summaryrefslogtreecommitdiff
path: root/arch/parisc/lib/io.c
blob: 7c1406ff825e388d96c02d0ee4b946b429d43e13 (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
/*
 * arch/parisc/lib/io.c
 *
 * Copyright (c) Matthew Wilcox 2001 for Hewlett-Packard
 * Copyright (c) Randolph Chung 2001 <tausq@debian.org>
 *
 * IO accessing functions which shouldn't be inlined because they're too big
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <asm/io.h>

/* Copies a block of memory to a device in an efficient manner.
 * Assumes the device can cope with 32-bit transfers.  If it can't,
 * don't use this function.
 */
void memcpy_toio(volatile void __iomem *dst, const void *src, int count)
{
	if (((unsigned long)dst & 3) != ((unsigned long)src & 3))
		goto bytecopy;
	while ((unsigned long)dst & 3) {
		writeb(*(char *)src, dst++);
		src++;
		count--;
	}
	while (count > 3) {
		__raw_writel(*(u32 *)src, dst);
		src += 4;
		dst += 4;
		count -= 4;
	}
 bytecopy:
	while (count--) {
		writeb(*(char *)src, dst++);
		src++;
	}
}

/*
** Copies a block of memory from a device in an efficient manner.
** Assumes the device can cope with 32-bit transfers.  If it can't,
** don't use this function.
**
** CR16 counts on C3000 reading 256 bytes from Symbios 896 RAM:
**	27341/64    = 427 cyc per int
**	61311/128   = 478 cyc per short
**	122637/256  = 479 cyc per byte
** Ergo bus latencies dominant (not transfer size).
**      Minimize total number of transfers at cost of CPU cycles.
**	TODO: only look at src alignment and adjust the stores to dest.
*/
void memcpy_fromio(void *dst, const volatile void __iomem *src, int count)
{
	/* first compare alignment of src/dst */ 
	if ( (((unsigned long)dst ^ (unsigned long)src) & 1) || (count < 2) )
		goto bytecopy;

	if ( (((unsigned long)dst ^ (unsigned long)src) & 2) || (count < 4) )
		goto shortcopy;

	/* Then check for misaligned start address */
	if ((unsigned long)src & 1) {
		*(u8 *)dst = readb(src);
		src++;
		dst++;
		count--;
		if (count < 2) goto bytecopy;
	}

	if ((unsigned long)src & 2) {
		*(u16 *)dst = __raw_readw(src);
		src += 2;
		dst += 2;
		count -= 2;
	}

	while (count > 3) {
		*(u32 *)dst = __raw_readl(src);
		dst += 4;
		src += 4;
		count -= 4;
	}

 shortcopy:
	while (count > 1) {
		*(u16 *)dst = __raw_readw(src);
		src += 2;
		dst += 2;
		count -= 2;
	}

 bytecopy:
	while (count--) {
		*(char *)dst = readb(src);
		src++;
		dst++;
	}
}

/* Sets a block of memory on a device to a given value.
 * Assumes the device can cope with 32-bit transfers.  If it can't,
 * don't use this function.
 */
void memset_io(volatile void __iomem *addr, unsigned char val, int count)
{
	u32 val32 = (val << 24) | (val << 16) | (val << 8) | val;
	while ((unsigned long)addr & 3) {
		writeb(val, addr++);
		count--;
	}
	while (count > 3) {
		__raw_writel(val32, addr);
		addr += 4;
		count -= 4;
	}
	while (count--) {
		writeb(val, addr++);
	}
}

/*
 * Read COUNT 8-bit bytes from port PORT into memory starting at
 * SRC.
 */
void insb (unsigned long port, void *dst, unsigned long count)
{
	unsigned char *p;

	p = (unsigned char *)dst;

	while (((unsigned long)p) & 0x3) {
		if (!count)
			return;
		count--;
		*p = inb(port);
		p++;
	}

	while (count >= 4) {
		unsigned int w;
		count -= 4;
		w = inb(port) << 24;
		w |= inb(port) << 16;
		w |= inb(port) << 8;
		w |= inb(port);
		*(unsigned int *) p = w;
		p += 4;
	}

