aboutsummaryrefslogtreecommitdiff
path: root/drivers/usb/host/ehci-q.c
blob: 140bfa423e07d39b18815221723b5181654ebd15 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
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
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
/*
 * Copyright (C) 2001-2004 by David Brownell
 *
 * 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., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/* this file is part of ehci-hcd.c */

/*-------------------------------------------------------------------------*/

/*
 * EHCI hardware queue manipulation ... the core.  QH/QTD manipulation.
 *
 * Control, bulk, and interrupt traffic all use "qh" lists.  They list "qtd"
 * entries describing USB transactions, max 16-20kB/entry (with 4kB-aligned
 * buffers needed for the larger number).  We use one QH per endpoint, queue
 * multiple urbs (all three types) per endpoint.  URBs may need several qtds.
 *
 * ISO traffic uses "ISO TD" (itd, and sitd) records, and (along with
 * interrupts) needs careful scheduling.  Performance improvements can be
 * an ongoing challenge.  That's in "ehci-sched.c".
 *
 * USB 1.1 devices are handled (a) by "companion" OHCI or UHCI root hubs,
 * or otherwise through transaction translators (TTs) in USB 2.0 hubs using
 * (b) special fields in qh entries or (c) split iso entries.  TTs will
 * buffer low/full speed data so the host collects it at high speed.
 */

/*-------------------------------------------------------------------------*/

/* fill a qtd, returning how much of the buffer we were able to queue up */

static int
qtd_fill(struct ehci_hcd *ehci, struct ehci_qtd *qtd, dma_addr_t buf,
		  size_t len, int token, int maxpacket)
{
	int	i, count;
	u64	addr = buf;

	/* one buffer entry per 4K ... first might be short or unaligned */
	qtd->hw_buf[0] = cpu_to_hc32(ehci, (u32)addr);
	qtd->hw_buf_hi[0] = cpu_to_hc32(ehci, (u32)(addr >> 32));
	count = 0x1000 - (buf & 0x0fff);	/* rest of that page */
	if (likely (len < count))		/* ... iff needed */
		count = len;
	else {
		buf +=  0x1000;
		buf &= ~0x0fff;

		/* per-qtd limit: from 16K to 20K (best alignment) */
		for (i = 1; count < len && i < 5; i++) {
			addr = buf;
			qtd->hw_buf[i] = cpu_to_hc32(ehci, (u32)addr);
			qtd->hw_buf_hi[i] = cpu_to_hc32(ehci,
					(u32)(addr >> 32));
			buf += 0x1000;
			if ((count + 0x1000) < len)
				count += 0x1000;
			else
				count = len;
		}

		/* short packets may only terminate transfers */
		if (count != len)
			count -= (count % maxpacket);
	}
	qtd->hw_token = cpu_to_hc32(ehci, (count << 16) | token);
	qtd->length = count;

	return count;
}

/*-------------------------------------------------------------------------*/

static inline void
qh_update (struct ehci_hcd *ehci, struct ehci_qh *qh, struct ehci_qtd *qtd)
{
	/* writes to an active overlay are unsafe */
	BUG_ON(qh->qh_state != QH_STATE_IDLE);

	qh->hw_qtd_next = QTD_NEXT(ehci, qtd->qtd_dma);
	qh->hw_alt_next = EHCI_LIST_END(ehci);

	/* Except for control endpoints, we make hardware maintain data
	 * toggle (like OHCI) ... here (re)initialize the toggle in the QH,
	 * and set the pseudo-toggle in udev. Only usb_clear_halt() will
	 * ever clear it.
	 */
	if (!(qh->hw_info1 & cpu_to_hc32(ehci, 1 << 14))) {
		unsigned	is_out, epnum;

		is_out = !(qtd->hw_token & cpu_to_hc32(ehci, 1 << 8));
		epnum = (hc32_to_cpup(ehci, &qh->hw_info1) >> 8) & 0x0f;
		if (unlikely (!usb_gettoggle (qh->dev, epnum, is_out))) {
			qh->hw_token &= ~cpu_to_hc32(ehci, QTD_TOGGLE);
			usb_settoggle (qh->dev, epnum, is_out, 1);
		}
	}

	/* HC must see latest qtd and qh data before we clear ACTIVE+HALT */
	wmb ();
	qh->hw_token &= cpu_to_hc32(ehci, QTD_TOGGLE | QTD_STS_PING);
}

/* if it weren't for a common silicon quirk (writing the dummy into the qh
 * overlay, so qh->hw_token wrongly becomes inactive/halted), only fault
 * recovery (including urb dequeue) would need software changes to a QH...
 */
static void
qh_refresh (struct ehci_hcd *ehci, struct ehci_qh *qh)
{
	struct ehci_qtd *qtd;

	if (list_empty (&qh->qtd_list))
		qtd = qh->dummy;
	else {
		qtd = list_entry (qh->qtd_list.next,
				struct ehci_qtd, qtd_list);
		/* first qtd may already be partially processed */
		if (cpu_to_hc32(ehci, qtd->qtd_dma) == qh->hw_current)
			qtd = NULL;
	}

	if (qtd)
		qh_update (ehci, qh, qtd);
}

/*-------------------------------------------------------------------------*/

static void qtd_copy_status (
	struct ehci_hcd *ehci,
	struct urb *urb,
	size_t length,
	u32 token
)
{
	/* count IN/OUT bytes, not SETUP (even short packets) */
	if (likely (QTD_PID (token) != 2))
		urb->actual_length += length - QTD_LENGTH (token);

