aboutsummaryrefslogtreecommitdiff
path: root/arch/sparc64/kernel/iommu.c
blob: 070a4846c0cb7d7dcb06be192bcd8879733f5112 (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
/* iommu.c: Generic sparc64 IOMMU support.
 *
 * Copyright (C) 1999, 2007 David S. Miller (davem@davemloft.net)
 * Copyright (C) 1999, 2000 Jakub Jelinek (jakub@redhat.com)
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/delay.h>
#include <linux/device.h>
#include <linux/dma-mapping.h>
#include <linux/errno.h>

#ifdef CONFIG_PCI
#include <linux/pci.h>
#endif

#include <asm/iommu.h>

#include "iommu_common.h"

#define STC_CTXMATCH_ADDR(STC, CTX)	\
	((STC)->strbuf_ctxmatch_base + ((CTX) << 3))
#define STC_FLUSHFLAG_INIT(STC) \
	(*((STC)->strbuf_flushflag) = 0UL)
#define STC_FLUSHFLAG_SET(STC) \
	(*((STC)->strbuf_flushflag) != 0UL)

#define iommu_read(__reg) \
({	u64 __ret; \
	__asm__ __volatile__("ldxa [%1] %2, %0" \
			     : "=r" (__ret) \
			     : "r" (__reg), "i" (ASI_PHYS_BYPASS_EC_E) \
			     : "memory"); \
	__ret; \
})
#define iommu_write(__reg, __val) \
	__asm__ __volatile__("stxa %0, [%1] %2" \
			     : /* no outputs */ \
			     : "r" (__val), "r" (__reg), \
			       "i" (ASI_PHYS_BYPASS_EC_E))

/* Must be invoked under the IOMMU lock. */
static void __iommu_flushall(struct iommu *iommu)
{
	if (iommu->iommu_flushinv) {
		iommu_write(iommu->iommu_flushinv, ~(u64)0);
	} else {
		unsigned long tag;
		int entry;

		tag = iommu->iommu_tags;
		for (entry = 0; entry < 16; entry++) {
			iommu_write(tag, 0);
			tag += 8;
		}

		/* Ensure completion of previous PIO writes. */
		(void) iommu_read(iommu->write_complete_reg);
	}
}

#define IOPTE_CONSISTENT(CTX) \
	(IOPTE_VALID | IOPTE_CACHE | \
	 (((CTX) << 47) & IOPTE_CONTEXT))

#define IOPTE_STREAMING(CTX) \
	(IOPTE_CONSISTENT(CTX) | IOPTE_STBUF)

/* Existing mappings are never marked invalid, instead they
 * are pointed to a dummy page.
 */
#define IOPTE_IS_DUMMY(iommu, iopte)	\
	((iopte_val(*iopte) & IOPTE_PAGE) == (iommu)->dummy_page_pa)

static inline void iopte_make_dummy(struct iommu *iommu, iopte_t *iopte)
{
	unsigned long val = iopte_val(*iopte);

	val &= ~IOPTE_PAGE;
	val |= iommu->dummy_page_pa;

	iopte_val(*iopte) = val;
}

/* Based largely upon the ppc64 iommu allocator.  */
static long arena_alloc(struct iommu *iommu, unsigned long npages)
{
	struct iommu_arena *arena = &iommu->arena;
	unsigned long n, i, start, end, limit;
	int pass;

	limit = arena->limit;
	start = arena->hint;
	pass = 0;

again:
	n = find_next_zero_bit(arena->map, limit, start);
	end = n + npages;
	if (unlikely(end >= limit)) {
		if (likely(pass < 1)) {
			limit = start;
			start = 0;
			__iommu_flushall(iommu);
			pass++;
			goto again;
		} else {
			/* Scanned the whole thing, give up. */
			return -1;
		}
	}

	for (i = n; i < end; i++) {
		if (test_bit(i, arena->map)) {
			start = i + 1;
			goto again;
		}
	}

	for (i = n; i < end; i++)
		__set_bit(i, arena->map);

	arena->hint = end;

	return n;
}

static void arena_free(struct iommu_arena *arena, unsigned long base, unsigned long npages)
{
	unsigned long i;

	for (i = base; i < (base + npages); i++)
		__clear_bit(i, arena->map);
}

int iommu_table_init(struct iommu *iommu, int tsbsize,
		     u32 dma_offset, u32 dma_addr_mask)
{
	unsigned long i, tsbbase, order, sz, num_tsb_entries;

	num_tsb_entries = tsbsize / sizeof(iopte_t);