	while (count) {
		--count;
		*p = inb(port);
		p++;
	}
}


/*
 * Read COUNT 16-bit words from port PORT into memory starting at
 * SRC.  SRC must be at least short aligned.  This is used by the
 * IDE driver to read disk sectors.  Performance is important, but
 * the interfaces seems to be slow: just using the inlined version
 * of the inw() breaks things.
 */
void insw (unsigned long port, void *dst, unsigned long count)
{
	unsigned int l = 0, l2;
	unsigned char *p;

	p = (unsigned char *)dst;
	
	if (!count)
		return;
	
	switch (((unsigned long)p) & 0x3)
	{
	 case 0x00:			/* Buffer 32-bit aligned */
		while (count>=2) {
			
			count -= 2;
			l = cpu_to_le16(inw(port)) << 16;
			l |= cpu_to_le16(inw(port));
			*(unsigned int *)p = l;
			p += 4;
		}
		if (count) {
			*(unsigned short *)p = cpu_to_le16(inw(port));
		}
		break;
	
	 case 0x02:			/* Buffer 16-bit aligned */
		*(unsigned short *)p = cpu_to_le16(inw(port));
		p += 2;
		count--;
		while (count>=2) {
			
			count -= 2;
			l = cpu_to_le16(inw(port)) << 16;
			l |= cpu_to_le16(inw(port));
			*(unsigned int *)p = l;
			p += 4;
		}
		if (count) {
			*(unsigned short *)p = cpu_to_le16(inw(port));
		}
		break;
		
	 case 0x01:			/* Buffer 8-bit aligned */
	 case 0x03:
		/* I don't bother with 32bit transfers
		 * in this case, 16bit will have to do -- DE */
		--count;
		
		l = cpu_to_le16(inw(port));
		*p = l >> 8;
		p++;
		while (count--)
		{
			l2 = cpu_to_le16(inw(port));
			*(unsigned short *)p = (l & 0xff) << 8 | (l2 >> 8);
			p += 2;
			l = l2;
		}
		*p = l & 0xff;
		break;
	}
}



/*
 * Read COUNT 32-bit words from port PORT into memory starting at
 * SRC. Now works with any alignment in SRC. Performance is important,
 * but the interfaces seems to be slow: just using the inlined version
 * of the inl() breaks things.
 */
void insl (unsigned long port, void *dst, unsigned long count)
{
	unsigned int l = 0, l2;
	unsigned char *p;

	p = (unsigned char *)dst;
	
	if (!count)
		return;
	
	switch (((unsigned long) dst) & 0x3)
	{
	 case 0x00:			/* Buffer 32-bit aligned */
		while (count--)
		{
			*(unsigned int *)p = cpu_to_le32(inl(port));
			p += 4;
		}
		break;
	
	 case 0x02:			/* Buffer 16-bit aligned */
		--count;
		
		l = cpu_to_le32(inl(port));
		*(unsigned short *)p = l >> 16;
		p += 2;
		
		while (count--)
		{
			l2 = cpu_to_le32(inl(port));
			*(unsigned int *)p = (l & 0xffff) << 16 | (l2 >> 16);
			p += 4;
			l = l2;
		}
		*(unsigned short *)p = l & 0xffff;
		break;
	 case 0x01:			/* Buffer 8-bit aligned */
		--count;
		
		l = cpu_to_le32(inl(port));
		*(unsigned char *)p = l >> 24;
		p++;
		*(unsigned short *)p = (l >> 8) & 0xffff;
		p += 2;
		while (count--)
		{
			l2 = cpu_to_le32(inl(port));
			*(unsigned int *)p = (l & 0xff) << 24 | (l2 >> 8);
			p += 4;
			l = l2;
		}
		*p = l & 0xff;
		break;
	 case 0x03:			/* Buffer 8-bit aligned */
		--count;
		
		l = cpu_to_le32(inl(port));
		*p = l >> 24;
		p++;
		while (count--)
		{
			l2 = cpu_to_le32(inl(port));
			*(unsigned int *)p = (l & 0xffffff) << 8 | l2 >> 24;
			p += 4;
			l = l2;
		}
		*(unsigned short *)p = (l >> 8) & 0xffff;
		p += 2;
		*p = l & 0xff;
		break;
	}
}


/*
 * Like insb but in the opposite direction.
 * Don't worry as much about doing aligned memory transfers:
 * doing byte reads the "slow" way isn't nearly as slow as
 * doing byte writes the slow way (no r-m-w cycle).
 */
void outsb(unsigned long port, const void * src, unsigned long count)
{
	const unsigned char *p;

	p = (const unsigned char *)src;
	while (count) {
		count--;
		outb(*p, port);
		p++;
	}
}