	/* don't modify error codes */
	if (unlikely (urb->status != -EINPROGRESS))
		return;

	/* force cleanup after short read; not always an error */
	if (unlikely (IS_SHORT_READ (token)))
		urb->status = -EREMOTEIO;

	/* serious "can't proceed" faults reported by the hardware */
	if (token & QTD_STS_HALT) {
		if (token & QTD_STS_BABBLE) {
			/* FIXME "must" disable babbling device's port too */
			urb->status = -EOVERFLOW;
		} else if (token & QTD_STS_MMF) {
			/* fs/ls interrupt xfer missed the complete-split */
			urb->status = -EPROTO;
		} else if (token & QTD_STS_DBE) {
			urb->status = (QTD_PID (token) == 1) /* IN ? */
				? -ENOSR  /* hc couldn't read data */
				: -ECOMM; /* hc couldn't write data */
		} else if (token & QTD_STS_XACT) {
			/* timeout, bad crc, wrong PID, etc; retried */
			if (QTD_CERR (token))
				urb->status = -EPIPE;
			else {
				ehci_dbg (ehci, "devpath %s ep%d%s 3strikes\n",
					urb->dev->devpath,
					usb_pipeendpoint (urb->pipe),
					usb_pipein (urb->pipe) ? "in" : "out");
				urb->status = -EPROTO;
			}
		/* CERR nonzero + no errors + halt --> stall */
		} else if (QTD_CERR (token))
			urb->status = -EPIPE;
		else	/* unknown */
			urb->status = -EPROTO;

		ehci_vdbg (ehci,
			"dev%d ep%d%s qtd token %08x --> status %d\n",
			usb_pipedevice (urb->pipe),
			usb_pipeendpoint (urb->pipe),
			usb_pipein (urb->pipe) ? "in" : "out",
			token, urb->status);

		/* if async CSPLIT failed, try cleaning out the TT buffer */
		if (urb->status != -EPIPE
				&& urb->dev->tt && !usb_pipeint (urb->pipe)
				&& ((token & QTD_STS_MMF) != 0
					|| QTD_CERR(token) == 0)
				&& (!ehci_is_TDI(ehci)
			                || urb->dev->tt->hub !=
					   ehci_to_hcd(ehci)->self.root_hub)) {
#ifdef DEBUG
			struct usb_device *tt = urb->dev->tt->hub;
			dev_dbg (&tt->dev,
				"clear tt buffer port %d, a%d ep%d t%08x\n",
				urb->dev->ttport, urb->dev->devnum,
				usb_pipeendpoint (urb->pipe), token);
#endif /* DEBUG */
			usb_hub_tt_clear_buffer (urb->dev, urb->pipe);
		}
	}
}

static void
ehci_urb_done (struct ehci_hcd *ehci, struct urb *urb)
__releases(ehci->lock)
__acquires(ehci->lock)
{
	if (likely (urb->hcpriv != NULL)) {
		struct ehci_qh	*qh = (struct ehci_qh *) urb->hcpriv;

		/* S-mask in a QH means it's an interrupt urb */
		if ((qh->hw_info2 & cpu_to_hc32(ehci, QH_SMASK)) != 0) {

			/* ... update hc-wide periodic stats (for usbfs) */
			ehci_to_hcd(ehci)->self.bandwidth_int_reqs--;
		}
		qh_put (qh);
	}

	spin_lock (&urb->lock);
	urb->hcpriv = NULL;
	switch (urb->status) {
	case -EINPROGRESS:		/* success */
		urb->status = 0;
	default:			/* fault */
		COUNT (ehci->stats.complete);
		break;
	case -EREMOTEIO:		/* fault or normal */
		if (!(urb->transfer_flags & URB_SHORT_NOT_OK))
			urb->status = 0;
		COUNT (ehci->stats.complete);
		break;
	case -ECONNRESET:		/* canceled */
	case -ENOENT:
		COUNT (ehci->stats.unlink);
		break;
	}
	spin_unlock (&urb->lock);

#ifdef EHCI_URB_TRACE
	ehci_dbg (ehci,
		"%s %s urb %p ep%d%s status %d len %d/%d\n",
		__FUNCTION__, urb->dev->devpath, urb,
		usb_pipeendpoint (urb->pipe),
		usb_pipein (urb->pipe) ? "in" : "out",
		urb->status,
		urb->actual_length, urb->transfer_buffer_length);
#endif

