1/* 2 * Copyright (c) 2004, 2005 Topspin Communications. All rights reserved. 3 * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. 4 * Copyright (c) 2004 Voltaire, Inc. All rights reserved. 5 * 6 * This software is available to you under a choice of one of two 7 * licenses. You may choose to be licensed under the terms of the GNU 8 * General Public License (GPL) Version 2, available from the file 9 * COPYING in the main directory of this source tree, or the 10 * OpenIB.org BSD license below: 11 * 12 * Redistribution and use in source and binary forms, with or 13 * without modification, are permitted provided that the following 14 * conditions are met: 15 * 16 * - Redistributions of source code must retain the above 17 * copyright notice, this list of conditions and the following 18 * disclaimer. 19 * 20 * - Redistributions in binary form must reproduce the above 21 * copyright notice, this list of conditions and the following 22 * disclaimer in the documentation and/or other materials 23 * provided with the distribution. 24 * 25 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 26 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 27 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 28 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 29 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 30 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 31 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 32 * SOFTWARE. 33 */ 34 35#include <linux/skbuff.h> 36#include <linux/rtnetlink.h> 37#include <linux/moduleparam.h> 38#include <linux/ip.h> 39#include <linux/in.h> 40#include <linux/igmp.h> 41#include <linux/inetdevice.h> 42#include <linux/delay.h> 43#include <linux/completion.h> 44#include <linux/slab.h> 45 46#include <net/dst.h> 47 48#include "ipoib.h" 49 50#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG 51static int mcast_debug_level; 52 53module_param(mcast_debug_level, int, 0644); 54MODULE_PARM_DESC(mcast_debug_level, 55 "Enable multicast debug tracing if > 0"); 56#endif 57 58struct ipoib_mcast_iter { 59 struct net_device *dev; 60 union ib_gid mgid; 61 unsigned long created; 62 unsigned int queuelen; 63 unsigned int complete; 64 unsigned int send_only; 65}; 66 67/* join state that allows creating mcg with sendonly member request */ 68#define SENDONLY_FULLMEMBER_JOIN 8 69 70/* 71 * This should be called with the priv->lock held 72 */ 73static void __ipoib_mcast_schedule_join_thread(struct ipoib_dev_priv *priv, 74 struct ipoib_mcast *mcast, 75 bool delay) 76{ 77 if (!test_bit(IPOIB_FLAG_OPER_UP, &priv->flags)) 78 return; 79 80 /* 81 * We will be scheduling *something*, so cancel whatever is 82 * currently scheduled first 83 */ 84 cancel_delayed_work(&priv->mcast_task); 85 if (mcast && delay) { 86 /* 87 * We had a failure and want to schedule a retry later 88 */ 89 mcast->backoff *= 2; 90 if (mcast->backoff > IPOIB_MAX_BACKOFF_SECONDS) 91 mcast->backoff = IPOIB_MAX_BACKOFF_SECONDS; 92 mcast->delay_until = jiffies + (mcast->backoff * HZ); 93 /* 94 * Mark this mcast for its delay, but restart the 95 * task immediately. The join task will make sure to 96 * clear out all entries without delays, and then 97 * schedule itself to run again when the earliest 98 * delay expires 99 */ 100 queue_delayed_work(priv->wq, &priv->mcast_task, 0); 101 } else if (delay) { 102 /* 103 * Special case of retrying after a failure to 104 * allocate the broadcast multicast group, wait 105 * 1 second and try again 106 */ 107 queue_delayed_work(priv->wq, &priv->mcast_task, HZ); 108 } else 109 queue_delayed_work(priv->wq, &priv->mcast_task, 0); 110} 111 112static void ipoib_mcast_free(struct ipoib_mcast *mcast) 113{ 114 struct net_device *dev = mcast->dev; 115 int tx_dropped = 0; 116 117 ipoib_dbg_mcast(ipoib_priv(dev), "deleting multicast group %pI6\n", 118 mcast->mcmember.mgid.raw); 119 120 /* remove all neigh connected to this mcast */ 121 ipoib_del_neighs_by_gid(dev, mcast->mcmember.mgid.raw); 