xref: /kernel/linux/linux-6.6/net/bridge/br_mrp.c (revision 62306a36)
1// SPDX-License-Identifier: GPL-2.0-or-later
2
3#include <linux/mrp_bridge.h>
4#include "br_private_mrp.h"
5
6static const u8 mrp_test_dmac[ETH_ALEN] = { 0x1, 0x15, 0x4e, 0x0, 0x0, 0x1 };
7static const u8 mrp_in_test_dmac[ETH_ALEN] = { 0x1, 0x15, 0x4e, 0x0, 0x0, 0x3 };
8
9static int br_mrp_process(struct net_bridge_port *p, struct sk_buff *skb);
10
11static struct br_frame_type mrp_frame_type __read_mostly = {
12	.type = cpu_to_be16(ETH_P_MRP),
13	.frame_handler = br_mrp_process,
14};
15
16static bool br_mrp_is_ring_port(struct net_bridge_port *p_port,
17				struct net_bridge_port *s_port,
18				struct net_bridge_port *port)
19{
20	if (port == p_port ||
21	    port == s_port)
22		return true;
23
24	return false;
25}
26
27static bool br_mrp_is_in_port(struct net_bridge_port *i_port,
28			      struct net_bridge_port *port)
29{
30	if (port == i_port)
31		return true;
32
33	return false;
34}
35
36static struct net_bridge_port *br_mrp_get_port(struct net_bridge *br,
37					       u32 ifindex)
38{
39	struct net_bridge_port *res = NULL;
40	struct net_bridge_port *port;
41
42	list_for_each_entry(port, &br->port_list, list) {
43		if (port->dev->ifindex == ifindex) {
44			res = port;
45			break;
46		}
47	}
48
49	return res;
50}
51
52static struct br_mrp *br_mrp_find_id(struct net_bridge *br, u32 ring_id)
53{
54	struct br_mrp *res = NULL;
55	struct br_mrp *mrp;
56
57	hlist_for_each_entry_rcu(mrp, &br->mrp_list, list,
58				 lockdep_rtnl_is_held()) {
59		if (mrp->ring_id == ring_id) {
60			res = mrp;
61			break;
62		}
63	}
64
65	return res;
66}
67
68static struct br_mrp *br_mrp_find_in_id(struct net_bridge *br, u32 in_id)
69{
70	struct br_mrp *res = NULL;
71	struct br_mrp *mrp;
72
73	hlist_for_each_entry_rcu(mrp, &br->mrp_list, list,
74				 lockdep_rtnl_is_held()) {
75		if (mrp->in_id == in_id) {
76			res = mrp;
77			break;
78		}
79	}
80
81	return res;
82}
83
84static bool br_mrp_unique_ifindex(struct net_bridge *br, u32 ifindex)
85{
86	struct br_mrp *mrp;
87
88	hlist_for_each_entry_rcu(mrp, &br->mrp_list, list,
89				 lockdep_rtnl_is_held()) {
90		struct net_bridge_port *p;
91
92		p = rtnl_dereference(mrp->p_port);
93		if (p && p->dev->ifindex == ifindex)
94			return false;
95
96		p = rtnl_dereference(mrp->s_port);
97		if (p && p->dev->ifindex == ifindex)
98			return false;
99
100		p = rtnl_dereference(mrp->i_port);
101		if (p && p->dev->ifindex == ifindex)
102			return false;
103	}
104
105	return true;
106}
107
108static struct br_mrp *br_mrp_find_port(struct net_bridge *br,
109				       struct net_bridge_port *p)
110{
111	struct br_mrp *res = NULL;
112	struct br_mrp *mrp;
113
114	hlist_for_each_entry_rcu(mrp, &br->mrp_list, list,
115				 lockdep_rtnl_is_held()) {
116		if (rcu_access_pointer(mrp->p_port) == p ||
117		    rcu_access_pointer(mrp->s_port) == p ||
118		    rcu_access_pointer(mrp->i_port) == p) {
119			res = mrp;
120			break;
121		}
122	}
123
124	return res;
125}
126
127static int br_mrp_next_seq(struct br_mrp *mrp)
128{
129	mrp->seq_id++;
130	return mrp->seq_id;
131}
132
133static struct sk_buff *br_mrp_skb_alloc(struct net_bridge_port *p,
134					const u8 *src, const u8 *dst)
135{
136	struct ethhdr *eth_hdr;
137	struct sk_buff *skb;
138	__be16 *version;
139
140	skb = dev_alloc_skb(MRP_MAX_FRAME_LENGTH);
141	if (!skb)
142		return NULL;
143
144	skb->dev = p->dev;
145	skb->protocol = htons(ETH_P_MRP);
146	skb->priority = MRP_FRAME_PRIO;
147	skb_reserve(skb, sizeof(*eth_hdr));
148
149	eth_hdr = skb_push(skb, sizeof(*eth_hdr));
150	ether_addr_copy(eth_hdr->h_dest, dst);
151	ether_addr_copy(eth_hdr->h_source, src);
152	eth_hdr->h_proto = htons(ETH_P_MRP);
153
154	version = skb_put(skb, sizeof(*version));
155	*version = cpu_to_be16(MRP_VERSION);
156
157	return skb;
158}
159
160static void br_mrp_skb_tlv(struct sk_buff *skb,
161			   enum br_mrp_tlv_header_type type,
162			   u8 length)
163{
164	struct br_mrp_tlv_hdr *hdr;
165
166	hdr = skb_put(skb, sizeof(*hdr));
167	hdr->type = type;
168	hdr->length = length;
169}
170
171static void br_mrp_skb_common(struct sk_buff *skb, struct br_mrp *mrp)
172{
173	struct br_mrp_common_hdr *hdr;
174
175	br_mrp_skb_tlv(skb, BR_MRP_TLV_HEADER_COMMON, sizeof(*hdr));
176
177	hdr = skb_put(skb, sizeof(*hdr));
178	hdr->seq_id = cpu_to_be16(br_mrp_next_seq(mrp));
