aboutsummaryrefslogtreecommitdiffstats
path: root/net/mac80211/agg-tx.c
blob: b09948ceec4ae1c031754c08c59d425de3e5ab7b (plain) (blame)
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
466
467
468
469
470
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
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
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
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
/*
 * HT handling
 *
 * Copyright 2003, Jouni Malinen <jkmaline@cc.hut.fi>
 * Copyright 2002-2005, Instant802 Networks, Inc.
 * Copyright 2005-2006, Devicescape Software, Inc.
 * Copyright 2006-2007	Jiri Benc <jbenc@suse.cz>
 * Copyright 2007, Michael Wu <flamingice@sourmilk.net>
 * Copyright 2007-2009, Intel Corporation
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */

#include <linux/ieee80211.h>
#include <net/mac80211.h>
#include "ieee80211_i.h"
#include "driver-ops.h"
#include "wme.h"

/**
 * DOC: TX aggregation
 *
 * Aggregation on the TX side requires setting the hardware flag
 * %IEEE80211_HW_AMPDU_AGGREGATION as well as, if present, the @ampdu_queues
 * hardware parameter to the number of hardware AMPDU queues. If there are no
 * hardware queues then the driver will (currently) have to do all frame
 * buffering.
 *
 * When TX aggregation is started by some subsystem (usually the rate control
 * algorithm would be appropriate) by calling the
 * ieee80211_start_tx_ba_session() function, the driver will be notified via
 * its @ampdu_action function, with the %IEEE80211_AMPDU_TX_START action.
 *
 * In response to that, the driver is later required to call the
 * ieee80211_start_tx_ba_cb() (or ieee80211_start_tx_ba_cb_irqsafe())
 * function, which will start the aggregation session.
 *
 * Similarly, when the aggregation session is stopped by
 * ieee80211_stop_tx_ba_session(), the driver's @ampdu_action function will
 * be called with the action %IEEE80211_AMPDU_TX_STOP. In this case, the
 * call must not fail, and the driver must later call ieee80211_stop_tx_ba_cb()
 * (or ieee80211_stop_tx_ba_cb_irqsafe()).
 */

static void ieee80211_send_addba_request(struct ieee80211_sub_if_data *sdata,
					 const u8 *da, u16 tid,
					 u8 dialog_token, u16 start_seq_num,
					 u16 agg_size, u16 timeout)
{
	struct ieee80211_local *local = sdata->local;
	struct sk_buff *skb;
	struct ieee80211_mgmt *mgmt;
	u16 capab;

	skb = dev_alloc_skb(sizeof(*mgmt) + local->hw.extra_tx_headroom);

	if (!skb) {
		printk(KERN_ERR "%s: failed to allocate buffer "
				"for addba request frame\n", sdata->dev->name);
		return;
	}
	skb_reserve(skb, local->hw.extra_tx_headroom);
	mgmt = (struct ieee80211_mgmt *) skb_put(skb, 24);
	memset(mgmt, 0, 24);
	memcpy(mgmt->da, da, ETH_ALEN);
	memcpy(mgmt->sa, sdata->dev->dev_addr, ETH_ALEN);
	if (sdata->vif.type == NL80211_IFTYPE_AP ||
	    sdata->vif.type == NL80211_IFTYPE_AP_VLAN)
		memcpy(mgmt->bssid, sdata->dev->dev_addr, ETH_ALEN);
	else if (sdata->vif.type == NL80211_IFTYPE_STATION)
		memcpy(mgmt->bssid, sdata->u.mgd.bssid, ETH_ALEN);

	mgmt->frame_control = cpu_to_le16(IEEE80211_FTYPE_MGMT |
					  IEEE80211_STYPE_ACTION);

	skb_put(skb, 1 + sizeof(mgmt->u.action.u.addba_req));

	mgmt->u.action.category = WLAN_CATEGORY_BACK;
	mgmt->u.action.u.addba_req.action_code = WLAN_ACTION_ADDBA_REQ;

	mgmt->u.action.u.addba_req.dialog_token = dialog_token;
	capab = (u16)(1 << 1);		/* bit 1 aggregation policy */
	capab |= (u16)(tid << 2); 	/* bit 5:2 TID number */
	capab |= (u16)(agg_size << 6);	/* bit 15:6 max size of aggergation */

	mgmt->u.action.u.addba_req.capab = cpu_to_le16(capab);

	mgmt->u.action.u.addba_req.timeout = cpu_to_le16(timeout);
	mgmt->u.action.u.addba_req.start_seq_num =
					cpu_to_le16(start_seq_num << 4);

	ieee80211_tx_skb(sdata, skb, 1);
}

void ieee80211_send_bar(struct ieee80211_sub_if_data *sdata, u8 *ra, u16 tid, u16 ssn)
{
	struct ieee80211_local *local = sdata->local;
	struct sk_buff *skb;
	struct ieee80211_bar *bar;
	u16 bar_control = 0;

	skb = dev_alloc_skb(sizeof(*bar) + local->hw.extra_tx_headroom);
	if (!skb) {
		printk(KERN_ERR "%s: failed to allocate buffer for "
			"bar frame\n", sdata->dev->name);
		return;
	}
	skb_reserve(skb, local->hw.extra_tx_headroom);
	bar = (struct ieee80211_bar *)skb_put(skb, sizeof(*bar));
	memset(bar, 0, sizeof(*bar));
	bar->frame_control = cpu_to_le16(IEEE80211_FTYPE_CTL |
					 IEEE80211_STYPE_BACK_REQ);
	memcpy(bar->ra, ra, ETH_ALEN);
	memcpy(bar->ta, sdata->dev->dev_addr, ETH_ALEN);
	bar_control |= (u16)IEEE80211_BAR_CTRL_ACK_POLICY_NORMAL;
	bar_control |= (u16)IEEE80211_BAR_CTRL_CBMTID_COMPRESSED_BA;
	bar_control |= (u16)(tid << 12);
	bar->control = cpu_to_le16(bar_control);
	bar->start_seq_num = cpu_to_le16(ssn);

	ieee80211_tx_skb(sdata, skb, 0);
}

static int ___ieee80211_stop_tx_ba_session(struct sta_info *sta, u16 tid,
					   enum ieee80211_back_parties initiator)
{
	struct ieee80211_local *local = sta->local;
	int ret;
	u8 *state;

	state = &sta->ampdu_mlme.tid_state_tx[tid];

	if (*state == HT_AGG_STATE_OPERATIONAL)
		sta->ampdu_mlme.addba_req_num[tid] = 0;

	*state = HT_AGG_STATE_REQ_STOP_BA_MSK |
		(initiator << HT_AGG_STATE_INITIATOR_SHIFT);

	ret = drv_ampdu_action(local, IEEE80211_AMPDU_TX_STOP,
			       &sta->sta, tid, NULL);

	/* HW shall not deny going back to legacy */
	if (WARN_ON(ret)) {
		*state = HT_AGG_STATE_OPERATIONAL;
		/*
		 * We may have pending packets get stuck in this case...
		 * Not bothering with a workaround for now.
		 */
	}

	return ret;
}

/*
 * After sending add Block Ack request we activated a timer until
 * add Block Ack response will arrive from the recipient.
 * If this timer expires sta_addba_resp_timer_expired will be executed.
 */
static void sta_addba_resp_timer_expired(unsigned long data)
{
	/* not an elegant detour, but there is no choice as the timer passes
	 * only one argument, and both sta_info and TID are needed, so init
	 * flow in sta_info_create gives the TID as data, while the timer_to_id
	 * array gives the sta through container_of */
	u16 tid = *(u8 *)data;
	struct sta_info *sta = container_of((void *)data,
		struct sta_info, timer_to_tid[tid]);
	u8 *state;

	state = &sta->ampdu_mlme.tid_state_tx[tid];

	/* check if the TID waits for addBA response */
	spin_lock_bh(&sta->lock);
	if (!(*state & HT_ADDBA_REQUESTED_MSK)) {
		spin_unlock_bh(&sta->lock);
		*state = HT_AGG_STATE_IDLE;
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "timer expired on tid %d but we are not "
				"expecting addBA response there", tid);
#endif
		return;
	}

#ifdef CONFIG_MAC80211_HT_DEBUG
	printk(KERN_DEBUG "addBA response timer expired on tid %d\n", tid);
#endif

	___ieee80211_stop_tx_ba_session(sta, tid, WLAN_BACK_INITIATOR);
	spin_unlock_bh(&sta->lock);
}

static inline int ieee80211_ac_from_tid(int tid)
{
	return ieee802_1d_to_ac[tid & 7];
}

int ieee80211_start_tx_ba_session(struct ieee80211_hw *hw, u8 *ra, u16 tid)
{
	struct ieee80211_local *local = hw_to_local(hw);
	struct sta_info *sta;
	struct ieee80211_sub_if_data *sdata;
	u8 *state;
	int ret = 0;
	u16 start_seq_num;

	if (WARN_ON(!local->ops->ampdu_action))
		return -EINVAL;

	if ((tid >= STA_TID_NUM) || !(hw->flags & IEEE80211_HW_AMPDU_AGGREGATION))
		return -EINVAL;

#ifdef CONFIG_MAC80211_HT_DEBUG
	printk(KERN_DEBUG "Open BA session requested for %pM tid %u\n",
	       ra, tid);
#endif /* CONFIG_MAC80211_HT_DEBUG */

	rcu_read_lock();

	sta = sta_info_get(local, ra);
	if (!sta) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "Could not find the station\n");
#endif
		ret = -ENOENT;
		goto unlock;
	}

	/*
	 * The aggregation code is not prepared to handle
	 * anything but STA/AP due to the BSSID handling.
	 * IBSS could work in the code but isn't supported
	 * by drivers or the standard.
	 */
	if (sta->sdata->vif.type != NL80211_IFTYPE_STATION &&
	    sta->sdata->vif.type != NL80211_IFTYPE_AP_VLAN &&
	    sta->sdata->vif.type != NL80211_IFTYPE_AP) {
		ret = -EINVAL;
		goto unlock;
	}

	if (test_sta_flags(sta, WLAN_STA_SUSPEND)) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "Suspend in progress. "
		       "Denying BA session request\n");
#endif
		ret = -EINVAL;
		goto unlock;
	}

	spin_lock_bh(&sta->lock);
	spin_lock(&local->ampdu_lock);

	sdata = sta->sdata;

	/* we have tried too many times, receiver does not want A-MPDU */
	if (sta->ampdu_mlme.addba_req_num[tid] > HT_AGG_MAX_RETRIES) {
		ret = -EBUSY;
		goto err_unlock_sta;
	}

	state = &sta->ampdu_mlme.tid_state_tx[tid];
	/* check if the TID is not in aggregation flow already */
	if (*state != HT_AGG_STATE_IDLE) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "BA request denied - session is not "
				 "idle on tid %u\n", tid);
#endif /* CONFIG_MAC80211_HT_DEBUG */
		ret = -EAGAIN;
		goto err_unlock_sta;
	}

	/*
	 * While we're asking the driver about the aggregation,
	 * stop the AC queue so that we don't have to worry
	 * about frames that came in while we were doing that,
	 * which would require us to put them to the AC pending
	 * afterwards which just makes the code more complex.
	 */
	ieee80211_stop_queue_by_reason(
		&local->hw, ieee80211_ac_from_tid(tid),
		IEEE80211_QUEUE_STOP_REASON_AGGREGATION);

	/* prepare A-MPDU MLME for Tx aggregation */
	sta->ampdu_mlme.tid_tx[tid] =
			kmalloc(sizeof(struct tid_ampdu_tx), GFP_ATOMIC);
	if (!sta->ampdu_mlme.tid_tx[tid]) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		if (net_ratelimit())
			printk(KERN_ERR "allocate tx mlme to tid %d failed\n",
					tid);
#endif
		ret = -ENOMEM;
		goto err_wake_queue;
	}

	skb_queue_head_init(&sta->ampdu_mlme.tid_tx[tid]->pending);

	/* Tx timer */
	sta->ampdu_mlme.tid_tx[tid]->addba_resp_timer.function =
			sta_addba_resp_timer_expired;
	sta->ampdu_mlme.tid_tx[tid]->addba_resp_timer.data =
			(unsigned long)&sta->timer_to_tid[tid];
	init_timer(&sta->ampdu_mlme.tid_tx[tid]->addba_resp_timer);

	/* Ok, the Addba frame hasn't been sent yet, but if the driver calls the
	 * call back right away, it must see that the flow has begun */
	*state |= HT_ADDBA_REQUESTED_MSK;

	start_seq_num = sta->tid_seq[tid];

	ret = drv_ampdu_action(local, IEEE80211_AMPDU_TX_START,
			       &sta->sta, tid, &start_seq_num);

	if (ret) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "BA request denied - HW unavailable for"
					" tid %d\n", tid);
#endif /* CONFIG_MAC80211_HT_DEBUG */
		*state = HT_AGG_STATE_IDLE;
		goto err_free;
	}

	/* Driver vetoed or OKed, but we can take packets again now */
	ieee80211_wake_queue_by_reason(
		&local->hw, ieee80211_ac_from_tid(tid),
		IEEE80211_QUEUE_STOP_REASON_AGGREGATION);

	spin_unlock(&local->ampdu_lock);
	spin_unlock_bh(&sta->lock);

	/* send an addBA request */
	sta->ampdu_mlme.dialog_token_allocator++;
	sta->ampdu_mlme.tid_tx[tid]->dialog_token =
			sta->ampdu_mlme.dialog_token_allocator;
	sta->ampdu_mlme.tid_tx[tid]->ssn = start_seq_num;

	ieee80211_send_addba_request(sta->sdata, ra, tid,
			 sta->ampdu_mlme.tid_tx[tid]->dialog_token,
			 sta->ampdu_mlme.tid_tx[tid]->ssn,
			 0x40, 5000);
	sta->ampdu_mlme.addba_req_num[tid]++;
	/* activate the timer for the recipient's addBA response */
	sta->ampdu_mlme.tid_tx[tid]->addba_resp_timer.expires =
				jiffies + ADDBA_RESP_INTERVAL;
	add_timer(&sta->ampdu_mlme.tid_tx[tid]->addba_resp_timer);
#ifdef CONFIG_MAC80211_HT_DEBUG
	printk(KERN_DEBUG "activated addBA response timer on tid %d\n", tid);
#endif
	goto unlock;

 err_free:
	kfree(sta->ampdu_mlme.tid_tx[tid]);
	sta->ampdu_mlme.tid_tx[tid] = NULL;
 err_wake_queue:
	ieee80211_wake_queue_by_reason(
		&local->hw, ieee80211_ac_from_tid(tid),
		IEEE80211_QUEUE_STOP_REASON_AGGREGATION);
 err_unlock_sta:
	spin_unlock(&local->ampdu_lock);
	spin_unlock_bh(&sta->lock);
 unlock:
	rcu_read_unlock();
	return ret;
}
EXPORT_SYMBOL(ieee80211_start_tx_ba_session);

/*
 * splice packets from the STA's pending to the local pending,
 * requires a call to ieee80211_agg_splice_finish and holding
 * local->ampdu_lock across both calls.
 */
static void ieee80211_agg_splice_packets(struct ieee80211_local *local,
					 struct sta_info *sta, u16 tid)
{
	unsigned long flags;
	u16 queue = ieee80211_ac_from_tid(tid);

	ieee80211_stop_queue_by_reason(
		&local->hw, queue,
		IEEE80211_QUEUE_STOP_REASON_AGGREGATION);

	if (!(sta->ampdu_mlme.tid_state_tx[tid] & HT_ADDBA_REQUESTED_MSK))
		return;

	if (WARN(!sta->ampdu_mlme.tid_tx[tid],
		 "TID %d gone but expected when splicing aggregates from"
		 "the pending queue\n", tid))
		return;

	if (!skb_queue_empty(&sta->ampdu_mlme.tid_tx[tid]->pending)) {
		spin_lock_irqsave(&local->queue_stop_reason_lock, flags);
		/* copy over remaining packets */
		skb_queue_splice_tail_init(
			&sta->ampdu_mlme.tid_tx[tid]->pending,
			&local->pending[queue]);
		spin_unlock_irqrestore(&local->queue_stop_reason_lock, flags);
	}
}

static void ieee80211_agg_splice_finish(struct ieee80211_local *local,
					struct sta_info *sta, u16 tid)
{
	u16 queue = ieee80211_ac_from_tid(tid);

	ieee80211_wake_queue_by_reason(
		&local->hw, queue,
		IEEE80211_QUEUE_STOP_REASON_AGGREGATION);
}

/* caller must hold sta->lock */
static void ieee80211_agg_tx_operational(struct ieee80211_local *local,
					 struct sta_info *sta, u16 tid)
{
#ifdef CONFIG_MAC80211_HT_DEBUG
	printk(KERN_DEBUG "Aggregation is on for tid %d \n", tid);
#endif

	spin_lock(&local->ampdu_lock);
	ieee80211_agg_splice_packets(local, sta, tid);
	/*
	 * NB: we rely on sta->lock being taken in the TX
	 * processing here when adding to the pending queue,
	 * otherwise we could only change the state of the
	 * session to OPERATIONAL _here_.
	 */
	ieee80211_agg_splice_finish(local, sta, tid);
	spin_unlock(&local->ampdu_lock);

	drv_ampdu_action(local, IEEE80211_AMPDU_TX_OPERATIONAL,
			 &sta->sta, tid, NULL);
}

void ieee80211_start_tx_ba_cb(struct ieee80211_hw *hw, u8 *ra, u16 tid)
{
	struct ieee80211_local *local = hw_to_local(hw);
	struct sta_info *sta;
	u8 *state;

	if (tid >= STA_TID_NUM) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "Bad TID value: tid = %d (>= %d)\n",
				tid, STA_TID_NUM);
#endif
		return;
	}

	rcu_read_lock();
	sta = sta_info_get(local, ra);
	if (!sta) {
		rcu_read_unlock();
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "Could not find station: %pM\n", ra);
#endif
		return;
	}

	state = &sta->ampdu_mlme.tid_state_tx[tid];
	spin_lock_bh(&sta->lock);

	if (WARN_ON(!(*state & HT_ADDBA_REQUESTED_MSK))) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "addBA was not requested yet, state is %d\n",
				*state);
#endif
		spin_unlock_bh(&sta->lock);
		rcu_read_unlock();
		return;
	}

	if (WARN_ON(*state & HT_ADDBA_DRV_READY_MSK))
		goto out;

	*state |= HT_ADDBA_DRV_READY_MSK;

	if (*state == HT_AGG_STATE_OPERATIONAL)
		ieee80211_agg_tx_operational(local, sta, tid);

 out:
	spin_unlock_bh(&sta->lock);
	rcu_read_unlock();
}
EXPORT_SYMBOL(ieee80211_start_tx_ba_cb);

void ieee80211_start_tx_ba_cb_irqsafe(struct ieee80211_hw *hw,
				      const u8 *ra, u16 tid)
{
	struct ieee80211_local *local = hw_to_local(hw);
	struct ieee80211_ra_tid *ra_tid;
	struct sk_buff *skb = dev_alloc_skb(0);

	if (unlikely(!skb)) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		if (net_ratelimit())
			printk(KERN_WARNING "%s: Not enough memory, "
			       "dropping start BA session", skb->dev->name);
#endif
		return;
	}
	ra_tid = (struct ieee80211_ra_tid *) &skb->cb;
	memcpy(&ra_tid->ra, ra, ETH_ALEN);
	ra_tid->tid = tid;

	skb->pkt_type = IEEE80211_ADDBA_MSG;
	skb_queue_tail(&local->skb_queue, skb);
	tasklet_schedule(&local->tasklet);
}
EXPORT_SYMBOL(ieee80211_start_tx_ba_cb_irqsafe);

int __ieee80211_stop_tx_ba_session(struct sta_info *sta, u16 tid,
				   enum ieee80211_back_parties initiator)
{
	u8 *state;
	int ret;

	/* check if the TID is in aggregation */
	state = &sta->ampdu_mlme.tid_state_tx[tid];
	spin_lock_bh(&sta->lock);

	if (*state != HT_AGG_STATE_OPERATIONAL) {
		ret = -ENOENT;
		goto unlock;
	}

#ifdef CONFIG_MAC80211_HT_DEBUG
	printk(KERN_DEBUG "Tx BA session stop requested for %pM tid %u\n",
	       sta->sta.addr, tid);
#endif /* CONFIG_MAC80211_HT_DEBUG */

	ret = ___ieee80211_stop_tx_ba_session(sta, tid, initiator);

 unlock:
	spin_unlock_bh(&sta->lock);
	return ret;
}

int ieee80211_stop_tx_ba_session(struct ieee80211_hw *hw,
				 u8 *ra, u16 tid,
				 enum ieee80211_back_parties initiator)
{
	struct ieee80211_local *local = hw_to_local(hw);
	struct sta_info *sta;
	int ret = 0;

	if (WARN_ON(!local->ops->ampdu_action))
		return -EINVAL;

	if (tid >= STA_TID_NUM)
		return -EINVAL;

	rcu_read_lock();
	sta = sta_info_get(local, ra);
	if (!sta) {
		rcu_read_unlock();
		return -ENOENT;
	}

	ret = __ieee80211_stop_tx_ba_session(sta, tid, initiator);
	rcu_read_unlock();
	return ret;
}
EXPORT_SYMBOL(ieee80211_stop_tx_ba_session);

void ieee80211_stop_tx_ba_cb(struct ieee80211_hw *hw, u8 *ra, u8 tid)
{
	struct ieee80211_local *local = hw_to_local(hw);
	struct sta_info *sta;
	u8 *state;

	if (tid >= STA_TID_NUM) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "Bad TID value: tid = %d (>= %d)\n",
				tid, STA_TID_NUM);
#endif
		return;
	}

#ifdef CONFIG_MAC80211_HT_DEBUG
	printk(KERN_DEBUG "Stopping Tx BA session for %pM tid %d\n",
	       ra, tid);
#endif /* CONFIG_MAC80211_HT_DEBUG */

	rcu_read_lock();
	sta = sta_info_get(local, ra);
	if (!sta) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "Could not find station: %pM\n", ra);
#endif
		rcu_read_unlock();
		return;
	}
	state = &sta->ampdu_mlme.tid_state_tx[tid];

	/* NOTE: no need to use sta->lock in this state check, as
	 * ieee80211_stop_tx_ba_session will let only one stop call to
	 * pass through per sta/tid
	 */
	if ((*state & HT_AGG_STATE_REQ_STOP_BA_MSK) == 0) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "unexpected callback to A-MPDU stop\n");
#endif
		rcu_read_unlock();
		return;
	}

	if (*state & HT_AGG_STATE_INITIATOR_MSK)
		ieee80211_send_delba(sta->sdata, ra, tid,
			WLAN_BACK_INITIATOR, WLAN_REASON_QSTA_NOT_USE);

	spin_lock_bh(&sta->lock);
	spin_lock(&local->ampdu_lock);

	ieee80211_agg_splice_packets(local, sta, tid);

	*state = HT_AGG_STATE_IDLE;
	/* from now on packets are no longer put onto sta->pending */
	kfree(sta->ampdu_mlme.tid_tx[tid]);
	sta->ampdu_mlme.tid_tx[tid] = NULL;

	ieee80211_agg_splice_finish(local, sta, tid);

	spin_unlock(&local->ampdu_lock);
	spin_unlock_bh(&sta->lock);

	rcu_read_unlock();
}
EXPORT_SYMBOL(ieee80211_stop_tx_ba_cb);

void ieee80211_stop_tx_ba_cb_irqsafe(struct ieee80211_hw *hw,
				     const u8 *ra, u16 tid)
{
	struct ieee80211_local *local = hw_to_local(hw);
	struct ieee80211_ra_tid *ra_tid;
	struct sk_buff *skb = dev_alloc_skb(0);

	if (unlikely(!skb)) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		if (net_ratelimit())
			printk(KERN_WARNING "%s: Not enough memory, "
			       "dropping stop BA session", skb->dev->name);
#endif
		return;
	}
	ra_tid = (struct ieee80211_ra_tid *) &skb->cb;
	memcpy(&ra_tid->ra, ra, ETH_ALEN);
	ra_tid->tid = tid;

	skb->pkt_type = IEEE80211_DELBA_MSG;
	skb_queue_tail(&local->skb_queue, skb);
	tasklet_schedule(&local->tasklet);
}
EXPORT_SYMBOL(ieee80211_stop_tx_ba_cb_irqsafe);


void ieee80211_process_addba_resp(struct ieee80211_local *local,
				  struct sta_info *sta,
				  struct ieee80211_mgmt *mgmt,
				  size_t len)
{
	u16 capab, tid;
	u8 *state;

	capab = le16_to_cpu(mgmt->u.action.u.addba_resp.capab);
	tid = (capab & IEEE80211_ADDBA_PARAM_TID_MASK) >> 2;

	state = &sta->ampdu_mlme.tid_state_tx[tid];

	del_timer_sync(&sta->ampdu_mlme.tid_tx[tid]->addba_resp_timer);

	spin_lock_bh(&sta->lock);

	if (!(*state & HT_ADDBA_REQUESTED_MSK))
		goto timer_still_needed;

	if (mgmt->u.action.u.addba_resp.dialog_token !=
		sta->ampdu_mlme.tid_tx[tid]->dialog_token) {
#ifdef CONFIG_MAC80211_HT_DEBUG
		printk(KERN_DEBUG "wrong addBA response token, tid %d\n", tid);
#endif /* CONFIG_MAC80211_HT_DEBUG */
		goto timer_still_needed;
	}

