net/switchdev/switchdev.c

   1 /*
   2  * net/switchdev/switchdev.c - Switch device API
   3  * Copyright (c) 2014 Jiri Pirko <jiri@resnulli.us>
   4  * Copyright (c) 2014-2015 Scott Feldman <sfeldma@gmail.com>
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License as published by
   8  * the Free Software Foundation; either version 2 of the License, or
   9  * (at your option) any later version.
  10  */
  11
  12 #include <linux/kernel.h>
  13 #include <linux/types.h>
  14 #include <linux/init.h>
  15 #include <linux/mutex.h>
  16 #include <linux/notifier.h>
  17 #include <linux/netdevice.h>
  18 #include <linux/if_bridge.h>
  19 #include <net/ip_fib.h>
  20 #include <net/switchdev.h>
  21
  22 /**
  23  *      switchdev_port_attr_get - Get port attribute
  24  *
  25  *      @dev: port device
  26  *      @attr: attribute to get
  27  */
  28 int switchdev_port_attr_get(struct net_device *dev, struct switchdev_attr *attr)
  29 {
  30         const struct switchdev_ops *ops = dev->switchdev_ops;
  31         struct net_device *lower_dev;
  32         struct list_head *iter;
  33         struct switchdev_attr first = {
  34                 .id = SWITCHDEV_ATTR_UNDEFINED
  35         };
  36         int err = -EOPNOTSUPP;
  37
  38         if (ops && ops->switchdev_port_attr_get)
  39                 return ops->switchdev_port_attr_get(dev, attr);
  40
  41         if (attr->flags & SWITCHDEV_F_NO_RECURSE)
  42                 return err;
  43
  44         /* Switch device port(s) may be stacked under
  45          * bond/team/vlan dev, so recurse down to get attr on
  46          * each port.  Return -ENODATA if attr values don't
  47          * compare across ports.
  48          */
  49
  50         netdev_for_each_lower_dev(dev, lower_dev, iter) {
  51                 err = switchdev_port_attr_get(lower_dev, attr);
  52                 if (err)
  53                         break;
  54                 if (first.id == SWITCHDEV_ATTR_UNDEFINED)
  55                         first = *attr;
  56                 else if (memcmp(&first, attr, sizeof(*attr)))
  57                         return -ENODATA;
  58         }
  59
  60         return err;
  61 }
  62 EXPORT_SYMBOL_GPL(switchdev_port_attr_get);
  63
  64 static int __switchdev_port_attr_set(struct net_device *dev,
  65                                      struct switchdev_attr *attr)
  66 {
  67         const struct switchdev_ops *ops = dev->switchdev_ops;
  68         struct net_device *lower_dev;
  69         struct list_head *iter;
  70         int err = -EOPNOTSUPP;
  71
  72         if (ops && ops->switchdev_port_attr_set)
  73                 return ops->switchdev_port_attr_set(dev, attr);
  74
  75         if (attr->flags & SWITCHDEV_F_NO_RECURSE)
  76                 return err;
  77
  78         /* Switch device port(s) may be stacked under
  79          * bond/team/vlan dev, so recurse down to set attr on
  80          * each port.
  81          */
  82
  83         netdev_for_each_lower_dev(dev, lower_dev, iter) {
  84                 err = __switchdev_port_attr_set(lower_dev, attr);
  85                 if (err)
  86                         break;
  87         }
  88
  89         return err;
  90 }
  91
  92 struct switchdev_attr_set_work {
  93         struct work_struct work;
  94         struct net_device *dev;
  95         struct switchdev_attr attr;
  96 };
  97
  98 static void switchdev_port_attr_set_work(struct work_struct *work)
  99 {
 100         struct switchdev_attr_set_work *asw =
 101                 container_of(work, struct switchdev_attr_set_work, work);
 102         int err;
 103
 104         rtnl_lock();
 105         err = switchdev_port_attr_set(asw->dev, &asw->attr);
 106         if (err && err != -EOPNOTSUPP)
 107                 netdev_err(asw->dev, "failed (err=%d) to set attribute (id=%d)\n",
 108                            err, asw->attr.id);
 109         rtnl_unlock();
 110
 111         dev_put(asw->dev);
 112         kfree(work);
 113 }
 114
 115 static int switchdev_port_attr_set_defer(struct net_device *dev,
 116                                          struct switchdev_attr *attr)
 117 {
 118         struct switchdev_attr_set_work *asw;
 119
 120         asw = kmalloc(sizeof(*asw), GFP_ATOMIC);
 121         if (!asw)
 122                 return -ENOMEM;
 123
 124         INIT_WORK(&asw->work, switchdev_port_attr_set_work);
 125
 126         dev_hold(dev);
 127         asw->dev = dev;
 128         memcpy(&asw->attr, attr, sizeof(asw->attr));
 129
 130         schedule_work(&asw->work);
 131
 132         return 0;
 133 }
 134
 135 /**
 136  *      switchdev_port_attr_set - Set port attribute
 137  *
 138  *      @dev: port device
 139  *      @attr: attribute to set
 140  *
 141  *      Use a 2-phase prepare-commit transaction model to ensure
 142  *      system is not left in a partially updated state due to
 143  *      failure from driver/device.
