summaryrefslogtreecommitdiffstats
path: root/src/plugins/dpdk/buffer.c
blob: 8b4b4a926b3b98b613a5c95bde1988502a7de097 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
/*
 * Copyright (c) 2017-2019 Cisco and/or its affiliates.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include <unistd.h>
#include <errno.h>

#include <rte_config.h>
#include <rte_mbuf.h>
#include <rte_ethdev.h>
#include <rte_cryptodev.h>
#include <rte_vfio.h>
#include <rte_version.h>

#include <vlib/vlib.h>
#include <dpdk/buffer.h>

STATIC_ASSERT (VLIB_BUFFER_PRE_DATA_SIZE == RTE_PKTMBUF_HEADROOM,
	       "VLIB_BUFFER_PRE_DATA_SIZE must be equal to RTE_PKTMBUF_HEADROOM");

extern struct rte_mbuf *dpdk_mbuf_template_by_pool_index;
#ifndef CLIB_MARCH_VARIANT
struct rte_mempool **dpdk_mempool_by_buffer_pool_index = 0;
struct rte_mempool **dpdk_no_cache_mempool_by_buffer_pool_index = 0;
struct rte_mbuf *dpdk_mbuf_template_by_pool_index = 0;

clib_error_t *
dpdk_buffer_pool_init (vlib_main_t * vm, vlib_buffer_pool_t * bp)
{
  uword buffer_mem_start = vm->buffer_main->buffer_mem_start;
  struct rte_mempool *mp, *nmp;
  struct rte_pktmbuf_pool_private priv;
  enum rte_iova_mode iova_mode;
  u32 i;
  u8 *name = 0;

  u32 elt_size =
    sizeof (struct rte_mbuf) + sizeof (vlib_buffer_t) + bp->data_size;

  /* create empty mempools */
  vec_validate_aligned (dpdk_mempool_by_buffer_pool_index, bp->index,
			CLIB_CACHE_LINE_BYTES);
  vec_validate_aligned (dpdk_no_cache_mempool_by_buffer_pool_index, bp->index,
			CLIB_CACHE_LINE_BYTES);

  /* normal mempool */
  name = format (name, "vpp pool %u%c", bp->index, 0);
  mp = rte_mempool_create_empty ((char *) name, bp->n_buffers,
				 elt_size, 512, sizeof (priv),
				 bp->numa_node, 0);
  if (!mp)
    {
      vec_free (name);
      return clib_error_return (0,
				"failed to create normal mempool for numa node %u",
				bp->index);
    }
  vec_reset_length (name);

  /* non-cached mempool */
  name = format (name, "vpp pool %u (no cache)%c", bp->index, 0);
  nmp = rte_mempool_create_empty ((char *) name, bp->n_buffers,
				  elt_size, 0, sizeof (priv),
				  bp->numa_node, 0);
  if (!nmp)
    {
      rte_mempool_free (mp);
      vec_free (name);
      return clib_error_return (0,
				"failed to create non-cache mempool for numa nude %u",
				bp->index);
    }
  vec_free (name);

  dpdk_mempool_by_buffer_pool_index[bp->index] = mp;
  dpdk_no_cache_mempool_by_buffer_pool_index[bp->index] = nmp;

  mp->pool_id = nmp->pool_id = bp->index;

  rte_mempool_set_ops_byname (mp, "vpp", NULL);
  rte_mempool_set_ops_byname (nmp, "vpp-no-cache", NULL);

  /* Call the mempool priv initializer */
  memset (&priv, 0, sizeof (priv));
  priv.mbuf_data_room_size = VLIB_BUFFER_PRE_DATA_SIZE +
    vlib_buffer_get_default_data_size (vm);
  priv.mbuf_priv_size = VLIB_BUFFER_HDR_SIZE;
  rte_pktmbuf_pool_init (mp, &priv);
  rte_pktmbuf_pool_init (nmp, &priv);

  iova_mode = rte_eal_iova_mode ();