#ifdef CONFIG_MAC80211_HT_DEBUG
	printk(KERN_DEBUG "switched off addBA timer for tid %d \n", tid);
#endif /* CONFIG_MAC80211_HT_DEBUG */

	if (le16_to_cpu(mgmt->u.action.u.addba_resp.status)
			== WLAN_STATUS_SUCCESS) {
		u8 curstate = *state;

		*state |= HT_ADDBA_RECEIVED_MSK;

		if (*state != curstate && *state == HT_AGG_STATE_OPERATIONAL)
			ieee80211_agg_tx_operational(local, sta, tid);

		sta->ampdu_mlme.addba_req_num[tid] = 0;
	} else {
		___ieee80211_stop_tx_ba_session(sta, tid, WLAN_BACK_INITIATOR);
	}

	goto out;

 timer_still_needed:
	add_timer(&sta->ampdu_mlme.tid_tx[tid]->addba_resp_timer);
 out:
	spin_unlock_bh(&sta->lock);
}
ef='#n2742'>2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371
/* cassini.c: Sun Microsystems Cassini(+) ethernet driver.
 *
 * Copyright (C) 2004 Sun Microsystems Inc.
 * Copyright (C) 2003 Adrian Sun (asun@darksunrising.com)
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation; either version 2 of the
 * License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
 * 02111-1307, USA.
 *
 * This driver uses the sungem driver (c) David Miller
 * (davem@redhat.com) as its basis.
 *
 * The cassini chip has a number of features that distinguish it from
 * the gem chip:
 *  4 transmit descriptor rings that are used for either QoS (VLAN) or
 *      load balancing (non-VLAN mode)
 *  batching of multiple packets
 *  multiple CPU dispatching
 *  page-based RX descriptor engine with separate completion rings
 *  Gigabit support (GMII and PCS interface)
 *  MIF link up/down detection works
 *
 * RX is handled by page sized buffers that are attached as fragments to
 * the skb. here's what's done:
 *  -- driver allocates pages at a time and keeps reference counts
 *     on them.
 *  -- the upper protocol layers assume that the header is in the skb
 *     itself. as a result, cassini will copy a small amount (64 bytes)
 *     to make them happy.
 *  -- driver appends the rest of the data pages as frags to skbuffs
 *     and increments the reference count
 *  -- on page reclamation, the driver swaps the page with a spare page.
 *     if that page is still in use, it frees its reference to that page,
 *     and allocates a new page for use. otherwise, it just recycles the
 *     the page.
 *
 * NOTE: cassini can parse the header. however, it's not worth it
 *       as long as the network stack requires a header copy.
 *
 * TX has 4 queues. currently these queues are used in a round-robin
 * fashion for load balancing. They can also be used for QoS. for that
 * to work, however, QoS information needs to be exposed down to the driver
 * level so that subqueues get targetted to particular transmit rings.
 * alternatively, the queues can be configured via use of the all-purpose
 * ioctl.
 *
 * RX DATA: the rx completion ring has all the info, but the rx desc
 * ring has all of the data. RX can conceivably come in under multiple
 * interrupts, but the INT# assignment needs to be set up properly by
 * the BIOS and conveyed to the driver. PCI BIOSes don't know how to do
 * that. also, the two descriptor rings are designed to distinguish between
 * encrypted and non-encrypted packets, but we use them for buffering
 * instead.
 *
 * by default, the selective clear mask is set up to process rx packets.
 */


#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/types.h>
#include <linux/compiler.h>
#include <linux/slab.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/vmalloc.h>
#include <linux/ioport.h>
#include <linux/pci.h>
#include <linux/mm.h>
#include <linux/highmem.h>
#include <linux/list.h>
#include <linux/dma-mapping.h>

#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <linux/ethtool.h>
#include <linux/crc32.h>
#include <linux/random.h>
#include <linux/mii.h>
#include <linux/ip.h>
#include <linux/tcp.h>
#include <linux/mutex.h>
#include <linux/firmware.h>

#include <net/checksum.h>

#include <asm/atomic.h>
#include <asm/system.h>
#include <asm/io.h>
#include <asm/byteorder.h>
#include <asm/uaccess.h>

#define cas_page_map(x)      kmap_atomic((x), KM_SKB_DATA_SOFTIRQ)
#define cas_page_unmap(x)    kunmap_atomic((x), KM_SKB_DATA_SOFTIRQ)
#define CAS_NCPUS            num_online_cpus()

#if defined(CONFIG_CASSINI_NAPI) && defined(HAVE_NETDEV_POLL)
#define USE_NAPI
#define cas_skb_release(x)  netif_receive_skb(x)
#else
#define cas_skb_release(x)  netif_rx(x)
#endif

/* select which firmware to use */
#define USE_HP_WORKAROUND
#define HP_WORKAROUND_DEFAULT /* select which firmware to use as default */
#define CAS_HP_ALT_FIRMWARE   cas_prog_null /* alternate firmware */

#include "cassini.h"

#define USE_TX_COMPWB      /* use completion writeback registers */
#define USE_CSMA_CD_PROTO  /* standard CSMA/CD */
#define USE_RX_BLANK       /* hw interrupt mitigation */
#undef USE_ENTROPY_DEV     /* don't test for entropy device */

/* NOTE: these aren't useable unless PCI interrupts can be assigned.
 * also, we need to make cp->lock finer-grained.
 */
#undef  USE_PCI_INTB
#undef  USE_PCI_INTC
#undef  USE_PCI_INTD
#undef  USE_QOS

#undef  USE_VPD_DEBUG       /* debug vpd information if defined */

/* rx processing options */
#define USE_PAGE_ORDER      /* specify to allocate large rx pages */
#define RX_DONT_BATCH  0    /* if 1, don't batch flows */
#define RX_COPY_ALWAYS 0    /* if 0, use frags */
#define RX_COPY_MIN    64   /* copy a little to make upper layers happy */
#undef  RX_COUNT_BUFFERS    /* define to calculate RX buffer stats */

#define DRV_MODULE_NAME		"cassini"
#define PFX DRV_MODULE_NAME	": "
#define DRV_MODULE_VERSION	"1.6"
#define DRV_MODULE_RELDATE	"21 May 2008"

#define CAS_DEF_MSG_ENABLE	  \
	(NETIF_MSG_DRV		| \
	 NETIF_MSG_PROBE	| \
	 NETIF_MSG_LINK		| \
	 NETIF_MSG_TIMER	| \
	 NETIF_MSG_IFDOWN	| \
	 NETIF_MSG_IFUP		| \
	 NETIF_MSG_RX_ERR	| \
	 NETIF_MSG_TX_ERR)

/* length of time before we decide the hardware is borked,
 * and dev->tx_timeout() should be called to fix the problem
 */
#define CAS_TX_TIMEOUT			(HZ)
#define CAS_LINK_TIMEOUT                (22*HZ/10)
#define CAS_LINK_FAST_TIMEOUT           (1)

/* timeout values for state changing. these specify the number
 * of 10us delays to be used before giving up.
 */
#define STOP_TRIES_PHY 1000
#define STOP_TRIES     5000

/* specify a minimum frame size to deal with some fifo issues
 * max mtu == 2 * page size - ethernet header - 64 - swivel =
 *            2 * page_size - 0x50
 */
#define CAS_MIN_FRAME			97
#define CAS_1000MB_MIN_FRAME            255
#define CAS_MIN_MTU                     60
#define CAS_MAX_MTU                     min(((cp->page_size << 1) - 0x50), 9000)

#if 1
/*
 * Eliminate these and use separate atomic counters for each, to
 * avoid a race condition.
 */
#else
#define CAS_RESET_MTU                   1
#define CAS_RESET_ALL                   2
#define CAS_RESET_SPARE                 3
#endif

static char version[] __devinitdata =
	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";

static int cassini_debug = -1;	/* -1 == use CAS_DEF_MSG_ENABLE as value */
static int link_mode;

MODULE_AUTHOR("Adrian Sun (asun@darksunrising.com)");
MODULE_DESCRIPTION("Sun Cassini(+) ethernet driver");
MODULE_LICENSE("GPL");
MODULE_FIRMWARE("sun/cassini.bin");
module_param(cassini_debug, int, 0);
MODULE_PARM_DESC(cassini_debug, "Cassini bitmapped debugging message enable value");
module_param(link_mode, int, 0);
MODULE_PARM_DESC(link_mode, "default link mode");

/*
 * Work around for a PCS bug in which the link goes down due to the chip
 * being confused and never showing a link status of "up."
 */
#define DEFAULT_LINKDOWN_TIMEOUT 5
/*
 * Value in seconds, for user input.
 */
static int linkdown_timeout = DEFAULT_LINKDOWN_TIMEOUT;
module_param(linkdown_timeout, int, 0);
MODULE_PARM_DESC(linkdown_timeout,
"min reset interval in sec. for PCS linkdown issue; disabled if not positive");

/*
 * value in 'ticks' (units used by jiffies). Set when we init the
 * module because 'HZ' in actually a function call on some flavors of
 * Linux.  This will default to DEFAULT_LINKDOWN_TIMEOUT * HZ.
 */
static int link_transition_timeout;



static u16 link_modes[] __devinitdata = {
	BMCR_ANENABLE,			 /* 0 : autoneg */
	0,				 /* 1 : 10bt half duplex */
	BMCR_SPEED100,			 /* 2 : 100bt half duplex */
	BMCR_FULLDPLX,			 /* 3 : 10bt full duplex */
	BMCR_SPEED100|BMCR_FULLDPLX,	 /* 4 : 100bt full duplex */
	CAS_BMCR_SPEED1000|BMCR_FULLDPLX /* 5 : 1000bt full duplex */
};

static struct pci_device_id cas_pci_tbl[] __devinitdata = {
	{ PCI_VENDOR_ID_SUN, PCI_DEVICE_ID_SUN_CASSINI,
	  PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0UL },
	{ PCI_VENDOR_ID_NS, PCI_DEVICE_ID_NS_SATURN,
	  PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0UL },
	{ 0, }
};

MODULE_DEVICE_TABLE(pci, cas_pci_tbl);

static void cas_set_link_modes(struct cas *cp);

static inline void cas_lock_tx(struct cas *cp)
{
	int i;

	for (i = 0; i < N_TX_RINGS; i++)
		spin_lock(&cp->tx_lock[i]);
}

static inline void cas_lock_all(struct cas *cp)
{
	spin_lock_irq(&cp->lock);
	cas_lock_tx(cp);
}

/* WTZ: QA was finding deadlock problems with the previous
 * versions after long test runs with multiple cards per machine.
 * See if replacing cas_lock_all with safer versions helps. The
 * symptoms QA is reporting match those we'd expect if interrupts
 * aren't being properly restored, and we fixed a previous deadlock
 * with similar symptoms by using save/restore versions in other
 * places.
 */
#define cas_lock_all_save(cp, flags) \
do { \
	struct cas *xxxcp = (cp); \
	spin_lock_irqsave(&xxxcp->lock, flags); \
	cas_lock_tx(xxxcp); \
} while (0)

static inline void cas_unlock_tx(struct cas *cp)
{
	int i;

	for (i = N_TX_RINGS; i > 0; i--)
		spin_unlock(&cp->tx_lock[i - 1]);
}

static inline void cas_unlock_all(struct cas *cp)
{
	cas_unlock_tx(cp);
	spin_unlock_irq(&cp->lock);
}

#define cas_unlock_all_restore(cp, flags) \
do { \
	struct cas *xxxcp = (cp); \
	cas_unlock_tx(xxxcp); \
	spin_unlock_irqrestore(&xxxcp->lock, flags); \
} while (0)

static void cas_disable_irq(struct cas *cp, const int ring)
{
	/* Make sure we won't get any more interrupts */
	if (ring == 0) {
		writel(0xFFFFFFFF, cp->regs + REG_INTR_MASK);
		return;
	}

	/* disable completion interrupts and selectively mask */
	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
		switch (ring) {
#if defined (USE_PCI_INTB) || defined(USE_PCI_INTC) || defined(USE_PCI_INTD)
#ifdef USE_PCI_INTB
		case 1:
#endif
#ifdef USE_PCI_INTC
		case 2:
#endif
#ifdef USE_PCI_INTD
		case 3:
#endif
			writel(INTRN_MASK_CLEAR_ALL | INTRN_MASK_RX_EN,
			       cp->regs + REG_PLUS_INTRN_MASK(ring));
			break;
#endif
		default:
			writel(INTRN_MASK_CLEAR_ALL, cp->regs +
			       REG_PLUS_INTRN_MASK(ring));
			break;
		}
	}
}

static inline void cas_mask_intr(struct cas *cp)
{
	int i;

	for (i = 0; i < N_RX_COMP_RINGS; i++)
		cas_disable_irq(cp, i);
}

static void cas_enable_irq(struct cas *cp, const int ring)
{
	if (ring == 0) { /* all but TX_DONE */
		writel(INTR_TX_DONE, cp->regs + REG_INTR_MASK);
		return;
	}

	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
		switch (ring) {
#if defined (USE_PCI_INTB) || defined(USE_PCI_INTC) || defined(USE_PCI_INTD)
#ifdef USE_PCI_INTB
		case 1:
#endif
#ifdef USE_PCI_INTC
		case 2:
#endif
#ifdef USE_PCI_INTD
		case 3:
#endif
			writel(INTRN_MASK_RX_EN, cp->regs +
			       REG_PLUS_INTRN_MASK(ring));
			break;
#endif
		default:
			break;
		}
	}
}

static inline void cas_unmask_intr(struct cas *cp)
{
	int i;

	for (i = 0; i < N_RX_COMP_RINGS; i++)
		cas_enable_irq(cp, i);
}

static inline void cas_entropy_gather(struct cas *cp)
{
#ifdef USE_ENTROPY_DEV
	if ((cp->cas_flags & CAS_FLAG_ENTROPY_DEV) == 0)
		return;

	batch_entropy_store(readl(cp->regs + REG_ENTROPY_IV),
			    readl(cp->regs + REG_ENTROPY_IV),
			    sizeof(uint64_t)*8);
#endif
}

static inline void cas_entropy_reset(struct cas *cp)
{
#ifdef USE_ENTROPY_DEV
	if ((cp->cas_flags & CAS_FLAG_ENTROPY_DEV) == 0)
		return;

	writel(BIM_LOCAL_DEV_PAD | BIM_LOCAL_DEV_PROM | BIM_LOCAL_DEV_EXT,
	       cp->regs + REG_BIM_LOCAL_DEV_EN);
	writeb(ENTROPY_RESET_STC_MODE, cp->regs + REG_ENTROPY_RESET);
	writeb(0x55, cp->regs + REG_ENTROPY_RAND_REG);

	/* if we read back 0x0, we don't have an entropy device */
	if (readb(cp->regs + REG_ENTROPY_RAND_REG) == 0)
		cp->cas_flags &= ~CAS_FLAG_ENTROPY_DEV;
#endif
}

/* access to the phy. the following assumes that we've initialized the MIF to
 * be in frame rather than bit-bang mode
 */
static u16 cas_phy_read(struct cas *cp, int reg)
{
	u32 cmd;
	int limit = STOP_TRIES_PHY;

	cmd = MIF_FRAME_ST | MIF_FRAME_OP_READ;
	cmd |= CAS_BASE(MIF_FRAME_PHY_ADDR, cp->phy_addr);
	cmd |= CAS_BASE(MIF_FRAME_REG_ADDR, reg);
	cmd |= MIF_FRAME_TURN_AROUND_MSB;
	writel(cmd, cp->regs + REG_MIF_FRAME);

	/* poll for completion */
	while (limit-- > 0) {
		udelay(10);
		cmd = readl(cp->regs + REG_MIF_FRAME);
		if (cmd & MIF_FRAME_TURN_AROUND_LSB)
			return (cmd & MIF_FRAME_DATA_MASK);
	}
	return 0xFFFF; /* -1 */
}

static int cas_phy_write(struct cas *cp, int reg, u16 val)
{
	int limit = STOP_TRIES_PHY;
	u32 cmd;

	cmd = MIF_FRAME_ST | MIF_FRAME_OP_WRITE;
	cmd |= CAS_BASE(MIF_FRAME_PHY_ADDR, cp->phy_addr);
	cmd |= CAS_BASE(MIF_FRAME_REG_ADDR, reg);
	cmd |= MIF_FRAME_TURN_AROUND_MSB;
	cmd |= val & MIF_FRAME_DATA_MASK;
	writel(cmd, cp->regs + REG_MIF_FRAME);

	/* poll for completion */
	while (limit-- > 0) {
		udelay(10);
		cmd = readl(cp->regs + REG_MIF_FRAME);
		if (cmd & MIF_FRAME_TURN_AROUND_LSB)
			return 0;
	}
	return -1;
}

static void cas_phy_powerup(struct cas *cp)
{
	u16 ctl = cas_phy_read(cp, MII_BMCR);

	if ((ctl & BMCR_PDOWN) == 0)
		return;
	ctl &= ~BMCR_PDOWN;
	cas_phy_write(cp, MII_BMCR, ctl);
}

static void cas_phy_powerdown(struct cas *cp)
{
	u16 ctl = cas_phy_read(cp, MII_BMCR);

	if (ctl & BMCR_PDOWN)
		return;
	ctl |= BMCR_PDOWN;
	cas_phy_write(cp, MII_BMCR, ctl);
}

/* cp->lock held. note: the last put_page will free the buffer */
static int cas_page_free(struct cas *cp, cas_page_t *page)
{
	pci_unmap_page(cp->pdev, page->dma_addr, cp->page_size,
		       PCI_DMA_FROMDEVICE);
	__free_pages(page->buffer, cp->page_order);
	kfree(page);
	return 0;
}

#ifdef RX_COUNT_BUFFERS
#define RX_USED_ADD(x, y)       ((x)->used += (y))
#define RX_USED_SET(x, y)       ((x)->used  = (y))
#else
#define RX_USED_ADD(x, y)
#define RX_USED_SET(x, y)
#endif

/* local page allocation routines for the receive buffers. jumbo pages
 * require at least 8K contiguous and 8K aligned buffers.
 */
static cas_page_t *cas_page_alloc(struct cas *cp, const gfp_t flags)
{
	cas_page_t *page;

	page = kmalloc(sizeof(cas_page_t), flags);
	if (!page)
		return NULL;

	INIT_LIST_HEAD(&page->list);
	RX_USED_SET(page, 0);
	page->buffer = alloc_pages(flags, cp->page_order);
	if (!page->buffer)
		goto page_err;
	page->dma_addr = pci_map_page(cp->pdev, page->buffer, 0,
				      cp->page_size, PCI_DMA_FROMDEVICE);
	return page;

page_err:
	kfree(page);
	return NULL;
}

/* initialize spare pool of rx buffers, but allocate during the open */
static void cas_spare_init(struct cas *cp)
{
  	spin_lock(&cp->rx_inuse_lock);
	INIT_LIST_HEAD(&cp->rx_inuse_list);
	spin_unlock(&cp->rx_inuse_lock);

	spin_lock(&cp->rx_spare_lock);
	INIT_LIST_HEAD(&cp->rx_spare_list);
	cp->rx_spares_needed = RX_SPARE_COUNT;
	spin_unlock(&cp->rx_spare_lock);
}

/* used on close. free all the spare buffers. */
static void cas_spare_free(struct cas *cp)
{
	struct list_head list, *elem, *tmp;

	/* free spare buffers */
	INIT_LIST_HEAD(&list);
	spin_lock(&cp->rx_spare_lock);
	list_splice_init(&cp->rx_spare_list, &list);
	spin_unlock(&cp->rx_spare_lock);
	list_for_each_safe(elem, tmp, &list) {
		cas_page_free(cp, list_entry(elem, cas_page_t, list));
	}

	INIT_LIST_HEAD(&list);
#if 1
	/*
	 * Looks like Adrian had protected this with a different
	 * lock than used everywhere else to manipulate this list.
	 */
	spin_lock(&cp->rx_inuse_lock);
	list_splice_init(&cp->rx_inuse_list, &list);
	spin_unlock(&cp->rx_inuse_lock);
#else
	spin_lock(&cp->rx_spare_lock);
	list_splice_init(&cp->rx_inuse_list, &list);
	spin_unlock(&cp->rx_spare_lock);
#endif
	list_for_each_safe(elem, tmp, &list) {
		cas_page_free(cp, list_entry(elem, cas_page_t, list));
	}
}

/* replenish spares if needed */
static void cas_spare_recover(struct cas *cp, const gfp_t flags)
{
	struct list_head list, *elem, *tmp;
	int needed, i;

	/* check inuse list. if we don't need any more free buffers,
	 * just free it
	 */

	/* make a local copy of the list */
	INIT_LIST_HEAD(&list);
	spin_lock(&cp->rx_inuse_lock);
	list_splice_init(&cp->rx_inuse_list, &list);
	spin_unlock(&cp->rx_inuse_lock);

	list_for_each_safe(elem, tmp, &list) {
		cas_page_t *page = list_entry(elem, cas_page_t, list);

		/*
		 * With the lockless pagecache, cassini buffering scheme gets
		 * slightly less accurate: we might find that a page has an
		 * elevated reference count here, due to a speculative ref,
		 * and skip it as in-use. Ideally we would be able to reclaim
		 * it. However this would be such a rare case, it doesn't
		 * matter too much as we should pick it up the next time round.
		 *
		 * Importantly, if we find that the page has a refcount of 1
		 * here (our refcount), then we know it is definitely not inuse
		 * so we can reuse it.
		 */
		if (page_count(page->buffer) > 1)
			continue;

		list_del(elem);
		spin_lock(&cp->rx_spare_lock);
		if (cp->rx_spares_needed > 0) {
			list_add(elem, &cp->rx_spare_list);
			cp->rx_spares_needed--;
			spin_unlock(&cp->rx_spare_lock);
		} else {
			spin_unlock(&cp->rx_spare_lock);
			cas_page_free(cp, page);
		}
	}

	/* put any inuse buffers back on the list */
	if (!list_empty(&list)) {
		spin_lock(&cp->rx_inuse_lock);
		list_splice(&list, &cp->rx_inuse_list);
		spin_unlock(&cp->rx_inuse_lock);
	}

	spin_lock(&cp->rx_spare_lock);
	needed = cp->rx_spares_needed;
	spin_unlock(&cp->rx_spare_lock);
	if (!needed)
		return;

	/* we still need spares, so try to allocate some */
	INIT_LIST_HEAD(&list);
	i = 0;
	while (i < needed) {
		cas_page_t *spare = cas_page_alloc(cp, flags);
		if (!spare)
			break;
		list_add(&spare->list, &list);
		i++;
	}

	spin_lock(&cp->rx_spare_lock);
	list_splice(&list, &cp->rx_spare_list);
	cp->rx_spares_needed -= i;
	spin_unlock(&cp->rx_spare_lock);
}

/* pull a page from the list. */
static cas_page_t *cas_page_dequeue(struct cas *cp)
{
	struct list_head *entry;
	int recover;

	spin_lock(&cp->rx_spare_lock);
	if (list_empty(&cp->rx_spare_list)) {
		/* try to do a quick recovery */
		spin_unlock(&cp->rx_spare_lock);
		cas_spare_recover(cp, GFP_ATOMIC);
		spin_lock(&cp->rx_spare_lock);
		if (list_empty(&cp->rx_spare_list)) {
			if (netif_msg_rx_err(cp))
				printk(KERN_ERR "%s: no spare buffers "
				       "available.\n", cp->dev->name);
			spin_unlock(&cp->rx_spare_lock);
			return NULL;
		}
	}

	entry = cp->rx_spare_list.next;
	list_del(entry);
	recover = ++cp->rx_spares_needed;
	spin_unlock(&cp->rx_spare_lock);

	/* trigger the timer to do the recovery */
	if ((recover & (RX_SPARE_RECOVER_VAL - 1)) == 0) {
#if 1
		atomic_inc(&cp->reset_task_pending);
		atomic_inc(&cp->reset_task_pending_spare);
		schedule_work(&cp->reset_task);
#else
		atomic_set(&cp->reset_task_pending, CAS_RESET_SPARE);
		schedule_work(&cp->reset_task);
#endif
	}
	return list_entry(entry, cas_page_t, list);
}


static void cas_mif_poll(struct cas *cp, const int enable)
{
	u32 cfg;

	cfg  = readl(cp->regs + REG_MIF_CFG);
	cfg &= (MIF_CFG_MDIO_0 | MIF_CFG_MDIO_1);

	if (cp->phy_type & CAS_PHY_MII_MDIO1)
		cfg |= MIF_CFG_PHY_SELECT;

	/* poll and interrupt on link status change. */
	if (enable) {
		cfg |= MIF_CFG_POLL_EN;
		cfg |= CAS_BASE(MIF_CFG_POLL_REG, MII_BMSR);
		cfg |= CAS_BASE(MIF_CFG_POLL_PHY, cp->phy_addr);
	}
	writel((enable) ? ~(BMSR_LSTATUS | BMSR_ANEGCOMPLETE) : 0xFFFF,
	       cp->regs + REG_MIF_MASK);
	writel(cfg, cp->regs + REG_MIF_CFG);
}

