aboutsummaryrefslogtreecommitdiffstats
path: root/src/vnet/devices/virtio/vhost_user_input.c
blob: 465ec4462a83b5631e70fcd054e56601422ee832 (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
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
/*
 *------------------------------------------------------------------
 * vhost-user-input
 *
 * Copyright (c) 2014-2018 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 <fcntl.h>		/* for open */
#include <sys/ioctl.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/uio.h>		/* for iovec */
#include <netinet/in.h>
#include <sys/vfs.h>

#include <linux/if_arp.h>
#include <linux/if_tun.h>

#include <vlib/vlib.h>
#include <vlib/unix/unix.h>

#include <vnet/ip/ip.h>

#include <vnet/ethernet/ethernet.h>
#include <vnet/devices/devices.h>
#include <vnet/feature/feature.h>

#include <vnet/devices/virtio/vhost_user.h>
#include <vnet/devices/virtio/vhost_user_inline.h>

/*
 * When an RX queue is down but active, received packets
 * must be discarded. This value controls up to how many
 * packets will be discarded during each round.
 */
#define VHOST_USER_DOWN_DISCARD_COUNT 256

/*
 * When the number of available buffers gets under this threshold,
 * RX node will start discarding packets.
 */
#define VHOST_USER_RX_BUFFER_STARVATION 32

/*
 * On the receive side, the host should free descriptors as soon
 * as possible in order to avoid TX drop in the VM.
 * This value controls the number of copy operations that are stacked
 * before copy is done for all and descriptors are given back to
 * the guest.
 * The value 64 was obtained by testing (48 and 128 were not as good).
 */
#define VHOST_USER_RX_COPY_THRESHOLD 64

vlib_node_registration_t vhost_user_input_node;

#define foreach_vhost_user_input_func_error      \
  _(NO_ERROR, "no error")  \
  _(NO_BUFFER, "no available buffer")  \
  _(MMAP_FAIL, "mmap failure")  \
  _(INDIRECT_OVERFLOW, "indirect descriptor overflows table")  \
  _(UNDERSIZED_FRAME, "undersized ethernet frame received (< 14 bytes)") \
  _(FULL_RX_QUEUE, "full rx queue (possible driver tx drop)")

typedef enum
{
#define _(f,s) VHOST_USER_INPUT_FUNC_ERROR_##f,
  foreach_vhost_user_input_func_error
#undef _
    VHOST_USER_INPUT_FUNC_N_ERROR,
} vhost_user_input_func_error_t;

static __clib_unused char *vhost_user_input_func_error_strings[] = {
#define _(n,s) s,
  foreach_vhost_user_input_func_error
#undef _
};

static_always_inline void
vhost_user_rx_trace (vhost_trace_t * t,
		     vhost_user_intf_t * vui, u16 qid,
		     vlib_buffer_t * b, vhost_user_vring_t * txvq,
		     u16 last_avail_idx)
{
  vhost_user_main_t *vum = &vhost_user_main;
  u32 desc_current = txvq->avail->ring[last_avail_idx & txvq->qsz_mask];
  vring_desc_t *hdr_desc = 0;
  virtio_net_hdr_mrg_rxbuf_t *hdr;
  u32 hint = 0;

  clib_memset (t, 0, sizeof (*t));
  t->device_index = vui - vum->vhost_user_interfaces;
  t->qid = qid;

  hdr_desc = &txvq->desc[desc_current];
  if (txvq->desc[desc_current].flags & VIRTQ_DESC_F_INDIRECT)
    {
      t->virtio_ring_flags |= 1 << VIRTIO_TRACE_F_INDIRECT;
      /* Header is the first here */
      hdr_desc = map_guest_mem (vui, txvq->desc[desc_current].addr, &hint);
    }
  if (txvq->desc[desc_current].flags & VIRTQ_DESC_F_NEXT)
    {
      t->virtio_ring_flags |= 1 << VIRTIO_TRACE_F_SIMPLE_CHAINED;
    }
  if (!(txvq->desc[desc_current].flags & VIRTQ_DESC_F_NEXT) &&
      !(txvq->desc[desc_current].flags & VIRTQ_DESC_F_INDIRECT))
    {
      t->virtio_ring_flags |= 1 << VIRTIO_TRACE_F_SINGLE_DESC;
    }

