roce_gid_mgmt.c 19.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39
/*
 * Copyright (c) 2015, Mellanox Technologies inc.  All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#include "core_priv.h"

#include <linux/in.h>
#include <linux/in6.h>

/* For in6_dev_get/in6_dev_put */
#include <net/addrconf.h>
40
#include <net/bonding.h>
41 42 43 44

#include <rdma/ib_cache.h>
#include <rdma/ib_addr.h>

45 46
static struct workqueue_struct *gid_cache_wq;

47 48
static struct workqueue_struct *gid_cache_wq;

49 50 51 52 53 54 55 56 57 58 59 60
enum gid_op_type {
	GID_DEL = 0,
	GID_ADD
};

struct update_gid_event_work {
	struct work_struct work;
	union ib_gid       gid;
	struct ib_gid_attr gid_attr;
	enum gid_op_type gid_op;
};

61
#define ROCE_NETDEV_CALLBACK_SZ		3
62 63 64
struct netdev_event_work_cmd {
	roce_netdev_callback	cb;
	roce_netdev_filter	filter;
65 66
	struct net_device	*ndev;
	struct net_device	*filter_ndev;
67 68 69 70 71 72 73
};

struct netdev_event_work {
	struct work_struct		work;
	struct netdev_event_work_cmd	cmds[ROCE_NETDEV_CALLBACK_SZ];
};

74 75 76 77
static const struct {
	bool (*is_supported)(const struct ib_device *device, u8 port_num);
	enum ib_gid_type gid_type;
} PORT_CAP_TO_GID_TYPE[] = {
78 79
	{rdma_protocol_roce_eth_encap, IB_GID_TYPE_ROCE},
	{rdma_protocol_roce_udp_encap, IB_GID_TYPE_ROCE_UDP_ENCAP},
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
};

#define CAP_TO_GID_TABLE_SIZE	ARRAY_SIZE(PORT_CAP_TO_GID_TYPE)

unsigned long roce_gid_type_mask_support(struct ib_device *ib_dev, u8 port)
{
	int i;
	unsigned int ret_flags = 0;

	if (!rdma_protocol_roce(ib_dev, port))
		return 1UL << IB_GID_TYPE_IB;

	for (i = 0; i < CAP_TO_GID_TABLE_SIZE; i++)
		if (PORT_CAP_TO_GID_TYPE[i].is_supported(ib_dev, port))
			ret_flags |= 1UL << PORT_CAP_TO_GID_TYPE[i].gid_type;

	return ret_flags;
}
EXPORT_SYMBOL(roce_gid_type_mask_support);

100 101 102 103
static void update_gid(enum gid_op_type gid_op, struct ib_device *ib_dev,
		       u8 port, union ib_gid *gid,
		       struct ib_gid_attr *gid_attr)
{
104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
	int i;
	unsigned long gid_type_mask = roce_gid_type_mask_support(ib_dev, port);

	for (i = 0; i < IB_GID_TYPE_SIZE; i++) {
		if ((1UL << i) & gid_type_mask) {
			gid_attr->gid_type = i;
			switch (gid_op) {
			case GID_ADD:
				ib_cache_gid_add(ib_dev, port,
						 gid, gid_attr);
				break;
			case GID_DEL:
				ib_cache_gid_del(ib_dev, port,
						 gid, gid_attr);
				break;
			}
		}
121 122 123
	}
}

124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
enum bonding_slave_state {
	BONDING_SLAVE_STATE_ACTIVE	= 1UL << 0,
	BONDING_SLAVE_STATE_INACTIVE	= 1UL << 1,
	/* No primary slave or the device isn't a slave in bonding */
	BONDING_SLAVE_STATE_NA		= 1UL << 2,
};

static enum bonding_slave_state is_eth_active_slave_of_bonding_rcu(struct net_device *dev,
								   struct net_device *upper)
{
	if (upper && netif_is_bond_master(upper)) {
		struct net_device *pdev =
			bond_option_active_slave_get_rcu(netdev_priv(upper));