	/* Setup initial software IOMMU state. */
	spin_lock_init(&iommu->lock);
	iommu->ctx_lowest_free = 1;
	iommu->page_table_map_base = dma_offset;
	iommu->dma_addr_mask = dma_addr_mask;

	/* Allocate and initialize the free area map.  */
	sz = num_tsb_entries / 8;
	sz = (sz + 7UL) & ~7UL;
	iommu->arena.map = kzalloc(sz, GFP_KERNEL);
	if (!iommu->arena.map) {
		printk(KERN_ERR "IOMMU: Error, kmalloc(arena.map) failed.\n");
		return -ENOMEM;
	}
	iommu->arena.limit = num_tsb_entries;

	/* Allocate and initialize the dummy page which we
	 * set inactive IO PTEs to point to.
	 */
	iommu->dummy_page = __get_free_pages(GFP_KERNEL, 0);
	if (!iommu->dummy_page) {
		printk(KERN_ERR "IOMMU: Error, gfp(dummy_page) failed.\n");
		goto out_free_map;
	}
	memset((void *)iommu->dummy_page, 0, PAGE_SIZE);
	iommu->dummy_page_pa = (unsigned long) __pa(iommu->dummy_page);

	/* Now allocate and setup the IOMMU page table itself.  */
	order = get_order(tsbsize);
	tsbbase = __get_free_pages(GFP_KERNEL, order);
	if (!tsbbase) {
		printk(KERN_ERR "IOMMU: Error, gfp(tsb) failed.\n");
		goto out_free_dummy_page;
	}
	iommu->page_table = (iopte_t *)tsbbase;

	for (i = 0; i < num_tsb_entries; i++)
		iopte_make_dummy(iommu, &iommu->page_table[i]);

	return 0;

out_free_dummy_page:
	free_page(iommu->dummy_page);
	iommu->dummy_page = 0UL;

out_free_map:
	kfree(iommu->arena.map);
	iommu->arena.map = NULL;

	return -ENOMEM;
}

static inline iopte_t *alloc_npages(struct iommu *iommu, unsigned long npages)
{
	long entry;

	entry = arena_alloc(iommu, npages);
	if (unlikely(entry < 0))
		return NULL;

	return iommu->page_table + entry;
}

static inline void free_npages(struct iommu *iommu, dma_addr_t base, unsigned long npages)
{
	arena_free(&iommu->arena, base >> IO_PAGE_SHIFT, npages);
}

static int iommu_alloc_ctx(struct iommu *iommu)
{
	int lowest = iommu->ctx_lowest_free;
	int sz = IOMMU_NUM_CTXS - lowest;
	int n = find_next_zero_bit(iommu->ctx_bitmap, sz, lowest);

	if (unlikely(n == sz)) {
		n = find_next_zero_bit(iommu->ctx_bitmap, lowest, 1);
		if (unlikely(n == lowest)) {
			printk(KERN_WARNING "IOMMU: Ran out of contexts.\n");
			n = 0;
		}
	}
	if (n)
		__set_bit(n, iommu->ctx_bitmap);

	return n;
}

static inline void iommu_free_ctx(struct iommu *iommu, int ctx)
{
	if (likely(ctx)) {
		__clear_bit(ctx, iommu->ctx_bitmap);
		if (ctx < iommu->ctx_lowest_free)
			iommu->ctx_lowest_free = ctx;
	}
}