122 123 if (mcast->ah) 124 ipoib_put_ah(mcast->ah); 125 126 while (!skb_queue_empty(&mcast->pkt_queue)) { 127 ++tx_dropped; 128 dev_kfree_skb_any(skb_dequeue(&mcast->pkt_queue)); 129 } 130 131 netif_tx_lock_bh(dev); 132 dev->stats.tx_dropped += tx_dropped; 133 netif_tx_unlock_bh(dev); 134 135 kfree(mcast); 136} 137 138static struct ipoib_mcast *ipoib_mcast_alloc(struct net_device *dev) 139{ 140 struct ipoib_mcast *mcast; 141 142 mcast = kzalloc(sizeof(*mcast), GFP_ATOMIC); 143 if (!mcast) 144 return NULL; 145 146 mcast->dev = dev; 147 mcast->created = jiffies; 148 mcast->delay_until = jiffies; 149 mcast->backoff = 1; 150 151 INIT_LIST_HEAD(&mcast->list); 152 INIT_LIST_HEAD(&mcast->neigh_list); 153 skb_queue_head_init(&mcast->pkt_queue); 154 155 return mcast; 156} 157 158static struct ipoib_mcast *__ipoib_mcast_find(struct net_device *dev, void *mgid) 159{ 160 struct ipoib_dev_priv *priv = ipoib_priv(dev); 161 struct rb_node *n = priv->multicast_tree.rb_node; 162 163 while (n) { 164 struct ipoib_mcast *mcast; 165 int ret; 166 167 mcast = rb_entry(n, struct ipoib_mcast, rb_node); 168 169 ret = memcmp(mgid, mcast->mcmember.mgid.raw, 170 sizeof (union ib_gid)); 171 if (ret < 0) 172 n = n->rb_left; 173 else if (ret > 0) 174 n = n->rb_right; 175 else 176 return mcast; 177 } 178 179 return NULL; 180} 181 182static int __ipoib_mcast_add(struct net_device *dev, struct ipoib_mcast *mcast) 183{ 184 struct ipoib_dev_priv *priv = ipoib_priv(dev); 185 struct rb_node **n = &priv->multicast_tree.rb_node, *pn = NULL; 186 187 while (*n) { 188 struct ipoib_mcast *tmcast; 189 int ret; 190 191 pn = *n; 192 tmcast = rb_entry(pn, struct ipoib_mcast, rb_node); 193 194 ret = memcmp(mcast->mcmember.mgid.raw, tmcast->mcmember.mgid.raw, 195 sizeof (union ib_gid)); 196 if (ret < 0) 197 n = &pn->rb_left; 198 else if (ret > 0) 199 n = &pn->rb_right; 200 else 201 return -EEXIST; 202 } 203 204 rb_link_node(&mcast->rb_node, pn, n); 205 rb_insert_color(&mcast->rb_node, &priv->multicast_tree); 206 207 return 0; 208} 209 210static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast, 211 struct ib_sa_mcmember_rec *mcmember) 212{ 213 struct net_device *dev = mcast->dev; 214 struct ipoib_dev_priv *priv = ipoib_priv(dev); 215 struct rdma_netdev *rn = netdev_priv(dev); 216 struct ipoib_ah *ah; 217 struct rdma_ah_attr av; 218 int ret; 219 int set_qkey = 0; 220 int mtu; 221 222 mcast->mcmember = *mcmember; 223 224 /* Set the multicast MTU and cached Q_Key before we attach if it's 225 * the broadcast group. 226 */ 227 if (!memcmp(mcast->mcmember.mgid.raw, priv->dev->broadcast + 4, 228 sizeof (union ib_gid))) { 229 spin_lock_irq(&priv->lock); 230 if (!priv->broadcast) { 231 spin_unlock_irq(&priv->lock); 232 return -EAGAIN; 233 } 234 /*update priv member according to the new mcast*/ 235 priv->broadcast->mcmember.qkey = mcmember->qkey; 236 priv->broadcast->mcmember.mtu = mcmember->mtu; 237 priv->broadcast->mcmember.traffic_class = mcmember->traffic_class; 238 priv->broadcast->mcmember.rate = mcmember->rate; 239 priv->broadcast->mcmember.sl = mcmember->sl; 240 priv->broadcast->mcmember.flow_label = mcmember->flow_label; 241 priv->broadcast->mcmember.hop_limit = mcmember->hop_limit; 242 /* assume if the admin and the mcast are the same both can be changed */ 243 mtu = rdma_mtu_enum_to_int(priv->ca, priv->port, 244 priv->broadcast->mcmember.mtu); 245 if (priv->mcast_mtu == priv->admin_mtu) 246 priv->admin_mtu = IPOIB_UD_MTU(mtu); 247 priv->mcast_mtu = IPOIB_UD_MTU(mtu); 248 rn->mtu = priv->mcast_mtu; 249 250 priv->qkey = be32_to_cpu(priv->broadcast->mcmember.qkey); 251 spin_unlock_irq(&priv->lock); 252 priv->tx_wr.remote_qkey = priv->qkey; 253 set_qkey = 1; 254 } 255 256 if (!test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) { 257 if (test_and_set_bit(IPOIB_MCAST_FLAG_ATTACHED, &mcast->flags)) { 258 ipoib_warn(priv, "multicast group %pI6 already attached\n", 259 mcast->mcmember.mgid.raw); 260 261 return 0; 262 } 263 264 