 144  */
 145 int switchdev_port_attr_set(struct net_device *dev, struct switchdev_attr *attr)
 146 {
 147         int err;
 148
 149         if (!rtnl_is_locked()) {
 150                 /* Running prepare-commit transaction across stacked
 151                  * devices requires nothing moves, so if rtnl_lock is
 152                  * not held, schedule a worker thread to hold rtnl_lock
 153                  * while setting attr.
 154                  */
 155
 156                 return switchdev_port_attr_set_defer(dev, attr);
 157         }
 158
 159         /* Phase I: prepare for attr set. Driver/device should fail
 160          * here if there are going to be issues in the commit phase,
 161          * such as lack of resources or support.  The driver/device
 162          * should reserve resources needed for the commit phase here,
 163          * but should not commit the attr.
 164          */
 165
 166         attr->trans = SWITCHDEV_TRANS_PREPARE;
 167         err = __switchdev_port_attr_set(dev, attr);
 168         if (err) {
 169                 /* Prepare phase failed: abort the transaction.  Any
 170                  * resources reserved in the prepare phase are
 171                  * released.
 172                  */
 173
 174                 attr->trans = SWITCHDEV_TRANS_ABORT;
 175                 __switchdev_port_attr_set(dev, attr);
 176
 177                 return err;
 178         }
 179
 180         /* Phase II: commit attr set.  This cannot fail as a fault
 181          * of driver/device.  If it does, it's a bug in the driver/device
 182          * because the driver said everythings was OK in phase I.
 183          */
 184
 185         attr->trans = SWITCHDEV_TRANS_COMMIT;
 186         err = __switchdev_port_attr_set(dev, attr);
 187         BUG_ON(err);
 188
 189         return err;
 190 }
 191 EXPORT_SYMBOL_GPL(switchdev_port_attr_set);
 192
 193 static int __switchdev_port_obj_add(struct net_device *dev,
 194                                     struct switchdev_obj *obj)
 195 {
 196         const struct switchdev_ops *ops = dev->switchdev_ops;
 197         struct net_device *lower_dev;
 198         struct list_head *iter;
 199         int err = -EOPNOTSUPP;
 200
 201         if (ops && ops->switchdev_port_obj_add)
 202                 return ops->switchdev_port_obj_add(dev, obj);
 203
 204         /* Switch device port(s) may be stacked under
 205          * bond/team/vlan dev, so recurse down to add object on
 206          * each port.
 207          */
 208
 209         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 210                 err = __switchdev_port_obj_add(lower_dev, obj);
 211                 if (err)
 212                         break;
 213         }
 214
 215         return err;
 216 }
 217
 218 /**
 219  *      switchdev_port_obj_add - Add port object
 220  *
 221  *      @dev: port device
 222  *      @obj: object to add
 223  *
 224  *      Use a 2-phase prepare-commit transaction model to ensure
 225  *      system is not left in a partially updated state due to
 226  *      failure from driver/device.
 227  *
 228  *      rtnl_lock must be held.
 229  */
 230 int switchdev_port_obj_add(struct net_device *dev, struct switchdev_obj *obj)
 231 {
 232         int err;
 233
 234         ASSERT_RTNL();
 235
 236         /* Phase I: prepare for obj add. Driver/device should fail
 237          * here if there are going to be issues in the commit phase,
 238          * such as lack of resources or support.  The driver/device
 239          * should reserve resources needed for the commit phase here,
 240          * but should not commit the obj.