  /* populate mempool object buffer header */
  for (i = 0; i < bp->n_buffers; i++)
    {
      struct rte_mempool_objhdr *hdr;
      vlib_buffer_t *b = vlib_get_buffer (vm, bp->buffers[i]);
      struct rte_mbuf *mb = rte_mbuf_from_vlib_buffer (b);
      hdr = (struct rte_mempool_objhdr *) RTE_PTR_SUB (mb, sizeof (*hdr));
      hdr->mp = mp;
      hdr->iova = (iova_mode == RTE_IOVA_VA) ?
	pointer_to_uword (mb) : vlib_physmem_get_pa (vm, mb);
      STAILQ_INSERT_TAIL (&mp->elt_list, hdr, next);
      STAILQ_INSERT_TAIL (&nmp->elt_list, hdr, next);
      mp->populated_size++;
      nmp->populated_size++;
    }
#if RTE_VERSION >= RTE_VERSION_NUM(22, 3, 0, 0)
  mp->flags &= ~RTE_MEMPOOL_F_NON_IO;
#endif

  /* call the object initializers */
  rte_mempool_obj_iter (mp, rte_pktmbuf_init, 0);

  /* create mbuf header tempate from the first buffer in the pool */
  vec_validate_aligned (dpdk_mbuf_template_by_pool_index, bp->index,
			CLIB_CACHE_LINE_BYTES);
  clib_memcpy (vec_elt_at_index (dpdk_mbuf_template_by_pool_index, bp->index),
	       rte_mbuf_from_vlib_buffer (vlib_buffer_ptr_from_index
					  (buffer_mem_start, *bp->buffers,
					   0)), sizeof (struct rte_mbuf));

  for (i = 0; i < bp->n_buffers; i++)
    {
      vlib_buffer_t *b;
      b = vlib_buffer_ptr_from_index (buffer_mem_start, bp->buffers[i], 0);
      vlib_buffer_copy_template (b, &bp->buffer_template);
    }

  /* map DMA pages if at least one physical device exists */
  if (rte_eth_dev_count_avail () || rte_cryptodev_count ())
    {
      uword i;
      size_t page_sz;
      vlib_physmem_map_t *pm;
      int do_vfio_map = 1;

      pm = vlib_physmem_get_map (vm, bp->physmem_map_index);
      page_sz = 1ULL << pm->log2_page_size;

      for (i = 0; i < pm->n_pages; i++)
	{
	  char *va = ((char *) pm->base) + i * page_sz;
	  uword pa = (iova_mode == RTE_IOVA_VA) ?
	    pointer_to_uword (va) : pm->page_table[i];

	  if (do_vfio_map &&
#if RTE_VERSION < RTE_VERSION_NUM(19, 11, 0, 0)
	      rte_vfio_dma_map (pointer_to_uword (va), pa, page_sz))
#else
	      rte_vfio_container_dma_map (RTE_VFIO_DEFAULT_CONTAINER_FD,
					  pointer_to_uword (va), pa, page_sz))
#endif
	    do_vfio_map = 0;

	  struct rte_mempool_memhdr *memhdr;
	  memhdr = clib_mem_alloc (sizeof (*memhdr));
	  memhdr->mp = mp;
	  memhdr->addr = va;
	  memhdr->iova = pa;
	  memhdr->len = page_sz;
	  memhdr->free_cb = 0;
	  memhdr->opaque = 0;

	  STAILQ_INSERT_TAIL (&mp->mem_list, memhdr, next);
	  mp->nb_mem_chunks++;
	}
    }

  return 0;
}

static int
dpdk_ops_vpp_alloc (struct rte_mempool *mp)
{
  clib_warning ("");
  return 0;
}

static void
dpdk_ops_vpp_free (struct rte_mempool *mp)
{
  clib_warning ("");
}

#endif

static_always_inline void
dpdk_ops_vpp_enqueue_one (vlib_buffer_t * bt, void *obj)
{
  /* Only non-replicated packets (b->ref_count == 1) expected */

  struct rte_mbuf *mb = obj;
  vlib_buffer_t *b = vlib_buffer_from_rte_mbuf (mb);
  ASSERT (b->ref_count == 1);
  ASSERT (b->buffer_pool_index == bt->buffer_pool_index);
  vlib_buffer_copy_template (b, bt);
}

int
CLIB_MULTIARCH_FN (dpdk_ops_vpp_enqueue) (struct rte_mempool * mp,
					  void *const *obj_table, unsigned n)
{
  const int batch_size = 32;
  vlib_main_t *vm = vlib_get_main ();
  vlib_buffer_t bt;
  u8 buffer_pool_index = mp->pool_id;
  vlib_buffer_pool_t *bp = vlib_get_buffer_pool (vm, buffer_pool_index);
  u32 bufs[batch_size];
  u32 n_left = n;
  void *const *obj = obj_table;

