summaryrefslogtreecommitdiffstats
path: root/drivers/net/ethernet/chelsio/cxgb3/l2t.c
blob: b3e4118a15e700364292aba4c343f3dc67d9e849 (plain)
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
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
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
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
/*
 * Copyright (c) 2003-2008 Chelsio, 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 <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/if.h>
#include <linux/if_vlan.h>
#include <linux/jhash.h>
#include <linux/slab.h>
#include <linux/export.h>
#include <net/neighbour.h>
#include "common.h"
#include "t3cdev.h"
#include "cxgb3_defs.h"
#include "l2t.h"
#include "t3_cpl.h"
#include "firmware_exports.h"

#define VLAN_NONE 0xfff

/*
 * Module locking notes:  There is a RW lock protecting the L2 table as a
 * whole plus a spinlock per L2T entry.  Entry lookups and allocations happen
 * under the protection of the table lock, individual entry changes happen
 * while holding that entry's spinlock.  The table lock nests outside the
 * entry locks.  Allocations of new entries take the table lock as writers so
 * no other lookups can happen while allocating new entries.  Entry updates
 * take the table lock as readers so multiple entries can be updated in
 * parallel.  An L2T entry can be dropped by decrementing its reference count
 * and therefore can happen in parallel with entry allocation but no entry
 * can change state or increment its ref count during allocation as both of
 * these perform lookups.
 */

static inline unsigned int vlan_prio(const struct l2t_entry *e)
{
	return e->vlan >> 13;
}

static inline unsigned int arp_hash(u32 key, int ifindex,
				    const struct l2t_data *d)
{
	return jhash_2words(key, ifindex, 0) & (d->nentries - 1);
}

static inline void neigh_replace(struct l2t_entry *e, struct neighbour *n)
{
	neigh_hold(n);
	if (e->neigh)
		neigh_release(e->neigh);
	e->neigh = n;
}

/*
 * Set up an L2T entry and send any packets waiting in the arp queue.  The
 * supplied skb is used for the CPL_L2T_WRITE_REQ.  Must be called with the
 * entry locked.
 */
static int setup_l2e_send_pending(struct t3cdev *dev, struct sk_buff *skb,
				  struct l2t_entry *e)
{
	struct cpl_l2t_write_req *req;
	struct sk_buff *tmp;

	if (!skb) {
		skb = alloc_skb(sizeof(*req), GFP_ATOMIC);
		if (!skb)
			return -ENOMEM;
	}

	req = __skb_put(skb, sizeof(*req));
	req->wr.wr_hi = htonl(V_WR_OP(FW_WROPCODE_FORWARD));
	OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_L2T_WRITE_REQ, e->idx));
	req->params = htonl(V_L2T_W_IDX(e->idx) | V_L2T_W_IFF(e->smt_idx) |
			    V_L2T_W_VLAN(e->vlan & VLAN_VID_MASK) |
			    V_L2T_W_PRIO(vlan_prio(e)));
	memcpy(e->dmac, e->neigh->ha, sizeof(e->dmac));
	memcpy(req->dst_mac, e->dmac, sizeof(req->dst_mac));
	skb->priority = CPL_PRIORITY_CONTROL;
	cxgb3_ofld_send(dev, skb);

	skb_queue_walk_safe(&e->arpq, skb, tmp) {
		__skb_unlink(skb, &e->arpq);
		cxgb3_ofld_send(dev, skb);
	}
	e->state = L2T_STATE_VALID;

	return 0;
}

/*
 * Add a packet to the an L2T entry's queue of packets awaiting resolution.
 * Must be called with the entry's lock held.
 */
static inline void arpq_enqueue(struct l2t_entry *e, struct sk_buff *skb)
{
	__skb_queue_tail(&e->arpq, skb);
}

int t3_l2t_send_slow(struct t3cdev *dev, struct sk_buff *skb,
		     struct l2t_entry *e)
{
again:
	switch (e->state) {
	case L2T_STATE_STALE:	/* entry is stale, kick off revalidation */
		neigh_event_send(e->neigh, NULL);
		spin_lock_bh(&e->lock);
		if (e->state == L2T_STATE_STALE)
			e->state = L2T_STATE_VALID;
		spin_unlock_bh(&e->lock);
		/* fall through */
	case L2T_STATE_VALID:	/* fast-path, send the packet on */
		return cxgb3_ofld_send(dev, skb);
	case L2T_STATE_RESOLVING:
		spin_lock_bh(&e->lock);
		if (e->state != L2T_STATE_RESOLVING) {
			/* ARP already completed */
			spin_unlock_bh(&e->lock);
			goto again;
		}
		arpq_enqueue(e, skb);
		spin_unlock_bh(&e->lock);