 241          */
 242
 243         obj->trans = SWITCHDEV_TRANS_PREPARE;
 244         err = __switchdev_port_obj_add(dev, obj);
 245         if (err) {
 246                 /* Prepare phase failed: abort the transaction.  Any
 247                  * resources reserved in the prepare phase are
 248                  * released.
 249                  */
 250
 251                 obj->trans = SWITCHDEV_TRANS_ABORT;
 252                 __switchdev_port_obj_add(dev, obj);
 253
 254                 return err;
 255         }
 256
 257         /* Phase II: commit obj add.  This cannot fail as a fault
 258          * of driver/device.  If it does, it's a bug in the driver/device
 259          * because the driver said everythings was OK in phase I.
 260          */
 261
 262         obj->trans = SWITCHDEV_TRANS_COMMIT;
 263         err = __switchdev_port_obj_add(dev, obj);
 264         WARN(err, "%s: Commit of object (id=%d) failed.\n", dev->name, obj->id);
 265
 266         return err;
 267 }
 268 EXPORT_SYMBOL_GPL(switchdev_port_obj_add);
 269
 270 /**
 271  *      switchdev_port_obj_del - Delete port object
 272  *
 273  *      @dev: port device
 274  *      @obj: object to delete
 275  */
 276 int switchdev_port_obj_del(struct net_device *dev, struct switchdev_obj *obj)
 277 {
 278         const struct switchdev_ops *ops = dev->switchdev_ops;
 279         struct net_device *lower_dev;
 280         struct list_head *iter;
 281         int err = -EOPNOTSUPP;
 282
 283         if (ops && ops->switchdev_port_obj_del)
 284                 return ops->switchdev_port_obj_del(dev, obj);
 285
 286         /* Switch device port(s) may be stacked under
 287          * bond/team/vlan dev, so recurse down to delete object on
 288          * each port.
 289          */
 290
 291         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 292                 err = switchdev_port_obj_del(lower_dev, obj);
 293                 if (err)
 294                         break;
 295         }
 296
 297         return err;
 298 }
 299 EXPORT_SYMBOL_GPL(switchdev_port_obj_del);
 300
 301 /**
 302  *      switchdev_port_obj_dump - Dump port objects
 303  *
 304  *      @dev: port device
 305  *      @obj: object to dump
 306  */
 307 int switchdev_port_obj_dump(struct net_device *dev, struct switchdev_obj *obj)
 308 {
 309         const struct switchdev_ops *ops = dev->switchdev_ops;
 310         struct net_device *lower_dev;
 311         struct list_head *iter;
 312         int err = -EOPNOTSUPP;
 313
 314         if (ops && ops->switchdev_port_obj_dump)
 315                 return ops->switchdev_port_obj_dump(dev, obj);
 316
 317         /* Switch device port(s) may be stacked under
 318          * bond/team/vlan dev, so recurse down to dump objects on
 319          * first port at bottom of stack.
 320          */
 321
 322         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 323                 err = switchdev_port_obj_dump(lower_dev, obj);
 324                 break;
 325         }
 326
 327         return err;
 328 }
 329 EXPORT_SYMBOL_GPL(switchdev_port_obj_dump);
 330
 331 static DEFINE_MUTEX(switchdev_mutex);
 332 static RAW_NOTIFIER_HEAD(switchdev_notif_chain);
 333
 334 /**
 335  *      register_switchdev_notifier - Register notifier
 336  *      @nb: notifier_block
 337  *
 338  *      Register switch device notifier. This should be used by code
 339  *      which needs to monitor events happening in particular device.
 340  *      Return values are same as for atomic_notifier_chain_register().
 341  */
 342 int register_switchdev_notifier(struct notifier_block *nb)
 343 {
 344         int err;
 345
 346         mutex_lock(&switchdev_mutex);
 347         err = raw_notifier_chain_register(&switchdev_notif_chain, nb);
 348         mutex_unlock(&switchdev_mutex);
 349         return err;
 350 }
 351 EXPORT_SYMBOL_GPL(register_switchdev_notifier);
 352
 353 /**
 354  *      unregister_switchdev_notifier - Unregister notifier
 355  *      @nb: notifier_block
 356  *
 357  *      Unregister switch device notifier.
