PROTECT_INPUT_ERROR_##sym,
48  foreach_ipsec_tun_protect_input_error
49#undef _
50    IPSEC_TUN_PROTECT_INPUT_N_ERROR,
51} ipsec_tun_protect_input_error_t;
52
53typedef enum ipsec_tun_next_t_
54{
55#define _(v, s) IPSEC_TUN_PROTECT_NEXT_##v,
56  foreach_ipsec_input_next
57#undef _
58    IPSEC_TUN_PROTECT_NEXT_DECRYPT,
59  IPSEC_TUN_PROTECT_N_NEXT,
60} ipsec_tun_next_t;
61
62typedef struct
63{
64  union
65  {
66    ipsec4_tunnel_key_t key4;
67    ipsec6_tunnel_key_t key6;
68  };
69  u8 is_ip6;
70  u32 seq;
71  u32 sa_index;
72} ipsec_tun_protect_input_trace_t;
73
74static u8 *
75format_ipsec_tun_protect_input_trace (u8 * s, va_list * args)
76{
77  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
78  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
79  ipsec_tun_protect_input_trace_t *t =
80    va_arg (*args, ipsec_tun_protect_input_trace_t *);
81
82  if (t->is_ip6)
83    s = format (s, "IPSec: %U seq %u sa %d",
84		format_ipsec6_tunnel_key, &t->key6, t->seq, t->sa_index);
85  else
86    s = format (s, "IPSec: %U seq %u sa %d",
87		format_ipsec4_tunnel_key, &t->key4, t->seq, t->sa_index);
88  return s;
89}
90
91always_inline u16
92ipsec_ip4_if_no_tunnel (vlib_node_runtime_t * node,
93			vlib_buffer_t * b,
94			const esp_header_t * esp, const ip4_header_t * ip4)
95{
96  if (PREDICT_FALSE (0 == esp->spi))
97    {
98      b->error = node->errors[IPSEC_TUN_PROTECT_INPUT_ERROR_SPI_0];
99      b->punt_reason = ipsec_punt_reason[(ip4->protocol == IP_PROTOCOL_UDP ?
100					  IPSEC_PUNT_IP4_SPI_UDP_0 :
101					  IPSEC_PUNT_IP4_NO_SUCH_TUNNEL)];
102    }
103  else
104    {
105      b->error = node->errors[IPSEC_TUN_PROTECT_INPUT_ERROR_NO_TUNNEL];
106      b->punt_reason = ipsec_punt_reason[IPSEC_PUNT_IP4_NO_SUCH_TUNNEL];
107    }
108  return IPSEC_INPUT_NEXT_PUNT;
109}
110
111always_inline u16
112ipsec_ip6_if_no_tunnel (vlib_node_runtime_t * node,
113			vlib_buffer_t * b, const esp_header_t * esp)
114{
115  b->error = node->errors[IPSEC_TUN_PROTECT_INPUT_ERROR_NO_TUNNEL];
116  b->punt_reason = ipsec_punt_reason[IPSEC_PUNT_IP6_NO_SUCH_TUNNEL];
117
118  return (IPSEC_INPUT_NEXT_PUNT);
119}
120
121always_inline uword
122ipsec_tun_protect_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
123				vlib_frame_t * from_frame, int is_ip6)
124{
125  ipsec_main_t *im = &ipsec_main;
126  vnet_main_t *vnm = im->vnet_main;
127  vnet_interface_main_t *vim = &vnm->interface_main;
128
129  int is_trace = node->flags & VLIB_NODE_FLAG_TRACE;
130  u32 thread_index = vm->thread_index;
131
132  u32 n_left_from, *from;
133  u16 nexts[VLIB_FRAME_SIZE], *next;
134  vlib_buffer_t *bufs[VLIB_FRAME_SIZE], **b;
135
136  from = vlib_frame_vector_args (from_frame);
137  n_left_from = from_frame->n_vectors;
138
139  vlib_get_buffers (vm, from, bufs, n_left_from);
140  b = bufs;
141  next = nexts;
142
143  clib_memset_u16 (nexts, im->esp4_decrypt_next_index, n_left_from);
144
145  u64 n_bytes = 0, n_packets = 0;
146  u32 n_disabled = 0, n_no_tunnel = 0;
147
148  u32 last_sw_if_index = ~0;
149  ipsec_tun_lkup_result_t last_result = {
150    .tun_index = ~0
151  };
152  ipsec4_tunnel_key_t last_key4;
153  ipsec6_tunnel_key_t last_key6;
154
155  vlib_combined_counter_main_t *rx_counter;
156  vlib_combined_counter_main_t *drop_counter;
157  ipsec_tun_protect_t *itp0;
158
159  if (is_ip6)
160    clib_memset (&last_key6, 0xff, sizeof (last_key6));
161  else
162    last_key4.as_u64 = ~0;
163
164  rx_counter = vim->combined_sw_if_counters + VNET_INTERFACE_COUNTER_RX;
165  drop_counter = vim->combined_sw_if_counters + VNET_INTERFACE_COUNTER_DROP;
166
167  while (n_left_from > 0)
168    {
169      u32 sw_if_index0, len0, hdr_sz0;
170      ipsec_tun_lkup_result_t itr0;
171      ipsec4_tunnel_key_t key40;
172      ipsec6_tunnel_key_t key60;
173      ip4_header_t *ip40;
174      ip6_header_t *ip60;
175      esp_header_t *esp0;
176      u16 buf_rewind0;
177
178      ip40 =
179	(ip4_header_t *) (b[0]->data + vnet_buffer (b[0])->l3_hdr_offset);
180
181      if (is_ip6)
182	{
183	  ip60 = (ip6_header_t *) ip40;
