drivers/infiniband/core/multicast.c

   1 /*
   2  * Copyright (c) 2006 Intel Corporation.  All rights reserved.
   3  *
   4  * This software is available to you under a choice of one of two
   5  * licenses.  You may choose to be licensed under the terms of the GNU
   6  * General Public License (GPL) Version 2, available from the file
   7  * COPYING in the main directory of this source tree, or the
   8  * OpenIB.org BSD license below:
   9  *
  10  *     Redistribution and use in source and binary forms, with or
  11  *     without modification, are permitted provided that the following
  12  *     conditions are met:
  13  *
  14  *      - Redistributions of source code must retain the above
  15  *        copyright notice, this list of conditions and the following
  16  *        disclaimer.
  17  *
  18  *      - Redistributions in binary form must reproduce the above
  19  *        copyright notice, this list of conditions and the following
  20  *        disclaimer in the documentation and/or other materials
  21  *        provided with the distribution.
  22  *
  23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  30  * SOFTWARE.
  31  */
  32
  33 #include <linux/completion.h>
  34 #include <linux/dma-mapping.h>
  35 #include <linux/err.h>
  36 #include <linux/interrupt.h>
  37 #include <linux/export.h>
  38 #include <linux/slab.h>
  39 #include <linux/bitops.h>
  40 #include <linux/random.h>
  41
  42 #include <rdma/ib_cache.h>
  43 #include "sa.h"
  44
  45 static void mcast_add_one(struct ib_device *device);
  46 static void mcast_remove_one(struct ib_device *device, void *client_data);
  47
  48 static struct ib_client mcast_client = {
  49         .name   = "ib_multicast",
  50         .add    = mcast_add_one,
  51         .remove = mcast_remove_one
  52 };
  53
  54 static struct ib_sa_client      sa_client;
  55 static struct workqueue_struct  *mcast_wq;
  56 static union ib_gid mgid0;
  57
  58 struct mcast_device;
  59
  60 struct mcast_port {
  61         struct mcast_device     *dev;
  62         spinlock_t              lock;
  63         struct rb_root          table;
  64         atomic_t                refcount;
  65         struct completion       comp;
  66         u8                      port_num;
  67 };
  68
  69 struct mcast_device {
  70         struct ib_device        *device;
  71         struct ib_event_handler event_handler;
  72         int                     start_port;
  73         int                     end_port;
  74         struct mcast_port       port[0];
  75 };
  76
  77 enum mcast_state {
  78         MCAST_JOINING,
  79         MCAST_MEMBER,
  80         MCAST_ERROR,
  81 };
  82
  83 enum mcast_group_state {
  84         MCAST_IDLE,
  85         MCAST_BUSY,
  86         MCAST_GROUP_ERROR,
  87         MCAST_PKEY_EVENT
  88 };
  89
  90 enum {
  91         MCAST_INVALID_PKEY_INDEX = 0xFFFF
  92 };
  93
  94 struct mcast_member;
  95
  96 struct mcast_group {
  97         struct ib_sa_mcmember_rec rec;
  98         struct rb_node          node;
  99         struct mcast_port       *port;
 100         spinlock_t              lock;
 101         struct work_struct      work;
 102         struct list_head        pending_list;
 103         struct list_head        active_list;
 104         struct mcast_member     *last_join;
 105         int                     members[NUM_JOIN_MEMBERSHIP_TYPES];
 106         atomic_t                refcount;
 107         enum mcast_group_state  state;
 108         struct ib_sa_query      *query;
 109         u16                     pkey_index;
 110         u8                      leave_state;
 111         int                     retries;
 112 };
 113
 114 struct mcast_member {
 115         struct ib_sa_multicast  multicast;
 116         struct ib_sa_client     *client;
 117         struct mcast_group      *group;
 118         struct list_head        list;
 119         enum mcast_state        state;
 120         atomic_t                refcount;
 121         struct completion       comp;
 122 };
 123
 124 static void join_handler(int status, struct ib_sa_mcmember_rec *rec,
 125                          void *context);
 126 static void leave_handler(int status, struct ib_sa_mcmember_rec *rec,
 127                           void *context);
 128