ret = rn->attach_mcast(dev, priv->ca, &mcast->mcmember.mgid, 265 be16_to_cpu(mcast->mcmember.mlid), 266 set_qkey, priv->qkey); 267 if (ret < 0) { 268 ipoib_warn(priv, "couldn't attach QP to multicast group %pI6\n", 269 mcast->mcmember.mgid.raw); 270 271 clear_bit(IPOIB_MCAST_FLAG_ATTACHED, &mcast->flags); 272 return ret; 273 } 274 } 275 276 memset(&av, 0, sizeof(av)); 277 av.type = rdma_ah_find_type(priv->ca, priv->port); 278 rdma_ah_set_dlid(&av, be16_to_cpu(mcast->mcmember.mlid)); 279 rdma_ah_set_port_num(&av, priv->port); 280 rdma_ah_set_sl(&av, mcast->mcmember.sl); 281 rdma_ah_set_static_rate(&av, mcast->mcmember.rate); 282 283 rdma_ah_set_grh(&av, &mcast->mcmember.mgid, 284 be32_to_cpu(mcast->mcmember.flow_label), 285 0, mcast->mcmember.hop_limit, 286 mcast->mcmember.traffic_class); 287 288 ah = ipoib_create_ah(dev, priv->pd, &av); 289 if (IS_ERR(ah)) { 290 ipoib_warn(priv, "ib_address_create failed %ld\n", 291 -PTR_ERR(ah)); 292 /* use original error */ 293 return PTR_ERR(ah); 294 } 295 spin_lock_irq(&priv->lock); 296 mcast->ah = ah; 297 spin_unlock_irq(&priv->lock); 298 299 ipoib_dbg_mcast(priv, "MGID %pI6 AV %p, LID 0x%04x, SL %d\n", 300 mcast->mcmember.mgid.raw, 301 mcast->ah->ah, 302 be16_to_cpu(mcast->mcmember.mlid), 303 mcast->mcmember.sl); 304 305 /* actually send any queued packets */ 306 netif_tx_lock_bh(dev); 307 while (!skb_queue_empty(&mcast->pkt_queue)) { 308 struct sk_buff *skb = skb_dequeue(&mcast->pkt_queue); 309 310 netif_tx_unlock_bh(dev); 311 312 skb->dev = dev; 313 314 ret = dev_queue_xmit(skb); 315 if (ret) 316 ipoib_warn(priv, "%s:dev_queue_xmit failed to re-queue packet, ret:%d\n", 317 __func__, ret); 318 netif_tx_lock_bh(dev); 319 } 320 netif_tx_unlock_bh(dev); 321 322 return 0; 323} 324 325void ipoib_mcast_carrier_on_task(struct work_struct *work) 326{ 327 struct ipoib_dev_priv *priv = container_of(work, struct ipoib_dev_priv, 328 carrier_on_task); 329 struct ib_port_attr attr; 330 331 if (ib_query_port(priv->ca, priv->port, &attr) || 332 attr.state != IB_PORT_ACTIVE) { 333 ipoib_dbg(priv, "Keeping carrier off until IB port is active\n"); 334 return; 335 } 336 /* 337 * Take rtnl_lock to avoid racing with ipoib_stop() and 338 * turning the carrier back on while a device is being 339 * removed. However, ipoib_stop() will attempt to flush 340 * the workqueue while holding the rtnl lock, so loop 341 * on trylock until either we get the lock or we see 342 * FLAG_OPER_UP go away as that signals that we are bailing 343 * and can safely ignore the carrier on work. 344 */ 345 while (!rtnl_trylock()) { 346 if (!test_bit(IPOIB_FLAG_OPER_UP, &priv->flags)) 347 return; 348 else 349 msleep(20); 350 } 351 if (!ipoib_cm_admin_enabled(priv->dev)) 352 dev_set_mtu(priv->dev, min(priv->mcast_mtu, priv->admin_mtu)); 353 netif_carrier_on(priv->dev); 354 rtnl_unlock(); 355} 356 357static int ipoib_mcast_join_complete(int status, 358 struct ib_sa_multicast *multicast) 359{ 360 struct ipoib_mcast *mcast = multicast->context; 361 struct net_device *dev = mcast->dev; 362 struct ipoib_dev_priv *priv = ipoib_priv(dev); 363 364 ipoib_dbg_mcast(priv, "%sjoin completion for %pI6 (status %d)\n", 365 test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags) ? 366 "sendonly " : "", 367 mcast->mcmember.mgid.raw, status); 368 369 /* We trap for port events ourselves. */ 370 if (status == -ENETRESET) { 371 status = 0; 372 goto out; 373 } 374 375 if (!status) 376 status = ipoib_mcast_join_finish(mcast, &multicast->rec); 377 378 if (!status) { 379 mcast->backoff = 1; 380 mcast->delay_until = jiffies; 381 382 /* 383 * Defer carrier on work to priv->wq to avoid a 384 * deadlock on rtnl_lock here. Requeue our multicast 385 * work too, which will end up happening right after 386 * our carrier on task work and will allow us to 387 * send out all of the non-broadcast joins 388 */ 389 if (mcast == priv->broadcast) { 390 spin_lock_irq(&priv->lock); 391 queue_work(priv->wq, &priv->carrier_on_task); 392 __ipoib_mcast_schedule_join_thread(priv, NULL, 0); 393 goto out_locked; 394 } 395 } else { 396 bool silent_fail = 397 test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags) && 398 status == -EINVAL; 399 400 if (mcast->logcount < 20) { 401 if (status == -ETIMEDOUT || status == -EAGAIN || 402 silent_fail) { 403 ipoib_dbg_mcast(priv, "%smulticast join failed for %pI6, status %d\n", 404 test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags) ? "sendonly " : "", 405 mcast->mcmember.mgid.raw, status); 406 } else { 407 ipoib_warn(priv, "%smulticast join failed for %pI6, status %d\n", 408 test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags) ? "sendonly " : "", 409 mcast->mcmember.mgid.raw, status); 410 } 411 412 if (!silent_fail) 413 mcast->logcount++; 414 } 415 416 if (test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags) && 417 mcast->backoff >= 2) { 418 /* 419 * We only retry sendonly joins once before we drop 420 * the packet and quit trying to deal with the 421 * group. However, we leave the group in the 422 * mcast list as an unjoined group. If we want to 423 * try joining again, we simply queue up a packet 424 * and restart the join thread. The empty queue 425 * is why the join thread ignores this group. 426 */ 427 mcast->backoff = 1; 428 netif_tx_lock_bh(dev); 429 while (!skb_queue_empty(&mcast->pkt_queue)) { 430 ++dev->stats.tx_dropped; 431 dev_kfree_skb_any(skb_dequeue(&mcast->pkt_queue)); 432 } 433 netif_tx_unlock_bh(dev); 434 } else { 435 spin_lock_irq(&priv->lock); 436 /* Requeue this join task with a backoff delay */ 437 __ipoib_mcast_schedule_join_thread(priv, mcast, 1); 438 goto out_locked; 439 } 440 } 441out: 442 spin_lock_irq(&priv->lock); 443out_locked: 444 /* 445 * Make sure to set mcast->mc before we clear the busy flag to avoid 446 * racing with code that checks for BUSY before checking mcast->mc 447 */ 448 if (status) 449 mcast->mc = NULL; 450 else 451 mcast->mc = multicast; 452 clear_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags); 453 spin_unlock_irq(&priv->lock); 454 complete(&mcast->done); 455 456 return status; 457} 458 459/* 460 * Caller must hold 'priv->lock' 461 */ 462static int ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast) 463{ 464 struct ipoib_dev_priv *priv = ipoib_priv(dev); 465 struct ib_sa_multicast *multicast; 466 struct ib_sa_mcmember_rec rec = { 467 .join_state = 1 468 }; 469 ib_sa_comp_mask comp_mask; 470 int ret = 0; 471 472 if (!priv->broadcast || 473 !test_bit(IPOIB_FLAG_OPER_UP, &priv->flags)) 474 return -EINVAL; 475 476 init_completion(&mcast->done); 477 set_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags); 478 479 ipoib_dbg_mcast(priv, "joining MGID %pI6\n", mcast->mcmember.mgid.raw); 480 481 rec.mgid = mcast->mcmember.mgid; 482 rec.port_gid = priv->local_gid; 483 rec.pkey = cpu_to_be16(priv->pkey); 484 485 comp_mask = 486 IB_SA_MCMEMBER_REC_MGID | 487 IB_SA_MCMEMBER_REC_PORT_GID | 488 IB_SA_MCMEMBER_REC_PKEY | 489 IB_SA_MCMEMBER_REC_JOIN_STATE; 490 491 if (mcast != priv->broadcast) { 492 /* 493 * RFC 4391: 494 * The MGID MUST use the same P_Key, Q_Key, SL, MTU, 495 * and HopLimit as those used in the broadcast-GID. The rest 496 * of attributes SHOULD follow the values used in the 497 * broadcast-GID as well. 498 */ 499 comp_mask |= 500 IB_SA_MCMEMBER_REC_QKEY | 501 IB_SA_MCMEMBER_REC_MTU_SELECTOR | 502 IB_SA_MCMEMBER_REC_MTU | 503 IB_SA_MCMEMBER_REC_TRAFFIC_CLASS | 504 IB_SA_MCMEMBER_REC_RATE_SELECTOR | 505 IB_SA_MCMEMBER_REC_RATE | 506 IB_SA_MCMEMBER_REC_SL | 507 IB_SA_MCMEMBER_REC_FLOW_LABEL | 508 IB_SA_MCMEMBER_REC_HOP_LIMIT; 509 510 rec.qkey = priv->broadcast->mcmember.qkey; 511 rec.mtu_selector = IB_SA_EQ; 512 rec.mtu = priv->broadcast->mcmember.mtu; 513 rec.traffic_class = priv->broadcast->mcmember.traffic_class; 514 rec.rate_selector = IB_SA_EQ; 515 rec.rate = priv->broadcast->mcmember.rate; 516 rec.sl = priv->broadcast->mcmember.sl; 517 rec.flow_label = priv->broadcast->mcmember.flow_label; 518 rec.hop_limit = priv->broadcast->mcmember.hop_limit; 519 520 /* 521 * Send-only IB Multicast joins work at the core IB layer but 522 * require specific SM support. 