		if (pdev)
			return dev == pdev ? BONDING_SLAVE_STATE_ACTIVE :
				BONDING_SLAVE_STATE_INACTIVE;
	}

	return BONDING_SLAVE_STATE_NA;
}

#define REQUIRED_BOND_STATES		(BONDING_SLAVE_STATE_ACTIVE |	\
					 BONDING_SLAVE_STATE_NA)
148 149 150 151
static int is_eth_port_of_netdev(struct ib_device *ib_dev, u8 port,
				 struct net_device *rdma_ndev, void *cookie)
{
	struct net_device *real_dev;
152 153 154 155 156 157
	int res;

	if (!rdma_ndev)
		return 0;

	rcu_read_lock();
158
	real_dev = rdma_vlan_dev_real_dev(cookie);
159
	if (!real_dev)
160
		real_dev = cookie;
161

162
	res = ((rdma_is_upper_dev_rcu(rdma_ndev, cookie) &&
163 164 165 166 167 168 169 170 171 172 173
	       (is_eth_active_slave_of_bonding_rcu(rdma_ndev, real_dev) &
		REQUIRED_BOND_STATES)) ||
	       real_dev == rdma_ndev);

	rcu_read_unlock();
	return res;
}

static int is_eth_port_inactive_slave(struct ib_device *ib_dev, u8 port,
				      struct net_device *rdma_ndev, void *cookie)
{
174 175 176 177 178 179 180 181
	struct net_device *master_dev;
	int res;

	if (!rdma_ndev)
		return 0;

	rcu_read_lock();
	master_dev = netdev_master_upper_dev_get_rcu(rdma_ndev);
182 183
	res = is_eth_active_slave_of_bonding_rcu(rdma_ndev, master_dev) ==
		BONDING_SLAVE_STATE_INACTIVE;
184 185 186 187 188 189 190 191 192 193 194
	rcu_read_unlock();

	return res;
}

static int pass_all_filter(struct ib_device *ib_dev, u8 port,
			   struct net_device *rdma_ndev, void *cookie)
{
	return 1;
}

195 196 197 198 199 200 201 202
static int upper_device_filter(struct ib_device *ib_dev, u8 port,
			       struct net_device *rdma_ndev, void *cookie)
{
	int res;

	if (!rdma_ndev)
		return 0;

203
	if (rdma_ndev == cookie)
204 205 206
		return 1;

	rcu_read_lock();
207
	res = rdma_is_upper_dev_rcu(rdma_ndev, cookie);
208 209 210 211 212
	rcu_read_unlock();

	return res;
}

213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
static void update_gid_ip(enum gid_op_type gid_op,
			  struct ib_device *ib_dev,
			  u8 port, struct net_device *ndev,
			  struct sockaddr *addr)
{
	union ib_gid gid;
	struct ib_gid_attr gid_attr;

	rdma_ip2gid(addr, &gid);
	memset(&gid_attr, 0, sizeof(gid_attr));
	gid_attr.ndev = ndev;

	update_gid(gid_op, ib_dev, port, &gid, &gid_attr);
}

static void enum_netdev_default_gids(struct ib_device *ib_dev,
				     u8 port, struct net_device *event_ndev,
				     struct net_device *rdma_ndev)
{
232 233
	unsigned long gid_type_mask;

234 235 236
	rcu_read_lock();
	if (!rdma_ndev ||
	    ((rdma_ndev != event_ndev &&
237
	      !rdma_is_upper_dev_rcu(rdma_ndev, event_ndev)) ||
238 239 240 241
	     is_eth_active_slave_of_bonding_rcu(rdma_ndev,
						netdev_master_upper_dev_get_rcu(rdma_ndev)) ==
	     BONDING_SLAVE_STATE_INACTIVE)) {
		rcu_read_unlock();
242
		return;
243 244
	}
	rcu_read_unlock();
245

246 247 248
	gid_type_mask = roce_gid_type_mask_support(ib_dev, port);

	ib_cache_gid_set_default_gid(ib_dev, port, rdma_ndev, gid_type_mask,
249 250 251
				     IB_CACHE_GID_DEFAULT_MODE_SET);
}