 358  *      Return values are same as for atomic_notifier_chain_unregister().
 359  */
 360 int unregister_switchdev_notifier(struct notifier_block *nb)
 361 {
 362         int err;
 363
 364         mutex_lock(&switchdev_mutex);
 365         err = raw_notifier_chain_unregister(&switchdev_notif_chain, nb);
 366         mutex_unlock(&switchdev_mutex);
 367         return err;
 368 }
 369 EXPORT_SYMBOL_GPL(unregister_switchdev_notifier);
 370
 371 /**
 372  *      call_switchdev_notifiers - Call notifiers
 373  *      @val: value passed unmodified to notifier function
 374  *      @dev: port device
 375  *      @info: notifier information data
 376  *
 377  *      Call all network notifier blocks. This should be called by driver
 378  *      when it needs to propagate hardware event.
 379  *      Return values are same as for atomic_notifier_call_chain().
 380  */
 381 int call_switchdev_notifiers(unsigned long val, struct net_device *dev,
 382                              struct switchdev_notifier_info *info)
 383 {
 384         int err;
 385
 386         info->dev = dev;
 387         mutex_lock(&switchdev_mutex);
 388         err = raw_notifier_call_chain(&switchdev_notif_chain, val, info);
 389         mutex_unlock(&switchdev_mutex);
 390         return err;
 391 }
 392 EXPORT_SYMBOL_GPL(call_switchdev_notifiers);
 393
 394 /**
 395  *      switchdev_port_bridge_getlink - Get bridge port attributes
 396  *
 397  *      @dev: port device
 398  *
 399  *      Called for SELF on rtnl_bridge_getlink to get bridge port
 400  *      attributes.
 401  */
 402 int switchdev_port_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
 403                                   struct net_device *dev, u32 filter_mask,
 404                                   int nlflags)
 405 {
 406         struct switchdev_attr attr = {
 407                 .id = SWITCHDEV_ATTR_PORT_BRIDGE_FLAGS,
 408         };
 409         u16 mode = BRIDGE_MODE_UNDEF;
 410         u32 mask = BR_LEARNING | BR_LEARNING_SYNC;
 411         int err;
 412
 413         err = switchdev_port_attr_get(dev, &attr);
 414         if (err)
 415                 return err;
 416
 417         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode,
 418                                        attr.u.brport_flags, mask, nlflags);
 419 }
 420 EXPORT_SYMBOL_GPL(switchdev_port_bridge_getlink);
 421
 422 static int switchdev_port_br_setflag(struct net_device *dev,
 423                                      struct nlattr *nlattr,
 424                                      unsigned long brport_flag)
 425 {
 426         struct switchdev_attr attr = {
 427                 .id = SWITCHDEV_ATTR_PORT_BRIDGE_FLAGS,
 428         };
 429         u8 flag = nla_get_u8(nlattr);
 430         int err;
 431
 432         err = switchdev_port_attr_get(dev, &attr);
 433         if (err)
 434                 return err;
 435
 436         if (flag)
 437                 attr.u.brport_flags |= brport_flag;
 438         else
 439                 attr.u.brport_flags &= ~brport_flag;
 440
 441         return switchdev_port_attr_set(dev, &attr);
 442 }
 443
 444 static const struct nla_policy
 445 switchdev_port_bridge_policy[IFLA_BRPORT_MAX + 1] = {
 446         [IFLA_BRPORT_STATE]             = { .type = NLA_U8 },
 447         [IFLA_BRPORT_COST]              = { .type = NLA_U32 },
 448         [IFLA_BRPORT_PRIORITY]          = { .type = NLA_U16 },
 449         [IFLA_BRPORT_MODE]              = { .type = NLA_U8 },
 450         [IFLA_BRPORT_GUARD]             = { .type = NLA_U8 },
 451         [IFLA_BRPORT_PROTECT]           = { .type = NLA_U8 },
 452         [IFLA_BRPORT_FAST_LEAVE]        = { .type = NLA_U8 },
 453         [IFLA_BRPORT_LEARNING]          = { .type = NLA_U8 },
 454         [IFLA_BRPORT_LEARNING_SYNC]     = { .type = NLA_U8 },
 455         [IFLA_BRPORT_UNICAST_FLOOD]     = { .type = NLA_U8 },
 456 };
 457
 458 static int switchdev_port_br_setlink_protinfo(struct net_device *dev,
 459                                               struct nlattr *protinfo)
 460 {
 461         struct nlattr *attr;
 462         int rem;
 463         int err;
 464
 465         err = nla_validate_nested(protinfo, IFLA_BRPORT_MAX,
 466                                   switchdev_port_bridge_policy);
 467         if (err)
 468                 return err;
 469
 470         nla_for_each_nested(attr, protinfo, rem) {
 471                 switch (nla_type(attr)) {
 472                 case IFLA_BRPORT_LEARNING:
