xref: /linux/drivers/gpu/drm/i915/display/intel_overlay.c (revision bba2c3615bd6cfee7456d1130f2e6b01b3f4e9ba)
1 /*
2  * Copyright © 2009
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining a
5  * copy of this software and associated documentation files (the "Software"),
6  * to deal in the Software without restriction, including without limitation
7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8  * and/or sell copies of the Software, and to permit persons to whom the
9  * Software is furnished to do so, subject to the following conditions:
10  *
11  * The above copyright notice and this permission notice (including the next
12  * paragraph) shall be included in all copies or substantial portions of the
13  * Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21  * SOFTWARE.
22  *
23  * Authors:
24  *    Daniel Vetter <daniel@ffwll.ch>
25  *
26  * Derived from Xorg ddx, xf86-video-intel, src/i830_video.c
27  */
28 
29 #include <drm/drm_fourcc.h>
30 #include <drm/drm_gem.h>
31 #include <drm/drm_print.h>
32 
33 #include "intel_color_regs.h"
34 #include "intel_de.h"
35 #include "intel_display_regs.h"
36 #include "intel_display_types.h"
37 #include "intel_frontbuffer.h"
38 #include "intel_overlay.h"
39 #include "intel_parent.h"
40 #include "intel_pfit_regs.h"
41 
42 /* Limits for overlay size. According to intel doc, the real limits are:
43  * Y width: 4095, UV width (planar): 2047, Y height: 2047,
44  * UV width (planar): * 1023. But the xorg thinks 2048 for height and width. Use
45  * the minimum of both.
46  */
47 #define IMAGE_MAX_WIDTH		2048
48 #define IMAGE_MAX_HEIGHT	2046 /* 2 * 1023 */
49 /* on 830 and 845 these large limits result in the card hanging */
50 #define IMAGE_MAX_WIDTH_LEGACY	1024
51 #define IMAGE_MAX_HEIGHT_LEGACY	1088
52 
53 /* overlay register definitions */
54 /* OCMD register */
55 #define OCMD_TILED_SURFACE	(0x1<<19)
56 #define OCMD_MIRROR_MASK	(0x3<<17)
57 #define OCMD_MIRROR_MODE	(0x3<<17)
58 #define OCMD_MIRROR_HORIZONTAL	(0x1<<17)
59 #define OCMD_MIRROR_VERTICAL	(0x2<<17)
60 #define OCMD_MIRROR_BOTH	(0x3<<17)
61 #define OCMD_BYTEORDER_MASK	(0x3<<14) /* zero for YUYV or FOURCC YUY2 */
62 #define OCMD_UV_SWAP		(0x1<<14) /* YVYU */
63 #define OCMD_Y_SWAP		(0x2<<14) /* UYVY or FOURCC UYVY */
64 #define OCMD_Y_AND_UV_SWAP	(0x3<<14) /* VYUY */
65 #define OCMD_SOURCE_FORMAT_MASK (0xf<<10)
66 #define OCMD_RGB_888		(0x1<<10) /* not in i965 Intel docs */
67 #define OCMD_RGB_555		(0x2<<10) /* not in i965 Intel docs */
68 #define OCMD_RGB_565		(0x3<<10) /* not in i965 Intel docs */
69 #define OCMD_YUV_422_PACKED	(0x8<<10)
70 #define OCMD_YUV_411_PACKED	(0x9<<10) /* not in i965 Intel docs */
71 #define OCMD_YUV_420_PLANAR	(0xc<<10)
72 #define OCMD_YUV_422_PLANAR	(0xd<<10)
73 #define OCMD_YUV_410_PLANAR	(0xe<<10) /* also 411 */
74 #define OCMD_TVSYNCFLIP_PARITY	(0x1<<9)
75 #define OCMD_TVSYNCFLIP_ENABLE	(0x1<<7)
76 #define OCMD_BUF_TYPE_MASK	(0x1<<5)