179	memset(hdr->domain, 0xff, MRP_DOMAIN_UUID_LENGTH);
180}
181
182static struct sk_buff *br_mrp_alloc_test_skb(struct br_mrp *mrp,
183					     struct net_bridge_port *p,
184					     enum br_mrp_port_role_type port_role)
185{
186	struct br_mrp_ring_test_hdr *hdr = NULL;
187	struct sk_buff *skb = NULL;
188
189	if (!p)
190		return NULL;
191
192	skb = br_mrp_skb_alloc(p, p->dev->dev_addr, mrp_test_dmac);
193	if (!skb)
194		return NULL;
195
196	br_mrp_skb_tlv(skb, BR_MRP_TLV_HEADER_RING_TEST, sizeof(*hdr));
197	hdr = skb_put(skb, sizeof(*hdr));
198
199	hdr->prio = cpu_to_be16(mrp->prio);
200	ether_addr_copy(hdr->sa, p->br->dev->dev_addr);
201	hdr->port_role = cpu_to_be16(port_role);
202	hdr->state = cpu_to_be16(mrp->ring_state);
203	hdr->transitions = cpu_to_be16(mrp->ring_transitions);
204	hdr->timestamp = cpu_to_be32(jiffies_to_msecs(jiffies));
205
206	br_mrp_skb_common(skb, mrp);
207
208	/* In case the node behaves as MRA then the Test frame needs to have
209	 * an Option TLV which includes eventually a sub-option TLV that has
210	 * the type AUTO_MGR
211	 */
212	if (mrp->ring_role == BR_MRP_RING_ROLE_MRA) {
213		struct br_mrp_sub_option1_hdr *sub_opt = NULL;
214		struct br_mrp_tlv_hdr *sub_tlv = NULL;
215		struct br_mrp_oui_hdr *oui = NULL;
216		u8 length;
217
218		length = sizeof(*sub_opt) + sizeof(*sub_tlv) + sizeof(oui) +
219			MRP_OPT_PADDING;
220		br_mrp_skb_tlv(skb, BR_MRP_TLV_HEADER_OPTION, length);
221
222		oui = skb_put(skb, sizeof(*oui));
223		memset(oui, 0x0, sizeof(*oui));
224		sub_opt = skb_put(skb, sizeof(*sub_opt));
225		memset(sub_opt, 0x0, sizeof(*sub_opt));
226
227		sub_tlv = skb_put(skb, sizeof(*sub_tlv));
228		sub_tlv->type = BR_MRP_SUB_TLV_HEADER_TEST_AUTO_MGR;
229
230		/* 32 bit alligment shall be ensured therefore add 2 bytes */
231		skb_put(skb, MRP_OPT_PADDING);
232	}
233
234	br_mrp_skb_tlv(skb, BR_MRP_TLV_HEADER_END, 0x0);
235
236	return skb;
237}
238
239static struct sk_buff *br_mrp_alloc_in_test_skb(struct br_mrp *mrp,
240						struct net_bridge_port *p,
241						enum br_mrp_port_role_type port_role)
242{
243	struct br_mrp_in_test_hdr *hdr = NULL;
244	struct sk_buff *skb = NULL;
245
246	if (!p)
247		return NULL;
248
249	skb = br_mrp_skb_alloc(p, p->dev->dev_addr, mrp_in_test_dmac);
250	if (!skb)
251		return NULL;
252
253	br_mrp_skb_tlv(skb, BR_MRP_TLV_HEADER_IN_TEST, sizeof(*hdr));
254	hdr = skb_put(skb, sizeof(*hdr));
255
256	hdr->id = cpu_to_be16(mrp->in_id);
257	ether_addr_copy(hdr->sa, p->br->dev->dev_addr);
258	hdr->port_role = cpu_to_be16(port_role);
259	hdr->state = cpu_to_be16(mrp->in_state);
260	hdr->transitions = cpu_to_be16(mrp->in_transitions);
261	hdr->timestamp = cpu_to_be32(jiffies_to_msecs(jiffies));
262
263	br_mrp_skb_common(skb, mrp);
264	br_mrp_skb_tlv(skb, BR_MRP_TLV_HEADER_END, 0x0);
265
266	return skb;
267}
268
269/* This function is continuously called in the following cases:
270 * - when node role is MRM, in this case test_monitor is always set to false
271 *   because it needs to notify the userspace that the ring is open and needs to
272 *   send MRP_Test frames
273 * - when node role is MRA, there are 2 subcases:
274 *     - when MRA behaves as MRM, in this case is similar with MRM role
275 *     - when MRA behaves as MRC, in this case test_monitor is set to true,
276 *       because it needs to detect when it stops seeing MRP_Test frames
277 *       from MRM node but it doesn't need to send MRP_Test frames.
278 */
279static void br_mrp_test_work_expired(struct work_struct *work)
280{
281	struct delayed_work *del_work = to_delayed_work(work);
282	struct br_mrp *mrp = container_of(del_work, struct br_mrp, test_work);
283	struct net_bridge_port *p;
284	bool notify_open = false;
285	struct sk_buff *skb;
286
287	if (time_before_eq(mrp->test_end, jiffies))
288		return;
289
290	if (mrp->test_count_miss < mrp->test_max_miss) {
291		mrp->test_count_miss++;
292	} else {
293		/* Notify that the ring is open only if the ring state is
294		 * closed, otherwise it would continue to notify at every
295		 * interval.
296		 * Also notify that the ring is open when the node has the
297		 * role MRA and behaves as MRC. The reason is that the
298		 * userspace needs to know when the MRM stopped sending
299		 * MRP_Test frames so that the current node to try to take
300		 * the role of a MRM.