523 * We can use such joins here only if the current SM supports that feature. 524 * However, if not, we emulate an Ethernet multicast send, 525 * which does not require a multicast subscription and will 526 * still send properly. The most appropriate thing to 527 * do is to create the group if it doesn't exist as that 528 * most closely emulates the behavior, from a user space 529 * application perspective, of Ethernet multicast operation. 530 */ 531 if (test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) 532 rec.join_state = SENDONLY_FULLMEMBER_JOIN; 533 } 534 535 multicast = ib_sa_join_multicast(&ipoib_sa_client, priv->ca, priv->port, 536 &rec, comp_mask, GFP_ATOMIC, 537 ipoib_mcast_join_complete, mcast); 538 if (IS_ERR(multicast)) { 539 ret = PTR_ERR(multicast); 540 ipoib_warn(priv, "ib_sa_join_multicast failed, status %d\n", ret); 541 /* Requeue this join task with a backoff delay */ 542 __ipoib_mcast_schedule_join_thread(priv, mcast, 1); 543 clear_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags); 544 complete(&mcast->done); 545 return ret; 546 } 547 return 0; 548} 549 550void ipoib_mcast_join_task(struct work_struct *work) 551{ 552 struct ipoib_dev_priv *priv = 553 container_of(work, struct ipoib_dev_priv, mcast_task.work); 554 struct net_device *dev = priv->dev; 555 struct ib_port_attr port_attr; 556 unsigned long delay_until = 0; 557 struct ipoib_mcast *mcast = NULL; 558 559 if (!test_bit(IPOIB_FLAG_OPER_UP, &priv->flags)) 560 return; 561 562 if (ib_query_port(priv->ca, priv->port, &port_attr)) { 563 ipoib_dbg(priv, "ib_query_port() failed\n"); 564 return; 565 } 566 if (port_attr.state != IB_PORT_ACTIVE) { 567 ipoib_dbg(priv, "port state is not ACTIVE (state = %d) suspending join task\n", 568 port_attr.state); 569 return; 570 } 571 priv->local_lid = port_attr.lid; 572 netif_addr_lock_bh(dev); 573 574 if (!test_bit(IPOIB_FLAG_DEV_ADDR_SET, &priv->flags)) { 575 netif_addr_unlock_bh(dev); 576 return; 577 } 578 netif_addr_unlock_bh(dev); 579 580 spin_lock_irq(&priv->lock); 581 if (!test_bit(IPOIB_FLAG_OPER_UP, &priv->flags)) 582 goto out; 583 584 if (!priv->broadcast) { 585 struct ipoib_mcast *broadcast; 586 587 broadcast = ipoib_mcast_alloc(dev); 588 if (!broadcast) { 589 ipoib_warn(priv, "failed to allocate broadcast group\n"); 590 /* 591 * Restart us after a 1 second delay to retry 592 * creating our broadcast group and attaching to 593 * it. Until this succeeds, this ipoib dev is 594 * completely stalled (multicast wise). 595 */ 596 __ipoib_mcast_schedule_join_thread(priv, NULL, 1); 597 goto out; 598 } 599 600 memcpy(broadcast->mcmember.mgid.raw, priv->dev->broadcast + 4, 601 sizeof (union ib_gid)); 602 priv->broadcast = broadcast; 603 604 __ipoib_mcast_add(dev, priv->broadcast); 605 } 606 607 if (!test_bit(IPOIB_MCAST_FLAG_ATTACHED, &priv->broadcast->flags)) { 608 if (IS_ERR_OR_NULL(priv->broadcast->mc) && 609 !test_bit(IPOIB_MCAST_FLAG_BUSY, &priv->broadcast->flags)) { 610 mcast = priv->broadcast; 611 if (mcast->backoff > 1 && 612 time_before(jiffies, mcast->delay_until)) { 613 delay_until = mcast->delay_until; 614 mcast = NULL; 615 } 616 } 617 goto out; 618 } 619 620 /* 621 * We'll never get here until the broadcast group is both allocated 622 * and attached 623 */ 624 list_for_each_entry(mcast, &priv->multicast_list, list) { 625 if (IS_ERR_OR_NULL(mcast->mc) && 626 !test_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags) && 627 (!test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags) || 628 !skb_queue_empty(&mcast->pkt_queue))) { 629 if (mcast->backoff == 1 || 630 time_after_eq(jiffies, mcast->delay_until)) { 631 /* Found the next unjoined group */ 632 if (ipoib_mcast_join(dev, mcast)) { 633 spin_unlock_irq(&priv->lock); 634 return; 635 } 636 } else if (!delay_until || 637 