summaryrefslogtreecommitdiff
path: root/drivers/dma/ioat_dma.c
blob: 7e4a785c2dff6fdb5d1640e9b0499974f27f8cc5 (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
/*
 * Intel I/OAT DMA Linux driver
 * Copyright(c) 2004 - 2007 Intel Corporation.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * 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.,
 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * The full GNU General Public License is included in this distribution in
 * the file called "COPYING".
 *
 */

/*
 * This driver supports an Intel I/OAT DMA engine, which does asynchronous
 * copy operations.
 */

#include <linux/init.h>
#include <linux/module.h>
#include <linux/pci.h>
#include <linux/interrupt.h>
#include <linux/dmaengine.h>
#include <linux/delay.h>
#include <linux/dma-mapping.h>
#include "ioatdma.h"
#include "ioatdma_registers.h"
#include "ioatdma_hw.h"

#define INITIAL_IOAT_DESC_COUNT 128

#define to_ioat_chan(chan) container_of(chan, struct ioat_dma_chan, common)
#define to_ioatdma_device(dev) container_of(dev, struct ioatdma_device, common)
#define to_ioat_desc(lh) container_of(lh, struct ioat_desc_sw, node)
#define tx_to_ioat_desc(tx) container_of(tx, struct ioat_desc_sw, async_tx)

/* internal functions */
static void ioat_dma_start_null_desc(struct ioat_dma_chan *ioat_chan);
static void ioat_dma_memcpy_cleanup(struct ioat_dma_chan *ioat_chan);
static struct ioat_desc_sw *
ioat_dma_get_next_descriptor(struct ioat_dma_chan *ioat_chan);

static inline struct ioat_dma_chan *ioat_lookup_chan_by_index(
						struct ioatdma_device *device,
						int index)
{
	return device->idx[index];
}

/**
 * ioat_dma_do_interrupt - handler used for single vector interrupt mode
 * @irq: interrupt id
 * @data: interrupt data
 */
static irqreturn_t ioat_dma_do_interrupt(int irq, void *data)
{
	struct ioatdma_device *instance = data;
	struct ioat_dma_chan *ioat_chan;
	unsigned long attnstatus;
	int bit;
	u8 intrctrl;

	intrctrl = readb(instance->reg_base + IOAT_INTRCTRL_OFFSET);

	if (!(intrctrl & IOAT_INTRCTRL_MASTER_INT_EN))
		return IRQ_NONE;

	if (!(intrctrl & IOAT_INTRCTRL_INT_STATUS)) {
		writeb(intrctrl, instance->reg_base + IOAT_INTRCTRL_OFFSET);
		return IRQ_NONE;
	}

	attnstatus = readl(instance->reg_base + IOAT_ATTNSTATUS_OFFSET);
	for_each_bit(bit, &attnstatus, BITS_PER_LONG) {
		ioat_chan = ioat_lookup_chan_by_index(instance, bit);
		tasklet_schedule(&ioat_chan->cleanup_task);
	}

	writeb(intrctrl, instance->reg_base + IOAT_INTRCTRL_OFFSET);
	return IRQ_HANDLED;
}

/**
 * ioat_dma_do_interrupt_msix - handler used for vector-per-channel interrupt mode
 * @irq: interrupt id
 * @data: interrupt data
 */
static irqreturn_t ioat_dma_do_interrupt_msix(int irq, void *data)
{
	struct ioat_dma_chan *ioat_chan = data;

	tasklet_schedule(&ioat_chan->cleanup_task);

	return IRQ_HANDLED;
}

static void ioat_dma_cleanup_tasklet(unsigned long data);

/**
 * ioat_dma_enumerate_channels - find and initialize the device's channels
 * @device: the device to be enumerated
 */
static int ioat_dma_enumerate_channels(struct ioatdma_device *device)
{
	u8 xfercap_scale;
	u32 xfercap;
	int i;
	struct ioat_dma_chan *ioat_chan;

	device->common.chancnt = readb(device->reg_base + IOAT_CHANCNT_OFFSET);
	xfercap_scale = readb(device->reg_base + IOAT_XFERCAP_OFFSET);
	xfercap = (xfercap_scale == 0 ? -1 : (1UL << xfercap_scale));

	for (i = 0; i < device->common.chancnt; i++) {
		ioat_chan = kzalloc(sizeof(*ioat_chan), GFP_KERNEL);
		if (!ioat_chan) {
			device->common.chancnt = i;
			break;
		}

		ioat_chan->device = device;
		ioat_chan->reg_base = device->reg_base + (0x80 * (i + 1));
		ioat_chan->xfercap = xfercap;
		spin_lock_init(&ioat_chan->cleanup_lock);
		spin_lock_init(&ioat_chan->desc_lock);
		INIT_LIST_HEAD(&ioat_chan->free_desc);
		INIT_LIST_HEAD(&ioat_chan->used_desc);
		/* This should be made common somewhere in dmaengine.c */
		ioat_chan->common.device = &device->common;
		list_add_tail(&ioat_chan->common.device_node,
			      &device->common.channels);
		device->idx[i] = ioat_chan;
		tasklet_init(&ioat_chan->cleanup_task,
			     ioat_dma_cleanup_tasklet,
			     (unsigned long) ioat_chan);
		tasklet_disable(&ioat_chan->cleanup_task);
	}
	return device->common.chancnt;
}

static void ioat_set_src(dma_addr_t addr,
			 struct dma_async_tx_descriptor *tx,
			 int index)
{
	tx_to_ioat_desc(tx)->src = addr;
}

static void ioat_set_dest(dma_addr_t addr,
			  struct dma_async_tx_descriptor *tx,
			  int index)
{
	tx_to_ioat_desc(tx)->dst = addr;
}

static dma_cookie_t ioat_tx_submit(struct dma_async_tx_descriptor *tx)
{
	struct ioat_dma_chan *ioat_chan = to_ioat_chan(tx->chan);
	struct ioat_desc_sw *first = tx_to_ioat_desc(tx);
	struct ioat_desc_sw *prev, *new;
	struct ioat_dma_descriptor *hw;
	int append = 0;
	dma_cookie_t cookie;
	LIST_HEAD(new_chain);
	u32 copy;
	size_t len;
	dma_addr_t src, dst;
	int orig_ack;
	unsigned int desc_count = 0;

	/* src and dest and len are stored in the initial descriptor */
	len = first->len;
	src = first->src;
	dst = first->dst;
	orig_ack = first->async_tx.ack;
	new = first;

	spin_lock_bh(&ioat_chan->desc_lock);
	prev = to_ioat_desc(ioat_chan->used_desc.prev);
	prefetch(prev->hw);
	do {
		copy = min((u32) len, ioat_chan->xfercap);

		new->async_tx.ack = 1;

		hw = new->hw;
		hw->size = copy;
		hw->ctl = 0;
		hw->src_addr = src;
		hw->dst_addr = dst;
		hw->next = 0;

		/* chain together the physical address list for the HW */
		wmb();
		prev->hw->next = (u64) new->async_tx.phys;

		len -= copy;
		dst += copy;
		src += copy;

		list_add_tail(&new->node, &new_chain);
		desc_count++;
		prev = new;
	} while (len && (new = ioat_dma_get_next_descriptor(ioat_chan)));

	hw->ctl = IOAT_DMA_DESCRIPTOR_CTL_CP_STS;
	if (new->async_tx.callback) {
		hw->ctl |= IOAT_DMA_DESCRIPTOR_CTL_INT_GN;
		if (first != new) {
			/* move callback into to last desc */
			new->async_tx.callback = first->async_tx.callback;
			new->async_tx.callback_param
					= first->async_tx.callback_param;
			first->async_tx.callback = NULL;
			first->async_tx.callback_param = NULL;
		}
	}

	new->tx_cnt = desc_count;
	new->async_tx.ack = orig_ack; /* client is in control of this ack */

	/* store the original values for use in later cleanup */
	if (new != first) {
		new->src = first->src;
		new->dst = first->dst;
		new->len = first->len;
	}

	/* cookie incr and addition to used_list must be atomic */
	cookie = ioat_chan->common.cookie;
	cookie++;
	if (cookie < 0)
		cookie = 1;
	ioat_chan->common.cookie = new->async_tx.cookie = cookie;

	/* write address into NextDescriptor field of last desc in chain */
	to_ioat_desc(ioat_chan->used_desc.prev)->hw->next =
							first->async_tx.phys;
	__list_splice(&new_chain, ioat_chan->used_desc.prev);

	ioat_chan->pending += desc_count;
	if (ioat_chan->pending >= 4) {
		append = 1;
		ioat_chan->pending = 0;
	}
	spin_unlock_bh(&ioat_chan->desc_lock);

	if (append)
		writeb(IOAT_CHANCMD_APPEND,
			ioat_chan->reg_base + IOAT_CHANCMD_OFFSET);

	return cookie;
}

static struct ioat_desc_sw *ioat_dma_alloc_descriptor(
					struct ioat_dma_chan *ioat_chan,
					gfp_t flags)
{
	struct ioat_dma_descriptor *desc;
	struct ioat_desc_sw *desc_sw;
	struct ioatdma_device *ioatdma_device;
	dma_addr_t phys;

	ioatdma_device = to_ioatdma_device(ioat_chan->common.device);
	desc = pci_pool_alloc(ioatdma_device->dma_pool, flags, &phys);
	if (unlikely(!desc))
		return NULL;

	desc_sw = kzalloc(sizeof(*desc_sw), flags);
	if (unlikely(!desc_sw)) {
		pci_pool_free(ioatdma_device->dma_pool, desc, phys);
		return NULL;
	}

	memset(desc, 0, sizeof(*desc));
	dma_async_tx_descriptor_init(&desc_sw->async_tx, &ioat_chan->common);
	desc_sw->async_tx.tx_set_src = ioat_set_src;
	desc_sw->async_tx.tx_set_dest = ioat_set_dest;
	desc_sw->async_tx.tx_submit = ioat_tx_submit;
	INIT_LIST_HEAD(&desc_sw->async_tx.tx_list);
	desc_sw->hw = desc;
	desc_sw->async_tx.phys = phys;

	return desc_sw;
}

/* returns the actual number of allocated descriptors */
static int ioat_dma_alloc_chan_resources(struct dma_chan *chan)
{
	struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan);
	struct ioat_desc_sw *desc = NULL;
	u16 chanctrl;
	u32 chanerr;
	int i;
	LIST_HEAD(tmp_list);

	/* have we already been set up? */
	if (!list_empty(&ioat_chan->free_desc))
		return INITIAL_IOAT_DESC_COUNT;

	/* Setup register to interrupt and write completion status on error */
	chanctrl = IOAT_CHANCTRL_ERR_INT_EN |
		IOAT_CHANCTRL_ANY_ERR_ABORT_EN |
		IOAT_CHANCTRL_ERR_COMPLETION_EN;
	writew(chanctrl, ioat_chan->reg_base + IOAT_CHANCTRL_OFFSET);

	chanerr = readl(ioat_chan->reg_base + IOAT_CHANERR_OFFSET);
	if (chanerr) {
		dev_err(&ioat_chan->device->pdev->dev,
			"CHANERR = %x, clearing\n", chanerr);
		writel(chanerr, ioat_chan->reg_base + IOAT_CHANERR_OFFSET);
	}

	/* Allocate descriptors */
	for (i = 0; i < INITIAL_IOAT_DESC_COUNT; i++) {
		desc = ioat_dma_alloc_descriptor(ioat_chan, GFP_KERNEL);
		if (!desc) {
			dev_err(&ioat_chan->device->pdev->dev,
				"Only %d initial descriptors\n", i);
			break;
		}
		list_add_tail(&desc->node, &tmp_list);
	}
	spin_lock_bh(&ioat_chan->desc_lock);
	list_splice(&tmp_list, &ioat_chan->free_desc);
	spin_unlock_bh(&ioat_chan->desc_lock);

	/* allocate a completion writeback area */
	/* doing 2 32bit writes to mmio since 1 64b write doesn't work */
	ioat_chan->completion_virt =
		pci_pool_alloc(ioat_chan->device->completion_pool,
			       GFP_KERNEL,
			       &ioat_chan->completion_addr);
	memset(ioat_chan->completion_virt, 0,
	       sizeof(*ioat_chan->completion_virt));
	writel(((u64) ioat_chan->completion_addr) & 0x00000000FFFFFFFF,
	       ioat_chan->reg_base + IOAT_CHANCMP_OFFSET_LOW);
	writel(((u64) ioat_chan->completion_addr) >> 32,
	       ioat_chan->reg_base + IOAT_CHANCMP_OFFSET_HIGH);

	tasklet_enable(&ioat_chan->cleanup_task);
	ioat_dma_start_null_desc(ioat_chan);
	return i;
}

static void ioat_dma_free_chan_resources(struct dma_chan *chan)
{
	struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan);
	struct ioatdma_device *ioatdma_device = to_ioatdma_device(chan->device);
	struct ioat_desc_sw *desc, *_desc;
	int in_use_descs = 0;

	tasklet_disable(&ioat_chan->cleanup_task);
	ioat_dma_memcpy_cleanup(ioat_chan);

	/* Delay 100ms after reset to allow internal DMA logic to quiesce
	 * before removing DMA descriptor resources.
	 */
	writeb(IOAT_CHANCMD_RESET, ioat_chan->reg_base + IOAT_CHANCMD_OFFSET);
	mdelay(100);

	spin_lock_bh(&ioat_chan->desc_lock);
	list_for_each_entry_safe(desc, _desc, &ioat_chan->used_desc, node) {
		in_use_descs++;
		list_del(&desc->node);
		pci_pool_free(ioatdma_device->dma_pool, desc->hw,
			      desc->async_tx.phys);
		kfree(desc);
	}
	list_for_each_entry_safe(desc, _desc, &ioat_chan->free_desc, node) {
		list_del(&desc->node);
		pci_pool_free(ioatdma_device->dma_pool, desc->hw,
			      desc->async_tx.phys);
		kfree(desc);
	}
	spin_unlock_bh(&ioat_chan->desc_lock);

	pci_pool_free(ioatdma_device->completion_pool,
		      ioat_chan->completion_virt,
		      ioat_chan->completion_addr);

	/* one is ok since we left it on there on purpose */
	if (in_use_descs > 1)
		dev_err(&ioat_chan->device->pdev->dev,
			"Freeing %d in use descriptors!\n",
			in_use_descs - 1);

	ioat_chan->last_completion = ioat_chan->completion_addr = 0;
	ioat_chan->pending = 0;
}

/**
 * ioat_dma_get_next_descriptor - return the next available descriptor
 * @ioat_chan: IOAT DMA channel handle
 *
 * Gets the next descriptor from the chain, and must be called with the
 * channel's desc_lock held.  Allocates more descriptors if the channel
 * has run out.
 */
static struct ioat_desc_sw *
ioat_dma_get_next_descriptor(struct ioat_dma_chan *ioat_chan)
{
	struct ioat_desc_sw *new = NULL;

	if (!list_empty(&ioat_chan->free_desc)) {
		new = to_ioat_desc(ioat_chan->free_desc.next);
		list_del(&new->node);
	} else {
		/* try to get another desc */
		new = ioat_dma_alloc_descriptor(ioat_chan, GFP_ATOMIC);
		/* will this ever happen? */
		/* TODO add upper limit on these */
		BUG_ON(!new);
	}

	prefetch(new->hw);
	return new;
}

static struct dma_async_tx_descriptor *ioat_dma_prep_memcpy(
						struct dma_chan *chan,
						size_t len,
						int int_en)
{
	struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan);
	struct ioat_desc_sw *new;

	spin_lock_bh(&ioat_chan->desc_lock);
	new = ioat_dma_get_next_descriptor(ioat_chan);
	new->len = len;
	spin_unlock_bh(&ioat_chan->desc_lock);

	return new ? &new->async_tx : NULL;
}

/**
 * ioat_dma_memcpy_issue_pending - push potentially unrecognized appended
 *                                 descriptors to hw
 * @chan: DMA channel handle
 */
static void ioat_dma_memcpy_issue_pending(struct dma_chan *chan)
{
	struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan);

	if (ioat_chan->pending != 0) {
		ioat_chan->pending = 0;
		writeb(IOAT_CHANCMD_APPEND,
		       ioat_chan->reg_base + IOAT_CHANCMD_OFFSET);
	}
}

static void ioat_dma_cleanup_tasklet(unsigned long data)
{
	struct ioat_dma_chan *chan = (void *)data;
	ioat_dma_memcpy_cleanup(chan);
	writew(IOAT_CHANCTRL_INT_DISABLE,
	       chan->reg_base + IOAT_CHANCTRL_OFFSET);
}

static void ioat_dma_memcpy_cleanup(struct ioat_dma_chan *ioat_chan)
{
	unsigned long phys_complete;
	struct ioat_desc_sw *desc, *_desc;
	dma_cookie_t cookie = 0;

	prefetch(ioat_chan->completion_virt);

	if (!spin_trylock_bh(&ioat_chan->cleanup_lock))
		return;

	/* The completion writeback can happen at any time,
	   so reads by the driver need to be atomic operations
	   The descriptor physical addresses are limited to 32-bits
	   when the CPU can only do a 32-bit mov */

#if (BITS_PER_LONG == 64)
	phys_complete =
		ioat_chan->completion_virt->full
		& IOAT_CHANSTS_COMPLETED_DESCRIPTOR_ADDR;
#else
	phys_complete =
		ioat_chan->completion_virt->low & IOAT_LOW_COMPLETION_MASK;
#endif

	if ((ioat_chan->completion_virt->full
		& IOAT_CHANSTS_DMA_TRANSFER_STATUS) ==
				IOAT_CHANSTS_DMA_TRANSFER_STATUS_HALTED) {
		dev_err(&ioat_chan->device->pdev->dev,
			"Channel halted, chanerr = %x\n",
			readl(ioat_chan->reg_base + IOAT_CHANERR_OFFSET));

		/* TODO do something to salvage the situation */
	}

	if (phys_complete == ioat_chan->last_completion) {
		spin_unlock_bh(&ioat_chan->cleanup_lock);
		return;
	}

	cookie = 0;
	spin_lock_bh(&ioat_chan->desc_lock);
	list_for_each_entry_safe(desc, _desc, &ioat_chan->used_desc, node) {

		/*
		 * Incoming DMA requests may use multiple descriptors, due to
		 * exceeding xfercap, perhaps. If so, only the last one will
		 * have a cookie, and require unmapping.
		 */
		if (desc->async_tx.cookie) {
			cookie = desc->async_tx.cookie;

			/*
			 * yes we are unmapping both _page and _single alloc'd
			 * regions with unmap_page. Is this *really* that bad?
			 */
			pci_unmap_page(ioat_chan->device->pdev,
					pci_unmap_addr(desc, dst),
					pci_unmap_len(desc, len),
					PCI_DMA_FROMDEVICE);
			pci_unmap_page(ioat_chan->device->pdev,
					pci_unmap_addr(desc, src),
					pci_unmap_len(desc, len),
					PCI_DMA_TODEVICE);
			if (desc->async_tx.callback) {
				desc->async_tx.callback(
						desc->async_tx.callback_param);
				desc->async_tx.callback = NULL;
			}
		}

		if (desc->async_tx.phys != phys_complete) {
			/*
			 * a completed entry, but not the last, so cleanup
			 * if the client is done with the descriptor
			 */
			if (desc->async_tx.ack) {
				list_del(&desc->node);
				list_add_tail(&desc->node,
					      &ioat_chan->free_desc);
			} else
				desc->async_tx.cookie = 0;
		} else {
			/*
			 * last used desc. Do not remove, so we can append from
			 * it, but don't look at it next time, either
			 */
			desc->async_tx.cookie = 0;

			/* TODO check status bits? */
			break;
		}
	}

	spin_unlock_bh(&ioat_chan->desc_lock);

	ioat_chan->last_completion = phys_complete;
	if (cookie != 0)
		ioat_chan->completed_cookie = cookie;

	spin_unlock_bh(&ioat_chan->cleanup_lock);
}

static void ioat_dma_dependency_added(struct dma_chan *chan)
{
	struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan);
	spin_lock_bh(&ioat_chan->desc_lock);
	if (ioat_chan->pending == 0) {
		spin_unlock_bh(&ioat_chan->desc_lock);
		ioat_dma_memcpy_cleanup(ioat_chan);
	} else
		spin_unlock_bh(&ioat_chan->desc_lock);
}

/**
 * ioat_dma_is_complete - poll the status of a IOAT DMA transaction
 * @chan: IOAT DMA channel handle
 * @cookie: DMA transaction identifier
 * @done: if not %NULL, updated with last completed transaction
 * @used: if not %NULL, updated with last used transaction
 */
static enum dma_status ioat_dma_is_complete(struct dma_chan *chan,
					    dma_cookie_t cookie,
					    dma_cookie_t *done,
					    dma_cookie_t *used)
{
	struct ioat_dma_chan *ioat_chan = to_ioat_chan(chan);
	dma_cookie_t last_used;
	dma_cookie_t last_complete;
	enum dma_status ret;

	last_used = chan->cookie;
	last_complete = ioat_chan->completed_cookie;

	if (done)
		*done = last_complete;
	if (used)
		*used = last_used;

	ret = dma_async_is_complete(cookie, last_complete, last_used);
	if (ret == DMA_SUCCESS)
		return ret;

	ioat_dma_memcpy_cleanup(ioat_chan);

	last_used = chan->cookie;
	last_complete = ioat_chan->completed_cookie;

	if (done)
		*done = last_complete;
	if (used)
		*used = last_used;

	return dma_async_is_complete(cookie, last_complete, last_used);
}

/* PCI API */

static void ioat_dma_start_null_desc(struct ioat_dma_chan *ioat_chan)
{
	struct ioat_desc_sw *desc;

	spin_lock_bh(&ioat_chan->desc_lock);

	desc = ioat_dma_get_next_descriptor(ioat_chan);
	desc->hw->ctl = IOAT_DMA_DESCRIPTOR_NUL
				| IOAT_DMA_DESCRIPTOR_CTL_INT_GN
				| IOAT_DMA_DESCRIPTOR_CTL_CP_STS;
	desc->hw->next = 0;
	desc->hw->size = 0;
	desc->hw->src_addr = 0;
	desc->hw->dst_addr = 0;
	desc->async_tx.ack = 1;

	list_add_tail(&desc->node, &ioat_chan->used_desc);
	spin_unlock_bh(&ioat_chan->desc_lock);

	writel(((u64) desc->async_tx.phys) & 0x00000000FFFFFFFF,
	       ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_LOW);
	writel(((u64) desc->async_tx.phys) >> 32,
	       ioat_chan->reg_base + IOAT_CHAINADDR_OFFSET_HIGH);

	writeb(IOAT_CHANCMD_START, ioat_chan->reg_base + IOAT_CHANCMD_OFFSET);
}

/*
 * Perform a IOAT transaction to verify the HW works.
 */
#define IOAT_TEST_SIZE 2000

static void ioat_dma_test_callback(void *dma_async_param)
{
	printk(KERN_ERR "ioatdma: ioat_dma_test_callback(%p)\n",
			dma_async_param);
}

/**
 * ioat_dma_self_test - Perform a IOAT transaction to verify the HW works.
 * @device: device to be tested
 */
static int ioat_dma_self_test(struct ioatdma_device *device)
{
	int i;
	u8 *src;
	u8 *dest;
	struct dma_chan *dma_chan;
	struct dma_async_tx_descriptor *tx = NULL;
	dma_addr_t addr;
	dma_cookie_t cookie;
	int err = 0;

	src = kzalloc(sizeof(u8) * IOAT_TEST_SIZE, GFP_KERNEL);
	if (!src)
		return -ENOMEM;
	dest = kzalloc(sizeof(u8) * IOAT_TEST_SIZE, GFP_KERNEL);
	if (!dest) {
		kfree(src);
		return -ENOMEM;
	}

	/* Fill in src buffer */
	for (i = 0; i < IOAT_TEST_SIZE; i++)
		src[i] = (u8)i;

	/* Start copy, using first DMA channel */
	dma_chan = container_of(device->common.channels.next,
				struct dma_chan,
				device_node);
	if (ioat_dma_alloc_chan_resources(dma_chan) < 1) {
		dev_err(&device->pdev->dev,
			"selftest cannot allocate chan resource\n");
		err = -ENODEV;
		goto out;
	}

	tx = ioat_dma_prep_memcpy(dma_chan, IOAT_TEST_SIZE, 0);
	if (!tx) {
		dev_err(&device->pdev->dev,
			"Self-test prep failed, disabling\n");
		err = -ENODEV;
		goto free_resources;
	}

	async_tx_ack(tx);
	addr = dma_map_single(dma_chan->device->dev, src, IOAT_TEST_SIZE,
			DMA_TO_DEVICE);
	ioat_set_src(addr, tx, 0);
	addr = dma_map_single(dma_chan->device->dev, dest, IOAT_TEST_SIZE,
			DMA_FROM_DEVICE);
	ioat_set_dest(addr, tx, 0);
	tx->callback = ioat_dma_test_callback;
	tx->callback_param = (void *)0x8086;
	cookie = ioat_tx_submit(tx);
	if (cookie < 0) {
		dev_err(&device->pdev->dev,
			"Self-test setup failed, disabling\n");
		err = -ENODEV;
		goto free_resources;
	}
	ioat_dma_memcpy_issue_pending(dma_chan);
	msleep(1);

	if (ioat_dma_is_complete(dma_chan, cookie, NULL, NULL) != DMA_SUCCESS) {
		dev_err(&device->pdev->dev,
			"Self-test copy timed out, disabling\n");
		err = -ENODEV;
		goto free_resources;
	}
	if (memcmp(src, dest, IOAT_TEST_SIZE)) {
		dev_err(&device->pdev->dev,
			"Self-test copy failed compare, disabling\n");
		err = -ENODEV;
		goto free_resources;
	}

free_resources:
	ioat_dma_free_chan_resources(dma_chan);
out:
	kfree(src);
	kfree(dest);
	return err;
}

static char ioat_interrupt_style[32] = "msix";
module_param_string(ioat_interrupt_style, ioat_interrupt_style,
		    sizeof(ioat_interrupt_style), 0644);
MODULE_PARM_DESC(ioat_interrupt_style,
		 "set ioat interrupt style: msix (default), "
		 "msix-single-vector, msi, intx)");

