xref: /linux/drivers/media/common/saa7146/saa7146_hlp.c (revision 26fbb4c8c7c3ee9a4c3b4de555a8587b5a19154e)
1 // SPDX-License-Identifier: GPL-2.0-only
2 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
3 
4 #include <linux/kernel.h>
5 #include <linux/export.h>
6 #include <media/drv-intf/saa7146_vv.h>
7 
8 static void calculate_output_format_register(struct saa7146_dev* saa, u32 palette, u32* clip_format)
9 {
10 	/* clear out the necessary bits */
11 	*clip_format &= 0x0000ffff;
12 	/* set these bits new */
13 	*clip_format |=  (( ((palette&0xf00)>>8) << 30) | ((palette&0x00f) << 24) | (((palette&0x0f0)>>4) << 16));
14 }
15 
16 static void calculate_hps_source_and_sync(struct saa7146_dev *dev, int source, int sync, u32* hps_ctrl)
17 {
18 	*hps_ctrl &= ~(MASK_30 | MASK_31 | MASK_28);
19 	*hps_ctrl |= (source << 30) | (sync << 28);
20 }
21 
22 static void calculate_hxo_and_hyo(struct saa7146_vv *vv, u32* hps_h_scale, u32* hps_ctrl)
23 {
24 	int hyo = 0, hxo = 0;
25 
26 	hyo = vv->standard->v_offset;
27 	hxo = vv->standard->h_offset;
28 
29 	*hps_h_scale	&= ~(MASK_B0 | 0xf00);
30 	*hps_h_scale	|= (hxo <<  0);
31 
32 	*hps_ctrl	&= ~(MASK_W0 | MASK_B2);
33 	*hps_ctrl	|= (hyo << 12);
34 }
35 
36 /* helper functions for the calculation of the horizontal- and vertical
37    scaling registers, clip-format-register etc ...
38    these functions take pointers to the (most-likely read-out
39    original-values) and manipulate them according to the requested
40    changes.
41 */
42 
43 /* hps_coeff used for CXY and CXUV; scale 1/1 -> scale 1/64 */
44 static struct {
45 	u16 hps_coeff;
46 	u16 weight_sum;
47 } hps_h_coeff_tab [] = {
48 	{0x00,   2}, {0x02,   4}, {0x00,   4}, {0x06,   8}, {0x02,   8},
49 	{0x08,   8}, {0x00,   8}, {0x1E,  16}, {0x0E,   8}, {0x26,   8},
50 	{0x06,   8}, {0x42,   8}, {0x02,   8}, {0x80,   8}, {0x00,   8},
51 	{0xFE,  16}, {0xFE,   8}, {0x7E,   8}, {0x7E,   8}, {0x3E,   8},
52 	{0x3E,   8}, {0x1E,   8}, {0x1E,   8}, {0x0E,   8}, {0x0E,   8},
53 	{0x06,   8}, {0x06,   8}, {0x02,   8}, {0x02,   8}, {0x00,   8},
54 	{0x00,   8}, {0xFE,  16}, {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}, {0xFE,   8},
57 	{0xFE,   8}, {0xFE,   8}, {0xFE,   8}, {0xFE,   8}, {0x7E,   8},
58 	{0x7E,   8}, {0x3E,   8}, {0x3E,   8}, {0x1E,   8}, {0x1E,   8},
59 	{0x0E,   8}, {0x0E,   8}, {0x06,   8}, {0x06,   8}, {0x02,   8},
60 	{0x02,   8}, {0x00,   8}, {0x00,   8}, {0xFE,  16}
61 };
62 
63 /* table of attenuation values for horizontal scaling */
64 static u8 h_attenuation[] = { 1, 2, 4, 8, 2, 4, 8, 16, 0};
65 
66 /* calculate horizontal scale registers */
67 static int calculate_h_scale_registers(struct saa7146_dev *dev,
68 	int in_x, int out_x, int flip_lr,
69 	u32* hps_ctrl, u32* hps_v_gain, u32* hps_h_prescale, u32* hps_h_scale)
70 {
71 	/* horizontal prescaler */
72 	u32 dcgx = 0, xpsc = 0, xacm = 0, cxy = 0, cxuv = 0;
73 	/* horizontal scaler */
74 	u32 xim = 0, xp = 0, xsci =0;
75 	/* vertical scale & gain */
76 	u32 pfuv = 0;
77 
78 	/* helper variables */
79 	u32 h_atten = 0, i = 0;
80 
81 	if ( 0 == out_x ) {
82 		return -EINVAL;
83 	}
84 
85 	/* mask out vanity-bit */
86 	*hps_ctrl &= ~MASK_29;
87 
88 	/* calculate prescale-(xspc)-value:	[n   .. 1/2) : 1
89 						[1/2 .. 1/3) : 2
90 						[1/3 .. 1/4) : 3
91 						...		*/
92 	if (in_x > out_x) {
93 		xpsc = in_x / out_x;
94 	}
95 	else {
96 		/* zooming */
97 		xpsc = 1;
98 	}
99 
100 	/* if flip_lr-bit is set, number of pixels after
101 	   horizontal prescaling must be < 384 */
102 	if ( 0 != flip_lr ) {
103 
104 		/* set vanity bit */
105 		*hps_ctrl |= MASK_29;
106 
107 		while (in_x / xpsc >= 384 )
108 			xpsc++;
109 	}
110 	/* if zooming is wanted, number of pixels after
111 	   horizontal prescaling must be < 768 */
112 	else {
113 		while ( in_x / xpsc >= 768 )
114 			xpsc++;
115 	}
116 
117 	/* maximum prescale is 64 (p.69) */
118 	if ( xpsc > 64 )
119 		xpsc = 64;
120 
121 	/* keep xacm clear*/
122 	xacm = 0;
123 
124 	/* set horizontal filter parameters (CXY = CXUV) */
125 	cxy = hps_h_coeff_tab[( (xpsc - 1) < 63 ? (xpsc - 1) : 63 )].hps_coeff;
126 	cxuv = cxy;
127 
128 	/* calculate and set horizontal fine scale (xsci) */
129 
130 	/* bypass the horizontal scaler ? */
131 	if ( (in_x == out_x) && ( 1 == xpsc ) )
132 		xsci = 0x400;
133 	else
134 		xsci = ( (1024 * in_x) / (out_x * xpsc) ) + xpsc;
135 
136 	/* set start phase for horizontal fine scale (xp) to 0 */
137 	xp = 0;
138 
139 	/* set xim, if we bypass the horizontal scaler */
140 	if ( 0x400 == xsci )
141 		xim = 1;
142 	else
143 		xim = 0;
144 
145 	/* if the prescaler is bypassed, enable horizontal
146 	   accumulation mode (xacm) and clear dcgx */
147 	if( 1 == xpsc ) {
148 		xacm = 1;
149 		dcgx = 0;
150 	} else {
151 		xacm = 0;
152 		/* get best match in the table of attenuations
153 		   for horizontal scaling */
154 		h_atten = hps_h_coeff_tab[( (xpsc - 1) < 63 ? (xpsc - 1) : 63 )].weight_sum;
155 
156 		for (i = 0; h_attenuation[i] != 0; i++) {
157 			if (h_attenuation[i] >= h_atten)
158 				break;
159 		}
160 
161 		dcgx = i;
162 	}
163 
164 	/* the horizontal scaling increment controls the UV filter
165 	   to reduce the bandwidth to improve the display quality,
166 	   so set it ... */
167 	if ( xsci == 0x400)
168 		pfuv = 0x00;
169 	else if ( xsci < 0x600)
170 		pfuv = 0x01;
171 	else if ( xsci < 0x680)
172 		pfuv = 0x11;
173 	else if ( xsci < 0x700)
174 		pfuv = 0x22;
175 	else
176 		pfuv = 0x33;
177 
178 
179 	*hps_v_gain  &= MASK_W0|MASK_B2;
180 	*hps_v_gain  |= (pfuv << 24);
181 
182 	*hps_h_scale	&= ~(MASK_W1 | 0xf000);
183 	*hps_h_scale	|= (xim << 31) | (xp << 24) | (xsci << 12);
184 
185 	*hps_h_prescale	|= (dcgx << 27) | ((xpsc-1) << 18) | (xacm << 17) | (cxy << 8) | (cxuv << 0);
186 
187 	return 0;
188 }
189 
190 static struct {
191 	u16 hps_coeff;
192 	u16 weight_sum;
193 } hps_v_coeff_tab [] = {
194  {0x0100,   2},  {0x0102,   4},  {0x0300,   4},  {0x0106,   8},  {0x0502,   8},
195  {0x0708,   8},  {0x0F00,   8},  {0x011E,  16},  {0x110E,  16},  {0x1926,  16},
196  {0x3906,  16},  {0x3D42,  16},  {0x7D02,  16},  {0x7F80,  16},  {0xFF00,  16},
197  {0x01FE,  32},  {0x01FE,  32},  {0x817E,  32},  {0x817E,  32},  {0xC13E,  32},
198  {0xC13E,  32},  {0xE11E,  32},  {0xE11E,  32},  {0xF10E,  32},  {0xF10E,  32},
199  {0xF906,  32},  {0xF906,  32},  {0xFD02,  32},  {0xFD02,  32},  {0xFF00,  32},
200  {0xFF00,  32},  {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},  {0x01FE,  64},
203  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x01FE,  64},  {0x817E,  64},
204  {0x817E,  64},  {0xC13E,  64},  {0xC13E,  64},  {0xE11E,  64},  {0xE11E,  64},
205  {0xF10E,  64},  {0xF10E,  64},  {0xF906,  64},  {0xF906,  64},  {0xFD02,  64},
206  {0xFD02,  64},  {0xFF00,  64},  {0xFF00,  64},  {0x01FE, 128}
207 };
208 
209 /* table of attenuation values for vertical scaling */
210 static u16 v_attenuation[] = { 2, 4, 8, 16, 32, 64, 128, 256, 0};
211 
212 /* calculate vertical scale registers */
213 static int calculate_v_scale_registers(struct saa7146_dev *dev, enum v4l2_field field,
214 	int in_y, int out_y, u32* hps_v_scale, u32* hps_v_gain)
215 {
216 	int lpi = 0;
217 
218 	/* vertical scaling */
219 	u32 yacm = 0, ysci = 0, yacl = 0, ypo = 0, ype = 0;
220 	/* vertical scale & gain */
221 	u32 dcgy = 0, cya_cyb = 0;
222 
223 	/* helper variables */
224 	u32 v_atten = 0, i = 0;
225 
226 	/* error, if vertical zooming */
227 	if ( in_y < out_y ) {
228 		return -EINVAL;
229 	}
230 
231 	/* linear phase interpolation may be used
232 	   if scaling is between 1 and 1/2 (both fields used)
233 	   or scaling is between 1/2 and 1/4 (if only one field is used) */
234 
235 	if (V4L2_FIELD_HAS_BOTH(field)) {
236 		if( 2*out_y >= in_y) {
237 			lpi = 1;
238 		}
239 	} else if (field == V4L2_FIELD_TOP
240 		|| field == V4L2_FIELD_ALTERNATE
241 		|| field == V4L2_FIELD_BOTTOM) {
242 		if( 4*out_y >= in_y ) {
243 			lpi = 1;
244 		}
245 		out_y *= 2;
246 	}
247 	if( 0 != lpi ) {
248 
249 		yacm = 0;
250 		yacl = 0;
251 		cya_cyb = 0x00ff;
252 
253 		/* calculate scaling increment */
254 		if ( in_y > out_y )
255 			ysci = ((1024 * in_y) / (out_y + 1)) - 1024;
256 		else
257 			ysci = 0;
258 
259 		dcgy = 0;
260 
261 		/* calculate ype and ypo */
262 		ype = ysci / 16;
263 		ypo = ype + (ysci / 64);
264 
265 	} else {
266 		yacm = 1;
267 
268 		/* calculate scaling increment */
269 		ysci = (((10 * 1024 * (in_y - out_y - 1)) / in_y) + 9) / 10;
270 
271 		/* calculate ype and ypo */
272 		ypo = ype = ((ysci + 15) / 16);
273 
274 		/* the sequence length interval (yacl) has to be set according
275 		   to the prescale value, e.g.	[n   .. 1/2) : 0
276 						[1/2 .. 1/3) : 1
277 						[1/3 .. 1/4) : 2
278 						... */
279 		if ( ysci < 512) {
280 			yacl = 0;
281 		} else {
282 			yacl = ( ysci / (1024 - ysci) );
283 		}
284 
285 		/* get filter coefficients for cya, cyb from table hps_v_coeff_tab */
286 		cya_cyb = hps_v_coeff_tab[ (yacl < 63 ? yacl : 63 ) ].hps_coeff;
287 
288 		/* get best match in the table of attenuations for vertical scaling */
289 		v_atten = hps_v_coeff_tab[ (yacl < 63 ? yacl : 63 ) ].weight_sum;
290 
291 		for (i = 0; v_attenuation[i] != 0; i++) {
292 			if (v_attenuation[i] >= v_atten)
293 				break;
294 		}
295 
296 		dcgy = i;
297 	}
298 
299 	/* ypo and ype swapped in spec ? */
300 	*hps_v_scale	|= (yacm << 31) | (ysci << 21) | (yacl << 15) | (ypo << 8 ) | (ype << 1);
301 
302 	*hps_v_gain	&= ~(MASK_W0|MASK_B2);
303 	*hps_v_gain	|= (dcgy << 16) | (cya_cyb << 0);
304 
305 	return 0;
306 }
307 
308 /* simple bubble-sort algorithm with duplicate elimination */
309 static int sort_and_eliminate(u32* values, int* count)
310 {
311 	int low = 0, high = 0, top = 0;
312 	int cur = 0, next = 0;
313 
314 	/* sanity checks */
315 	if( (0 > *count) || (NULL == values) ) {
316 		return -EINVAL;
317 	}
318 
319 	/* bubble sort the first @count items of the array @values */
320 	for( top = *count; top > 0; top--) {
321 		for( low = 0, high = 1; high < top; low++, high++) {
322 			if( values[low] > values[high] )
323 				swap(values[low], values[high]);
324 		}
325 	}
326 
327 	/* remove duplicate items */
328 	for( cur = 0, next = 1; next < *count; next++) {
329 		if( values[cur] != values[next])
330 			values[++cur] = values[next];
331 	}
332 
333 	*count = cur + 1;
334 
335 	return 0;
336 }
337 
338 static void calculate_clipping_registers_rect(struct saa7146_dev *dev, struct saa7146_fh *fh,
339 	struct saa7146_video_dma *vdma2, u32* clip_format, u32* arbtr_ctrl, enum v4l2_field field)
340 {
341 	struct saa7146_vv *vv = dev->vv_data;
342 	__le32 *clipping = vv->d_clipping.cpu_addr;
343 
344 	int width = vv->ov.win.w.width;
345 	int height =  vv->ov.win.w.height;
346 	int clipcount = vv->ov.nclips;
347 
348 	u32 line_list[32];
349 	u32 pixel_list[32];
350 	int numdwords = 0;
351 
352 	int i = 0, j = 0;
353 	int cnt_line = 0, cnt_pixel = 0;
354 
355 	int x[32], y[32], w[32], h[32];
356 
357 	/* clear out memory */
358 	memset(&line_list[0],  0x00, sizeof(u32)*32);
359 	memset(&pixel_list[0], 0x00, sizeof(u32)*32);
360 	memset(clipping,  0x00, SAA7146_CLIPPING_MEM);
361 
362 	/* fill the line and pixel-lists */
363 	for(i = 0; i < clipcount; i++) {
364 		int l = 0, r = 0, t = 0, b = 0;
365 
366 		x[i] = vv->ov.clips[i].c.left;
367 		y[i] = vv->ov.clips[i].c.top;
368 		w[i] = vv->ov.clips[i].c.width;
369 		h[i] = vv->ov.clips[i].c.height;
370 
371 		if( w[i] < 0) {
372 			x[i] += w[i]; w[i] = -w[i];
373 		}
374 		if( h[i] < 0) {
375 			y[i] += h[i]; h[i] = -h[i];
376 		}
377 		if( x[i] < 0) {
378 			w[i] += x[i]; x[i] = 0;
379 		}
380 		if( y[i] < 0) {
381 			h[i] += y[i]; y[i] = 0;
382 		}
383 		if( 0 != vv->vflip ) {
384 			y[i] = height - y[i] - h[i];
385 		}
386 
387 		l = x[i];
388 		r = x[i]+w[i];
389 		t = y[i];
390 		b = y[i]+h[i];
391 
392 		/* insert left/right coordinates */
393 		pixel_list[ 2*i   ] = min_t(int, l, width);
394 		pixel_list[(2*i)+1] = min_t(int, r, width);
395 		/* insert top/bottom coordinates */
396 		line_list[ 2*i   ] = min_t(int, t, height);
397 		line_list[(2*i)+1] = min_t(int, b, height);
398 	}
399 
400 	/* sort and eliminate lists */
401 	cnt_line = cnt_pixel = 2*clipcount;
402 	sort_and_eliminate( &pixel_list[0], &cnt_pixel );
403 	sort_and_eliminate( &line_list[0], &cnt_line );
404 
405 	/* calculate the number of used u32s */
406 	numdwords = max_t(int, (cnt_line+1), (cnt_pixel+1))*2;
407 	numdwords = max_t(int, 4, numdwords);
408 	numdwords = min_t(int, 64, numdwords);
409 
410 	/* fill up cliptable */
411 	for(i = 0; i < cnt_pixel; i++) {
412 		clipping[2*i] |= cpu_to_le32(pixel_list[i] << 16);
413 	}
414 	for(i = 0; i < cnt_line; i++) {
415 		clipping[(2*i)+1] |= cpu_to_le32(line_list[i] << 16);
416 	}
417 
418 	/* fill up cliptable with the display infos */
419 	for(j = 0; j < clipcount; j++) {
420 
421 		for(i = 0; i < cnt_pixel; i++) {
422 
423 			if( x[j] < 0)
424 				x[j] = 0;
425 
426 			if( pixel_list[i] < (x[j] + w[j])) {
427 
428 				if ( pixel_list[i] >= x[j] ) {
429 					clipping[2*i] |= cpu_to_le32(1 << j);
430 				}
431 			}
432 		}
433 		for(i = 0; i < cnt_line; i++) {
434 
435 			if( y[j] < 0)
436 				y[j] = 0;
437 
438 			if( line_list[i] < (y[j] + h[j]) ) {
439 
440 				if( line_list[i] >= y[j] ) {
441 					clipping[(2*i)+1] |= cpu_to_le32(1 << j);
442 				}
443 			}
444 		}
445 	}
446 
447 	/* adjust arbitration control register */
448 	*arbtr_ctrl &= 0xffff00ff;
449 	*arbtr_ctrl |= 0x00001c00;
450 
451 	vdma2->base_even	= vv->d_clipping.dma_handle;
452 	vdma2->base_odd		= vv->d_clipping.dma_handle;
453 	vdma2->prot_addr	= vv->d_clipping.dma_handle+((sizeof(u32))*(numdwords));
454 	vdma2->base_page	= 0x04;
455 	vdma2->pitch		= 0x00;
456 	vdma2->num_line_byte	= (0 << 16 | (sizeof(u32))*(numdwords-1) );
457 
458 	/* set clipping-mode. this depends on the field(s) used */
459 	*clip_format &= 0xfffffff7;
460 	if (V4L2_FIELD_HAS_BOTH(field)) {
461 		*clip_format |= 0x00000008;
462 	} else {
463 		*clip_format |= 0x00000000;
464 	}
465 }
466 
467 /* disable clipping */
468 static void saa7146_disable_clipping(struct saa7146_dev *dev)
469 {
470 	u32 clip_format	= saa7146_read(dev, CLIP_FORMAT_CTRL);
471 
472 	/* mask out relevant bits (=lower word)*/
473 	clip_format &= MASK_W1;
474 
475 	/* upload clipping-registers*/
476 	saa7146_write(dev, CLIP_FORMAT_CTRL,clip_format);
477 	saa7146_write(dev, MC2, (MASK_05 | MASK_21));
478 
479 	/* disable video dma2 */
480 	saa7146_write(dev, MC1, MASK_21);
481 }
482 
483 static void saa7146_set_clipping_rect(struct saa7146_fh *fh)
484 {
485 	struct saa7146_dev *dev = fh->dev;
486 	struct saa7146_vv *vv = dev->vv_data;
487 	enum v4l2_field field = vv->ov.win.field;
488 	struct	saa7146_video_dma vdma2;
489 	u32 clip_format;
490 	u32 arbtr_ctrl;
491 
492 	/* check clipcount, disable clipping if clipcount == 0*/
493 	if (vv->ov.nclips == 0) {
494 		saa7146_disable_clipping(dev);
495 		return;
496 	}
497 
498 	clip_format = saa7146_read(dev, CLIP_FORMAT_CTRL);
499 	arbtr_ctrl = saa7146_read(dev, PCI_BT_V1);
500 
501 	calculate_clipping_registers_rect(dev, fh, &vdma2, &clip_format, &arbtr_ctrl, field);
502 
503 	/* set clipping format */
504 	clip_format &= 0xffff0008;
505 	clip_format |= (SAA7146_CLIPPING_RECT << 4);
506 
507 	/* prepare video dma2 */
508 	saa7146_write(dev, BASE_EVEN2,		vdma2.base_even);
509 	saa7146_write(dev, BASE_ODD2,		vdma2.base_odd);
510 	saa7146_write(dev, PROT_ADDR2,		vdma2.prot_addr);
511 	saa7146_write(dev, BASE_PAGE2,		vdma2.base_page);
512 	saa7146_write(dev, PITCH2,		vdma2.pitch);
513 	saa7146_write(dev, NUM_LINE_BYTE2,	vdma2.num_line_byte);
514 
515 	/* prepare the rest */
516 	saa7146_write(dev, CLIP_FORMAT_CTRL,clip_format);
517 	saa7146_write(dev, PCI_BT_V1, arbtr_ctrl);
518 
519 	/* upload clip_control-register, clipping-registers, enable video dma2 */
520 	saa7146_write(dev, MC2, (MASK_05 | MASK_21 | MASK_03 | MASK_19));
521 	saa7146_write(dev, MC1, (MASK_05 | MASK_21));
522 }
523 
524 static void saa7146_set_window(struct saa7146_dev *dev, int width, int height, enum v4l2_field field)
525 {
526 	struct saa7146_vv *vv = dev->vv_data;
527 
528 	int source = vv->current_hps_source;
529 	int sync = vv->current_hps_sync;
530 
531 	u32 hps_v_scale = 0, hps_v_gain  = 0, hps_ctrl = 0, hps_h_prescale = 0, hps_h_scale = 0;
532 
533 	/* set vertical scale */
534 	hps_v_scale = 0; /* all bits get set by the function-call */
535 	hps_v_gain  = 0; /* fixme: saa7146_read(dev, HPS_V_GAIN);*/
536 	calculate_v_scale_registers(dev, field, vv->standard->v_field*2, height, &hps_v_scale, &hps_v_gain);
537 
538 	/* set horizontal scale */
539 	hps_ctrl	= 0;
540 	hps_h_prescale	= 0; /* all bits get set in the function */
541 	hps_h_scale	= 0;
542 	calculate_h_scale_registers(dev, vv->standard->h_pixels, width, vv->hflip, &hps_ctrl, &hps_v_gain, &hps_h_prescale, &hps_h_scale);
543 
544 	/* set hyo and hxo */
545 	calculate_hxo_and_hyo(vv, &hps_h_scale, &hps_ctrl);
546 	calculate_hps_source_and_sync(dev, source, sync, &hps_ctrl);
547 
548 	/* write out new register contents */
549 	saa7146_write(dev, HPS_V_SCALE,	hps_v_scale);
550 	saa7146_write(dev, HPS_V_GAIN,	hps_v_gain);
551 	saa7146_write(dev, HPS_CTRL,	hps_ctrl);
552 	saa7146_write(dev, HPS_H_PRESCALE,hps_h_prescale);
553 	saa7146_write(dev, HPS_H_SCALE,	hps_h_scale);
554 
555 	/* upload shadow-ram registers */
556 	saa7146_write(dev, MC2, (MASK_05 | MASK_06 | MASK_21 | MASK_22) );
557 }
558 
559 /* calculate the new memory offsets for a desired position */
560 static void saa7146_set_position(struct saa7146_dev *dev, int w_x, int w_y, int w_height, enum v4l2_field field, u32 pixelformat)
561 {
562 	struct saa7146_vv *vv = dev->vv_data;
563 	struct saa7146_format *sfmt = saa7146_format_by_fourcc(dev, pixelformat);
564 
565 	int b_depth = vv->ov_fmt->depth;
566 	int b_bpl = vv->ov_fb.fmt.bytesperline;
567 	/* The unsigned long cast is to remove a 64-bit compile warning since
568 	   it looks like a 64-bit address is cast to a 32-bit value, even
569 	   though the base pointer is really a 32-bit physical address that
570 	   goes into a 32-bit DMA register.
571 	   FIXME: might not work on some 64-bit platforms, but see the FIXME
572 	   in struct v4l2_framebuffer (videodev2.h) for that.
573 	 */
574 	u32 base = (u32)(unsigned long)vv->ov_fb.base;
575 
576 	struct	saa7146_video_dma vdma1;
577 
578 	/* calculate memory offsets for picture, look if we shall top-down-flip */
579 	vdma1.pitch	= 2*b_bpl;
580 	if ( 0 == vv->vflip ) {
581 		vdma1.base_even = base + (w_y * (vdma1.pitch/2)) + (w_x * (b_depth / 8));
582 		vdma1.base_odd  = vdma1.base_even + (vdma1.pitch / 2);
583 		vdma1.prot_addr = vdma1.base_even + (w_height * (vdma1.pitch / 2));
584 	}
585 	else {
586 		vdma1.base_even = base + ((w_y+w_height) * (vdma1.pitch/2)) + (w_x * (b_depth / 8));
587 		vdma1.base_odd  = vdma1.base_even - (vdma1.pitch / 2);
588 		vdma1.prot_addr = vdma1.base_odd - (w_height * (vdma1.pitch / 2));
589 	}
590 
591 	if (V4L2_FIELD_HAS_BOTH(field)) {
592 	} else if (field == V4L2_FIELD_ALTERNATE) {
593 		/* fixme */
594 		vdma1.base_odd = vdma1.prot_addr;
595 		vdma1.pitch /= 2;
596 	} else if (field == V4L2_FIELD_TOP) {
597 		vdma1.base_odd = vdma1.prot_addr;
598 		vdma1.pitch /= 2;
599 	} else if (field == V4L2_FIELD_BOTTOM) {
600 		vdma1.base_odd = vdma1.base_even;
601 		vdma1.base_even = vdma1.prot_addr;
602 		vdma1.pitch /= 2;
603 	}
604 
605 	if ( 0 != vv->vflip ) {
606 		vdma1.pitch *= -1;
607 	}
608 
609 	vdma1.base_page = sfmt->swap;
610 	vdma1.num_line_byte = (vv->standard->v_field<<16)+vv->standard->h_pixels;
611 
612 	saa7146_write_out_dma(dev, 1, &vdma1);
613 }
614 
615 static void saa7146_set_output_format(struct saa7146_dev *dev, unsigned long palette)
616 {
617 	u32 clip_format = saa7146_read(dev, CLIP_FORMAT_CTRL);
618 
619 	/* call helper function */
620 	calculate_output_format_register(dev,palette,&clip_format);
621 
622 	/* update the hps registers */
623 	saa7146_write(dev, CLIP_FORMAT_CTRL, clip_format);
624 	saa7146_write(dev, MC2, (MASK_05 | MASK_21));
625 }
626 
627 /* select input-source */
628 void saa7146_set_hps_source_and_sync(struct saa7146_dev *dev, int source, int sync)
629 {
630 	struct saa7146_vv *vv = dev->vv_data;
631 	u32 hps_ctrl = 0;
632 
633 	/* read old state */
634 	hps_ctrl = saa7146_read(dev, HPS_CTRL);
635 
636 	hps_ctrl &= ~( MASK_31 | MASK_30 | MASK_28 );
637 	hps_ctrl |= (source << 30) | (sync << 28);
638 
639 	/* write back & upload register */
640 	saa7146_write(dev, HPS_CTRL, hps_ctrl);
641 	saa7146_write(dev, MC2, (MASK_05 | MASK_21));
642 
643 	vv->current_hps_source = source;
644 	vv->current_hps_sync = sync;
645 }
646 EXPORT_SYMBOL_GPL(saa7146_set_hps_source_and_sync);
647 
648 int saa7146_enable_overlay(struct saa7146_fh *fh)
649 {
650 	struct saa7146_dev *dev = fh->dev;
651 	struct saa7146_vv *vv = dev->vv_data;
652 
653 	saa7146_set_window(dev, vv->ov.win.w.width, vv->ov.win.w.height, vv->ov.win.field);
654 	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);
655 	saa7146_set_output_format(dev, vv->ov_fmt->trans);
656 	saa7146_set_clipping_rect(fh);
657 
658 	/* enable video dma1 */
659 	saa7146_write(dev, MC1, (MASK_06 | MASK_22));
660 	return 0;
661 }
662 
663 void saa7146_disable_overlay(struct saa7146_fh *fh)
664 {
665 	struct saa7146_dev *dev = fh->dev;
666 
667 	/* disable clipping + video dma1 */
668 	saa7146_disable_clipping(dev);
669 	saa7146_write(dev, MC1, MASK_22);
670 }
671 
672 void saa7146_write_out_dma(struct saa7146_dev* dev, int which, struct saa7146_video_dma* vdma)
673 {
674 	int where = 0;
675 
676 	if( which < 1 || which > 3) {
677 		return;
678 	}
679 
680 	/* calculate starting address */
681 	where  = (which-1)*0x18;
682 
683 	saa7146_write(dev, where,	vdma->base_odd);
684 	saa7146_write(dev, where+0x04,	vdma->base_even);
685 	saa7146_write(dev, where+0x08,	vdma->prot_addr);
686 	saa7146_write(dev, where+0x0c,	vdma->pitch);
687 	saa7146_write(dev, where+0x10,	vdma->base_page);
688 	saa7146_write(dev, where+0x14,	vdma->num_line_byte);
689 
690 	/* upload */
691 	saa7146_write(dev, MC2, (MASK_02<<(which-1))|(MASK_18<<(which-1)));
692 /*
693 	printk("vdma%d.base_even:     0x%08x\n", which,vdma->base_even);
694 	printk("vdma%d.base_odd:      0x%08x\n", which,vdma->base_odd);
695 	printk("vdma%d.prot_addr:     0x%08x\n", which,vdma->prot_addr);
696 	printk("vdma%d.base_page:     0x%08x\n", which,vdma->base_page);
697 	printk("vdma%d.pitch:         0x%08x\n", which,vdma->pitch);
698 	printk("vdma%d.num_line_byte: 0x%08x\n", which,vdma->num_line_byte);
699 */
700 }
701 
702 static int calculate_video_dma_grab_packed(struct saa7146_dev* dev, struct saa7146_buf *buf)
703 {
704 	struct saa7146_vv *vv = dev->vv_data;
705 	struct saa7146_video_dma vdma1;
706 
707 	struct saa7146_format *sfmt = saa7146_format_by_fourcc(dev,buf->fmt->pixelformat);
708 
709 	int width = buf->fmt->width;
710 	int height = buf->fmt->height;
711 	int bytesperline = buf->fmt->bytesperline;
712 	enum v4l2_field field = buf->fmt->field;
713 
714 	int depth = sfmt->depth;
715 
716 	DEB_CAP("[size=%dx%d,fields=%s]\n",
717 		width, height, v4l2_field_names[field]);
718 
719 	if( bytesperline != 0) {
720 		vdma1.pitch = bytesperline*2;
721 	} else {
722 		vdma1.pitch = (width*depth*2)/8;
723 	}
724 	vdma1.num_line_byte	= ((vv->standard->v_field<<16) + vv->standard->h_pixels);
725 	vdma1.base_page		= buf->pt[0].dma | ME1 | sfmt->swap;
726 
727 	if( 0 != vv->vflip ) {
728 		vdma1.prot_addr	= buf->pt[0].offset;
729 		vdma1.base_even	= buf->pt[0].offset+(vdma1.pitch/2)*height;
730 		vdma1.base_odd	= vdma1.base_even - (vdma1.pitch/2);
731 	} else {
732 		vdma1.base_even	= buf->pt[0].offset;
733 		vdma1.base_odd	= vdma1.base_even + (vdma1.pitch/2);
734 		vdma1.prot_addr	= buf->pt[0].offset+(vdma1.pitch/2)*height;
735 	}
736 
737 	if (V4L2_FIELD_HAS_BOTH(field)) {
738 	} else if (field == V4L2_FIELD_ALTERNATE) {
739 		/* fixme */
740 		if ( vv->last_field == V4L2_FIELD_TOP ) {
741 			vdma1.base_odd	= vdma1.prot_addr;
742 			vdma1.pitch /= 2;
743 		} else if ( vv->last_field == V4L2_FIELD_BOTTOM ) {
744 			vdma1.base_odd	= vdma1.base_even;
745 			vdma1.base_even = vdma1.prot_addr;
746 			vdma1.pitch /= 2;
747 		}
748 	} else if (field == V4L2_FIELD_TOP) {
749 		vdma1.base_odd	= vdma1.prot_addr;
750 		vdma1.pitch /= 2;
751 	} else if (field == V4L2_FIELD_BOTTOM) {
752 		vdma1.base_odd	= vdma1.base_even;
753 		vdma1.base_even = vdma1.prot_addr;
754 		vdma1.pitch /= 2;
755 	}
756 
757 	if( 0 != vv->vflip ) {
758 		vdma1.pitch *= -1;
759 	}
760 
761 	saa7146_write_out_dma(dev, 1, &vdma1);
762 	return 0;
763 }
764 
765 static int calc_planar_422(struct saa7146_vv *vv, struct saa7146_buf *buf, struct saa7146_video_dma *vdma2, struct saa7146_video_dma *vdma3)
766 {
767 	int height = buf->fmt->height;
768 	int width = buf->fmt->width;
769 
770 	vdma2->pitch	= width;
771 	vdma3->pitch	= width;
772 
773 	/* fixme: look at bytesperline! */
774 
775 	if( 0 != vv->vflip ) {
776 		vdma2->prot_addr	= buf->pt[1].offset;
777 		vdma2->base_even	= ((vdma2->pitch/2)*height)+buf->pt[1].offset;
778 		vdma2->base_odd		= vdma2->base_even - (vdma2->pitch/2);
779 
780 		vdma3->prot_addr	= buf->pt[2].offset;
781 		vdma3->base_even	= ((vdma3->pitch/2)*height)+buf->pt[2].offset;
782 		vdma3->base_odd		= vdma3->base_even - (vdma3->pitch/2);
783 	} else {
784 		vdma3->base_even	= buf->pt[2].offset;
785 		vdma3->base_odd		= vdma3->base_even + (vdma3->pitch/2);
786 		vdma3->prot_addr	= (vdma3->pitch/2)*height+buf->pt[2].offset;
787 
788 		vdma2->base_even	= buf->pt[1].offset;
789 		vdma2->base_odd		= vdma2->base_even + (vdma2->pitch/2);
790 		vdma2->prot_addr	= (vdma2->pitch/2)*height+buf->pt[1].offset;
791 	}
792 
793 	return 0;
794 }
795 
796 static int calc_planar_420(struct saa7146_vv *vv, struct saa7146_buf *buf, struct saa7146_video_dma *vdma2, struct saa7146_video_dma *vdma3)
797 {
798 	int height = buf->fmt->height;
799 	int width = buf->fmt->width;
800 
801 	vdma2->pitch	= width/2;
802 	vdma3->pitch	= width/2;
803 
804 	if( 0 != vv->vflip ) {
805 		vdma2->prot_addr	= buf->pt[2].offset;
806 		vdma2->base_even	= ((vdma2->pitch/2)*height)+buf->pt[2].offset;
807 		vdma2->base_odd		= vdma2->base_even - (vdma2->pitch/2);
808 
809 		vdma3->prot_addr	= buf->pt[1].offset;
810 		vdma3->base_even	= ((vdma3->pitch/2)*height)+buf->pt[1].offset;
811 		vdma3->base_odd		= vdma3->base_even - (vdma3->pitch/2);
812 
813 	} else {
814 		vdma3->base_even	= buf->pt[2].offset;
815 		vdma3->base_odd		= vdma3->base_even + (vdma3->pitch);
816 		vdma3->prot_addr	= (vdma3->pitch/2)*height+buf->pt[2].offset;
817 
818 		vdma2->base_even	= buf->pt[1].offset;
819 		vdma2->base_odd		= vdma2->base_even + (vdma2->pitch);
820 		vdma2->prot_addr	= (vdma2->pitch/2)*height+buf->pt[1].offset;
821 	}
822 	return 0;
823 }
824 
825 static int calculate_video_dma_grab_planar(struct saa7146_dev* dev, struct saa7146_buf *buf)
826 {
827 	struct saa7146_vv *vv = dev->vv_data;
828 	struct saa7146_video_dma vdma1;
829 	struct saa7146_video_dma vdma2;
830 	struct saa7146_video_dma vdma3;
831 
832 	struct saa7146_format *sfmt = saa7146_format_by_fourcc(dev,buf->fmt->pixelformat);
833 
834 	int width = buf->fmt->width;
835 	int height = buf->fmt->height;
836 	enum v4l2_field field = buf->fmt->field;
837 
838 	BUG_ON(0 == buf->pt[0].dma);
839 	BUG_ON(0 == buf->pt[1].dma);
840 	BUG_ON(0 == buf->pt[2].dma);
841 
842 	DEB_CAP("[size=%dx%d,fields=%s]\n",
843 		width, height, v4l2_field_names[field]);
844 
845 	/* fixme: look at bytesperline! */
846 
847 	/* fixme: what happens for user space buffers here?. The offsets are
848 	   most likely wrong, this version here only works for page-aligned
849 	   buffers, modifications to the pagetable-functions are necessary...*/
850 
851 	vdma1.pitch		= width*2;
852 	vdma1.num_line_byte	= ((vv->standard->v_field<<16) + vv->standard->h_pixels);
853 	vdma1.base_page		= buf->pt[0].dma | ME1;
854 
855 	if( 0 != vv->vflip ) {
856 		vdma1.prot_addr	= buf->pt[0].offset;
857 		vdma1.base_even	= ((vdma1.pitch/2)*height)+buf->pt[0].offset;
858 		vdma1.base_odd	= vdma1.base_even - (vdma1.pitch/2);
859 	} else {
860 		vdma1.base_even	= buf->pt[0].offset;
861 		vdma1.base_odd	= vdma1.base_even + (vdma1.pitch/2);
862 		vdma1.prot_addr	= (vdma1.pitch/2)*height+buf->pt[0].offset;
863 	}
864 
865 	vdma2.num_line_byte	= 0; /* unused */
866 	vdma2.base_page		= buf->pt[1].dma | ME1;
867 
868 	vdma3.num_line_byte	= 0; /* unused */
869 	vdma3.base_page		= buf->pt[2].dma | ME1;
870 
871 	switch( sfmt->depth ) {
872 		case 12: {
873 			calc_planar_420(vv,buf,&vdma2,&vdma3);
874 			break;
875 		}
876 		case 16: {
877 			calc_planar_422(vv,buf,&vdma2,&vdma3);
878 			break;
879 		}
880 		default: {
881 			return -1;
882 		}
883 	}
884 
885 	if (V4L2_FIELD_HAS_BOTH(field)) {
886 	} else if (field == V4L2_FIELD_ALTERNATE) {
887 		/* fixme */
888 		vdma1.base_odd	= vdma1.prot_addr;
889 		vdma1.pitch /= 2;
890 		vdma2.base_odd	= vdma2.prot_addr;
891 		vdma2.pitch /= 2;
892 		vdma3.base_odd	= vdma3.prot_addr;
893 		vdma3.pitch /= 2;
894 	} else if (field == V4L2_FIELD_TOP) {
895 		vdma1.base_odd	= vdma1.prot_addr;
896 		vdma1.pitch /= 2;
897 		vdma2.base_odd	= vdma2.prot_addr;
898 		vdma2.pitch /= 2;
899 		vdma3.base_odd	= vdma3.prot_addr;
900 		vdma3.pitch /= 2;
901 	} else if (field == V4L2_FIELD_BOTTOM) {
902 		vdma1.base_odd	= vdma1.base_even;
903 		vdma1.base_even = vdma1.prot_addr;
904 		vdma1.pitch /= 2;
905 		vdma2.base_odd	= vdma2.base_even;
906 		vdma2.base_even = vdma2.prot_addr;
907 		vdma2.pitch /= 2;
908 		vdma3.base_odd	= vdma3.base_even;
909 		vdma3.base_even = vdma3.prot_addr;
910 		vdma3.pitch /= 2;
911 	}
912 
913 	if( 0 != vv->vflip ) {
914 		vdma1.pitch *= -1;
915 		vdma2.pitch *= -1;
916 		vdma3.pitch *= -1;
917 	}
918 
919 	saa7146_write_out_dma(dev, 1, &vdma1);
920 	if( (sfmt->flags & FORMAT_BYTE_SWAP) != 0 ) {
921 		saa7146_write_out_dma(dev, 3, &vdma2);
922 		saa7146_write_out_dma(dev, 2, &vdma3);
923 	} else {
924 		saa7146_write_out_dma(dev, 2, &vdma2);
925 		saa7146_write_out_dma(dev, 3, &vdma3);
926 	}
927 	return 0;
928 }
929 
930 static void program_capture_engine(struct saa7146_dev *dev, int planar)
931 {
932 	struct saa7146_vv *vv = dev->vv_data;
933 	int count = 0;
934 
935 	unsigned long e_wait = vv->current_hps_sync == SAA7146_HPS_SYNC_PORT_A ? CMD_E_FID_A : CMD_E_FID_B;
936 	unsigned long o_wait = vv->current_hps_sync == SAA7146_HPS_SYNC_PORT_A ? CMD_O_FID_A : CMD_O_FID_B;
937 
938 	/* wait for o_fid_a/b / e_fid_a/b toggle only if rps register 0 is not set*/
939 	WRITE_RPS0(CMD_PAUSE | CMD_OAN | CMD_SIG0 | o_wait);
