summaryrefslogtreecommitdiffstats
path: root/src/vnet/adj/adj_midchain_delegate.c
AgeCommit message (Expand)AuthorFilesLines
2019-07-19fib: FIB Entry trackingNeale Ranns1-9/+7
2019-03-26ADJ: midchain delegate to performing stackingNeale Ranns1-0/+183
79' href='#n79'>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
/*
 * Copyright (c) 2017 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 <vlib/vlib.h>
#include <vnet/vnet.h>
#include <vnet/pg/pg.h>
#include <vnet/ip/ip.h>

#include <vnet/tcp/tcp.h>

#include <vppinfra/hash.h>
#include <vppinfra/error.h>
#include <vppinfra/elog.h>
#include <vlibmemory/unix_shared_memory_queue.h>

#include <vnet/udp/udp_packet.h>
#include <vnet/lisp-cp/packets.h>
#include <math.h>

vlib_node_registration_t session_queue_node;

typedef struct
{
  u32 session_index;
  u32 server_thread_index;
} session_queue_trace_t;

/* packet trace format function */
static u8 *
format_session_queue_trace (u8 * s, va_list * args)
{
  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
  session_queue_trace_t *t = va_arg (*args, session_queue_trace_t *);

  s = format (s, "SESSION_QUEUE: session index %d, server thread index %d",
	      t->session_index, t->server_thread_index);
  return s;
}

vlib_node_registration_t session_queue_node;

#define foreach_session_queue_error                 \
_(TX, "Packets transmitted")                    \
_(TIMER, "Timer events")

typedef enum
{
#define _(sym,str) SESSION_QUEUE_ERROR_##sym,
  foreach_session_queue_error
#undef _
    SESSION_QUEUE_N_ERROR,
} session_queue_error_t;

static char *session_queue_error_strings[] = {
#define _(sym,string) string,
  foreach_session_queue_error
#undef _
};

static u32 session_type_to_next[] = {
  SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT,
  SESSION_QUEUE_NEXT_IP4_LOOKUP,
  SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT,
  SESSION_QUEUE_NEXT_IP6_LOOKUP,
};

always_inline int
session_fifo_rx_i (vlib_main_t * vm, vlib_node_runtime_t * node,
		   session_manager_main_t * smm, session_fifo_event_t * e0,
		   stream_session_t * s0, u32 thread_index, int *n_tx_packets,
		   u8 peek_data)
{
  u32 n_trace = vlib_get_trace_count (vm, node);
  u32 left_to_snd0, max_len_to_snd0, len_to_deq0, n_bufs, snd_space0;
  u32 n_frame_bytes, n_frames_per_evt;
  transport_connection_t *tc0;
  transport_proto_vft_t *transport_vft;
  u32 next_index, next0, *to_next, n_left_to_next, bi0;
  vlib_buffer_t *b0;
  u32 rx_offset;
  u16 snd_mss0;
  u8 *data0;
  int i;

  next_index = next0 = session_type_to_next[s0->session_type];

  transport_vft = session_get_transport_vft (s0->session_type);
  tc0 = transport_vft->get_connection (s0->connection_index, thread_index);

  /* Make sure we have space to send and there's something to dequeue */
  snd_space0 = transport_vft->send_space (tc0);
  snd_mss0 = transport_vft->send_mss (tc0);

  if (snd_space0 == 0 || svm_fifo_max_dequeue (s0->server_tx_fifo) == 0
      || snd_mss0 == 0)
    return 0;

  ASSERT (e0->enqueue_length > 0);

  /* Ensure we're not writing more than transport window allows */
  max_len_to_snd0 = clib_min (e0->enqueue_length, snd_space0);

  if (peek_data)
    {
      /* Offset in rx fifo from where to peek data  */
      rx_offset = transport_vft->rx_fifo_offset (tc0);
    }

  /* TODO check if transport is willing to send len_to_snd0
   * bytes (Nagle) */

  n_frame_bytes = snd_mss0 * VLIB_FRAME_SIZE;
  n_frames_per_evt = ceil ((double) max_len_to_snd0 / n_frame_bytes);

  n_bufs = vec_len (smm->tx_buffers[thread_index]);
  left_to_snd0 = max_len_to_snd0;
  for (i = 0; i < n_frames_per_evt; i++)
    {
      /* Make sure we have at least one full frame of buffers ready */
      if (PREDICT_FALSE (n_bufs < VLIB_FRAME_SIZE))
	{
	  vec_validate (smm->tx_buffers[thread_index],
			n_bufs + VLIB_FRAME_SIZE - 1);
	  n_bufs +=
	    vlib_buffer_alloc (vm, &smm->tx_buffers[thread_index][n_bufs],
			       VLIB_FRAME_SIZE);