  t->first_desc_len = hdr_desc ? hdr_desc->len : 0;

  if (!hdr_desc || !(hdr = map_guest_mem (vui, hdr_desc->addr, &hint)))
    {
      t->virtio_ring_flags |= 1 << VIRTIO_TRACE_F_MAP_ERROR;
    }
  else
    {
      u32 len = vui->virtio_net_hdr_sz;
      memcpy (&t->hdr, hdr, len > hdr_desc->len ? hdr_desc->len : len);
    }
}

static_always_inline u32
vhost_user_input_copy (vhost_user_intf_t * vui, vhost_copy_t * cpy,
		       u16 copy_len, u32 * map_hint)
{
  void *src0, *src1, *src2, *src3;
  if (PREDICT_TRUE (copy_len >= 4))
    {
      if (PREDICT_FALSE (!(src2 = map_guest_mem (vui, cpy[0].src, map_hint))))
	return 1;
      if (PREDICT_FALSE (!(src3 = map_guest_mem (vui, cpy[1].src, map_hint))))
	return 1;

      while (PREDICT_TRUE (copy_len >= 4))
	{
	  src0 = src2;
	  src1 = src3;

	  if (PREDICT_FALSE
	      (!(src2 = map_guest_mem (vui, cpy[2].src, map_hint))))
	    return 1;
	  if (PREDICT_FALSE
	      (!(src3 = map_guest_mem (vui, cpy[3].src, map_hint))))
	    return 1;

	  CLIB_PREFETCH (src2, 64, LOAD);
	  CLIB_PREFETCH (src3, 64, LOAD);

	  clib_memcpy_fast ((void *) cpy[0].dst, src0, cpy[0].len);
	  clib_memcpy_fast ((void *) cpy[1].dst, src1, cpy[1].len);
	  copy_len -= 2;
	  cpy += 2;
	}
    }
  while (copy_len)
    {
      if (PREDICT_FALSE (!(src0 = map_guest_mem (vui, cpy->src, map_hint))))
	return 1;
      clib_memcpy_fast ((void *) cpy->dst, src0, cpy->len);
      copy_len -= 1;
      cpy += 1;
    }
  return 0;
}

/**
 * Try to discard packets from the tx ring (VPP RX path).
 * Returns the number of discarded packets.
 */
static_always_inline u32
vhost_user_rx_discard_packet (vlib_main_t * vm,
			      vhost_user_intf_t * vui,
			      vhost_user_vring_t * txvq, u32 discard_max)
{
  /*
   * On the RX side, each packet corresponds to one descriptor
   * (it is the same whether it is a shallow descriptor, chained, or indirect).
   * Therefore, discarding a packet is like discarding a descriptor.
   */
  u32 discarded_packets = 0;
  u32 avail_idx = txvq->avail->idx;
  u16 mask = txvq->qsz_mask;
  u16 last_avail_idx = txvq->last_avail_idx;
  u16 last_used_idx = txvq->last_used_idx;
  while (discarded_packets != discard_max)
    {
      if (avail_idx == txvq->last_avail_idx)
	goto out;

      u16 desc_chain_head = txvq->avail->ring[last_avail_idx & mask];
      last_avail_idx++;
      txvq->used->ring[last_used_idx & mask].id = desc_chain_head;
      txvq->used->ring[last_used_idx & mask].len = 0;
      vhost_user_log_dirty_ring (vui, txvq, ring[last_used_idx & mask]);
      last_used_idx++;
      discarded_packets++;
    }

out:
  txvq->last_avail_idx = last_avail_idx;
  txvq->last_used_idx = last_used_idx;
  CLIB_MEMORY_STORE_BARRIER ();
  txvq->used->idx = txvq->last_used_idx;
  vhost_user_log_dirty_ring (vui, txvq, idx);
  return discarded_packets;
}