	/* complete() can reenter this HCD */
	spin_unlock (&ehci->lock);
	usb_hcd_giveback_urb (ehci_to_hcd(ehci), urb);
	spin_lock (&ehci->lock);
}

static void start_unlink_async (struct ehci_hcd *ehci, struct ehci_qh *qh);
static void unlink_async (struct ehci_hcd *ehci, struct ehci_qh *qh);

static void intr_deschedule (struct ehci_hcd *ehci, struct ehci_qh *qh);
static int qh_schedule (struct ehci_hcd *ehci, struct ehci_qh *qh);

/*
 * Process and free completed qtds for a qh, returning URBs to drivers.
 * Chases up to qh->hw_current.  Returns number of completions called,
 * indicating how much "real" work we did.
 */
static unsigned
qh_completions (struct ehci_hcd *ehci, struct ehci_qh *qh)
{
	struct ehci_qtd		*last = NULL, *end = qh->dummy;
	struct list_head	*entry, *tmp;
	int			stopped;
	unsigned		count = 0;
	int			do_status = 0;
	u8			state;
	u32			halt = HALT_BIT(ehci);

	if (unlikely (list_empty (&qh->qtd_list)))
		return count;

	/* completions (or tasks on other cpus) must never clobber HALT
	 * till we've gone through and cleaned everything up, even when
	 * they add urbs to this qh's queue or mark them for unlinking.
	 *
	 * NOTE:  unlinking expects to be done in queue order.
	 */
	state = qh->qh_state;
	qh->qh_state = QH_STATE_COMPLETING;
	stopped = (state == QH_STATE_IDLE);

	/* remove de-activated QTDs from front of queue.
	 * after faults (including short reads), cleanup this urb
	 * then let the queue advance.
	 * if queue is stopped, handles unlinks.
	 */
	list_for_each_safe (entry, tmp, &qh->qtd_list) {
		struct ehci_qtd	*qtd;
		struct urb	*urb;
		u32		token = 0;

		qtd = list_entry (entry, struct ehci_qtd, qtd_list);
		urb = qtd->urb;

		/* clean up any state from previous QTD ...*/
		if (last) {
			if (likely (last->urb != urb)) {
				ehci_urb_done (ehci, last->urb);
				count++;
			}
			ehci_qtd_free (ehci, last);
			last = NULL;
		}

		/* ignore urbs submitted during completions we reported */
		if (qtd == end)
			break;

		/* hardware copies qtd out of qh overlay */
		rmb ();
		token = hc32_to_cpu(ehci, qtd->hw_token);

		/* always clean up qtds the hc de-activated */
		if ((token & QTD_STS_ACTIVE) == 0) {

			if ((token & QTD_STS_HALT) != 0) {
				stopped = 1;

			/* magic dummy for some short reads; qh won't advance.
			 * that silicon quirk can kick in with this dummy too.
			 */
			} else if (IS_SHORT_READ (token)
					&& !(qtd->hw_alt_next
						& EHCI_LIST_END(ehci))) {
				stopped = 1;
				goto halt;
			}

		/* stop scanning when we reach qtds the hc is using */
		} else if (likely (!stopped
				&& HC_IS_RUNNING (ehci_to_hcd(ehci)->state))) {
			break;

		} else {
			stopped = 1;

			if (unlikely (!HC_IS_RUNNING (ehci_to_hcd(ehci)->state)))
				urb->status = -ESHUTDOWN;

			/* ignore active urbs unless some previous qtd
			 * for the urb faulted (including short read) or
			 * its urb was canceled.  we may patch qh or qtds.
			 */
			if (likely (urb->status == -EINPROGRESS))
				continue;

			/* issue status after short control reads */
			if (unlikely (do_status != 0)
					&& QTD_PID (token) == 0 /* OUT */) {
				do_status = 0;
				continue;
			}

			/* token in overlay may be most current */
			if (state == QH_STATE_IDLE
					&& cpu_to_hc32(ehci, qtd->qtd_dma)
						== qh->hw_current)
				token = hc32_to_cpu(ehci, qh->hw_token);

			/* force halt for unlinked or blocked qh, so we'll
			 * patch the qh later and so that completions can't
			 * activate it while we "know" it's stopped.
			 */
			if ((halt & qh->hw_token) == 0) {
halt:
				qh->hw_token |= halt;
				wmb ();
			}
		}

		/* remove it from the queue */
		spin_lock (&urb->lock);
		qtd_copy_status (ehci, urb, qtd->length, token);
		do_status = (urb->status == -EREMOTEIO)
				&& usb_pipecontrol (urb->pipe);
		spin_unlock (&urb->lock);

		if (stopped && qtd->qtd_list.prev != &qh->qtd_list) {
			last = list_entry (qtd->qtd_list.prev,
					struct ehci_qtd, qtd_list);
			last->hw_next = qtd->hw_next;
		}
		list_del (&qtd->qtd_list);
		last = qtd;
	}

	/* last urb's completion might still need calling */
	if (likely (last != NULL)) {
		ehci_urb_done (ehci, last->urb);
		count++;
		ehci_qtd_free (ehci, last);
	}