252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
static void bond_delete_netdev_default_gids(struct ib_device *ib_dev,
					    u8 port,
					    struct net_device *event_ndev,
					    struct net_device *rdma_ndev)
{
	struct net_device *real_dev = rdma_vlan_dev_real_dev(event_ndev);

	if (!rdma_ndev)
		return;

	if (!real_dev)
		real_dev = event_ndev;

	rcu_read_lock();

267
	if (rdma_is_upper_dev_rcu(rdma_ndev, event_ndev) &&
268 269
	    is_eth_active_slave_of_bonding_rcu(rdma_ndev, real_dev) ==
	    BONDING_SLAVE_STATE_INACTIVE) {
270 271
		unsigned long gid_type_mask;

272 273
		rcu_read_unlock();

274 275
		gid_type_mask = roce_gid_type_mask_support(ib_dev, port);

276
		ib_cache_gid_set_default_gid(ib_dev, port, rdma_ndev,
277
					     gid_type_mask,
278 279 280 281 282 283
					     IB_CACHE_GID_DEFAULT_MODE_DELETE);
	} else {
		rcu_read_unlock();
	}
}

284 285 286 287
static void enum_netdev_ipv4_ips(struct ib_device *ib_dev,
				 u8 port, struct net_device *ndev)
{
	struct in_device *in_dev;
M
Matan Barak 已提交
288 289 290 291 292 293 294 295
	struct sin_list {
		struct list_head	list;
		struct sockaddr_in	ip;
	};
	struct sin_list *sin_iter;
	struct sin_list *sin_temp;

	LIST_HEAD(sin_list);
296 297 298
	if (ndev->reg_state >= NETREG_UNREGISTERING)
		return;

M
Matan Barak 已提交
299 300 301 302
	rcu_read_lock();
	in_dev = __in_dev_get_rcu(ndev);
	if (!in_dev) {
		rcu_read_unlock();
303
		return;
M
Matan Barak 已提交
304
	}
305 306

	for_ifa(in_dev) {
M
Matan Barak 已提交
307 308
		struct sin_list *entry = kzalloc(sizeof(*entry), GFP_ATOMIC);

309
		if (!entry)
M
Matan Barak 已提交
310
			continue;
311

M
Matan Barak 已提交
312 313 314
		entry->ip.sin_family = AF_INET;
		entry->ip.sin_addr.s_addr = ifa->ifa_address;
		list_add_tail(&entry->list, &sin_list);
315 316
	}
	endfor_ifa(in_dev);
M
Matan Barak 已提交
317
	rcu_read_unlock();
318

M
Matan Barak 已提交
319 320 321 322 323 324
	list_for_each_entry_safe(sin_iter, sin_temp, &sin_list, list) {
		update_gid_ip(GID_ADD, ib_dev, port, ndev,
			      (struct sockaddr *)&sin_iter->ip);
		list_del(&sin_iter->list);
		kfree(sin_iter);
	}
325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
}

static void enum_netdev_ipv6_ips(struct ib_device *ib_dev,
				 u8 port, struct net_device *ndev)
{
	struct inet6_ifaddr *ifp;
	struct inet6_dev *in6_dev;
	struct sin6_list {
		struct list_head	list;
		struct sockaddr_in6	sin6;
	};
	struct sin6_list *sin6_iter;
	struct sin6_list *sin6_temp;
	struct ib_gid_attr gid_attr = {.ndev = ndev};
	LIST_HEAD(sin6_list);

	if (ndev->reg_state >= NETREG_UNREGISTERING)
		return;

	in6_dev = in6_dev_get(ndev);
	if (!in6_dev)
		return;

	read_lock_bh(&in6_dev->lock);
	list_for_each_entry(ifp, &in6_dev->addr_list, if_list) {
		struct sin6_list *entry = kzalloc(sizeof(*entry), GFP_ATOMIC);