		/*
		 * Only the first packet added to the arpq should kick off
		 * resolution.  However, because the alloc_skb below can fail,
		 * we allow each packet added to the arpq to retry resolution
		 * as a way of recovering from transient memory exhaustion.
		 * A better way would be to use a work request to retry L2T
		 * entries when there's no memory.
		 */
		if (!neigh_event_send(e->neigh, NULL)) {
			skb = alloc_skb(sizeof(struct cpl_l2t_write_req),
					GFP_ATOMIC);
			if (!skb)
				break;

			spin_lock_bh(&e->lock);
			if (!skb_queue_empty(&e->arpq))
				setup_l2e_send_pending(dev, skb, e);
			else	/* we lost the race */
				__kfree_skb(skb);
			spin_unlock_bh(&e->lock);
		}
	}
	return 0;
}

EXPORT_SYMBOL(t3_l2t_send_slow);

void t3_l2t_send_event(struct t3cdev *dev, struct l2t_entry *e)
{
again:
	switch (e->state) {
	case L2T_STATE_STALE:	/* entry is stale, kick off revalidation */
		neigh_event_send(e->neigh, NULL);
		spin_lock_bh(&e->lock);
		if (e->state == L2T_STATE_STALE) {
			e->state = L2T_STATE_VALID;
		}
		spin_unlock_bh(&e->lock);
		return;
	case L2T_STATE_VALID:	/* fast-path, send the packet on */
		return;
	case L2T_STATE_RESOLVING:
		spin_lock_bh(&e->lock);
		if (e->state != L2T_STATE_RESOLVING) {
			/* ARP already completed */
			spin_unlock_bh(&e->lock);
			goto again;
		}
		spin_unlock_bh(&e->lock);

		/*
		 * Only the first packet added to the arpq should kick off
		 * resolution.  However, because the alloc_skb below can fail,
		 * we allow each packet added to the arpq to retry resolution
		 * as a way of recovering from transient memory exhaustion.
		 * A better way would be to use a work request to retry L2T
		 * entries when there's no memory.
		 */
		neigh_event_send(e->neigh, NULL);
	}
}

EXPORT_SYMBOL(t3_l2t_send_event);

/*
 * Allocate a free L2T entry.  Must be called with l2t_data.lock held.
 */
static struct l2t_entry *alloc_l2e(struct l2t_data *d)
{
	struct l2t_entry *end, *e, **p;

	if (!atomic_read(&d->nfree))
		return NULL;

	/* there's definitely a free entry */
	for (e = d->rover, end = &d->l2tab[d->nentries]; e != end; ++e)
		if (atomic_read(&e->refcnt) == 0)
			goto found;

	for (e = &d->l2tab[1]; atomic_read(&e->refcnt); ++e) ;
found:
	d->rover = e + 1;
	atomic_dec(&d->nfree);

	/*
	 * The entry we found may be an inactive entry that is
	 * presently in the hash table.  We need to remove it.
	 */
	if (e->state != L2T_STATE_UNUSED) {
		int hash = arp_hash(e->addr, e->ifindex, d);

		for (p = &d->l2tab[hash].first; *p; p = &(*p)->next)
			if (*p == e) {
				*p = e->next;
				break;
			}
		e->state = L2T_STATE_UNUSED;
	}
	return e;
}

/*
 * Called when an L2T entry has no more users.  The entry is left in the hash
 * table since it is likely to be reused but we also bump nfree to indicate
 * that the entry can be reallocated for a different neighbor.  We also drop
 * the existing neighbor reference in case the neighbor is going away and is
 * waiting on our reference.
 *
 * Because entries can be reallocated to other neighbors once their ref count
 * drops to 0 we need to take the entry's lock to avoid races with a new
 * incarnation.
 */
void t3_l2e_free(struct l2t_data *d, struct l2t_entry *e)
{
	spin_lock_bh(&e->lock);
	if (atomic_read(&e->refcnt) == 0) {	/* hasn't been recycled */
		if (e->neigh) {
			neigh_release(e->neigh);
			e->neigh = NULL;
		}
	}
	spin_unlock_bh(&e->lock);
	atomic_inc(&d->nfree);
}

EXPORT_SYMBOL(t3_l2e_free);

/*
 * Update an L2T entry that was previously used for the same next hop as neigh.
 * Must be called with softirqs disabled.
 */
static inline void reuse_entry(struct l2t_entry *e, struct neighbour *neigh)
{
	unsigned int nud_state;

	spin_lock(&e->lock);	/* avoid race with t3_l2t_free */

	if (neigh != e->neigh)
		neigh_replace(e, neigh);
	nud_state = neigh->nud_state;
	if (memcmp(e->dmac, neigh->ha, sizeof(e->dmac)) ||
	    !(nud_state & NUD_VALID))
		e->state = L2T_STATE_RESOLVING;
	else if (nud_state & NUD_CONNECTED)
		e->state = L2T_STATE_VALID;
	else
		e->state = L2T_STATE_STALE;
	spin_unlock(&e->lock);
}

struct l2t_entry *t3_l2t_get(struct t3cdev *cdev, struct dst_entry *dst,
			     struct net_device *dev, const void *daddr)
{
	struct l2t_entry *e = NULL;
	struct neighbour *neigh;
	struct port_info *p;
	struct l2t_data *d;
	int hash;
	u32 addr;
	int ifidx;
	int smt_idx;

	rcu_read_lock();
	neigh = dst_neigh_lookup(dst, daddr);
	if (!neigh)
		goto done_rcu;