 129 static struct mcast_group *mcast_find(struct mcast_port *port,
 130                                       union ib_gid *mgid)
 131 {
 132         struct rb_node *node = port->table.rb_node;
 133         struct mcast_group *group;
 134         int ret;
 135
 136         while (node) {
 137                 group = rb_entry(node, struct mcast_group, node);
 138                 ret = memcmp(mgid->raw, group->rec.mgid.raw, sizeof *mgid);
 139                 if (!ret)
 140                         return group;
 141
 142                 if (ret < 0)
 143                         node = node->rb_left;
 144                 else
 145                         node = node->rb_right;
 146         }
 147         return NULL;
 148 }
 149
 150 static struct mcast_group *mcast_insert(struct mcast_port *port,
 151                                         struct mcast_group *group,
 152                                         int allow_duplicates)
 153 {
 154         struct rb_node **link = &port->table.rb_node;
 155         struct rb_node *parent = NULL;
 156         struct mcast_group *cur_group;
 157         int ret;
 158
 159         while (*link) {
 160                 parent = *link;
 161                 cur_group = rb_entry(parent, struct mcast_group, node);
 162
 163                 ret = memcmp(group->rec.mgid.raw, cur_group->rec.mgid.raw,
 164                              sizeof group->rec.mgid);
 165                 if (ret < 0)
 166                         link = &(*link)->rb_left;
 167                 else if (ret > 0)
 168                         link = &(*link)->rb_right;
 169                 else if (allow_duplicates)
 170                         link = &(*link)->rb_left;
 171                 else
 172                         return cur_group;
 173         }
 174         rb_link_node(&group->node, parent, link);
 175         rb_insert_color(&group->node, &port->table);
 176         return NULL;
 177 }
 178
 179 static void deref_port(struct mcast_port *port)
 180 {
 181         if (atomic_dec_and_test(&port->refcount))
 182                 complete(&port->comp);
 183 }
 184
 185 static void release_group(struct mcast_group *group)
 186 {
 187         struct mcast_port *port = group->port;
 188         unsigned long flags;
 189
 190         spin_lock_irqsave(&port->lock, flags);
 191         if (atomic_dec_and_test(&group->refcount)) {
 192                 rb_erase(&group->node, &port->table);
 193                 spin_unlock_irqrestore(&port->lock, flags);
 194                 kfree(group);
 195                 deref_port(port);
 196         } else
 197                 spin_unlock_irqrestore(&port->lock, flags);
 198 }
 199
 200 static void deref_member(struct mcast_member *member)
 201 {
 202         if (atomic_dec_and_test(&member->refcount))
 203                 complete(&member->comp);
 204 }
 205
 206 static void queue_join(struct mcast_member *member)
 207 {
 208         struct mcast_group *group = member->group;
 209         unsigned long flags;
 210
 211         spin_lock_irqsave(&group->lock, flags);
 212         list_add_tail(&member->list, &group->pending_list);
 213         if (group->state == MCAST_IDLE) {
 214                 group->state = MCAST_BUSY;
 215                 atomic_inc(&group->refcount);
 216                 queue_work(mcast_wq, &group->work);
 217         }
 218         spin_unlock_irqrestore(&group->lock, flags);
 219 }
 220
 221 /*
 222  * A multicast group has four types of members: full member, non member,
 223  * sendonly non member and sendonly full member.
 224  * We need to keep track of the number of members of each
 225  * type based on their join state.  Adjust the number of members the belong to
 226  * the specified join states.
 227  */
 228 static void adjust_membership(struct mcast_group *group, u8 join_state, int inc)
 229 {
 230         int i;
 231
 232         for (i = 0; i < NUM_JOIN_MEMBERSHIP_TYPES; i++, join_state >>= 1)
 233                 if (join_state & 0x1)
 234                         group->members[i] += inc;
 235 }
 236
 237 /*
 238  * If a multicast group has zero members left for a particular join state, but
 239  * the group is still a member with the SA, we need to leave that join state.
 240  * Determine which join states we still belong to, but that do not have any
 241  * active members.