static void *dma_4u_alloc_coherent(struct device *dev, size_t size,
				   dma_addr_t *dma_addrp, gfp_t gfp)
{
	struct iommu *iommu;
	iopte_t *iopte;
	unsigned long flags, order, first_page;
	void *ret;
	int npages;

	size = IO_PAGE_ALIGN(size);
	order = get_order(size);
	if (order >= 10)
		return NULL;

	first_page = __get_free_pages(gfp, order);
	if (first_page == 0UL)
		return NULL;
	memset((char *)first_page, 0, PAGE_SIZE << order);

	iommu = dev->archdata.iommu;

	spin_lock_irqsave(&iommu->lock, flags);
	iopte = alloc_npages(iommu, size >> IO_PAGE_SHIFT);
	spin_unlock_irqrestore(&iommu->lock, flags);

	if (unlikely(iopte == NULL)) {
		free_pages(first_page, order);
		return NULL;
	}

	*dma_addrp = (iommu->page_table_map_base +
		      ((iopte - iommu->page_table) << IO_PAGE_SHIFT));
	ret = (void *) first_page;
	npages = size >> IO_PAGE_SHIFT;
	first_page = __pa(first_page);
	while (npages--) {
		iopte_val(*iopte) = (IOPTE_CONSISTENT(0UL) |
				     IOPTE_WRITE |
				     (first_page & IOPTE_PAGE));
		iopte++;
		first_page += IO_PAGE_SIZE;
	}

	return ret;
}

static void dma_4u_free_coherent(struct device *dev, size_t size,
				 void *cpu, dma_addr_t dvma)
{
	struct iommu *iommu;
	iopte_t *iopte;
	unsigned long flags, order, npages;

	npages = IO_PAGE_ALIGN(size) >> IO_PAGE_SHIFT;
	iommu = dev->archdata.iommu;
	iopte = iommu->page_table +
		((dvma - iommu->page_table_map_base) >> IO_PAGE_SHIFT);

	spin_lock_irqsave(&iommu->lock, flags);

	free_npages(iommu, dvma - iommu->page_table_map_base, npages);

	spin_unlock_irqrestore(&iommu->lock, flags);

	order = get_order(size);
	if (order < 10)
		free_pages((unsigned long)cpu, order);
}

static dma_addr_t dma_4u_map_single(struct device *dev, void *ptr, size_t sz,
				    enum dma_data_direction direction)
{
	struct iommu *iommu;
	struct strbuf *strbuf;
	iopte_t *base;
	unsigned long flags, npages, oaddr;
	unsigned long i, base_paddr, ctx;
	u32 bus_addr, ret;
	unsigned long iopte_protection;

	iommu = dev->archdata.iommu;
	strbuf = dev->archdata.stc;

	if (unlikely(direction == DMA_NONE))
		goto bad_no_ctx;

	oaddr = (unsigned long)ptr;
	npages = IO_PAGE_ALIGN(oaddr + sz) - (oaddr & IO_PAGE_MASK);
	npages >>= IO_PAGE_SHIFT;

	spin_lock_irqsave(&iommu->lock, flags);
	base = alloc_npages(iommu, npages);
	ctx = 0;
	if (iommu->iommu_ctxflush)
		ctx = iommu_alloc_ctx(iommu);
	spin_unlock_irqrestore(&iommu->lock, flags);

	if (unlikely(!base))
		goto bad;

	bus_addr = (iommu->page_table_map_base +
		    ((base - iommu->page_table) << IO_PAGE_SHIFT));
	ret = bus_addr | (oaddr & ~IO_PAGE_MASK);
	base_paddr = __pa(oaddr & IO_PAGE_MASK);
	if (strbuf->strbuf_enabled)
		iopte_protection = IOPTE_STREAMING(ctx);
	else
		iopte_protection = IOPTE_CONSISTENT(ctx);
	if (direction != DMA_TO_DEVICE)
		iopte_protection |= IOPTE_WRITE;