77 #define OCMD_BUF_TYPE_FRAME	(0x0<<5)
78 #define OCMD_BUF_TYPE_FIELD	(0x1<<5)
79 #define OCMD_TEST_MODE		(0x1<<4)
80 #define OCMD_BUFFER_SELECT	(0x3<<2)
81 #define OCMD_BUFFER0		(0x0<<2)
82 #define OCMD_BUFFER1		(0x1<<2)
83 #define OCMD_FIELD_SELECT	(0x1<<2)
84 #define OCMD_FIELD0		(0x0<<1)
85 #define OCMD_FIELD1		(0x1<<1)
86 #define OCMD_ENABLE		(0x1<<0)
87 
88 /* OCONFIG register */
89 #define OCONF_PIPE_MASK		(0x1<<18)
90 #define OCONF_PIPE_A		(0x0<<18)
91 #define OCONF_PIPE_B		(0x1<<18)
92 #define OCONF_GAMMA2_ENABLE	(0x1<<16)
93 #define OCONF_CSC_MODE_BT601	(0x0<<5)
94 #define OCONF_CSC_MODE_BT709	(0x1<<5)
95 #define OCONF_CSC_BYPASS	(0x1<<4)
96 #define OCONF_CC_OUT_8BIT	(0x1<<3)
97 #define OCONF_TEST_MODE		(0x1<<2)
98 #define OCONF_THREE_LINE_BUFFER	(0x1<<0)
99 #define OCONF_TWO_LINE_BUFFER	(0x0<<0)
100 
101 /* DCLRKM (dst-key) register */
102 #define DST_KEY_ENABLE		(0x1<<31)
103 #define CLK_RGB24_MASK		0x0
104 #define CLK_RGB16_MASK		0x070307
105 #define CLK_RGB15_MASK		0x070707
106 
107 #define RGB30_TO_COLORKEY(c) \
108 	((((c) & 0x3fc00000) >> 6) | (((c) & 0x000ff000) >> 4) | (((c) & 0x000003fc) >> 2))
109 #define RGB16_TO_COLORKEY(c) \
110 	((((c) & 0xf800) << 8) | (((c) & 0x07e0) << 5) | (((c) & 0x001f) << 3))
111 #define RGB15_TO_COLORKEY(c) \
112 	((((c) & 0x7c00) << 9) | (((c) & 0x03e0) << 6) | (((c) & 0x001f) << 3))
113 #define RGB8I_TO_COLORKEY(c) \
114 	((((c) & 0xff) << 16) | (((c) & 0xff) << 8) | (((c) & 0xff) << 0))
115 
116 /* polyphase filter coefficients */
117 #define N_HORIZ_Y_TAPS          5
118 #define N_VERT_Y_TAPS           3
119 #define N_HORIZ_UV_TAPS         3
120 #define N_VERT_UV_TAPS          3
121 #define N_PHASES                17
122 #define MAX_TAPS                5
123 
124 /* memory bufferd overlay registers */
125 struct overlay_registers {
126 	u32 OBUF_0Y;
127 	u32 OBUF_1Y;
128 	u32 OBUF_0U;
129 	u32 OBUF_0V;
130 	u32 OBUF_1U;
131 	u32 OBUF_1V;
132 	u32 OSTRIDE;
133 	u32 YRGB_VPH;
134 	u32 UV_VPH;
135 	u32 HORZ_PH;
136 	u32 INIT_PHS;
137 	u32 DWINPOS;
138 	u32 DWINSZ;
139 	u32 SWIDTH;
140 	u32 SWIDTHSW;
141 	u32 SHEIGHT;
142 	u32 YRGBSCALE;
143 	u32 UVSCALE;
144 	u32 OCLRC0;
145 	u32 OCLRC1;
146 	u32 DCLRKV;
147 	u32 DCLRKM;
148 	u32 SCLRKVH;
149 	u32 SCLRKVL;
150 	u32 SCLRKEN;
151 	u32 OCONFIG;
152 	u32 OCMD;
153 	u32 RESERVED1; /* 0x6C */
154 	u32 OSTART_0Y;
155 	u32 OSTART_1Y;
156 	u32 OSTART_0U;
157 	u32 OSTART_0V;
158 	u32 OSTART_1U;
159 	u32 OSTART_1V;
160 	u32 OTILEOFF_0Y;
161 	u32 OTILEOFF_1Y;
162 	u32 OTILEOFF_0U;
163 	u32 OTILEOFF_0V;
164 	u32 OTILEOFF_1U;
165 	u32 OTILEOFF_1V;
166 	u32 FASTHSCALE; /* 0xA0 */
167 	u32 UVSCALEV; /* 0xA4 */
168 	u32 RESERVEDC[(0x200 - 0xA8) / 4]; /* 0xA8 - 0x1FC */
169 	u16 Y_VCOEFS[N_VERT_Y_TAPS * N_PHASES]; /* 0x200 */
170 	u16 RESERVEDD[0x100 / 2 - N_VERT_Y_TAPS * N_PHASES];
171 	u16 Y_HCOEFS[N_HORIZ_Y_TAPS * N_PHASES]; /* 0x300 */
172 	u16 RESERVEDE[0x200 / 2 - N_HORIZ_Y_TAPS * N_PHASES];