	/* restore original state; caller must unlink or relink */
	qh->qh_state = state;

	/* be sure the hardware's done with the qh before refreshing
	 * it after fault cleanup, or recovering from silicon wrongly
	 * overlaying the dummy qtd (which reduces DMA chatter).
	 */
	if (stopped != 0 || qh->hw_qtd_next == EHCI_LIST_END(ehci)) {
		switch (state) {
		case QH_STATE_IDLE:
			qh_refresh(ehci, qh);
			break;
		case QH_STATE_LINKED:
			/* should be rare for periodic transfers,
			 * except maybe high bandwidth ...
			 */
			if ((cpu_to_hc32(ehci, QH_SMASK)
					& qh->hw_info2) != 0) {
				intr_deschedule (ehci, qh);
				(void) qh_schedule (ehci, qh);
			} else
				unlink_async (ehci, qh);
			break;
		/* otherwise, unlink already started */
		}
	}

	return count;
}

/*-------------------------------------------------------------------------*/

// high bandwidth multiplier, as encoded in highspeed endpoint descriptors
#define hb_mult(wMaxPacketSize) (1 + (((wMaxPacketSize) >> 11) & 0x03))
// ... and packet size, for any kind of endpoint descriptor
#define max_packet(wMaxPacketSize) ((wMaxPacketSize) & 0x07ff)

/*
 * reverse of qh_urb_transaction:  free a list of TDs.
 * used for cleanup after errors, before HC sees an URB's TDs.
 */
static void qtd_list_free (
	struct ehci_hcd		*ehci,
	struct urb		*urb,
	struct list_head	*qtd_list
) {
	struct list_head	*entry, *temp;

	list_for_each_safe (entry, temp, qtd_list) {
		struct ehci_qtd	*qtd;

		qtd = list_entry (entry, struct ehci_qtd, qtd_list);
		list_del (&qtd->qtd_list);
		ehci_qtd_free (ehci, qtd);
	}
}

/*
 * create a list of filled qtds for this URB; won't link into qh.
 */
static struct list_head *
qh_urb_transaction (
	struct ehci_hcd		*ehci,
	struct urb		*urb,
	struct list_head	*head,
	gfp_t			flags
) {
	struct ehci_qtd		*qtd, *qtd_prev;
	dma_addr_t		buf;
	int			len, maxpacket;
	int			is_input;
	u32			token;

	/*
	 * URBs map to sequences of QTDs:  one logical transaction
	 */
	qtd = ehci_qtd_alloc (ehci, flags);
	if (unlikely (!qtd))
		return NULL;
	list_add_tail (&qtd->qtd_list, head);
	qtd->urb = urb;

	token = QTD_STS_ACTIVE;
	token |= (EHCI_TUNE_CERR << 10);
	/* for split transactions, SplitXState initialized to zero */

	len = urb->transfer_buffer_length;
	is_input = usb_pipein (urb->pipe);
	if (usb_pipecontrol (urb->pipe)) {
		/* SETUP pid */
		qtd_fill(ehci, qtd, urb->setup_dma,
				sizeof (struct usb_ctrlrequest),
				token | (2 /* "setup" */ << 8), 8);

		/* ... and always at least one more pid */
		token ^= QTD_TOGGLE;
		qtd_prev = qtd;
		qtd = ehci_qtd_alloc (ehci, flags);
		if (unlikely (!qtd))
			goto cleanup;
		qtd->urb = urb;
		qtd_prev->hw_next = QTD_NEXT(ehci, qtd->qtd_dma);
		list_add_tail (&qtd->qtd_list, head);

		/* for zero length DATA stages, STATUS is always IN */
		if (len == 0)
			token |= (1 /* "in" */ << 8);
	}

	/*
	 * data transfer stage:  buffer setup
	 */
	buf = urb->transfer_dma;

	if (is_input)
		token |= (1 /* "in" */ << 8);
	/* else it's already initted to "out" pid (0 << 8) */

	maxpacket = max_packet(usb_maxpacket(urb->dev, urb->pipe, !is_input));

	/*
	 * buffer gets wrapped in one or more qtds;
	 * last one may be "short" (including zero len)
	 * and may serve as a control status ack
	 */
	for (;;) {
		int this_qtd_len;

		this_qtd_len = qtd_fill(ehci, qtd, buf, len, token, maxpacket);
		len -= this_qtd_len;
		buf += this_qtd_len;
		if (is_input)
			qtd->hw_alt_next = ehci->async->hw_alt_next;

		/* qh makes control packets use qtd toggle; maybe switch it */
		if ((maxpacket & (this_qtd_len + (maxpacket - 1))) == 0)
			token ^= QTD_TOGGLE;

		if (likely (len <= 0))
			break;

		qtd_prev = qtd;
		qtd = ehci_qtd_alloc (ehci, flags);
		if (unlikely (!qtd))
			goto cleanup;
		qtd->urb = urb;
		qtd_prev->hw_next = QTD_NEXT(ehci, qtd->qtd_dma);
		list_add_tail (&qtd->qtd_list, head);
	}

	/* unless the bulk/interrupt caller wants a chance to clean
	 * up after short reads, hc should advance qh past this urb
	 */
	if (likely ((urb->transfer_flags & URB_SHORT_NOT_OK) == 0
				|| usb_pipecontrol (urb->pipe)))
		qtd->hw_alt_next = EHCI_LIST_END(ehci);

	/*
	 * control requests may need a terminating data "status" ack;
	 * bulk ones may need a terminating short packet (zero length).
	 */
	if (likely (urb->transfer_buffer_length != 0)) {
		int	one_more = 0;

		if (usb_pipecontrol (urb->pipe)) {
			one_more = 1;
			token ^= 0x0100;	/* "in" <--> "out"  */
			token |= QTD_TOGGLE;	/* force DATA1 */
		} else if (usb_pipebulk (urb->pipe)
				&& (urb->transfer_flags & URB_ZERO_PACKET)
				&& !(urb->transfer_buffer_length % maxpacket)) {
			one_more = 1;
		}
		if (one_more) {
			qtd_prev = qtd;
			qtd = ehci_qtd_alloc (ehci, flags);
			if (unlikely (!qtd))
				goto cleanup;
			qtd->urb = urb;
			qtd_prev->hw_next = QTD_NEXT(ehci, qtd->qtd_dma);
			list_add_tail (&qtd->qtd_list, head);

			/* never any data in such packets */
			qtd_fill(ehci, qtd, 0, 0, token, 0);
		}
	}

	/* by default, enable interrupt on urb completion */
	if (likely (!(urb->transfer_flags & URB_NO_INTERRUPT)))
		qtd->hw_token |= cpu_to_hc32(ehci, QTD_IOC);
	return head;

cleanup:
	qtd_list_free (ehci, urb, head);
	return NULL;
}

/*-------------------------------------------------------------------------*/

// Would be best to create all qh's from config descriptors,
// when each interface/altsetting is established.  Unlink
// any previous qh and cancel its urbs first; endpoints are
// implicitly reset then (data toggle too).
// That'd mean updating how usbcore talks to HCDs. (2.7?)


/*
 * Each QH holds a qtd list; a QH is used for everything except iso.
 *
 * For interrupt urbs, the scheduler must set the microframe scheduling
 * mask(s) each time the QH gets scheduled.  For highspeed, that's
 * just one microframe in the s-mask.  For split interrupt transactions
 * there are additional complications: c-mask, maybe FSTNs.
 */
static struct ehci_qh *
qh_make (
	struct ehci_hcd		*ehci,
	struct urb		*urb,
	gfp_t			flags
) {
	struct ehci_qh		*qh = ehci_qh_alloc (ehci, flags);
	u32			info1 = 0, info2 = 0;
	int			is_input, type;
	int			maxp = 0;

	if (!qh)
		return qh;

	/*
	 * init endpoint/device data for this QH
	 */
	info1 |= usb_pipeendpoint (urb->pipe) << 8;
	info1 |= usb_pipedevice (urb->pipe) << 0;

	is_input = usb_pipein (urb->pipe);
	type = usb_pipetype (urb->pipe);
	maxp = usb_maxpacket (urb->dev, urb->pipe, !is_input);

	/* Compute interrupt scheduling parameters just once, and save.
	 * - allowing for high bandwidth, how many nsec/uframe are used?
	 * - split transactions need a second CSPLIT uframe; same question
	 * - splits also need a schedule gap (for full/low speed I/O)
	 * - qh has a polling interval
	 *
	 * For control/bulk requests, the HC or TT handles these.
	 */
	if (type == PIPE_INTERRUPT) {
		qh->usecs = NS_TO_US (usb_calc_bus_time (USB_SPEED_HIGH, is_input, 0,
				hb_mult (maxp) * max_packet (maxp)));
		qh->start = NO_FRAME;

		if (urb->dev->speed == USB_SPEED_HIGH) {
			qh->c_usecs = 0;
			qh->gap_uf = 0;

			qh->period = urb->interval >> 3;
			if (qh->period == 0 && urb->interval != 1) {
				/* NOTE interval 2 or 4 uframes could work.
				 * But interval 1 scheduling is simpler, and
				 * includes high bandwidth.
				 */
				dbg ("intr period %d uframes, NYET!",
						urb->interval);
				goto done;
			}
		} else {
			struct usb_tt	*tt = urb->dev->tt;
			int		think_time;

			/* gap is f(FS/LS transfer times) */
			qh->gap_uf = 1 + usb_calc_bus_time (urb->dev->speed,
					is_input, 0, maxp) / (125 * 1000);