	for (i = 0; i < npages; i++, base++, base_paddr += IO_PAGE_SIZE)
		iopte_val(*base) = iopte_protection | base_paddr;

	return ret;

bad:
	iommu_free_ctx(iommu, ctx);
bad_no_ctx:
	if (printk_ratelimit())
		WARN_ON(1);
	return DMA_ERROR_CODE;
}

static void strbuf_flush(struct strbuf *strbuf, struct iommu *iommu,
			 u32 vaddr, unsigned long ctx, unsigned long npages,
			 enum dma_data_direction direction)
{
	int limit;

	if (strbuf->strbuf_ctxflush &&
	    iommu->iommu_ctxflush) {
		unsigned long matchreg, flushreg;
		u64 val;

		flushreg = strbuf->strbuf_ctxflush;
		matchreg = STC_CTXMATCH_ADDR(strbuf, ctx);

		iommu_write(flushreg, ctx);
		val = iommu_read(matchreg);
		val &= 0xffff;
		if (!val)
			goto do_flush_sync;

		while (val) {
			if (val & 0x1)
				iommu_write(flushreg, ctx);
			val >>= 1;
		}
		val = iommu_read(matchreg);
		if (unlikely(val)) {
			printk(KERN_WARNING "strbuf_flush: ctx flush "
			       "timeout matchreg[%lx] ctx[%lx]\n",
			       val, ctx);
			goto do_page_flush;
		}
	} else {
		unsigned long i;

	do_page_flush:
		for (i = 0; i < npages; i++, vaddr += IO_PAGE_SIZE)
			iommu_write(strbuf->strbuf_pflush, vaddr);
	}

do_flush_sync:
	/* If the device could not have possibly put dirty data into
	 * the streaming cache, no flush-flag synchronization needs
	 * to be performed.
	 */
	if (direction == DMA_TO_DEVICE)
		return;

	STC_FLUSHFLAG_INIT(strbuf);
	iommu_write(strbuf->strbuf_fsync, strbuf->strbuf_flushflag_pa);
	(void) iommu_read(iommu->write_complete_reg);

	limit = 100000;
	while (!STC_FLUSHFLAG_SET(strbuf)) {
		limit--;
		if (!limit)
			break;
		udelay(1);
		rmb();
	}
	if (!limit)
		printk(KERN_WARNING "strbuf_flush: flushflag timeout "
		       "vaddr[%08x] ctx[%lx] npages[%ld]\n",
		       vaddr, ctx, npages);
}

static void dma_4u_unmap_single(struct device *dev, dma_addr_t bus_addr,
				size_t sz, enum dma_data_direction direction)
{
	struct iommu *iommu;
	struct strbuf *strbuf;
	iopte_t *base;
	unsigned long flags, npages, ctx, i;

	if (unlikely(direction == DMA_NONE)) {
		if (printk_ratelimit())
			WARN_ON(1);
		return;
	}

	iommu = dev->archdata.iommu;
	strbuf = dev->archdata.stc;

	npages = IO_PAGE_ALIGN(bus_addr + sz) - (bus_addr & IO_PAGE_MASK);
	npages >>= IO_PAGE_SHIFT;
	base = iommu->page_table +
		((bus_addr - iommu->page_table_map_base) >> IO_PAGE_SHIFT);
	bus_addr &= IO_PAGE_MASK;

	spin_lock_irqsave(&iommu->lock, flags);

	/* Record the context, if any. */
	ctx = 0;
	if (iommu->iommu_ctxflush)
		ctx = (iopte_val(*base) & IOPTE_CONTEXT) >> 47UL;

	/* Step 1: Kick data out of streaming buffers if necessary. */
	if (strbuf->strbuf_enabled)
		strbuf_flush(strbuf, iommu, bus_addr, ctx,
			     npages, direction);