/*
 * In case of overflow, we need to rewind the array of allocated buffers.
 */
static_always_inline void
vhost_user_input_rewind_buffers (vlib_main_t * vm,
				 vhost_cpu_t * cpu, vlib_buffer_t * b_head)
{
  u32 bi_current = cpu->rx_buffers[cpu->rx_buffers_len];
  vlib_buffer_t *b_current = vlib_get_buffer (vm, bi_current);
  b_current->current_length = 0;
  b_current->flags = 0;
  while (b_current != b_head)
    {
      cpu->rx_buffers_len++;
      bi_current = cpu->rx_buffers[cpu->rx_buffers_len];
      b_current = vlib_get_buffer (vm, bi_current);
      b_current->current_length = 0;
      b_current->flags = 0;
    }
  cpu->rx_buffers_len++;
}

static_always_inline u32
vhost_user_if_input (vlib_main_t * vm,
		     vhost_user_main_t * vum,
		     vhost_user_intf_t * vui,
		     u16 qid, vlib_node_runtime_t * node,
		     vnet_hw_interface_rx_mode mode)
{
  vhost_user_vring_t *txvq = &vui->vrings[VHOST_VRING_IDX_TX (qid)];
  vnet_feature_main_t *fm = &feature_main;
  u16 n_rx_packets = 0;
  u32 n_rx_bytes = 0;
  u16 n_left;
  u32 n_left_to_next, *to_next;
  u32 next_index = VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT;
  u32 n_trace = vlib_get_trace_count (vm, node);
  u32 map_hint = 0;
  vhost_cpu_t *cpu = &vum->cpus[vm->thread_index];
  u16 copy_len = 0;
  u8 feature_arc_idx = fm->device_input_feature_arc_index;
  u32 current_config_index = ~(u32) 0;
  u16 mask = txvq->qsz_mask;

  /* The descriptor table is not ready yet */
  if (PREDICT_FALSE (txvq->avail == 0))
    goto done;

  {
    /* do we have pending interrupts ? */
    vhost_user_vring_t *rxvq = &vui->vrings[VHOST_VRING_IDX_RX (qid)];
    f64 now = vlib_time_now (vm);

    if ((txvq->n_since_last_int) && (txvq->int_deadline < now))
      vhost_user_send_call (vm, txvq);

    if ((rxvq->n_since_last_int) && (rxvq->int_deadline < now))
      vhost_user_send_call (vm, rxvq);
  }

  /*
   * For adaptive mode, it is optimized to reduce interrupts.
   * If the scheduler switches the input node to polling due
   * to burst of traffic, we tell the driver no interrupt.
   * When the traffic subsides, the scheduler switches the node back to
   * interrupt mode. We must tell the driver we want interrupt.
   */
  if (PREDICT_FALSE (mode == VNET_HW_INTERFACE_RX_MODE_ADAPTIVE))
    {
      if ((node->flags &
	   VLIB_NODE_FLAG_SWITCH_FROM_POLLING_TO_INTERRUPT_MODE) ||
	  !(node->flags &
	    VLIB_NODE_FLAG_SWITCH_FROM_INTERRUPT_TO_POLLING_MODE))
	/* Tell driver we want notification */
	txvq->used->flags = 0;
      else
	/* Tell driver we don't want notification */
	txvq->used->flags = VRING_USED_F_NO_NOTIFY;
    }

  if (PREDICT_FALSE (txvq->avail->flags & 0xFFFE))
    goto done;

  n_left = (u16) (txvq->avail->idx - txvq->last_avail_idx);

  /* nothing to do */
  if (PREDICT_FALSE (n_left == 0))
    goto done;

  if (PREDICT_FALSE (!vui->admin_up || !(txvq->enabled)))
    {
      /*
       * Discard input packet if interface is admin down or vring is not
       * enabled.
       * "For example, for a networking device, in the disabled state
       * client must not supply any new RX packets, but must process
       * and discard any TX packets."
       */
      vhost_user_rx_discard_packet (vm, vui, txvq,
				    VHOST_USER_DOWN_DISCARD_COUNT);
      goto done;
    }

  if (PREDICT_FALSE (n_left == (mask + 1)))
    {
      /*
       * Informational error logging when VPP is not
       * receiving packets fast enough.
       */
      vlib_error_count (vm, node->node_index,
			VHOST_USER_INPUT_FUNC_ERROR_FULL_RX_QUEUE, 1);
    }

  if (n_left > VLIB_FRAME_SIZE)
    n_left = VLIB_FRAME_SIZE;