	  /* buffer shortage
	   * XXX 0.9 because when debugging we might not get a full frame */
	  if (PREDICT_FALSE (n_bufs < 0.9 * VLIB_FRAME_SIZE))
	    {
	      /* Keep track of how much we've dequeued and exit */
	      e0->enqueue_length -= max_len_to_snd0 - left_to_snd0;
	      return -1;
	    }

	  _vec_len (smm->tx_buffers[thread_index]) = n_bufs;
	}

      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
      while (left_to_snd0 && n_left_to_next)
	{
	  /* Get free buffer */
	  n_bufs--;
	  bi0 = smm->tx_buffers[thread_index][n_bufs];
	  _vec_len (smm->tx_buffers[thread_index]) = n_bufs;

	  b0 = vlib_get_buffer (vm, bi0);
	  b0->error = 0;
	  b0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID
	    | VNET_BUFFER_LOCALLY_ORIGINATED;
	  b0->current_data = 0;

	  /* RX on the local interface. tx in default fib */
	  vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0;
	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;

	  /* usual speculation, or the enqueue_x1 macro will barf */
	  to_next[0] = bi0;
	  to_next += 1;
	  n_left_to_next -= 1;

	  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b0);
	  if (PREDICT_FALSE (n_trace > 0))
	    {
	      session_queue_trace_t *t0;
	      vlib_trace_buffer (vm, node, next_index, b0,
				 1 /* follow_chain */ );
	      vlib_set_trace_count (vm, node, --n_trace);
	      t0 = vlib_add_trace (vm, node, b0, sizeof (*t0));
	      t0->session_index = s0->session_index;
	      t0->server_thread_index = s0->thread_index;
	    }

	  if (1)
	    {
	      ELOG_TYPE_DECLARE (e) =
	      {
	      .format = "evt-dequeue: id %d length %d",.format_args =
		  "i4i4",};
	      struct
	      {
		u32 data[2];
	      } *ed;
	      ed = ELOG_DATA (&vm->elog_main, e);
	      ed->data[0] = e0->event_id;
	      ed->data[1] = e0->enqueue_length;
	    }

	  len_to_deq0 = (left_to_snd0 < snd_mss0) ? left_to_snd0 : snd_mss0;

	  /* Make room for headers */
	  data0 = vlib_buffer_make_headroom (b0, MAX_HDRS_LEN);

	  /* Dequeue the data
	   * TODO 1) peek instead of dequeue
	   *      2) buffer chains */
	  if (peek_data)
	    {
	      int n_bytes_read;
	      n_bytes_read = svm_fifo_peek (s0->server_tx_fifo, s0->pid,
					    rx_offset, len_to_deq0, data0);
	      if (n_bytes_read < 0)
		goto dequeue_fail;

	      /* Keep track of progress locally, transport is also supposed to
	       * increment it independently when pushing header */
	      rx_offset += n_bytes_read;
	    }
	  else
	    {
	      if (svm_fifo_dequeue_nowait (s0->server_tx_fifo, s0->pid,
					   len_to_deq0, data0) < 0)
		goto dequeue_fail;
	    }

	  b0->current_length = len_to_deq0;

	  /* Ask transport to push header */
	  transport_vft->push_header (tc0, b0);

	  left_to_snd0 -= len_to_deq0;
	  *n_tx_packets = *n_tx_packets + 1;

	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
					   to_next, n_left_to_next,
					   bi0, next0);
	}
      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
    }

  /* If we couldn't dequeue all bytes store progress */
  if (max_len_to_snd0 < e0->enqueue_length)
    {
      e0->enqueue_length -= max_len_to_snd0;
      vec_add1 (smm->evts_partially_read[thread_index], *e0);
    }
  return 0;

dequeue_fail:
  /* Can't read from fifo. Store event rx progress, save as partially read,
   * return buff to free list and return  */
  e0->enqueue_length -= max_len_to_snd0 - left_to_snd0;
  vec_add1 (smm->evts_partially_read[thread_index], *e0);

  to_next -= 1;
  n_left_to_next += 1;
  _vec_len (smm->tx_buffers[thread_index]) += 1;

  clib_warning ("dequeue fail");
  return 0;
}

int
session_fifo_rx_peek (vlib_main_t * vm, vlib_node_runtime_t * node,
		      session_manager_main_t * smm, session_fifo_event_t * e0,
		      stream_session_t * s0, u32 thread_index, int *n_tx_pkts)
{
  return session_fifo_rx_i (vm, node, smm, e0, s0, thread_index, n_tx_pkts,
			    1);
}

int
session_fifo_rx_dequeue (vlib_main_t * vm, vlib_node_runtime_t * node,
			 session_manager_main_t * smm,
			 session_fifo_event_t * e0, stream_session_t * s0,
			 u32 thread_index, int *n_tx_pkts)
{
  return session_fifo_rx_i (vm, node, smm, e0, s0, thread_index, n_tx_pkts,
			    0);
}

static uword
session_queue_node_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
		       vlib_frame_t * frame)
{
  session_manager_main_t *smm = vnet_get_session_manager_main ();
  session_fifo_event_t *my_fifo_events, *e;
  u32 n_to_dequeue;
  unix_shared_memory_queue_t *q;
  int n_tx_packets = 0;
  u32 my_thread_index = vm->cpu_index;
  int i, rv;

