aboutsummaryrefslogtreecommitdiffstats
path: root/arch/blackfin/kernel/bfin_dma.c
blob: 4a32f2dd5ddc39120bb66c1c49e4bc49c630c940 (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
/*
 * bfin_dma.c - Blackfin DMA implementation
 *
 * Copyright 2004-2008 Analog Devices Inc.
 *
 * Licensed under the GPL-2 or later.
 */

#include <linux/errno.h>
#include <linux/interrupt.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/param.h>
#include <linux/proc_fs.h>
#include <linux/sched.h>
#include <linux/seq_file.h>
#include <linux/spinlock.h>

#include <asm/blackfin.h>
#include <asm/cacheflush.h>
#include <asm/dma.h>
#include <asm/uaccess.h>
#include <asm/early_printk.h>

/*
 * To make sure we work around 05000119 - we always check DMA_DONE bit,
 * never the DMA_RUN bit
 */

struct dma_channel dma_ch[MAX_DMA_CHANNELS];
EXPORT_SYMBOL(dma_ch);

static int __init blackfin_dma_init(void)
{
	int i;

	printk(KERN_INFO "Blackfin DMA Controller\n");


#if ANOMALY_05000480
	bfin_write_DMAC_TC_PER(0x0111);
#endif

	for (i = 0; i < MAX_DMA_CHANNELS; i++) {
		atomic_set(&dma_ch[i].chan_status, 0);
		dma_ch[i].regs = dma_io_base_addr[i];
	}
#if defined(CH_MEM_STREAM3_SRC) && defined(CONFIG_BF60x)
	/* Mark MEMDMA Channel 3 as requested since we're using it internally */
	request_dma(CH_MEM_STREAM3_DEST, "Blackfin dma_memcpy");
	request_dma(CH_MEM_STREAM3_SRC, "Blackfin dma_memcpy");
#else
	/* Mark MEMDMA Channel 0 as requested since we're using it internally */
	request_dma(CH_MEM_STREAM0_DEST, "Blackfin dma_memcpy");
	request_dma(CH_MEM_STREAM0_SRC, "Blackfin dma_memcpy");
#endif

#if defined(CONFIG_DEB_DMA_URGENT)
	bfin_write_EBIU_DDRQUE(bfin_read_EBIU_DDRQUE()
			 | DEB1_URGENT | DEB2_URGENT | DEB3_URGENT);
#endif

	return 0;
}
arch_initcall(blackfin_dma_init);

#ifdef CONFIG_PROC_FS
static int proc_dma_show(struct seq_file *m, void *v)
{
	int i;

	for (i = 0; i < MAX_DMA_CHANNELS; ++i)
		if (dma_channel_active(i))
			seq_printf(m, "%2d: %s\n", i, dma_ch[i].device_id);

	return 0;
}

static int proc_dma_open(struct inode *inode, struct file *file)
{
	return single_open(file, proc_dma_show, NULL);
}

static const struct file_operations proc_dma_operations = {
	.open		= proc_dma_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};

static int __init proc_dma_init(void)
{
	proc_create("dma", 0, NULL, &proc_dma_operations);
	return 0;
}
late_initcall(proc_dma_init);
#endif

static void set_dma_peripheral_map(unsigned int channel, const char *device_id)
{
#ifdef CONFIG_BF54x
	unsigned int per_map;

	switch (channel) {
		case CH_UART2_RX: per_map = 0xC << 12; break;
		case CH_UART2_TX: per_map = 0xD << 12; break;
		case CH_UART3_RX: per_map = 0xE << 12; break;
		case CH_UART3_TX: per_map = 0xF << 12; break;
		default:          return;
	}

	if (strncmp(device_id, "BFIN_UART", 9) == 0)
		dma_ch[channel].regs->peripheral_map = per_map;
#endif
}

/**
 *	request_dma - request a DMA channel
 *
 * Request the specific DMA channel from the system if it's available.
 */
int request_dma(unsigned int channel, const char *device_id)
{
	pr_debug("request_dma() : BEGIN\n");

	if (device_id == NULL)
		printk(KERN_WARNING "request_dma(%u): no device_id given\n", channel);