301		 */
302		if (mrp->ring_state == BR_MRP_RING_STATE_CLOSED ||
303		    mrp->test_monitor)
304			notify_open = true;
305	}
306
307	rcu_read_lock();
308
309	p = rcu_dereference(mrp->p_port);
310	if (p) {
311		if (!mrp->test_monitor) {
312			skb = br_mrp_alloc_test_skb(mrp, p,
313						    BR_MRP_PORT_ROLE_PRIMARY);
314			if (!skb)
315				goto out;
316
317			skb_reset_network_header(skb);
318			dev_queue_xmit(skb);
319		}
320
321		if (notify_open && !mrp->ring_role_offloaded)
322			br_mrp_ring_port_open(p->dev, true);
323	}
324
325	p = rcu_dereference(mrp->s_port);
326	if (p) {
327		if (!mrp->test_monitor) {
328			skb = br_mrp_alloc_test_skb(mrp, p,
329						    BR_MRP_PORT_ROLE_SECONDARY);
330			if (!skb)
331				goto out;
332
333			skb_reset_network_header(skb);
334			dev_queue_xmit(skb);
335		}
336
337		if (notify_open && !mrp->ring_role_offloaded)
338			br_mrp_ring_port_open(p->dev, true);
339	}
340
341out:
342	rcu_read_unlock();
343
344	queue_delayed_work(system_wq, &mrp->test_work,
345			   usecs_to_jiffies(mrp->test_interval));
346}
347
348/* This function is continuously called when the node has the interconnect role
349 * MIM. It would generate interconnect test frames and will send them on all 3
350 * ports. But will also check if it stop receiving interconnect test frames.
351 */
352static void br_mrp_in_test_work_expired(struct work_struct *work)
353{
354	struct delayed_work *del_work = to_delayed_work(work);
355	struct br_mrp *mrp = container_of(del_work, struct br_mrp, in_test_work);
356	struct net_bridge_port *p;
357	bool notify_open = false;
358	struct sk_buff *skb;
359
360	if (time_before_eq(mrp->in_test_end, jiffies))
361		return;
362
363	if (mrp->in_test_count_miss < mrp->in_test_max_miss) {
364		mrp->in_test_count_miss++;
365	} else {
366		/* Notify that the interconnect ring is open only if the
367		 * interconnect ring state is closed, otherwise it would
368		 * continue to notify at every interval.
369		 */
370		if (mrp->in_state == BR_MRP_IN_STATE_CLOSED)
371			notify_open = true;
372	}
373
374	rcu_read_lock();
375
376	p = rcu_dereference(mrp->p_port);
377	if (p) {
378		skb = br_mrp_alloc_in_test_skb(mrp, p,
379					       BR_MRP_PORT_ROLE_PRIMARY);
380		if (!skb)
381			goto out;
382
383		skb_reset_network_header(skb);
384		dev_queue_xmit(skb);
385
386		if (notify_open && !mrp->in_role_offloaded)
387			br_mrp_in_port_open(p->dev, true);
388	}
389
390	p = rcu_dereference(mrp->s_port);
391	if (p) {
392		skb = br_mrp_alloc_in_test_skb(mrp, p,
393					       BR_MRP_PORT_ROLE_SECONDARY);
394		if (!skb)
395			goto out;
396
397		skb_reset_network_header(skb);
398		dev_queue_xmit(skb);
399
400		if (notify_open && !mrp->in_role_offloaded)
401			br_mrp_in_port_open(p->dev, true);
402	}
403
404	p = rcu_dereference(mrp->i_port);
405	if (p) {
406		skb = br_mrp_alloc_in_test_skb(mrp, p,
407					       BR_MRP_PORT_ROLE_INTER);
408		if (!skb)
409			goto out;
410
411		skb_reset_network_header(skb);
412		dev_queue_xmit(skb);
413
414		if (notify_open && !mrp->in_role_offloaded)
415			br_mrp_in_port_open(p->dev, true);
416	}
417
418out:
419	rcu_read_unlock();
420
421	queue_delayed_work(system_wq, &mrp->in_test_work,
422			   usecs_to_jiffies(mrp->in_test_interval));
423}
424
425/* Deletes the MRP instance.
426 * note: called under rtnl_lock
427 */
428static void br_mrp_del_impl(struct net_bridge *br, struct br_mrp *mrp)
429{
430	struct net_bridge_port *p;
431	u8 state;
432
433	/* Stop sending MRP_Test frames */
434	cancel_delayed_work_sync(&mrp->test_work);
435	br_mrp_switchdev_send_ring_test(br, mrp, 0, 0, 0, 0);
436
437	/* Stop sending MRP_InTest frames if has an interconnect role */
438	cancel_delayed_work_sync(&mrp->in_test_work);
439	br_mrp_switchdev_send_in_test(br, mrp, 0, 0, 0);
440
441	/* Disable the roles */
442	br_mrp_switchdev_set_ring_role(br, mrp, BR_MRP_RING_ROLE_DISABLED);
443	p = rtnl_dereference(mrp->i_port);
444	if (p)
445		br_mrp_switchdev_set_in_role(br, mrp, mrp->in_id, mrp->ring_id,
446					     BR_MRP_IN_ROLE_DISABLED);
447
448	br_mrp_switchdev_del(br, mrp);
449
450	/* Reset the ports */
451	p = rtnl_dereference(mrp->p_port);
452	if (p) {
453		spin_lock_bh(&br->lock);
454		state = netif_running(br->dev) ?
455				BR_STATE_FORWARDING : BR_STATE_DISABLED;
456		p->state = state;
457		p->flags &= ~BR_MRP_AWARE;
458		spin_unlock_bh(&br->lock);
459		br_mrp_port_switchdev_set_state(p, state);
460		rcu_assign_pointer(mrp->p_port, NULL);
461	}
462
463	p = rtnl_dereference(mrp->s_port);
464	if (p) {
465		spin_lock_bh(&br->lock);
466		state = netif_running(br->dev) ?