352
		if (!entry)
353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372
			continue;

		entry->sin6.sin6_family = AF_INET6;
		entry->sin6.sin6_addr = ifp->addr;
		list_add_tail(&entry->list, &sin6_list);
	}
	read_unlock_bh(&in6_dev->lock);

	in6_dev_put(in6_dev);

	list_for_each_entry_safe(sin6_iter, sin6_temp, &sin6_list, list) {
		union ib_gid	gid;

		rdma_ip2gid((struct sockaddr *)&sin6_iter->sin6, &gid);
		update_gid(GID_ADD, ib_dev, port, &gid, &gid_attr);
		list_del(&sin6_iter->list);
		kfree(sin6_iter);
	}
}

373 374 375 376 377 378 379 380
static void _add_netdev_ips(struct ib_device *ib_dev, u8 port,
			    struct net_device *ndev)
{
	enum_netdev_ipv4_ips(ib_dev, port, ndev);
	if (IS_ENABLED(CONFIG_IPV6))
		enum_netdev_ipv6_ips(ib_dev, port, ndev);
}

381 382 383
static void add_netdev_ips(struct ib_device *ib_dev, u8 port,
			   struct net_device *rdma_ndev, void *cookie)
{
384 385
	enum_netdev_default_gids(ib_dev, port, cookie, rdma_ndev);
	_add_netdev_ips(ib_dev, port, cookie);
386 387 388 389 390
}

static void del_netdev_ips(struct ib_device *ib_dev, u8 port,
			   struct net_device *rdma_ndev, void *cookie)
{
391
	ib_cache_gid_del_all_netdev_gids(ib_dev, port, cookie);
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412
}

static void enum_all_gids_of_dev_cb(struct ib_device *ib_dev,
				    u8 port,
				    struct net_device *rdma_ndev,
				    void *cookie)
{
	struct net *net;
	struct net_device *ndev;

	/* Lock the rtnl to make sure the netdevs does not move under
	 * our feet
	 */
	rtnl_lock();
	for_each_net(net)
		for_each_netdev(net, ndev)
			if (is_eth_port_of_netdev(ib_dev, port, rdma_ndev, ndev))
				add_netdev_ips(ib_dev, port, rdma_ndev, ndev);
	rtnl_unlock();
}

413 414 415 416 417 418 419
/**
 * rdma_roce_rescan_device - Rescan all of the network devices in the system
 * and add their gids, as needed, to the relevant RoCE devices.
 *
 * @device:         the rdma device
 */
void rdma_roce_rescan_device(struct ib_device *ib_dev)
420 421 422 423
{
	ib_enum_roce_netdev(ib_dev, pass_all_filter, NULL,
			    enum_all_gids_of_dev_cb, NULL);
}
424
EXPORT_SYMBOL(rdma_roce_rescan_device);
425 426 427 428 429 430 431 432 433 434 435 436 437

static void callback_for_addr_gid_device_scan(struct ib_device *device,
					      u8 port,
					      struct net_device *rdma_ndev,
					      void *cookie)
{
	struct update_gid_event_work *parsed = cookie;

	return update_gid(parsed->gid_op, device,
			  port, &parsed->gid,
			  &parsed->gid_attr);
}

438 439 440 441 442 443 444 445 446 447
struct upper_list {
	struct list_head list;
	struct net_device *upper;
};

static int netdev_upper_walk(struct net_device *upper, void *data)
{
	struct upper_list *entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
	struct list_head *upper_list = data;