time_before(mcast->delay_until, delay_until)) 638 delay_until = mcast->delay_until; 639 } 640 } 641 642 mcast = NULL; 643 ipoib_dbg_mcast(priv, "successfully started all multicast joins\n"); 644 645out: 646 if (delay_until) { 647 cancel_delayed_work(&priv->mcast_task); 648 queue_delayed_work(priv->wq, &priv->mcast_task, 649 delay_until - jiffies); 650 } 651 if (mcast) 652 ipoib_mcast_join(dev, mcast); 653 654 spin_unlock_irq(&priv->lock); 655} 656 657void ipoib_mcast_start_thread(struct net_device *dev) 658{ 659 struct ipoib_dev_priv *priv = ipoib_priv(dev); 660 unsigned long flags; 661 662 ipoib_dbg_mcast(priv, "starting multicast thread\n"); 663 664 spin_lock_irqsave(&priv->lock, flags); 665 __ipoib_mcast_schedule_join_thread(priv, NULL, 0); 666 spin_unlock_irqrestore(&priv->lock, flags); 667} 668 669void ipoib_mcast_stop_thread(struct net_device *dev) 670{ 671 struct ipoib_dev_priv *priv = ipoib_priv(dev); 672 673 ipoib_dbg_mcast(priv, "stopping multicast thread\n"); 674 675 cancel_delayed_work_sync(&priv->mcast_task); 676} 677 678static int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast) 679{ 680 struct ipoib_dev_priv *priv = ipoib_priv(dev); 681 struct rdma_netdev *rn = netdev_priv(dev); 682 int ret = 0; 683 684 if (test_and_clear_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags)) 685 ipoib_warn(priv, "ipoib_mcast_leave on an in-flight join\n"); 686 687 if (!IS_ERR_OR_NULL(mcast->mc)) 688 ib_sa_free_multicast(mcast->mc); 689 690 if (test_and_clear_bit(IPOIB_MCAST_FLAG_ATTACHED, &mcast->flags)) { 691 ipoib_dbg_mcast(priv, "leaving MGID %pI6\n", 692 mcast->mcmember.mgid.raw); 693 694 /* Remove ourselves from the multicast group */ 695 ret = rn->detach_mcast(dev, priv->ca, &mcast->mcmember.mgid, 696 be16_to_cpu(mcast->mcmember.mlid)); 697 if (ret) 698 ipoib_warn(priv, "ib_detach_mcast failed (result = %d)\n", ret); 699 } else if (!test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) 700 ipoib_dbg(priv, "leaving with no mcmember but not a " 701 "SENDONLY join\n"); 702 703 return 0; 704} 705 706/* 707 * Check if the multicast group is sendonly. If so remove it from the maps 708 * and add to the remove list 709 */ 710void ipoib_check_and_add_mcast_sendonly(struct ipoib_dev_priv *priv, u8 *mgid, 711 struct list_head *remove_list) 712{ 713 /* Is this multicast ? */ 714 if (*mgid == 0xff) { 715 struct ipoib_mcast *mcast = __ipoib_mcast_find(priv->dev, mgid); 716 717 if (mcast && test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) { 718 list_del(&mcast->list); 719 rb_erase(&mcast->rb_node, &priv->multicast_tree); 720 list_add_tail(&mcast->list, remove_list); 721 } 722 } 723} 724 725void ipoib_mcast_remove_list(struct list_head *remove_list) 726{ 727 struct ipoib_mcast *mcast, *tmcast; 728 729 /* 730 * make sure the in-flight joins have finished before we attempt 731 * to leave 732 */ 733 list_for_each_entry_safe(mcast, tmcast, remove_list, list) 734 if (test_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags)) 735 wait_for_completion(&mcast->done); 736 737 list_for_each_entry_safe(mcast, tmcast, remove_list, list) { 738 ipoib_mcast_leave(mcast->dev, mcast); 739 ipoib_mcast_free(mcast); 740 } 741} 742 743void ipoib_mcast_send(struct net_device *dev, u8 *daddr, struct sk_buff *skb) 744{ 745 struct ipoib_dev_priv *priv = ipoib_priv(dev); 746 struct rdma_netdev *rn = netdev_priv(dev); 747 struct ipoib_mcast *mcast; 748 unsigned long flags; 749 void *mgid = daddr + 4; 750 751 spin_lock_irqsave(&priv->lock, flags); 752 753 if (!test_bit(IPOIB_FLAG_OPER_UP, &priv->flags) || 754 !priv->broadcast || 755 !test_bit(IPOIB_MCAST_FLAG_ATTACHED, &priv->broadcast->flags)) { 756 ++dev->stats.tx_dropped; 757 dev_kfree_skb_any(skb); 758 goto unlock; 759 } 760 761 mcast = __ipoib_mcast_find(dev, mgid); 762 if (!mcast || !mcast->ah) { 763 if (!mcast) { 764 /* Let's create a new send only group now */ 765 ipoib_dbg_mcast(priv, "setting up send only multicast group for %pI6\n", 766 mgid); 