173 	u16 UV_VCOEFS[N_VERT_UV_TAPS * N_PHASES]; /* 0x500 */
174 	u16 RESERVEDF[0x100 / 2 - N_VERT_UV_TAPS * N_PHASES];
175 	u16 UV_HCOEFS[N_HORIZ_UV_TAPS * N_PHASES]; /* 0x600 */
176 	u16 RESERVEDG[0x100 / 2 - N_HORIZ_UV_TAPS * N_PHASES];
177 };
178 
179 struct intel_overlay {
180 	struct intel_display *display;
181 	struct intel_crtc *crtc;
182 	bool pfit_active;
183 	u32 pfit_vscale_ratio; /* shifted-point number, (1<<12) == 1.0 */
184 	u32 color_key:24;
185 	u32 color_key_enabled:1;
186 	u32 brightness, contrast, saturation;
187 	u32 old_xscale, old_yscale;
188 	struct overlay_registers __iomem *regs;
189 };
190 
191 void intel_overlay_reset(struct intel_display *display)
192 {
193 	struct intel_overlay *overlay = display->overlay;
194 
195 	if (!overlay)
196 		return;
197 
198 	overlay->old_xscale = 0;
199 	overlay->old_yscale = 0;
200 	overlay->crtc = NULL;
201 
202 	intel_parent_overlay_reset(display);
203 }
204 
205 static int packed_depth_bytes(u32 format)
206 {
207 	switch (format & I915_OVERLAY_DEPTH_MASK) {
208 	case I915_OVERLAY_YUV422:
209 		return 4;
210 	case I915_OVERLAY_YUV411:
211 		/* return 6; not implemented */
212 	default:
213 		return -EINVAL;
214 	}
215 }
216 
217 static int packed_width_bytes(u32 format, short width)
218 {
219 	switch (format & I915_OVERLAY_DEPTH_MASK) {
220 	case I915_OVERLAY_YUV422:
221 		return width << 1;
222 	default:
223 		return -EINVAL;
224 	}
225 }
226 
227 static int uv_hsubsampling(u32 format)
228 {
229 	switch (format & I915_OVERLAY_DEPTH_MASK) {
230 	case I915_OVERLAY_YUV422:
231 	case I915_OVERLAY_YUV420:
232 		return 2;
233 	case I915_OVERLAY_YUV411:
234 	case I915_OVERLAY_YUV410:
235 		return 4;
236 	default:
237 		return -EINVAL;
238 	}
239 }
240 
241 static int uv_vsubsampling(u32 format)
242 {
243 	switch (format & I915_OVERLAY_DEPTH_MASK) {
244 	case I915_OVERLAY_YUV420:
245 	case I915_OVERLAY_YUV410:
246 		return 2;
247 	case I915_OVERLAY_YUV422:
248 	case I915_OVERLAY_YUV411:
249 		return 1;
250 	default:
251 		return -EINVAL;
252 	}
253 }
254 
255 static u32 calc_swidthsw(struct intel_display *display, u32 offset, u32 width)
256 {
257 	u32 sw;
258 
259 	if (DISPLAY_VER(display) == 2)
260 		sw = ALIGN((offset & 31) + width, 32);
261 	else
262 		sw = ALIGN((offset & 63) + width, 64);
263 
264 	if (sw == 0)
265 		return 0;
266 
267 	return (sw - 32) >> 3;
268 }
269 
270 static const u16 y_static_hcoeffs[N_PHASES][N_HORIZ_Y_TAPS] = {
271 	[ 0] = { 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0, },
272 	[ 1] = { 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440, },
273 	[ 2] = { 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0, },
274 	[ 3] = { 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380, },
275 	[ 4] = { 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320, },
276 	[ 5] = { 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0, },
277 	[ 6] = { 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260, },