/* Must be invoked under cp->lock */
static void cas_begin_auto_negotiation(struct cas *cp, struct ethtool_cmd *ep)
{
	u16 ctl;
#if 1
	int lcntl;
	int changed = 0;
	int oldstate = cp->lstate;
	int link_was_not_down = !(oldstate == link_down);
#endif
	/* Setup link parameters */
	if (!ep)
		goto start_aneg;
	lcntl = cp->link_cntl;
	if (ep->autoneg == AUTONEG_ENABLE)
		cp->link_cntl = BMCR_ANENABLE;
	else {
		cp->link_cntl = 0;
		if (ep->speed == SPEED_100)
			cp->link_cntl |= BMCR_SPEED100;
		else if (ep->speed == SPEED_1000)
			cp->link_cntl |= CAS_BMCR_SPEED1000;
		if (ep->duplex == DUPLEX_FULL)
			cp->link_cntl |= BMCR_FULLDPLX;
	}
#if 1
	changed = (lcntl != cp->link_cntl);
#endif
start_aneg:
	if (cp->lstate == link_up) {
		printk(KERN_INFO "%s: PCS link down.\n",
		       cp->dev->name);
	} else {
		if (changed) {
			printk(KERN_INFO "%s: link configuration changed\n",
			       cp->dev->name);
		}
	}
	cp->lstate = link_down;
	cp->link_transition = LINK_TRANSITION_LINK_DOWN;
	if (!cp->hw_running)
		return;
#if 1
	/*
	 * WTZ: If the old state was link_up, we turn off the carrier
	 * to replicate everything we do elsewhere on a link-down
	 * event when we were already in a link-up state..
	 */
	if (oldstate == link_up)
		netif_carrier_off(cp->dev);
	if (changed  && link_was_not_down) {
		/*
		 * WTZ: This branch will simply schedule a full reset after
		 * we explicitly changed link modes in an ioctl. See if this
		 * fixes the link-problems we were having for forced mode.
		 */
		atomic_inc(&cp->reset_task_pending);
		atomic_inc(&cp->reset_task_pending_all);
		schedule_work(&cp->reset_task);
		cp->timer_ticks = 0;
		mod_timer(&cp->link_timer, jiffies + CAS_LINK_TIMEOUT);
		return;
	}
#endif
	if (cp->phy_type & CAS_PHY_SERDES) {
		u32 val = readl(cp->regs + REG_PCS_MII_CTRL);

		if (cp->link_cntl & BMCR_ANENABLE) {
			val |= (PCS_MII_RESTART_AUTONEG | PCS_MII_AUTONEG_EN);
			cp->lstate = link_aneg;
		} else {
			if (cp->link_cntl & BMCR_FULLDPLX)
				val |= PCS_MII_CTRL_DUPLEX;
			val &= ~PCS_MII_AUTONEG_EN;
			cp->lstate = link_force_ok;
		}
		cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
		writel(val, cp->regs + REG_PCS_MII_CTRL);

	} else {
		cas_mif_poll(cp, 0);
		ctl = cas_phy_read(cp, MII_BMCR);
		ctl &= ~(BMCR_FULLDPLX | BMCR_SPEED100 |
			 CAS_BMCR_SPEED1000 | BMCR_ANENABLE);
		ctl |= cp->link_cntl;
		if (ctl & BMCR_ANENABLE) {
			ctl |= BMCR_ANRESTART;
			cp->lstate = link_aneg;
		} else {
			cp->lstate = link_force_ok;
		}
		cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
		cas_phy_write(cp, MII_BMCR, ctl);
		cas_mif_poll(cp, 1);
	}

	cp->timer_ticks = 0;
	mod_timer(&cp->link_timer, jiffies + CAS_LINK_TIMEOUT);
}

/* Must be invoked under cp->lock. */
static int cas_reset_mii_phy(struct cas *cp)
{
	int limit = STOP_TRIES_PHY;
	u16 val;

	cas_phy_write(cp, MII_BMCR, BMCR_RESET);
	udelay(100);
	while (--limit) {
		val = cas_phy_read(cp, MII_BMCR);
		if ((val & BMCR_RESET) == 0)
			break;
		udelay(10);
	}
	return (limit <= 0);
}

static int cas_saturn_firmware_init(struct cas *cp)
{
	const struct firmware *fw;
	const char fw_name[] = "sun/cassini.bin";
	int err;

	if (PHY_NS_DP83065 != cp->phy_id)
		return 0;

	err = request_firmware(&fw, fw_name, &cp->pdev->dev);
	if (err) {
		printk(KERN_ERR "cassini: Failed to load firmware \"%s\"\n",
		       fw_name);
		return err;
	}
	if (fw->size < 2) {
		printk(KERN_ERR "cassini: bogus length %zu in \"%s\"\n",
		       fw->size, fw_name);
		err = -EINVAL;
		goto out;
	}
	cp->fw_load_addr= fw->data[1] << 8 | fw->data[0];
	cp->fw_size = fw->size - 2;
	cp->fw_data = vmalloc(cp->fw_size);
	if (!cp->fw_data) {
		err = -ENOMEM;
		printk(KERN_ERR "cassini: \"%s\" Failed %d\n", fw_name, err);
		goto out;
	}
	memcpy(cp->fw_data, &fw->data[2], cp->fw_size);
out:
	release_firmware(fw);
	return err;
}

static void cas_saturn_firmware_load(struct cas *cp)
{
	int i;

	cas_phy_powerdown(cp);

	/* expanded memory access mode */
	cas_phy_write(cp, DP83065_MII_MEM, 0x0);

	/* pointer configuration for new firmware */
	cas_phy_write(cp, DP83065_MII_REGE, 0x8ff9);
	cas_phy_write(cp, DP83065_MII_REGD, 0xbd);
	cas_phy_write(cp, DP83065_MII_REGE, 0x8ffa);
	cas_phy_write(cp, DP83065_MII_REGD, 0x82);
	cas_phy_write(cp, DP83065_MII_REGE, 0x8ffb);
	cas_phy_write(cp, DP83065_MII_REGD, 0x0);
	cas_phy_write(cp, DP83065_MII_REGE, 0x8ffc);
	cas_phy_write(cp, DP83065_MII_REGD, 0x39);

	/* download new firmware */
	cas_phy_write(cp, DP83065_MII_MEM, 0x1);
	cas_phy_write(cp, DP83065_MII_REGE, cp->fw_load_addr);
	for (i = 0; i < cp->fw_size; i++)
		cas_phy_write(cp, DP83065_MII_REGD, cp->fw_data[i]);

	/* enable firmware */
	cas_phy_write(cp, DP83065_MII_REGE, 0x8ff8);
	cas_phy_write(cp, DP83065_MII_REGD, 0x1);
}


/* phy initialization */
static void cas_phy_init(struct cas *cp)
{
	u16 val;

	/* if we're in MII/GMII mode, set up phy */
	if (CAS_PHY_MII(cp->phy_type)) {
		writel(PCS_DATAPATH_MODE_MII,
		       cp->regs + REG_PCS_DATAPATH_MODE);

		cas_mif_poll(cp, 0);
		cas_reset_mii_phy(cp); /* take out of isolate mode */

		if (PHY_LUCENT_B0 == cp->phy_id) {
			/* workaround link up/down issue with lucent */
			cas_phy_write(cp, LUCENT_MII_REG, 0x8000);
			cas_phy_write(cp, MII_BMCR, 0x00f1);
			cas_phy_write(cp, LUCENT_MII_REG, 0x0);

		} else if (PHY_BROADCOM_B0 == (cp->phy_id & 0xFFFFFFFC)) {
			/* workarounds for broadcom phy */
			cas_phy_write(cp, BROADCOM_MII_REG8, 0x0C20);
			cas_phy_write(cp, BROADCOM_MII_REG7, 0x0012);
			cas_phy_write(cp, BROADCOM_MII_REG5, 0x1804);
			cas_phy_write(cp, BROADCOM_MII_REG7, 0x0013);
			cas_phy_write(cp, BROADCOM_MII_REG5, 0x1204);
			cas_phy_write(cp, BROADCOM_MII_REG7, 0x8006);
			cas_phy_write(cp, BROADCOM_MII_REG5, 0x0132);
			cas_phy_write(cp, BROADCOM_MII_REG7, 0x8006);
			cas_phy_write(cp, BROADCOM_MII_REG5, 0x0232);
			cas_phy_write(cp, BROADCOM_MII_REG7, 0x201F);
			cas_phy_write(cp, BROADCOM_MII_REG5, 0x0A20);

		} else if (PHY_BROADCOM_5411 == cp->phy_id) {
			val = cas_phy_read(cp, BROADCOM_MII_REG4);
			val = cas_phy_read(cp, BROADCOM_MII_REG4);
			if (val & 0x0080) {
				/* link workaround */
				cas_phy_write(cp, BROADCOM_MII_REG4,
					      val & ~0x0080);
			}

		} else if (cp->cas_flags & CAS_FLAG_SATURN) {
			writel((cp->phy_type & CAS_PHY_MII_MDIO0) ?
			       SATURN_PCFG_FSI : 0x0,
			       cp->regs + REG_SATURN_PCFG);

			/* load firmware to address 10Mbps auto-negotiation
			 * issue. NOTE: this will need to be changed if the
			 * default firmware gets fixed.
			 */
			if (PHY_NS_DP83065 == cp->phy_id) {
				cas_saturn_firmware_load(cp);
			}
			cas_phy_powerup(cp);
		}

		/* advertise capabilities */
		val = cas_phy_read(cp, MII_BMCR);
		val &= ~BMCR_ANENABLE;
		cas_phy_write(cp, MII_BMCR, val);
		udelay(10);

		cas_phy_write(cp, MII_ADVERTISE,
			      cas_phy_read(cp, MII_ADVERTISE) |
			      (ADVERTISE_10HALF | ADVERTISE_10FULL |
			       ADVERTISE_100HALF | ADVERTISE_100FULL |
			       CAS_ADVERTISE_PAUSE |
			       CAS_ADVERTISE_ASYM_PAUSE));

		if (cp->cas_flags & CAS_FLAG_1000MB_CAP) {
			/* make sure that we don't advertise half
			 * duplex to avoid a chip issue
			 */
			val  = cas_phy_read(cp, CAS_MII_1000_CTRL);
			val &= ~CAS_ADVERTISE_1000HALF;
			val |= CAS_ADVERTISE_1000FULL;
			cas_phy_write(cp, CAS_MII_1000_CTRL, val);
		}

	} else {
		/* reset pcs for serdes */
		u32 val;
		int limit;

		writel(PCS_DATAPATH_MODE_SERDES,
		       cp->regs + REG_PCS_DATAPATH_MODE);

		/* enable serdes pins on saturn */
		if (cp->cas_flags & CAS_FLAG_SATURN)
			writel(0, cp->regs + REG_SATURN_PCFG);

		/* Reset PCS unit. */
		val = readl(cp->regs + REG_PCS_MII_CTRL);
		val |= PCS_MII_RESET;
		writel(val, cp->regs + REG_PCS_MII_CTRL);

		limit = STOP_TRIES;
		while (--limit > 0) {
			udelay(10);
			if ((readl(cp->regs + REG_PCS_MII_CTRL) &
			     PCS_MII_RESET) == 0)
				break;
		}
		if (limit <= 0)
			printk(KERN_WARNING "%s: PCS reset bit would not "
			       "clear [%08x].\n", cp->dev->name,
			       readl(cp->regs + REG_PCS_STATE_MACHINE));

		/* Make sure PCS is disabled while changing advertisement
		 * configuration.
		 */
		writel(0x0, cp->regs + REG_PCS_CFG);

		/* Advertise all capabilities except half-duplex. */
		val  = readl(cp->regs + REG_PCS_MII_ADVERT);
		val &= ~PCS_MII_ADVERT_HD;
		val |= (PCS_MII_ADVERT_FD | PCS_MII_ADVERT_SYM_PAUSE |
			PCS_MII_ADVERT_ASYM_PAUSE);
		writel(val, cp->regs + REG_PCS_MII_ADVERT);

		/* enable PCS */
		writel(PCS_CFG_EN, cp->regs + REG_PCS_CFG);

		/* pcs workaround: enable sync detect */
		writel(PCS_SERDES_CTRL_SYNCD_EN,
		       cp->regs + REG_PCS_SERDES_CTRL);
	}
}


static int cas_pcs_link_check(struct cas *cp)
{
	u32 stat, state_machine;
	int retval = 0;

	/* The link status bit latches on zero, so you must
	 * read it twice in such a case to see a transition
	 * to the link being up.
	 */
	stat = readl(cp->regs + REG_PCS_MII_STATUS);
	if ((stat & PCS_MII_STATUS_LINK_STATUS) == 0)
		stat = readl(cp->regs + REG_PCS_MII_STATUS);

	/* The remote-fault indication is only valid
	 * when autoneg has completed.
	 */
	if ((stat & (PCS_MII_STATUS_AUTONEG_COMP |
		     PCS_MII_STATUS_REMOTE_FAULT)) ==
	    (PCS_MII_STATUS_AUTONEG_COMP | PCS_MII_STATUS_REMOTE_FAULT)) {
		if (netif_msg_link(cp))
			printk(KERN_INFO "%s: PCS RemoteFault\n",
			       cp->dev->name);
	}

	/* work around link detection issue by querying the PCS state
	 * machine directly.
	 */
	state_machine = readl(cp->regs + REG_PCS_STATE_MACHINE);
	if ((state_machine & PCS_SM_LINK_STATE_MASK) != SM_LINK_STATE_UP) {
		stat &= ~PCS_MII_STATUS_LINK_STATUS;
	} else if (state_machine & PCS_SM_WORD_SYNC_STATE_MASK) {
		stat |= PCS_MII_STATUS_LINK_STATUS;
	}

	if (stat & PCS_MII_STATUS_LINK_STATUS) {
		if (cp->lstate != link_up) {
			if (cp->opened) {
				cp->lstate = link_up;
				cp->link_transition = LINK_TRANSITION_LINK_UP;

				cas_set_link_modes(cp);
				netif_carrier_on(cp->dev);
			}
		}
	} else if (cp->lstate == link_up) {
		cp->lstate = link_down;
		if (link_transition_timeout != 0 &&
		    cp->link_transition != LINK_TRANSITION_REQUESTED_RESET &&
		    !cp->link_transition_jiffies_valid) {
			/*
			 * force a reset, as a workaround for the
			 * link-failure problem. May want to move this to a
			 * point a bit earlier in the sequence. If we had
			 * generated a reset a short time ago, we'll wait for
			 * the link timer to check the status until a
			 * timer expires (link_transistion_jiffies_valid is
			 * true when the timer is running.)  Instead of using
			 * a system timer, we just do a check whenever the
			 * link timer is running - this clears the flag after
			 * a suitable delay.
			 */
			retval = 1;
			cp->link_transition = LINK_TRANSITION_REQUESTED_RESET;
			cp->link_transition_jiffies = jiffies;
			cp->link_transition_jiffies_valid = 1;
		} else {
			cp->link_transition = LINK_TRANSITION_ON_FAILURE;
		}
		netif_carrier_off(cp->dev);
		if (cp->opened && netif_msg_link(cp)) {
			printk(KERN_INFO "%s: PCS link down.\n",
			       cp->dev->name);
		}

		/* Cassini only: if you force a mode, there can be
		 * sync problems on link down. to fix that, the following
		 * things need to be checked:
		 * 1) read serialink state register
		 * 2) read pcs status register to verify link down.
		 * 3) if link down and serial link == 0x03, then you need
		 *    to global reset the chip.
		 */
		if ((cp->cas_flags & CAS_FLAG_REG_PLUS) == 0) {
			/* should check to see if we're in a forced mode */
			stat = readl(cp->regs + REG_PCS_SERDES_STATE);
			if (stat == 0x03)
				return 1;
		}
	} else if (cp->lstate == link_down) {
		if (link_transition_timeout != 0 &&
		    cp->link_transition != LINK_TRANSITION_REQUESTED_RESET &&
		    !cp->link_transition_jiffies_valid) {
			/* force a reset, as a workaround for the
			 * link-failure problem.  May want to move
			 * this to a point a bit earlier in the
			 * sequence.
			 */
			retval = 1;
			cp->link_transition = LINK_TRANSITION_REQUESTED_RESET;
			cp->link_transition_jiffies = jiffies;
			cp->link_transition_jiffies_valid = 1;
		} else {
			cp->link_transition = LINK_TRANSITION_STILL_FAILED;
		}
	}

	return retval;
}

static int cas_pcs_interrupt(struct net_device *dev,
			     struct cas *cp, u32 status)
{
	u32 stat = readl(cp->regs + REG_PCS_INTR_STATUS);

	if ((stat & PCS_INTR_STATUS_LINK_CHANGE) == 0)
		return 0;
	return cas_pcs_link_check(cp);
}

static int cas_txmac_interrupt(struct net_device *dev,
			       struct cas *cp, u32 status)
{
	u32 txmac_stat = readl(cp->regs + REG_MAC_TX_STATUS);

	if (!txmac_stat)
		return 0;

	if (netif_msg_intr(cp))
		printk(KERN_DEBUG "%s: txmac interrupt, txmac_stat: 0x%x\n",
			cp->dev->name, txmac_stat);

	/* Defer timer expiration is quite normal,
	 * don't even log the event.
	 */
	if ((txmac_stat & MAC_TX_DEFER_TIMER) &&
	    !(txmac_stat & ~MAC_TX_DEFER_TIMER))
		return 0;

	spin_lock(&cp->stat_lock[0]);
	if (txmac_stat & MAC_TX_UNDERRUN) {
		printk(KERN_ERR "%s: TX MAC xmit underrun.\n",
		       dev->name);
		cp->net_stats[0].tx_fifo_errors++;
	}

	if (txmac_stat & MAC_TX_MAX_PACKET_ERR) {
		printk(KERN_ERR "%s: TX MAC max packet size error.\n",
		       dev->name);
		cp->net_stats[0].tx_errors++;
	}

	/* The rest are all cases of one of the 16-bit TX
	 * counters expiring.
	 */
	if (txmac_stat & MAC_TX_COLL_NORMAL)
		cp->net_stats[0].collisions += 0x10000;

	if (txmac_stat & MAC_TX_COLL_EXCESS) {
		cp->net_stats[0].tx_aborted_errors += 0x10000;
		cp->net_stats[0].collisions += 0x10000;
	}

	if (txmac_stat & MAC_TX_COLL_LATE) {
		cp->net_stats[0].tx_aborted_errors += 0x10000;
		cp->net_stats[0].collisions += 0x10000;
	}
	spin_unlock(&cp->stat_lock[0]);

	/* We do not keep track of MAC_TX_COLL_FIRST and
	 * MAC_TX_PEAK_ATTEMPTS events.
	 */
	return 0;
}

static void cas_load_firmware(struct cas *cp, cas_hp_inst_t *firmware)
{
	cas_hp_inst_t *inst;
	u32 val;
	int i;

	i = 0;
	while ((inst = firmware) && inst->note) {
		writel(i, cp->regs + REG_HP_INSTR_RAM_ADDR);

		val = CAS_BASE(HP_INSTR_RAM_HI_VAL, inst->val);
		val |= CAS_BASE(HP_INSTR_RAM_HI_MASK, inst->mask);
		writel(val, cp->regs + REG_HP_INSTR_RAM_DATA_HI);

		val = CAS_BASE(HP_INSTR_RAM_MID_OUTARG, inst->outarg >> 10);
		val |= CAS_BASE(HP_INSTR_RAM_MID_OUTOP, inst->outop);
		val |= CAS_BASE(HP_INSTR_RAM_MID_FNEXT, inst->fnext);
		val |= CAS_BASE(HP_INSTR_RAM_MID_FOFF, inst->foff);
		val |= CAS_BASE(HP_INSTR_RAM_MID_SNEXT, inst->snext);
		val |= CAS_BASE(HP_INSTR_RAM_MID_SOFF, inst->soff);
		val |= CAS_BASE(HP_INSTR_RAM_MID_OP, inst->op);
		writel(val, cp->regs + REG_HP_INSTR_RAM_DATA_MID);

		val = CAS_BASE(HP_INSTR_RAM_LOW_OUTMASK, inst->outmask);
		val |= CAS_BASE(HP_INSTR_RAM_LOW_OUTSHIFT, inst->outshift);
		val |= CAS_BASE(HP_INSTR_RAM_LOW_OUTEN, inst->outenab);
		val |= CAS_BASE(HP_INSTR_RAM_LOW_OUTARG, inst->outarg);
		writel(val, cp->regs + REG_HP_INSTR_RAM_DATA_LOW);
		++firmware;
		++i;
	}
}

static void cas_init_rx_dma(struct cas *cp)
{
	u64 desc_dma = cp->block_dvma;
	u32 val;
	int i, size;

	/* rx free descriptors */
	val = CAS_BASE(RX_CFG_SWIVEL, RX_SWIVEL_OFF_VAL);
	val |= CAS_BASE(RX_CFG_DESC_RING, RX_DESC_RINGN_INDEX(0));
	val |= CAS_BASE(RX_CFG_COMP_RING, RX_COMP_RINGN_INDEX(0));
	if ((N_RX_DESC_RINGS > 1) &&
	    (cp->cas_flags & CAS_FLAG_REG_PLUS))  /* do desc 2 */
		val |= CAS_BASE(RX_CFG_DESC_RING1, RX_DESC_RINGN_INDEX(1));
	writel(val, cp->regs + REG_RX_CFG);

	val = (unsigned long) cp->init_rxds[0] -
		(unsigned long) cp->init_block;
	writel((desc_dma + val) >> 32, cp->regs + REG_RX_DB_HI);
	writel((desc_dma + val) & 0xffffffff, cp->regs + REG_RX_DB_LOW);
	writel(RX_DESC_RINGN_SIZE(0) - 4, cp->regs + REG_RX_KICK);

	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
		/* rx desc 2 is for IPSEC packets. however,
		 * we don't it that for that purpose.
		 */
		val = (unsigned long) cp->init_rxds[1] -
			(unsigned long) cp->init_block;
		writel((desc_dma + val) >> 32, cp->regs + REG_PLUS_RX_DB1_HI);
		writel((desc_dma + val) & 0xffffffff, cp->regs +
		       REG_PLUS_RX_DB1_LOW);
		writel(RX_DESC_RINGN_SIZE(1) - 4, cp->regs +
		       REG_PLUS_RX_KICK1);
	}

	/* rx completion registers */
	val = (unsigned long) cp->init_rxcs[0] -
		(unsigned long) cp->init_block;
	writel((desc_dma + val) >> 32, cp->regs + REG_RX_CB_HI);
	writel((desc_dma + val) & 0xffffffff, cp->regs + REG_RX_CB_LOW);

	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
		/* rx comp 2-4 */
		for (i = 1; i < MAX_RX_COMP_RINGS; i++) {
			val = (unsigned long) cp->init_rxcs[i] -
				(unsigned long) cp->init_block;
			writel((desc_dma + val) >> 32, cp->regs +
			       REG_PLUS_RX_CBN_HI(i));
			writel((desc_dma + val) & 0xffffffff, cp->regs +
			       REG_PLUS_RX_CBN_LOW(i));
		}
	}

	/* read selective clear regs to prevent spurious interrupts
	 * on reset because complete == kick.
	 * selective clear set up to prevent interrupts on resets
	 */
	readl(cp->regs + REG_INTR_STATUS_ALIAS);
	writel(INTR_RX_DONE | INTR_RX_BUF_UNAVAIL, cp->regs + REG_ALIAS_CLEAR);
	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
		for (i = 1; i < N_RX_COMP_RINGS; i++)
			readl(cp->regs + REG_PLUS_INTRN_STATUS_ALIAS(i));

		/* 2 is different from 3 and 4 */
		if (N_RX_COMP_RINGS > 1)
			writel(INTR_RX_DONE_ALT | INTR_RX_BUF_UNAVAIL_1,
			       cp->regs + REG_PLUS_ALIASN_CLEAR(1));

		for (i = 2; i < N_RX_COMP_RINGS; i++)
			writel(INTR_RX_DONE_ALT,
			       cp->regs + REG_PLUS_ALIASN_CLEAR(i));
	}

	/* set up pause thresholds */
	val  = CAS_BASE(RX_PAUSE_THRESH_OFF,
			cp->rx_pause_off / RX_PAUSE_THRESH_QUANTUM);
	val |= CAS_BASE(RX_PAUSE_THRESH_ON,
			cp->rx_pause_on / RX_PAUSE_THRESH_QUANTUM);
	writel(val, cp->regs + REG_RX_PAUSE_THRESH);

	/* zero out dma reassembly buffers */
	for (i = 0; i < 64; i++) {
		writel(i, cp->regs + REG_RX_TABLE_ADDR);
		writel(0x0, cp->regs + REG_RX_TABLE_DATA_LOW);
		writel(0x0, cp->regs + REG_RX_TABLE_DATA_MID);
		writel(0x0, cp->regs + REG_RX_TABLE_DATA_HI);
	}

	/* make sure address register is 0 for normal operation */
	writel(0x0, cp->regs + REG_RX_CTRL_FIFO_ADDR);
	writel(0x0, cp->regs + REG_RX_IPP_FIFO_ADDR);

	/* interrupt mitigation */
#ifdef USE_RX_BLANK
	val = CAS_BASE(RX_BLANK_INTR_TIME, RX_BLANK_INTR_TIME_VAL);
	val |= CAS_BASE(RX_BLANK_INTR_PKT, RX_BLANK_INTR_PKT_VAL);
	writel(val, cp->regs + REG_RX_BLANK);
#else
	writel(0x0, cp->regs + REG_RX_BLANK);
#endif