 473                         err = switchdev_port_br_setflag(dev, attr,
 474                                                         BR_LEARNING);
 475                         break;
 476                 case IFLA_BRPORT_LEARNING_SYNC:
 477                         err = switchdev_port_br_setflag(dev, attr,
 478                                                         BR_LEARNING_SYNC);
 479                         break;
 480                 default:
 481                         err = -EOPNOTSUPP;
 482                         break;
 483                 }
 484                 if (err)
 485                         return err;
 486         }
 487
 488         return 0;
 489 }
 490
 491 static int switchdev_port_br_afspec(struct net_device *dev,
 492                                     struct nlattr *afspec,
 493                                     int (*f)(struct net_device *dev,
 494                                              struct switchdev_obj *obj))
 495 {
 496         struct nlattr *attr;
 497         struct bridge_vlan_info *vinfo;
 498         struct switchdev_obj obj = {
 499                 .id = SWITCHDEV_OBJ_PORT_VLAN,
 500         };
 501         struct switchdev_obj_vlan *vlan = &obj.u.vlan;
 502         int rem;
 503         int err;
 504
 505         nla_for_each_nested(attr, afspec, rem) {
 506                 if (nla_type(attr) != IFLA_BRIDGE_VLAN_INFO)
 507                         continue;
 508                 if (nla_len(attr) != sizeof(struct bridge_vlan_info))
 509                         return -EINVAL;
 510                 vinfo = nla_data(attr);
 511                 vlan->flags = vinfo->flags;
 512                 if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_BEGIN) {
 513                         if (vlan->vid_begin)
 514                                 return -EINVAL;
 515                         vlan->vid_begin = vinfo->vid;
 516                 } else if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_END) {
 517                         if (!vlan->vid_begin)
 518                                 return -EINVAL;
 519                         vlan->vid_end = vinfo->vid;
 520                         if (vlan->vid_end <= vlan->vid_begin)
 521                                 return -EINVAL;
 522                         err = f(dev, &obj);
 523                         if (err)
 524                                 return err;
 525                         memset(vlan, 0, sizeof(*vlan));
 526                 } else {
 527                         if (vlan->vid_begin)
 528                                 return -EINVAL;
 529                         vlan->vid_begin = vinfo->vid;
 530                         vlan->vid_end = vinfo->vid;
 531                         err = f(dev, &obj);
 532                         if (err)
 533                                 return err;
 534                         memset(vlan, 0, sizeof(*vlan));
 535                 }
 536         }
 537
 538         return 0;
 539 }
 540
 541 /**
 542  *      switchdev_port_bridge_setlink - Set bridge port attributes
 543  *
 544  *      @dev: port device
 545  *      @nlh: netlink header
 546  *      @flags: netlink flags
 547  *
 548  *      Called for SELF on rtnl_bridge_setlink to set bridge port
 549  *      attributes.
 550  */
 551 int switchdev_port_bridge_setlink(struct net_device *dev,
 552                                   struct nlmsghdr *nlh, u16 flags)
 553 {
 554         struct nlattr *protinfo;
 555         struct nlattr *afspec;
 556         int err = 0;
 557
 558         protinfo = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
 559                                    IFLA_PROTINFO);
 560         if (protinfo) {
 561                 err = switchdev_port_br_setlink_protinfo(dev, protinfo);
 562                 if (err)
 563                         return err;
 564         }
 565
 566         afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
 567                                  IFLA_AF_SPEC);
 568         if (afspec)
 569                 err = switchdev_port_br_afspec(dev, afspec,
 570                                                switchdev_port_obj_add);
 571
 572         return err;
 573 }
 574 EXPORT_SYMBOL_GPL(switchdev_port_bridge_setlink);
 575
 576 /**
 577  *      switchdev_port_bridge_dellink - Set bridge port attributes
 578  *
 579  *      @dev: port device
 580  *      @nlh: netlink header
 581  *      @flags: netlink flags
 582  *
 583  *      Called for SELF on rtnl_bridge_dellink to set bridge port
 584  *      attributes.