/*
 * Like insw but in the opposite direction.  This is used by the IDE
 * driver to write disk sectors.  Performance is important, but the
 * interfaces seems to be slow: just using the inlined version of the
 * outw() breaks things.
 */
void outsw (unsigned long port, const void *src, unsigned long count)
{
	unsigned int l = 0, l2;
	const unsigned char *p;

	p = (const unsigned char *)src;
	
	if (!count)
		return;
	
	switch (((unsigned long)p) & 0x3)
	{
	 case 0x00:			/* Buffer 32-bit aligned */
		while (count>=2) {
			count -= 2;
			l = *(unsigned int *)p;
			p += 4;
			outw(le16_to_cpu(l >> 16), port);
			outw(le16_to_cpu(l & 0xffff), port);
		}
		if (count) {
			outw(le16_to_cpu(*(unsigned short*)p), port);
		}
		break;
	
	 case 0x02:			/* Buffer 16-bit aligned */
		
		outw(le16_to_cpu(*(unsigned short*)p), port);
		p += 2;
		count--;
		
		while (count>=2) {
			count -= 2;
			l = *(unsigned int *)p;
			p += 4;
			outw(le16_to_cpu(l >> 16), port);
			outw(le16_to_cpu(l & 0xffff), port);
		}
		if (count) {
			outw(le16_to_cpu(*(unsigned short *)p), port);
		}
		break;
		
	 case 0x01:			/* Buffer 8-bit aligned */	
		/* I don't bother with 32bit transfers
		 * in this case, 16bit will have to do -- DE */
		
		l  = *p << 8;
		p++;
		count--;
		while (count)
		{
			count--;
			l2 = *(unsigned short *)p;
			p += 2;
			outw(le16_to_cpu(l | l2 >> 8), port);
		        l = l2 << 8;
		}
		l2 = *(unsigned char *)p;
		outw (le16_to_cpu(l | l2>>8), port);
		break;
	
	}
}


/*
 * Like insl but in the opposite direction.  This is used by the IDE
 * driver to write disk sectors.  Works with any alignment in SRC.
 *  Performance is important, but the interfaces seems to be slow:
 * just using the inlined version of the outl() breaks things.
 */
void outsl (unsigned long port, const void *src, unsigned long count)
{
	unsigned int l = 0, l2;
	const unsigned char *p;

	p = (const unsigned char *)src;
	
	if (!count)
		return;
	
	switch (((unsigned long)p) & 0x3)
	{
	 case 0x00:			/* Buffer 32-bit aligned */
		while (count--)
		{
			outl(le32_to_cpu(*(unsigned int *)p), port);
			p += 4;
		}
		break;
	
	 case 0x02:			/* Buffer 16-bit aligned */
		--count;
		
		l = *(unsigned short *)p;
		p += 2;
		
		while (count--)
		{
			l2 = *(unsigned int *)p;
			p += 4;
			outl (le32_to_cpu(l << 16 | l2 >> 16), port);
			l = l2;
		}
		l2 = *(unsigned short *)p;
		outl (le32_to_cpu(l << 16 | l2), port);
		break;
	 case 0x01:			/* Buffer 8-bit aligned */
		--count;

		l = *p << 24;
		p++;
		l |= *(unsigned short *)p << 8;
		p += 2;

		while (count--)
		{
			l2 = *(unsigned int *)p;
			p += 4;
			outl (le32_to_cpu(l | l2 >> 24), port);
			l = l2 << 8;
		}
		l2 = *p;
		outl (le32_to_cpu(l | l2), port);
		break;
	 case 0x03:			/* Buffer 8-bit aligned */
		--count;
		
		l = *p << 24;
		p++;

		while (count--)
		{
			l2 = *(unsigned int *)p;
			p += 4;
			outl (le32_to_cpu(l | l2 >> 8), port);
			l = l2 << 24;
		}
		l2 = *(unsigned short *)p << 16;
		p += 2;
		l2 |= *p;
		outl (le32_to_cpu(l | l2), port);
		break;
	}
}

EXPORT_SYMBOL(insb);
EXPORT_SYMBOL(insw);
EXPORT_SYMBOL(insl);
EXPORT_SYMBOL(outsb);
EXPORT_SYMBOL(outsw);
EXPORT_SYMBOL(outsl);