	/* interrupt generation as a function of low water marks for
	 * free desc and completion entries. these are used to trigger
	 * housekeeping for rx descs. we don't use the free interrupt
	 * as it's not very useful
	 */
	/* val = CAS_BASE(RX_AE_THRESH_FREE, RX_AE_FREEN_VAL(0)); */
	val = CAS_BASE(RX_AE_THRESH_COMP, RX_AE_COMP_VAL);
	writel(val, cp->regs + REG_RX_AE_THRESH);
	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
		val = CAS_BASE(RX_AE1_THRESH_FREE, RX_AE_FREEN_VAL(1));
		writel(val, cp->regs + REG_PLUS_RX_AE1_THRESH);
	}

	/* Random early detect registers. useful for congestion avoidance.
	 * this should be tunable.
	 */
	writel(0x0, cp->regs + REG_RX_RED);

	/* receive page sizes. default == 2K (0x800) */
	val = 0;
	if (cp->page_size == 0x1000)
		val = 0x1;
	else if (cp->page_size == 0x2000)
		val = 0x2;
	else if (cp->page_size == 0x4000)
		val = 0x3;

	/* round mtu + offset. constrain to page size. */
	size = cp->dev->mtu + 64;
	if (size > cp->page_size)
		size = cp->page_size;

	if (size <= 0x400)
		i = 0x0;
	else if (size <= 0x800)
		i = 0x1;
	else if (size <= 0x1000)
		i = 0x2;
	else
		i = 0x3;

	cp->mtu_stride = 1 << (i + 10);
	val  = CAS_BASE(RX_PAGE_SIZE, val);
	val |= CAS_BASE(RX_PAGE_SIZE_MTU_STRIDE, i);
	val |= CAS_BASE(RX_PAGE_SIZE_MTU_COUNT, cp->page_size >> (i + 10));
	val |= CAS_BASE(RX_PAGE_SIZE_MTU_OFF, 0x1);
	writel(val, cp->regs + REG_RX_PAGE_SIZE);

	/* enable the header parser if desired */
	if (CAS_HP_FIRMWARE == cas_prog_null)
		return;

	val = CAS_BASE(HP_CFG_NUM_CPU, CAS_NCPUS > 63 ? 0 : CAS_NCPUS);
	val |= HP_CFG_PARSE_EN | HP_CFG_SYN_INC_MASK;
	val |= CAS_BASE(HP_CFG_TCP_THRESH, HP_TCP_THRESH_VAL);
	writel(val, cp->regs + REG_HP_CFG);
}

static inline void cas_rxc_init(struct cas_rx_comp *rxc)
{
	memset(rxc, 0, sizeof(*rxc));
	rxc->word4 = cpu_to_le64(RX_COMP4_ZERO);
}

/* NOTE: we use the ENC RX DESC ring for spares. the rx_page[0,1]
 * flipping is protected by the fact that the chip will not
 * hand back the same page index while it's being processed.
 */
static inline cas_page_t *cas_page_spare(struct cas *cp, const int index)
{
	cas_page_t *page = cp->rx_pages[1][index];
	cas_page_t *new;

	if (page_count(page->buffer) == 1)
		return page;

	new = cas_page_dequeue(cp);
	if (new) {
		spin_lock(&cp->rx_inuse_lock);
		list_add(&page->list, &cp->rx_inuse_list);
		spin_unlock(&cp->rx_inuse_lock);
	}
	return new;
}

/* this needs to be changed if we actually use the ENC RX DESC ring */
static cas_page_t *cas_page_swap(struct cas *cp, const int ring,
				 const int index)
{
	cas_page_t **page0 = cp->rx_pages[0];
	cas_page_t **page1 = cp->rx_pages[1];

	/* swap if buffer is in use */
	if (page_count(page0[index]->buffer) > 1) {
		cas_page_t *new = cas_page_spare(cp, index);
		if (new) {
			page1[index] = page0[index];
			page0[index] = new;
		}
	}
	RX_USED_SET(page0[index], 0);
	return page0[index];
}

static void cas_clean_rxds(struct cas *cp)
{
	/* only clean ring 0 as ring 1 is used for spare buffers */
        struct cas_rx_desc *rxd = cp->init_rxds[0];
	int i, size;

	/* release all rx flows */
	for (i = 0; i < N_RX_FLOWS; i++) {
		struct sk_buff *skb;
		while ((skb = __skb_dequeue(&cp->rx_flows[i]))) {
			cas_skb_release(skb);
		}
	}

	/* initialize descriptors */
	size = RX_DESC_RINGN_SIZE(0);
	for (i = 0; i < size; i++) {
		cas_page_t *page = cas_page_swap(cp, 0, i);
		rxd[i].buffer = cpu_to_le64(page->dma_addr);
		rxd[i].index  = cpu_to_le64(CAS_BASE(RX_INDEX_NUM, i) |
					    CAS_BASE(RX_INDEX_RING, 0));
	}

	cp->rx_old[0]  = RX_DESC_RINGN_SIZE(0) - 4;
	cp->rx_last[0] = 0;
	cp->cas_flags &= ~CAS_FLAG_RXD_POST(0);
}

static void cas_clean_rxcs(struct cas *cp)
{
	int i, j;

	/* take ownership of rx comp descriptors */
	memset(cp->rx_cur, 0, sizeof(*cp->rx_cur)*N_RX_COMP_RINGS);
	memset(cp->rx_new, 0, sizeof(*cp->rx_new)*N_RX_COMP_RINGS);
	for (i = 0; i < N_RX_COMP_RINGS; i++) {
		struct cas_rx_comp *rxc = cp->init_rxcs[i];
		for (j = 0; j < RX_COMP_RINGN_SIZE(i); j++) {
			cas_rxc_init(rxc + j);
		}
	}
}

#if 0
/* When we get a RX fifo overflow, the RX unit is probably hung
 * so we do the following.
 *
 * If any part of the reset goes wrong, we return 1 and that causes the
 * whole chip to be reset.
 */
static int cas_rxmac_reset(struct cas *cp)
{
	struct net_device *dev = cp->dev;
	int limit;
	u32 val;

	/* First, reset MAC RX. */
	writel(cp->mac_rx_cfg & ~MAC_RX_CFG_EN, cp->regs + REG_MAC_RX_CFG);
	for (limit = 0; limit < STOP_TRIES; limit++) {
		if (!(readl(cp->regs + REG_MAC_RX_CFG) & MAC_RX_CFG_EN))
			break;
		udelay(10);
	}
	if (limit == STOP_TRIES) {
		printk(KERN_ERR "%s: RX MAC will not disable, resetting whole "
		       "chip.\n", dev->name);
		return 1;
	}

	/* Second, disable RX DMA. */
	writel(0, cp->regs + REG_RX_CFG);
	for (limit = 0; limit < STOP_TRIES; limit++) {
		if (!(readl(cp->regs + REG_RX_CFG) & RX_CFG_DMA_EN))
			break;
		udelay(10);
	}
	if (limit == STOP_TRIES) {
		printk(KERN_ERR "%s: RX DMA will not disable, resetting whole "
		       "chip.\n", dev->name);
		return 1;
	}

	mdelay(5);

	/* Execute RX reset command. */
	writel(SW_RESET_RX, cp->regs + REG_SW_RESET);
	for (limit = 0; limit < STOP_TRIES; limit++) {
		if (!(readl(cp->regs + REG_SW_RESET) & SW_RESET_RX))
			break;
		udelay(10);
	}
	if (limit == STOP_TRIES) {
		printk(KERN_ERR "%s: RX reset command will not execute, "
		       "resetting whole chip.\n", dev->name);
		return 1;
	}

	/* reset driver rx state */
	cas_clean_rxds(cp);
	cas_clean_rxcs(cp);

	/* Now, reprogram the rest of RX unit. */
	cas_init_rx_dma(cp);

	/* re-enable */
	val = readl(cp->regs + REG_RX_CFG);
	writel(val | RX_CFG_DMA_EN, cp->regs + REG_RX_CFG);
	writel(MAC_RX_FRAME_RECV, cp->regs + REG_MAC_RX_MASK);
	val = readl(cp->regs + REG_MAC_RX_CFG);
	writel(val | MAC_RX_CFG_EN, cp->regs + REG_MAC_RX_CFG);
	return 0;
}
#endif

static int cas_rxmac_interrupt(struct net_device *dev, struct cas *cp,
			       u32 status)
{
	u32 stat = readl(cp->regs + REG_MAC_RX_STATUS);

	if (!stat)
		return 0;

	if (netif_msg_intr(cp))
		printk(KERN_DEBUG "%s: rxmac interrupt, stat: 0x%x\n",
			cp->dev->name, stat);

	/* these are all rollovers */
	spin_lock(&cp->stat_lock[0]);
	if (stat & MAC_RX_ALIGN_ERR)
		cp->net_stats[0].rx_frame_errors += 0x10000;

	if (stat & MAC_RX_CRC_ERR)
		cp->net_stats[0].rx_crc_errors += 0x10000;

	if (stat & MAC_RX_LEN_ERR)
		cp->net_stats[0].rx_length_errors += 0x10000;

	if (stat & MAC_RX_OVERFLOW) {
		cp->net_stats[0].rx_over_errors++;
		cp->net_stats[0].rx_fifo_errors++;
	}

	/* We do not track MAC_RX_FRAME_COUNT and MAC_RX_VIOL_ERR
	 * events.
	 */
	spin_unlock(&cp->stat_lock[0]);
	return 0;
}

static int cas_mac_interrupt(struct net_device *dev, struct cas *cp,
			     u32 status)
{
	u32 stat = readl(cp->regs + REG_MAC_CTRL_STATUS);

	if (!stat)
		return 0;

	if (netif_msg_intr(cp))
		printk(KERN_DEBUG "%s: mac interrupt, stat: 0x%x\n",
			cp->dev->name, stat);

	/* This interrupt is just for pause frame and pause
	 * tracking.  It is useful for diagnostics and debug
	 * but probably by default we will mask these events.
	 */
	if (stat & MAC_CTRL_PAUSE_STATE)
		cp->pause_entered++;

	if (stat & MAC_CTRL_PAUSE_RECEIVED)
		cp->pause_last_time_recvd = (stat >> 16);

	return 0;
}


/* Must be invoked under cp->lock. */
static inline int cas_mdio_link_not_up(struct cas *cp)
{
	u16 val;

	switch (cp->lstate) {
	case link_force_ret:
		if (netif_msg_link(cp))
			printk(KERN_INFO "%s: Autoneg failed again, keeping"
				" forced mode\n", cp->dev->name);
		cas_phy_write(cp, MII_BMCR, cp->link_fcntl);
		cp->timer_ticks = 5;
		cp->lstate = link_force_ok;
		cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
		break;

	case link_aneg:
		val = cas_phy_read(cp, MII_BMCR);

		/* Try forced modes. we try things in the following order:
		 * 1000 full -> 100 full/half -> 10 half
		 */
		val &= ~(BMCR_ANRESTART | BMCR_ANENABLE);
		val |= BMCR_FULLDPLX;
		val |= (cp->cas_flags & CAS_FLAG_1000MB_CAP) ?
			CAS_BMCR_SPEED1000 : BMCR_SPEED100;
		cas_phy_write(cp, MII_BMCR, val);
		cp->timer_ticks = 5;
		cp->lstate = link_force_try;
		cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
		break;

	case link_force_try:
		/* Downgrade from 1000 to 100 to 10 Mbps if necessary. */
		val = cas_phy_read(cp, MII_BMCR);
		cp->timer_ticks = 5;
		if (val & CAS_BMCR_SPEED1000) { /* gigabit */
			val &= ~CAS_BMCR_SPEED1000;
			val |= (BMCR_SPEED100 | BMCR_FULLDPLX);
			cas_phy_write(cp, MII_BMCR, val);
			break;
		}

		if (val & BMCR_SPEED100) {
			if (val & BMCR_FULLDPLX) /* fd failed */
				val &= ~BMCR_FULLDPLX;
			else { /* 100Mbps failed */
				val &= ~BMCR_SPEED100;
			}
			cas_phy_write(cp, MII_BMCR, val);
			break;
		}
	default:
		break;
	}
	return 0;
}


/* must be invoked with cp->lock held */
static int cas_mii_link_check(struct cas *cp, const u16 bmsr)
{
	int restart;

	if (bmsr & BMSR_LSTATUS) {
		/* Ok, here we got a link. If we had it due to a forced
		 * fallback, and we were configured for autoneg, we
		 * retry a short autoneg pass. If you know your hub is
		 * broken, use ethtool ;)
		 */
		if ((cp->lstate == link_force_try) &&
		    (cp->link_cntl & BMCR_ANENABLE)) {
			cp->lstate = link_force_ret;
			cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
			cas_mif_poll(cp, 0);
			cp->link_fcntl = cas_phy_read(cp, MII_BMCR);
			cp->timer_ticks = 5;
			if (cp->opened && netif_msg_link(cp))
				printk(KERN_INFO "%s: Got link after fallback, retrying"
				       " autoneg once...\n", cp->dev->name);
			cas_phy_write(cp, MII_BMCR,
				      cp->link_fcntl | BMCR_ANENABLE |
				      BMCR_ANRESTART);
			cas_mif_poll(cp, 1);

		} else if (cp->lstate != link_up) {
			cp->lstate = link_up;
			cp->link_transition = LINK_TRANSITION_LINK_UP;

			if (cp->opened) {
				cas_set_link_modes(cp);
				netif_carrier_on(cp->dev);
			}
		}
		return 0;
	}

	/* link not up. if the link was previously up, we restart the
	 * whole process
	 */
	restart = 0;
	if (cp->lstate == link_up) {
		cp->lstate = link_down;
		cp->link_transition = LINK_TRANSITION_LINK_DOWN;

		netif_carrier_off(cp->dev);
		if (cp->opened && netif_msg_link(cp))
			printk(KERN_INFO "%s: Link down\n",
			       cp->dev->name);
		restart = 1;

	} else if (++cp->timer_ticks > 10)
		cas_mdio_link_not_up(cp);

	return restart;
}

static int cas_mif_interrupt(struct net_device *dev, struct cas *cp,
			     u32 status)
{
	u32 stat = readl(cp->regs + REG_MIF_STATUS);
	u16 bmsr;

	/* check for a link change */
	if (CAS_VAL(MIF_STATUS_POLL_STATUS, stat) == 0)
		return 0;

	bmsr = CAS_VAL(MIF_STATUS_POLL_DATA, stat);
	return cas_mii_link_check(cp, bmsr);
}

static int cas_pci_interrupt(struct net_device *dev, struct cas *cp,
			     u32 status)
{
	u32 stat = readl(cp->regs + REG_PCI_ERR_STATUS);

	if (!stat)
		return 0;

	printk(KERN_ERR "%s: PCI error [%04x:%04x] ", dev->name, stat,
	       readl(cp->regs + REG_BIM_DIAG));

	/* cassini+ has this reserved */
	if ((stat & PCI_ERR_BADACK) &&
	    ((cp->cas_flags & CAS_FLAG_REG_PLUS) == 0))
		printk("<No ACK64# during ABS64 cycle> ");

	if (stat & PCI_ERR_DTRTO)
		printk("<Delayed transaction timeout> ");
	if (stat & PCI_ERR_OTHER)
		printk("<other> ");
	if (stat & PCI_ERR_BIM_DMA_WRITE)
		printk("<BIM DMA 0 write req> ");
	if (stat & PCI_ERR_BIM_DMA_READ)
		printk("<BIM DMA 0 read req> ");
	printk("\n");

	if (stat & PCI_ERR_OTHER) {
		u16 cfg;

		/* Interrogate PCI config space for the
		 * true cause.
		 */
		pci_read_config_word(cp->pdev, PCI_STATUS, &cfg);
		printk(KERN_ERR "%s: Read PCI cfg space status [%04x]\n",
		       dev->name, cfg);
		if (cfg & PCI_STATUS_PARITY)
			printk(KERN_ERR "%s: PCI parity error detected.\n",
			       dev->name);
		if (cfg & PCI_STATUS_SIG_TARGET_ABORT)
			printk(KERN_ERR "%s: PCI target abort.\n",
			       dev->name);
		if (cfg & PCI_STATUS_REC_TARGET_ABORT)
			printk(KERN_ERR "%s: PCI master acks target abort.\n",
			       dev->name);
		if (cfg & PCI_STATUS_REC_MASTER_ABORT)
			printk(KERN_ERR "%s: PCI master abort.\n", dev->name);
		if (cfg & PCI_STATUS_SIG_SYSTEM_ERROR)
			printk(KERN_ERR "%s: PCI system error SERR#.\n",
			       dev->name);
		if (cfg & PCI_STATUS_DETECTED_PARITY)
			printk(KERN_ERR "%s: PCI parity error.\n",
			       dev->name);

		/* Write the error bits back to clear them. */
		cfg &= (PCI_STATUS_PARITY |
			PCI_STATUS_SIG_TARGET_ABORT |
			PCI_STATUS_REC_TARGET_ABORT |
			PCI_STATUS_REC_MASTER_ABORT |
			PCI_STATUS_SIG_SYSTEM_ERROR |
			PCI_STATUS_DETECTED_PARITY);
		pci_write_config_word(cp->pdev, PCI_STATUS, cfg);
	}

	/* For all PCI errors, we should reset the chip. */
	return 1;
}

/* All non-normal interrupt conditions get serviced here.
 * Returns non-zero if we should just exit the interrupt
 * handler right now (ie. if we reset the card which invalidates
 * all of the other original irq status bits).
 */
static int cas_abnormal_irq(struct net_device *dev, struct cas *cp,
			    u32 status)
{
	if (status & INTR_RX_TAG_ERROR) {
		/* corrupt RX tag framing */
		if (netif_msg_rx_err(cp))
			printk(KERN_DEBUG "%s: corrupt rx tag framing\n",
				cp->dev->name);
		spin_lock(&cp->stat_lock[0]);
		cp->net_stats[0].rx_errors++;
		spin_unlock(&cp->stat_lock[0]);
		goto do_reset;
	}

	if (status & INTR_RX_LEN_MISMATCH) {
		/* length mismatch. */
		if (netif_msg_rx_err(cp))
			printk(KERN_DEBUG "%s: length mismatch for rx frame\n",
				cp->dev->name);
		spin_lock(&cp->stat_lock[0]);
		cp->net_stats[0].rx_errors++;
		spin_unlock(&cp->stat_lock[0]);
		goto do_reset;
	}

	if (status & INTR_PCS_STATUS) {
		if (cas_pcs_interrupt(dev, cp, status))
			goto do_reset;
	}

	if (status & INTR_TX_MAC_STATUS) {
		if (cas_txmac_interrupt(dev, cp, status))
			goto do_reset;
	}

	if (status & INTR_RX_MAC_STATUS) {
		if (cas_rxmac_interrupt(dev, cp, status))
			goto do_reset;
	}

	if (status & INTR_MAC_CTRL_STATUS) {
		if (cas_mac_interrupt(dev, cp, status))
			goto do_reset;
	}

	if (status & INTR_MIF_STATUS) {
		if (cas_mif_interrupt(dev, cp, status))
			goto do_reset;
	}

	if (status & INTR_PCI_ERROR_STATUS) {
		if (cas_pci_interrupt(dev, cp, status))
			goto do_reset;
	}
	return 0;

do_reset:
#if 1
	atomic_inc(&cp->reset_task_pending);
	atomic_inc(&cp->reset_task_pending_all);
	printk(KERN_ERR "%s:reset called in cas_abnormal_irq [0x%x]\n",
	       dev->name, status);
	schedule_work(&cp->reset_task);
#else
	atomic_set(&cp->reset_task_pending, CAS_RESET_ALL);
	printk(KERN_ERR "reset called in cas_abnormal_irq\n");
	schedule_work(&cp->reset_task);
#endif
	return 1;
}

/* NOTE: CAS_TABORT returns 1 or 2 so that it can be used when
 *       determining whether to do a netif_stop/wakeup
 */
#define CAS_TABORT(x)      (((x)->cas_flags & CAS_FLAG_TARGET_ABORT) ? 2 : 1)
#define CAS_ROUND_PAGE(x)  (((x) + PAGE_SIZE - 1) & PAGE_MASK)
static inline int cas_calc_tabort(struct cas *cp, const unsigned long addr,
				  const int len)
{
	unsigned long off = addr + len;

	if (CAS_TABORT(cp) == 1)
		return 0;
	if ((CAS_ROUND_PAGE(off) - off) > TX_TARGET_ABORT_LEN)
		return 0;
	return TX_TARGET_ABORT_LEN;
}

static inline void cas_tx_ringN(struct cas *cp, int ring, int limit)
{
	struct cas_tx_desc *txds;
	struct sk_buff **skbs;
	struct net_device *dev = cp->dev;
	int entry, count;

	spin_lock(&cp->tx_lock[ring]);
	txds = cp->init_txds[ring];
	skbs = cp->tx_skbs[ring];
	entry = cp->tx_old[ring];

	count = TX_BUFF_COUNT(ring, entry, limit);
	while (entry != limit) {
		struct sk_buff *skb = skbs[entry];
		dma_addr_t daddr;
		u32 dlen;
		int frag;

		if (!skb) {
			/* this should never occur */
			entry = TX_DESC_NEXT(ring, entry);
			continue;
		}

		/* however, we might get only a partial skb release. */
		count -= skb_shinfo(skb)->nr_frags +
			+ cp->tx_tiny_use[ring][entry].nbufs + 1;
		if (count < 0)
			break;

		if (netif_msg_tx_done(cp))
			printk(KERN_DEBUG "%s: tx[%d] done, slot %d\n",
			       cp->dev->name, ring, entry);

		skbs[entry] = NULL;
		cp->tx_tiny_use[ring][entry].nbufs = 0;

		for (frag = 0; frag <= skb_shinfo(skb)->nr_frags; frag++) {
			struct cas_tx_desc *txd = txds + entry;

			daddr = le64_to_cpu(txd->buffer);
			dlen = CAS_VAL(TX_DESC_BUFLEN,
				       le64_to_cpu(txd->control));
			pci_unmap_page(cp->pdev, daddr, dlen,
				       PCI_DMA_TODEVICE);
			entry = TX_DESC_NEXT(ring, entry);

			/* tiny buffer may follow */
			if (cp->tx_tiny_use[ring][entry].used) {
				cp->tx_tiny_use[ring][entry].used = 0;
				entry = TX_DESC_NEXT(ring, entry);
			}
		}

		spin_lock(&cp->stat_lock[ring]);
		cp->net_stats[ring].tx_packets++;
		cp->net_stats[ring].tx_bytes += skb->len;
		spin_unlock(&cp->stat_lock[ring]);
		dev_kfree_skb_irq(skb);
	}
	cp->tx_old[ring] = entry;

	/* this is wrong for multiple tx rings. the net device needs
	 * multiple queues for this to do the right thing.  we wait
	 * for 2*packets to be available when using tiny buffers
	 */
	if (netif_queue_stopped(dev) &&
	    (TX_BUFFS_AVAIL(cp, ring) > CAS_TABORT(cp)*(MAX_SKB_FRAGS + 1)))
		netif_wake_queue(dev);
	spin_unlock(&cp->tx_lock[ring]);
}

static void cas_tx(struct net_device *dev, struct cas *cp,
		   u32 status)
{
        int limit, ring;
#ifdef USE_TX_COMPWB
	u64 compwb = le64_to_cpu(cp->init_block->tx_compwb);
#endif
	if (netif_msg_intr(cp))
		printk(KERN_DEBUG "%s: tx interrupt, status: 0x%x, %llx\n",
			cp->dev->name, status, (unsigned long long)compwb);
	/* process all the rings */
	for (ring = 0; ring < N_TX_RINGS; ring++) {
#ifdef USE_TX_COMPWB
		/* use the completion writeback registers */
		limit = (CAS_VAL(TX_COMPWB_MSB, compwb) << 8) |
			CAS_VAL(TX_COMPWB_LSB, compwb);
		compwb = TX_COMPWB_NEXT(compwb);
#else
		limit = readl(cp->regs + REG_TX_COMPN(ring));
#endif
		if (cp->tx_old[ring] != limit)
			cas_tx_ringN(cp, ring, limit);
	}
}


static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc,
			      int entry, const u64 *words,
			      struct sk_buff **skbref)
{
	int dlen, hlen, len, i, alloclen;
	int off, swivel = RX_SWIVEL_OFF_VAL;
	struct cas_page *page;
	struct sk_buff *skb;
	void *addr, *crcaddr;
	__sum16 csum;
	char *p;

	hlen = CAS_VAL(RX_COMP2_HDR_SIZE, words[1]);
	dlen = CAS_VAL(RX_COMP1_DATA_SIZE, words[0]);
	len  = hlen + dlen;

	if (RX_COPY_ALWAYS || (words[2] & RX_COMP3_SMALL_PKT))
		alloclen = len;
	else
		alloclen = max(hlen, RX_COPY_MIN);

	skb = dev_alloc_skb(alloclen + swivel + cp->crc_size);
	if (skb == NULL)
		return -1;

	*skbref = skb;
	skb_reserve(skb, swivel);

	p = skb->data;
	addr = crcaddr = NULL;
	if (hlen) { /* always copy header pages */
		i = CAS_VAL(RX_COMP2_HDR_INDEX, words[1]);
		page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
		off = CAS_VAL(RX_COMP2_HDR_OFF, words[1]) * 0x100 +
			swivel;

		i = hlen;
		if (!dlen) /* attach FCS */
			i += cp->crc_size;
		pci_dma_sync_single_for_cpu(cp->pdev, page->dma_addr + off, i,
				    PCI_DMA_FROMDEVICE);
		addr = cas_page_map(page->buffer);
		memcpy(p, addr + off, i);
		pci_dma_sync_single_for_device(cp->pdev, page->dma_addr + off, i,
				    PCI_DMA_FROMDEVICE);
		cas_page_unmap(addr);
		RX_USED_ADD(page, 0x100);
		p += hlen;
		swivel = 0;
	}


	if (alloclen < (hlen + dlen)) {
		skb_frag_t *frag = skb_shinfo(skb)->frags;