767 768 mcast = ipoib_mcast_alloc(dev); 769 if (!mcast) { 770 ipoib_warn(priv, "unable to allocate memory " 771 "for multicast structure\n"); 772 ++dev->stats.tx_dropped; 773 dev_kfree_skb_any(skb); 774 goto unlock; 775 } 776 777 set_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags); 778 memcpy(mcast->mcmember.mgid.raw, mgid, 779 sizeof (union ib_gid)); 780 __ipoib_mcast_add(dev, mcast); 781 list_add_tail(&mcast->list, &priv->multicast_list); 782 } 783 if (skb_queue_len(&mcast->pkt_queue) < IPOIB_MAX_MCAST_QUEUE) { 784 /* put pseudoheader back on for next time */ 785 skb_push(skb, sizeof(struct ipoib_pseudo_header)); 786 skb_queue_tail(&mcast->pkt_queue, skb); 787 } else { 788 ++dev->stats.tx_dropped; 789 dev_kfree_skb_any(skb); 790 } 791 if (!test_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags)) { 792 __ipoib_mcast_schedule_join_thread(priv, NULL, 0); 793 } 794 } else { 795 struct ipoib_neigh *neigh; 796 797 spin_unlock_irqrestore(&priv->lock, flags); 798 neigh = ipoib_neigh_get(dev, daddr); 799 spin_lock_irqsave(&priv->lock, flags); 800 if (!neigh) { 801 neigh = ipoib_neigh_alloc(daddr, dev); 802 /* Make sure that the neigh will be added only 803 * once to mcast list. 804 */ 805 if (neigh && list_empty(&neigh->list)) { 806 kref_get(&mcast->ah->ref); 807 neigh->ah = mcast->ah; 808 neigh->ah->valid = 1; 809 list_add_tail(&neigh->list, &mcast->neigh_list); 810 } 811 } 812 spin_unlock_irqrestore(&priv->lock, flags); 813 mcast->ah->last_send = rn->send(dev, skb, mcast->ah->ah, 814 IB_MULTICAST_QPN); 815 if (neigh) 816 ipoib_neigh_put(neigh); 817 return; 818 } 819 820unlock: 821 spin_unlock_irqrestore(&priv->lock, flags); 822} 823 824void ipoib_mcast_dev_flush(struct net_device *dev) 825{ 826 struct ipoib_dev_priv *priv = ipoib_priv(dev); 827 LIST_HEAD(remove_list); 828 struct ipoib_mcast *mcast, *tmcast; 829 unsigned long flags; 830 831 mutex_lock(&priv->mcast_mutex); 832 ipoib_dbg_mcast(priv, "flushing multicast list\n"); 833 834 spin_lock_irqsave(&priv->lock, flags); 835 836 list_for_each_entry_safe(mcast, tmcast, &priv->multicast_list, list) { 837 list_del(&mcast->list); 838 rb_erase(&mcast->rb_node, &priv->multicast_tree); 839 list_add_tail(&mcast->list, &remove_list); 840 } 841 842 if (priv->broadcast) { 843 rb_erase(&priv->broadcast->rb_node, &priv->multicast_tree); 844 list_add_tail(&priv->broadcast->list, &remove_list); 845 priv->broadcast = NULL; 846 } 847 848 spin_unlock_irqrestore(&priv->lock, flags); 849 850 ipoib_mcast_remove_list(&remove_list); 851 mutex_unlock(&priv->mcast_mutex); 852} 853 854static int ipoib_mcast_addr_is_valid(const u8 *addr, const u8 *broadcast) 855{ 856 /* reserved QPN, prefix, scope */ 857 if (memcmp(addr, broadcast, 6)) 858 return 0; 859 /* signature lower, pkey */ 860 if (memcmp(addr + 7, broadcast + 7, 3)) 861 return 0; 862 return 1; 863} 864 865void ipoib_mcast_restart_task(struct work_struct *work) 866{ 867 struct ipoib_dev_priv *priv = 868 container_of(work, struct ipoib_dev_priv, restart_task); 869 struct net_device *dev = priv->dev; 870 struct netdev_hw_addr *ha; 871 struct ipoib_mcast *mcast, *tmcast; 872 LIST_HEAD(remove_list); 873 struct ib_sa_mcmember_rec rec; 874 875 if (!test_bit(IPOIB_FLAG_OPER_UP, &priv->flags)) 876 /* 877 * shortcut...on shutdown flush is called next, just 878 * let it do all the work 879 */ 880 return; 881 882 ipoib_dbg_mcast(priv, "restarting multicast task\n"); 883 884 netif_addr_lock_bh(dev); 885 spin_lock_irq(&priv->lock); 886 887 /* 888 * Unfortunately, the networking core only gives us a list of all of 889 * the multicast hardware addresses. We need to figure out which ones 890 * are new and which ones have been removed 891 */ 892 893 /* Clear out the found flag */ 894 list_for_each_entry(mcast, &priv->multicast_list, list) 895 clear_bit(IPOIB_MCAST_FLAG_FOUND, &mcast->flags); 896 897 /* Mark all of the entries that are found or don't exist */ 898 netdev_for_each_mc_addr(ha, dev) { 899 union ib_gid