  vlib_buffer_copy_template (&bt, &bp->buffer_template);

  while (n_left >= 4)
    {
      dpdk_ops_vpp_enqueue_one (&bt, obj[0]);
      dpdk_ops_vpp_enqueue_one (&bt, obj[1]);
      dpdk_ops_vpp_enqueue_one (&bt, obj[2]);
      dpdk_ops_vpp_enqueue_one (&bt, obj[3]);
      obj += 4;
      n_left -= 4;
    }

  while (n_left)
    {
      dpdk_ops_vpp_enqueue_one (&bt, obj[0]);
      obj += 1;
      n_left -= 1;
    }

  while (n >= batch_size)
    {
      vlib_get_buffer_indices_with_offset (vm, (void **) obj_table, bufs,
					   batch_size,
					   sizeof (struct rte_mbuf));
      vlib_buffer_pool_put (vm, buffer_pool_index, bufs, batch_size);
      n -= batch_size;
      obj_table += batch_size;
    }

  if (n)
    {
      vlib_get_buffer_indices_with_offset (vm, (void **) obj_table, bufs,
					   n, sizeof (struct rte_mbuf));
      vlib_buffer_pool_put (vm, buffer_pool_index, bufs, n);
    }

  return 0;
}

CLIB_MARCH_FN_REGISTRATION (dpdk_ops_vpp_enqueue);

static_always_inline void
dpdk_ops_vpp_enqueue_no_cache_one (vlib_main_t * vm, struct rte_mempool *old,
				   struct rte_mempool *new, void *obj,
				   vlib_buffer_t * bt)
{
  struct rte_mbuf *mb = obj;
  vlib_buffer_t *b = vlib_buffer_from_rte_mbuf (mb);

  if (clib_atomic_sub_fetch (&b->ref_count, 1) == 0)
    {
      u32 bi = vlib_get_buffer_index (vm, b);
      vlib_buffer_copy_template (b, bt);
      vlib_buffer_pool_put (vm, bt->buffer_pool_index, &bi, 1);
      return;
    }
}

int
CLIB_MULTIARCH_FN (dpdk_ops_vpp_enqueue_no_cache) (struct rte_mempool * cmp,
						   void *const *obj_table,
						   unsigned n)
{
  vlib_main_t *vm = vlib_get_main ();
  vlib_buffer_t bt;
  struct rte_mempool *mp;
  mp = dpdk_mempool_by_buffer_pool_index[cmp->pool_id];
  u8 buffer_pool_index = cmp->pool_id;
  vlib_buffer_pool_t *bp = vlib_get_buffer_pool (vm, buffer_pool_index);
  vlib_buffer_copy_template (&bt, &bp->buffer_template);

  while (n >= 4)
    {
      dpdk_ops_vpp_enqueue_no_cache_one (vm, cmp, mp, obj_table[0], &bt);
      dpdk_ops_vpp_enqueue_no_cache_one (vm, cmp, mp, obj_table[1], &bt);
      dpdk_ops_vpp_enqueue_no_cache_one (vm, cmp, mp, obj_table[2], &bt);
      dpdk_ops_vpp_enqueue_no_cache_one (vm, cmp, mp, obj_table[3], &bt);
      obj_table += 4;
      n -= 4;
    }

  while (n)
    {
      dpdk_ops_vpp_enqueue_no_cache_one (vm, cmp, mp, obj_table[0], &bt);
      obj_table += 1;
      n -= 1;
    }

  return 0;
}

CLIB_MARCH_FN_REGISTRATION (dpdk_ops_vpp_enqueue_no_cache);

static_always_inline void
dpdk_mbuf_init_from_template (struct rte_mbuf **mba, struct rte_mbuf *mt,
			      int count)
{
  /* Assumptions about rte_mbuf layout */
  STATIC_ASSERT_OFFSET_OF (struct rte_mbuf, buf_addr, 0);
  STATIC_ASSERT_OFFSET_OF (struct rte_mbuf, buf_iova, 8);
  STATIC_ASSERT_SIZEOF_ELT (struct rte_mbuf, buf_iova, 8);
  STATIC_ASSERT_SIZEOF_ELT (struct rte_mbuf, buf_iova, 8);
  STATIC_ASSERT_SIZEOF (struct rte_mbuf, 128);

  while (count--)
    {
      struct rte_mbuf *mb = mba[0];
      int i;
      /* bytes 0 .. 15 hold buf_addr and buf_iova which we need to preserve */
      /* copy bytes 16 .. 31 */
      *((u8x16 *) mb + 1) = *((u8x16 *) mt + 1);