467				BR_STATE_FORWARDING : BR_STATE_DISABLED;
468		p->state = state;
469		p->flags &= ~BR_MRP_AWARE;
470		spin_unlock_bh(&br->lock);
471		br_mrp_port_switchdev_set_state(p, state);
472		rcu_assign_pointer(mrp->s_port, NULL);
473	}
474
475	p = rtnl_dereference(mrp->i_port);
476	if (p) {
477		spin_lock_bh(&br->lock);
478		state = netif_running(br->dev) ?
479				BR_STATE_FORWARDING : BR_STATE_DISABLED;
480		p->state = state;
481		p->flags &= ~BR_MRP_AWARE;
482		spin_unlock_bh(&br->lock);
483		br_mrp_port_switchdev_set_state(p, state);
484		rcu_assign_pointer(mrp->i_port, NULL);
485	}
486
487	hlist_del_rcu(&mrp->list);
488	kfree_rcu(mrp, rcu);
489
490	if (hlist_empty(&br->mrp_list))
491		br_del_frame(br, &mrp_frame_type);
492}
493
494/* Adds a new MRP instance.
495 * note: called under rtnl_lock
496 */
497int br_mrp_add(struct net_bridge *br, struct br_mrp_instance *instance)
498{
499	struct net_bridge_port *p;
500	struct br_mrp *mrp;
501	int err;
502
503	/* If the ring exists, it is not possible to create another one with the
504	 * same ring_id
505	 */
506	mrp = br_mrp_find_id(br, instance->ring_id);
507	if (mrp)
508		return -EINVAL;
509
510	if (!br_mrp_get_port(br, instance->p_ifindex) ||
511	    !br_mrp_get_port(br, instance->s_ifindex))
512		return -EINVAL;
513
514	/* It is not possible to have the same port part of multiple rings */
515	if (!br_mrp_unique_ifindex(br, instance->p_ifindex) ||
516	    !br_mrp_unique_ifindex(br, instance->s_ifindex))
517		return -EINVAL;
518
519	mrp = kzalloc(sizeof(*mrp), GFP_KERNEL);
520	if (!mrp)
521		return -ENOMEM;
522
523	mrp->ring_id = instance->ring_id;
524	mrp->prio = instance->prio;
525
526	p = br_mrp_get_port(br, instance->p_ifindex);
527	spin_lock_bh(&br->lock);
528	p->state = BR_STATE_FORWARDING;
529	p->flags |= BR_MRP_AWARE;
530	spin_unlock_bh(&br->lock);
531	rcu_assign_pointer(mrp->p_port, p);
532
533	p = br_mrp_get_port(br, instance->s_ifindex);
534	spin_lock_bh(&br->lock);
535	p->state = BR_STATE_FORWARDING;
536	p->flags |= BR_MRP_AWARE;
537	spin_unlock_bh(&br->lock);
538	rcu_assign_pointer(mrp->s_port, p);
539
540	if (hlist_empty(&br->mrp_list))
541		br_add_frame(br, &mrp_frame_type);
542
543	INIT_DELAYED_WORK(&mrp->test_work, br_mrp_test_work_expired);
544	INIT_DELAYED_WORK(&mrp->in_test_work, br_mrp_in_test_work_expired);
545	hlist_add_tail_rcu(&mrp->list, &br->mrp_list);
546
547	err = br_mrp_switchdev_add(br, mrp);
548	if (err)
549		goto delete_mrp;
550
551	return 0;
552
553delete_mrp:
554	br_mrp_del_impl(br, mrp);
555
556	return err;
557}
558
559/* Deletes the MRP instance from which the port is part of
560 * note: called under rtnl_lock
561 */
562void br_mrp_port_del(struct net_bridge *br, struct net_bridge_port *p)
563{
564	struct br_mrp *mrp = br_mrp_find_port(br, p);
565
566	/* If the port is not part of a MRP instance just bail out */
567	if (!mrp)
568		return;
569
570	br_mrp_del_impl(br, mrp);
571}
572
573/* Deletes existing MRP instance based on ring_id
574 * note: called under rtnl_lock
575 */
576int br_mrp_del(struct net_bridge *br, struct br_mrp_instance *instance)
577{
578	struct br_mrp *mrp = br_mrp_find_id(br, instance->ring_id);
579
580	if (!mrp)
581		return -EINVAL;
582
583	br_mrp_del_impl(br, mrp);
584
585	return 0;
586}
587
588/* Set port state, port state can be forwarding, blocked or disabled
589 * note: already called with rtnl_lock
590 */
591int br_mrp_set_port_state(struct net_bridge_port *p,
592			  enum br_mrp_port_state_type state)
593{
594	u32 port_state;
595
596	if (!p || !(p->flags & BR_MRP_AWARE))
597		return -EINVAL;
598
599	spin_lock_bh(&p->br->lock);
600
601	if (state == BR_MRP_PORT_STATE_FORWARDING)
602		port_state = BR_STATE_FORWARDING;
603	else
604		port_state = BR_STATE_BLOCKING;
605
606	p->state = port_state;
607	spin_unlock_bh(&p->br->lock);
608
609	br_mrp_port_switchdev_set_state(p, port_state);
610
611	return 0;
612}
613
614/* Set port role, port role can be primary or secondary
615 * note: already called with rtnl_lock
616 */
617int br_mrp_set_port_role(struct net_bridge_port *p,
618			 enum br_mrp_port_role_type role)
619{
620	struct br_mrp *mrp;
621
622	if (!p || !(p->flags & BR_MRP_AWARE))
623		return -EINVAL;
624
625	mrp = br_mrp_find_port(p->br, p);
626
627	if (!mrp)
628		return -EINVAL;
629
630	switch (role) {
631	case BR_MRP_PORT_ROLE_PRIMARY:
632		rcu_assign_pointer(mrp->p_port, p);
633		break;
634	case BR_MRP_PORT_ROLE_SECONDARY:
635		rcu_assign_pointer(mrp->s_port, p);
636		break;
637	default:
638		return -EINVAL;
639	}
640
641	br_mrp_port_switchdev_set_role(p, role);
642
643	return 0;
644}
645
646/* Set ring state, ring state can be only Open or Closed
647 * note: already called with rtnl_lock
648 */
649int br_mrp_set_ring_state(struct net_bridge *br,
650			  struct br_mrp_ring_state *state)
651{
652	struct br_mrp *mrp = br_mrp_find_id(br, state->ring_id);
653
654	if (!mrp)
655		return -EINVAL;
656
657	if (mrp->ring_state != state->ring_state)
658		mrp->ring_transitions++;
659
660	mrp->ring_state = state->ring_state;
661
662	br_mrp_switchdev_set_ring_state(br, mrp, state->ring_state);
663
664	return 0;
665}
666
667/* Set ring role, ring role can be only MRM(Media Redundancy Manager) or
668 * MRC(Media Redundancy Client).