184	  esp0 = (esp_header_t *) (ip60 + 1);
185	  hdr_sz0 = sizeof (ip6_header_t);
186	}
187      else
188	{
189	  /* NAT UDP port 4500 case, don't advance any more */
190	  if (ip40->protocol == IP_PROTOCOL_UDP)
191	    {
192	      esp0 =
193		(esp_header_t *) ((u8 *) ip40 + ip4_header_bytes (ip40) +
194				  sizeof (udp_header_t));
195	      hdr_sz0 = 0;
196	      buf_rewind0 = ip4_header_bytes (ip40) + sizeof (udp_header_t);
197	    }
198	  else
199	    {
200	      esp0 = (esp_header_t *) ((u8 *) ip40 + ip4_header_bytes (ip40));
201	      buf_rewind0 = hdr_sz0 = ip4_header_bytes (ip40);
202	    }
203	}
204
205      /* stats for the tunnel include all the data after the IP header
206         just like a norml IP-IP tunnel */
207      vlib_buffer_advance (b[0], hdr_sz0);
208      len0 = vlib_buffer_length_in_chain (vm, b[0]);
209
210      if (len0 < sizeof (esp_header_t))
211	{
212	  if (esp0->spi_bytes[0] == 0xff)
213	    b[0]->error =
214	      node->errors[IPSEC_TUN_PROTECT_INPUT_ERROR_NAT_KEEPALIVE];
215	  else
216	    b[0]->error =
217	      node->errors[IPSEC_TUN_PROTECT_INPUT_ERROR_TOO_SHORT];
218
219	  next[0] = IPSEC_INPUT_NEXT_DROP;
220	  goto trace00;
221	}
222
223      if (is_ip6)
224	{
225	  key60.remote_ip = ip60->src_address;
226	  key60.spi = esp0->spi;
227
228	  if (memcmp (&key60, &last_key6, sizeof (last_key6)) == 0)
229	    {
230	      itr0 = last_result;
231	    }
232	  else
233	    {
234	      uword *p = hash_get_mem (im->tun6_protect_by_key, &key60);
235	      if (p)
236		{
237		  itr0.as_u64 = p[0];
238		  last_result = itr0;
239		  clib_memcpy_fast (&last_key6, &key60, sizeof (key60));
240		}
241	      else
242		{
243		  next[0] = ipsec_ip6_if_no_tunnel (node, b[0], esp0);
244		  n_no_tunnel++;
245		  goto trace00;
246		}
247	    }
248	}
249      else
250	{
251	  key40.remote_ip = ip40->src_address;
252	  key40.spi = esp0->spi;
253
254	  if (key40.as_u64 == last_key4.as_u64)
255	    {
256	      itr0 = last_result;
257	    }
258	  else
259	    {
260	      uword *p = hash_get (im->tun4_protect_by_key, key40.as_u64);
261	      if (p)
262		{
263		  itr0.as_u64 = p[0];
264		  last_result = itr0;
265		  last_key4.as_u64 = key40.as_u64;
266		}
267	      else
268		{
269		  next[0] = ipsec_ip4_if_no_tunnel (node, b[0], esp0, ip40);
270		  vlib_buffer_advance (b[0], -buf_rewind0);
271		  n_no_tunnel++;
272		  goto trace00;
273		}
274	    }
275	}
276
277      itp0 = pool_elt_at_index (ipsec_protect_pool, itr0.tun_index);
278      vnet_buffer (b[0])->ipsec.sad_index = itr0.sa_index;
279      vnet_buffer (b[0])->ipsec.protect_index = itr0.tun_index;
280
281      sw_if_index0 = itp0->itp_sw_if_index;
282      vnet_buffer (b[0])->sw_if_index[VLIB_RX] = sw_if_index0;
283
284      if (PREDICT_FALSE (!vnet_sw_interface_is_admin_up (vnm, sw_if_index0)))
285	{
286	  vlib_increment_combined_counter
287	    (drop_counter, thread_index, sw_if_index0, 1, len0);
288	  n_disabled++;
289	  b[0]->error = node->errors[IPSEC_TUN_PROTECT_INPUT_ERROR_DISABLED];
290	  next[0] = IPSEC_INPUT_NEXT_DROP;
291	  goto trace00;
292	}
293      else
294	{
295	  if (PREDICT_TRUE (sw_if_index0 == last_sw_if_index))
296	    {
297	      n_packets++;
298	      n_bytes += len0;
299	    }
300	  else
301	    {
302	      if (n_packets && !(itp0->itp_flags & IPSEC_PROTECT_ENCAPED))
303		{
304		  vlib_increment_combined_counter
305		    (rx_counter, thread_index, last_sw_if_index,
306		     n_packets, n_bytes);
307		}
308
309	      last_sw_if_index = sw_if_index0;
310	      n_packets = 1;
311	      n_bytes = len0;
312	    }
313
314	  next[0] = IPSEC_TUN_PROTECT_NEXT_DECRYPT;
315	}
316    trace00:
317      if (PREDICT_FALSE (is_trace))
318	{
319	  if (b[0]->flags & VLIB_BUFFER_IS_TRACED)
320	    {
321	      ipsec_tun_protect_input_trace_t *tr =
322		vlib_add_trace (vm, node, b[0], sizeof (*tr));
323	      if (is_ip6)
324		clib_memcpy (&tr->key6, &key60, sizeof (tr->key6));
325	      else
326		clib_memcpy (&tr->key4, &key40, sizeof (tr->key4));
327	      tr->is_ip6 = is_ip6;
328	      tr->seq = (len0 >= sizeof (*esp0) ?