			/* FIXME this just approximates SPLIT/CSPLIT times */
			if (is_input) {		// SPLIT, gap, CSPLIT+DATA
				qh->c_usecs = qh->usecs + HS_USECS (0);
				qh->usecs = HS_USECS (1);
			} else {		// SPLIT+DATA, gap, CSPLIT
				qh->usecs += HS_USECS (1);
				qh->c_usecs = HS_USECS (0);
			}

			think_time = tt ? tt->think_time : 0;
			qh->tt_usecs = NS_TO_US (think_time +
					usb_calc_bus_time (urb->dev->speed,
					is_input, 0, max_packet (maxp)));
			qh->period = urb->interval;
		}
	}

	/* support for tt scheduling, and access to toggles */
	qh->dev = urb->dev;

	/* using TT? */
	switch (urb->dev->speed) {
	case USB_SPEED_LOW:
		info1 |= (1 << 12);	/* EPS "low" */
		/* FALL THROUGH */

	case USB_SPEED_FULL:
		/* EPS 0 means "full" */
		if (type != PIPE_INTERRUPT)
			info1 |= (EHCI_TUNE_RL_TT << 28);
		if (type == PIPE_CONTROL) {
			info1 |= (1 << 27);	/* for TT */
			info1 |= 1 << 14;	/* toggle from qtd */
		}
		info1 |= maxp << 16;

		info2 |= (EHCI_TUNE_MULT_TT << 30);

		/* Some Freescale processors have an erratum in which the
		 * port number in the queue head was 0..N-1 instead of 1..N.
		 */
		if (ehci_has_fsl_portno_bug(ehci))
			info2 |= (urb->dev->ttport-1) << 23;
		else
			info2 |= urb->dev->ttport << 23;

		/* set the address of the TT; for TDI's integrated
		 * root hub tt, leave it zeroed.
		 */
		if (!ehci_is_TDI(ehci)
				|| urb->dev->tt->hub !=
					ehci_to_hcd(ehci)->self.root_hub)
			info2 |= urb->dev->tt->hub->devnum << 16;

		/* NOTE:  if (PIPE_INTERRUPT) { scheduler sets c-mask } */

		break;

	case USB_SPEED_HIGH:		/* no TT involved */
		info1 |= (2 << 12);	/* EPS "high" */
		if (type == PIPE_CONTROL) {
			info1 |= (EHCI_TUNE_RL_HS << 28);
			info1 |= 64 << 16;	/* usb2 fixed maxpacket */
			info1 |= 1 << 14;	/* toggle from qtd */
			info2 |= (EHCI_TUNE_MULT_HS << 30);
		} else if (type == PIPE_BULK) {
			info1 |= (EHCI_TUNE_RL_HS << 28);
			info1 |= 512 << 16;	/* usb2 fixed maxpacket */
			info2 |= (EHCI_TUNE_MULT_HS << 30);
		} else {		/* PIPE_INTERRUPT */
			info1 |= max_packet (maxp) << 16;
			info2 |= hb_mult (maxp) << 30;
		}
		break;
	default:
		dbg ("bogus dev %p speed %d", urb->dev, urb->dev->speed);
done:
		qh_put (qh);
		return NULL;
	}

	/* NOTE:  if (PIPE_INTERRUPT) { scheduler sets s-mask } */

	/* init as live, toggle clear, advance to dummy */
	qh->qh_state = QH_STATE_IDLE;
	qh->hw_info1 = cpu_to_hc32(ehci, info1);
	qh->hw_info2 = cpu_to_hc32(ehci, info2);
	usb_settoggle (urb->dev, usb_pipeendpoint (urb->pipe), !is_input, 1);
	qh_refresh (ehci, qh);
	return qh;
}

/*-------------------------------------------------------------------------*/

/* move qh (and its qtds) onto async queue; maybe enable queue.  */

static void qh_link_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
{
	__hc32		dma = QH_NEXT(ehci, qh->qh_dma);
	struct ehci_qh	*head;

	/* (re)start the async schedule? */
	head = ehci->async;
	timer_action_done (ehci, TIMER_ASYNC_OFF);
	if (!head->qh_next.qh) {
		u32	cmd = ehci_readl(ehci, &ehci->regs->command);

		if (!(cmd & CMD_ASE)) {
			/* in case a clear of CMD_ASE didn't take yet */
			(void)handshake(ehci, &ehci->regs->status,
					STS_ASS, 0, 150);
			cmd |= CMD_ASE | CMD_RUN;
			ehci_writel(ehci, cmd, &ehci->regs->command);
			ehci_to_hcd(ehci)->state = HC_STATE_RUNNING;
			/* posted write need not be known to HC yet ... */
		}
	}

	/* clear halt and/or toggle; and maybe recover from silicon quirk */
	if (qh->qh_state == QH_STATE_IDLE)
		qh_refresh (ehci, qh);

	/* splice right after start */
	qh->qh_next = head->qh_next;
	qh->hw_next = head->hw_next;
	wmb ();

	head->qh_next.qh = qh;
	head->hw_next = dma;

	qh->qh_state = QH_STATE_LINKED;
	/* qtd completions reported later by interrupt */
}