  /*
   * For small packets (<2kB), we will not need more than one vlib buffer
   * per packet. In case packets are bigger, we will just yeld at some point
   * in the loop and come back later. This is not an issue as for big packet,
   * processing cost really comes from the memory copy.
   * The assumption is that big packets will fit in 40 buffers.
   */
  if (PREDICT_FALSE (cpu->rx_buffers_len < n_left + 1 ||
		     cpu->rx_buffers_len < 40))
    {
      u32 curr_len = cpu->rx_buffers_len;
      cpu->rx_buffers_len +=
	vlib_buffer_alloc (vm, cpu->rx_buffers + curr_len,
			   VHOST_USER_RX_BUFFERS_N - curr_len);

      if (PREDICT_FALSE
	  (cpu->rx_buffers_len < VHOST_USER_RX_BUFFER_STARVATION))
	{
	  /* In case of buffer starvation, discard some packets from the queue
	   * and log the event.
	   * We keep doing best effort for the remaining packets. */
	  u32 flush = (n_left + 1 > cpu->rx_buffers_len) ?
	    n_left + 1 - cpu->rx_buffers_len : 1;
	  flush = vhost_user_rx_discard_packet (vm, vui, txvq, flush);

	  n_left -= flush;
	  vlib_increment_simple_counter (vnet_main.
					 interface_main.sw_if_counters +
					 VNET_INTERFACE_COUNTER_DROP,
					 vm->thread_index, vui->sw_if_index,
					 flush);

	  vlib_error_count (vm, vhost_user_input_node.index,
			    VHOST_USER_INPUT_FUNC_ERROR_NO_BUFFER, flush);
	}
    }

  if (PREDICT_FALSE (vnet_have_features (feature_arc_idx, vui->sw_if_index)))
    {
      vnet_feature_config_main_t *cm;
      cm = &fm->feature_config_mains[feature_arc_idx];
      current_config_index = vec_elt (cm->config_index_by_sw_if_index,
				      vui->sw_if_index);
      vnet_get_config_data (&cm->config_main, &current_config_index,
			    &next_index, 0);
    }

  u16 last_avail_idx = txvq->last_avail_idx;
  u16 last_used_idx = txvq->last_used_idx;

  vlib_get_new_next_frame (vm, node, next_index, to_next, n_left_to_next);

  if (next_index == VNET_DEVICE_INPUT_NEXT_ETHERNET_INPUT)
    {
      /* give some hints to ethernet-input */
      vlib_next_frame_t *nf;
      vlib_frame_t *f;
      ethernet_input_frame_t *ef;
      nf = vlib_node_runtime_get_next_frame (vm, node, next_index);
      f = vlib_get_frame (vm, nf->frame_index);
      f->flags = ETH_INPUT_FRAME_F_SINGLE_SW_IF_IDX;

      ef = vlib_frame_scalar_args (f);
      ef->sw_if_index = vui->sw_if_index;
      ef->hw_if_index = vui->hw_if_index;
    }

  while (n_left > 0)
    {
      vlib_buffer_t *b_head, *b_current;
      u32 bi_current;
      u16 desc_current;
      u32 desc_data_offset;
      vring_desc_t *desc_table = txvq->desc;

      if (PREDICT_FALSE (cpu->rx_buffers_len <= 1))
	{
	  /* Not enough rx_buffers
	   * Note: We yeld on 1 so we don't need to do an additional
	   * check for the next buffer prefetch.
	   */
	  n_left = 0;
	  break;
	}

      desc_current = txvq->avail->ring[last_avail_idx & mask];
      cpu->rx_buffers_len--;
      bi_current = cpu->rx_buffers[cpu->rx_buffers_len];
      b_head = b_current = vlib_get_buffer (vm, bi_current);
      to_next[0] = bi_current;	//We do that now so we can forget about bi_current
      to_next++;
      n_left_to_next--;

      vlib_prefetch_buffer_with_index
	(vm, cpu->rx_buffers[cpu->rx_buffers_len - 1], LOAD);

      /* Just preset the used descriptor id and length for later */
      txvq->used->ring[last_used_idx & mask].id = desc_current;
      txvq->used->ring[last_used_idx & mask].len = 0;
      vhost_user_log_dirty_ring (vui, txvq, ring[last_used_idx & mask]);