448
	if (!entry)
449 450 451 452 453 454 455 456 457
		return 0;

	list_add_tail(&entry->list, upper_list);
	dev_hold(upper);
	entry->upper = upper;

	return 0;
}

458 459 460 461 462 463
static void handle_netdev_upper(struct ib_device *ib_dev, u8 port,
				void *cookie,
				void (*handle_netdev)(struct ib_device *ib_dev,
						      u8 port,
						      struct net_device *ndev))
{
464
	struct net_device *ndev = cookie;
465 466 467 468 469
	struct upper_list *upper_iter;
	struct upper_list *upper_temp;
	LIST_HEAD(upper_list);

	rcu_read_lock();
470
	netdev_walk_all_upper_dev_rcu(ndev, netdev_upper_walk, &upper_list);
471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
	rcu_read_unlock();

	handle_netdev(ib_dev, port, ndev);
	list_for_each_entry_safe(upper_iter, upper_temp, &upper_list,
				 list) {
		handle_netdev(ib_dev, port, upper_iter->upper);
		dev_put(upper_iter->upper);
		list_del(&upper_iter->list);
		kfree(upper_iter);
	}
}

static void _roce_del_all_netdev_gids(struct ib_device *ib_dev, u8 port,
				      struct net_device *event_ndev)
{
	ib_cache_gid_del_all_netdev_gids(ib_dev, port, event_ndev);
}

static void del_netdev_upper_ips(struct ib_device *ib_dev, u8 port,
				 struct net_device *rdma_ndev, void *cookie)
{
	handle_netdev_upper(ib_dev, port, cookie, _roce_del_all_netdev_gids);
}

static void add_netdev_upper_ips(struct ib_device *ib_dev, u8 port,
				 struct net_device *rdma_ndev, void *cookie)
{
	handle_netdev_upper(ib_dev, port, cookie, _add_netdev_ips);
}

static void del_netdev_default_ips_join(struct ib_device *ib_dev, u8 port,
					struct net_device *rdma_ndev,
					void *cookie)
{
	struct net_device *master_ndev;

	rcu_read_lock();
	master_ndev = netdev_master_upper_dev_get_rcu(rdma_ndev);
	if (master_ndev)
		dev_hold(master_ndev);
	rcu_read_unlock();

	if (master_ndev) {
		bond_delete_netdev_default_gids(ib_dev, port, master_ndev,
						rdma_ndev);
		dev_put(master_ndev);
	}
}

static void del_netdev_default_ips(struct ib_device *ib_dev, u8 port,
				   struct net_device *rdma_ndev, void *cookie)
{
523
	bond_delete_netdev_default_gids(ib_dev, port, cookie, rdma_ndev);
524 525
}

526 527 528 529 530 531 532 533 534 535 536
/* The following functions operate on all IB devices. netdevice_event and
 * addr_event execute ib_enum_all_roce_netdevs through a work.
 * ib_enum_all_roce_netdevs iterates through all IB devices.
 */

static void netdevice_event_work_handler(struct work_struct *_work)
{
	struct netdev_event_work *work =
		container_of(_work, struct netdev_event_work, work);
	unsigned int i;

537 538 539 540 541 542 543 544
	for (i = 0; i < ARRAY_SIZE(work->cmds) && work->cmds[i].cb; i++) {
		ib_enum_all_roce_netdevs(work->cmds[i].filter,
					 work->cmds[i].filter_ndev,
					 work->cmds[i].cb,
					 work->cmds[i].ndev);
		dev_put(work->cmds[i].ndev);
		dev_put(work->cmds[i].filter_ndev);
	}
545 546 547 548 549 550 551

	kfree(work);
}

static int netdevice_queue_work(struct netdev_event_work_cmd *cmds,
				struct net_device *ndev)
{
552
	unsigned int i;
553 554 555
	struct netdev_event_work *ndev_work =
		kmalloc(sizeof(*ndev_work), GFP_KERNEL);

556
	if (!ndev_work)
557 558 559
		return NOTIFY_DONE;

	memcpy(ndev_work->cmds, cmds, sizeof(ndev_work->cmds));
560 561 562 563 564 565 566 567
	for (i = 0; i < ARRAY_SIZE(ndev_work->cmds) && ndev_work->cmds[i].cb; i++) {
		if (!ndev_work->cmds[i].ndev)
			ndev_work->cmds[i].ndev = ndev;
		if (!ndev_work->cmds[i].filter_ndev)
			ndev_work->cmds[i].filter_ndev = ndev;
		dev_hold(ndev_work->cmds[i].ndev);
		dev_hold(ndev_work->cmds[i].filter_ndev);
	}
568 569
	INIT_WORK(&ndev_work->work, netdevice_event_work_handler);