 585  */
 586 int switchdev_port_bridge_dellink(struct net_device *dev,
 587                                   struct nlmsghdr *nlh, u16 flags)
 588 {
 589         struct nlattr *afspec;
 590
 591         afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
 592                                  IFLA_AF_SPEC);
 593         if (afspec)
 594                 return switchdev_port_br_afspec(dev, afspec,
 595                                                 switchdev_port_obj_del);
 596
 597         return 0;
 598 }
 599 EXPORT_SYMBOL_GPL(switchdev_port_bridge_dellink);
 600
 601 /**
 602  *      switchdev_port_fdb_add - Add FDB (MAC/VLAN) entry to port
 603  *
 604  *      @ndmsg: netlink hdr
 605  *      @nlattr: netlink attributes
 606  *      @dev: port device
 607  *      @addr: MAC address to add
 608  *      @vid: VLAN to add
 609  *
 610  *      Add FDB entry to switch device.
 611  */
 612 int switchdev_port_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
 613                            struct net_device *dev, const unsigned char *addr,
 614                            u16 vid, u16 nlm_flags)
 615 {
 616         struct switchdev_obj obj = {
 617                 .id = SWITCHDEV_OBJ_PORT_FDB,
 618                 .u.fdb = {
 619                         .addr = addr,
 620                         .vid = vid,
 621                 },
 622         };
 623
 624         return switchdev_port_obj_add(dev, &obj);
 625 }
 626 EXPORT_SYMBOL_GPL(switchdev_port_fdb_add);
 627
 628 /**
 629  *      switchdev_port_fdb_del - Delete FDB (MAC/VLAN) entry from port
 630  *
 631  *      @ndmsg: netlink hdr
 632  *      @nlattr: netlink attributes
 633  *      @dev: port device
 634  *      @addr: MAC address to delete
 635  *      @vid: VLAN to delete
 636  *
 637  *      Delete FDB entry from switch device.
 638  */
 639 int switchdev_port_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
 640                            struct net_device *dev, const unsigned char *addr,
 641                            u16 vid)
 642 {
 643         struct switchdev_obj obj = {
 644                 .id = SWITCHDEV_OBJ_PORT_FDB,
 645                 .u.fdb = {
 646                         .addr = addr,
 647                         .vid = vid,
 648                 },
 649         };
 650
 651         return switchdev_port_obj_del(dev, &obj);
 652 }
 653 EXPORT_SYMBOL_GPL(switchdev_port_fdb_del);
 654
 655 struct switchdev_fdb_dump {
 656         struct switchdev_obj obj;
 657         struct sk_buff *skb;
 658         struct netlink_callback *cb;
 659         int idx;
 660 };
 661
 662 static int switchdev_port_fdb_dump_cb(struct net_device *dev,
 663                                       struct switchdev_obj *obj)
 664 {
 665         struct switchdev_fdb_dump *dump =
 666                 container_of(obj, struct switchdev_fdb_dump, obj);
 667         u32 portid = NETLINK_CB(dump->cb->skb).portid;
 668         u32 seq = dump->cb->nlh->nlmsg_seq;
 669         struct nlmsghdr *nlh;
 670         struct ndmsg *ndm;
 671
 672         if (dump->idx < dump->cb->args[0])
 673                 goto skip;
 674
 675         nlh = nlmsg_put(dump->skb, portid, seq, RTM_NEWNEIGH,
 676                         sizeof(*ndm), NLM_F_MULTI);
 677         if (!nlh)
 678                 return -EMSGSIZE;
 679
 680         ndm = nlmsg_data(nlh);
 681         ndm->ndm_family  = AF_BRIDGE;
 682         ndm->ndm_pad1    = 0;
 683         ndm->ndm_pad2    = 0;
 684         ndm->ndm_flags   = NTF_SELF;
 685         ndm->ndm_type    = 0;
 686         ndm->ndm_ifindex = dev->ifindex;
 687         ndm->ndm_state   = NUD_REACHABLE;
 688
 689         if (nla_put(dump->skb, NDA_LLADDR, ETH_ALEN, obj->u.fdb.addr))
 690                 goto nla_put_failure;
 691
 692         if (obj->u.fdb.vid && nla_put_u16(dump->skb, NDA_VLAN, obj->u.fdb.vid))
 693                 goto nla_put_failure;
 694
 695         nlmsg_end(dump->skb, nlh);
 696
 697 skip:
 698         dump->idx++;
 699         return 0;
 700
 701 nla_put_failure:
 702         nlmsg_cancel(dump->skb, nlh);
 703         return -EMSGSIZE;
 704 }
 705
 706 /**
 707  *      switchdev_port_fdb_dump - Dump port FDB (MAC/VLAN) entries
 708  *
 709  *      @skb: netlink skb
 710  *      @cb: netlink callback
 711  *      @dev: port device
 712  *      @filter_dev: filter device
 713  *      @idx:
 714  *
 715  *      Delete FDB entry from switch device.