		/* normal or jumbo packets. we use frags */
		i = CAS_VAL(RX_COMP1_DATA_INDEX, words[0]);
		page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
		off = CAS_VAL(RX_COMP1_DATA_OFF, words[0]) + swivel;

		hlen = min(cp->page_size - off, dlen);
		if (hlen < 0) {
			if (netif_msg_rx_err(cp)) {
				printk(KERN_DEBUG "%s: rx page overflow: "
				       "%d\n", cp->dev->name, hlen);
			}
			dev_kfree_skb_irq(skb);
			return -1;
		}
		i = hlen;
		if (i == dlen)  /* attach FCS */
			i += cp->crc_size;
		pci_dma_sync_single_for_cpu(cp->pdev, page->dma_addr + off, i,
				    PCI_DMA_FROMDEVICE);

		/* make sure we always copy a header */
		swivel = 0;
		if (p == (char *) skb->data) { /* not split */
			addr = cas_page_map(page->buffer);
			memcpy(p, addr + off, RX_COPY_MIN);
			pci_dma_sync_single_for_device(cp->pdev, page->dma_addr + off, i,
					PCI_DMA_FROMDEVICE);
			cas_page_unmap(addr);
			off += RX_COPY_MIN;
			swivel = RX_COPY_MIN;
			RX_USED_ADD(page, cp->mtu_stride);
		} else {
			RX_USED_ADD(page, hlen);
		}
		skb_put(skb, alloclen);

		skb_shinfo(skb)->nr_frags++;
		skb->data_len += hlen - swivel;
		skb->truesize += hlen - swivel;
		skb->len      += hlen - swivel;

		get_page(page->buffer);
		frag->page = page->buffer;
		frag->page_offset = off;
		frag->size = hlen - swivel;

		/* any more data? */
		if ((words[0] & RX_COMP1_SPLIT_PKT) && ((dlen -= hlen) > 0)) {
			hlen = dlen;
			off = 0;

			i = CAS_VAL(RX_COMP2_NEXT_INDEX, words[1]);
			page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
			pci_dma_sync_single_for_cpu(cp->pdev, page->dma_addr,
					    hlen + cp->crc_size,
					    PCI_DMA_FROMDEVICE);
			pci_dma_sync_single_for_device(cp->pdev, page->dma_addr,
					    hlen + cp->crc_size,
					    PCI_DMA_FROMDEVICE);

			skb_shinfo(skb)->nr_frags++;
			skb->data_len += hlen;
			skb->len      += hlen;
			frag++;

			get_page(page->buffer);
			frag->page = page->buffer;
			frag->page_offset = 0;
			frag->size = hlen;
			RX_USED_ADD(page, hlen + cp->crc_size);
		}

		if (cp->crc_size) {
			addr = cas_page_map(page->buffer);
			crcaddr  = addr + off + hlen;
		}

	} else {
		/* copying packet */
		if (!dlen)
			goto end_copy_pkt;

		i = CAS_VAL(RX_COMP1_DATA_INDEX, words[0]);
		page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
		off = CAS_VAL(RX_COMP1_DATA_OFF, words[0]) + swivel;
		hlen = min(cp->page_size - off, dlen);
		if (hlen < 0) {
			if (netif_msg_rx_err(cp)) {
				printk(KERN_DEBUG "%s: rx page overflow: "
				       "%d\n", cp->dev->name, hlen);
			}
			dev_kfree_skb_irq(skb);
			return -1;
		}
		i = hlen;
		if (i == dlen) /* attach FCS */
			i += cp->crc_size;
		pci_dma_sync_single_for_cpu(cp->pdev, page->dma_addr + off, i,
				    PCI_DMA_FROMDEVICE);
		addr = cas_page_map(page->buffer);
		memcpy(p, addr + off, i);
		pci_dma_sync_single_for_device(cp->pdev, page->dma_addr + off, i,
				    PCI_DMA_FROMDEVICE);
		cas_page_unmap(addr);
		if (p == (char *) skb->data) /* not split */
			RX_USED_ADD(page, cp->mtu_stride);
		else
			RX_USED_ADD(page, i);

		/* any more data? */
		if ((words[0] & RX_COMP1_SPLIT_PKT) && ((dlen -= hlen) > 0)) {
			p += hlen;
			i = CAS_VAL(RX_COMP2_NEXT_INDEX, words[1]);
			page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
			pci_dma_sync_single_for_cpu(cp->pdev, page->dma_addr,
					    dlen + cp->crc_size,
					    PCI_DMA_FROMDEVICE);
			addr = cas_page_map(page->buffer);
			memcpy(p, addr, dlen + cp->crc_size);
			pci_dma_sync_single_for_device(cp->pdev, page->dma_addr,
					    dlen + cp->crc_size,
					    PCI_DMA_FROMDEVICE);
			cas_page_unmap(addr);
			RX_USED_ADD(page, dlen + cp->crc_size);
		}
end_copy_pkt:
		if (cp->crc_size) {
			addr    = NULL;
			crcaddr = skb->data + alloclen;
		}
		skb_put(skb, alloclen);
	}

	csum = (__force __sum16)htons(CAS_VAL(RX_COMP4_TCP_CSUM, words[3]));
	if (cp->crc_size) {
		/* checksum includes FCS. strip it out. */
		csum = csum_fold(csum_partial(crcaddr, cp->crc_size,
					      csum_unfold(csum)));
		if (addr)
			cas_page_unmap(addr);
	}
	skb->protocol = eth_type_trans(skb, cp->dev);
	if (skb->protocol == htons(ETH_P_IP)) {
		skb->csum = csum_unfold(~csum);
		skb->ip_summed = CHECKSUM_COMPLETE;
	} else
		skb->ip_summed = CHECKSUM_NONE;
	return len;
}


/* we can handle up to 64 rx flows at a time. we do the same thing
 * as nonreassm except that we batch up the buffers.
 * NOTE: we currently just treat each flow as a bunch of packets that
 *       we pass up. a better way would be to coalesce the packets
 *       into a jumbo packet. to do that, we need to do the following:
 *       1) the first packet will have a clean split between header and
 *          data. save both.
 *       2) each time the next flow packet comes in, extend the
 *          data length and merge the checksums.
 *       3) on flow release, fix up the header.
 *       4) make sure the higher layer doesn't care.
 * because packets get coalesced, we shouldn't run into fragment count
 * issues.
 */
static inline void cas_rx_flow_pkt(struct cas *cp, const u64 *words,
				   struct sk_buff *skb)
{
	int flowid = CAS_VAL(RX_COMP3_FLOWID, words[2]) & (N_RX_FLOWS - 1);
	struct sk_buff_head *flow = &cp->rx_flows[flowid];

	/* this is protected at a higher layer, so no need to
	 * do any additional locking here. stick the buffer
	 * at the end.
	 */
	__skb_queue_tail(flow, skb);
	if (words[0] & RX_COMP1_RELEASE_FLOW) {
		while ((skb = __skb_dequeue(flow))) {
			cas_skb_release(skb);
		}
	}
}

/* put rx descriptor back on ring. if a buffer is in use by a higher
 * layer, this will need to put in a replacement.
 */
static void cas_post_page(struct cas *cp, const int ring, const int index)
{
	cas_page_t *new;
	int entry;

	entry = cp->rx_old[ring];

	new = cas_page_swap(cp, ring, index);
	cp->init_rxds[ring][entry].buffer = cpu_to_le64(new->dma_addr);
	cp->init_rxds[ring][entry].index  =
		cpu_to_le64(CAS_BASE(RX_INDEX_NUM, index) |
			    CAS_BASE(RX_INDEX_RING, ring));

	entry = RX_DESC_ENTRY(ring, entry + 1);
	cp->rx_old[ring] = entry;

	if (entry % 4)
		return;

	if (ring == 0)
		writel(entry, cp->regs + REG_RX_KICK);
	else if ((N_RX_DESC_RINGS > 1) &&
		 (cp->cas_flags & CAS_FLAG_REG_PLUS))
		writel(entry, cp->regs + REG_PLUS_RX_KICK1);
}


/* only when things are bad */
static int cas_post_rxds_ringN(struct cas *cp, int ring, int num)
{
	unsigned int entry, last, count, released;
	int cluster;
	cas_page_t **page = cp->rx_pages[ring];

	entry = cp->rx_old[ring];

	if (netif_msg_intr(cp))
		printk(KERN_DEBUG "%s: rxd[%d] interrupt, done: %d\n",
		       cp->dev->name, ring, entry);

	cluster = -1;
	count = entry & 0x3;
	last = RX_DESC_ENTRY(ring, num ? entry + num - 4: entry - 4);
	released = 0;
	while (entry != last) {
		/* make a new buffer if it's still in use */
		if (page_count(page[entry]->buffer) > 1) {
			cas_page_t *new = cas_page_dequeue(cp);
			if (!new) {
				/* let the timer know that we need to
				 * do this again
				 */
				cp->cas_flags |= CAS_FLAG_RXD_POST(ring);
				if (!timer_pending(&cp->link_timer))
					mod_timer(&cp->link_timer, jiffies +
						  CAS_LINK_FAST_TIMEOUT);
				cp->rx_old[ring]  = entry;
				cp->rx_last[ring] = num ? num - released : 0;
				return -ENOMEM;
			}
			spin_lock(&cp->rx_inuse_lock);
			list_add(&page[entry]->list, &cp->rx_inuse_list);
			spin_unlock(&cp->rx_inuse_lock);
			cp->init_rxds[ring][entry].buffer =
				cpu_to_le64(new->dma_addr);
			page[entry] = new;

		}

		if (++count == 4) {
			cluster = entry;
			count = 0;
		}
		released++;
		entry = RX_DESC_ENTRY(ring, entry + 1);
	}
	cp->rx_old[ring] = entry;

	if (cluster < 0)
		return 0;

	if (ring == 0)
		writel(cluster, cp->regs + REG_RX_KICK);
	else if ((N_RX_DESC_RINGS > 1) &&
		 (cp->cas_flags & CAS_FLAG_REG_PLUS))
		writel(cluster, cp->regs + REG_PLUS_RX_KICK1);
	return 0;
}


/* process a completion ring. packets are set up in three basic ways:
 * small packets: should be copied header + data in single buffer.
 * large packets: header and data in a single buffer.
 * split packets: header in a separate buffer from data.
 *                data may be in multiple pages. data may be > 256
 *                bytes but in a single page.
 *
 * NOTE: RX page posting is done in this routine as well. while there's
 *       the capability of using multiple RX completion rings, it isn't
 *       really worthwhile due to the fact that the page posting will
 *       force serialization on the single descriptor ring.
 */
static int cas_rx_ringN(struct cas *cp, int ring, int budget)
{
	struct cas_rx_comp *rxcs = cp->init_rxcs[ring];
	int entry, drops;
	int npackets = 0;

	if (netif_msg_intr(cp))
		printk(KERN_DEBUG "%s: rx[%d] interrupt, done: %d/%d\n",
		       cp->dev->name, ring,
		       readl(cp->regs + REG_RX_COMP_HEAD),
		       cp->rx_new[ring]);

	entry = cp->rx_new[ring];
	drops = 0;
	while (1) {
		struct cas_rx_comp *rxc = rxcs + entry;
		struct sk_buff *uninitialized_var(skb);
		int type, len;
		u64 words[4];
		int i, dring;

		words[0] = le64_to_cpu(rxc->word1);
		words[1] = le64_to_cpu(rxc->word2);
		words[2] = le64_to_cpu(rxc->word3);
		words[3] = le64_to_cpu(rxc->word4);

		/* don't touch if still owned by hw */
		type = CAS_VAL(RX_COMP1_TYPE, words[0]);
		if (type == 0)
			break;

		/* hw hasn't cleared the zero bit yet */
		if (words[3] & RX_COMP4_ZERO) {
			break;
		}

		/* get info on the packet */
		if (words[3] & (RX_COMP4_LEN_MISMATCH | RX_COMP4_BAD)) {
			spin_lock(&cp->stat_lock[ring]);
			cp->net_stats[ring].rx_errors++;
			if (words[3] & RX_COMP4_LEN_MISMATCH)
				cp->net_stats[ring].rx_length_errors++;
			if (words[3] & RX_COMP4_BAD)
				cp->net_stats[ring].rx_crc_errors++;
			spin_unlock(&cp->stat_lock[ring]);

			/* We'll just return it to Cassini. */
		drop_it:
			spin_lock(&cp->stat_lock[ring]);
			++cp->net_stats[ring].rx_dropped;
			spin_unlock(&cp->stat_lock[ring]);
			goto next;
		}

		len = cas_rx_process_pkt(cp, rxc, entry, words, &skb);
		if (len < 0) {
			++drops;
			goto drop_it;
		}

		/* see if it's a flow re-assembly or not. the driver
		 * itself handles release back up.
		 */
		if (RX_DONT_BATCH || (type == 0x2)) {
			/* non-reassm: these always get released */
			cas_skb_release(skb);
		} else {
			cas_rx_flow_pkt(cp, words, skb);
		}

		spin_lock(&cp->stat_lock[ring]);
		cp->net_stats[ring].rx_packets++;
		cp->net_stats[ring].rx_bytes += len;
		spin_unlock(&cp->stat_lock[ring]);

	next:
		npackets++;

		/* should it be released? */
		if (words[0] & RX_COMP1_RELEASE_HDR) {
			i = CAS_VAL(RX_COMP2_HDR_INDEX, words[1]);
			dring = CAS_VAL(RX_INDEX_RING, i);
			i = CAS_VAL(RX_INDEX_NUM, i);
			cas_post_page(cp, dring, i);
		}

		if (words[0] & RX_COMP1_RELEASE_DATA) {
			i = CAS_VAL(RX_COMP1_DATA_INDEX, words[0]);
			dring = CAS_VAL(RX_INDEX_RING, i);
			i = CAS_VAL(RX_INDEX_NUM, i);
			cas_post_page(cp, dring, i);
		}

		if (words[0] & RX_COMP1_RELEASE_NEXT) {
			i = CAS_VAL(RX_COMP2_NEXT_INDEX, words[1]);
			dring = CAS_VAL(RX_INDEX_RING, i);
			i = CAS_VAL(RX_INDEX_NUM, i);
			cas_post_page(cp, dring, i);
		}

		/* skip to the next entry */
		entry = RX_COMP_ENTRY(ring, entry + 1 +
				      CAS_VAL(RX_COMP1_SKIP, words[0]));
#ifdef USE_NAPI
		if (budget && (npackets >= budget))
			break;
#endif
	}
	cp->rx_new[ring] = entry;

	if (drops)
		printk(KERN_INFO "%s: Memory squeeze, deferring packet.\n",
		       cp->dev->name);
	return npackets;
}


/* put completion entries back on the ring */
static void cas_post_rxcs_ringN(struct net_device *dev,
				struct cas *cp, int ring)
{
	struct cas_rx_comp *rxc = cp->init_rxcs[ring];
	int last, entry;

	last = cp->rx_cur[ring];
	entry = cp->rx_new[ring];
	if (netif_msg_intr(cp))
		printk(KERN_DEBUG "%s: rxc[%d] interrupt, done: %d/%d\n",
		       dev->name, ring, readl(cp->regs + REG_RX_COMP_HEAD),
		       entry);

	/* zero and re-mark descriptors */
	while (last != entry) {
		cas_rxc_init(rxc + last);
		last = RX_COMP_ENTRY(ring, last + 1);
	}
	cp->rx_cur[ring] = last;

	if (ring == 0)
		writel(last, cp->regs + REG_RX_COMP_TAIL);
	else if (cp->cas_flags & CAS_FLAG_REG_PLUS)
		writel(last, cp->regs + REG_PLUS_RX_COMPN_TAIL(ring));
}



/* cassini can use all four PCI interrupts for the completion ring.
 * rings 3 and 4 are identical
 */
#if defined(USE_PCI_INTC) || defined(USE_PCI_INTD)
static inline void cas_handle_irqN(struct net_device *dev,
				   struct cas *cp, const u32 status,
				   const int ring)
{
	if (status & (INTR_RX_COMP_FULL_ALT | INTR_RX_COMP_AF_ALT))
		cas_post_rxcs_ringN(dev, cp, ring);
}

static irqreturn_t cas_interruptN(int irq, void *dev_id)
{
	struct net_device *dev = dev_id;
	struct cas *cp = netdev_priv(dev);
	unsigned long flags;
	int ring;
	u32 status = readl(cp->regs + REG_PLUS_INTRN_STATUS(ring));

	/* check for shared irq */
	if (status == 0)
		return IRQ_NONE;

	ring = (irq == cp->pci_irq_INTC) ? 2 : 3;
	spin_lock_irqsave(&cp->lock, flags);
	if (status & INTR_RX_DONE_ALT) { /* handle rx separately */
#ifdef USE_NAPI
		cas_mask_intr(cp);
		napi_schedule(&cp->napi);
#else
		cas_rx_ringN(cp, ring, 0);
#endif
		status &= ~INTR_RX_DONE_ALT;
	}

	if (status)
		cas_handle_irqN(dev, cp, status, ring);
	spin_unlock_irqrestore(&cp->lock, flags);
	return IRQ_HANDLED;
}
#endif

#ifdef USE_PCI_INTB
/* everything but rx packets */
static inline void cas_handle_irq1(struct cas *cp, const u32 status)
{
	if (status & INTR_RX_BUF_UNAVAIL_1) {
		/* Frame arrived, no free RX buffers available.
		 * NOTE: we can get this on a link transition. */
		cas_post_rxds_ringN(cp, 1, 0);
		spin_lock(&cp->stat_lock[1]);
		cp->net_stats[1].rx_dropped++;
		spin_unlock(&cp->stat_lock[1]);
	}

	if (status & INTR_RX_BUF_AE_1)
		cas_post_rxds_ringN(cp, 1, RX_DESC_RINGN_SIZE(1) -
				    RX_AE_FREEN_VAL(1));

	if (status & (INTR_RX_COMP_AF | INTR_RX_COMP_FULL))
		cas_post_rxcs_ringN(cp, 1);
}

/* ring 2 handles a few more events than 3 and 4 */
static irqreturn_t cas_interrupt1(int irq, void *dev_id)
{
	struct net_device *dev = dev_id;
	struct cas *cp = netdev_priv(dev);
	unsigned long flags;
	u32 status = readl(cp->regs + REG_PLUS_INTRN_STATUS(1));

	/* check for shared interrupt */
	if (status == 0)
		return IRQ_NONE;

	spin_lock_irqsave(&cp->lock, flags);
	if (status & INTR_RX_DONE_ALT) { /* handle rx separately */
#ifdef USE_NAPI
		cas_mask_intr(cp);
		napi_schedule(&cp->napi);
#else
		cas_rx_ringN(cp, 1, 0);
#endif
		status &= ~INTR_RX_DONE_ALT;
	}
	if (status)
		cas_handle_irq1(cp, status);
	spin_unlock_irqrestore(&cp->lock, flags);
	return IRQ_HANDLED;
}
#endif

static inline void cas_handle_irq(struct net_device *dev,
				  struct cas *cp, const u32 status)
{
	/* housekeeping interrupts */
	if (status & INTR_ERROR_MASK)
		cas_abnormal_irq(dev, cp, status);

	if (status & INTR_RX_BUF_UNAVAIL) {
		/* Frame arrived, no free RX buffers available.
		 * NOTE: we can get this on a link transition.
		 */
		cas_post_rxds_ringN(cp, 0, 0);
		spin_lock(&cp->stat_lock[0]);
		cp->net_stats[0].rx_dropped++;
		spin_unlock(&cp->stat_lock[0]);
	} else if (status & INTR_RX_BUF_AE) {
		cas_post_rxds_ringN(cp, 0, RX_DESC_RINGN_SIZE(0) -
				    RX_AE_FREEN_VAL(0));
	}

	if (status & (INTR_RX_COMP_AF | INTR_RX_COMP_FULL))
		cas_post_rxcs_ringN(dev, cp, 0);
}

static irqreturn_t cas_interrupt(int irq, void *dev_id)
{
	struct net_device *dev = dev_id;
	struct cas *cp = netdev_priv(dev);
	unsigned long flags;
	u32 status = readl(cp->regs + REG_INTR_STATUS);

	if (status == 0)
		return IRQ_NONE;

	spin_lock_irqsave(&cp->lock, flags);
	if (status & (INTR_TX_ALL | INTR_TX_INTME)) {
		cas_tx(dev, cp, status);
		status &= ~(INTR_TX_ALL | INTR_TX_INTME);
	}

	if (status & INTR_RX_DONE) {
#ifdef USE_NAPI
		cas_mask_intr(cp);
		napi_schedule(&cp->napi);
#else
		cas_rx_ringN(cp, 0, 0);
#endif
		status &= ~INTR_RX_DONE;
	}

	if (status)
		cas_handle_irq(dev, cp, status);
	spin_unlock_irqrestore(&cp->lock, flags);
	return IRQ_HANDLED;
}


#ifdef USE_NAPI
static int cas_poll(struct napi_struct *napi, int budget)
{
	struct cas *cp = container_of(napi, struct cas, napi);
	struct net_device *dev = cp->dev;
	int i, enable_intr, credits;
	u32 status = readl(cp->regs + REG_INTR_STATUS);
	unsigned long flags;

	spin_lock_irqsave(&cp->lock, flags);
	cas_tx(dev, cp, status);
	spin_unlock_irqrestore(&cp->lock, flags);

	/* NAPI rx packets. we spread the credits across all of the
	 * rxc rings
	 *
	 * to make sure we're fair with the work we loop through each
	 * ring N_RX_COMP_RING times with a request of
	 * budget / N_RX_COMP_RINGS
	 */
	enable_intr = 1;
	credits = 0;
	for (i = 0; i < N_RX_COMP_RINGS; i++) {
		int j;
		for (j = 0; j < N_RX_COMP_RINGS; j++) {
			credits += cas_rx_ringN(cp, j, budget / N_RX_COMP_RINGS);
			if (credits >= budget) {
				enable_intr = 0;
				goto rx_comp;
			}
		}
	}

rx_comp:
	/* final rx completion */
	spin_lock_irqsave(&cp->lock, flags);
	if (status)
		cas_handle_irq(dev, cp, status);

#ifdef USE_PCI_INTB
	if (N_RX_COMP_RINGS > 1) {
		status = readl(cp->regs + REG_PLUS_INTRN_STATUS(1));
		if (status)
			cas_handle_irq1(dev, cp, status);
	}
#endif

#ifdef USE_PCI_INTC
	if (N_RX_COMP_RINGS > 2) {
		status = readl(cp->regs + REG_PLUS_INTRN_STATUS(2));
		if (status)
			cas_handle_irqN(dev, cp, status, 2);
	}
#endif

#ifdef USE_PCI_INTD
	if (N_RX_COMP_RINGS > 3) {
		status = readl(cp->regs + REG_PLUS_INTRN_STATUS(3));
		if (status)
			cas_handle_irqN(dev, cp, status, 3);
	}
#endif
	spin_unlock_irqrestore(&cp->lock, flags);
	if (enable_intr) {
		napi_complete(napi);
		cas_unmask_intr(cp);
	}
	return credits;
}
#endif

#ifdef CONFIG_NET_POLL_CONTROLLER
static void cas_netpoll(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);

	cas_disable_irq(cp, 0);
	cas_interrupt(cp->pdev->irq, dev);
	cas_enable_irq(cp, 0);

#ifdef USE_PCI_INTB
	if (N_RX_COMP_RINGS > 1) {
		/* cas_interrupt1(); */
	}
#endif
#ifdef USE_PCI_INTC
	if (N_RX_COMP_RINGS > 2) {
		/* cas_interruptN(); */
	}
#endif
#ifdef USE_PCI_INTD
	if (N_RX_COMP_RINGS > 3) {
		/* cas_interruptN(); */
	}
#endif
}
#endif

static void cas_tx_timeout(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);

	printk(KERN_ERR "%s: transmit timed out, resetting\n", dev->name);
	if (!cp->hw_running) {
		printk("%s: hrm.. hw not running!\n", dev->name);
		return;
	}

	printk(KERN_ERR "%s: MIF_STATE[%08x]\n",
	       dev->name, readl(cp->regs + REG_MIF_STATE_MACHINE));

	printk(KERN_ERR "%s: MAC_STATE[%08x]\n",
	       dev->name, readl(cp->regs + REG_MAC_STATE_MACHINE));

	printk(KERN_ERR "%s: TX_STATE[%08x:%08x:%08x] "
	       "FIFO[%08x:%08x:%08x] SM1[%08x] SM2[%08x]\n",
	       dev->name,
	       readl(cp->regs + REG_TX_CFG),
	       readl(cp->regs + REG_MAC_TX_STATUS),
	       readl(cp->regs + REG_MAC_TX_CFG),
	       readl(cp->regs + REG_TX_FIFO_PKT_CNT),
	       readl(cp->regs + REG_TX_FIFO_WRITE_PTR),
	       readl(cp->regs + REG_TX_FIFO_READ_PTR),
	       readl(cp->regs + REG_TX_SM_1),
	       readl(cp->regs + REG_TX_SM_2));

	printk(KERN_ERR "%s: RX_STATE[%08x:%08x:%08x]\n",
	       dev->name,
	       readl(cp->regs + REG_RX_CFG),
	       readl(cp->regs + REG_MAC_RX_STATUS),
	       readl(cp->regs + REG_MAC_RX_CFG));

	printk(KERN_ERR "%s: HP_STATE[%08x:%08x:%08x:%08x]\n",
	       dev->name,
	       readl(cp->regs + REG_HP_STATE_MACHINE),
	       readl(cp->regs + REG_HP_STATUS0),
	       readl(cp->regs + REG_HP_STATUS1),
	       readl(cp->regs + REG_HP_STATUS2));