#if defined(CONFIG_BF561) && ANOMALY_05000182
	if (channel >= CH_IMEM_STREAM0_DEST && channel <= CH_IMEM_STREAM1_DEST) {
		if (get_cclk() > 500000000) {
			printk(KERN_WARNING
			       "Request IMDMA failed due to ANOMALY 05000182\n");
			return -EFAULT;
		}
	}
#endif

	if (atomic_cmpxchg(&dma_ch[channel].chan_status, 0, 1)) {
		pr_debug("DMA CHANNEL IN USE\n");
		return -EBUSY;
	}

	set_dma_peripheral_map(channel, device_id);
	dma_ch[channel].device_id = device_id;
	dma_ch[channel].irq = 0;

	/* This is to be enabled by putting a restriction -
	 * you have to request DMA, before doing any operations on
	 * descriptor/channel
	 */
	pr_debug("request_dma() : END\n");
	return 0;
}
EXPORT_SYMBOL(request_dma);

int set_dma_callback(unsigned int channel, irq_handler_t callback, void *data)
{
	int ret;
	unsigned int irq;

	BUG_ON(channel >= MAX_DMA_CHANNELS || !callback ||
			!atomic_read(&dma_ch[channel].chan_status));

	irq = channel2irq(channel);
	ret = request_irq(irq, callback, 0, dma_ch[channel].device_id, data);
	if (ret)
		return ret;

	dma_ch[channel].irq = irq;
	dma_ch[channel].data = data;

	return 0;
}
EXPORT_SYMBOL(set_dma_callback);

/**
 *	clear_dma_buffer - clear DMA fifos for specified channel
 *
 * Set the Buffer Clear bit in the Configuration register of specific DMA
 * channel. This will stop the descriptor based DMA operation.
 */
static void clear_dma_buffer(unsigned int channel)
{
	dma_ch[channel].regs->cfg |= RESTART;
	SSYNC();
	dma_ch[channel].regs->cfg &= ~RESTART;
}

void free_dma(unsigned int channel)
{
	pr_debug("freedma() : BEGIN\n");
	BUG_ON(channel >= MAX_DMA_CHANNELS ||
			!atomic_read(&dma_ch[channel].chan_status));

	/* Halt the DMA */
	disable_dma(channel);
	clear_dma_buffer(channel);

	if (dma_ch[channel].irq)
		free_irq(dma_ch[channel].irq, dma_ch[channel].data);

	/* Clear the DMA Variable in the Channel */
	atomic_set(&dma_ch[channel].chan_status, 0);

	pr_debug("freedma() : END\n");
}
EXPORT_SYMBOL(free_dma);

#ifdef CONFIG_PM
# ifndef MAX_DMA_SUSPEND_CHANNELS
#  define MAX_DMA_SUSPEND_CHANNELS MAX_DMA_CHANNELS
# endif
# ifndef CONFIG_BF60x
int blackfin_dma_suspend(void)
{
	int i;

	for (i = 0; i < MAX_DMA_CHANNELS; ++i) {
		if (dma_ch[i].regs->cfg & DMAEN) {
			printk(KERN_ERR "DMA Channel %d failed to suspend\n", i);
			return -EBUSY;
		}
		if (i < MAX_DMA_SUSPEND_CHANNELS)
			dma_ch[i].saved_peripheral_map = dma_ch[i].regs->peripheral_map;
	}

#if ANOMALY_05000480
	bfin_write_DMAC_TC_PER(0x0);
#endif
	return 0;
}

void blackfin_dma_resume(void)
{
	int i;

	for (i = 0; i < MAX_DMA_CHANNELS; ++i) {
		dma_ch[i].regs->cfg = 0;
		if (i < MAX_DMA_SUSPEND_CHANNELS)
			dma_ch[i].regs->peripheral_map = dma_ch[i].saved_peripheral_map;
	}
#if ANOMALY_05000480
	bfin_write_DMAC_TC_PER(0x0111);
#endif
}
# else
int blackfin_dma_suspend(void)
{
	return 0;
}