  /*
   *  Update TCP time
   */
  tcp_update_time (vlib_time_now (vm), my_thread_index);

  /*
   * Get vpp queue events
   */
  q = smm->vpp_event_queues[my_thread_index];
  if (PREDICT_FALSE (q == 0))
    return 0;

  /* min number of events we can dequeue without blocking */
  n_to_dequeue = q->cursize;
  if (n_to_dequeue == 0)
    return 0;

  my_fifo_events = smm->fifo_events[my_thread_index];

  /* If we didn't manage to process previous events try going
   * over them again without dequeuing new ones.
   * XXX: Block senders to sessions that can't keep up */
  if (vec_len (my_fifo_events) >= 100)
    goto skip_dequeue;

  /* See you in the next life, don't be late */
  if (pthread_mutex_trylock (&q->mutex))
    return 0;

  for (i = 0; i < n_to_dequeue; i++)
    {
      vec_add2 (my_fifo_events, e, 1);
      unix_shared_memory_queue_sub_raw (q, (u8 *) e);
    }

  /* The other side of the connection is not polling */
  if (q->cursize < (q->maxsize / 8))
    (void) pthread_cond_broadcast (&q->condvar);
  pthread_mutex_unlock (&q->mutex);

  smm->fifo_events[my_thread_index] = my_fifo_events;

skip_dequeue:

  for (i = 0; i < n_to_dequeue; i++)
    {
      svm_fifo_t *f0;		/* $$$ prefetch 1 ahead maybe */
      stream_session_t *s0;
      u32 server_session_index0, server_thread_index0;
      session_fifo_event_t *e0;

      e0 = &my_fifo_events[i];
      f0 = e0->fifo;
      server_session_index0 = f0->server_session_index;
      server_thread_index0 = f0->server_thread_index;

      /* $$$ add multiple event queues, per vpp worker thread */
      ASSERT (server_thread_index0 == my_thread_index);

      s0 = pool_elt_at_index (smm->sessions[my_thread_index],
			      server_session_index0);

      ASSERT (s0->thread_index == my_thread_index);

      switch (e0->event_type)
	{
	case FIFO_EVENT_SERVER_TX:
	  /* Spray packets in per session type frames, since they go to
	   * different nodes */
	  rv = (smm->session_rx_fns[s0->session_type]) (vm, node, smm, e0, s0,
							my_thread_index,
							&n_tx_packets);
	  if (rv < 0)
	    goto done;

	  break;

	default:
	  clib_warning ("unhandled event type %d", e0->event_type);
	}
    }

done:

  /* Couldn't process all events. Probably out of buffers */
  if (PREDICT_FALSE (i < n_to_dequeue))
    {
      session_fifo_event_t *partially_read =
	smm->evts_partially_read[my_thread_index];
      vec_add (partially_read, &my_fifo_events[i], n_to_dequeue - i);
      vec_free (my_fifo_events);
      smm->fifo_events[my_thread_index] = partially_read;
      smm->evts_partially_read[my_thread_index] = 0;
    }
  else
    {
      vec_free (smm->fifo_events[my_thread_index]);
      smm->fifo_events[my_thread_index] =
	smm->evts_partially_read[my_thread_index];
      smm->evts_partially_read[my_thread_index] = 0;
    }

  vlib_node_increment_counter (vm, session_queue_node.index,
			       SESSION_QUEUE_ERROR_TX, n_tx_packets);

  return n_tx_packets;
}

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (session_queue_node) =
{
  .function = session_queue_node_fn,
  .name = "session-queue",
  .format_trace = format_session_queue_trace,
  .type = VLIB_NODE_TYPE_INPUT,
  .n_errors = ARRAY_LEN (session_queue_error_strings),
  .error_strings = session_queue_error_strings,
  .n_next_nodes = SESSION_QUEUE_N_NEXT,
  /* .state = VLIB_NODE_STATE_DISABLED, enable on-demand? */
  /* edit / add dispositions here */
  .next_nodes =
  {
      [SESSION_QUEUE_NEXT_DROP] = "error-drop",
      [SESSION_QUEUE_NEXT_IP4_LOOKUP] = "ip4-lookup",
      [SESSION_QUEUE_NEXT_IP6_LOOKUP] = "ip6-lookup",
      [SESSION_QUEUE_NEXT_TCP_IP4_OUTPUT] = "tcp4-output",
      [SESSION_QUEUE_NEXT_TCP_IP6_OUTPUT] = "tcp6-output",
  },
};
/* *INDENT-ON* */

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