 242  */
 243 static u8 get_leave_state(struct mcast_group *group)
 244 {
 245         u8 leave_state = 0;
 246         int i;
 247
 248         for (i = 0; i < NUM_JOIN_MEMBERSHIP_TYPES; i++)
 249                 if (!group->members[i])
 250                         leave_state |= (0x1 << i);
 251
 252         return leave_state & group->rec.join_state;
 253 }
 254
 255 static int check_selector(ib_sa_comp_mask comp_mask,
 256                           ib_sa_comp_mask selector_mask,
 257                           ib_sa_comp_mask value_mask,
 258                           u8 selector, u8 src_value, u8 dst_value)
 259 {
 260         int err;
 261
 262         if (!(comp_mask & selector_mask) || !(comp_mask & value_mask))
 263                 return 0;
 264
 265         switch (selector) {
 266         case IB_SA_GT:
 267                 err = (src_value <= dst_value);
 268                 break;
 269         case IB_SA_LT:
 270                 err = (src_value >= dst_value);
 271                 break;
 272         case IB_SA_EQ:
 273                 err = (src_value != dst_value);
 274                 break;
 275         default:
 276                 err = 0;
 277                 break;
 278         }
 279
 280         return err;
 281 }
 282
 283 static int cmp_rec(struct ib_sa_mcmember_rec *src,
 284                    struct ib_sa_mcmember_rec *dst, ib_sa_comp_mask comp_mask)
 285 {
 286         /* MGID must already match */
 287
 288         if (comp_mask & IB_SA_MCMEMBER_REC_PORT_GID &&
 289             memcmp(&src->port_gid, &dst->port_gid, sizeof src->port_gid))
 290                 return -EINVAL;
 291         if (comp_mask & IB_SA_MCMEMBER_REC_QKEY && src->qkey != dst->qkey)
 292                 return -EINVAL;
 293         if (comp_mask & IB_SA_MCMEMBER_REC_MLID && src->mlid != dst->mlid)
 294                 return -EINVAL;
 295         if (check_selector(comp_mask, IB_SA_MCMEMBER_REC_MTU_SELECTOR,
 296                            IB_SA_MCMEMBER_REC_MTU, dst->mtu_selector,
 297                            src->mtu, dst->mtu))
 298                 return -EINVAL;
 299         if (comp_mask & IB_SA_MCMEMBER_REC_TRAFFIC_CLASS &&
 300             src->traffic_class != dst->traffic_class)
 301                 return -EINVAL;
 302         if (comp_mask & IB_SA_MCMEMBER_REC_PKEY && src->pkey != dst->pkey)
 303                 return -EINVAL;
 304         if (check_selector(comp_mask, IB_SA_MCMEMBER_REC_RATE_SELECTOR,
 305                            IB_SA_MCMEMBER_REC_RATE, dst->rate_selector,
 306                            src->rate, dst->rate))
 307                 return -EINVAL;
 308         if (check_selector(comp_mask,
 309                            IB_SA_MCMEMBER_REC_PACKET_LIFE_TIME_SELECTOR,
 310                            IB_SA_MCMEMBER_REC_PACKET_LIFE_TIME,
 311                            dst->packet_life_time_selector,
 312                            src->packet_life_time, dst->packet_life_time))
 313                 return -EINVAL;
 314         if (comp_mask & IB_SA_MCMEMBER_REC_SL && src->sl != dst->sl)
 315                 return -EINVAL;
 316         if (comp_mask & IB_SA_MCMEMBER_REC_FLOW_LABEL &&
 317             src->flow_label != dst->flow_label)
 318                 return -EINVAL;
 319         if (comp_mask & IB_SA_MCMEMBER_REC_HOP_LIMIT &&
 320             src->hop_limit != dst->hop_limit)
 321                 return -EINVAL;
 322         if (comp_mask & IB_SA_MCMEMBER_REC_SCOPE && src->scope != dst->scope)
 323                 return -EINVAL;
 324
 325         /* join_state checked separately, proxy_join ignored */
 326
 327         return 0;
 328 }
 329
 330 static int send_join(struct mcast_group *group, struct mcast_member *member)
 331 {
 332         struct mcast_port *port = group->port;
 333         int ret;
 334
 335         group->last_join = member;
 336         ret = ib_sa_mcmember_rec_query(&sa_client, port->dev->device,
 337                                        port->port_num, IB_MGMT_METHOD_SET,
 338                                        &member->multicast.rec,
 339                                        member->multicast.comp_mask,
 340                                        3000, GFP_KERNEL, join_handler, group,
 341                                        &group->query);