	/* Step 2: Clear out TSB entries. */
	for (i = 0; i < npages; i++)
		iopte_make_dummy(iommu, base + i);

	free_npages(iommu, bus_addr - iommu->page_table_map_base, npages);

	iommu_free_ctx(iommu, ctx);

	spin_unlock_irqrestore(&iommu->lock, flags);
}

#define SG_ENT_PHYS_ADDRESS(SG)	(__pa(sg_virt((SG))))

static void fill_sg(iopte_t *iopte, struct scatterlist *sg,
		    int nused, int nelems,
		    unsigned long iopte_protection)
{
	struct scatterlist *dma_sg = sg;
	int i;

	for (i = 0; i < nused; i++) {
		unsigned long pteval = ~0UL;
		u32 dma_npages;

		dma_npages = ((dma_sg->dma_address & (IO_PAGE_SIZE - 1UL)) +
			      dma_sg->dma_length +
			      ((IO_PAGE_SIZE - 1UL))) >> IO_PAGE_SHIFT;
		do {
			unsigned long offset;
			signed int len;

			/* If we are here, we know we have at least one
			 * more page to map.  So walk forward until we
			 * hit a page crossing, and begin creating new
			 * mappings from that spot.
			 */
			for (;;) {
				unsigned long tmp;

				tmp = SG_ENT_PHYS_ADDRESS(sg);
				len = sg->length;
				if (((tmp ^ pteval) >> IO_PAGE_SHIFT) != 0UL) {
					pteval = tmp & IO_PAGE_MASK;
					offset = tmp & (IO_PAGE_SIZE - 1UL);
					break;
				}
				if (((tmp ^ (tmp + len - 1UL)) >> IO_PAGE_SHIFT) != 0UL) {
					pteval = (tmp + IO_PAGE_SIZE) & IO_PAGE_MASK;
					offset = 0UL;
					len -= (IO_PAGE_SIZE - (tmp & (IO_PAGE_SIZE - 1UL)));
					break;
				}
				sg = sg_next(sg);
				nelems--;
			}

			pteval = iopte_protection | (pteval & IOPTE_PAGE);
			while (len > 0) {
				*iopte++ = __iopte(pteval);
				pteval += IO_PAGE_SIZE;
				len -= (IO_PAGE_SIZE - offset);
				offset = 0;
				dma_npages--;
			}

			pteval = (pteval & IOPTE_PAGE) + len;
			sg = sg_next(sg);
			nelems--;

			/* Skip over any tail mappings we've fully mapped,
			 * adjusting pteval along the way.  Stop when we
			 * detect a page crossing event.
			 */
			while (nelems &&
			       (pteval << (64 - IO_PAGE_SHIFT)) != 0UL &&
			       (pteval == SG_ENT_PHYS_ADDRESS(sg)) &&
			       ((pteval ^
				 (SG_ENT_PHYS_ADDRESS(sg) + sg->length - 1UL)) >> IO_PAGE_SHIFT) == 0UL) {
				pteval += sg->length;
				sg = sg_next(sg);
				nelems--;
			}
			if ((pteval << (64 - IO_PAGE_SHIFT)) == 0UL)
				pteval = ~0UL;
		} while (dma_npages != 0);
		dma_sg = sg_next(dma_sg);
	}
}

static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
			 int nelems, enum dma_data_direction direction)
{
	struct iommu *iommu;
	struct strbuf *strbuf;
	unsigned long flags, ctx, npages, iopte_protection;
	iopte_t *base;
	u32 dma_base;
	struct scatterlist *sgtmp;
	int used;

	/* Fast path single entry scatterlists. */
	if (nelems == 1) {
		sglist->dma_address =
			dma_4u_map_single(dev, sg_virt(sglist),
					  sglist->length, direction);
		if (unlikely(sglist->dma_address == DMA_ERROR_CODE))
			return 0;
		sglist->dma_length = sglist->length;
		return 1;
	}

	iommu = dev->archdata.iommu;
	strbuf = dev->archdata.stc;

	if (unlikely(direction == DMA_NONE))
		goto bad_no_ctx;

	/* Step 1: Prepare scatter list. */

	npages = prepare_sg(sglist, nelems);