void blackfin_dma_resume(void)
{
}
#endif
#endif

/**
 *	blackfin_dma_early_init - minimal DMA init
 *
 * Setup a few DMA registers so we can safely do DMA transfers early on in
 * the kernel booting process.  Really this just means using dma_memcpy().
 */
void __init blackfin_dma_early_init(void)
{
	early_shadow_stamp();
	bfin_write_MDMA_S0_CONFIG(0);
	bfin_write_MDMA_S1_CONFIG(0);
}

void __init early_dma_memcpy(void *pdst, const void *psrc, size_t size)
{
	unsigned long dst = (unsigned long)pdst;
	unsigned long src = (unsigned long)psrc;
	struct dma_register *dst_ch, *src_ch;

	early_shadow_stamp();

	/* We assume that everything is 4 byte aligned, so include
	 * a basic sanity check
	 */
	BUG_ON(dst % 4);
	BUG_ON(src % 4);
	BUG_ON(size % 4);

	src_ch = 0;
	/* Find an avalible memDMA channel */
	while (1) {
		if (src_ch == (struct dma_register *)MDMA_S0_NEXT_DESC_PTR) {
			dst_ch = (struct dma_register *)MDMA_D1_NEXT_DESC_PTR;
			src_ch = (struct dma_register *)MDMA_S1_NEXT_DESC_PTR;
		} else {
			dst_ch = (struct dma_register *)MDMA_D0_NEXT_DESC_PTR;
			src_ch = (struct dma_register *)MDMA_S0_NEXT_DESC_PTR;
		}

		if (!DMA_MMR_READ(&src_ch->cfg))
			break;
		else if (DMA_MMR_READ(&dst_ch->irq_status) & DMA_DONE) {
			DMA_MMR_WRITE(&src_ch->cfg, 0);
			break;
		}
	}

	/* Force a sync in case a previous config reset on this channel
	 * occurred.  This is needed so subsequent writes to DMA registers
	 * are not spuriously lost/corrupted.
	 */
	__builtin_bfin_ssync();

	/* Destination */
	bfin_write32(&dst_ch->start_addr, dst);
	DMA_MMR_WRITE(&dst_ch->x_count, size >> 2);
	DMA_MMR_WRITE(&dst_ch->x_modify, 1 << 2);
	DMA_MMR_WRITE(&dst_ch->irq_status, DMA_DONE | DMA_ERR);

	/* Source */
	bfin_write32(&src_ch->start_addr, src);
	DMA_MMR_WRITE(&src_ch->x_count, size >> 2);
	DMA_MMR_WRITE(&src_ch->x_modify, 1 << 2);
	DMA_MMR_WRITE(&src_ch->irq_status, DMA_DONE | DMA_ERR);

	/* Enable */
	DMA_MMR_WRITE(&src_ch->cfg, DMAEN | WDSIZE_32);
	DMA_MMR_WRITE(&dst_ch->cfg, WNR | DI_EN_X | DMAEN | WDSIZE_32);

	/* Since we are atomic now, don't use the workaround ssync */
	__builtin_bfin_ssync();

#ifdef CONFIG_BF60x
	/* Work around a possible MDMA anomaly. Running 2 MDMA channels to
	 * transfer DDR data to L1 SRAM may corrupt data.
	 * Should be reverted after this issue is root caused.
	 */
	while (!(DMA_MMR_READ(&dst_ch->irq_status) & DMA_DONE))
		continue;
#endif
}

void __init early_dma_memcpy_done(void)
{
	early_shadow_stamp();

	while ((bfin_read_MDMA_S0_CONFIG() && !(bfin_read_MDMA_D0_IRQ_STATUS() & DMA_DONE)) ||
	       (bfin_read_MDMA_S1_CONFIG() && !(bfin_read_MDMA_D1_IRQ_STATUS() & DMA_DONE)))
		continue;

	bfin_write_MDMA_D0_IRQ_STATUS(DMA_DONE | DMA_ERR);
	bfin_write_MDMA_D1_IRQ_STATUS(DMA_DONE | DMA_ERR);
	/*
	 * Now that DMA is done, we would normally flush cache, but
	 * i/d cache isn't running this early, so we don't bother,
	 * and just clear out the DMA channel for next time
	 */
	bfin_write_MDMA_S0_CONFIG(0);
	bfin_write_MDMA_S1_CONFIG(0);
	bfin_write_MDMA_D0_CONFIG(0);
	bfin_write_MDMA_D1_CONFIG(0);