      /* copy bytes 32 .. 127 */
#ifdef CLIB_HAVE_VEC256
      for (i = 1; i < 4; i++)
	*((u8x32 *) mb + i) = *((u8x32 *) mt + i);
#else
      for (i = 2; i < 8; i++)
	*((u8x16 *) mb + i) = *((u8x16 *) mt + i);
#endif
      mba++;
    }
}

int
CLIB_MULTIARCH_FN (dpdk_ops_vpp_dequeue) (struct rte_mempool * mp,
					  void **obj_table, unsigned n)
{
  const int batch_size = 32;
  vlib_main_t *vm = vlib_get_main ();
  u32 bufs[batch_size], total = 0, n_alloc = 0;
  u8 buffer_pool_index = mp->pool_id;
  void **obj = obj_table;
  struct rte_mbuf t = dpdk_mbuf_template_by_pool_index[buffer_pool_index];

  while (n >= batch_size)
    {
      n_alloc = vlib_buffer_alloc_from_pool (vm, bufs, batch_size,
					     buffer_pool_index);
      if (n_alloc != batch_size)
	goto alloc_fail;

      vlib_get_buffers_with_offset (vm, bufs, obj, batch_size,
				    -(i32) sizeof (struct rte_mbuf));
      dpdk_mbuf_init_from_template ((struct rte_mbuf **) obj, &t, batch_size);
      total += batch_size;
      obj += batch_size;
      n -= batch_size;
    }

  if (n)
    {
      n_alloc = vlib_buffer_alloc_from_pool (vm, bufs, n, buffer_pool_index);

      if (n_alloc != n)
	goto alloc_fail;

      vlib_get_buffers_with_offset (vm, bufs, obj, n,
				    -(i32) sizeof (struct rte_mbuf));
      dpdk_mbuf_init_from_template ((struct rte_mbuf **) obj, &t, n);
    }

  return 0;

alloc_fail:
  /* dpdk doesn't support partial alloc, so we need to return what we
     already got */
  if (n_alloc)
    vlib_buffer_pool_put (vm, buffer_pool_index, bufs, n_alloc);
  obj = obj_table;
  while (total)
    {
      vlib_get_buffer_indices_with_offset (vm, obj, bufs, batch_size,
					   sizeof (struct rte_mbuf));
      vlib_buffer_pool_put (vm, buffer_pool_index, bufs, batch_size);

      obj += batch_size;
      total -= batch_size;
    }
  return -ENOENT;
}

CLIB_MARCH_FN_REGISTRATION (dpdk_ops_vpp_dequeue);

#ifndef CLIB_MARCH_VARIANT

static int
dpdk_ops_vpp_dequeue_no_cache (struct rte_mempool *mp, void **obj_table,
			       unsigned n)
{
  clib_error ("bug");
  return 0;
}

static unsigned
dpdk_ops_vpp_get_count (const struct rte_mempool *mp)
{
  vlib_main_t *vm = vlib_get_main ();
  if (mp)
    {
      vlib_buffer_pool_t *pool = vlib_get_buffer_pool (vm, mp->pool_id);
      if (pool)
	{
	  return pool->n_avail;
	}
    }
  return 0;
}

static unsigned
dpdk_ops_vpp_get_count_no_cache (const struct rte_mempool *mp)
{
  struct rte_mempool *cmp;
  cmp = dpdk_no_cache_mempool_by_buffer_pool_index[mp->pool_id];
  return dpdk_ops_vpp_get_count (cmp);
}

clib_error_t *
dpdk_buffer_pools_create (vlib_main_t * vm)
{
  clib_error_t *err;
  vlib_buffer_pool_t *bp;

  struct rte_mempool_ops ops = { };