 342         return (ret > 0) ? 0 : ret;
 343 }
 344
 345 static int send_leave(struct mcast_group *group, u8 leave_state)
 346 {
 347         struct mcast_port *port = group->port;
 348         struct ib_sa_mcmember_rec rec;
 349         int ret;
 350
 351         rec = group->rec;
 352         rec.join_state = leave_state;
 353         group->leave_state = leave_state;
 354
 355         ret = ib_sa_mcmember_rec_query(&sa_client, port->dev->device,
 356                                        port->port_num, IB_SA_METHOD_DELETE, &rec,
 357                                        IB_SA_MCMEMBER_REC_MGID     |
 358                                        IB_SA_MCMEMBER_REC_PORT_GID |
 359                                        IB_SA_MCMEMBER_REC_JOIN_STATE,
 360                                        3000, GFP_KERNEL, leave_handler,
 361                                        group, &group->query);
 362         return (ret > 0) ? 0 : ret;
 363 }
 364
 365 static void join_group(struct mcast_group *group, struct mcast_member *member,
 366                        u8 join_state)
 367 {
 368         member->state = MCAST_MEMBER;
 369         adjust_membership(group, join_state, 1);
 370         group->rec.join_state |= join_state;
 371         member->multicast.rec = group->rec;
 372         member->multicast.rec.join_state = join_state;
 373         list_move(&member->list, &group->active_list);
 374 }
 375
 376 static int fail_join(struct mcast_group *group, struct mcast_member *member,
 377                      int status)
 378 {
 379         spin_lock_irq(&group->lock);
 380         list_del_init(&member->list);
 381         spin_unlock_irq(&group->lock);
 382         return member->multicast.callback(status, &member->multicast);
 383 }
 384
 385 static void process_group_error(struct mcast_group *group)
 386 {
 387         struct mcast_member *member;
 388         int ret = 0;
 389         u16 pkey_index;
 390
 391         if (group->state == MCAST_PKEY_EVENT)
 392                 ret = ib_find_pkey(group->port->dev->device,
 393                                    group->port->port_num,
 394                                    be16_to_cpu(group->rec.pkey), &pkey_index);
 395
 396         spin_lock_irq(&group->lock);
 397         if (group->state == MCAST_PKEY_EVENT && !ret &&
 398             group->pkey_index == pkey_index)
 399                 goto out;
 400
 401         while (!list_empty(&group->active_list)) {
 402                 member = list_entry(group->active_list.next,
 403                                     struct mcast_member, list);
 404                 atomic_inc(&member->refcount);
 405                 list_del_init(&member->list);
 406                 adjust_membership(group, member->multicast.rec.join_state, -1);
 407                 member->state = MCAST_ERROR;
 408                 spin_unlock_irq(&group->lock);
 409
 410                 ret = member->multicast.callback(-ENETRESET,
 411                                                  &member->multicast);
 412                 deref_member(member);
 413                 if (ret)
 414                         ib_sa_free_multicast(&member->multicast);
 415                 spin_lock_irq(&group->lock);
 416         }
 417
 418         group->rec.join_state = 0;
 419 out:
 420         group->state = MCAST_BUSY;
 421         spin_unlock_irq(&group->lock);
 422 }
 423
 424 static void mcast_work_handler(struct work_struct *work)
 425 {
 426         struct mcast_group *group;
 427         struct mcast_member *member;
 428         struct ib_sa_multicast *multicast;
 429         int status, ret;
 430         u8 join_state;
 431
 432         group = container_of(work, typeof(*group), work);
 433 retest:
 434         spin_lock_irq(&group->lock);
 435         while (!list_empty(&group->pending_list) ||
 436                (group->state != MCAST_BUSY)) {
 437
 438                 if (group->state != MCAST_BUSY) {
 439                         spin_unlock_irq(&group->lock);
 440                         process_group_error(group);
 441                         goto retest;
 442                 }
 443
 444                 member = list_entry(group->pending_list.next,
 445                                     struct mcast_member, list);