mgid; 900 901 if (!ipoib_mcast_addr_is_valid(ha->addr, dev->broadcast)) 902 continue; 903 904 memcpy(mgid.raw, ha->addr + 4, sizeof(mgid)); 905 906 mcast = __ipoib_mcast_find(dev, &mgid); 907 if (!mcast || test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) { 908 struct ipoib_mcast *nmcast; 909 910 /* ignore group which is directly joined by userspace */ 911 if (test_bit(IPOIB_FLAG_UMCAST, &priv->flags) && 912 !ib_sa_get_mcmember_rec(priv->ca, priv->port, &mgid, &rec)) { 913 ipoib_dbg_mcast(priv, "ignoring multicast entry for mgid %pI6\n", 914 mgid.raw); 915 continue; 916 } 917 918 /* Not found or send-only group, let's add a new entry */ 919 ipoib_dbg_mcast(priv, "adding multicast entry for mgid %pI6\n", 920 mgid.raw); 921 922 nmcast = ipoib_mcast_alloc(dev); 923 if (!nmcast) { 924 ipoib_warn(priv, "unable to allocate memory for multicast structure\n"); 925 continue; 926 } 927 928 set_bit(IPOIB_MCAST_FLAG_FOUND, &nmcast->flags); 929 930 nmcast->mcmember.mgid = mgid; 931 932 if (mcast) { 933 /* Destroy the send only entry */ 934 list_move_tail(&mcast->list, &remove_list); 935 936 rb_replace_node(&mcast->rb_node, 937 &nmcast->rb_node, 938 &priv->multicast_tree); 939 } else 940 __ipoib_mcast_add(dev, nmcast); 941 942 list_add_tail(&nmcast->list, &priv->multicast_list); 943 } 944 945 if (mcast) 946 set_bit(IPOIB_MCAST_FLAG_FOUND, &mcast->flags); 947 } 948 949 /* Remove all of the entries don't exist anymore */ 950 list_for_each_entry_safe(mcast, tmcast, &priv->multicast_list, list) { 951 if (!test_bit(IPOIB_MCAST_FLAG_FOUND, &mcast->flags) && 952 !test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) { 953 ipoib_dbg_mcast(priv, "deleting multicast group %pI6\n", 954 mcast->mcmember.mgid.raw); 955 956 rb_erase(&mcast->rb_node, &priv->multicast_tree); 957 958 /* Move to the remove list */ 959 list_move_tail(&mcast->list, &remove_list); 960 } 961 } 962 963 spin_unlock_irq(&priv->lock); 964 netif_addr_unlock_bh(dev); 965 966 ipoib_mcast_remove_list(&remove_list); 967 968 /* 969 * Double check that we are still up 970 */ 971 if (test_bit(IPOIB_FLAG_OPER_UP, &priv->flags)) { 972 spin_lock_irq(&priv->lock); 973 __ipoib_mcast_schedule_join_thread(priv, NULL, 0); 974 spin_unlock_irq(&priv->lock); 975 } 976} 977 978#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG 979 980struct ipoib_mcast_iter *ipoib_mcast_iter_init(struct net_device *dev) 981{ 982 struct ipoib_mcast_iter *iter; 983 984 iter = kmalloc(sizeof(*iter), GFP_KERNEL); 985 if (!iter) 986 return NULL; 987 988 iter->dev = dev; 989 memset(iter->mgid.raw, 0, 16); 990 991 if (ipoib_mcast_iter_next(iter)) { 992 kfree(iter); 993 return NULL; 994 } 995 996 return iter; 997} 998 999int ipoib_mcast_iter_next(struct ipoib_mcast_iter *iter) 1000{ 1001 struct ipoib_dev_priv *priv = ipoib_priv(iter->dev); 1002 struct rb_node *n; 1003 struct ipoib_mcast *mcast; 1004 int ret = 1; 1005 1006 spin_lock_irq(&priv->lock); 1007 1008 n = rb_first(&priv->multicast_tree); 1009 1010 while (n) { 1011 mcast = rb_entry(n, struct ipoib_mcast, rb_node); 1012 1013 if (memcmp(iter->mgid.raw, mcast->mcmember.mgid.raw, 1014 sizeof (union ib_gid)) < 0) { 1015 iter->mgid = mcast->mcmember.mgid; 1016 iter->created = mcast->created; 1017 iter->queuelen = skb_queue_len(&mcast->pkt_queue); 1018 iter->complete = !!mcast->ah; 1019 iter->send_only = !!(mcast->flags & (1 << IPOIB_MCAST_FLAG_SENDONLY)); 1020 1021 ret = 0; 1022 1023 break; 1024 } 1025 1026 n = rb_next(n); 1027 } 1028 1029 spin_unlock_irq(&priv->lock); 1030 1031 return ret; 1032} 1033 1034void ipoib_mcast_iter_read(struct ipoib_mcast_iter *iter, 1035 union ib_gid *mgid, 1036 unsigned long *created, 1037 unsigned int *queuelen, 1038 unsigned int *complete, 1039 unsigned int *send_only) 1040{ 1041 *mgid = iter->mgid; 1042 *created = iter->created; 1043 *queuelen = iter->queuelen; 1044 *complete = iter->complete; 1045 *send_only = iter->send_only; 1046} 1047 1048#endif /* CONFIG_INFINIBAND_IPOIB_DEBUG */ 1049