      /* The buffer should already be initialized */
      b_head->total_length_not_including_first_buffer = 0;
      b_head->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID;

      if (PREDICT_FALSE (n_trace))
	{
	  //TODO: next_index is not exactly known at that point
	  vlib_trace_buffer (vm, node, next_index, b_head,
			     /* follow_chain */ 0);
	  vhost_trace_t *t0 =
	    vlib_add_trace (vm, node, b_head, sizeof (t0[0]));
	  vhost_user_rx_trace (t0, vui, qid, b_head, txvq, last_avail_idx);
	  n_trace--;
	  vlib_set_trace_count (vm, node, n_trace);
	}

      /* This depends on the setup but is very consistent
       * So I think the CPU branch predictor will make a pretty good job
       * at optimizing the decision. */
      if (txvq->desc[desc_current].flags & VIRTQ_DESC_F_INDIRECT)
	{
	  desc_table = map_guest_mem (vui, txvq->desc[desc_current].addr,
				      &map_hint);
	  desc_current = 0;
	  if (PREDICT_FALSE (desc_table == 0))
	    {
	      vlib_error_count (vm, node->node_index,
				VHOST_USER_INPUT_FUNC_ERROR_MMAP_FAIL, 1);
	      goto out;
	    }
	}

      if (PREDICT_TRUE (vui->is_any_layout) ||
	  (!(desc_table[desc_current].flags & VIRTQ_DESC_F_NEXT)))
	{
	  /* ANYLAYOUT or single buffer */
	  desc_data_offset = vui->virtio_net_hdr_sz;
	}
      else
	{
	  /* CSR case without ANYLAYOUT, skip 1st buffer */
	  desc_data_offset = desc_table[desc_current].len;
	}

      while (1)
	{
	  /* Get more input if necessary. Or end of packet. */
	  if (desc_data_offset == desc_table[desc_current].len)
	    {
	      if (PREDICT_FALSE (desc_table[desc_current].flags &
				 VIRTQ_DESC_F_NEXT))
		{
		  desc_current = desc_table[desc_current].next;
		  desc_data_offset = 0;
		}
	      else
		{
		  goto out;
		}
	    }

	  /* Get more output if necessary. Or end of packet. */
	  if (PREDICT_FALSE
	      (b_current->current_length == vlib_bufer_get_default_size (vm)))
	    {
	      if (PREDICT_FALSE (cpu->rx_buffers_len == 0))
		{
		  /* Cancel speculation */
		  to_next--;
		  n_left_to_next++;

		  /*
		   * Checking if there are some left buffers.
		   * If not, just rewind the used buffers and stop.
		   * Note: Scheduled copies are not cancelled. This is
		   * not an issue as they would still be valid. Useless,
		   * but valid.
		   */
		  vhost_user_input_rewind_buffers (vm, cpu, b_head);
		  n_left = 0;
		  goto stop;
		}

	      /* Get next output */
	      cpu->rx_buffers_len--;
	      u32 bi_next = cpu->rx_buffers[cpu->rx_buffers_len];
	      b_current->next_buffer = bi_next;
	      b_current->flags |= VLIB_BUFFER_NEXT_PRESENT;
	      bi_current = bi_next;
	      b_current = vlib_get_buffer (vm, bi_current);
	    }

	  /* Prepare a copy order executed later for the data */
	  vhost_copy_t *cpy = &cpu->copy[copy_len];
	  copy_len++;
	  u32 desc_data_l = desc_table[desc_current].len - desc_data_offset;
	  cpy->len =
	    vlib_bufer_get_default_size (vm) - b_current->current_length;
	  cpy->len = (cpy->len > desc_data_l) ? desc_data_l : cpy->len;
	  cpy->dst = (uword) (vlib_buffer_get_current (b_current) +
			      b_current->current_length);
	  cpy->src = desc_table[desc_current].addr + desc_data_offset;

	  desc_data_offset += cpy->len;

	  b_current->current_length += cpy->len;
	  b_head->total_length_not_including_first_buffer += cpy->len;
	}

    out:

      n_rx_bytes += b_head->total_length_not_including_first_buffer;
      n_rx_packets++;

      b_head->total_length_not_including_first_buffer -=
	b_head->current_length;

      /* consume the descriptor and return it as used */
      last_avail_idx++;
      last_used_idx++;

      VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b_head);

      vnet_buffer (b_head)->sw_if_index[VLIB_RX] = vui->sw_if_index;
      vnet_buffer (b_head)->sw_if_index[VLIB_TX] = (u32) ~ 0;
      b_head->error = 0;

      if (current_config_index != ~(u32) 0)
	{
	  b_head->current_config_index = current_config_index;
	  vnet_buffer (b_head)->feature_arc_index = feature_arc_idx;
	}

      n_left--;

      /*
       * Although separating memory copies from virtio ring parsing
       * is beneficial, we can offer to perform the copies from time
       * to time in order to free some space in the ring.
       */
      if (PREDICT_FALSE (copy_len >= VHOST_USER_RX_COPY_THRESHOLD))
	{
	  if (PREDICT_FALSE (vhost_user_input_copy (vui, cpu->copy,
						    copy_len, &map_hint)))
	    {
	      vlib_error_count (vm, node->node_index,
				VHOST_USER_INPUT_FUNC_ERROR_MMAP_FAIL, 1);
	    }
	  copy_len = 0;

	  /* give buffers back to driver */
	  CLIB_MEMORY_STORE_BARRIER ();
	  txvq->used->idx = last_used_idx;
	  vhost_user_log_dirty_ring (vui, txvq, idx);
	}
    }
stop:
  vlib_put_next_frame (vm, node, next_index, n_left_to_next);

  txvq->last_used_idx = last_used_idx;
  txvq->last_avail_idx = last_avail_idx;

  /* Do the memory copies */
  if (PREDICT_FALSE (vhost_user_input_copy (vui, cpu->copy, copy_len,
					    &map_hint)))
    {
      vlib_error_count (vm, node->node_index,
			VHOST_USER_INPUT_FUNC_ERROR_MMAP_FAIL, 1);
    }

  /* give buffers back to driver */
  CLIB_MEMORY_STORE_BARRIER ();
  txvq->used->idx = txvq->last_used_idx;
  vhost_user_log_dirty_ring (vui, txvq, idx);

  /* interrupt (call) handling */
  if ((txvq->callfd_idx != ~0) &&
      !(txvq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT))
    {
      txvq->n_since_last_int += n_rx_packets;

      if (txvq->n_since_last_int > vum->coalesce_frames)
	vhost_user_send_call (vm, txvq);
    }

  /* increase rx counters */
  vlib_increment_combined_counter
    (vnet_main.interface_main.combined_sw_if_counters
     + VNET_INTERFACE_COUNTER_RX, vm->thread_index, vui->sw_if_index,
     n_rx_packets, n_rx_bytes);

  vnet_device_increment_rx_packets (vm->thread_index, n_rx_packets);

done:
  return n_rx_packets;
}

VLIB_NODE_FN (vhost_user_input_node) (vlib_main_t * vm,
				      vlib_node_runtime_t * node,
				      vlib_frame_t * frame)
{
  vhost_user_main_t *vum = &vhost_user_main;
  uword n_rx_packets = 0;
  vhost_user_intf_t *vui;
  vnet_device_input_runtime_t *rt =
    (vnet_device_input_runtime_t *) node->runtime_data;
  vnet_device_and_queue_t *dq;

  vec_foreach (dq, rt->devices_and_queues)
  {
    if ((node->state == VLIB_NODE_STATE_POLLING) ||
	clib_atomic_swap_acq_n (&dq->interrupt_pending, 0))
      {
	vui =
	  pool_elt_at_index (vum->vhost_user_interfaces, dq->dev_instance);
	n_rx_packets += vhost_user_if_input (vm, vum, vui, dq->queue_id, node,
					     dq->mode);
      }
  }

  return n_rx_packets;
}

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (vhost_user_input_node) = {
  .type = VLIB_NODE_TYPE_INPUT,
  .name = "vhost-user-input",
  .sibling_of = "device-input",

  /* Will be enabled if/when hardware is detected. */
  .state = VLIB_NODE_STATE_DISABLED,

  .format_buffer = format_ethernet_header_with_length,
  .format_trace = format_vhost_trace,

  .n_errors = VHOST_USER_INPUT_FUNC_N_ERROR,
  .error_strings = vhost_user_input_func_error_strings,
};
/* *INDENT-ON* */

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */