1 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
2 
3 #include <linux/kernel.h>
4 #include <linux/export.h>
5 #include <media/saa7146_vv.h>
6 
7 static void calculate_output_format_register(struct saa7146_dev* saa, u32 palette, u32* clip_format)
8 {
9 	/* clear out the necessary bits */
10 	*clip_format &= 0x0000ffff;
11 	/* set these bits new */
12 	*clip_format |=  (( ((palette&0xf00)>>8) << 30) | ((palette&0x00f) << 24) | (((palette&0x0f0)>>4) << 16));
13 }
14 
15 static void calculate_hps_source_and_sync(struct saa7146_dev *dev, int source, int sync, u32* hps_ctrl)
16 {
17 	*hps_ctrl &= ~(MASK_30 | MASK_31 | MASK_28);
18 	*hps_ctrl |= (source << 30) | (sync << 28);
19 }
20 
21 static void calculate_hxo_and_hyo(struct saa7146_vv *vv, u32* hps_h_scale, u32* hps_ctrl)
22 {
23 	int hyo = 0, hxo = 0;
24 
25 	hyo = vv->standard->v_offset;
26 	hxo = vv->standard->h_offset;
27 
28 	*hps_h_scale	&= ~(MASK_B0 | 0xf00);
29 	*hps_h_scale	|= (hxo <<  0);
30 
31 	*hps_ctrl	&= ~(MASK_W0 | MASK_B2);
32 	*hps_ctrl	|= (hyo << 12);
33 }
34 
35 /* helper functions for the calculation of the horizontal- and vertical
36    scaling registers, clip-format-register etc ...
37    these functions take pointers to the (most-likely read-out
38    original-values) and manipulate them according to the requested
39    changes.
40 */
41 
42 /* hps_coeff used for CXY and CXUV; scale 1/1 -> scale 1/64 */
43 static struct {
44 	u16 hps_coeff;
45 	u16 weight_sum;
46 } hps_h_coeff_tab [] = {
47 	{0x00,   2}, {0x02,   4}, {0x00,   4}, {0x06,   8}, {0x02,   8},
48 	{0x08,   8}, {0x00,   8}, {0x1E,  16}, {0x0E,   8}, {0x26,   8},
49 	{0x06,   8}, {0x42,   8}, {0x02,   8}, {0x80,   8}, {0x00,   8},
50 	{0xFE,  16}, {0xFE,   8}, {0x7E,   8}, {0x7E,   8}, {0x3E,   8},
51 	{0x3E,   8}, {0x1E,   8}, {0x1E,   8}, {0x0E,   8}, {0x0E,   8},
52 	{0x06,   8}, {0x06,   8}, {0x02,   8}, {0x02,   8}, {0x00,   8},
53 	{0x00,   8}, {0xFE,  16}, {0xFE,   8}, {0xFE,   8}, {0xFE,   8},
54 	{0xFE,   8}, {0xFE,   8}, {0xFE,   8}, {0xFE,   8}, {0xFE,   8},
55 	{0xFE,   8}, {0xFE,   8}, {0xFE,   8}, {0xFE,   8}, {0xFE,   8},
56 	{0xFE,   8}, {0xFE,   8}, {0xFE,   8}, {0xFE,   8}, {0x7E,   8},
57 	{0x7E,   8}, {0x3E,   8}, {0x3E,   8}, {0x1E,   8}, {0x1E,   8},
58 	{0x0E,   8}, {0x0E,   8}, {0x06,   8}, {0x06,   8}, {0x02,   8},
59 	{0x02,   8}, {0x00,   8}, {0x00,   8}, {0xFE,  16}
60 };
61 
62 /* table of attenuation values for horizontal scaling */
63 static u8 h_attenuation[] = { 1, 2, 4, 8, 2, 4, 8, 16, 0};
64 
65 /* calculate horizontal scale registers */
66 static int calculate_h_scale_registers(struct saa7146_dev *dev,
67 	int in_x, int out_x, int flip_lr,
68 	u32* hps_ctrl, u32* hps_v_gain, u32* hps_h_prescale, u32* hps_h_scale)
69 {
70 	/* horizontal prescaler */
71 	u32 dcgx = 0, xpsc = 0, xacm = 0, cxy = 0, cxuv = 0;
72 	/* horizontal scaler */
73 	u32 xim = 0, xp = 0, xsci =0;
74 	/* vertical scale & gain */
75 	u32 pfuv = 0;
76 
77 	/* helper variables */
78 	u32 h_atten = 0, i = 0;
79 
80 	if ( 0 == out_x ) {
81 		return -EINVAL;
82 	}
83 
84 	/* mask out vanity-bit */
85 	*hps_ctrl &= ~MASK_29;
86 
87 	/* calculate prescale-(xspc)-value:	[n   .. 1/2) : 1
88 						[1/2 .. 1/3) : 2
89 						[1/3 .. 1/4) : 3
90 						...		*/
91 	if (in_x > out_x) {
92 		xpsc = in_x / out_x;
93 	}
94 	else {
95 		/* zooming */
96 		xpsc = 1;
97 	}
98 
99 	/* if flip_lr-bit is set, number of pixels after
100 	   horizontal prescaling must be < 384 */
101 	if ( 0 != flip_lr ) {
102 
103 		/* set vanity bit */
104 		*hps_ctrl |= MASK_29;
105 
106 		while (in_x / xpsc >= 384 )
107 			xpsc++;
108 	}
109 	/* if zooming is wanted, number of pixels after
110 	   horizontal prescaling must be < 768 */
111 	else {
112 		while ( in_x / xpsc >= 768 )
113 			xpsc++;
114 	}
115 
116 	/* maximum prescale is 64 (p.69) */
117 	if ( xpsc > 64 )
118 		xpsc = 64;
119 
120 	/* keep xacm clear*/
121 	xacm = 0;
122 
123 	/* set horizontal filter parameters (CXY = CXUV) */
124 	cxy = hps_h_coeff_tab[( (xpsc - 1) < 63 ? (xpsc - 1) : 63 )].hps_coeff;
125 	cxuv = cxy;
126 
127 	/* calculate and set horizontal fine scale (xsci) */
128 
129 	/* bypass the horizontal scaler ? */
130 	if ( (in_x == out_x) && ( 1 == xpsc ) )
131 		xsci = 0x400;
132 	else
133 		xsci = ( (1024 * in_x) / (out_x * xpsc) ) + xpsc;
134 
135 	/* set start phase for horizontal fine scale (xp) to 0 */
136 	xp = 0;
137 
138 	/* set xim, if we bypass the horizontal scaler */
139 	if ( 0x400 == xsci )
140 		xim = 1;
141 	else
142 		xim = 0;
143 
144 	/* if the prescaler is bypassed, enable horizontal
145 	   accumulation mode (xacm) and clear dcgx */
146 	if( 1 == xpsc ) {
147 		xacm = 1;
148 		dcgx = 0;
149 	} else {
150 		xacm = 0;
151 		/* get best match in the table of attenuations
152 		   for horizontal scaling */
153 		h_atten = hps_h_coeff_tab[( (xpsc - 1) < 63 ? (xpsc - 1) : 63 )].weight_sum;
154 
155 		for (i = 0; h_attenuation[i] != 0; i++) {
156 			if (h_attenuation[i] >= h_atten)
157 				break;
158 		}
159 
160 		dcgx = i;
161 	}
162 
163 	/* the horizontal scaling increment controls the UV filter
164 	   to reduce the bandwidth to improve the display quality,
165 	   so set it ... */
166 	if ( xsci == 0x400)
167 		pfuv = 0x00;
168 	else if ( xsci < 0x600)
169 		pfuv = 0x01;
170 	else if ( xsci < 0x680)
171 		pfuv = 0x11;
172 	else if ( xsci < 0x700)
173 		pfuv = 0x22;
174 	else
175 		pfuv = 0x33;
176 
177 
178 	*hps_v_gain  &= MASK_W0|MASK_B2;
179 	*hps_v_gain  |= (pfuv << 24);
180 
181 	*hps_h_scale	&= ~(MASK_W1 | 0xf000);
182 	*hps_h_scale	|= (xim << 31) | (xp << 24) | (xsci << 12);
183 
184 	*hps_h_prescale	|= (dcgx << 27) | ((xpsc-1) << 18) | (xacm << 17) | (cxy << 8) | (cxuv << 0);
185 
186 	return 0;
187 }
188 
189 static struct {
190 	u16 hps_coeff;
191 	u16 weight_sum;
192 } hps_v_coeff_tab [] = {
193  {0x0100,   2},  {0x0102,   4},  {0x0300,   4},  {0x0106,   8},  {0x0502,   8},
194  {0x0708,   8},  {0x0F00,   8},  {0x011E,  16},  {0x110E,  16},  {0x1926,  16},
195  {0x3906,  16},  {0x3D42,  16},  {0x7D02,  16},  {0x7F80,  16},  {0xFF00,  16},
196  {0x01FE,  32},  {0x01FE,  32},  {0x817E,  32},  {0x817E,  32},  {0xC13E,  32},
197  {0xC13E,  32},  {0xE11E,  32},  {0xE11E,  32},  {0xF10E,  32},  {0xF10E,  32},
198  {0xF906,  32},  {0xF906,  32},  {0xFD02,  32},  {0xFD02,  32},  {0xFF00,  32},
199  {0xFF00,  32},  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},
200  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},
201  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},
202  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x817E,  64},
203  {0x817E,  64},  {0xC13E,  64},  {0xC13E,  64},  {0xE11E,  64},  {0xE11E,  64},
204  {0xF10E,  64},  {0xF10E,  64},  {0xF906,  64},  {0xF906,  64},  {0xFD02,  64},
205  {0xFD02,  64},  {0xFF00,  64},  {0xFF00,  64},  {0x01FE, 128}
206 };
207 
208 /* table of attenuation values for vertical scaling */
209 static u16 v_attenuation[] = { 2, 4, 8, 16, 32, 64, 128, 256, 0};
210 
211 /* calculate vertical scale registers */
212 static int calculate_v_scale_registers(struct saa7146_dev *dev, enum v4l2_field field,
213 	int in_y, int out_y, u32* hps_v_scale, u32* hps_v_gain)
214 {
215 	int lpi = 0;
216 
217 	/* vertical scaling */
218 	u32 yacm = 0, ysci = 0, yacl = 0, ypo = 0, ype = 0;
219 	/* vertical scale & gain */
220 	u32 dcgy = 0, cya_cyb = 0;
221 
222 	/* helper variables */
223 	u32 v_atten = 0, i = 0;
224 
225 	/* error, if vertical zooming */
226 	if ( in_y < out_y ) {
227 		return -EINVAL;
228 	}
229 
230 	/* linear phase interpolation may be used
231 	   if scaling is between 1 and 1/2 (both fields used)
232 	   or scaling is between 1/2 and 1/4 (if only one field is used) */
233 
234 	if (V4L2_FIELD_HAS_BOTH(field)) {
235 		if( 2*out_y >= in_y) {
236 			lpi = 1;
237 		}
238 	} else if (field == V4L2_FIELD_TOP
239 		|| field == V4L2_FIELD_ALTERNATE
240 		|| field == V4L2_FIELD_BOTTOM) {
241 		if( 4*out_y >= in_y ) {
242 			lpi = 1;
243 		}
244 		out_y *= 2;
245 	}
246 	if( 0 != lpi ) {
247 
248 		yacm = 0;
249 		yacl = 0;
250 		cya_cyb = 0x00ff;
251 
252 		/* calculate scaling increment */
253 		if ( in_y > out_y )
254 			ysci = ((1024 * in_y) / (out_y + 1)) - 1024;
255 		else
256 			ysci = 0;
257 
258 		dcgy = 0;
259 
260 		/* calculate ype and ypo */
261 		ype = ysci / 16;
262 		ypo = ype + (ysci / 64);
263 
264 	} else {
265 		yacm = 1;
266 
267 		/* calculate scaling increment */
268 		ysci = (((10 * 1024 * (in_y - out_y - 1)) / in_y) + 9) / 10;
269 
270 		/* calculate ype and ypo */
271 		ypo = ype = ((ysci + 15) / 16);
272 
273 		/* the sequence length interval (yacl) has to be set according
274 		   to the prescale value, e.g.	[n   .. 1/2) : 0
275 						[1/2 .. 1/3) : 1
276 						[1/3 .. 1/4) : 2
277 						... */
278 		if ( ysci < 512) {
279 			yacl = 0;
280 		} else {
281 			yacl = ( ysci / (1024 - ysci) );
282 		}
283 
284 		/* get filter coefficients for cya, cyb from table hps_v_coeff_tab */
285 		cya_cyb = hps_v_coeff_tab[ (yacl < 63 ? yacl : 63 ) ].hps_coeff;
286 
287 		/* get best match in the table of attenuations for vertical scaling */
288 		v_atten = hps_v_coeff_tab[ (yacl < 63 ? yacl : 63 ) ].weight_sum;
289 
290 		for (i = 0; v_attenuation[i] != 0; i++) {
291 			if (v_attenuation[i] >= v_atten)
292 				break;
293 		}
294 
295 		dcgy = i;
296 	}
297 
298 	/* ypo and ype swapped in spec ? */
299 	*hps_v_scale	|= (yacm << 31) | (ysci << 21) | (yacl << 15) | (ypo << 8 ) | (ype << 1);
300 
301 	*hps_v_gain	&= ~(MASK_W0|MASK_B2);
302 	*hps_v_gain	|= (dcgy << 16) | (cya_cyb << 0);
303 
304 	return 0;
305 }
306 
307 /* simple bubble-sort algorithm with duplicate elimination */
308 static int sort_and_eliminate(u32* values, int* count)
309 {
310 	int low = 0, high = 0, top = 0, temp = 0;
311 	int cur = 0, next = 0;
312 
313 	/* sanity checks */
314 	if( (0 > *count) || (NULL == values) ) {
315 		return -EINVAL;
316 	}
317 
318 	/* bubble sort the first @count items of the array @values */
319 	for( top = *count; top > 0; top--) {
320 		for( low = 0, high = 1; high < top; low++, high++) {
321 			if( values[low] > values[high] ) {
322 				temp = values[low];
323 				values[low] = values[high];
324 				values[high] = temp;
325 			}
326 		}
327 	}
328 
329 	/* remove duplicate items */
330 	for( cur = 0, next = 1; next < *count; next++) {
331 		if( values[cur] != values[next])
332 			values[++cur] = values[next];
333 	}
334 
335 	*count = cur + 1;
336 
337 	return 0;
338 }
339 
340 static void calculate_clipping_registers_rect(struct saa7146_dev *dev, struct saa7146_fh *fh,
341 	struct saa7146_video_dma *vdma2, u32* clip_format, u32* arbtr_ctrl, enum v4l2_field field)
342 {
343 	struct saa7146_vv *vv = dev->vv_data;
344 	__le32 *clipping = vv->d_clipping.cpu_addr;
345 
346 	int width = vv->ov.win.w.width;
347 	int height =  vv->ov.win.w.height;
348 	int clipcount = vv->ov.nclips;
349 
350 	u32 line_list[32];
351 	u32 pixel_list[32];
352 	int numdwords = 0;
353 
354 	int i = 0, j = 0;
355 	int cnt_line = 0, cnt_pixel = 0;
356 
357 	int x[32], y[32], w[32], h[32];
358 
359 	/* clear out memory */
360 	memset(&line_list[0],  0x00, sizeof(u32)*32);
361 	memset(&pixel_list[0], 0x00, sizeof(u32)*32);
362 	memset(clipping,  0x00, SAA7146_CLIPPING_MEM);
363 
364 	/* fill the line and pixel-lists */
365 	for(i = 0; i < clipcount; i++) {
366 		int l = 0, r = 0, t = 0, b = 0;
367 
368 		x[i] = vv->ov.clips[i].c.left;
369 		y[i] = vv->ov.clips[i].c.top;
370 		w[i] = vv->ov.clips[i].c.width;
371 		h[i] = vv->ov.clips[i].c.height;
372 
373 		if( w[i] < 0) {
374 			x[i] += w[i]; w[i] = -w[i];
375 		}
376 		if( h[i] < 0) {
377 			y[i] += h[i]; h[i] = -h[i];
378 		}
379 		if( x[i] < 0) {
380 			w[i] += x[i]; x[i] = 0;
381 		}
382 		if( y[i] < 0) {
383 			h[i] += y[i]; y[i] = 0;
384 		}
385 		if( 0 != vv->vflip ) {
386 			y[i] = height - y[i] - h[i];
387 		}
388 
389 		l = x[i];
390 		r = x[i]+w[i];
391 		t = y[i];
392 		b = y[i]+h[i];
393 
394 		/* insert left/right coordinates */
395 		pixel_list[ 2*i   ] = min_t(int, l, width);
396 		pixel_list[(2*i)+1] = min_t(int, r, width);
397 		/* insert top/bottom coordinates */
398 		line_list[ 2*i   ] = min_t(int, t, height);
399 		line_list[(2*i)+1] = min_t(int, b, height);
400 	}
401 
402 	/* sort and eliminate lists */
403 	cnt_line = cnt_pixel = 2*clipcount;
404 	sort_and_eliminate( &pixel_list[0], &cnt_pixel );
405 	sort_and_eliminate( &line_list[0], &cnt_line );
406 
407 	/* calculate the number of used u32s */
408 	numdwords = max_t(int, (cnt_line+1), (cnt_pixel+1))*2;
409 	numdwords = max_t(int, 4, numdwords);
410 	numdwords = min_t(int, 64, numdwords);
411 
412 	/* fill up cliptable */
413 	for(i = 0; i < cnt_pixel; i++) {
414 		clipping[2*i] |= cpu_to_le32(pixel_list[i] << 16);
415 	}
416 	for(i = 0; i < cnt_line; i++) {
417 		clipping[(2*i)+1] |= cpu_to_le32(line_list[i] << 16);
418 	}
419 
420 	/* fill up cliptable with the display infos */
421 	for(j = 0; j < clipcount; j++) {
422 
423 		for(i = 0; i < cnt_pixel; i++) {
424 
425 			if( x[j] < 0)
426 				x[j] = 0;
427 
428 			if( pixel_list[i] < (x[j] + w[j])) {
429 
430 				if ( pixel_list[i] >= x[j] ) {
431 					clipping[2*i] |= cpu_to_le32(1 << j);
432 				}
433 			}
434 		}
435 		for(i = 0; i < cnt_line; i++) {
436 
437 			if( y[j] < 0)
438 				y[j] = 0;
439 
440 			if( line_list[i] < (y[j] + h[j]) ) {
441 
442 				if( line_list[i] >= y[j] ) {
443 					clipping[(2*i)+1] |= cpu_to_le32(1 << j);
444 				}
445 			}
446 		}
447 	}
448 
449 	/* adjust arbitration control register */
450 	*arbtr_ctrl &= 0xffff00ff;
451 	*arbtr_ctrl |= 0x00001c00;
452 
453 	vdma2->base_even	= vv->d_clipping.dma_handle;
454 	vdma2->base_odd		= vv->d_clipping.dma_handle;
455 	vdma2->prot_addr	= vv->d_clipping.dma_handle+((sizeof(u32))*(numdwords));
456 	vdma2->base_page	= 0x04;
457 	vdma2->pitch		= 0x00;
458 	vdma2->num_line_byte	= (0 << 16 | (sizeof(u32))*(numdwords-1) );
459 
460 	/* set clipping-mode. this depends on the field(s) used */
461 	*clip_format &= 0xfffffff7;
462 	if (V4L2_FIELD_HAS_BOTH(field)) {
463 		*clip_format |= 0x00000008;
464 	} else {
465 		*clip_format |= 0x00000000;
466 	}
467 }
468 
469 /* disable clipping */
470 static void saa7146_disable_clipping(struct saa7146_dev *dev)
471 {
472 	u32 clip_format	= saa7146_read(dev, CLIP_FORMAT_CTRL);
473 
474 	/* mask out relevant bits (=lower word)*/
475 	clip_format &= MASK_W1;
476 
477 	/* upload clipping-registers*/
478 	saa7146_write(dev, CLIP_FORMAT_CTRL,clip_format);
479 	saa7146_write(dev, MC2, (MASK_05 | MASK_21));
480 
481 	/* disable video dma2 */
482 	saa7146_write(dev, MC1, MASK_21);
483 }
484 
485 static void saa7146_set_clipping_rect(struct saa7146_fh *fh)
486 {
487 	struct saa7146_dev *dev = fh->dev;
488 	struct saa7146_vv *vv = dev->vv_data;
489 	enum v4l2_field field = vv->ov.win.field;
490 	struct	saa7146_video_dma vdma2;
491 	u32 clip_format;
492 	u32 arbtr_ctrl;
493 
494 	/* check clipcount, disable clipping if clipcount == 0*/
495 	if (vv->ov.nclips == 0) {
496 		saa7146_disable_clipping(dev);
497 		return;
498 	}
499 
500 	clip_format = saa7146_read(dev, CLIP_FORMAT_CTRL);
501 	arbtr_ctrl = saa7146_read(dev, PCI_BT_V1);
502 
503 	calculate_clipping_registers_rect(dev, fh, &vdma2, &clip_format, &arbtr_ctrl, field);
504 
505 	/* set clipping format */
506 	clip_format &= 0xffff0008;
507 	clip_format |= (SAA7146_CLIPPING_RECT << 4);
508 
509 	/* prepare video dma2 */
510 	saa7146_write(dev, BASE_EVEN2,		vdma2.base_even);
511 	saa7146_write(dev, BASE_ODD2,		vdma2.base_odd);
512 	saa7146_write(dev, PROT_ADDR2,		vdma2.prot_addr);
513 	saa7146_write(dev, BASE_PAGE2,		vdma2.base_page);
514 	saa7146_write(dev, PITCH2,		vdma2.pitch);
515 	saa7146_write(dev, NUM_LINE_BYTE2,	vdma2.num_line_byte);
516 
517 	/* prepare the rest */
518 	saa7146_write(dev, CLIP_FORMAT_CTRL,clip_format);
519 	saa7146_write(dev, PCI_BT_V1, arbtr_ctrl);
520 
521 	/* upload clip_control-register, clipping-registers, enable video dma2 */
522 	saa7146_write(dev, MC2, (MASK_05 | MASK_21 | MASK_03 | MASK_19));
523 	saa7146_write(dev, MC1, (MASK_05 | MASK_21));
524 }
525 
526 static void saa7146_set_window(struct saa7146_dev *dev, int width, int height, enum v4l2_field field)
527 {
528 	struct saa7146_vv *vv = dev->vv_data;
529 
530 	int source = vv->current_hps_source;
531 	int sync = vv->current_hps_sync;
532 
533 	u32 hps_v_scale = 0, hps_v_gain  = 0, hps_ctrl = 0, hps_h_prescale = 0, hps_h_scale = 0;
534 
535 	/* set vertical scale */
536 	hps_v_scale = 0; /* all bits get set by the function-call */
537 	hps_v_gain  = 0; /* fixme: saa7146_read(dev, HPS_V_GAIN);*/
538 	calculate_v_scale_registers(dev, field, vv->standard->v_field*2, height, &hps_v_scale, &hps_v_gain);
539 
540 	/* set horizontal scale */
541 	hps_ctrl	= 0;
542 	hps_h_prescale	= 0; /* all bits get set in the function */
543 	hps_h_scale	= 0;
544 	calculate_h_scale_registers(dev, vv->standard->h_pixels, width, vv->hflip, &hps_ctrl, &hps_v_gain, &hps_h_prescale, &hps_h_scale);
545 
546 	/* set hyo and hxo */
547 	calculate_hxo_and_hyo(vv, &hps_h_scale, &hps_ctrl);
548 	calculate_hps_source_and_sync(dev, source, sync, &hps_ctrl);
549 
550 	/* write out new register contents */
551 	saa7146_write(dev, HPS_V_SCALE,	hps_v_scale);
552 	saa7146_write(dev, HPS_V_GAIN,	hps_v_gain);
553 	saa7146_write(dev, HPS_CTRL,	hps_ctrl);
554 	saa7146_write(dev, HPS_H_PRESCALE,hps_h_prescale);
555 	saa7146_write(dev, HPS_H_SCALE,	hps_h_scale);
556 
557 	/* upload shadow-ram registers */
558 	saa7146_write(dev, MC2, (MASK_05 | MASK_06 | MASK_21 | MASK_22) );
559 }
560 
561 /* calculate the new memory offsets for a desired position */
562 static void saa7146_set_position(struct saa7146_dev *dev, int w_x, int w_y, int w_height, enum v4l2_field field, u32 pixelformat)
563 {
564 	struct saa7146_vv *vv = dev->vv_data;
565 	struct saa7146_format *sfmt = saa7146_format_by_fourcc(dev, pixelformat);
566 
567 	int b_depth = vv->ov_fmt->depth;
568 	int b_bpl = vv->ov_fb.fmt.bytesperline;
569 	/* The unsigned long cast is to remove a 64-bit compile warning since
570 	   it looks like a 64-bit address is cast to a 32-bit value, even
571 	   though the base pointer is really a 32-bit physical address that
572 	   goes into a 32-bit DMA register.
573 	   FIXME: might not work on some 64-bit platforms, but see the FIXME
574 	   in struct v4l2_framebuffer (videodev2.h) for that.
575 	 */
576 	u32 base = (u32)(unsigned long)vv->ov_fb.base;
577 
578 	struct	saa7146_video_dma vdma1;
579 
580 	/* calculate memory offsets for picture, look if we shall top-down-flip */
581 	vdma1.pitch	= 2*b_bpl;
582 	if ( 0 == vv->vflip ) {
583 		vdma1.base_even = base + (w_y * (vdma1.pitch/2)) + (w_x * (b_depth / 8));
584 		vdma1.base_odd  = vdma1.base_even + (vdma1.pitch / 2);
585 		vdma1.prot_addr = vdma1.base_even + (w_height * (vdma1.pitch / 2));
586 	}
587 	else {
588 		vdma1.base_even = base + ((w_y+w_height) * (vdma1.pitch/2)) + (w_x * (b_depth / 8));
589 		vdma1.base_odd  = vdma1.base_even - (vdma1.pitch / 2);
590 		vdma1.prot_addr = vdma1.base_odd - (w_height * (vdma1.pitch / 2));
591 	}
592 
593 	if (V4L2_FIELD_HAS_BOTH(field)) {
594 	} else if (field == V4L2_FIELD_ALTERNATE) {
595 		/* fixme */
596 		vdma1.base_odd = vdma1.prot_addr;
597 		vdma1.pitch /= 2;
598 	} else if (field == V4L2_FIELD_TOP) {
599 		vdma1.base_odd = vdma1.prot_addr;
600 		vdma1.pitch /= 2;
601 	} else if (field == V4L2_FIELD_BOTTOM) {
602 		vdma1.base_odd = vdma1.base_even;
603 		vdma1.base_even = vdma1.prot_addr;
604 		vdma1.pitch /= 2;
605 	}
606 
607 	if ( 0 != vv->vflip ) {
608 		vdma1.pitch *= -1;
609 	}
610 
611 	vdma1.base_page = sfmt->swap;
612 	vdma1.num_line_byte = (vv->standard->v_field<<16)+vv->standard->h_pixels;
613 
614 	saa7146_write_out_dma(dev, 1, &vdma1);
615 }
616 
617 static void saa7146_set_output_format(struct saa7146_dev *dev, unsigned long palette)
618 {
619 	u32 clip_format = saa7146_read(dev, CLIP_FORMAT_CTRL);
620 
621 	/* call helper function */
622 	calculate_output_format_register(dev,palette,&clip_format);
623 
624 	/* update the hps registers */
625 	saa7146_write(dev, CLIP_FORMAT_CTRL, clip_format);
626 	saa7146_write(dev, MC2, (MASK_05 | MASK_21));
627 }
628 
629 /* select input-source */
630 void saa7146_set_hps_source_and_sync(struct saa7146_dev *dev, int source, int sync)
631 {
632 	struct saa7146_vv *vv = dev->vv_data;
633 	u32 hps_ctrl = 0;
634 
635 	/* read old state */
636 	hps_ctrl = saa7146_read(dev, HPS_CTRL);
637 
638 	hps_ctrl &= ~( MASK_31 | MASK_30 | MASK_28 );
639 	hps_ctrl |= (source << 30) | (sync << 28);
640 
641 	/* write back & upload register */
642 	saa7146_write(dev, HPS_CTRL, hps_ctrl);
643 	saa7146_write(dev, MC2, (MASK_05 | MASK_21));
644 
645 	vv->current_hps_source = source;
646 	vv->current_hps_sync = sync;
647 }
648 EXPORT_SYMBOL_GPL(saa7146_set_hps_source_and_sync);
649 
650 int saa7146_enable_overlay(struct saa7146_fh *fh)
651 {
652 	struct saa7146_dev *dev = fh->dev;
653 	struct saa7146_vv *vv = dev->vv_data;
654 
655 	saa7146_set_window(dev, vv->ov.win.w.width, vv->ov.win.w.height, vv->ov.win.field);
656 	saa7146_set_position(dev, vv->ov.win.w.left, vv->ov.win.w.top, vv->ov.win.w.height, vv->ov.win.field, vv->ov_fmt->pixelformat);
657 	saa7146_set_output_format(dev, vv->ov_fmt->trans);
658 	saa7146_set_clipping_rect(fh);
659 
660 	/* enable video dma1 */
661 	saa7146_write(dev, MC1, (MASK_06 | MASK_22));
662 	return 0;
663 }
664 
665 void saa7146_disable_overlay(struct saa7146_fh *fh)
666 {
667 	struct saa7146_dev *dev = fh->dev;
668 
669 	/* disable clipping + video dma1 */
670 	saa7146_disable_clipping(dev);
671 	saa7146_write(dev, MC1, MASK_22);
672 }
673 
674 void saa7146_write_out_dma(struct saa7146_dev* dev, int which, struct saa7146_video_dma* vdma)
675 {
676 	int where = 0;
677 
678 	if( which < 1 || which > 3) {
679 		return;
680 	}
681 
682 	/* calculate starting address */
683 	where  = (which-1)*0x18;
684 
685 	saa7146_write(dev, where,	vdma->base_odd);
686 	saa7146_write(dev, where+0x04,	vdma->base_even);
687 	saa7146_write(dev, where+0x08,	vdma->prot_addr);
688 	saa7146_write(dev, where+0x0c,	vdma->pitch);
689 	saa7146_write(dev, where+0x10,	vdma->base_page);
690 	saa7146_write(dev, where+0x14,	vdma->num_line_byte);
691 
692 	/* upload */
693 	saa7146_write(dev, MC2, (MASK_02<<(which-1))|(MASK_18<<(which-1)));
694 /*
695 	printk("vdma%d.base_even:     0x%08x\n", which,vdma->base_even);
696 	printk("vdma%d.base_odd:      0x%08x\n", which,vdma->base_odd);
697 	printk("vdma%d.prot_addr:     0x%08x\n", which,vdma->prot_addr);
698 	printk("vdma%d.base_page:     0x%08x\n", which,vdma->base_page);
699 	printk("vdma%d.pitch:         0x%08x\n", which,vdma->pitch);
700 	printk("vdma%d.num_line_byte: 0x%08x\n", which,vdma->num_line_byte);
701 */
702 }
703 
704 static int calculate_video_dma_grab_packed(struct saa7146_dev* dev, struct saa7146_buf *buf)
705 {
706 	struct saa7146_vv *vv = dev->vv_data;
707 	struct saa7146_video_dma vdma1;
708 
709 	struct saa7146_format *sfmt = saa7146_format_by_fourcc(dev,buf->fmt->pixelformat);
710 
711 	int width = buf->fmt->width;
712 	int height = buf->fmt->height;
713 	int bytesperline = buf->fmt->bytesperline;
714 	enum v4l2_field field = buf->fmt->field;
715 
716 	int depth = sfmt->depth;
717 
718 	DEB_CAP("[size=%dx%d,fields=%s]\n",
719 		width, height, v4l2_field_names[field]);
720 
721 	if( bytesperline != 0) {
722 		vdma1.pitch = bytesperline*2;
723 	} else {
724 		vdma1.pitch = (width*depth*2)/8;
725 	}
726 	vdma1.num_line_byte	= ((vv->standard->v_field<<16) + vv->standard->h_pixels);
727 	vdma1.base_page		= buf->pt[0].dma | ME1 | sfmt->swap;
728 
729 	if( 0 != vv->vflip ) {
730 		vdma1.prot_addr	= buf->pt[0].offset;
731 		vdma1.base_even	= buf->pt[0].offset+(vdma1.pitch/2)*height;
732 		vdma1.base_odd	= vdma1.base_even - (vdma1.pitch/2);
733 	} else {
734 		vdma1.base_even	= buf->pt[0].offset;
735 		vdma1.base_odd	= vdma1.base_even + (vdma1.pitch/2);
736 		vdma1.prot_addr	= buf->pt[0].offset+(vdma1.pitch/2)*height;
737 	}
738 
739 	if (V4L2_FIELD_HAS_BOTH(field)) {
740 	} else if (field == V4L2_FIELD_ALTERNATE) {
741 		/* fixme */
742 		if ( vv->last_field == V4L2_FIELD_TOP ) {
743 			vdma1.base_odd	= vdma1.prot_addr;
744 			vdma1.pitch /= 2;
745 		} else if ( vv->last_field == V4L2_FIELD_BOTTOM ) {
746 			vdma1.base_odd	= vdma1.base_even;
747 			vdma1.base_even = vdma1.prot_addr;
748 			vdma1.pitch /= 2;
749 		}
750 	} else if (field == V4L2_FIELD_TOP) {
751 		vdma1.base_odd	= vdma1.prot_addr;
752 		vdma1.pitch /= 2;
753 	} else if (field == V4L2_FIELD_BOTTOM) {
754 		vdma1.base_odd	= vdma1.base_even;
755 		vdma1.base_even = vdma1.prot_addr;
756 		vdma1.pitch /= 2;
757 	}
758 
759 	if( 0 != vv->vflip ) {
760 		vdma1.pitch *= -1;
761 	}
762 
763 	saa7146_write_out_dma(dev, 1, &vdma1);
764 	return 0;
765 }
766 
767 static int calc_planar_422(struct saa7146_vv *vv, struct saa7146_buf *buf, struct saa7146_video_dma *vdma2, struct saa7146_video_dma *vdma3)
768 {
769 	int height = buf->fmt->height;
770 	int width = buf->fmt->width;
771 
772 	vdma2->pitch	= width;
773 	vdma3->pitch	= width;
774 
775 	/* fixme: look at bytesperline! */
776 
777 	if( 0 != vv->vflip ) {
778 		vdma2->prot_addr	= buf->pt[1].offset;
779 		vdma2->base_even	= ((vdma2->pitch/2)*height)+buf->pt[1].offset;
780 		vdma2->base_odd		= vdma2->base_even - (vdma2->pitch/2);
781 
782 		vdma3->prot_addr	= buf->pt[2].offset;
783 		vdma3->base_even	= ((vdma3->pitch/2)*height)+buf->pt[2].offset;
784 		vdma3->base_odd		= vdma3->base_even - (vdma3->pitch/2);
785 	} else {
786 		vdma3->base_even	= buf->pt[2].offset;
787 		vdma3->base_odd		= vdma3->base_even + (vdma3->pitch/2);
788 		vdma3->prot_addr	= (vdma3->pitch/2)*height+buf->pt[2].offset;
789 
790 		vdma2->base_even	= buf->pt[1].offset;
791 		vdma2->base_odd		= vdma2->base_even + (vdma2->pitch/2);
792 		vdma2->prot_addr	= (vdma2->pitch/2)*height+buf->pt[1].offset;
793 	}
794 
795 	return 0;
796 }
797 
798 static int calc_planar_420(struct saa7146_vv *vv, struct saa7146_buf *buf, struct saa7146_video_dma *vdma2, struct saa7146_video_dma *vdma3)
799 {
800 	int height = buf->fmt->height;
801 	int width = buf->fmt->width;
802 
803 	vdma2->pitch	= width/2;
804 	vdma3->pitch	= width/2;
805 
806 	if( 0 != vv->vflip ) {
807 		vdma2->prot_addr	= buf->pt[2].offset;
808 		vdma2->base_even	= ((vdma2->pitch/2)*height)+buf->pt[2].offset;
809 		vdma2->base_odd		= vdma2->base_even - (vdma2->pitch/2);
810 
811 		vdma3->prot_addr	= buf->pt[1].offset;
812 		vdma3->base_even	= ((vdma3->pitch/2)*height)+buf->pt[1].offset;
813 		vdma3->base_odd		= vdma3->base_even - (vdma3->pitch/2);
814 
815 	} else {
816 		vdma3->base_even	= buf->pt[2].offset;
817 		vdma3->base_odd		= vdma3->base_even + (vdma3->pitch);
818 		vdma3->prot_addr	= (vdma3->pitch/2)*height+buf->pt[2].offset;
819 
820 		vdma2->base_even	= buf->pt[1].offset;
821 		vdma2->base_odd		= vdma2->base_even + (vdma2->pitch);
822 		vdma2->prot_addr	= (vdma2->pitch/2)*height+buf->pt[1].offset;
823 	}
824 	return 0;
825 }
826 
827 static int calculate_video_dma_grab_planar(struct saa7146_dev* dev, struct saa7146_buf *buf)
828 {
829 	struct saa7146_vv *vv = dev->vv_data;
830 	struct saa7146_video_dma vdma1;
831 	struct saa7146_video_dma vdma2;
832 	struct saa7146_video_dma vdma3;
833 
834 	struct saa7146_format *sfmt = saa7146_format_by_fourcc(dev,buf->fmt->pixelformat);
835 
836 	int width = buf->fmt->width;
837 	int height = buf->fmt->height;
838 	enum v4l2_field field = buf->fmt->field;
839 
840 	BUG_ON(0 == buf->pt[0].dma);
841 	BUG_ON(0 == buf->pt[1].dma);
842 	BUG_ON(0 == buf->pt[2].dma);
843 
844 	DEB_CAP("[size=%dx%d,fields=%s]\n",
845 		width, height, v4l2_field_names[field]);
846 
847 	/* fixme: look at bytesperline! */
848 
849 	/* fixme: what happens for user space buffers here?. The offsets are
850 	   most likely wrong, this version here only works for page-aligned
851 	   buffers, modifications to the pagetable-functions are necessary...*/
852 
853 	vdma1.pitch		= width*2;
854 	vdma1.num_line_byte	= ((vv->standard->v_field<<16) + vv->standard->h_pixels);
855 	vdma1.base_page		= buf->pt[0].dma | ME1;
856 
857 	if( 0 != vv->vflip ) {
858 		vdma1.prot_addr	= buf->pt[0].offset;
859 		vdma1.base_even	= ((vdma1.pitch/2)*height)+buf->pt[0].offset;
860 		vdma1.base_odd	= vdma1.base_even - (vdma1.pitch/2);
861 	} else {
862 		vdma1.base_even	= buf->pt[0].offset;
863 		vdma1.base_odd	= vdma1.base_even + (vdma1.pitch/2);
864 		vdma1.prot_addr	= (vdma1.pitch/2)*height+buf->pt[0].offset;
865 	}
866 
867 	vdma2.num_line_byte	= 0; /* unused */
868 	vdma2.base_page		= buf->pt[1].dma | ME1;
869 
870 	vdma3.num_line_byte	= 0; /* unused */
871 	vdma3.base_page		= buf->pt[2].dma | ME1;
872 
873 	switch( sfmt->depth ) {
874 		case 12: {
875 			calc_planar_420(vv,buf,&vdma2,&vdma3);
876 			break;
877 		}
878 		case 16: {
879 			calc_planar_422(vv,buf,&vdma2,&vdma3);
880 			break;
881 		}
882 		default: {
883 			return -1;
884 		}
885 	}
886 
887 	if (V4L2_FIELD_HAS_BOTH(field)) {
888 	} else if (field == V4L2_FIELD_ALTERNATE) {
889 		/* fixme */
890 		vdma1.base_odd	= vdma1.prot_addr;
891 		vdma1.pitch /= 2;
892 		vdma2.base_odd	= vdma2.prot_addr;
893 		vdma2.pitch /= 2;
894 		vdma3.base_odd	= vdma3.prot_addr;
895 		vdma3.pitch /= 2;
896 	} else if (field == V4L2_FIELD_TOP) {
897 		vdma1.base_odd	= vdma1.prot_addr;
898 		vdma1.pitch /= 2;
899 		vdma2.base_odd	= vdma2.prot_addr;
900 		vdma2.pitch /= 2;
901 		vdma3.base_odd	= vdma3.prot_addr;
902 		vdma3.pitch /= 2;
903 	} else if (field == V4L2_FIELD_BOTTOM) {
904 		vdma1.base_odd	= vdma1.base_even;
905 		vdma1.base_even = vdma1.prot_addr;
906 		vdma1.pitch /= 2;
907 		vdma2.base_odd	= vdma2.base_even;
908 		vdma2.base_even = vdma2.prot_addr;
909 		vdma2.pitch /= 2;
910 		vdma3.base_odd	= vdma3.base_even;
911 		vdma3.base_even = vdma3.prot_addr;
912 		vdma3.pitch /= 2;
913 	}
914 
915 	if( 0 != vv->vflip ) {
916 		vdma1.pitch *= -1;
917 		vdma2.pitch *= -1;
918 		vdma3.pitch *= -1;
919 	}
920 
921 	saa7146_write_out_dma(dev, 1, &vdma1);
922 	if( (sfmt->flags & FORMAT_BYTE_SWAP) != 0 ) {
923 		saa7146_write_out_dma(dev, 3, &vdma2);
924 		saa7146_write_out_dma(dev, 2, &vdma3);
925 	} else {
926 		saa7146_write_out_dma(dev, 2, &vdma2);
927 		saa7146_write_out_dma(dev, 3, &vdma3);
928 	}
929 	return 0;
930 }
931 
932 static void program_capture_engine(struct saa7146_dev *dev, int planar)
933 {
934 	struct saa7146_vv *vv = dev->vv_data;
935 	int count = 0;
936 
937 	unsigned long e_wait = vv->current_hps_sync == SAA7146_HPS_SYNC_PORT_A ? CMD_E_FID_A : CMD_E_FID_B;
938 	unsigned long o_wait = vv->current_hps_sync == SAA7146_HPS_SYNC_PORT_A ? CMD_O_FID_A : CMD_O_FID_B;
939 
940 	/* wait for o_fid_a/b / e_fid_a/b toggle only if rps register 0 is not set*/
941 	WRITE_RPS0(CMD_PAUSE | CMD_OAN | CMD_SIG0 | o_wait);
942 	WRITE_RPS0(CMD_PAUSE | CMD_OAN | CMD_SIG0 | e_wait);
943 
944 	/* set rps register 0 */
945 	WRITE_RPS0(CMD_WR_REG | (1 << 8) | (MC2/4));
946 	WRITE_RPS0(MASK_27 | MASK_11);
947 
948 	/* turn on video-dma1 */
949 	WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
950 	WRITE_RPS0(MASK_06 | MASK_22);			/* => mask */
951 	WRITE_RPS0(MASK_06 | MASK_22);			/* => values */
952 	if( 0 != planar ) {
953 		/* turn on video-dma2 */
954 		WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
955 		WRITE_RPS0(MASK_05 | MASK_21);			/* => mask */
956 		WRITE_RPS0(MASK_05 | MASK_21);			/* => values */
957 
958 		/* turn on video-dma3 */
959 		WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
960 		WRITE_RPS0(MASK_04 | MASK_20);			/* => mask */
961 		WRITE_RPS0(MASK_04 | MASK_20);			/* => values */
962 	}
963 
964 	/* wait for o_fid_a/b / e_fid_a/b toggle */
965 	if ( vv->last_field == V4L2_FIELD_INTERLACED ) {
966 		WRITE_RPS0(CMD_PAUSE | o_wait);
967 		WRITE_RPS0(CMD_PAUSE | e_wait);
968 	} else if ( vv->last_field == V4L2_FIELD_TOP ) {
969 		WRITE_RPS0(CMD_PAUSE | (vv->current_hps_sync == SAA7146_HPS_SYNC_PORT_A ? MASK_10 : MASK_09));
970 		WRITE_RPS0(CMD_PAUSE | o_wait);
971 	} else if ( vv->last_field == V4L2_FIELD_BOTTOM ) {
972 		WRITE_RPS0(CMD_PAUSE | (vv->current_hps_sync == SAA7146_HPS_SYNC_PORT_A ? MASK_10 : MASK_09));
973 		WRITE_RPS0(CMD_PAUSE | e_wait);
974 	}
975 
976 	/* turn off video-dma1 */
977 	WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
978 	WRITE_RPS0(MASK_22 | MASK_06);			/* => mask */
979 	WRITE_RPS0(MASK_22);				/* => values */
980 	if( 0 != planar ) {
981 		/* turn off video-dma2 */
982 		WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
983 		WRITE_RPS0(MASK_05 | MASK_21);			/* => mask */
984 		WRITE_RPS0(MASK_21);				/* => values */
985 
986 		/* turn off video-dma3 */
987 		WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
988 		WRITE_RPS0(MASK_04 | MASK_20);			/* => mask */
989 		WRITE_RPS0(MASK_20);				/* => values */
990 	}
991 
992 	/* generate interrupt */
993 	WRITE_RPS0(CMD_INTERRUPT);
994 
995 	/* stop */
996 	WRITE_RPS0(CMD_STOP);
997 }
998 
999 void saa7146_set_capture(struct saa7146_dev *dev, struct saa7146_buf *buf, struct saa7146_buf *next)
1000 {
1001 	struct saa7146_format *sfmt = saa7146_format_by_fourcc(dev,buf->fmt->pixelformat);
1002 	struct saa7146_vv *vv = dev->vv_data;
1003 	u32 vdma1_prot_addr;
1004 
1005 	DEB_CAP("buf:%p, next:%p\n", buf, next);
1006 
1007 	vdma1_prot_addr = saa7146_read(dev, PROT_ADDR1);
1008 	if( 0 == vdma1_prot_addr ) {
1009 		/* clear out beginning of streaming bit (rps register 0)*/
1010 		DEB_CAP("forcing sync to new frame\n");
1011 		saa7146_write(dev, MC2, MASK_27 );
1012 	}
1013 
1014 	saa7146_set_window(dev, buf->fmt->width, buf->fmt->height, buf->fmt->field);
1015 	saa7146_set_output_format(dev, sfmt->trans);
1016 	saa7146_disable_clipping(dev);
1017 
1018 	if ( vv->last_field == V4L2_FIELD_INTERLACED ) {
1019 	} else if ( vv->last_field == V4L2_FIELD_TOP ) {
1020 		vv->last_field = V4L2_FIELD_BOTTOM;
1021 	} else if ( vv->last_field == V4L2_FIELD_BOTTOM ) {
1022 		vv->last_field = V4L2_FIELD_TOP;
1023 	}
1024 
1025 	if( 0 != IS_PLANAR(sfmt->trans)) {
1026 		calculate_video_dma_grab_planar(dev, buf);
1027 		program_capture_engine(dev,1);
1028 	} else {
1029 		calculate_video_dma_grab_packed(dev, buf);
1030 		program_capture_engine(dev,0);
1031 	}
1032 
1033 /*
1034 	printk("vdma%d.base_even:     0x%08x\n", 1,saa7146_read(dev,BASE_EVEN1));
1035 	printk("vdma%d.base_odd:      0x%08x\n", 1,saa7146_read(dev,BASE_ODD1));
1036 	printk("vdma%d.prot_addr:     0x%08x\n", 1,saa7146_read(dev,PROT_ADDR1));
1037 	printk("vdma%d.base_page:     0x%08x\n", 1,saa7146_read(dev,BASE_PAGE1));
1038 	printk("vdma%d.pitch:         0x%08x\n", 1,saa7146_read(dev,PITCH1));
1039 	printk("vdma%d.num_line_byte: 0x%08x\n", 1,saa7146_read(dev,NUM_LINE_BYTE1));
1040 	printk("vdma%d => vptr      : 0x%08x\n", 1,saa7146_read(dev,PCI_VDP1));
1041 */
1042 
1043 	/* write the address of the rps-program */
1044 	saa7146_write(dev, RPS_ADDR0, dev->d_rps0.dma_handle);
1045 
1046 	/* turn on rps */
1047 	saa7146_write(dev, MC1, (MASK_12 | MASK_28));
1048 }
1049