669 * note: already called with rtnl_lock
670 */
671int br_mrp_set_ring_role(struct net_bridge *br,
672			 struct br_mrp_ring_role *role)
673{
674	struct br_mrp *mrp = br_mrp_find_id(br, role->ring_id);
675	enum br_mrp_hw_support support;
676
677	if (!mrp)
678		return -EINVAL;
679
680	mrp->ring_role = role->ring_role;
681
682	/* If there is an error just bailed out */
683	support = br_mrp_switchdev_set_ring_role(br, mrp, role->ring_role);
684	if (support == BR_MRP_NONE)
685		return -EOPNOTSUPP;
686
687	/* Now detect if the HW actually applied the role or not. If the HW
688	 * applied the role it means that the SW will not to do those operations
689	 * anymore. For example if the role ir MRM then the HW will notify the
690	 * SW when ring is open, but if the is not pushed to the HW the SW will
691	 * need to detect when the ring is open
692	 */
693	mrp->ring_role_offloaded = support == BR_MRP_SW ? 0 : 1;
694
695	return 0;
696}
697
698/* Start to generate or monitor MRP test frames, the frames are generated by
699 * HW and if it fails, they are generated by the SW.
700 * note: already called with rtnl_lock
701 */
702int br_mrp_start_test(struct net_bridge *br,
703		      struct br_mrp_start_test *test)
704{
705	struct br_mrp *mrp = br_mrp_find_id(br, test->ring_id);
706	enum br_mrp_hw_support support;
707
708	if (!mrp)
709		return -EINVAL;
710
711	/* Try to push it to the HW and if it fails then continue with SW
712	 * implementation and if that also fails then return error.
713	 */
714	support = br_mrp_switchdev_send_ring_test(br, mrp, test->interval,
715						  test->max_miss, test->period,
716						  test->monitor);
717	if (support == BR_MRP_NONE)
718		return -EOPNOTSUPP;
719
720	if (support == BR_MRP_HW)
721		return 0;
722
723	mrp->test_interval = test->interval;
724	mrp->test_end = jiffies + usecs_to_jiffies(test->period);
725	mrp->test_max_miss = test->max_miss;
726	mrp->test_monitor = test->monitor;
727	mrp->test_count_miss = 0;
728	queue_delayed_work(system_wq, &mrp->test_work,
729			   usecs_to_jiffies(test->interval));
730
731	return 0;
732}
733
734/* Set in state, int state can be only Open or Closed
735 * note: already called with rtnl_lock
736 */
737int br_mrp_set_in_state(struct net_bridge *br, struct br_mrp_in_state *state)
738{
739	struct br_mrp *mrp = br_mrp_find_in_id(br, state->in_id);
740
741	if (!mrp)
742		return -EINVAL;
743
744	if (mrp->in_state != state->in_state)
745		mrp->in_transitions++;
746
747	mrp->in_state = state->in_state;
748
749	br_mrp_switchdev_set_in_state(br, mrp, state->in_state);
750
751	return 0;
752}
753
754/* Set in role, in role can be only MIM(Media Interconnection Manager) or
755 * MIC(Media Interconnection Client).
756 * note: already called with rtnl_lock
757 */
758int br_mrp_set_in_role(struct net_bridge *br, struct br_mrp_in_role *role)
759{
760	struct br_mrp *mrp = br_mrp_find_id(br, role->ring_id);
761	enum br_mrp_hw_support support;
762	struct net_bridge_port *p;
763
764	if (!mrp)
765		return -EINVAL;
766
767	if (!br_mrp_get_port(br, role->i_ifindex))
768		return -EINVAL;
769
770	if (role->in_role == BR_MRP_IN_ROLE_DISABLED) {
771		u8 state;
772
773		/* It is not allowed to disable a port that doesn't exist */
774		p = rtnl_dereference(mrp->i_port);
775		if (!p)
776			return -EINVAL;
777
778		/* Stop the generating MRP_InTest frames */
779		cancel_delayed_work_sync(&mrp->in_test_work);
780		br_mrp_switchdev_send_in_test(br, mrp, 0, 0, 0);
781
782		/* Remove the port */
783		spin_lock_bh(&br->lock);
784		state = netif_running(br->dev) ?