  strncpy (ops.name, "vpp", 4);
  ops.alloc = dpdk_ops_vpp_alloc;
  ops.free = dpdk_ops_vpp_free;
  ops.get_count = dpdk_ops_vpp_get_count;
  ops.enqueue = CLIB_MARCH_FN_POINTER (dpdk_ops_vpp_enqueue);
  ops.dequeue = CLIB_MARCH_FN_POINTER (dpdk_ops_vpp_dequeue);
  rte_mempool_register_ops (&ops);

  strncpy (ops.name, "vpp-no-cache", 13);
  ops.get_count = dpdk_ops_vpp_get_count_no_cache;
  ops.enqueue = CLIB_MARCH_FN_POINTER (dpdk_ops_vpp_enqueue_no_cache);
  ops.dequeue = dpdk_ops_vpp_dequeue_no_cache;
  rte_mempool_register_ops (&ops);

  /* *INDENT-OFF* */
  vec_foreach (bp, vm->buffer_main->buffer_pools)
    if (bp->start && (err = dpdk_buffer_pool_init (vm, bp)))
      return err;
  /* *INDENT-ON* */
  return 0;
}

VLIB_BUFFER_SET_EXT_HDR_SIZE (sizeof (struct rte_mempool_objhdr) +
			      sizeof (struct rte_mbuf));