570
	queue_work(gid_cache_wq, &ndev_work->work);
571 572 573 574

	return NOTIFY_DONE;
}

575 576 577 578 579
static const struct netdev_event_work_cmd add_cmd = {
	.cb = add_netdev_ips, .filter = is_eth_port_of_netdev};
static const struct netdev_event_work_cmd add_cmd_upper_ips = {
	.cb = add_netdev_upper_ips, .filter = is_eth_port_of_netdev};

580
static void netdevice_event_changeupper(struct netdev_notifier_changeupper_info *changeupper_info,
581 582 583 584 585 586 587
					struct netdev_event_work_cmd *cmds)
{
	static const struct netdev_event_work_cmd upper_ips_del_cmd = {
		.cb = del_netdev_upper_ips, .filter = upper_device_filter};
	static const struct netdev_event_work_cmd bonding_default_del_cmd = {
		.cb = del_netdev_default_ips, .filter = is_eth_port_inactive_slave};

588
	if (changeupper_info->linking == false) {
589
		cmds[0] = upper_ips_del_cmd;
590
		cmds[0].ndev = changeupper_info->upper_dev;
591
		cmds[1] = add_cmd;
592
	} else {
593
		cmds[0] = bonding_default_del_cmd;
594
		cmds[0].ndev = changeupper_info->upper_dev;
595
		cmds[1] = add_cmd_upper_ips;
596 597
		cmds[1].ndev = changeupper_info->upper_dev;
		cmds[1].filter_ndev = changeupper_info->upper_dev;
598 599 600
	}
}

601 602 603 604 605
static int netdevice_event(struct notifier_block *this, unsigned long event,
			   void *ptr)
{
	static const struct netdev_event_work_cmd del_cmd = {
		.cb = del_netdev_ips, .filter = pass_all_filter};
606 607 608 609 610 611
	static const struct netdev_event_work_cmd bonding_default_del_cmd_join = {
		.cb = del_netdev_default_ips_join, .filter = is_eth_port_inactive_slave};
	static const struct netdev_event_work_cmd default_del_cmd = {
		.cb = del_netdev_default_ips, .filter = pass_all_filter};
	static const struct netdev_event_work_cmd bonding_event_ips_del_cmd = {
		.cb = del_netdev_upper_ips, .filter = upper_device_filter};
612 613 614 615 616 617 618 619 620
	struct net_device *ndev = netdev_notifier_info_to_dev(ptr);
	struct netdev_event_work_cmd cmds[ROCE_NETDEV_CALLBACK_SZ] = { {NULL} };

	if (ndev->type != ARPHRD_ETHER)
		return NOTIFY_DONE;

	switch (event) {
	case NETDEV_REGISTER:
	case NETDEV_UP:
621 622
		cmds[0] = bonding_default_del_cmd_join;
		cmds[1] = add_cmd;
623 624 625 626 627 628 629 630 631 632
		break;

	case NETDEV_UNREGISTER:
		if (ndev->reg_state < NETREG_UNREGISTERED)
			cmds[0] = del_cmd;
		else
			return NOTIFY_DONE;
		break;

	case NETDEV_CHANGEADDR:
633
		cmds[0] = default_del_cmd;
634 635
		cmds[1] = add_cmd;
		break;
636 637 638

	case NETDEV_CHANGEUPPER:
		netdevice_event_changeupper(
639
			container_of(ptr, struct netdev_notifier_changeupper_info, info),
640 641 642 643 644 645 646 647 648
			cmds);
		break;

	case NETDEV_BONDING_FAILOVER:
		cmds[0] = bonding_event_ips_del_cmd;
		cmds[1] = bonding_default_del_cmd_join;
		cmds[2] = add_cmd_upper_ips;
		break;