 446                 multicast = &member->multicast;
 447                 join_state = multicast->rec.join_state;
 448                 atomic_inc(&member->refcount);
 449
 450                 if (join_state == (group->rec.join_state & join_state)) {
 451                         status = cmp_rec(&group->rec, &multicast->rec,
 452                                          multicast->comp_mask);
 453                         if (!status)
 454                                 join_group(group, member, join_state);
 455                         else
 456                                 list_del_init(&member->list);
 457                         spin_unlock_irq(&group->lock);
 458                         ret = multicast->callback(status, multicast);
 459                 } else {
 460                         spin_unlock_irq(&group->lock);
 461                         status = send_join(group, member);
 462                         if (!status) {
 463                                 deref_member(member);
 464                                 return;
 465                         }
 466                         ret = fail_join(group, member, status);
 467                 }
 468
 469                 deref_member(member);
 470                 if (ret)
 471                         ib_sa_free_multicast(&member->multicast);
 472                 spin_lock_irq(&group->lock);
 473         }
 474
 475         join_state = get_leave_state(group);
 476         if (join_state) {
 477                 group->rec.join_state &= ~join_state;
 478                 spin_unlock_irq(&group->lock);
 479                 if (send_leave(group, join_state))
 480                         goto retest;
 481         } else {
 482                 group->state = MCAST_IDLE;
 483                 spin_unlock_irq(&group->lock);
 484                 release_group(group);
 485         }
 486 }
 487
 488 /*
 489  * Fail a join request if it is still active - at the head of the pending queue.
 490  */
 491 static void process_join_error(struct mcast_group *group, int status)
 492 {
 493         struct mcast_member *member;
 494         int ret;
 495
 496         spin_lock_irq(&group->lock);
 497         member = list_entry(group->pending_list.next,
 498                             struct mcast_member, list);
 499         if (group->last_join == member) {
 500                 atomic_inc(&member->refcount);
 501                 list_del_init(&member->list);
 502                 spin_unlock_irq(&group->lock);
 503                 ret = member->multicast.callback(status, &member->multicast);
 504                 deref_member(member);
 505                 if (ret)
 506                         ib_sa_free_multicast(&member->multicast);
 507         } else
 508                 spin_unlock_irq(&group->lock);
 509 }
 510
 511 static void join_handler(int status, struct ib_sa_mcmember_rec *rec,
 512                          void *context)
 513 {
 514         struct mcast_group *group = context;
 515         u16 pkey_index = MCAST_INVALID_PKEY_INDEX;
 516
 517         if (status)
 518                 process_join_error(group, status);
 519         else {
 520                 int mgids_changed, is_mgid0;
 521
 522                 if (ib_find_pkey(group->port->dev->device,
 523                                  group->port->port_num, be16_to_cpu(rec->pkey),
 524                                  &pkey_index))
 525                         pkey_index = MCAST_INVALID_PKEY_INDEX;
 526
 527                 spin_lock_irq(&group->port->lock);
 528                 if (group->state == MCAST_BUSY &&
 529                     group->pkey_index == MCAST_INVALID_PKEY_INDEX)
 530                         group->pkey_index = pkey_index;
 531                 mgids_changed = memcmp(&rec->mgid, &group->rec.mgid,
 532                                        sizeof(group->rec.mgid));
 533                 group->rec = *rec;
 534                 if (mgids_changed) {
 535                         rb_erase(&group->node, &group->port->table);
 536                         is_mgid0 = !memcmp(&mgid0, &group->rec.mgid,
 537                                            sizeof(mgid0));
 538                         mcast_insert(group->port, group, is_mgid0);
 539                 }
 540                 spin_unlock_irq(&group->port->lock);