#endif

/** @endcond */
/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */
lass="p">{ u32 flen; u32 hw_if_index = ~0; u8 *new_name; vnet_hw_interface_t *hi; vnet_main_t *vnm = lgm->vnet_main; /* create hw lisp_gpeX iface if needed, otherwise reuse existing */ flen = vec_len (lgm->free_tunnel_hw_if_indices); if (flen > 0) { hw_if_index = lgm->free_tunnel_hw_if_indices[flen - 1]; vec_dec_len (lgm->free_tunnel_hw_if_indices, 1); hi = vnet_get_hw_interface (vnm, hw_if_index); /* rename interface */ new_name = format (0, "%U", dev_class->format_device_name, vni); vec_add1 (new_name, 0); vnet_rename_interface (vnm, hw_if_index, (char *) new_name); vec_free (new_name); /* clear old stats of freed interface before reuse */ vnet_interface_main_t *im = &vnm->interface_main; vnet_interface_counter_lock (im); vlib_zero_combined_counter (&im->combined_sw_if_counters [VNET_INTERFACE_COUNTER_TX], hi->sw_if_index); vlib_zero_combined_counter (&im->combined_sw_if_counters [VNET_INTERFACE_COUNTER_RX], hi->sw_if_index); vlib_zero_simple_counter (&im->sw_if_counters [VNET_INTERFACE_COUNTER_DROP], hi->sw_if_index); vnet_interface_counter_unlock (im); } else { hw_if_index = vnet_register_interface (vnm, dev_class->index, vni, lisp_gpe_hw_class.index, 0); hi = vnet_get_hw_interface (vnm, hw_if_index); } hash_set (tuns->hw_if_index_by_dp_table, dp_table, hw_if_index); /* set tunnel termination: post decap, packets are tagged as having been * originated by lisp-gpe interface */ hash_set (tuns->sw_if_index_by_vni, vni, hi->sw_if_index); hash_set (tuns->vni_by_sw_if_index, hi->sw_if_index, vni); return hi; } static void lisp_gpe_remove_iface (lisp_gpe_main_t * lgm, u32 hi_index, u32 dp_table, tunnel_lookup_t * tuns) { vnet_main_t *vnm = lgm->vnet_main; vnet_hw_interface_t *hi; uword *vnip; hi = vnet_get_hw_interface (vnm, hi_index); /* disable interface */ vnet_sw_interface_set_flags (vnm, hi->sw_if_index, 0 /* down */ ); vnet_hw_interface_set_flags (vnm, hi->hw_if_index, 0 /* down */ ); hash_unset (tuns->hw_if_index_by_dp_table, dp_table); vec_add1 (lgm->free_tunnel_hw_if_indices, hi->hw_if_index); /* clean tunnel termination and vni to sw_if_index binding */ vnip = hash_get (tuns->vni_by_sw_if_index, hi->sw_if_index); if (0 == vnip) { clib_warning ("No vni associated to interface %d", hi->sw_if_index); return; } hash_unset (tuns->sw_if_index_by_vni, vnip[0]); hash_unset (tuns->vni_by_sw_if_index, hi->sw_if_index); } static void lisp_gpe_iface_set_table (u32 sw_if_index, u32 table_id) { fib_node_index_t fib_index; fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP4, table_id, FIB_SOURCE_LISP); ip4_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; ip4_sw_interface_enable_disable (sw_if_index, 1); fib_index = fib_table_find_or_create_and_lock (FIB_PROTOCOL_IP6, table_id, FIB_SOURCE_LISP); ip6_main.fib_index_by_sw_if_index[sw_if_index] = fib_index; ip6_sw_interface_enable_disable (sw_if_index, 1); } static void lisp_gpe_tenant_del_default_routes (u32 table_id) { fib_protocol_t proto; FOR_EACH_FIB_IP_PROTOCOL (proto) { fib_prefix_t prefix = { .fp_proto = proto, }; u32 fib_index; fib_index = fib_table_find (prefix.fp_proto, table_id); fib_table_entry_special_remove (fib_index, &prefix, FIB_SOURCE_LISP); fib_table_unlock (fib_index, prefix.fp_proto, FIB_SOURCE_LISP); } } static void lisp_gpe_tenant_add_default_routes (u32 table_id) { fib_protocol_t proto; FOR_EACH_FIB_IP_PROTOCOL (proto) { fib_prefix_t prefix = { .fp_proto = proto, }; u32 fib_index; /* * Add a deafult route that results in a control plane punt DPO */ fib_index = fib_table_find_or_create_and_lock (prefix.fp_proto, table_id, FIB_SOURCE_LISP); fib_table_entry_special_dpo_add (fib_index, &prefix, FIB_SOURCE_LISP, FIB_ENTRY_FLAG_EXCLUSIVE, lisp_cp_dpo_get (fib_proto_to_dpo (proto))); } } /** * @brief Add/del LISP-GPE L3 interface. * * Creates LISP-GPE interface, sets ingress arcs from lisp_gpeX_lookup, * installs default routes that attract all traffic with no more specific * routes to lgpe-ipx-lookup, set egress arcs to ipx-lookup, sets * the interface in the right vrf and enables