	__builtin_bfin_ssync();
}

#if defined(CH_MEM_STREAM3_SRC) && defined(CONFIG_BF60x)
#define bfin_read_MDMA_S_CONFIG bfin_read_MDMA_S3_CONFIG
#define bfin_write_MDMA_S_CONFIG bfin_write_MDMA_S3_CONFIG
#define bfin_write_MDMA_S_START_ADDR bfin_write_MDMA_S3_START_ADDR
#define bfin_write_MDMA_S_IRQ_STATUS bfin_write_MDMA_S3_IRQ_STATUS
#define bfin_write_MDMA_S_X_COUNT bfin_write_MDMA_S3_X_COUNT
#define bfin_write_MDMA_S_X_MODIFY bfin_write_MDMA_S3_X_MODIFY
#define bfin_write_MDMA_S_Y_COUNT bfin_write_MDMA_S3_Y_COUNT
#define bfin_write_MDMA_S_Y_MODIFY bfin_write_MDMA_S3_Y_MODIFY
#define bfin_write_MDMA_D_CONFIG bfin_write_MDMA_D3_CONFIG
#define bfin_write_MDMA_D_START_ADDR bfin_write_MDMA_D3_START_ADDR
#define bfin_read_MDMA_D_IRQ_STATUS bfin_read_MDMA_D3_IRQ_STATUS
#define bfin_write_MDMA_D_IRQ_STATUS bfin_write_MDMA_D3_IRQ_STATUS
#define bfin_write_MDMA_D_X_COUNT bfin_write_MDMA_D3_X_COUNT
#define bfin_write_MDMA_D_X_MODIFY bfin_write_MDMA_D3_X_MODIFY
#define bfin_write_MDMA_D_Y_COUNT bfin_write_MDMA_D3_Y_COUNT
#define bfin_write_MDMA_D_Y_MODIFY bfin_write_MDMA_D3_Y_MODIFY
#else
#define bfin_read_MDMA_S_CONFIG bfin_read_MDMA_S0_CONFIG
#define bfin_write_MDMA_S_CONFIG bfin_write_MDMA_S0_CONFIG
#define bfin_write_MDMA_S_START_ADDR bfin_write_MDMA_S0_START_ADDR
#define bfin_write_MDMA_S_IRQ_STATUS bfin_write_MDMA_S0_IRQ_STATUS
#define bfin_write_MDMA_S_X_COUNT bfin_write_MDMA_S0_X_COUNT
#define bfin_write_MDMA_S_X_MODIFY bfin_write_MDMA_S0_X_MODIFY
#define bfin_write_MDMA_S_Y_COUNT bfin_write_MDMA_S0_Y_COUNT
#define bfin_write_MDMA_S_Y_MODIFY bfin_write_MDMA_S0_Y_MODIFY
#define bfin_write_MDMA_D_CONFIG bfin_write_MDMA_D0_CONFIG
#define bfin_write_MDMA_D_START_ADDR bfin_write_MDMA_D0_START_ADDR
#define bfin_read_MDMA_D_IRQ_STATUS bfin_read_MDMA_D0_IRQ_STATUS
#define bfin_write_MDMA_D_IRQ_STATUS bfin_write_MDMA_D0_IRQ_STATUS
#define bfin_write_MDMA_D_X_COUNT bfin_write_MDMA_D0_X_COUNT
#define bfin_write_MDMA_D_X_MODIFY bfin_write_MDMA_D0_X_MODIFY
#define bfin_write_MDMA_D_Y_COUNT bfin_write_MDMA_D0_Y_COUNT
#define bfin_write_MDMA_D_Y_MODIFY bfin_write_MDMA_D0_Y_MODIFY
#endif