329			 clib_host_to_net_u32 (esp0->seq) : ~0);
330	      tr->sa_index = vnet_buffer (b[0])->ipsec.sad_index;
331	    }
332	}
333
334      /* next */
335      b += 1;
336      next += 1;
337      n_left_from -= 1;
338    }
339
340  if (n_packets && !(itp0->itp_flags & IPSEC_PROTECT_ENCAPED))
341    {
342      vlib_increment_combined_counter (rx_counter,
343				       thread_index,
344				       last_sw_if_index, n_packets, n_bytes);
345    }
346
347  vlib_node_increment_counter (vm, node->node_index,
348			       IPSEC_TUN_PROTECT_INPUT_ERROR_RX,
349			       from_frame->n_vectors - (n_disabled +
350							n_no_tunnel));
351
352  vlib_buffer_enqueue_to_next (vm, node, from, nexts, from_frame->n_vectors);
353
354  return from_frame->n_vectors;
355}
356
357VLIB_NODE_FN (ipsec4_tun_input_node) (vlib_main_t * vm,
358				      vlib_node_runtime_t * node,
359				      vlib_frame_t * from_frame)
360{
361  return ipsec_tun_protect_input_inline (vm, node, from_frame,
362					 0 /* is_ip6 */ );
363}
364
365/* *INDENT-OFF* */
366VLIB_REGISTER_NODE (ipsec4_tun_input_node) = {
367  .name = "ipsec4-tun-input",
368  .vector_size = sizeof (u32),
369  .format_trace = format_ipsec_tun_protect_input_trace,
370  .type = VLIB_NODE_TYPE_INTERNAL,
371  .n_errors = ARRAY_LEN(ipsec_tun_protect_input_error_strings),
372  .error_strings = ipsec_tun_protect_input_error_strings,
373  .n_next_nodes = IPSEC_TUN_PROTECT_N_NEXT,
374  .next_nodes = {
375    [IPSEC_TUN_PROTECT_NEXT_DROP] = "ip4-drop",
376    [IPSEC_TUN_PROTECT_NEXT_PUNT] = "punt-dispatch",
377    [IPSEC_TUN_PROTECT_NEXT_DECRYPT] = "esp4-decrypt-tun",
378  }
379};
380/* *INDENT-ON* */
381
382VLIB_NODE_FN (ipsec6_tun_input_node) (vlib_main_t * vm,
383				      vlib_node_runtime_t * node,
384				      vlib_frame_t * from_frame)
385{
386  return ipsec_tun_protect_input_inline (vm, node, from_frame,
387					 1 /* is_ip6 */ );
388}
389
390/* *INDENT-OFF* */
391VLIB_REGISTER_NODE (ipsec6_tun_input_node) = {
392  .name = "ipsec6-tun-input",
393  .vector_size = sizeof (u32),
394  .format_trace = format_ipsec_tun_protect_input_trace,
395  .type = VLIB_NODE_TYPE_INTERNAL,
396  .n_errors = ARRAY_LEN(ipsec_tun_protect_input_error_strings),
397  .error_strings = ipsec_tun_protect_input_error_strings,
398  .n_next_nodes = IPSEC_TUN_PROTECT_N_NEXT,
399  .next_nodes = {
400    [IPSEC_TUN_PROTECT_NEXT_DROP] = "ip6-drop",
401    [IPSEC_TUN_PROTECT_NEXT_PUNT] = "punt-dispatch",
402    [IPSEC_TUN_PROTECT_NEXT_DECRYPT] = "esp6-decrypt-tun",
403  }
404};
405/* *INDENT-ON* */
406
407/*
408 * fd.io coding-style-patch-verification: ON
409 *
410 * Local Variables:
411 * eval: (c-set-style "gnu")
412 * End:
413 */
414