/**
 * ioat_dma_setup_interrupts - setup interrupt handler
 * @device: ioat device
 */
static int ioat_dma_setup_interrupts(struct ioatdma_device *device)
{
	struct ioat_dma_chan *ioat_chan;
	int err, i, j, msixcnt;
	u8 intrctrl = 0;

	if (!strcmp(ioat_interrupt_style, "msix"))
		goto msix;
	if (!strcmp(ioat_interrupt_style, "msix-single-vector"))
		goto msix_single_vector;
	if (!strcmp(ioat_interrupt_style, "msi"))
		goto msi;
	if (!strcmp(ioat_interrupt_style, "intx"))
		goto intx;
	dev_err(&device->pdev->dev, "invalid ioat_interrupt_style %s\n",
		ioat_interrupt_style);
	goto err_no_irq;

msix:
	/* The number of MSI-X vectors should equal the number of channels */
	msixcnt = device->common.chancnt;
	for (i = 0; i < msixcnt; i++)
		device->msix_entries[i].entry = i;

	err = pci_enable_msix(device->pdev, device->msix_entries, msixcnt);
	if (err < 0)
		goto msi;
	if (err > 0)
		goto msix_single_vector;

	for (i = 0; i < msixcnt; i++) {
		ioat_chan = ioat_lookup_chan_by_index(device, i);
		err = request_irq(device->msix_entries[i].vector,
				  ioat_dma_do_interrupt_msix,
				  0, "ioat-msix", ioat_chan);
		if (err) {
			for (j = 0; j < i; j++) {
				ioat_chan =
					ioat_lookup_chan_by_index(device, j);
				free_irq(device->msix_entries[j].vector,
					 ioat_chan);
			}
			goto msix_single_vector;
		}
	}
	intrctrl |= IOAT_INTRCTRL_MSIX_VECTOR_CONTROL;
	device->irq_mode = msix_multi_vector;
	goto done;

msix_single_vector:
	device->msix_entries[0].entry = 0;
	err = pci_enable_msix(device->pdev, device->msix_entries, 1);
	if (err)
		goto msi;

	err = request_irq(device->msix_entries[0].vector, ioat_dma_do_interrupt,
			  0, "ioat-msix", device);
	if (err) {
		pci_disable_msix(device->pdev);
		goto msi;
	}
	device->irq_mode = msix_single_vector;
	goto done;

msi:
	err = pci_enable_msi(device->pdev);
	if (err)
		goto intx;

	err = request_irq(device->pdev->irq, ioat_dma_do_interrupt,
			  0, "ioat-msi", device);
	if (err) {
		pci_disable_msi(device->pdev);
		goto intx;
	}
	/*
	 * CB 1.2 devices need a bit set in configuration space to enable MSI
	 */
	if (device->version == IOAT_VER_1_2) {
		u32 dmactrl;
		pci_read_config_dword(device->pdev,
				      IOAT_PCI_DMACTRL_OFFSET, &dmactrl);
		dmactrl |= IOAT_PCI_DMACTRL_MSI_EN;
		pci_write_config_dword(device->pdev,
				       IOAT_PCI_DMACTRL_OFFSET, dmactrl);
	}
	device->irq_mode = msi;
	goto done;

intx:
	err = request_irq(device->pdev->irq, ioat_dma_do_interrupt,
			  IRQF_SHARED, "ioat-intx", device);
	if (err)
		goto err_no_irq;
	device->irq_mode = intx;

done:
	intrctrl |= IOAT_INTRCTRL_MASTER_INT_EN;
	writeb(intrctrl, device->reg_base + IOAT_INTRCTRL_OFFSET);
	return 0;

err_no_irq:
	/* Disable all interrupt generation */
	writeb(0, device->reg_base + IOAT_INTRCTRL_OFFSET);
	dev_err(&device->pdev->dev, "no usable interrupts\n");
	device->irq_mode = none;
	return -1;
}