/**
 *	__dma_memcpy - program the MDMA registers
 *
 * Actually program MDMA0 and wait for the transfer to finish.  Disable IRQs
 * while programming registers so that everything is fully configured.  Wait
 * for DMA to finish with IRQs enabled.  If interrupted, the initial DMA_DONE
 * check will make sure we don't clobber any existing transfer.
 */
static void __dma_memcpy(u32 daddr, s16 dmod, u32 saddr, s16 smod, size_t cnt, u32 conf)
{
	static DEFINE_SPINLOCK(mdma_lock);
	unsigned long flags;

	spin_lock_irqsave(&mdma_lock, flags);

	/* Force a sync in case a previous config reset on this channel
	 * occurred.  This is needed so subsequent writes to DMA registers
	 * are not spuriously lost/corrupted.  Do it under irq lock and
	 * without the anomaly version (because we are atomic already).
	 */
	__builtin_bfin_ssync();

	if (bfin_read_MDMA_S_CONFIG())
		while (!(bfin_read_MDMA_D_IRQ_STATUS() & DMA_DONE))
			continue;

	if (conf & DMA2D) {
		/* For larger bit sizes, we've already divided down cnt so it
		 * is no longer a multiple of 64k.  So we have to break down
		 * the limit here so it is a multiple of the incoming size.
		 * There is no limitation here in terms of total size other
		 * than the hardware though as the bits lost in the shift are
		 * made up by MODIFY (== we can hit the whole address space).
		 * X: (2^(16 - 0)) * 1 == (2^(16 - 1)) * 2 == (2^(16 - 2)) * 4
		 */
		u32 shift = abs(dmod) >> 1;
		size_t ycnt = cnt >> (16 - shift);
		cnt = 1 << (16 - shift);
		bfin_write_MDMA_D_Y_COUNT(ycnt);
		bfin_write_MDMA_S_Y_COUNT(ycnt);
		bfin_write_MDMA_D_Y_MODIFY(dmod);
		bfin_write_MDMA_S_Y_MODIFY(smod);
	}

	bfin_write_MDMA_D_START_ADDR(daddr);
	bfin_write_MDMA_D_X_COUNT(cnt);
	bfin_write_MDMA_D_X_MODIFY(dmod);
	bfin_write_MDMA_D_IRQ_STATUS(DMA_DONE | DMA_ERR);

	bfin_write_MDMA_S_START_ADDR(saddr);
	bfin_write_MDMA_S_X_COUNT(cnt);
	bfin_write_MDMA_S_X_MODIFY(smod);
	bfin_write_MDMA_S_IRQ_STATUS(DMA_DONE | DMA_ERR);

	bfin_write_MDMA_S_CONFIG(DMAEN | conf);
	if (conf & DMA2D)
		bfin_write_MDMA_D_CONFIG(WNR | DI_EN_Y | DMAEN | conf);
	else
		bfin_write_MDMA_D_CONFIG(WNR | DI_EN_X | DMAEN | conf);

	spin_unlock_irqrestore(&mdma_lock, flags);

	SSYNC();

	while (!(bfin_read_MDMA_D_IRQ_STATUS() & DMA_DONE))
		if (bfin_read_MDMA_S_CONFIG())
			continue;
		else
			return;

	bfin_write_MDMA_D_IRQ_STATUS(DMA_DONE | DMA_ERR);

	bfin_write_MDMA_S_CONFIG(0);
	bfin_write_MDMA_D_CONFIG(0);
}

/**
 *	_dma_memcpy - translate C memcpy settings into MDMA settings
 *
 * Handle all the high level steps before we touch the MDMA registers.  So
 * handle direction, tweaking of sizes, and formatting of addresses.
 */
static void *_dma_memcpy(void *pdst, const void *psrc, size_t size)
{
	u32 conf, shift;
	s16 mod;
	unsigned long dst = (unsigned long)pdst;
	unsigned long src = (unsigned long)psrc;

	if (size == 0)
		return NULL;

	if (dst % 4 == 0 && src % 4 == 0 && size % 4 == 0) {
		conf = WDSIZE_32;
		shift = 2;
	} else if (dst % 2 == 0 && src % 2 == 0 && size % 2 == 0) {
		conf = WDSIZE_16;
		shift = 1;
	} else {
		conf = WDSIZE_8;
		shift = 0;
	}