it. * * @param[in] lgm Reference to @ref lisp_gpe_main_t. * @param[in] a Parameters to create interface. * * @return number of vectors in frame. */ u32 lisp_gpe_add_l3_iface (lisp_gpe_main_t * lgm, u32 vni, u32 table_id, u8 with_default_routes) { vnet_main_t *vnm = lgm->vnet_main; tunnel_lookup_t *l3_ifaces = &lgm->l3_ifaces; vnet_hw_interface_t *hi; uword *hip, *si; hip = hash_get (l3_ifaces->hw_if_index_by_dp_table, table_id); if (hip) { clib_warning ("vrf %d already mapped to a vni", table_id); return ~0; } si = hash_get (l3_ifaces->sw_if_index_by_vni, vni); if (si) { clib_warning ("Interface for vni %d already exists", vni); } /* create lisp iface and populate tunnel tables */ hi = lisp_gpe_create_iface (lgm, vni, table_id, &lisp_gpe_device_class, l3_ifaces); /* insert default routes that point to lisp-cp lookup */ lisp_gpe_iface_set_table (hi->sw_if_index, table_id); if (with_default_routes) lisp_gpe_tenant_add_default_routes (table_id); /* enable interface */ vnet_sw_interface_set_flags (vnm, hi->sw_if_index, VNET_SW_INTERFACE_FLAG_ADMIN_UP); vnet_hw_interface_set_flags (vnm, hi->hw_if_index, VNET_HW_INTERFACE_FLAG_LINK_UP); return (hi->sw_if_index); } void lisp_gpe_del_l3_iface (lisp_gpe_main_t * lgm, u32 vni, u32 table_id) { vnet_main_t *vnm = lgm->vnet_main; tunnel_lookup_t *l3_ifaces = &lgm->l3_ifaces; vnet_hw_interface_t *hi; uword *hip; hip = hash_get (l3_ifaces->hw_if_index_by_dp_table, table_id); if (hip == 0) { clib_warning ("The interface for vrf %d doesn't exist", table_id); return; } hi = vnet_get_hw_interface (vnm, hip[0]); lisp_gpe_remove_iface (lgm, hip[0], table_id, &lgm->l3_ifaces); /* unset default routes */ ip4_sw_interface_enable_disable (hi->sw_if_index, 0); ip6_sw_interface_enable_disable (hi->sw_if_index, 0); lisp_gpe_tenant_del_default_routes (table_id); } /** * @brief Add/del LISP-GPE L2 interface. * * Creates LISP-GPE interface, sets it in L2 mode in the appropriate * bridge domain, sets egress arcs and enables it. * * @param[in] lgm Reference to @ref lisp_gpe_main_t. * @param[in] a Parameters to create interface. * * @return number of vectors in frame. */ u32 lisp_gpe_add_l2_iface (lisp_gpe_main_t * lgm, u32 vni, u32 bd_id) { vnet_main_t *vnm = lgm->vnet_main; tunnel_lookup_t *l2_ifaces = &lgm->l2_ifaces; vnet_hw_interface_t *hi; uword *hip, *si; u16 bd_index; if (bd_id > L2_BD_ID_MAX) { clib_warning ("bridge domain ID %d exceed 16M limit", bd_id); return ~0; } bd_index = bd_find_or_add_bd_index (&bd_main, bd_id); hip = hash_get (l2_ifaces->hw_if_index_by_dp_table, bd_index); if (hip) { clib_warning ("bridge domain %d already mapped to a vni", bd_id); return ~0; } si = hash_get (l2_ifaces->sw_if_index_by_vni, vni); if (si) { clib_warning ("Interface for vni %d already exists", vni); return ~0; } /* create lisp iface and populate tunnel tables */ hi = lisp_gpe_create_iface (lgm, vni, bd_index, &l2_lisp_gpe_device_class, &lgm->l2_ifaces); /* enable interface */ vnet_sw_interface_set_flags (vnm, hi->sw_if_index, VNET_SW_INTERFACE_FLAG_ADMIN_UP); vnet_hw_interface_set_flags (vnm, hi->hw_if_index, VNET_HW_INTERFACE_FLAG_LINK_UP); l2_arc_to_lb = vlib_node_add_named_next (vlib_get_main (), hi->tx_node_index, "l2-load-balance"); /* we're ready. add iface to l2 bridge domain */ set_int_l2_mode (lgm->vlib_main, vnm, MODE_L2_BRIDGE, hi->sw_if_index, bd_index, L2_BD_PORT_TYPE_NORMAL, 0, 0); return (hi->sw_if_index); } /** * @brief Add/del LISP-GPE L2 interface. * * Creates LISP-GPE interface, sets it in L2 mode in the appropriate * bridge domain, sets egress arcs and enables it. * * @param[in] lgm Reference to @ref lisp_gpe_main_t. * @param[in] a Parameters to create interface. * * @return number of vectors in frame. */ void lisp_gpe_del_l2_iface (lisp_gpe_main_t * lgm, u32 vni, u32 bd_id) { tunnel_lookup_t *l2_ifaces = &lgm->l2_ifaces; vnet_hw_interface_t *hi; u32 bd_index = bd_find_index (&bd_main, bd_id); ASSERT (bd_index != ~0); uword *hip = hash_get (l2_ifaces->hw_if_index_by_dp_table, bd_index); if (hip == 0) { clib_warning ("The interface for bridge domain %d doesn't exist", bd_id); return; } /* Remove interface