linux/drivers/infiniband/core/multicast.c
<<
>>
Prefs
   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
  45static void mcast_add_one(struct ib_device *device);
  46static void mcast_remove_one(struct ib_device *device, void *client_data);
  47
  48static struct ib_client mcast_client = {
  49        .name   = "ib_multicast",
  50        .add    = mcast_add_one,
  51        .remove = mcast_remove_one
  52};
  53
  54static struct ib_sa_client      sa_client;
  55static struct workqueue_struct  *mcast_wq;
  56static union ib_gid mgid0;
  57
  58struct mcast_device;
  59
  60struct 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
  69struct 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
  77enum mcast_state {
  78        MCAST_JOINING,
  79        MCAST_MEMBER,
  80        MCAST_ERROR,
  81};
  82
  83enum mcast_group_state {
  84        MCAST_IDLE,
  85        MCAST_BUSY,
  86        MCAST_GROUP_ERROR,
  87        MCAST_PKEY_EVENT
  88};
  89
  90enum {
  91        MCAST_INVALID_PKEY_INDEX = 0xFFFF
  92};
  93
  94struct mcast_member;
  95
  96struct 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
 114struct 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
 124static void join_handler(int status, struct ib_sa_mcmember_rec *rec,
 125                         void *context);
 126static void leave_handler(int status, struct ib_sa_mcmember_rec *rec,
 127                          void *context);
 128
 129static 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
 150static 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
 179static void deref_port(struct mcast_port *port)
 180{
 181        if (atomic_dec_and_test(&port->refcount))
 182                complete(&port->comp);
 183}
 184
 185static 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
 200static void deref_member(struct mcast_member *member)
 201{
 202        if (atomic_dec_and_test(&member->refcount))
 203                complete(&member->comp);
 204}
 205
 206static 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 */
 228static 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 */
 243static 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
 255static 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
 283static 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
 330static 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
 345static 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
 365static 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
 376static 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
 385static 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;
 419out:
 420        group->state = MCAST_BUSY;
 421        spin_unlock_irq(&group->lock);
 422}
 423
 424static 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);
 433retest:
 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 */
 491static 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
 511static 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
 545static 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
 557static 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);
 593found:
 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 */
 606struct ib_sa_multicast *
 607ib_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
 655err:
 656        ib_sa_client_put(client);
 657        kfree(member);
 658        return ERR_PTR(ret);
 659}
 660EXPORT_SYMBOL(ib_sa_join_multicast);
 661
 662void 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}
 691EXPORT_SYMBOL(ib_sa_free_multicast);
 692
 693int 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}
 717EXPORT_SYMBOL(ib_sa_get_mcmember_rec);
 718
 719/**
 720 * ib_init_ah_from_mcmember - Initialize AH attribute from multicast
 721 * member record and gid of the device.
 722 * @device:     RDMA device
 723 * @port_num:   Port of the rdma device to consider
 724 * @ndev:       Optional netdevice, applicable only for RoCE
 725 * @gid_type:   GID type to consider
 726 * @ah_attr:    AH attribute to fillup on successful completion
 727 *
 728 * ib_init_ah_from_mcmember() initializes AH attribute based on multicast
 729 * member record and other device properties. On success the caller is
 730 * responsible to call rdma_destroy_ah_attr on the ah_attr. Returns 0 on
 731 * success or appropriate error code.
 732 *
 733 */
 734int ib_init_ah_from_mcmember(struct ib_device *device, u8 port_num,
 735                             struct ib_sa_mcmember_rec *rec,
 736                             struct net_device *ndev,
 737                             enum ib_gid_type gid_type,
 738                             struct rdma_ah_attr *ah_attr)
 739{
 740        const struct ib_gid_attr *sgid_attr;
 741
 742        /* GID table is not based on the netdevice for IB link layer,
 743         * so ignore ndev during search.