#if 1
	atomic_inc(&cp->reset_task_pending);
	atomic_inc(&cp->reset_task_pending_all);
	schedule_work(&cp->reset_task);
#else
	atomic_set(&cp->reset_task_pending, CAS_RESET_ALL);
	schedule_work(&cp->reset_task);
#endif
}

static inline int cas_intme(int ring, int entry)
{
	/* Algorithm: IRQ every 1/2 of descriptors. */
	if (!(entry & ((TX_DESC_RINGN_SIZE(ring) >> 1) - 1)))
		return 1;
	return 0;
}


static void cas_write_txd(struct cas *cp, int ring, int entry,
			  dma_addr_t mapping, int len, u64 ctrl, int last)
{
	struct cas_tx_desc *txd = cp->init_txds[ring] + entry;

	ctrl |= CAS_BASE(TX_DESC_BUFLEN, len);
	if (cas_intme(ring, entry))
		ctrl |= TX_DESC_INTME;
	if (last)
		ctrl |= TX_DESC_EOF;
	txd->control = cpu_to_le64(ctrl);
	txd->buffer = cpu_to_le64(mapping);
}

static inline void *tx_tiny_buf(struct cas *cp, const int ring,
				const int entry)
{
	return cp->tx_tiny_bufs[ring] + TX_TINY_BUF_LEN*entry;
}

static inline dma_addr_t tx_tiny_map(struct cas *cp, const int ring,
				     const int entry, const int tentry)
{
	cp->tx_tiny_use[ring][tentry].nbufs++;
	cp->tx_tiny_use[ring][entry].used = 1;
	return cp->tx_tiny_dvma[ring] + TX_TINY_BUF_LEN*entry;
}

static inline int cas_xmit_tx_ringN(struct cas *cp, int ring,
				    struct sk_buff *skb)
{
	struct net_device *dev = cp->dev;
	int entry, nr_frags, frag, tabort, tentry;
	dma_addr_t mapping;
	unsigned long flags;
	u64 ctrl;
	u32 len;

	spin_lock_irqsave(&cp->tx_lock[ring], flags);

	/* This is a hard error, log it. */
	if (TX_BUFFS_AVAIL(cp, ring) <=
	    CAS_TABORT(cp)*(skb_shinfo(skb)->nr_frags + 1)) {
		netif_stop_queue(dev);
		spin_unlock_irqrestore(&cp->tx_lock[ring], flags);
		printk(KERN_ERR PFX "%s: BUG! Tx Ring full when "
		       "queue awake!\n", dev->name);
		return 1;
	}

	ctrl = 0;
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
		const u64 csum_start_off = skb_transport_offset(skb);
		const u64 csum_stuff_off = csum_start_off + skb->csum_offset;

		ctrl =  TX_DESC_CSUM_EN |
			CAS_BASE(TX_DESC_CSUM_START, csum_start_off) |
			CAS_BASE(TX_DESC_CSUM_STUFF, csum_stuff_off);
	}

	entry = cp->tx_new[ring];
	cp->tx_skbs[ring][entry] = skb;

	nr_frags = skb_shinfo(skb)->nr_frags;
	len = skb_headlen(skb);
	mapping = pci_map_page(cp->pdev, virt_to_page(skb->data),
			       offset_in_page(skb->data), len,
			       PCI_DMA_TODEVICE);

	tentry = entry;
	tabort = cas_calc_tabort(cp, (unsigned long) skb->data, len);
	if (unlikely(tabort)) {
		/* NOTE: len is always >  tabort */
		cas_write_txd(cp, ring, entry, mapping, len - tabort,
			      ctrl | TX_DESC_SOF, 0);
		entry = TX_DESC_NEXT(ring, entry);

		skb_copy_from_linear_data_offset(skb, len - tabort,
			      tx_tiny_buf(cp, ring, entry), tabort);
		mapping = tx_tiny_map(cp, ring, entry, tentry);
		cas_write_txd(cp, ring, entry, mapping, tabort, ctrl,
			      (nr_frags == 0));
	} else {
		cas_write_txd(cp, ring, entry, mapping, len, ctrl |
			      TX_DESC_SOF, (nr_frags == 0));
	}
	entry = TX_DESC_NEXT(ring, entry);

	for (frag = 0; frag < nr_frags; frag++) {
		skb_frag_t *fragp = &skb_shinfo(skb)->frags[frag];

		len = fragp->size;
		mapping = pci_map_page(cp->pdev, fragp->page,
				       fragp->page_offset, len,
				       PCI_DMA_TODEVICE);

		tabort = cas_calc_tabort(cp, fragp->page_offset, len);
		if (unlikely(tabort)) {
			void *addr;

			/* NOTE: len is always > tabort */
			cas_write_txd(cp, ring, entry, mapping, len - tabort,
				      ctrl, 0);
			entry = TX_DESC_NEXT(ring, entry);

			addr = cas_page_map(fragp->page);
			memcpy(tx_tiny_buf(cp, ring, entry),
			       addr + fragp->page_offset + len - tabort,
			       tabort);
			cas_page_unmap(addr);
			mapping = tx_tiny_map(cp, ring, entry, tentry);
			len     = tabort;
		}

		cas_write_txd(cp, ring, entry, mapping, len, ctrl,
			      (frag + 1 == nr_frags));
		entry = TX_DESC_NEXT(ring, entry);
	}

	cp->tx_new[ring] = entry;
	if (TX_BUFFS_AVAIL(cp, ring) <= CAS_TABORT(cp)*(MAX_SKB_FRAGS + 1))
		netif_stop_queue(dev);

	if (netif_msg_tx_queued(cp))
		printk(KERN_DEBUG "%s: tx[%d] queued, slot %d, skblen %d, "
		       "avail %d\n",
		       dev->name, ring, entry, skb->len,
		       TX_BUFFS_AVAIL(cp, ring));
	writel(entry, cp->regs + REG_TX_KICKN(ring));
	spin_unlock_irqrestore(&cp->tx_lock[ring], flags);
	return 0;
}

static netdev_tx_t cas_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);

	/* this is only used as a load-balancing hint, so it doesn't
	 * need to be SMP safe
	 */
	static int ring;

	if (skb_padto(skb, cp->min_frame_size))
		return NETDEV_TX_OK;

	/* XXX: we need some higher-level QoS hooks to steer packets to
	 *      individual queues.
	 */
	if (cas_xmit_tx_ringN(cp, ring++ & N_TX_RINGS_MASK, skb))
		return NETDEV_TX_BUSY;
	dev->trans_start = jiffies;
	return NETDEV_TX_OK;
}

static void cas_init_tx_dma(struct cas *cp)
{
	u64 desc_dma = cp->block_dvma;
	unsigned long off;
	u32 val;
	int i;

	/* set up tx completion writeback registers. must be 8-byte aligned */
#ifdef USE_TX_COMPWB
	off = offsetof(struct cas_init_block, tx_compwb);
	writel((desc_dma + off) >> 32, cp->regs + REG_TX_COMPWB_DB_HI);
	writel((desc_dma + off) & 0xffffffff, cp->regs + REG_TX_COMPWB_DB_LOW);
#endif

	/* enable completion writebacks, enable paced mode,
	 * disable read pipe, and disable pre-interrupt compwbs
	 */
	val =   TX_CFG_COMPWB_Q1 | TX_CFG_COMPWB_Q2 |
		TX_CFG_COMPWB_Q3 | TX_CFG_COMPWB_Q4 |
		TX_CFG_DMA_RDPIPE_DIS | TX_CFG_PACED_MODE |
		TX_CFG_INTR_COMPWB_DIS;

	/* write out tx ring info and tx desc bases */
	for (i = 0; i < MAX_TX_RINGS; i++) {
		off = (unsigned long) cp->init_txds[i] -
			(unsigned long) cp->init_block;

		val |= CAS_TX_RINGN_BASE(i);
		writel((desc_dma + off) >> 32, cp->regs + REG_TX_DBN_HI(i));
		writel((desc_dma + off) & 0xffffffff, cp->regs +
		       REG_TX_DBN_LOW(i));
		/* don't zero out the kick register here as the system
		 * will wedge
		 */
	}
	writel(val, cp->regs + REG_TX_CFG);

	/* program max burst sizes. these numbers should be different
	 * if doing QoS.
	 */
#ifdef USE_QOS
	writel(0x800, cp->regs + REG_TX_MAXBURST_0);
	writel(0x1600, cp->regs + REG_TX_MAXBURST_1);
	writel(0x2400, cp->regs + REG_TX_MAXBURST_2);
	writel(0x4800, cp->regs + REG_TX_MAXBURST_3);
#else
	writel(0x800, cp->regs + REG_TX_MAXBURST_0);
	writel(0x800, cp->regs + REG_TX_MAXBURST_1);
	writel(0x800, cp->regs + REG_TX_MAXBURST_2);
	writel(0x800, cp->regs + REG_TX_MAXBURST_3);
#endif
}

/* Must be invoked under cp->lock. */
static inline void cas_init_dma(struct cas *cp)
{
	cas_init_tx_dma(cp);
	cas_init_rx_dma(cp);
}

/* Must be invoked under cp->lock. */
static u32 cas_setup_multicast(struct cas *cp)
{
	u32 rxcfg = 0;
	int i;

	if (cp->dev->flags & IFF_PROMISC) {
		rxcfg |= MAC_RX_CFG_PROMISC_EN;

	} else if (cp->dev->flags & IFF_ALLMULTI) {
	    	for (i=0; i < 16; i++)
			writel(0xFFFF, cp->regs + REG_MAC_HASH_TABLEN(i));
		rxcfg |= MAC_RX_CFG_HASH_FILTER_EN;

	} else {
		u16 hash_table[16];
		u32 crc;
		struct dev_mc_list *dmi = cp->dev->mc_list;
		int i;

		/* use the alternate mac address registers for the
		 * first 15 multicast addresses
		 */
		for (i = 1; i <= CAS_MC_EXACT_MATCH_SIZE; i++) {
			if (!dmi) {
				writel(0x0, cp->regs + REG_MAC_ADDRN(i*3 + 0));
				writel(0x0, cp->regs + REG_MAC_ADDRN(i*3 + 1));
				writel(0x0, cp->regs + REG_MAC_ADDRN(i*3 + 2));
				continue;
			}
			writel((dmi->dmi_addr[4] << 8) | dmi->dmi_addr[5],
			       cp->regs + REG_MAC_ADDRN(i*3 + 0));
			writel((dmi->dmi_addr[2] << 8) | dmi->dmi_addr[3],
			       cp->regs + REG_MAC_ADDRN(i*3 + 1));
			writel((dmi->dmi_addr[0] << 8) | dmi->dmi_addr[1],
			       cp->regs + REG_MAC_ADDRN(i*3 + 2));
			dmi = dmi->next;
		}

		/* use hw hash table for the next series of
		 * multicast addresses
		 */
		memset(hash_table, 0, sizeof(hash_table));
		while (dmi) {
 			crc = ether_crc_le(ETH_ALEN, dmi->dmi_addr);
			crc >>= 24;
			hash_table[crc >> 4] |= 1 << (15 - (crc & 0xf));
			dmi = dmi->next;
		}
	    	for (i=0; i < 16; i++)
			writel(hash_table[i], cp->regs +
			       REG_MAC_HASH_TABLEN(i));
		rxcfg |= MAC_RX_CFG_HASH_FILTER_EN;
	}

	return rxcfg;
}

/* must be invoked under cp->stat_lock[N_TX_RINGS] */
static void cas_clear_mac_err(struct cas *cp)
{
	writel(0, cp->regs + REG_MAC_COLL_NORMAL);
	writel(0, cp->regs + REG_MAC_COLL_FIRST);
	writel(0, cp->regs + REG_MAC_COLL_EXCESS);
	writel(0, cp->regs + REG_MAC_COLL_LATE);
	writel(0, cp->regs + REG_MAC_TIMER_DEFER);
	writel(0, cp->regs + REG_MAC_ATTEMPTS_PEAK);
	writel(0, cp->regs + REG_MAC_RECV_FRAME);
	writel(0, cp->regs + REG_MAC_LEN_ERR);
	writel(0, cp->regs + REG_MAC_ALIGN_ERR);
	writel(0, cp->regs + REG_MAC_FCS_ERR);
	writel(0, cp->regs + REG_MAC_RX_CODE_ERR);
}


static void cas_mac_reset(struct cas *cp)
{
	int i;

	/* do both TX and RX reset */
	writel(0x1, cp->regs + REG_MAC_TX_RESET);
	writel(0x1, cp->regs + REG_MAC_RX_RESET);

	/* wait for TX */
	i = STOP_TRIES;
	while (i-- > 0) {
		if (readl(cp->regs + REG_MAC_TX_RESET) == 0)
			break;
		udelay(10);
	}

	/* wait for RX */
	i = STOP_TRIES;
	while (i-- > 0) {
		if (readl(cp->regs + REG_MAC_RX_RESET) == 0)
			break;
		udelay(10);
	}

	if (readl(cp->regs + REG_MAC_TX_RESET) |
	    readl(cp->regs + REG_MAC_RX_RESET))
		printk(KERN_ERR "%s: mac tx[%d]/rx[%d] reset failed [%08x]\n",
		       cp->dev->name, readl(cp->regs + REG_MAC_TX_RESET),
		       readl(cp->regs + REG_MAC_RX_RESET),
		       readl(cp->regs + REG_MAC_STATE_MACHINE));
}


/* Must be invoked under cp->lock. */
static void cas_init_mac(struct cas *cp)
{
	unsigned char *e = &cp->dev->dev_addr[0];
	int i;
#ifdef CONFIG_CASSINI_MULTICAST_REG_WRITE
	u32 rxcfg;
#endif
	cas_mac_reset(cp);

	/* setup core arbitration weight register */
	writel(CAWR_RR_DIS, cp->regs + REG_CAWR);

	/* XXX Use pci_dma_burst_advice() */
#if !defined(CONFIG_SPARC64) && !defined(CONFIG_ALPHA)
	/* set the infinite burst register for chips that don't have
	 * pci issues.
	 */
	if ((cp->cas_flags & CAS_FLAG_TARGET_ABORT) == 0)
		writel(INF_BURST_EN, cp->regs + REG_INF_BURST);
#endif

	writel(0x1BF0, cp->regs + REG_MAC_SEND_PAUSE);

	writel(0x00, cp->regs + REG_MAC_IPG0);
	writel(0x08, cp->regs + REG_MAC_IPG1);
	writel(0x04, cp->regs + REG_MAC_IPG2);

	/* change later for 802.3z */
	writel(0x40, cp->regs + REG_MAC_SLOT_TIME);

	/* min frame + FCS */
	writel(ETH_ZLEN + 4, cp->regs + REG_MAC_FRAMESIZE_MIN);

	/* Ethernet payload + header + FCS + optional VLAN tag. NOTE: we
	 * specify the maximum frame size to prevent RX tag errors on
	 * oversized frames.
	 */
	writel(CAS_BASE(MAC_FRAMESIZE_MAX_BURST, 0x2000) |
	       CAS_BASE(MAC_FRAMESIZE_MAX_FRAME,
			(CAS_MAX_MTU + ETH_HLEN + 4 + 4)),
	       cp->regs + REG_MAC_FRAMESIZE_MAX);

	/* NOTE: crc_size is used as a surrogate for half-duplex.
	 * workaround saturn half-duplex issue by increasing preamble
	 * size to 65 bytes.
	 */
	if ((cp->cas_flags & CAS_FLAG_SATURN) && cp->crc_size)
		writel(0x41, cp->regs + REG_MAC_PA_SIZE);
	else
		writel(0x07, cp->regs + REG_MAC_PA_SIZE);
	writel(0x04, cp->regs + REG_MAC_JAM_SIZE);
	writel(0x10, cp->regs + REG_MAC_ATTEMPT_LIMIT);
	writel(0x8808, cp->regs + REG_MAC_CTRL_TYPE);

	writel((e[5] | (e[4] << 8)) & 0x3ff, cp->regs + REG_MAC_RANDOM_SEED);

	writel(0, cp->regs + REG_MAC_ADDR_FILTER0);
	writel(0, cp->regs + REG_MAC_ADDR_FILTER1);
	writel(0, cp->regs + REG_MAC_ADDR_FILTER2);
	writel(0, cp->regs + REG_MAC_ADDR_FILTER2_1_MASK);
	writel(0, cp->regs + REG_MAC_ADDR_FILTER0_MASK);

	/* setup mac address in perfect filter array */
	for (i = 0; i < 45; i++)
		writel(0x0, cp->regs + REG_MAC_ADDRN(i));

	writel((e[4] << 8) | e[5], cp->regs + REG_MAC_ADDRN(0));
	writel((e[2] << 8) | e[3], cp->regs + REG_MAC_ADDRN(1));
	writel((e[0] << 8) | e[1], cp->regs + REG_MAC_ADDRN(2));

	writel(0x0001, cp->regs + REG_MAC_ADDRN(42));
	writel(0xc200, cp->regs + REG_MAC_ADDRN(43));
	writel(0x0180, cp->regs + REG_MAC_ADDRN(44));

#ifndef CONFIG_CASSINI_MULTICAST_REG_WRITE
	cp->mac_rx_cfg = cas_setup_multicast(cp);
#else
	/* WTZ: Do what Adrian did in cas_set_multicast. Doing
	 * a writel does not seem to be necessary because Cassini
	 * seems to preserve the configuration when we do the reset.
	 * If the chip is in trouble, though, it is not clear if we
	 * can really count on this behavior. cas_set_multicast uses
	 * spin_lock_irqsave, but we are called only in cas_init_hw and
	 * cas_init_hw is protected by cas_lock_all, which calls
	 * spin_lock_irq (so it doesn't need to save the flags, and
	 * we should be OK for the writel, as that is the only
	 * difference).
	 */
	cp->mac_rx_cfg = rxcfg = cas_setup_multicast(cp);
	writel(rxcfg, cp->regs + REG_MAC_RX_CFG);
#endif
	spin_lock(&cp->stat_lock[N_TX_RINGS]);
	cas_clear_mac_err(cp);
	spin_unlock(&cp->stat_lock[N_TX_RINGS]);

	/* Setup MAC interrupts.  We want to get all of the interesting
	 * counter expiration events, but we do not want to hear about
	 * normal rx/tx as the DMA engine tells us that.
	 */
	writel(MAC_TX_FRAME_XMIT, cp->regs + REG_MAC_TX_MASK);
	writel(MAC_RX_FRAME_RECV, cp->regs + REG_MAC_RX_MASK);

	/* Don't enable even the PAUSE interrupts for now, we
	 * make no use of those events other than to record them.
	 */
	writel(0xffffffff, cp->regs + REG_MAC_CTRL_MASK);
}

/* Must be invoked under cp->lock. */
static void cas_init_pause_thresholds(struct cas *cp)
{
	/* Calculate pause thresholds.  Setting the OFF threshold to the
	 * full RX fifo size effectively disables PAUSE generation
	 */
	if (cp->rx_fifo_size <= (2 * 1024)) {
		cp->rx_pause_off = cp->rx_pause_on = cp->rx_fifo_size;
	} else {
		int max_frame = (cp->dev->mtu + ETH_HLEN + 4 + 4 + 64) & ~63;
		if (max_frame * 3 > cp->rx_fifo_size) {
			cp->rx_pause_off = 7104;
			cp->rx_pause_on  = 960;
		} else {
			int off = (cp->rx_fifo_size - (max_frame * 2));
			int on = off - max_frame;
			cp->rx_pause_off = off;
			cp->rx_pause_on = on;
		}
	}
}

static int cas_vpd_match(const void __iomem *p, const char *str)
{
	int len = strlen(str) + 1;
	int i;

	for (i = 0; i < len; i++) {
		if (readb(p + i) != str[i])
			return 0;
	}
	return 1;
}


/* get the mac address by reading the vpd information in the rom.
 * also get the phy type and determine if there's an entropy generator.
 * NOTE: this is a bit convoluted for the following reasons:
 *  1) vpd info has order-dependent mac addresses for multinic cards
 *  2) the only way to determine the nic order is to use the slot
 *     number.
 *  3) fiber cards don't have bridges, so their slot numbers don't
 *     mean anything.
 *  4) we don't actually know we have a fiber card until after
 *     the mac addresses are parsed.
 */
static int cas_get_vpd_info(struct cas *cp, unsigned char *dev_addr,
			    const int offset)
{
	void __iomem *p = cp->regs + REG_EXPANSION_ROM_RUN_START;
	void __iomem *base, *kstart;
	int i, len;
	int found = 0;
#define VPD_FOUND_MAC        0x01
#define VPD_FOUND_PHY        0x02

	int phy_type = CAS_PHY_MII_MDIO0; /* default phy type */
	int mac_off  = 0;

	/* give us access to the PROM */
	writel(BIM_LOCAL_DEV_PROM | BIM_LOCAL_DEV_PAD,
	       cp->regs + REG_BIM_LOCAL_DEV_EN);

	/* check for an expansion rom */
	if (readb(p) != 0x55 || readb(p + 1) != 0xaa)
		goto use_random_mac_addr;

	/* search for beginning of vpd */
	base = NULL;
	for (i = 2; i < EXPANSION_ROM_SIZE; i++) {
		/* check for PCIR */
		if ((readb(p + i + 0) == 0x50) &&
		    (readb(p + i + 1) == 0x43) &&
		    (readb(p + i + 2) == 0x49) &&
		    (readb(p + i + 3) == 0x52)) {
			base = p + (readb(p + i + 8) |
				    (readb(p + i + 9) << 8));
			break;
		}
	}

	if (!base || (readb(base) != 0x82))
		goto use_random_mac_addr;

	i = (readb(base + 1) | (readb(base + 2) << 8)) + 3;
	while (i < EXPANSION_ROM_SIZE) {
		if (readb(base + i) != 0x90) /* no vpd found */
			goto use_random_mac_addr;

		/* found a vpd field */
		len = readb(base + i + 1) | (readb(base + i + 2) << 8);

		/* extract keywords */
		kstart = base + i + 3;
		p = kstart;
		while ((p - kstart) < len) {
			int klen = readb(p + 2);
			int j;
			char type;

			p += 3;

			/* look for the following things:
			 * -- correct length == 29
			 * 3 (type) + 2 (size) +
			 * 18 (strlen("local-mac-address") + 1) +
			 * 6 (mac addr)
			 * -- VPD Instance 'I'
			 * -- VPD Type Bytes 'B'
			 * -- VPD data length == 6
			 * -- property string == local-mac-address
			 *
			 * -- correct length == 24
			 * 3 (type) + 2 (size) +
			 * 12 (strlen("entropy-dev") + 1) +
			 * 7 (strlen("vms110") + 1)
			 * -- VPD Instance 'I'
			 * -- VPD Type String 'B'
			 * -- VPD data length == 7
			 * -- property string == entropy-dev
			 *
			 * -- correct length == 18
			 * 3 (type) + 2 (size) +
			 * 9 (strlen("phy-type") + 1) +
			 * 4 (strlen("pcs") + 1)
			 * -- VPD Instance 'I'
			 * -- VPD Type String 'S'
			 * -- VPD data length == 4
			 * -- property string == phy-type
			 *
			 * -- correct length == 23
			 * 3 (type) + 2 (size) +
			 * 14 (strlen("phy-interface") + 1) +
			 * 4 (strlen("pcs") + 1)
			 * -- VPD Instance 'I'
			 * -- VPD Type String 'S'
			 * -- VPD data length == 4
			 * -- property string == phy-interface
			 */
			if (readb(p) != 'I')
				goto next;

			/* finally, check string and length */
			type = readb(p + 3);
			if (type == 'B') {
				if ((klen == 29) && readb(p + 4) == 6 &&
				    cas_vpd_match(p + 5,
						  "local-mac-address")) {
					if (mac_off++ > offset)
						goto next;

					/* set mac address */
					for (j = 0; j < 6; j++)
						dev_addr[j] =
							readb(p + 23 + j);
					goto found_mac;
				}
			}

			if (type != 'S')
				goto next;

#ifdef USE_ENTROPY_DEV
			if ((klen == 24) &&
			    cas_vpd_match(p + 5, "entropy-dev") &&
			    cas_vpd_match(p + 17, "vms110")) {
				cp->cas_flags |= CAS_FLAG_ENTROPY_DEV;
				goto next;
			}
#endif

			if (found & VPD_FOUND_PHY)
				goto next;

			if ((klen == 18) && readb(p + 4) == 4 &&
			    cas_vpd_match(p + 5, "phy-type")) {
				if (cas_vpd_match(p + 14, "pcs")) {
					phy_type = CAS_PHY_SERDES;
					goto found_phy;
				}
			}

			if ((klen == 23) && readb(p + 4) == 4 &&
			    cas_vpd_match(p + 5, "phy-interface")) {
				if (cas_vpd_match(p + 19, "pcs")) {
					phy_type = CAS_PHY_SERDES;
					goto found_phy;
				}
			}
found_mac:
			found |= VPD_FOUND_MAC;
			goto next;

found_phy:
			found |= VPD_FOUND_PHY;

next:
			p += klen;
		}
		i += len + 3;
	}

use_random_mac_addr:
	if (found & VPD_FOUND_MAC)
		goto done;