 541         }
 542         mcast_work_handler(&group->work);
 543 }
 544
 545 static void leave_handler(int status, struct ib_sa_mcmember_rec *rec,
 546                           void *context)
 547 {
 548         struct mcast_group *group = context;
 549
 550         if (status && group->retries > 0 &&
 551             !send_leave(group, group->leave_state))
 552                 group->retries--;
 553         else
 554                 mcast_work_handler(&group->work);
 555 }
 556
 557 static struct mcast_group *acquire_group(struct mcast_port *port,
 558                                          union ib_gid *mgid, gfp_t gfp_mask)
 559 {
 560         struct mcast_group *group, *cur_group;
 561         unsigned long flags;
 562         int is_mgid0;
 563
 564         is_mgid0 = !memcmp(&mgid0, mgid, sizeof mgid0);
 565         if (!is_mgid0) {
 566                 spin_lock_irqsave(&port->lock, flags);
 567                 group = mcast_find(port, mgid);
 568                 if (group)
 569                         goto found;
 570                 spin_unlock_irqrestore(&port->lock, flags);
 571         }
 572
 573         group = kzalloc(sizeof *group, gfp_mask);
 574         if (!group)
 575                 return NULL;
 576
 577         group->retries = 3;
 578         group->port = port;
 579         group->rec.mgid = *mgid;
 580         group->pkey_index = MCAST_INVALID_PKEY_INDEX;
 581         INIT_LIST_HEAD(&group->pending_list);
 582         INIT_LIST_HEAD(&group->active_list);
 583         INIT_WORK(&group->work, mcast_work_handler);
 584         spin_lock_init(&group->lock);
 585
 586         spin_lock_irqsave(&port->lock, flags);
 587         cur_group = mcast_insert(port, group, is_mgid0);
 588         if (cur_group) {
 589                 kfree(group);
 590                 group = cur_group;
 591         } else
 592                 atomic_inc(&port->refcount);
 593 found:
 594         atomic_inc(&group->refcount);
 595         spin_unlock_irqrestore(&port->lock, flags);
 596         return group;
 597 }
 598
 599 /*
 600  * We serialize all join requests to a single group to make our lives much
 601  * easier.  Otherwise, two users could try to join the same group
 602  * simultaneously, with different configurations, one could leave while the
 603  * join is in progress, etc., which makes locking around error recovery
 604  * difficult.
 605  */
 606 struct ib_sa_multicast *
 607 ib_sa_join_multicast(struct ib_sa_client *client,
 608                      struct ib_device *device, u8 port_num,
 609                      struct ib_sa_mcmember_rec *rec,
 610                      ib_sa_comp_mask comp_mask, gfp_t gfp_mask,
 611                      int (*callback)(int status,
 612                                      struct ib_sa_multicast *multicast),
 613                      void *context)
 614 {
 615         struct mcast_device *dev;
 616         struct mcast_member *member;
 617         struct ib_sa_multicast *multicast;
 618         int ret;
 619
 620         dev = ib_get_client_data(device, &mcast_client);
 621         if (!dev)
 622                 return ERR_PTR(-ENODEV);
 623
 624         member = kmalloc(sizeof *member, gfp_mask);
 625         if (!member)
 626                 return ERR_PTR(-ENOMEM);
 627
 628         ib_sa_client_get(client);
 629         member->client = client;
 630         member->multicast.rec = *rec;
 631         member->multicast.comp_mask = comp_mask;
 632         member->multicast.callback = callback;
 633         member->multicast.context = context;
 634         init_completion(&member->comp);
 635         atomic_set(&member->refcount, 1);
 636         member->state = MCAST_JOINING;
 637
 638         member->group = acquire_group(&dev->port[port_num - dev->start_port],
 639                                       &rec->mgid, gfp_mask);
 640         if (!member->group) {
 641                 ret = -ENOMEM;
 642                 goto err;
 643         }
 644
 645         /*
 646          * The user will get the multicast structure in their callback.  They
 647          * could then free the multicast structure before we can return from
 648          * this routine.  So we save the pointer to return before queuing
 649          * any callback.