 716  */
 717 int switchdev_port_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
 718                             struct net_device *dev,
 719                             struct net_device *filter_dev, int idx)
 720 {
 721         struct switchdev_fdb_dump dump = {
 722                 .obj = {
 723                         .id = SWITCHDEV_OBJ_PORT_FDB,
 724                         .cb = switchdev_port_fdb_dump_cb,
 725                 },
 726                 .skb = skb,
 727                 .cb = cb,
 728                 .idx = idx,
 729         };
 730         int err;
 731
 732         err = switchdev_port_obj_dump(dev, &dump.obj);
 733         if (err)
 734                 return err;
 735
 736         return dump.idx;
 737 }
 738 EXPORT_SYMBOL_GPL(switchdev_port_fdb_dump);
 739
 740 static struct net_device *switchdev_get_lowest_dev(struct net_device *dev)
 741 {
 742         const struct switchdev_ops *ops = dev->switchdev_ops;
 743         struct net_device *lower_dev;
 744         struct net_device *port_dev;
 745         struct list_head *iter;
 746
 747         /* Recusively search down until we find a sw port dev.
 748          * (A sw port dev supports switchdev_port_attr_get).
 749          */
 750
 751         if (ops && ops->switchdev_port_attr_get)
 752                 return dev;
 753
 754         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 755                 port_dev = switchdev_get_lowest_dev(lower_dev);
 756                 if (port_dev)
 757                         return port_dev;
 758         }
 759
 760         return NULL;
 761 }
 762
 763 static struct net_device *switchdev_get_dev_by_nhs(struct fib_info *fi)
 764 {
 765         struct switchdev_attr attr = {
 766                 .id = SWITCHDEV_ATTR_PORT_PARENT_ID,
 767         };
 768         struct switchdev_attr prev_attr;
 769         struct net_device *dev = NULL;
 770         int nhsel;
 771
 772         /* For this route, all nexthop devs must be on the same switch. */
 773
 774         for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) {
 775                 const struct fib_nh *nh = &fi->fib_nh[nhsel];
 776
 777                 if (!nh->nh_dev)
 778                         return NULL;
 779
 780                 dev = switchdev_get_lowest_dev(nh->nh_dev);
 781                 if (!dev)
 782                         return NULL;
 783
 784                 if (switchdev_port_attr_get(dev, &attr))
 785                         return NULL;
 786
 787                 if (nhsel > 0) {
 788                         if (prev_attr.u.ppid.id_len != attr.u.ppid.id_len)
 789                                 return NULL;
 790                         if (memcmp(prev_attr.u.ppid.id, attr.u.ppid.id,
 791                                    attr.u.ppid.id_len))
 792                                 return NULL;
 793                 }
 794
 795                 prev_attr = attr;
 796         }
 797
 798         return dev;
 799 }
 800
 801 /**
 802  *      switchdev_fib_ipv4_add - Add/modify switch IPv4 route entry
 803  *
 804  *      @dst: route's IPv4 destination address
 805  *      @dst_len: destination address length (prefix length)
 806  *      @fi: route FIB info structure
 807  *      @tos: route TOS
 808  *      @type: route type
 809  *      @nlflags: netlink flags passed in (NLM_F_*)
 810  *      @tb_id: route table ID
 811  *
 812  *      Add/modify switch IPv4 route entry.