	/* Sun MAC prefix then 3 random bytes. */
	printk(PFX "MAC address not found in ROM VPD\n");
	dev_addr[0] = 0x08;
	dev_addr[1] = 0x00;
	dev_addr[2] = 0x20;
	get_random_bytes(dev_addr + 3, 3);

done:
	writel(0, cp->regs + REG_BIM_LOCAL_DEV_EN);
	return phy_type;
}

/* check pci invariants */
static void cas_check_pci_invariants(struct cas *cp)
{
	struct pci_dev *pdev = cp->pdev;

	cp->cas_flags = 0;
	if ((pdev->vendor == PCI_VENDOR_ID_SUN) &&
	    (pdev->device == PCI_DEVICE_ID_SUN_CASSINI)) {
		if (pdev->revision >= CAS_ID_REVPLUS)
			cp->cas_flags |= CAS_FLAG_REG_PLUS;
		if (pdev->revision < CAS_ID_REVPLUS02u)
			cp->cas_flags |= CAS_FLAG_TARGET_ABORT;

		/* Original Cassini supports HW CSUM, but it's not
		 * enabled by default as it can trigger TX hangs.
		 */
		if (pdev->revision < CAS_ID_REV2)
			cp->cas_flags |= CAS_FLAG_NO_HW_CSUM;
	} else {
		/* Only sun has original cassini chips.  */
		cp->cas_flags |= CAS_FLAG_REG_PLUS;

		/* We use a flag because the same phy might be externally
		 * connected.
		 */
		if ((pdev->vendor == PCI_VENDOR_ID_NS) &&
		    (pdev->device == PCI_DEVICE_ID_NS_SATURN))
			cp->cas_flags |= CAS_FLAG_SATURN;
	}
}


static int cas_check_invariants(struct cas *cp)
{
	struct pci_dev *pdev = cp->pdev;
	u32 cfg;
	int i;

	/* get page size for rx buffers. */
	cp->page_order = 0;
#ifdef USE_PAGE_ORDER
	if (PAGE_SHIFT < CAS_JUMBO_PAGE_SHIFT) {
		/* see if we can allocate larger pages */
		struct page *page = alloc_pages(GFP_ATOMIC,
						CAS_JUMBO_PAGE_SHIFT -
						PAGE_SHIFT);
		if (page) {
			__free_pages(page, CAS_JUMBO_PAGE_SHIFT - PAGE_SHIFT);
			cp->page_order = CAS_JUMBO_PAGE_SHIFT - PAGE_SHIFT;
		} else {
			printk(PFX "MTU limited to %d bytes\n", CAS_MAX_MTU);
		}
	}
#endif
	cp->page_size = (PAGE_SIZE << cp->page_order);

	/* Fetch the FIFO configurations. */
	cp->tx_fifo_size = readl(cp->regs + REG_TX_FIFO_SIZE) * 64;
	cp->rx_fifo_size = RX_FIFO_SIZE;

	/* finish phy determination. MDIO1 takes precedence over MDIO0 if
	 * they're both connected.
	 */
	cp->phy_type = cas_get_vpd_info(cp, cp->dev->dev_addr,
					PCI_SLOT(pdev->devfn));
	if (cp->phy_type & CAS_PHY_SERDES) {
		cp->cas_flags |= CAS_FLAG_1000MB_CAP;
		return 0; /* no more checking needed */
	}

	/* MII */
	cfg = readl(cp->regs + REG_MIF_CFG);
	if (cfg & MIF_CFG_MDIO_1) {
		cp->phy_type = CAS_PHY_MII_MDIO1;
	} else if (cfg & MIF_CFG_MDIO_0) {
		cp->phy_type = CAS_PHY_MII_MDIO0;
	}

	cas_mif_poll(cp, 0);
	writel(PCS_DATAPATH_MODE_MII, cp->regs + REG_PCS_DATAPATH_MODE);

	for (i = 0; i < 32; i++) {
		u32 phy_id;
		int j;

		for (j = 0; j < 3; j++) {
			cp->phy_addr = i;
			phy_id = cas_phy_read(cp, MII_PHYSID1) << 16;
			phy_id |= cas_phy_read(cp, MII_PHYSID2);
			if (phy_id && (phy_id != 0xFFFFFFFF)) {
				cp->phy_id = phy_id;
				goto done;
			}
		}
	}
	printk(KERN_ERR PFX "MII phy did not respond [%08x]\n",
	       readl(cp->regs + REG_MIF_STATE_MACHINE));
	return -1;

done:
	/* see if we can do gigabit */
	cfg = cas_phy_read(cp, MII_BMSR);
	if ((cfg & CAS_BMSR_1000_EXTEND) &&
	    cas_phy_read(cp, CAS_MII_1000_EXTEND))
		cp->cas_flags |= CAS_FLAG_1000MB_CAP;
	return 0;
}

/* Must be invoked under cp->lock. */
static inline void cas_start_dma(struct cas *cp)
{
	int i;
	u32 val;
	int txfailed = 0;

	/* enable dma */
	val = readl(cp->regs + REG_TX_CFG) | TX_CFG_DMA_EN;
	writel(val, cp->regs + REG_TX_CFG);
	val = readl(cp->regs + REG_RX_CFG) | RX_CFG_DMA_EN;
	writel(val, cp->regs + REG_RX_CFG);

	/* enable the mac */
	val = readl(cp->regs + REG_MAC_TX_CFG) | MAC_TX_CFG_EN;
	writel(val, cp->regs + REG_MAC_TX_CFG);
	val = readl(cp->regs + REG_MAC_RX_CFG) | MAC_RX_CFG_EN;
	writel(val, cp->regs + REG_MAC_RX_CFG);

	i = STOP_TRIES;
	while (i-- > 0) {
		val = readl(cp->regs + REG_MAC_TX_CFG);
		if ((val & MAC_TX_CFG_EN))
			break;
		udelay(10);
	}
	if (i < 0) txfailed = 1;
	i = STOP_TRIES;
	while (i-- > 0) {
		val = readl(cp->regs + REG_MAC_RX_CFG);
		if ((val & MAC_RX_CFG_EN)) {
			if (txfailed) {
			  printk(KERN_ERR
				 "%s: enabling mac failed [tx:%08x:%08x].\n",
				 cp->dev->name,
				 readl(cp->regs + REG_MIF_STATE_MACHINE),
				 readl(cp->regs + REG_MAC_STATE_MACHINE));
			}
			goto enable_rx_done;
		}
		udelay(10);
	}
	printk(KERN_ERR "%s: enabling mac failed [%s:%08x:%08x].\n",
	       cp->dev->name,
	       (txfailed? "tx,rx":"rx"),
	       readl(cp->regs + REG_MIF_STATE_MACHINE),
	       readl(cp->regs + REG_MAC_STATE_MACHINE));

enable_rx_done:
	cas_unmask_intr(cp); /* enable interrupts */
	writel(RX_DESC_RINGN_SIZE(0) - 4, cp->regs + REG_RX_KICK);
	writel(0, cp->regs + REG_RX_COMP_TAIL);

	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
		if (N_RX_DESC_RINGS > 1)
			writel(RX_DESC_RINGN_SIZE(1) - 4,
			       cp->regs + REG_PLUS_RX_KICK1);

		for (i = 1; i < N_RX_COMP_RINGS; i++)
			writel(0, cp->regs + REG_PLUS_RX_COMPN_TAIL(i));
	}
}

/* Must be invoked under cp->lock. */
static void cas_read_pcs_link_mode(struct cas *cp, int *fd, int *spd,
				   int *pause)
{
	u32 val = readl(cp->regs + REG_PCS_MII_LPA);
	*fd     = (val & PCS_MII_LPA_FD) ? 1 : 0;
	*pause  = (val & PCS_MII_LPA_SYM_PAUSE) ? 0x01 : 0x00;
	if (val & PCS_MII_LPA_ASYM_PAUSE)
		*pause |= 0x10;
	*spd = 1000;
}

/* Must be invoked under cp->lock. */
static void cas_read_mii_link_mode(struct cas *cp, int *fd, int *spd,
				   int *pause)
{
	u32 val;

	*fd = 0;
	*spd = 10;
	*pause = 0;

	/* use GMII registers */
	val = cas_phy_read(cp, MII_LPA);
	if (val & CAS_LPA_PAUSE)
		*pause = 0x01;

	if (val & CAS_LPA_ASYM_PAUSE)
		*pause |= 0x10;

	if (val & LPA_DUPLEX)
		*fd = 1;
	if (val & LPA_100)
		*spd = 100;

	if (cp->cas_flags & CAS_FLAG_1000MB_CAP) {
		val = cas_phy_read(cp, CAS_MII_1000_STATUS);
		if (val & (CAS_LPA_1000FULL | CAS_LPA_1000HALF))
			*spd = 1000;
		if (val & CAS_LPA_1000FULL)
			*fd = 1;
	}
}

/* A link-up condition has occurred, initialize and enable the
 * rest of the chip.
 *
 * Must be invoked under cp->lock.
 */
static void cas_set_link_modes(struct cas *cp)
{
	u32 val;
	int full_duplex, speed, pause;

	full_duplex = 0;
	speed = 10;
	pause = 0;

	if (CAS_PHY_MII(cp->phy_type)) {
		cas_mif_poll(cp, 0);
		val = cas_phy_read(cp, MII_BMCR);
		if (val & BMCR_ANENABLE) {
			cas_read_mii_link_mode(cp, &full_duplex, &speed,
					       &pause);
		} else {
			if (val & BMCR_FULLDPLX)
				full_duplex = 1;

			if (val & BMCR_SPEED100)
				speed = 100;
			else if (val & CAS_BMCR_SPEED1000)
				speed = (cp->cas_flags & CAS_FLAG_1000MB_CAP) ?
					1000 : 100;
		}
		cas_mif_poll(cp, 1);

	} else {
		val = readl(cp->regs + REG_PCS_MII_CTRL);
		cas_read_pcs_link_mode(cp, &full_duplex, &speed, &pause);
		if ((val & PCS_MII_AUTONEG_EN) == 0) {
			if (val & PCS_MII_CTRL_DUPLEX)
				full_duplex = 1;
		}
	}

	if (netif_msg_link(cp))
		printk(KERN_INFO "%s: Link up at %d Mbps, %s-duplex.\n",
		       cp->dev->name, speed, (full_duplex ? "full" : "half"));

	val = MAC_XIF_TX_MII_OUTPUT_EN | MAC_XIF_LINK_LED;
	if (CAS_PHY_MII(cp->phy_type)) {
		val |= MAC_XIF_MII_BUFFER_OUTPUT_EN;
		if (!full_duplex)
			val |= MAC_XIF_DISABLE_ECHO;
	}
	if (full_duplex)
		val |= MAC_XIF_FDPLX_LED;
	if (speed == 1000)
		val |= MAC_XIF_GMII_MODE;
	writel(val, cp->regs + REG_MAC_XIF_CFG);

	/* deal with carrier and collision detect. */
	val = MAC_TX_CFG_IPG_EN;
	if (full_duplex) {
		val |= MAC_TX_CFG_IGNORE_CARRIER;
		val |= MAC_TX_CFG_IGNORE_COLL;
	} else {
#ifndef USE_CSMA_CD_PROTO
		val |= MAC_TX_CFG_NEVER_GIVE_UP_EN;
		val |= MAC_TX_CFG_NEVER_GIVE_UP_LIM;
#endif
	}
	/* val now set up for REG_MAC_TX_CFG */

	/* If gigabit and half-duplex, enable carrier extension
	 * mode.  increase slot time to 512 bytes as well.
	 * else, disable it and make sure slot time is 64 bytes.
	 * also activate checksum bug workaround
	 */
	if ((speed == 1000) && !full_duplex) {
		writel(val | MAC_TX_CFG_CARRIER_EXTEND,
		       cp->regs + REG_MAC_TX_CFG);

		val = readl(cp->regs + REG_MAC_RX_CFG);
		val &= ~MAC_RX_CFG_STRIP_FCS; /* checksum workaround */
		writel(val | MAC_RX_CFG_CARRIER_EXTEND,
		       cp->regs + REG_MAC_RX_CFG);

		writel(0x200, cp->regs + REG_MAC_SLOT_TIME);

		cp->crc_size = 4;
		/* minimum size gigabit frame at half duplex */
		cp->min_frame_size = CAS_1000MB_MIN_FRAME;

	} else {
		writel(val, cp->regs + REG_MAC_TX_CFG);

		/* checksum bug workaround. don't strip FCS when in
		 * half-duplex mode
		 */
		val = readl(cp->regs + REG_MAC_RX_CFG);
		if (full_duplex) {
			val |= MAC_RX_CFG_STRIP_FCS;
			cp->crc_size = 0;
			cp->min_frame_size = CAS_MIN_MTU;
		} else {
			val &= ~MAC_RX_CFG_STRIP_FCS;
			cp->crc_size = 4;
			cp->min_frame_size = CAS_MIN_FRAME;
		}
		writel(val & ~MAC_RX_CFG_CARRIER_EXTEND,
		       cp->regs + REG_MAC_RX_CFG);
		writel(0x40, cp->regs + REG_MAC_SLOT_TIME);
	}

	if (netif_msg_link(cp)) {
		if (pause & 0x01) {
			printk(KERN_INFO "%s: Pause is enabled "
			       "(rxfifo: %d off: %d on: %d)\n",
			       cp->dev->name,
			       cp->rx_fifo_size,
			       cp->rx_pause_off,
			       cp->rx_pause_on);
		} else if (pause & 0x10) {
			printk(KERN_INFO "%s: TX pause enabled\n",
			       cp->dev->name);
		} else {
			printk(KERN_INFO "%s: Pause is disabled\n",
			       cp->dev->name);
		}
	}

	val = readl(cp->regs + REG_MAC_CTRL_CFG);
	val &= ~(MAC_CTRL_CFG_SEND_PAUSE_EN | MAC_CTRL_CFG_RECV_PAUSE_EN);
	if (pause) { /* symmetric or asymmetric pause */
		val |= MAC_CTRL_CFG_SEND_PAUSE_EN;
		if (pause & 0x01) { /* symmetric pause */
			val |= MAC_CTRL_CFG_RECV_PAUSE_EN;
		}
	}
	writel(val, cp->regs + REG_MAC_CTRL_CFG);
	cas_start_dma(cp);
}

/* Must be invoked under cp->lock. */
static void cas_init_hw(struct cas *cp, int restart_link)
{
	if (restart_link)
		cas_phy_init(cp);

	cas_init_pause_thresholds(cp);
	cas_init_mac(cp);
	cas_init_dma(cp);

	if (restart_link) {
		/* Default aneg parameters */
		cp->timer_ticks = 0;
		cas_begin_auto_negotiation(cp, NULL);
	} else if (cp->lstate == link_up) {
		cas_set_link_modes(cp);
		netif_carrier_on(cp->dev);
	}
}

/* Must be invoked under cp->lock. on earlier cassini boards,
 * SOFT_0 is tied to PCI reset. we use this to force a pci reset,
 * let it settle out, and then restore pci state.
 */
static void cas_hard_reset(struct cas *cp)
{
	writel(BIM_LOCAL_DEV_SOFT_0, cp->regs + REG_BIM_LOCAL_DEV_EN);
	udelay(20);
	pci_restore_state(cp->pdev);
}


static void cas_global_reset(struct cas *cp, int blkflag)
{
	int limit;

	/* issue a global reset. don't use RSTOUT. */
	if (blkflag && !CAS_PHY_MII(cp->phy_type)) {
		/* For PCS, when the blkflag is set, we should set the
		 * SW_REST_BLOCK_PCS_SLINK bit to prevent the results of
		 * the last autonegotiation from being cleared.  We'll
		 * need some special handling if the chip is set into a
		 * loopback mode.
		 */
		writel((SW_RESET_TX | SW_RESET_RX | SW_RESET_BLOCK_PCS_SLINK),
		       cp->regs + REG_SW_RESET);
	} else {
		writel(SW_RESET_TX | SW_RESET_RX, cp->regs + REG_SW_RESET);
	}

	/* need to wait at least 3ms before polling register */
	mdelay(3);

	limit = STOP_TRIES;
	while (limit-- > 0) {
		u32 val = readl(cp->regs + REG_SW_RESET);
		if ((val & (SW_RESET_TX | SW_RESET_RX)) == 0)
			goto done;
		udelay(10);
	}
	printk(KERN_ERR "%s: sw reset failed.\n", cp->dev->name);

done:
	/* enable various BIM interrupts */
	writel(BIM_CFG_DPAR_INTR_ENABLE | BIM_CFG_RMA_INTR_ENABLE |
	       BIM_CFG_RTA_INTR_ENABLE, cp->regs + REG_BIM_CFG);

	/* clear out pci error status mask for handled errors.
	 * we don't deal with DMA counter overflows as they happen
	 * all the time.
	 */
	writel(0xFFFFFFFFU & ~(PCI_ERR_BADACK | PCI_ERR_DTRTO |
			       PCI_ERR_OTHER | PCI_ERR_BIM_DMA_WRITE |
			       PCI_ERR_BIM_DMA_READ), cp->regs +
	       REG_PCI_ERR_STATUS_MASK);

	/* set up for MII by default to address mac rx reset timeout
	 * issue
	 */
	writel(PCS_DATAPATH_MODE_MII, cp->regs + REG_PCS_DATAPATH_MODE);
}

static void cas_reset(struct cas *cp, int blkflag)
{
	u32 val;

	cas_mask_intr(cp);
	cas_global_reset(cp, blkflag);
	cas_mac_reset(cp);
	cas_entropy_reset(cp);

	/* disable dma engines. */
	val = readl(cp->regs + REG_TX_CFG);
	val &= ~TX_CFG_DMA_EN;
	writel(val, cp->regs + REG_TX_CFG);

	val = readl(cp->regs + REG_RX_CFG);
	val &= ~RX_CFG_DMA_EN;
	writel(val, cp->regs + REG_RX_CFG);

	/* program header parser */
	if ((cp->cas_flags & CAS_FLAG_TARGET_ABORT) ||
	    (CAS_HP_ALT_FIRMWARE == cas_prog_null)) {
		cas_load_firmware(cp, CAS_HP_FIRMWARE);
	} else {
		cas_load_firmware(cp, CAS_HP_ALT_FIRMWARE);
	}

	/* clear out error registers */
	spin_lock(&cp->stat_lock[N_TX_RINGS]);
	cas_clear_mac_err(cp);
	spin_unlock(&cp->stat_lock[N_TX_RINGS]);
}

/* Shut down the chip, must be called with pm_mutex held.  */
static void cas_shutdown(struct cas *cp)
{
	unsigned long flags;

	/* Make us not-running to avoid timers respawning */
	cp->hw_running = 0;

	del_timer_sync(&cp->link_timer);

	/* Stop the reset task */
#if 0
	while (atomic_read(&cp->reset_task_pending_mtu) ||
	       atomic_read(&cp->reset_task_pending_spare) ||
	       atomic_read(&cp->reset_task_pending_all))
		schedule();

#else
	while (atomic_read(&cp->reset_task_pending))
		schedule();
#endif
	/* Actually stop the chip */
	cas_lock_all_save(cp, flags);
	cas_reset(cp, 0);
	if (cp->cas_flags & CAS_FLAG_SATURN)
		cas_phy_powerdown(cp);
	cas_unlock_all_restore(cp, flags);
}

static int cas_change_mtu(struct net_device *dev, int new_mtu)
{
	struct cas *cp = netdev_priv(dev);

	if (new_mtu < CAS_MIN_MTU || new_mtu > CAS_MAX_MTU)
		return -EINVAL;

	dev->mtu = new_mtu;
	if (!netif_running(dev) || !netif_device_present(dev))
		return 0;

	/* let the reset task handle it */
#if 1
	atomic_inc(&cp->reset_task_pending);
	if ((cp->phy_type & CAS_PHY_SERDES)) {
		atomic_inc(&cp->reset_task_pending_all);
	} else {
		atomic_inc(&cp->reset_task_pending_mtu);
	}
	schedule_work(&cp->reset_task);
#else
	atomic_set(&cp->reset_task_pending, (cp->phy_type & CAS_PHY_SERDES) ?
		   CAS_RESET_ALL : CAS_RESET_MTU);
	printk(KERN_ERR "reset called in cas_change_mtu\n");
	schedule_work(&cp->reset_task);
#endif

	flush_scheduled_work();
	return 0;
}

static void cas_clean_txd(struct cas *cp, int ring)
{
	struct cas_tx_desc *txd = cp->init_txds[ring];
	struct sk_buff *skb, **skbs = cp->tx_skbs[ring];
	u64 daddr, dlen;
	int i, size;

	size = TX_DESC_RINGN_SIZE(ring);
	for (i = 0; i < size; i++) {
		int frag;

		if (skbs[i] == NULL)
			continue;

		skb = skbs[i];
		skbs[i] = NULL;

		for (frag = 0; frag <= skb_shinfo(skb)->nr_frags;  frag++) {
			int ent = i & (size - 1);

			/* first buffer is never a tiny buffer and so
			 * needs to be unmapped.
			 */
			daddr = le64_to_cpu(txd[ent].buffer);
			dlen  =  CAS_VAL(TX_DESC_BUFLEN,
					 le64_to_cpu(txd[ent].control));
			pci_unmap_page(cp->pdev, daddr, dlen,
				       PCI_DMA_TODEVICE);

			if (frag != skb_shinfo(skb)->nr_frags) {
				i++;

				/* next buffer might by a tiny buffer.
				 * skip past it.
				 */
				ent = i & (size - 1);
				if (cp->tx_tiny_use[ring][ent].used)
					i++;
			}
		}
		dev_kfree_skb_any(skb);
	}

	/* zero out tiny buf usage */
	memset(cp->tx_tiny_use[ring], 0, size*sizeof(*cp->tx_tiny_use[ring]));
}

/* freed on close */
static inline void cas_free_rx_desc(struct cas *cp, int ring)
{
	cas_page_t **page = cp->rx_pages[ring];
	int i, size;

	size = RX_DESC_RINGN_SIZE(ring);
	for (i = 0; i < size; i++) {
		if (page[i]) {
			cas_page_free(cp, page[i]);
			page[i] = NULL;
		}
	}
}

static void cas_free_rxds(struct cas *cp)
{
	int i;

	for (i = 0; i < N_RX_DESC_RINGS; i++)
		cas_free_rx_desc(cp, i);
}

/* Must be invoked under cp->lock. */
static void cas_clean_rings(struct cas *cp)
{
	int i;

	/* need to clean all tx rings */
	memset(cp->tx_old, 0, sizeof(*cp->tx_old)*N_TX_RINGS);
	memset(cp->tx_new, 0, sizeof(*cp->tx_new)*N_TX_RINGS);
	for (i = 0; i < N_TX_RINGS; i++)
		cas_clean_txd(cp, i);

	/* zero out init block */
	memset(cp->init_block, 0, sizeof(struct cas_init_block));
	cas_clean_rxds(cp);
	cas_clean_rxcs(cp);
}

/* allocated on open */
static inline int cas_alloc_rx_desc(struct cas *cp, int ring)
{
	cas_page_t **page = cp->rx_pages[ring];
	int size, i = 0;

	size = RX_DESC_RINGN_SIZE(ring);
	for (i = 0; i < size; i++) {
		if ((page[i] = cas_page_alloc(cp, GFP_KERNEL)) == NULL)
			return -1;
	}
	return 0;
}

static int cas_alloc_rxds(struct cas *cp)
{
	int i;

	for (i = 0; i < N_RX_DESC_RINGS; i++) {
		if (cas_alloc_rx_desc(cp, i) < 0) {
			cas_free_rxds(cp);
			return -1;
		}
	}
	return 0;
}

static void cas_reset_task(struct work_struct *work)
{
	struct cas *cp = container_of(work, struct cas, reset_task);
#if 0
	int pending = atomic_read(&cp->reset_task_pending);
#else
	int pending_all = atomic_read(&cp->reset_task_pending_all);
	int pending_spare = atomic_read(&cp->reset_task_pending_spare);
	int pending_mtu = atomic_read(&cp->reset_task_pending_mtu);

	if (pending_all == 0 && pending_spare == 0 && pending_mtu == 0) {
		/* We can have more tasks scheduled than actually
		 * needed.
		 */
		atomic_dec(&cp->reset_task_pending);
		return;
	}
#endif
	/* The link went down, we reset the ring, but keep
	 * DMA stopped. Use this function for reset
	 * on error as well.
	 */
	if (cp->hw_running) {
		unsigned long flags;