/*-------------------------------------------------------------------------*/

/*
 * For control/bulk/interrupt, return QH with these TDs appended.
 * Allocates and initializes the QH if necessary.
 * Returns null if it can't allocate a QH it needs to.
 * If the QH has TDs (urbs) already, that's great.
 */
static struct ehci_qh *qh_append_tds (
	struct ehci_hcd		*ehci,
	struct urb		*urb,
	struct list_head	*qtd_list,
	int			epnum,
	void			**ptr
)
{
	struct ehci_qh		*qh = NULL;
	u32			qh_addr_mask = cpu_to_hc32(ehci, 0x7f);

	qh = (struct ehci_qh *) *ptr;
	if (unlikely (qh == NULL)) {
		/* can't sleep here, we have ehci->lock... */
		qh = qh_make (ehci, urb, GFP_ATOMIC);
		*ptr = qh;
	}
	if (likely (qh != NULL)) {
		struct ehci_qtd	*qtd;

		if (unlikely (list_empty (qtd_list)))
			qtd = NULL;
		else
			qtd = list_entry (qtd_list->next, struct ehci_qtd,
					qtd_list);

		/* control qh may need patching ... */
		if (unlikely (epnum == 0)) {

                        /* usb_reset_device() briefly reverts to address 0 */
                        if (usb_pipedevice (urb->pipe) == 0)
                                qh->hw_info1 &= ~qh_addr_mask;
		}

		/* just one way to queue requests: swap with the dummy qtd.
		 * only hc or qh_refresh() ever modify the overlay.
		 */
		if (likely (qtd != NULL)) {
			struct ehci_qtd		*dummy;
			dma_addr_t		dma;
			__hc32			token;

			/* to avoid racing the HC, use the dummy td instead of
			 * the first td of our list (becomes new dummy).  both
			 * tds stay deactivated until we're done, when the
			 * HC is allowed to fetch the old dummy (4.10.2).
			 */
			token = qtd->hw_token;
			qtd->hw_token = HALT_BIT(ehci);
			wmb ();
			dummy = qh->dummy;

			dma = dummy->qtd_dma;
			*dummy = *qtd;
			dummy->qtd_dma = dma;

			list_del (&qtd->qtd_list);
			list_add (&dummy->qtd_list, qtd_list);
			__list_splice (qtd_list, qh->qtd_list.prev);

			ehci_qtd_init(ehci, qtd, qtd->qtd_dma);
			qh->dummy = qtd;

			/* hc must see the new dummy at list end */
			dma = qtd->qtd_dma;
			qtd = list_entry (qh->qtd_list.prev,
					struct ehci_qtd, qtd_list);
			qtd->hw_next = QTD_NEXT(ehci, dma);

			/* let the hc process these next qtds */
			wmb ();
			dummy->hw_token = token;

			urb->hcpriv = qh_get (qh);
		}
	}
	return qh;
}

/*-------------------------------------------------------------------------*/

static int
submit_async (
	struct ehci_hcd		*ehci,
	struct usb_host_endpoint *ep,
	struct urb		*urb,
	struct list_head	*qtd_list,
	gfp_t			mem_flags
) {
	struct ehci_qtd		*qtd;
	int			epnum;
	unsigned long		flags;
	struct ehci_qh		*qh = NULL;
	int			rc = 0;

	qtd = list_entry (qtd_list->next, struct ehci_qtd, qtd_list);
	epnum = ep->desc.bEndpointAddress;

#ifdef EHCI_URB_TRACE
	ehci_dbg (ehci,
		"%s %s urb %p ep%d%s len %d, qtd %p [qh %p]\n",
		__FUNCTION__, urb->dev->devpath, urb,
		epnum & 0x0f, (epnum & USB_DIR_IN) ? "in" : "out",
		urb->transfer_buffer_length,
		qtd, ep->hcpriv);
#endif

	spin_lock_irqsave (&ehci->lock, flags);
	if (unlikely(!test_bit(HCD_FLAG_HW_ACCESSIBLE,
			       &ehci_to_hcd(ehci)->flags))) {
		rc = -ESHUTDOWN;
		goto done;
	}

	qh = qh_append_tds (ehci, urb, qtd_list, epnum, &ep->hcpriv);
	if (unlikely(qh == NULL)) {
		rc = -ENOMEM;
		goto done;
	}

	/* Control/bulk operations through TTs don't need scheduling,
	 * the HC and TT handle it when the TT has a buffer ready.
	 */
	if (likely (qh->qh_state == QH_STATE_IDLE))
		qh_link_async (ehci, qh_get (qh));
 done:
	spin_unlock_irqrestore (&ehci->lock, flags);
	if (unlikely (qh == NULL))
		qtd_list_free (ehci, urb, qtd_list);
	return rc;
}

/*-------------------------------------------------------------------------*/

/* the async qh for the qtds being reclaimed are now unlinked from the HC */

static void end_unlink_async (struct ehci_hcd *ehci)
{
	struct ehci_qh		*qh = ehci->reclaim;
	struct ehci_qh		*next;

	timer_action_done (ehci, TIMER_IAA_WATCHDOG);