940 	WRITE_RPS0(CMD_PAUSE | CMD_OAN | CMD_SIG0 | e_wait);
941 
942 	/* set rps register 0 */
943 	WRITE_RPS0(CMD_WR_REG | (1 << 8) | (MC2/4));
944 	WRITE_RPS0(MASK_27 | MASK_11);
945 
946 	/* turn on video-dma1 */
947 	WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
948 	WRITE_RPS0(MASK_06 | MASK_22);			/* => mask */
949 	WRITE_RPS0(MASK_06 | MASK_22);			/* => values */
950 	if( 0 != planar ) {
951 		/* turn on video-dma2 */
952 		WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
953 		WRITE_RPS0(MASK_05 | MASK_21);			/* => mask */
954 		WRITE_RPS0(MASK_05 | MASK_21);			/* => values */
955 
956 		/* turn on video-dma3 */
957 		WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
958 		WRITE_RPS0(MASK_04 | MASK_20);			/* => mask */
959 		WRITE_RPS0(MASK_04 | MASK_20);			/* => values */
960 	}
961 
962 	/* wait for o_fid_a/b / e_fid_a/b toggle */
963 	if ( vv->last_field == V4L2_FIELD_INTERLACED ) {
964 		WRITE_RPS0(CMD_PAUSE | o_wait);
965 		WRITE_RPS0(CMD_PAUSE | e_wait);
966 	} else if ( vv->last_field == V4L2_FIELD_TOP ) {
967 		WRITE_RPS0(CMD_PAUSE | (vv->current_hps_sync == SAA7146_HPS_SYNC_PORT_A ? MASK_10 : MASK_09));
968 		WRITE_RPS0(CMD_PAUSE | o_wait);
969 	} else if ( vv->last_field == V4L2_FIELD_BOTTOM ) {
970 		WRITE_RPS0(CMD_PAUSE | (vv->current_hps_sync == SAA7146_HPS_SYNC_PORT_A ? MASK_10 : MASK_09));
971 		WRITE_RPS0(CMD_PAUSE | e_wait);
972 	}
973 
974 	/* turn off video-dma1 */
975 	WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
976 	WRITE_RPS0(MASK_22 | MASK_06);			/* => mask */
977 	WRITE_RPS0(MASK_22);				/* => values */
978 	if( 0 != planar ) {
979 		/* turn off video-dma2 */
980 		WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
981 		WRITE_RPS0(MASK_05 | MASK_21);			/* => mask */
982 		WRITE_RPS0(MASK_21);				/* => values */
983 
984 		/* turn off video-dma3 */
985 		WRITE_RPS0(CMD_WR_REG_MASK | (MC1/4));
986 		WRITE_RPS0(MASK_04 | MASK_20);			/* => mask */
987 		WRITE_RPS0(MASK_20);				/* => values */
988 	}
989 
990 	/* generate interrupt */
991 	WRITE_RPS0(CMD_INTERRUPT);
992 
993 	/* stop */
994 	WRITE_RPS0(CMD_STOP);
995 }
996 
997 void saa7146_set_capture(struct saa7146_dev *dev, struct saa7146_buf *buf, struct saa7146_buf *next)
998 {
999 	struct saa7146_format *sfmt = saa7146_format_by_fourcc(dev,buf->fmt->pixelformat);
1000 	struct saa7146_vv *vv = dev->vv_data;
1001 	u32 vdma1_prot_addr;
1002 
1003 	DEB_CAP("buf:%p, next:%p\n", buf, next);
1004 
1005 	vdma1_prot_addr = saa7146_read(dev, PROT_ADDR1);
1006 	if( 0 == vdma1_prot_addr ) {
1007 		/* clear out beginning of streaming bit (rps register 0)*/
1008 		DEB_CAP("forcing sync to new frame\n");
1009 		saa7146_write(dev, MC2, MASK_27 );
1010 	}
1011 
1012 	saa7146_set_window(dev, buf->fmt->width, buf->fmt->height, buf->fmt->field);
1013 	saa7146_set_output_format(dev, sfmt->trans);
1014 	saa7146_disable_clipping(dev);
1015 
1016 	if ( vv->last_field == V4L2_FIELD_INTERLACED ) {
1017 	} else if ( vv->last_field == V4L2_FIELD_TOP ) {
1018 		vv->last_field = V4L2_FIELD_BOTTOM;
1019 	} else if ( vv->last_field == V4L2_FIELD_BOTTOM ) {
1020 		vv->last_field = V4L2_FIELD_TOP;
1021 	}
1022 
1023 	if( 0 != IS_PLANAR(sfmt->trans)) {
1024 		calculate_video_dma_grab_planar(dev, buf);
1025 		program_capture_engine(dev,1);
1026 	} else {
1027 		calculate_video_dma_grab_packed(dev, buf);
1028 		program_capture_engine(dev,0);
1029 	}
1030 
1031 /*
1032 	printk("vdma%d.base_even:     0x%08x\n", 1,saa7146_read(dev,BASE_EVEN1));
1033 	printk("vdma%d.base_odd:      0x%08x\n", 1,saa7146_read(dev,BASE_ODD1));
1034 	printk("vdma%d.prot_addr:     0x%08x\n", 1,saa7146_read(dev,PROT_ADDR1));
1035 	printk("vdma%d.base_page:     0x%08x\n", 1,saa7146_read(dev,BASE_PAGE1));
1036 	printk("vdma%d.pitch:         0x%08x\n", 1,saa7146_read(dev,PITCH1));
1037 	printk("vdma%d.num_line_byte: 0x%08x\n", 1,saa7146_read(dev,NUM_LINE_BYTE1));
1038 	printk("vdma%d => vptr      : 0x%08x\n", 1,saa7146_read(dev,PCI_VDP1));
1039 */
1040 
1041 	/* write the address of the rps-program */
1042 	saa7146_write(dev, RPS_ADDR0, dev->d_rps0.dma_handle);
1043 
1044 	/* turn on rps */
1045 	saa7146_write(dev, MC1, (MASK_12 | MASK_28));
1046 }
1047