from bridge .. by enabling L3 mode */ hi = vnet_get_hw_interface (lgm->vnet_main, hip[0]); set_int_l2_mode (lgm->vlib_main, lgm->vnet_main, MODE_L3, hi->sw_if_index, 0, L2_BD_PORT_TYPE_NORMAL, 0, 0); lisp_gpe_remove_iface (lgm, hip[0], bd_index, &lgm->l2_ifaces); } /** * @brief Add LISP-GPE NSH interface. * * Creates LISP-GPE interface, sets it in L3 mode. * * @param[in] lgm Reference to @ref lisp_gpe_main_t. * @param[in] a Parameters to create interface. * * @return sw_if_index. */ u32 vnet_lisp_gpe_add_nsh_iface (lisp_gpe_main_t * lgm) { vnet_main_t *vnm = lgm->vnet_main; tunnel_lookup_t *nsh_ifaces = &lgm->nsh_ifaces; vnet_hw_interface_t *hi; uword *hip, *si; hip = hash_get (nsh_ifaces->hw_if_index_by_dp_table, 0); if (hip) { clib_warning ("NSH interface 0 already exists"); return ~0; } si = hash_get (nsh_ifaces->sw_if_index_by_vni, 0); if (si) { clib_warning ("NSH interface already exists"); return ~0; } /* create lisp iface and populate tunnel tables */ hi = lisp_gpe_create_iface (lgm, 0, 0, &nsh_lisp_gpe_device_class, &lgm->nsh_ifaces); /* enable interface */ vnet_sw_interface_set_flags (vnm, hi->sw_if_index, VNET_SW_INTERFACE_FLAG_ADMIN_UP); vnet_hw_interface_set_flags (vnm, hi->hw_if_index, VNET_HW_INTERFACE_FLAG_LINK_UP); return (hi->sw_if_index); } /** * @brief Del LISP-GPE NSH interface. * */ void vnet_lisp_gpe_del_nsh_iface (lisp_gpe_main_t * lgm) { tunnel_lookup_t *nsh_ifaces = &lgm->nsh_ifaces; uword *hip; hip = hash_get (nsh_ifaces->hw_if_index_by_dp_table, 0); if (hip == 0) { clib_warning ("The NSH 0 interface doesn't exist"); return; } lisp_gpe_remove_iface (lgm, hip[0], 0, &lgm->nsh_ifaces); } static clib_error_t * lisp_gpe_add_del_iface_command_fn (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd) { unformat_input_t _line_input, *line_input = &_line_input; u8 is_add = 1; u32 table_id, vni, bd_id; u8 vni_is_set = 0, vrf_is_set = 0, bd_index_is_set = 0; u8 nsh_iface = 0; clib_error_t *error = NULL; if (vnet_lisp_gpe_enable_disable_status () == 0) { return clib_error_return (0, "LISP is disabled"); } /* Get a line of input. */ if (!unformat_user (input, unformat_line_input, line_input)) return 0; while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT) { if (unformat (line_input, "add")) is_add = 1; else if (unformat (line_input, "del")) is_add = 0; else if (unformat (line_input, "vrf %d", &table_id)) { vrf_is_set = 1; } else if (unformat (line_input, "vni %d", &vni)) { vni_is_set = 1; } else if (unformat (line_input, "bd %d", &bd_id)) { bd_index_is_set = 1; } else if (unformat (line_input, "nsh")) { nsh_iface = 1; } else { error = clib_error_return (0, "parse error: '%U'", format_unformat_error, line_input); goto done; } } if (nsh_iface) { if (is_add) { if (~0 == vnet_lisp_gpe_add_nsh_iface (&lisp_gpe_main)) { error = clib_error_return (0, "NSH interface not created"); goto done; } } else { vnet_lisp_gpe_del_nsh_iface (&lisp_gpe_main); } goto done; } if (vrf_is_set && bd_index_is_set) { error = clib_error_return (0, "Cannot set both vrf and brdige domain index!"); goto done; } if (!vni_is_set) { error = clib_error_return (0, "vni must be set!"); goto done; } if (!vrf_is_set && !bd_index_is_set) { error = clib_error_return (0, "vrf or bridge domain index must be set!"); goto done; } if (bd_index_is_set) { if (is_add) { if (~0 == lisp_gpe_tenant_l2_iface_add_or_lock (vni, bd_id)) { error = clib_error_return (0, "L2 interface not created"); goto done; } } else lisp_gpe_tenant_l2_iface_unlock (vni); } else { if (is_add) { if (~0 == lisp_gpe_tenant_l3_iface_add_or_lock (vni, table_id, 1 /* with_default_route */ )) { error = clib_error_return (0, "L3 interface not created"); goto done; } } else lisp_gpe_tenant_l3_iface_unlock (vni); } done: unformat_free (line_input); return error; } /* *INDENT-OFF* */ VLIB_CLI_COMMAND (add_del_lisp_gpe_iface_command, static) = { .path = "gpe iface", .short_help = "gpe iface add/del vni <vni> vrf <vrf>", .function = lisp_gpe_add_del_iface_command_fn, }; /* *INDENT-ON* */ /* * fd.io coding-style-patch-verification: ON * * Local Variables: * eval: (c-set-style "gnu") * End: */