 744         */
 745        if (rdma_protocol_ib(device, port_num))
 746                ndev = NULL;
 747        else if (!rdma_protocol_roce(device, port_num))
 748                return -EINVAL;
 749
 750        sgid_attr = rdma_find_gid_by_port(device, &rec->port_gid,
 751                                          gid_type, port_num, ndev);
 752        if (IS_ERR(sgid_attr))
 753                return PTR_ERR(sgid_attr);
 754
 755        memset(ah_attr, 0, sizeof(*ah_attr));
 756        ah_attr->type = rdma_ah_find_type(device, port_num);
 757
 758        rdma_ah_set_dlid(ah_attr, be16_to_cpu(rec->mlid));
 759        rdma_ah_set_sl(ah_attr, rec->sl);
 760        rdma_ah_set_port_num(ah_attr, port_num);
 761        rdma_ah_set_static_rate(ah_attr, rec->rate);
 762        rdma_move_grh_sgid_attr(ah_attr, &rec->mgid,
 763                                be32_to_cpu(rec->flow_label),
 764                                rec->hop_limit, rec->traffic_class,
 765                                sgid_attr);
 766        return 0;
 767}
 768EXPORT_SYMBOL(ib_init_ah_from_mcmember);
 769
 770static void mcast_groups_event(struct mcast_port *port,
 771                               enum mcast_group_state state)
 772{
 773        struct mcast_group *group;
 774        struct rb_node *node;
 775        unsigned long flags;
 776
 777        spin_lock_irqsave(&port->lock, flags);
 778        for (node = rb_first(&port->table); node; node = rb_next(node)) {
 779                group = rb_entry(node, struct mcast_group, node);
 780                spin_lock(&group->lock);
 781                if (group->state == MCAST_IDLE) {
 782                        atomic_inc(&group->refcount);
 783                        queue_work(mcast_wq, &group->work);
 784                }
 785                if (group->state != MCAST_GROUP_ERROR)
 786                        group->state = state;
 787                spin_unlock(&group->lock);
 788        }
 789        spin_unlock_irqrestore(&port->lock, flags);
 790}
 791
 792static void mcast_event_handler(struct ib_event_handler *handler,
 793                                struct ib_event *event)
 794{
 795        struct mcast_device *dev;
 796        int index;
 797
 798        dev = container_of(handler, struct mcast_device, event_handler);
 799        if (!rdma_cap_ib_mcast(dev->device, event->element.port_num))
 800                return;
 801
 802        index = event->element.port_num - dev->start_port;
 803
 804        switch (event->event) {
 805        case IB_EVENT_PORT_ERR:
 806        case IB_EVENT_LID_CHANGE:
 807        case IB_EVENT_CLIENT_REREGISTER:
 808                mcast_groups_event(&dev->port[index], MCAST_GROUP_ERROR);
 809                break;
 810        case IB_EVENT_PKEY_CHANGE:
 811                mcast_groups_event(&dev->port[index], MCAST_PKEY_EVENT);
 812                break;
 813        default:
 814                break;
 815        }
 816}
 817
 818static void mcast_add_one(struct ib_device *device)
 819{
 820        struct mcast_device *dev;
 821        struct mcast_port *port;
 822        int i;
 823        int count = 0;
 824
 825        dev = kmalloc(struct_size(dev, port, device->phys_port_cnt),
 826                      GFP_KERNEL);
 827        if (!dev)
 828                return;
 829
 830        dev->start_port = rdma_start_port(device);
 831        dev->end_port = rdma_end_port(device);
 832
 833        for (i = 0; i <= dev->end_port - dev->start_port; i++) {
 834                if (!rdma_cap_ib_mcast(device, dev->start_port + i))
 835                        continue;
 836                port = &dev->port[i];
 837                port->dev = dev;
 838                port->port_num = dev->start_port + i;
 839                spin_lock_init(&port->lock);
 840                port->table = RB_ROOT;
 841                init_completion(&port->comp);
 842                atomic_set(&port->refcount, 1);
 843                ++count;
 844        }
 845
 846        if (!count) {
 847                kfree(dev);
 848                return;
 849        }
 850
 851        dev->device = device;
 852        ib_set_client_data(device, &mcast_client, dev);
 853
 854        INIT_IB_EVENT_HANDLER(&dev->event_handler, device, mcast_event_handler);
 855        ib_register_event_handler(&dev->event_handler);
 856}
 857
 858static void mcast_remove_one(struct ib_device *device, void *client_data)
 859{
 860        struct mcast_device *dev = client_data;
 861        struct mcast_port *port;
 862        int i;
 863
 864        if (!dev)
 865                return;
 866
 867        ib_unregister_event_handler(&dev->event_handler);
 868        flush_workqueue(mcast_wq);
 869
 870        for (i = 0; i <= dev->end_port - dev->start_port; i++) {
 871                if (rdma_cap_ib_mcast(device, dev->start_port + i)) {
 872                        port = &dev->port[i];
 873                        deref_port(port);
 874                        wait_for_completion(&port->comp);
 875                }
 876        }
 877
 878        kfree(dev);
 879}
 880
 881int mcast_init(void)
 882{
 883        int ret;
 884
 885        mcast_wq = alloc_ordered_workqueue("ib_mcast", WQ_MEM_RECLAIM);
 886        if (!mcast_wq)
 887                return -ENOMEM;
 888
 889        ib_sa_register_client(&sa_client);
 890
 891        ret = ib_register_client(&mcast_client);
 892        if (ret)
 893                goto err;
 894        return 0;
 895
 896err:
 897        ib_sa_unregister_client(&sa_client);
 898        destroy_workqueue(mcast_wq);
 899        return ret;
 900}
 901
 902void mcast_cleanup(void)
 903{
 904        ib_unregister_client(&mcast_client);
 905        ib_sa_unregister_client(&sa_client);
 906        destroy_workqueue(mcast_wq);
 907}
 908