785				BR_STATE_FORWARDING : BR_STATE_DISABLED;
786		p->state = state;
787		p->flags &= ~BR_MRP_AWARE;
788		spin_unlock_bh(&br->lock);
789		br_mrp_port_switchdev_set_state(p, state);
790		rcu_assign_pointer(mrp->i_port, NULL);
791
792		mrp->in_role = role->in_role;
793		mrp->in_id = 0;
794
795		return 0;
796	}
797
798	/* It is not possible to have the same port part of multiple rings */
799	if (!br_mrp_unique_ifindex(br, role->i_ifindex))
800		return -EINVAL;
801
802	/* It is not allowed to set a different interconnect port if the mrp
803	 * instance has already one. First it needs to be disabled and after
804	 * that set the new port
805	 */
806	if (rcu_access_pointer(mrp->i_port))
807		return -EINVAL;
808
809	p = br_mrp_get_port(br, role->i_ifindex);
810	spin_lock_bh(&br->lock);
811	p->state = BR_STATE_FORWARDING;
812	p->flags |= BR_MRP_AWARE;
813	spin_unlock_bh(&br->lock);
814	rcu_assign_pointer(mrp->i_port, p);
815
816	mrp->in_role = role->in_role;
817	mrp->in_id = role->in_id;
818
819	/* If there is an error just bailed out */
820	support = br_mrp_switchdev_set_in_role(br, mrp, role->in_id,
821					       role->ring_id, role->in_role);
822	if (support == BR_MRP_NONE)
823		return -EOPNOTSUPP;
824
825	/* Now detect if the HW actually applied the role or not. If the HW
826	 * applied the role it means that the SW will not to do those operations
827	 * anymore. For example if the role is MIM then the HW will notify the
828	 * SW when interconnect ring is open, but if the is not pushed to the HW
829	 * the SW will need to detect when the interconnect ring is open.
830	 */
831	mrp->in_role_offloaded = support == BR_MRP_SW ? 0 : 1;
832
833	return 0;
834}
835
836/* Start to generate MRP_InTest frames, the frames are generated by
837 * HW and if it fails, they are generated by the SW.
838 * note: already called with rtnl_lock
839 */
840int br_mrp_start_in_test(struct net_bridge *br,
841			 struct br_mrp_start_in_test *in_test)
842{
843	struct br_mrp *mrp = br_mrp_find_in_id(br, in_test->in_id);
844	enum br_mrp_hw_support support;
845
846	if (!mrp)
847		return -EINVAL;
848
849	if (mrp->in_role != BR_MRP_IN_ROLE_MIM)
850		return -EINVAL;
851
852	/* Try to push it to the HW and if it fails then continue with SW
853	 * implementation and if that also fails then return error.
854	 */
855	support =  br_mrp_switchdev_send_in_test(br, mrp, in_test->interval,
856						 in_test->max_miss,
857						 in_test->period);
858	if (support == BR_MRP_NONE)
859		return -EOPNOTSUPP;
860
861	if (support == BR_MRP_HW)
862		return 0;
863
864	mrp->in_test_interval = in_test->interval;
865	mrp->in_test_end = jiffies + usecs_to_jiffies(in_test->period);
866	mrp->in_test_max_miss = in_test->max_miss;
867	mrp->in_test_count_miss = 0;
868	queue_delayed_work(system_wq, &mrp->in_test_work,
869			   usecs_to_jiffies(in_test->interval));
870
871	return 0;
872}
873
874/* Determine if the frame type is a ring frame */
875static bool br_mrp_ring_frame(struct sk_buff *skb)
876{
877	const struct br_mrp_tlv_hdr *hdr;
878	struct br_mrp_tlv_hdr _hdr;
879
880	hdr = skb_header_pointer(skb, sizeof(uint16_t), sizeof(_hdr), &_hdr);
881	if (!hdr)
882		return false;
883
884	if (hdr->type == BR_MRP_TLV_HEADER_RING_TEST ||
885	    hdr->type == BR_MRP_TLV_HEADER_RING_TOPO ||
886	    hdr->type == BR_MRP_TLV_HEADER_RING_LINK_DOWN ||
887	    hdr->type == BR_MRP_TLV_HEADER_RING_LINK_UP ||
888	    hdr->type == BR_MRP_TLV_HEADER_OPTION)
889		return true;
890
891	return false;
892}
893
894/* Determine if the frame type is an interconnect frame */
895static bool br_mrp_in_frame(struct sk_buff *skb)
896{
897	const struct br_mrp_tlv_hdr *hdr;
898	struct br_mrp_tlv_hdr _hdr;
899
900	hdr = skb_header_pointer(skb, sizeof(uint16_t), sizeof(_hdr), &_hdr);
901	if (!hdr)
902		return false;
903
904	if (hdr->type == BR_MRP_TLV_HEADER_IN_TEST ||
905	    hdr->type == BR_MRP_TLV_HEADER_IN_TOPO ||
906	    hdr->type == BR_MRP_TLV_HEADER_IN_LINK_DOWN ||
907	    hdr->type == BR_MRP_TLV_HEADER_IN_LINK_UP ||
908	    hdr->type == BR_MRP_TLV_HEADER_IN_LINK_STATUS)
909		return true;
910
911	return false;
912}
913
914/* Process only MRP Test frame. All the other MRP frames are processed by
915 * userspace application
916 * note: already called with rcu_read_lock
917 */
918static void br_mrp_mrm_process(struct br_mrp *mrp, struct net_bridge_port *port,
919			       struct sk_buff *skb)
920{
921	const struct br_mrp_tlv_hdr *hdr;
922	struct br_mrp_tlv_hdr _hdr;
923
924	/* Each MRP header starts with a version field which is 16 bits.
925	 * Therefore skip the version and get directly the TLV header.
926	 */
927	hdr = skb_header_pointer(skb, sizeof(uint16_t), sizeof(_hdr), &_hdr);
928	if (!hdr)
929		return;
930
931	if (hdr->type != BR_MRP_TLV_HEADER_RING_TEST)
932		return;
933
934	mrp->test_count_miss = 0;
935
936	/* Notify the userspace that the ring is closed only when the ring is
937	 * not closed
938	 */
939	if (mrp->ring_state != BR_MRP_RING_STATE_CLOSED)
940		br_mrp_ring_port_open(port->dev, false);
941}
942
943/* Determine if the test hdr has a better priority than the node */
944static bool br_mrp_test_better_than_own(struct br_mrp *mrp,
945					struct net_bridge *br,
946					const struct br_mrp_ring_test_hdr *hdr)
947{
948	u16 prio = be16_to_cpu(hdr->prio);
949
950	if (prio < mrp->prio ||
951	    (prio == mrp->prio &&
952	    ether_addr_to_u64(hdr->sa) < ether_addr_to_u64(br->dev->dev_addr)))
953		return true;
954
955	return false;
956}
957
958/* Process only MRP Test frame. All the other MRP frames are processed by
959 * userspace application
960 * note: already called with rcu_read_lock
961 */
962static void br_mrp_mra_process(struct br_mrp *mrp, struct net_bridge *br,
963			       struct net_bridge_port *port,
964			       struct sk_buff *skb)
965{
966	const struct br_mrp_ring_test_hdr *test_hdr;
967	struct br_mrp_ring_test_hdr _test_hdr;
968	const struct br_mrp_tlv_hdr *hdr;
969	struct br_mrp_tlv_hdr _hdr;
970
971	/* Each MRP header starts with a version field which is 16 bits.