649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690
	default:
		return NOTIFY_DONE;
	}

	return netdevice_queue_work(cmds, ndev);
}

static void update_gid_event_work_handler(struct work_struct *_work)
{
	struct update_gid_event_work *work =
		container_of(_work, struct update_gid_event_work, work);

	ib_enum_all_roce_netdevs(is_eth_port_of_netdev, work->gid_attr.ndev,
				 callback_for_addr_gid_device_scan, work);

	dev_put(work->gid_attr.ndev);
	kfree(work);
}

static int addr_event(struct notifier_block *this, unsigned long event,
		      struct sockaddr *sa, struct net_device *ndev)
{
	struct update_gid_event_work *work;
	enum gid_op_type gid_op;

	if (ndev->type != ARPHRD_ETHER)
		return NOTIFY_DONE;

	switch (event) {
	case NETDEV_UP:
		gid_op = GID_ADD;
		break;

	case NETDEV_DOWN:
		gid_op = GID_DEL;
		break;

	default:
		return NOTIFY_DONE;
	}

	work = kmalloc(sizeof(*work), GFP_ATOMIC);
691
	if (!work)
692 693 694 695 696 697 698 699 700 701 702
		return NOTIFY_DONE;

	INIT_WORK(&work->work, update_gid_event_work_handler);

	rdma_ip2gid(sa, &work->gid);
	work->gid_op = gid_op;

	memset(&work->gid_attr, 0, sizeof(work->gid_attr));
	dev_hold(ndev);
	work->gid_attr.ndev   = ndev;

703
	queue_work(gid_cache_wq, &work->work);
704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749

	return NOTIFY_DONE;
}

static int inetaddr_event(struct notifier_block *this, unsigned long event,
			  void *ptr)
{
	struct sockaddr_in	in;
	struct net_device	*ndev;
	struct in_ifaddr	*ifa = ptr;

	in.sin_family = AF_INET;
	in.sin_addr.s_addr = ifa->ifa_address;
	ndev = ifa->ifa_dev->dev;

	return addr_event(this, event, (struct sockaddr *)&in, ndev);
}

static int inet6addr_event(struct notifier_block *this, unsigned long event,
			   void *ptr)
{
	struct sockaddr_in6	in6;
	struct net_device	*ndev;
	struct inet6_ifaddr	*ifa6 = ptr;

	in6.sin6_family = AF_INET6;
	in6.sin6_addr = ifa6->addr;
	ndev = ifa6->idev->dev;

	return addr_event(this, event, (struct sockaddr *)&in6, ndev);
}

static struct notifier_block nb_netdevice = {
	.notifier_call = netdevice_event
};

static struct notifier_block nb_inetaddr = {
	.notifier_call = inetaddr_event
};

static struct notifier_block nb_inet6addr = {
	.notifier_call = inet6addr_event
};

int __init roce_gid_mgmt_init(void)
{
750 751 752 753
	gid_cache_wq = alloc_ordered_workqueue("gid-cache-wq", 0);
	if (!gid_cache_wq)
		return -ENOMEM;

754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777
	register_inetaddr_notifier(&nb_inetaddr);
	if (IS_ENABLED(CONFIG_IPV6))
		register_inet6addr_notifier(&nb_inet6addr);
	/* We relay on the netdevice notifier to enumerate all
	 * existing devices in the system. Register to this notifier
	 * last to make sure we will not miss any IP add/del
	 * callbacks.
	 */
	register_netdevice_notifier(&nb_netdevice);

	return 0;
}

void __exit roce_gid_mgmt_cleanup(void)
{
	if (IS_ENABLED(CONFIG_IPV6))
		unregister_inet6addr_notifier(&nb_inet6addr);
	unregister_inetaddr_notifier(&nb_inetaddr);
	unregister_netdevice_notifier(&nb_netdevice);
	/* Ensure all gid deletion tasks complete before we go down,
	 * to avoid any reference to free'd memory. By the time
	 * ib-core is removed, all physical devices have been removed,
	 * so no issue with remaining hardware contexts.
	 */
778
	destroy_workqueue(gid_cache_wq);
779
}