	/* Step 2: Allocate a cluster and context, if necessary. */

	spin_lock_irqsave(&iommu->lock, flags);

	base = alloc_npages(iommu, npages);
	ctx = 0;
	if (iommu->iommu_ctxflush)
		ctx = iommu_alloc_ctx(iommu);

	spin_unlock_irqrestore(&iommu->lock, flags);

	if (base == NULL)
		goto bad;

	dma_base = iommu->page_table_map_base +
		((base - iommu->page_table) << IO_PAGE_SHIFT);

	/* Step 3: Normalize DMA addresses. */
	used = nelems;

	sgtmp = sglist;
	while (used && sgtmp->dma_length) {
		sgtmp->dma_address += dma_base;
		sgtmp = sg_next(sgtmp);
		used--;
	}
	used = nelems - used;

	/* Step 4: Create the mappings. */
	if (strbuf->strbuf_enabled)
		iopte_protection = IOPTE_STREAMING(ctx);
	else
		iopte_protection = IOPTE_CONSISTENT(ctx);
	if (direction != DMA_TO_DEVICE)
		iopte_protection |= IOPTE_WRITE;

	fill_sg(base, sglist, used, nelems, iopte_protection);

#ifdef VERIFY_SG
	verify_sglist(sglist, nelems, base, npages);
#endif

	return used;

bad:
	iommu_free_ctx(iommu, ctx);
bad_no_ctx:
	if (printk_ratelimit())
		WARN_ON(1);
	return 0;
}

static void dma_4u_unmap_sg(struct device *dev, struct scatterlist *sglist,
			    int nelems, enum dma_data_direction direction)
{
	struct iommu *iommu;
	struct strbuf *strbuf;
	iopte_t *base;
	unsigned long flags, ctx, i, npages;
	struct scatterlist *sg, *sgprv;
	u32 bus_addr;

	if (unlikely(direction == DMA_NONE)) {
		if (printk_ratelimit())
			WARN_ON(1);
	}

	iommu = dev->archdata.iommu;
	strbuf = dev->archdata.stc;

	bus_addr = sglist->dma_address & IO_PAGE_MASK;

	sgprv = NULL;
	for_each_sg(sglist, sg, nelems, i) {
		if (sg->dma_length == 0)
			break;
		sgprv = sg;
	}

	npages = (IO_PAGE_ALIGN(sgprv->dma_address + sgprv->dma_length) -
		  bus_addr) >> IO_PAGE_SHIFT;

	base = iommu->page_table +
		((bus_addr - iommu->page_table_map_base) >> IO_PAGE_SHIFT);

	spin_lock_irqsave(&iommu->lock, flags);

	/* Record the context, if any. */
	ctx = 0;
	if (iommu->iommu_ctxflush)
		ctx = (iopte_val(*base) & IOPTE_CONTEXT) >> 47UL;

	/* Step 1: Kick data out of streaming buffers if necessary. */
	if (strbuf->strbuf_enabled)
		strbuf_flush(strbuf, iommu, bus_addr, ctx, npages, direction);

	/* Step 2: Clear out the TSB entries. */
	for (i = 0; i < npages; i++)
		iopte_make_dummy(iommu, base + i);

	free_npages(iommu, bus_addr - iommu->page_table_map_base, npages);

	iommu_free_ctx(iommu, ctx);

	spin_unlock_irqrestore(&iommu->lock, flags);
}

static void dma_4u_sync_single_for_cpu(struct device *dev,
				       dma_addr_t bus_addr, size_t sz,
				       enum dma_data_direction direction)
{
	struct iommu *iommu;
	struct strbuf *strbuf;
	unsigned long flags, ctx, npages;

	iommu = dev->archdata.iommu;
	strbuf = dev->archdata.stc;

	if (!strbuf->strbuf_enabled)
		return;

	spin_lock_irqsave(&iommu->lock, flags);

	npages = IO_PAGE_ALIGN(bus_addr + sz) - (bus_addr & IO_PAGE_MASK);
	npages >>= IO_PAGE_SHIFT;
	bus_addr &= IO_PAGE_MASK;