	addr = *(u32 *) neigh->primary_key;
	ifidx = neigh->dev->ifindex;

	if (!dev)
		dev = neigh->dev;
	p = netdev_priv(dev);
	smt_idx = p->port_id;

	d = L2DATA(cdev);
	if (!d)
		goto done_rcu;

	hash = arp_hash(addr, ifidx, d);

	write_lock_bh(&d->lock);
	for (e = d->l2tab[hash].first; e; e = e->next)
		if (e->addr == addr && e->ifindex == ifidx &&
		    e->smt_idx == smt_idx) {
			l2t_hold(d, e);
			if (atomic_read(&e->refcnt) == 1)
				reuse_entry(e, neigh);
			goto done_unlock;
		}

	/* Need to allocate a new entry */
	e = alloc_l2e(d);
	if (e) {
		spin_lock(&e->lock);	/* avoid race with t3_l2t_free */
		e->next = d->l2tab[hash].first;
		d->l2tab[hash].first = e;
		e->state = L2T_STATE_RESOLVING;
		e->addr = addr;
		e->ifindex = ifidx;
		e->smt_idx = smt_idx;
		atomic_set(&e->refcnt, 1);
		neigh_replace(e, neigh);
		if (is_vlan_dev(neigh->dev))
			e->vlan = vlan_dev_vlan_id(neigh->dev);
		else
			e->vlan = VLAN_NONE;
		spin_unlock(&e->lock);
	}
done_unlock:
	write_unlock_bh(&d->lock);
done_rcu:
	if (neigh)
		neigh_release(neigh);
	rcu_read_unlock();
	return e;
}

EXPORT_SYMBOL(t3_l2t_get);

/*
 * Called when address resolution fails for an L2T entry to handle packets
 * on the arpq head.  If a packet specifies a failure handler it is invoked,
 * otherwise the packets is sent to the offload device.
 *
 * XXX: maybe we should abandon the latter behavior and just require a failure
 * handler.
 */
static void handle_failed_resolution(struct t3cdev *dev, struct sk_buff_head *arpq)
{
	struct sk_buff *skb, *tmp;

	skb_queue_walk_safe(arpq, skb, tmp) {
		struct l2t_skb_cb *cb = L2T_SKB_CB(skb);

		__skb_unlink(skb, arpq);
		if (cb->arp_failure_handler)
			cb->arp_failure_handler(dev, skb);
		else
			cxgb3_ofld_send(dev, skb);
	}
}

/*
 * Called when the host's ARP layer makes a change to some entry that is
 * loaded into the HW L2 table.
 */
void t3_l2t_update(struct t3cdev *dev, struct neighbour *neigh)
{
	struct sk_buff_head arpq;
	struct l2t_entry *e;
	struct l2t_data *d = L2DATA(dev);
	u32 addr = *(u32 *) neigh->primary_key;
	int ifidx = neigh->dev->ifindex;
	int hash = arp_hash(addr, ifidx, d);

	read_lock_bh(&d->lock);
	for (e = d->l2tab[hash].first; e; e = e->next)
		if (e->addr == addr && e->ifindex == ifidx) {
			spin_lock(&e->lock);
			goto found;
		}
	read_unlock_bh(&d->lock);
	return;

found:
	__skb_queue_head_init(&arpq);

	read_unlock(&d->lock);
	if (atomic_read(&e->refcnt)) {
		if (neigh != e->neigh)
			neigh_replace(e, neigh);

		if (e->state == L2T_STATE_RESOLVING) {
			if (neigh->nud_state & NUD_FAILED) {
				skb_queue_splice_init(&e->arpq, &arpq);
			} else if (neigh->nud_state & (NUD_CONNECTED|NUD_STALE))
				setup_l2e_send_pending(dev, NULL, e);
		} else {
			e->state = neigh->nud_state & NUD_CONNECTED ?
			    L2T_STATE_VALID : L2T_STATE_STALE;
			if (!ether_addr_equal(e->dmac, neigh->ha))
				setup_l2e_send_pending(dev, NULL, e);
		}
	}
	spin_unlock_bh(&e->lock);

	if (!skb_queue_empty(&arpq))
		handle_failed_resolution(dev, &arpq);
}

struct l2t_data *t3_init_l2t(unsigned int l2t_capacity)
{
	struct l2t_data *d;
	int i;

	d = kvzalloc(struct_size(d, l2tab, l2t_capacity), GFP_KERNEL);
	if (!d)
		return NULL;

	d->nentries = l2t_capacity;
	d->rover = &d->l2tab[1];	/* entry 0 is not used */
	atomic_set(&d->nfree, l2t_capacity - 1);
	rwlock_init(&d->lock);

	for (i = 0; i < l2t_capacity; ++i) {
		d->l2tab[i].idx = i;
		d->l2tab[i].state = L2T_STATE_UNUSED;
		__skb_queue_head_init(&d->l2tab[i].arpq);
		spin_lock_init(&d->l2tab[i].lock);
		atomic_set(&d->l2tab[i].refcnt, 0);
	}
	return d;
}