 813  */
 814 int switchdev_fib_ipv4_add(u32 dst, int dst_len, struct fib_info *fi,
 815                            u8 tos, u8 type, u32 nlflags, u32 tb_id)
 816 {
 817         struct switchdev_obj fib_obj = {
 818                 .id = SWITCHDEV_OBJ_IPV4_FIB,
 819                 .u.ipv4_fib = {
 820                         .dst = dst,
 821                         .dst_len = dst_len,
 822                         .fi = fi,
 823                         .tos = tos,
 824                         .type = type,
 825                         .nlflags = nlflags,
 826                         .tb_id = tb_id,
 827                 },
 828         };
 829         struct net_device *dev;
 830         int err = 0;
 831
 832         /* Don't offload route if using custom ip rules or if
 833          * IPv4 FIB offloading has been disabled completely.
 834          */
 835
 836 #ifdef CONFIG_IP_MULTIPLE_TABLES
 837         if (fi->fib_net->ipv4.fib_has_custom_rules)
 838                 return 0;
 839 #endif
 840
 841         if (fi->fib_net->ipv4.fib_offload_disabled)
 842                 return 0;
 843
 844         dev = switchdev_get_dev_by_nhs(fi);
 845         if (!dev)
 846                 return 0;
 847
 848         err = switchdev_port_obj_add(dev, &fib_obj);
 849         if (!err)
 850                 fi->fib_flags |= RTNH_F_OFFLOAD;
 851
 852         return err == -EOPNOTSUPP ? 0 : err;
 853 }
 854 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_add);
 855
 856 /**
 857  *      switchdev_fib_ipv4_del - Delete IPv4 route entry from switch
 858  *
 859  *      @dst: route's IPv4 destination address
 860  *      @dst_len: destination address length (prefix length)
 861  *      @fi: route FIB info structure
 862  *      @tos: route TOS
 863  *      @type: route type
 864  *      @tb_id: route table ID
 865  *
 866  *      Delete IPv4 route entry from switch device.
 867  */
 868 int switchdev_fib_ipv4_del(u32 dst, int dst_len, struct fib_info *fi,
 869                            u8 tos, u8 type, u32 tb_id)
 870 {
 871         struct switchdev_obj fib_obj = {
 872                 .id = SWITCHDEV_OBJ_IPV4_FIB,
 873                 .u.ipv4_fib = {
 874                         .dst = dst,
 875                         .dst_len = dst_len,
 876                         .fi = fi,
 877                         .tos = tos,
 878                         .type = type,
 879                         .nlflags = 0,
 880                         .tb_id = tb_id,
 881                 },
 882         };
 883         struct net_device *dev;
 884         int err = 0;
 885
 886         if (!(fi->fib_flags & RTNH_F_OFFLOAD))
 887                 return 0;
 888
 889         dev = switchdev_get_dev_by_nhs(fi);
 890         if (!dev)
 891                 return 0;
 892
 893         err = switchdev_port_obj_del(dev, &fib_obj);
 894         if (!err)
 895                 fi->fib_flags &= ~RTNH_F_OFFLOAD;
 896
 897         return err == -EOPNOTSUPP ? 0 : err;
 898 }
 899 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_del);
 900
 901 /**
 902  *      switchdev_fib_ipv4_abort - Abort an IPv4 FIB operation
 903  *
 904  *      @fi: route FIB info structure
 905  */
 906 void switchdev_fib_ipv4_abort(struct fib_info *fi)
 907 {
 908         /* There was a problem installing this route to the offload
 909          * device.  For now, until we come up with more refined
 910          * policy handling, abruptly end IPv4 fib offloading for
 911          * for entire net by flushing offload device(s) of all
 912          * IPv4 routes, and mark IPv4 fib offloading broken from
 913          * this point forward.
 914          */
 915
 916         fib_flush_external(fi->fib_net);
 917         fi->fib_net->ipv4.fib_offload_disabled = true;
 918 }
 919 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_abort);