972	 * Therefore skip the version and get directly the TLV header.
973	 */
974	hdr = skb_header_pointer(skb, sizeof(uint16_t), sizeof(_hdr), &_hdr);
975	if (!hdr)
976		return;
977
978	if (hdr->type != BR_MRP_TLV_HEADER_RING_TEST)
979		return;
980
981	test_hdr = skb_header_pointer(skb, sizeof(uint16_t) + sizeof(_hdr),
982				      sizeof(_test_hdr), &_test_hdr);
983	if (!test_hdr)
984		return;
985
986	/* Only frames that have a better priority than the node will
987	 * clear the miss counter because otherwise the node will need to behave
988	 * as MRM.
989	 */
990	if (br_mrp_test_better_than_own(mrp, br, test_hdr))
991		mrp->test_count_miss = 0;
992}
993
994/* Process only MRP InTest frame. All the other MRP frames are processed by
995 * userspace application
996 * note: already called with rcu_read_lock
997 */
998static bool br_mrp_mim_process(struct br_mrp *mrp, struct net_bridge_port *port,
999			       struct sk_buff *skb)
1000{
1001	const struct br_mrp_in_test_hdr *in_hdr;
1002	struct br_mrp_in_test_hdr _in_hdr;
1003	const struct br_mrp_tlv_hdr *hdr;
1004	struct br_mrp_tlv_hdr _hdr;
1005
1006	/* Each MRP header starts with a version field which is 16 bits.
1007	 * Therefore skip the version and get directly the TLV header.
1008	 */
1009	hdr = skb_header_pointer(skb, sizeof(uint16_t), sizeof(_hdr), &_hdr);
1010	if (!hdr)
1011		return false;
1012
1013	/* The check for InTest frame type was already done */
1014	in_hdr = skb_header_pointer(skb, sizeof(uint16_t) + sizeof(_hdr),
1015				    sizeof(_in_hdr), &_in_hdr);
1016	if (!in_hdr)
1017		return false;
1018
1019	/* It needs to process only it's own InTest frames. */
1020	if (mrp->in_id != ntohs(in_hdr->id))
1021		return false;
1022
1023	mrp->in_test_count_miss = 0;
1024
1025	/* Notify the userspace that the ring is closed only when the ring is
1026	 * not closed
1027	 */
1028	if (mrp->in_state != BR_MRP_IN_STATE_CLOSED)
1029		br_mrp_in_port_open(port->dev, false);
1030
1031	return true;
1032}
1033
1034/* Get the MRP frame type
1035 * note: already called with rcu_read_lock
1036 */
1037static u8 br_mrp_get_frame_type(struct sk_buff *skb)
1038{
1039	const struct br_mrp_tlv_hdr *hdr;
1040	struct br_mrp_tlv_hdr _hdr;
1041
1042	/* Each MRP header starts with a version field which is 16 bits.
1043	 * Therefore skip the version and get directly the TLV header.
1044	 */
1045	hdr = skb_header_pointer(skb, sizeof(uint16_t), sizeof(_hdr), &_hdr);
1046	if (!hdr)
1047		return 0xff;
1048
1049	return hdr->type;
1050}
1051
1052static bool br_mrp_mrm_behaviour(struct br_mrp *mrp)
1053{
1054	if (mrp->ring_role == BR_MRP_RING_ROLE_MRM ||
1055	    (mrp->ring_role == BR_MRP_RING_ROLE_MRA && !mrp->test_monitor))
1056		return true;
1057
1058	return false;
1059}
1060
1061static bool br_mrp_mrc_behaviour(struct br_mrp *mrp)
1062{
1063	if (mrp->ring_role == BR_MRP_RING_ROLE_MRC ||
1064	    (mrp->ring_role == BR_MRP_RING_ROLE_MRA && mrp->test_monitor))
1065		return true;
1066
1067	return false;
1068}
1069
1070/* This will just forward the frame to the other mrp ring ports, depending on
1071 * the frame type, ring role and interconnect role
1072 * note: already called with rcu_read_lock
1073 */
1074static int br_mrp_rcv(struct net_bridge_port *p,
1075		      struct sk_buff *skb, struct net_device *dev)
1076{
1077	struct net_bridge_port *p_port, *s_port, *i_port = NULL;
1078	struct net_bridge_port *p_dst, *s_dst, *i_dst = NULL;
1079	struct net_bridge *br;
1080	struct br_mrp *mrp;
1081
1082	/* If port is disabled don't accept any frames */
1083	if (p->state == BR_STATE_DISABLED)
1084		return 0;
1085
1086	br = p->br;
1087	mrp =  br_mrp_find_port(br, p);
1088	if (unlikely(!mrp))
1089		return 0;
1090
1091	p_port = rcu_dereference(mrp->p_port);
1092	if (!p_port)
1093		return 0;
1094	p_dst = p_port;
1095
1096	s_port = rcu_dereference(mrp->s_port);
1097	if (!s_port)
1098		return 0;
1099	s_dst = s_port;
1100
1101	/* If the frame is a ring frame then it is not required to check the
1102	 * interconnect role and ports to process or forward the frame
1103	 */
1104	if (br_mrp_ring_frame(skb)) {
1105		/* If the role is MRM then don't forward the frames */
1106		if (mrp->ring_role == BR_MRP_RING_ROLE_MRM) {
1107			br_mrp_mrm_process(mrp, p, skb);
1108			goto no_forward;
1109		}
1110
1111		/* If the role is MRA then don't forward the frames if it
1112		 * behaves as MRM node
1113		 */
1114		if (mrp->ring_role == BR_MRP_RING_ROLE_MRA) {
1115			if (!mrp->test_monitor) {
1116				br_mrp_mrm_process(mrp, p, skb);
1117				goto no_forward;
1118			}
1119
1120			br_mrp_mra_process(mrp, br, p, skb);
1121		}
1122
1123		goto forward;
1124	}
1125
1126	if (br_mrp_in_frame(skb)) {
1127		u8 in_type = br_mrp_get_frame_type(skb);
1128
1129		i_port = rcu_dereference(mrp->i_port);
1130		i_dst = i_port;
1131
1132		/* If the ring port is in block state it should not forward
1133		 * In_Test frames
1134		 */
1135		if (br_mrp_is_ring_port(p_port, s_port, p) &&
1136		    p->state == BR_STATE_BLOCKING &&
1137		    in_type == BR_MRP_TLV_HEADER_IN_TEST)
1138			goto no_forward;
1139
1140		/* Nodes that behaves as MRM needs to stop forwarding the
1141		 * frames in case the ring is closed, otherwise will be a loop.