		/* Make sure we don't get interrupts or tx packets */
		netif_device_detach(cp->dev);
		cas_lock_all_save(cp, flags);

		if (cp->opened) {
			/* We call cas_spare_recover when we call cas_open.
			 * but we do not initialize the lists cas_spare_recover
			 * uses until cas_open is called.
			 */
			cas_spare_recover(cp, GFP_ATOMIC);
		}
#if 1
		/* test => only pending_spare set */
		if (!pending_all && !pending_mtu)
			goto done;
#else
		if (pending == CAS_RESET_SPARE)
			goto done;
#endif
		/* when pending == CAS_RESET_ALL, the following
		 * call to cas_init_hw will restart auto negotiation.
		 * Setting the second argument of cas_reset to
		 * !(pending == CAS_RESET_ALL) will set this argument
		 * to 1 (avoiding reinitializing the PHY for the normal
		 * PCS case) when auto negotiation is not restarted.
		 */
#if 1
		cas_reset(cp, !(pending_all > 0));
		if (cp->opened)
			cas_clean_rings(cp);
		cas_init_hw(cp, (pending_all > 0));
#else
		cas_reset(cp, !(pending == CAS_RESET_ALL));
		if (cp->opened)
			cas_clean_rings(cp);
		cas_init_hw(cp, pending == CAS_RESET_ALL);
#endif

done:
		cas_unlock_all_restore(cp, flags);
		netif_device_attach(cp->dev);
	}
#if 1
	atomic_sub(pending_all, &cp->reset_task_pending_all);
	atomic_sub(pending_spare, &cp->reset_task_pending_spare);
	atomic_sub(pending_mtu, &cp->reset_task_pending_mtu);
	atomic_dec(&cp->reset_task_pending);
#else
	atomic_set(&cp->reset_task_pending, 0);
#endif
}

static void cas_link_timer(unsigned long data)
{
	struct cas *cp = (struct cas *) data;
	int mask, pending = 0, reset = 0;
	unsigned long flags;

	if (link_transition_timeout != 0 &&
	    cp->link_transition_jiffies_valid &&
	    ((jiffies - cp->link_transition_jiffies) >
	      (link_transition_timeout))) {
		/* One-second counter so link-down workaround doesn't
		 * cause resets to occur so fast as to fool the switch
		 * into thinking the link is down.
		 */
		cp->link_transition_jiffies_valid = 0;
	}

	if (!cp->hw_running)
		return;

	spin_lock_irqsave(&cp->lock, flags);
	cas_lock_tx(cp);
	cas_entropy_gather(cp);

	/* If the link task is still pending, we just
	 * reschedule the link timer
	 */
#if 1
	if (atomic_read(&cp->reset_task_pending_all) ||
	    atomic_read(&cp->reset_task_pending_spare) ||
	    atomic_read(&cp->reset_task_pending_mtu))
		goto done;
#else
	if (atomic_read(&cp->reset_task_pending))
		goto done;
#endif

	/* check for rx cleaning */
	if ((mask = (cp->cas_flags & CAS_FLAG_RXD_POST_MASK))) {
		int i, rmask;

		for (i = 0; i < MAX_RX_DESC_RINGS; i++) {
			rmask = CAS_FLAG_RXD_POST(i);
			if ((mask & rmask) == 0)
				continue;

			/* post_rxds will do a mod_timer */
			if (cas_post_rxds_ringN(cp, i, cp->rx_last[i]) < 0) {
				pending = 1;
				continue;
			}
			cp->cas_flags &= ~rmask;
		}
	}

	if (CAS_PHY_MII(cp->phy_type)) {
		u16 bmsr;
		cas_mif_poll(cp, 0);
		bmsr = cas_phy_read(cp, MII_BMSR);
		/* WTZ: Solaris driver reads this twice, but that
		 * may be due to the PCS case and the use of a
		 * common implementation. Read it twice here to be
		 * safe.
		 */
		bmsr = cas_phy_read(cp, MII_BMSR);
		cas_mif_poll(cp, 1);
		readl(cp->regs + REG_MIF_STATUS); /* avoid dups */
		reset = cas_mii_link_check(cp, bmsr);
	} else {
		reset = cas_pcs_link_check(cp);
	}

	if (reset)
		goto done;

	/* check for tx state machine confusion */
	if ((readl(cp->regs + REG_MAC_TX_STATUS) & MAC_TX_FRAME_XMIT) == 0) {
		u32 val = readl(cp->regs + REG_MAC_STATE_MACHINE);
		u32 wptr, rptr;
		int tlm  = CAS_VAL(MAC_SM_TLM, val);

		if (((tlm == 0x5) || (tlm == 0x3)) &&
		    (CAS_VAL(MAC_SM_ENCAP_SM, val) == 0)) {
			if (netif_msg_tx_err(cp))
				printk(KERN_DEBUG "%s: tx err: "
				       "MAC_STATE[%08x]\n",
				       cp->dev->name, val);
			reset = 1;
			goto done;
		}

		val  = readl(cp->regs + REG_TX_FIFO_PKT_CNT);
		wptr = readl(cp->regs + REG_TX_FIFO_WRITE_PTR);
		rptr = readl(cp->regs + REG_TX_FIFO_READ_PTR);
		if ((val == 0) && (wptr != rptr)) {
			if (netif_msg_tx_err(cp))
				printk(KERN_DEBUG "%s: tx err: "
				       "TX_FIFO[%08x:%08x:%08x]\n",
				       cp->dev->name, val, wptr, rptr);
			reset = 1;
		}

		if (reset)
			cas_hard_reset(cp);
	}

done:
	if (reset) {
#if 1
		atomic_inc(&cp->reset_task_pending);
		atomic_inc(&cp->reset_task_pending_all);
		schedule_work(&cp->reset_task);
#else
		atomic_set(&cp->reset_task_pending, CAS_RESET_ALL);
		printk(KERN_ERR "reset called in cas_link_timer\n");
		schedule_work(&cp->reset_task);
#endif
	}

	if (!pending)
		mod_timer(&cp->link_timer, jiffies + CAS_LINK_TIMEOUT);
	cas_unlock_tx(cp);
	spin_unlock_irqrestore(&cp->lock, flags);
}

/* tiny buffers are used to avoid target abort issues with
 * older cassini's
 */
static void cas_tx_tiny_free(struct cas *cp)
{
	struct pci_dev *pdev = cp->pdev;
	int i;

	for (i = 0; i < N_TX_RINGS; i++) {
		if (!cp->tx_tiny_bufs[i])
			continue;

		pci_free_consistent(pdev, TX_TINY_BUF_BLOCK,
				    cp->tx_tiny_bufs[i],
				    cp->tx_tiny_dvma[i]);
		cp->tx_tiny_bufs[i] = NULL;
	}
}

static int cas_tx_tiny_alloc(struct cas *cp)
{
	struct pci_dev *pdev = cp->pdev;
	int i;

	for (i = 0; i < N_TX_RINGS; i++) {
		cp->tx_tiny_bufs[i] =
			pci_alloc_consistent(pdev, TX_TINY_BUF_BLOCK,
					     &cp->tx_tiny_dvma[i]);
		if (!cp->tx_tiny_bufs[i]) {
			cas_tx_tiny_free(cp);
			return -1;
		}
	}
	return 0;
}


static int cas_open(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);
	int hw_was_up, err;
	unsigned long flags;

	mutex_lock(&cp->pm_mutex);

	hw_was_up = cp->hw_running;

	/* The power-management mutex protects the hw_running
	 * etc. state so it is safe to do this bit without cp->lock
	 */
	if (!cp->hw_running) {
		/* Reset the chip */
		cas_lock_all_save(cp, flags);
		/* We set the second arg to cas_reset to zero
		 * because cas_init_hw below will have its second
		 * argument set to non-zero, which will force
		 * autonegotiation to start.
		 */
		cas_reset(cp, 0);
		cp->hw_running = 1;
		cas_unlock_all_restore(cp, flags);
	}

	if (cas_tx_tiny_alloc(cp) < 0)
		return -ENOMEM;

	/* alloc rx descriptors */
	err = -ENOMEM;
	if (cas_alloc_rxds(cp) < 0)
		goto err_tx_tiny;

	/* allocate spares */
	cas_spare_init(cp);
	cas_spare_recover(cp, GFP_KERNEL);

	/* We can now request the interrupt as we know it's masked
	 * on the controller. cassini+ has up to 4 interrupts
	 * that can be used, but you need to do explicit pci interrupt
	 * mapping to expose them
	 */
	if (request_irq(cp->pdev->irq, cas_interrupt,
			IRQF_SHARED, dev->name, (void *) dev)) {
		printk(KERN_ERR "%s: failed to request irq !\n",
		       cp->dev->name);
		err = -EAGAIN;
		goto err_spare;
	}

#ifdef USE_NAPI
	napi_enable(&cp->napi);
#endif
	/* init hw */
	cas_lock_all_save(cp, flags);
	cas_clean_rings(cp);
	cas_init_hw(cp, !hw_was_up);
	cp->opened = 1;
	cas_unlock_all_restore(cp, flags);

	netif_start_queue(dev);
	mutex_unlock(&cp->pm_mutex);
	return 0;

err_spare:
	cas_spare_free(cp);
	cas_free_rxds(cp);
err_tx_tiny:
	cas_tx_tiny_free(cp);
	mutex_unlock(&cp->pm_mutex);
	return err;
}

static int cas_close(struct net_device *dev)
{
	unsigned long flags;
	struct cas *cp = netdev_priv(dev);

#ifdef USE_NAPI
	napi_disable(&cp->napi);
#endif
	/* Make sure we don't get distracted by suspend/resume */
	mutex_lock(&cp->pm_mutex);

	netif_stop_queue(dev);

	/* Stop traffic, mark us closed */
	cas_lock_all_save(cp, flags);
	cp->opened = 0;
	cas_reset(cp, 0);
	cas_phy_init(cp);
	cas_begin_auto_negotiation(cp, NULL);
	cas_clean_rings(cp);
	cas_unlock_all_restore(cp, flags);

	free_irq(cp->pdev->irq, (void *) dev);
	cas_spare_free(cp);
	cas_free_rxds(cp);
	cas_tx_tiny_free(cp);
	mutex_unlock(&cp->pm_mutex);
	return 0;
}

static struct {
	const char name[ETH_GSTRING_LEN];
} ethtool_cassini_statnames[] = {
	{"collisions"},
	{"rx_bytes"},
	{"rx_crc_errors"},
	{"rx_dropped"},
	{"rx_errors"},
	{"rx_fifo_errors"},
	{"rx_frame_errors"},
	{"rx_length_errors"},
	{"rx_over_errors"},
	{"rx_packets"},
	{"tx_aborted_errors"},
	{"tx_bytes"},
	{"tx_dropped"},
	{"tx_errors"},
	{"tx_fifo_errors"},
	{"tx_packets"}
};
#define CAS_NUM_STAT_KEYS ARRAY_SIZE(ethtool_cassini_statnames)

static struct {
	const int offsets;	/* neg. values for 2nd arg to cas_read_phy */
} ethtool_register_table[] = {
	{-MII_BMSR},
	{-MII_BMCR},
	{REG_CAWR},
	{REG_INF_BURST},
	{REG_BIM_CFG},
	{REG_RX_CFG},
	{REG_HP_CFG},
	{REG_MAC_TX_CFG},
	{REG_MAC_RX_CFG},
	{REG_MAC_CTRL_CFG},
	{REG_MAC_XIF_CFG},
	{REG_MIF_CFG},
	{REG_PCS_CFG},
	{REG_SATURN_PCFG},
	{REG_PCS_MII_STATUS},
	{REG_PCS_STATE_MACHINE},
	{REG_MAC_COLL_EXCESS},
	{REG_MAC_COLL_LATE}
};
#define CAS_REG_LEN 	ARRAY_SIZE(ethtool_register_table)
#define CAS_MAX_REGS 	(sizeof (u32)*CAS_REG_LEN)

static void cas_read_regs(struct cas *cp, u8 *ptr, int len)
{
	u8 *p;
	int i;
	unsigned long flags;

	spin_lock_irqsave(&cp->lock, flags);
	for (i = 0, p = ptr; i < len ; i ++, p += sizeof(u32)) {
		u16 hval;
		u32 val;
		if (ethtool_register_table[i].offsets < 0) {
			hval = cas_phy_read(cp,
				    -ethtool_register_table[i].offsets);
			val = hval;
		} else {
			val= readl(cp->regs+ethtool_register_table[i].offsets);
		}
		memcpy(p, (u8 *)&val, sizeof(u32));
	}
	spin_unlock_irqrestore(&cp->lock, flags);
}

static struct net_device_stats *cas_get_stats(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);
	struct net_device_stats *stats = cp->net_stats;
	unsigned long flags;
	int i;
	unsigned long tmp;

	/* we collate all of the stats into net_stats[N_TX_RING] */
	if (!cp->hw_running)
		return stats + N_TX_RINGS;

	/* collect outstanding stats */
	/* WTZ: the Cassini spec gives these as 16 bit counters but
	 * stored in 32-bit words.  Added a mask of 0xffff to be safe,
	 * in case the chip somehow puts any garbage in the other bits.
	 * Also, counter usage didn't seem to mach what Adrian did
	 * in the parts of the code that set these quantities. Made
	 * that consistent.
	 */
	spin_lock_irqsave(&cp->stat_lock[N_TX_RINGS], flags);
	stats[N_TX_RINGS].rx_crc_errors +=
	  readl(cp->regs + REG_MAC_FCS_ERR) & 0xffff;
	stats[N_TX_RINGS].rx_frame_errors +=
		readl(cp->regs + REG_MAC_ALIGN_ERR) &0xffff;
	stats[N_TX_RINGS].rx_length_errors +=
		readl(cp->regs + REG_MAC_LEN_ERR) & 0xffff;
#if 1
	tmp = (readl(cp->regs + REG_MAC_COLL_EXCESS) & 0xffff) +
		(readl(cp->regs + REG_MAC_COLL_LATE) & 0xffff);
	stats[N_TX_RINGS].tx_aborted_errors += tmp;
	stats[N_TX_RINGS].collisions +=
	  tmp + (readl(cp->regs + REG_MAC_COLL_NORMAL) & 0xffff);
#else
	stats[N_TX_RINGS].tx_aborted_errors +=
		readl(cp->regs + REG_MAC_COLL_EXCESS);
	stats[N_TX_RINGS].collisions += readl(cp->regs + REG_MAC_COLL_EXCESS) +
		readl(cp->regs + REG_MAC_COLL_LATE);
#endif
	cas_clear_mac_err(cp);

	/* saved bits that are unique to ring 0 */
	spin_lock(&cp->stat_lock[0]);
	stats[N_TX_RINGS].collisions        += stats[0].collisions;
	stats[N_TX_RINGS].rx_over_errors    += stats[0].rx_over_errors;
	stats[N_TX_RINGS].rx_frame_errors   += stats[0].rx_frame_errors;
	stats[N_TX_RINGS].rx_fifo_errors    += stats[0].rx_fifo_errors;
	stats[N_TX_RINGS].tx_aborted_errors += stats[0].tx_aborted_errors;
	stats[N_TX_RINGS].tx_fifo_errors    += stats[0].tx_fifo_errors;
	spin_unlock(&cp->stat_lock[0]);

	for (i = 0; i < N_TX_RINGS; i++) {
		spin_lock(&cp->stat_lock[i]);
		stats[N_TX_RINGS].rx_length_errors +=
			stats[i].rx_length_errors;
		stats[N_TX_RINGS].rx_crc_errors += stats[i].rx_crc_errors;
		stats[N_TX_RINGS].rx_packets    += stats[i].rx_packets;
		stats[N_TX_RINGS].tx_packets    += stats[i].tx_packets;
		stats[N_TX_RINGS].rx_bytes      += stats[i].rx_bytes;
		stats[N_TX_RINGS].tx_bytes      += stats[i].tx_bytes;
		stats[N_TX_RINGS].rx_errors     += stats[i].rx_errors;
		stats[N_TX_RINGS].tx_errors     += stats[i].tx_errors;
		stats[N_TX_RINGS].rx_dropped    += stats[i].rx_dropped;
		stats[N_TX_RINGS].tx_dropped    += stats[i].tx_dropped;
		memset(stats + i, 0, sizeof(struct net_device_stats));
		spin_unlock(&cp->stat_lock[i]);
	}
	spin_unlock_irqrestore(&cp->stat_lock[N_TX_RINGS], flags);
	return stats + N_TX_RINGS;
}


static void cas_set_multicast(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);
	u32 rxcfg, rxcfg_new;
	unsigned long flags;
	int limit = STOP_TRIES;

	if (!cp->hw_running)
		return;

	spin_lock_irqsave(&cp->lock, flags);
	rxcfg = readl(cp->regs + REG_MAC_RX_CFG);

	/* disable RX MAC and wait for completion */
	writel(rxcfg & ~MAC_RX_CFG_EN, cp->regs + REG_MAC_RX_CFG);
	while (readl(cp->regs + REG_MAC_RX_CFG) & MAC_RX_CFG_EN) {
		if (!limit--)
			break;
		udelay(10);
	}

	/* disable hash filter and wait for completion */
	limit = STOP_TRIES;
	rxcfg &= ~(MAC_RX_CFG_PROMISC_EN | MAC_RX_CFG_HASH_FILTER_EN);
	writel(rxcfg & ~MAC_RX_CFG_EN, cp->regs + REG_MAC_RX_CFG);
	while (readl(cp->regs + REG_MAC_RX_CFG) & MAC_RX_CFG_HASH_FILTER_EN) {
		if (!limit--)
			break;
		udelay(10);
	}

	/* program hash filters */
	cp->mac_rx_cfg = rxcfg_new = cas_setup_multicast(cp);
	rxcfg |= rxcfg_new;
	writel(rxcfg, cp->regs + REG_MAC_RX_CFG);
	spin_unlock_irqrestore(&cp->lock, flags);
}

static void cas_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
{
	struct cas *cp = netdev_priv(dev);
	strncpy(info->driver, DRV_MODULE_NAME, ETHTOOL_BUSINFO_LEN);
	strncpy(info->version, DRV_MODULE_VERSION, ETHTOOL_BUSINFO_LEN);
	info->fw_version[0] = '\0';
	strncpy(info->bus_info, pci_name(cp->pdev), ETHTOOL_BUSINFO_LEN);
	info->regdump_len = cp->casreg_len < CAS_MAX_REGS ?
		cp->casreg_len : CAS_MAX_REGS;
	info->n_stats = CAS_NUM_STAT_KEYS;
}

static int cas_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
{
	struct cas *cp = netdev_priv(dev);
	u16 bmcr;
	int full_duplex, speed, pause;
	unsigned long flags;
	enum link_state linkstate = link_up;

	cmd->advertising = 0;
	cmd->supported = SUPPORTED_Autoneg;
	if (cp->cas_flags & CAS_FLAG_1000MB_CAP) {
		cmd->supported |= SUPPORTED_1000baseT_Full;
		cmd->advertising |= ADVERTISED_1000baseT_Full;
	}

	/* Record PHY settings if HW is on. */
	spin_lock_irqsave(&cp->lock, flags);
	bmcr = 0;
	linkstate = cp->lstate;
	if (CAS_PHY_MII(cp->phy_type)) {
		cmd->port = PORT_MII;
		cmd->transceiver = (cp->cas_flags & CAS_FLAG_SATURN) ?
			XCVR_INTERNAL : XCVR_EXTERNAL;
		cmd->phy_address = cp->phy_addr;
		cmd->advertising |= ADVERTISED_TP | ADVERTISED_MII |
			ADVERTISED_10baseT_Half |
			ADVERTISED_10baseT_Full |
			ADVERTISED_100baseT_Half |
			ADVERTISED_100baseT_Full;

		cmd->supported |=
			(SUPPORTED_10baseT_Half |
			 SUPPORTED_10baseT_Full |
			 SUPPORTED_100baseT_Half |
			 SUPPORTED_100baseT_Full |
			 SUPPORTED_TP | SUPPORTED_MII);

		if (cp->hw_running) {
			cas_mif_poll(cp, 0);
			bmcr = cas_phy_read(cp, MII_BMCR);
			cas_read_mii_link_mode(cp, &full_duplex,
					       &speed, &pause);
			cas_mif_poll(cp, 1);
		}

	} else {
		cmd->port = PORT_FIBRE;
		cmd->transceiver = XCVR_INTERNAL;
		cmd->phy_address = 0;
		cmd->supported   |= SUPPORTED_FIBRE;
		cmd->advertising |= ADVERTISED_FIBRE;

		if (cp->hw_running) {
			/* pcs uses the same bits as mii */
			bmcr = readl(cp->regs + REG_PCS_MII_CTRL);
			cas_read_pcs_link_mode(cp, &full_duplex,
					       &speed, &pause);
		}
	}
	spin_unlock_irqrestore(&cp->lock, flags);

	if (bmcr & BMCR_ANENABLE) {
		cmd->advertising |= ADVERTISED_Autoneg;
		cmd->autoneg = AUTONEG_ENABLE;
		cmd->speed = ((speed == 10) ?
			      SPEED_10 :
			      ((speed == 1000) ?
			       SPEED_1000 : SPEED_100));
		cmd->duplex = full_duplex ? DUPLEX_FULL : DUPLEX_HALF;
	} else {
		cmd->autoneg = AUTONEG_DISABLE;
		cmd->speed =
			(bmcr & CAS_BMCR_SPEED1000) ?
			SPEED_1000 :
			((bmcr & BMCR_SPEED100) ? SPEED_100:
			 SPEED_10);
		cmd->duplex =
			(bmcr & BMCR_FULLDPLX) ?
			DUPLEX_FULL : DUPLEX_HALF;
	}
	if (linkstate != link_up) {
		/* Force these to "unknown" if the link is not up and
		 * autonogotiation in enabled. We can set the link
		 * speed to 0, but not cmd->duplex,
		 * because its legal values are 0 and 1.  Ethtool will
		 * print the value reported in parentheses after the
		 * word "Unknown" for unrecognized values.
		 *
		 * If in forced mode, we report the speed and duplex
		 * settings that we configured.
		 */
		if (cp->link_cntl & BMCR_ANENABLE) {
			cmd->speed = 0;
			cmd->duplex = 0xff;
		} else {
			cmd->speed = SPEED_10;
			if (cp->link_cntl & BMCR_SPEED100) {
				cmd->speed = SPEED_100;
			} else if (cp->link_cntl & CAS_BMCR_SPEED1000) {
				cmd->speed = SPEED_1000;
			}
			cmd->duplex = (cp->link_cntl & BMCR_FULLDPLX)?
				DUPLEX_FULL : DUPLEX_HALF;
		}
	}
	return 0;
}

static int cas_set_settings(struct net_device *dev, struct ethtool_cmd *cmd)
{
	struct cas *cp = netdev_priv(dev);
	unsigned long flags;

	/* Verify the settings we care about. */
	if (cmd->autoneg != AUTONEG_ENABLE &&
	    cmd->autoneg != AUTONEG_DISABLE)
		return -EINVAL;

	if (cmd->autoneg == AUTONEG_DISABLE &&
	    ((cmd->speed != SPEED_1000 &&
	      cmd->speed != SPEED_100 &&
	      cmd->speed != SPEED_10) ||
	     (cmd->duplex != DUPLEX_HALF &&
	      cmd->duplex != DUPLEX_FULL)))
		return -EINVAL;

	/* Apply settings and restart link process. */
	spin_lock_irqsave(&cp->lock, flags);
	cas_begin_auto_negotiation(cp, cmd);
	spin_unlock_irqrestore(&cp->lock, flags);
	return 0;
}

static int cas_nway_reset(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);
	unsigned long flags;

	if ((cp->link_cntl & BMCR_ANENABLE) == 0)
		return -EINVAL;

	/* Restart link process. */
	spin_lock_irqsave(&cp->lock, flags);
	cas_begin_auto_negotiation(cp, NULL);
	spin_unlock_irqrestore(&cp->lock, flags);

	return 0;
}

static u32 cas_get_link(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);
	return cp->lstate == link_up;
}

static u32 cas_get_msglevel(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);
	return cp->msg_enable;
}

static void cas_set_msglevel(struct net_device *dev, u32 value)
{
	struct cas *cp = netdev_priv(dev);
	cp->msg_enable = value;
}

static int cas_get_regs_len(struct net_device *dev)
{
	struct cas *cp = netdev_priv(dev);
	return cp->casreg_len < CAS_MAX_REGS ? cp->casreg_len: CAS_MAX_REGS;
}

static void cas_get_regs(struct net_device *dev, struct ethtool_regs *regs,
			     void *p)
{
	struct cas *cp = netdev_priv(dev);
	regs->version = 0;
	/* cas_read_regs handles locks (cp->lock).  */
	cas_read_regs(cp, p, regs->len / sizeof(u32));
}

static int cas_get_sset_count(struct net_device *dev, int sset)
{
	switch (sset) {
	case ETH_SS_STATS:
		return CAS_NUM_STAT_KEYS;
	default:
		return -EOPNOTSUPP;
	}
}

static void cas_get_strings(struct net_device *dev, u32 stringset, u8 *data)
{
	 memcpy(data, &ethtool_cassini_statnames,
					 CAS_NUM_STAT_KEYS * ETH_GSTRING_LEN);
}

static void cas_get_ethtool_stats(struct net_device *dev,
				      struct ethtool_stats *estats, u64 *data)
{
	struct cas *cp = netdev_priv(dev);
	struct net_device_stats *stats = cas_get_stats(cp->dev);
	int i = 0;
	data[i++] = stats->collisions;
	data[i++] = stats->rx_bytes;
	data[i++] = stats->rx_crc_errors;
	data[i++] = stats->rx_dropped;
	data[i++] = stats->rx_errors;
	data[i++] = stats->rx_fifo_errors;
	data[i++] = stats->rx_frame_errors;
	data[i++] = stats->rx_length_errors;
	data[i++] = stats->rx_over_errors;
	data[i++] = stats->rx_packets;
	data[i++] = stats->tx_aborted_errors;
	data[i++] = stats->tx_bytes;
	data[i++] = stats->tx_dropped;
	data[i++] = stats->tx_errors;
	data[i++] = stats->tx_fifo_errors;
	data[i++] = stats->tx_packets;