 650          */
 651         multicast = &member->multicast;
 652         queue_join(member);
 653         return multicast;
 654
 655 err:
 656         ib_sa_client_put(client);
 657         kfree(member);
 658         return ERR_PTR(ret);
 659 }
 660 EXPORT_SYMBOL(ib_sa_join_multicast);
 661
 662 void ib_sa_free_multicast(struct ib_sa_multicast *multicast)
 663 {
 664         struct mcast_member *member;
 665         struct mcast_group *group;
 666
 667         member = container_of(multicast, struct mcast_member, multicast);
 668         group = member->group;
 669
 670         spin_lock_irq(&group->lock);
 671         if (member->state == MCAST_MEMBER)
 672                 adjust_membership(group, multicast->rec.join_state, -1);
 673
 674         list_del_init(&member->list);
 675
 676         if (group->state == MCAST_IDLE) {
 677                 group->state = MCAST_BUSY;
 678                 spin_unlock_irq(&group->lock);
 679                 /* Continue to hold reference on group until callback */
 680                 queue_work(mcast_wq, &group->work);
 681         } else {
 682                 spin_unlock_irq(&group->lock);
 683                 release_group(group);
 684         }
 685
 686         deref_member(member);
 687         wait_for_completion(&member->comp);
 688         ib_sa_client_put(member->client);
 689         kfree(member);
 690 }
 691 EXPORT_SYMBOL(ib_sa_free_multicast);
 692
 693 int ib_sa_get_mcmember_rec(struct ib_device *device, u8 port_num,
 694                            union ib_gid *mgid, struct ib_sa_mcmember_rec *rec)
 695 {
 696         struct mcast_device *dev;
 697         struct mcast_port *port;
 698         struct mcast_group *group;
 699         unsigned long flags;
 700         int ret = 0;
 701
 702         dev = ib_get_client_data(device, &mcast_client);
 703         if (!dev)
 704                 return -ENODEV;
 705
 706         port = &dev->port[port_num - dev->start_port];
 707         spin_lock_irqsave(&port->lock, flags);
 708         group = mcast_find(port, mgid);
 709         if (group)
 710                 *rec = group->rec;
 711         else
 712                 ret = -EADDRNOTAVAIL;
 713         spin_unlock_irqrestore(&port->lock, flags);
 714
 715         return ret;
 716 }
 717 EXPORT_SYMBOL(ib_sa_get_mcmember_rec);
 718
 719 int ib_init_ah_from_mcmember(struct ib_device *device, u8 port_num,
 720                              struct ib_sa_mcmember_rec *rec,
 721                              struct net_device *ndev,
 722                              enum ib_gid_type gid_type,
 723                              struct ib_ah_attr *ah_attr)
 724 {
 725         int ret;
 726         u16 gid_index;
 727
 728         /* GID table is not based on the netdevice for IB link layer,
 729          * so ignore ndev during search.
 730          */
 731         if (rdma_protocol_ib(device, port_num))
 732                 ndev = NULL;
 733         else if (!rdma_protocol_roce(device, port_num))
 734                 return -EINVAL;
 735
 736         ret = ib_find_cached_gid_by_port(device, &rec->port_gid,
 737                                          gid_type, port_num,
 738                                          ndev,
 739                                          &gid_index);
 740         if (ret)
 741                 return ret;
 742
 743         memset(ah_attr, 0, sizeof *ah_attr);
 744         ah_attr->dlid = be16_to_cpu(rec->mlid);
 745         ah_attr->sl = rec->sl;
 746         ah_attr->port_num = port_num;
 747         ah_attr->static_rate = rec->rate;
 748
 749         ah_attr->ah_flags = IB_AH_GRH;
 750         ah_attr->grh.dgid = rec->mgid;
 751
 752         ah_attr->grh.sgid_index = (u8) gid_index;
 753         ah_attr->grh.flow_label = be32_to_cpu(rec->flow_label);
 754         ah_attr->grh.hop_limit = rec->hop_limit;
 755         ah_attr->grh.traffic_class = rec->traffic_class;
 756
 757         return 0;
 758 }
 759 EXPORT_SYMBOL(ib_init_ah_from_mcmember);
 760
 761 static void mcast_groups_event(struct mcast_port *port,
 762                                enum mcast_group_state state)
 763 {
 764         struct mcast_group *group;
 765         struct rb_node *node;
 766         unsigned long flags;
 767
 768         spin_lock_irqsave(&port->lock, flags);
 769         for (node = rb_first(&port->table); node; node = rb_next(node)) {
 770                 group = rb_entry(node, struct mcast_group, node);