1142		 * In this case the frame is no forward between the ring ports.
1143		 */
1144		if (br_mrp_mrm_behaviour(mrp) &&
1145		    br_mrp_is_ring_port(p_port, s_port, p) &&
1146		    (s_port->state != BR_STATE_FORWARDING ||
1147		     p_port->state != BR_STATE_FORWARDING)) {
1148			p_dst = NULL;
1149			s_dst = NULL;
1150		}
1151
1152		/* A node that behaves as MRC and doesn't have a interconnect
1153		 * role then it should forward all frames between the ring ports
1154		 * because it doesn't have an interconnect port
1155		 */
1156		if (br_mrp_mrc_behaviour(mrp) &&
1157		    mrp->in_role == BR_MRP_IN_ROLE_DISABLED)
1158			goto forward;
1159
1160		if (mrp->in_role == BR_MRP_IN_ROLE_MIM) {
1161			if (in_type == BR_MRP_TLV_HEADER_IN_TEST) {
1162				/* MIM should not forward it's own InTest
1163				 * frames
1164				 */
1165				if (br_mrp_mim_process(mrp, p, skb)) {
1166					goto no_forward;
1167				} else {
1168					if (br_mrp_is_ring_port(p_port, s_port,
1169								p))
1170						i_dst = NULL;
1171
1172					if (br_mrp_is_in_port(i_port, p))
1173						goto no_forward;
1174				}
1175			} else {
1176				/* MIM should forward IntLinkChange/Status and
1177				 * IntTopoChange between ring ports but MIM
1178				 * should not forward IntLinkChange/Status and
1179				 * IntTopoChange if the frame was received at
1180				 * the interconnect port
1181				 */
1182				if (br_mrp_is_ring_port(p_port, s_port, p))
1183					i_dst = NULL;
1184
1185				if (br_mrp_is_in_port(i_port, p))
1186					goto no_forward;
1187			}
1188		}
1189
1190		if (mrp->in_role == BR_MRP_IN_ROLE_MIC) {
1191			/* MIC should forward InTest frames on all ports
1192			 * regardless of the received port
1193			 */
1194			if (in_type == BR_MRP_TLV_HEADER_IN_TEST)
1195				goto forward;
1196
1197			/* MIC should forward IntLinkChange frames only if they
1198			 * are received on ring ports to all the ports
1199			 */
1200			if (br_mrp_is_ring_port(p_port, s_port, p) &&
1201			    (in_type == BR_MRP_TLV_HEADER_IN_LINK_UP ||
1202			     in_type == BR_MRP_TLV_HEADER_IN_LINK_DOWN))
1203				goto forward;
1204
1205			/* MIC should forward IntLinkStatus frames only to
1206			 * interconnect port if it was received on a ring port.
1207			 * If it is received on interconnect port then, it
1208			 * should be forward on both ring ports
1209			 */
1210			if (br_mrp_is_ring_port(p_port, s_port, p) &&
1211			    in_type == BR_MRP_TLV_HEADER_IN_LINK_STATUS) {
1212				p_dst = NULL;
1213				s_dst = NULL;
1214			}
1215
1216			/* Should forward the InTopo frames only between the
1217			 * ring ports
1218			 */
1219			if (in_type == BR_MRP_TLV_HEADER_IN_TOPO) {
1220				i_dst = NULL;
1221				goto forward;
1222			}
1223
1224			/* In all the other cases don't forward the frames */
1225			goto no_forward;
1226		}
1227	}
1228
1229forward:
1230	if (p_dst)
1231		br_forward(p_dst, skb, true, false);
1232	if (s_dst)
1233		br_forward(s_dst, skb, true, false);
1234	if (i_dst)
1235		br_forward(i_dst, skb, true, false);
1236
1237no_forward:
1238	return 1;
1239}
1240
1241/* Check if the frame was received on a port that is part of MRP ring
1242 * and if the frame has MRP eth. In that case process the frame otherwise do
1243 * normal forwarding.
1244 * note: already called with rcu_read_lock
1245 */
1246static int br_mrp_process(struct net_bridge_port *p, struct sk_buff *skb)
1247{
1248	/* If there is no MRP instance do normal forwarding */
1249	if (likely(!(p->flags & BR_MRP_AWARE)))
1250		goto out;
1251
1252	return br_mrp_rcv(p, skb, p->dev);
1253out:
1254	return 0;
1255}
1256
1257bool br_mrp_enabled(struct net_bridge *br)
1258{
1259	return !hlist_empty(&br->mrp_list);
1260}
1261