	/* Step 1: Record the context, if any. */
	ctx = 0;
	if (iommu->iommu_ctxflush &&
	    strbuf->strbuf_ctxflush) {
		iopte_t *iopte;

		iopte = iommu->page_table +
			((bus_addr - iommu->page_table_map_base)>>IO_PAGE_SHIFT);
		ctx = (iopte_val(*iopte) & IOPTE_CONTEXT) >> 47UL;
	}

	/* Step 2: Kick data out of streaming buffers. */
	strbuf_flush(strbuf, iommu, bus_addr, ctx, npages, direction);

	spin_unlock_irqrestore(&iommu->lock, flags);
}

static void dma_4u_sync_sg_for_cpu(struct device *dev,
				   struct scatterlist *sglist, int nelems,
				   enum dma_data_direction direction)
{
	struct iommu *iommu;
	struct strbuf *strbuf;
	unsigned long flags, ctx, npages, i;
	struct scatterlist *sg, *sgprv;
	u32 bus_addr;

	iommu = dev->archdata.iommu;
	strbuf = dev->archdata.stc;

	if (!strbuf->strbuf_enabled)
		return;

	spin_lock_irqsave(&iommu->lock, flags);

	/* Step 1: Record the context, if any. */
	ctx = 0;
	if (iommu->iommu_ctxflush &&
	    strbuf->strbuf_ctxflush) {
		iopte_t *iopte;

		iopte = iommu->page_table +
			((sglist[0].dma_address - iommu->page_table_map_base) >> IO_PAGE_SHIFT);
		ctx = (iopte_val(*iopte) & IOPTE_CONTEXT) >> 47UL;
	}

	/* Step 2: Kick data out of streaming buffers. */
	bus_addr = sglist[0].dma_address & IO_PAGE_MASK;
	sgprv = NULL;
	for_each_sg(sglist, sg, nelems, i) {
		if (sg->dma_length == 0)
			break;
		sgprv = sg;
	}

	npages = (IO_PAGE_ALIGN(sgprv->dma_address + sgprv->dma_length)
		  - bus_addr) >> IO_PAGE_SHIFT;
	strbuf_flush(strbuf, iommu, bus_addr, ctx, npages, direction);

	spin_unlock_irqrestore(&iommu->lock, flags);
}

const struct dma_ops sun4u_dma_ops = {
	.alloc_coherent		= dma_4u_alloc_coherent,
	.free_coherent		= dma_4u_free_coherent,
	.map_single		= dma_4u_map_single,
	.unmap_single		= dma_4u_unmap_single,
	.map_sg			= dma_4u_map_sg,
	.unmap_sg		= dma_4u_unmap_sg,
	.sync_single_for_cpu	= dma_4u_sync_single_for_cpu,
	.sync_sg_for_cpu	= dma_4u_sync_sg_for_cpu,
};

const struct dma_ops *dma_ops = &sun4u_dma_ops;
EXPORT_SYMBOL(dma_ops);

int dma_supported(struct device *dev, u64 device_mask)
{
	struct iommu *iommu = dev->archdata.iommu;
	u64 dma_addr_mask = iommu->dma_addr_mask;

	if (device_mask >= (1UL << 32UL))
		return 0;

	if ((device_mask & dma_addr_mask) == dma_addr_mask)
		return 1;

#ifdef CONFIG_PCI
	if (dev->bus == &pci_bus_type)
		return pci_dma_supported(to_pci_dev(dev), device_mask);
#endif

	return 0;
}
EXPORT_SYMBOL(dma_supported);

int dma_set_mask(struct device *dev, u64 dma_mask)
{
#ifdef CONFIG_PCI
	if (dev->bus == &pci_bus_type)
		return pci_set_dma_mask(to_pci_dev(dev), dma_mask);
#endif
	return -EINVAL;
}
EXPORT_SYMBOL(dma_set_mask);