 771                 spin_lock(&group->lock);
 772                 if (group->state == MCAST_IDLE) {
 773                         atomic_inc(&group->refcount);
 774                         queue_work(mcast_wq, &group->work);
 775                 }
 776                 if (group->state != MCAST_GROUP_ERROR)
 777                         group->state = state;
 778                 spin_unlock(&group->lock);
 779         }
 780         spin_unlock_irqrestore(&port->lock, flags);
 781 }
 782
 783 static void mcast_event_handler(struct ib_event_handler *handler,
 784                                 struct ib_event *event)
 785 {
 786         struct mcast_device *dev;
 787         int index;
 788
 789         dev = container_of(handler, struct mcast_device, event_handler);
 790         if (!rdma_cap_ib_mcast(dev->device, event->element.port_num))
 791                 return;
 792
 793         index = event->element.port_num - dev->start_port;
 794
 795         switch (event->event) {
 796         case IB_EVENT_PORT_ERR:
 797         case IB_EVENT_LID_CHANGE:
 798         case IB_EVENT_SM_CHANGE:
 799         case IB_EVENT_CLIENT_REREGISTER:
 800                 mcast_groups_event(&dev->port[index], MCAST_GROUP_ERROR);
 801                 break;
 802         case IB_EVENT_PKEY_CHANGE:
 803                 mcast_groups_event(&dev->port[index], MCAST_PKEY_EVENT);
 804                 break;
 805         default:
 806                 break;
 807         }
 808 }
 809
 810 static void mcast_add_one(struct ib_device *device)
 811 {
 812         struct mcast_device *dev;
 813         struct mcast_port *port;
 814         int i;
 815         int count = 0;
 816
 817         dev = kmalloc(sizeof *dev + device->phys_port_cnt * sizeof *port,
 818                       GFP_KERNEL);
 819         if (!dev)
 820                 return;
 821
 822         dev->start_port = rdma_start_port(device);
 823         dev->end_port = rdma_end_port(device);
 824
 825         for (i = 0; i <= dev->end_port - dev->start_port; i++) {
 826                 if (!rdma_cap_ib_mcast(device, dev->start_port + i))
 827                         continue;
 828                 port = &dev->port[i];
 829                 port->dev = dev;
 830                 port->port_num = dev->start_port + i;
 831                 spin_lock_init(&port->lock);
 832                 port->table = RB_ROOT;
 833                 init_completion(&port->comp);
 834                 atomic_set(&port->refcount, 1);
 835                 ++count;
 836         }
 837
 838         if (!count) {
 839                 kfree(dev);
 840                 return;
 841         }
 842
 843         dev->device = device;
 844         ib_set_client_data(device, &mcast_client, dev);
 845
 846         INIT_IB_EVENT_HANDLER(&dev->event_handler, device, mcast_event_handler);
 847         ib_register_event_handler(&dev->event_handler);
 848 }
 849
 850 static void mcast_remove_one(struct ib_device *device, void *client_data)
 851 {
 852         struct mcast_device *dev = client_data;
 853         struct mcast_port *port;
 854         int i;
 855
 856         if (!dev)
 857                 return;
 858
 859         ib_unregister_event_handler(&dev->event_handler);
 860         flush_workqueue(mcast_wq);
 861
 862         for (i = 0; i <= dev->end_port - dev->start_port; i++) {
 863                 if (rdma_cap_ib_mcast(device, dev->start_port + i)) {
 864                         port = &dev->port[i];
 865                         deref_port(port);
 866                         wait_for_completion(&port->comp);
 867                 }
 868         }
 869
 870         kfree(dev);
 871 }
 872
 873 int mcast_init(void)
 874 {
 875         int ret;
 876
 877         mcast_wq = alloc_ordered_workqueue("ib_mcast", WQ_MEM_RECLAIM);
 878         if (!mcast_wq)
 879                 return -ENOMEM;
 880
 881         ib_sa_register_client(&sa_client);
 882
 883         ret = ib_register_client(&mcast_client);
 884         if (ret)
 885                 goto err;
 886         return 0;
 887
 888 err:
 889         ib_sa_unregister_client(&sa_client);
 890         destroy_workqueue(mcast_wq);
 891         return ret;
 892 }
 893
 894 void mcast_cleanup(void)
 895 {
 896         ib_unregister_client(&mcast_client);
 897         ib_sa_unregister_client(&sa_client);
 898         destroy_workqueue(mcast_wq);
 899 }