/**
 * ioat_dma_remove_interrupts - remove whatever interrupts were set
 * @device: ioat device
 */
static void ioat_dma_remove_interrupts(struct ioatdma_device *device)
{
	struct ioat_dma_chan *ioat_chan;
	int i;

	/* Disable all interrupt generation */
	writeb(0, device->reg_base + IOAT_INTRCTRL_OFFSET);

	switch (device->irq_mode) {
	case msix_multi_vector:
		for (i = 0; i < device->common.chancnt; i++) {
			ioat_chan = ioat_lookup_chan_by_index(device, i);
			free_irq(device->msix_entries[i].vector, ioat_chan);
		}
		pci_disable_msix(device->pdev);
		break;
	case msix_single_vector:
		free_irq(device->msix_entries[0].vector, device);
		pci_disable_msix(device->pdev);
		break;
	case msi:
		free_irq(device->pdev->irq, device);
		pci_disable_msi(device->pdev);
		break;
	case intx:
		free_irq(device->pdev->irq, device);
		break;
	case none:
		dev_warn(&device->pdev->dev,
			 "call to %s without interrupts setup\n", __func__);
	}
	device->irq_mode = none;
}

struct ioatdma_device *ioat_dma_probe(struct pci_dev *pdev,
				      void __iomem *iobase)
{
	int err;
	struct ioatdma_device *device;

	device = kzalloc(sizeof(*device), GFP_KERNEL);
	if (!device) {
		err = -ENOMEM;
		goto err_kzalloc;
	}
	device->pdev = pdev;
	device->reg_base = iobase;
	device->version = readb(device->reg_base + IOAT_VER_OFFSET);

	/* DMA coherent memory pool for DMA descriptor allocations */
	device->dma_pool = pci_pool_create("dma_desc_pool", pdev,
					   sizeof(struct ioat_dma_descriptor),
					   64, 0);
	if (!device->dma_pool) {
		err = -ENOMEM;
		goto err_dma_pool;
	}

	device->completion_pool = pci_pool_create("completion_pool", pdev,
						  sizeof(u64), SMP_CACHE_BYTES,
						  SMP_CACHE_BYTES);
	if (!device->completion_pool) {
		err = -ENOMEM;
		goto err_completion_pool;
	}

	INIT_LIST_HEAD(&device->common.channels);
	ioat_dma_enumerate_channels(device);

	dma_cap_set(DMA_MEMCPY, device->common.cap_mask);
	device->common.device_alloc_chan_resources =
						ioat_dma_alloc_chan_resources;
	device->common.device_free_chan_resources =
						ioat_dma_free_chan_resources;
	device->common.device_prep_dma_memcpy = ioat_dma_prep_memcpy;
	device->common.device_is_tx_complete = ioat_dma_is_complete;
	device->common.device_issue_pending = ioat_dma_memcpy_issue_pending;
	device->common.device_dependency_added = ioat_dma_dependency_added;
	device->common.dev = &pdev->dev;
	dev_err(&device->pdev->dev,
		"Intel(R) I/OAT DMA Engine found,"
		" %d channels, device version 0x%02x, driver version %s\n",
		device->common.chancnt, device->version, IOAT_DMA_VERSION);

	err = ioat_dma_setup_interrupts(device);
	if (err)
		goto err_setup_interrupts;

	err = ioat_dma_self_test(device);
	if (err)
		goto err_self_test;

	dma_async_device_register(&device->common);

	return device;

err_self_test:
	ioat_dma_remove_interrupts(device);
err_setup_interrupts:
	pci_pool_destroy(device->completion_pool);
err_completion_pool:
	pci_pool_destroy(device->dma_pool);
err_dma_pool:
	kfree(device);
err_kzalloc:
	dev_err(&device->pdev->dev,
		"Intel(R) I/OAT DMA Engine initialization failed\n");
	return NULL;
}

void ioat_dma_remove(struct ioatdma_device *device)
{
	struct dma_chan *chan, *_chan;
	struct ioat_dma_chan *ioat_chan;

	ioat_dma_remove_interrupts(device);

	dma_async_device_unregister(&device->common);

	pci_pool_destroy(device->dma_pool);
	pci_pool_destroy(device->completion_pool);

	iounmap(device->reg_base);
	pci_release_regions(device->pdev);
	pci_disable_device(device->pdev);

	list_for_each_entry_safe(chan, _chan,
				 &device->common.channels, device_node) {
		ioat_chan = to_ioat_chan(chan);
		list_del(&chan->device_node);
		kfree(ioat_chan);
	}
	kfree(device);
}