	/* If the two memory regions have a chance of overlapping, make
	 * sure the memcpy still works as expected.  Do this by having the
	 * copy run backwards instead.
	 */
	mod = 1 << shift;
	if (src < dst) {
		mod *= -1;
		dst += size + mod;
		src += size + mod;
	}
	size >>= shift;

#ifndef DMA_MMR_SIZE_32
	if (size > 0x10000)
		conf |= DMA2D;
#endif

	__dma_memcpy(dst, mod, src, mod, size, conf);

	return pdst;
}

/**
 *	dma_memcpy - DMA memcpy under mutex lock
 *
 * Do not check arguments before starting the DMA memcpy.  Break the transfer
 * up into two pieces.  The first transfer is in multiples of 64k and the
 * second transfer is the piece smaller than 64k.
 */
void *dma_memcpy(void *pdst, const void *psrc, size_t size)
{
	unsigned long dst = (unsigned long)pdst;
	unsigned long src = (unsigned long)psrc;

	if (bfin_addr_dcacheable(src))
		blackfin_dcache_flush_range(src, src + size);

	if (bfin_addr_dcacheable(dst))
		blackfin_dcache_invalidate_range(dst, dst + size);

	return dma_memcpy_nocache(pdst, psrc, size);
}
EXPORT_SYMBOL(dma_memcpy);

/**
 *	dma_memcpy_nocache - DMA memcpy under mutex lock
 *	- No cache flush/invalidate
 *
 * Do not check arguments before starting the DMA memcpy.  Break the transfer
 * up into two pieces.  The first transfer is in multiples of 64k and the
 * second transfer is the piece smaller than 64k.
 */
void *dma_memcpy_nocache(void *pdst, const void *psrc, size_t size)
{
#ifdef DMA_MMR_SIZE_32
	_dma_memcpy(pdst, psrc, size);
#else
	size_t bulk, rest;

	bulk = size & ~0xffff;
	rest = size - bulk;
	if (bulk)
		_dma_memcpy(pdst, psrc, bulk);
	_dma_memcpy(pdst + bulk, psrc + bulk, rest);
#endif
	return pdst;
}
EXPORT_SYMBOL(dma_memcpy_nocache);

/**
 *	safe_dma_memcpy - DMA memcpy w/argument checking
 *
 * Verify arguments are safe before heading to dma_memcpy().
 */
void *safe_dma_memcpy(void *dst, const void *src, size_t size)
{
	if (!access_ok(VERIFY_WRITE, dst, size))
		return NULL;
	if (!access_ok(VERIFY_READ, src, size))
		return NULL;
	return dma_memcpy(dst, src, size);
}
EXPORT_SYMBOL(safe_dma_memcpy);

static void _dma_out(unsigned long addr, unsigned long buf, unsigned DMA_MMR_SIZE_TYPE len,
                     u16 size, u16 dma_size)
{
	blackfin_dcache_flush_range(buf, buf + len * size);
	__dma_memcpy(addr, 0, buf, size, len, dma_size);
}

static void _dma_in(unsigned long addr, unsigned long buf, unsigned DMA_MMR_SIZE_TYPE len,
                    u16 size, u16 dma_size)
{
	blackfin_dcache_invalidate_range(buf, buf + len * size);
	__dma_memcpy(buf, size, addr, 0, len, dma_size);
}

#define MAKE_DMA_IO(io, bwl, isize, dmasize, cnst) \
void dma_##io##s##bwl(unsigned long addr, cnst void *buf, unsigned DMA_MMR_SIZE_TYPE len) \
{ \
	_dma_##io(addr, (unsigned long)buf, len, isize, WDSIZE_##dmasize); \
} \
EXPORT_SYMBOL(dma_##io##s##bwl)
MAKE_DMA_IO(out, b, 1,  8, const);
MAKE_DMA_IO(in,  b, 1,  8, );
MAKE_DMA_IO(out, w, 2, 16, const);
MAKE_DMA_IO(in,  w, 2, 16, );
MAKE_DMA_IO(out, l, 4, 32, const);
MAKE_DMA_IO(in,  l, 4, 32, );