	// qh->hw_next = cpu_to_hc32(qh->qh_dma);
	qh->qh_state = QH_STATE_IDLE;
	qh->qh_next.qh = NULL;
	qh_put (qh);			// refcount from reclaim

	/* other unlink(s) may be pending (in QH_STATE_UNLINK_WAIT) */
	next = qh->reclaim;
	ehci->reclaim = next;
	ehci->reclaim_ready = 0;
	qh->reclaim = NULL;

	qh_completions (ehci, qh);

	if (!list_empty (&qh->qtd_list)
			&& HC_IS_RUNNING (ehci_to_hcd(ehci)->state))
		qh_link_async (ehci, qh);
	else {
		qh_put (qh);		// refcount from async list

		/* it's not free to turn the async schedule on/off; leave it
		 * active but idle for a while once it empties.
		 */
		if (HC_IS_RUNNING (ehci_to_hcd(ehci)->state)
				&& ehci->async->qh_next.qh == NULL)
			timer_action (ehci, TIMER_ASYNC_OFF);
	}

	if (next) {
		ehci->reclaim = NULL;
		start_unlink_async (ehci, next);
	}
}

/* makes sure the async qh will become idle */
/* caller must own ehci->lock */

static void start_unlink_async (struct ehci_hcd *ehci, struct ehci_qh *qh)
{
	int		cmd = ehci_readl(ehci, &ehci->regs->command);
	struct ehci_qh	*prev;

#ifdef DEBUG
	assert_spin_locked(&ehci->lock);
	if (ehci->reclaim
			|| (qh->qh_state != QH_STATE_LINKED
				&& qh->qh_state != QH_STATE_UNLINK_WAIT)
			)
		BUG ();
#endif

	/* stop async schedule right now? */
	if (unlikely (qh == ehci->async)) {
		/* can't get here without STS_ASS set */
		if (ehci_to_hcd(ehci)->state != HC_STATE_HALT
				&& !ehci->reclaim) {
			/* ... and CMD_IAAD clear */
			ehci_writel(ehci, cmd & ~CMD_ASE,
				    &ehci->regs->command);
			wmb ();
			// handshake later, if we need to
			timer_action_done (ehci, TIMER_ASYNC_OFF);
		}
		return;
	}

	qh->qh_state = QH_STATE_UNLINK;
	ehci->reclaim = qh = qh_get (qh);

	prev = ehci->async;
	while (prev->qh_next.qh != qh)
		prev = prev->qh_next.qh;

	prev->hw_next = qh->hw_next;
	prev->qh_next = qh->qh_next;
	wmb ();

	if (unlikely (ehci_to_hcd(ehci)->state == HC_STATE_HALT)) {
		/* if (unlikely (qh->reclaim != 0))
		 *	this will recurse, probably not much
		 */
		end_unlink_async (ehci);
		return;
	}

	ehci->reclaim_ready = 0;
	cmd |= CMD_IAAD;
	ehci_writel(ehci, cmd, &ehci->regs->command);
	(void)ehci_readl(ehci, &ehci->regs->command);
	timer_action (ehci, TIMER_IAA_WATCHDOG);
}

/*-------------------------------------------------------------------------*/

static void scan_async (struct ehci_hcd *ehci)
{
	struct ehci_qh		*qh;
	enum ehci_timer_action	action = TIMER_IO_WATCHDOG;

	if (!++(ehci->stamp))
		ehci->stamp++;
	timer_action_done (ehci, TIMER_ASYNC_SHRINK);
rescan:
	qh = ehci->async->qh_next.qh;
	if (likely (qh != NULL)) {
		do {
			/* clean any finished work for this qh */
			if (!list_empty (&qh->qtd_list)
					&& qh->stamp != ehci->stamp) {
				int temp;

				/* unlinks could happen here; completion
				 * reporting drops the lock.  rescan using
				 * the latest schedule, but don't rescan
				 * qhs we already finished (no looping).
				 */
				qh = qh_get (qh);
				qh->stamp = ehci->stamp;
				temp = qh_completions (ehci, qh);
				qh_put (qh);
				if (temp != 0) {
					goto rescan;
				}
			}

			/* unlink idle entries, reducing HC PCI usage as well
			 * as HCD schedule-scanning costs.  delay for any qh
			 * we just scanned, there's a not-unusual case that it
			 * doesn't stay idle for long.
			 * (plus, avoids some kind of re-activation race.)
			 */
			if (list_empty (&qh->qtd_list)) {
				if (qh->stamp == ehci->stamp)
					action = TIMER_ASYNC_SHRINK;
				else if (!ehci->reclaim
					    && qh->qh_state == QH_STATE_LINKED)
					start_unlink_async (ehci, qh);
			}

			qh = qh->qh_next.qh;
		} while (qh);
	}
	if (action == TIMER_ASYNC_SHRINK)
		timer_action (ehci, TIMER_ASYNC_SHRINK);
}