278 	[ 7] = { 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200, },
279 	[ 8] = { 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0, },
280 	[ 9] = { 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160, },
281 	[10] = { 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120, },
282 	[11] = { 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0, },
283 	[12] = { 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0, },
284 	[13] = { 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060, },
285 	[14] = { 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040, },
286 	[15] = { 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020, },
287 	[16] = { 0xb000, 0x3000, 0x0800, 0x3000, 0xb000, },
288 };
289 
290 static const u16 uv_static_hcoeffs[N_PHASES][N_HORIZ_UV_TAPS] = {
291 	[ 0] = { 0x3000, 0x1800, 0x1800, },
292 	[ 1] = { 0xb000, 0x18d0, 0x2e60, },
293 	[ 2] = { 0xb000, 0x1990, 0x2ce0, },
294 	[ 3] = { 0xb020, 0x1a68, 0x2b40, },
295 	[ 4] = { 0xb040, 0x1b20, 0x29e0, },
296 	[ 5] = { 0xb060, 0x1bd8, 0x2880, },
297 	[ 6] = { 0xb080, 0x1c88, 0x3e60, },
298 	[ 7] = { 0xb0a0, 0x1d28, 0x3c00, },
299 	[ 8] = { 0xb0c0, 0x1db8, 0x39e0, },
300 	[ 9] = { 0xb0e0, 0x1e40, 0x37e0, },
301 	[10] = { 0xb100, 0x1eb8, 0x3620, },
302 	[11] = { 0xb100, 0x1f18, 0x34a0, },
303 	[12] = { 0xb100, 0x1f68, 0x3360, },
304 	[13] = { 0xb0e0, 0x1fa8, 0x3240, },
305 	[14] = { 0xb0c0, 0x1fe0, 0x3140, },
306 	[15] = { 0xb060, 0x1ff0, 0x30a0, },
307 	[16] = { 0x3000, 0x0800, 0x3000, },
308 };
309 
310 static void update_polyphase_filter(struct overlay_registers __iomem *regs)
311 {
312 	memcpy_toio(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
313 	memcpy_toio(regs->UV_HCOEFS, uv_static_hcoeffs,
314 		    sizeof(uv_static_hcoeffs));
315 }
316 
317 static bool update_scaling_factors(struct intel_overlay *overlay,
318 				   struct overlay_registers __iomem *regs,
319 				   struct drm_intel_overlay_put_image *params)
320 {
321 	/* fixed point with a 12 bit shift */
322 	u32 xscale, yscale, xscale_UV, yscale_UV;
323 #define FP_SHIFT 12
324 #define FRACT_MASK 0xfff
325 	bool scale_changed = false;
326 	int uv_hscale = uv_hsubsampling(params->flags);
327 	int uv_vscale = uv_vsubsampling(params->flags);
328 
329 	if (params->dst_width > 1)
330 		xscale = ((params->src_scan_width - 1) << FP_SHIFT) /
331 			params->dst_width;
332 	else
333 		xscale = 1 << FP_SHIFT;
334 
335 	if (params->dst_height > 1)
336 		yscale = ((params->src_scan_height - 1) << FP_SHIFT) /
337 			params->dst_height;
338 	else
339 		yscale = 1 << FP_SHIFT;
340 
341 	/*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
342 	xscale_UV = xscale/uv_hscale;
343 	yscale_UV = yscale/uv_vscale;
344 	/* make the Y scale to UV scale ratio an exact multiply */
345 	xscale = xscale_UV * uv_hscale;
346 	yscale = yscale_UV * uv_vscale;
347 	/*} else {
348 	  xscale_UV = 0;
349 	  yscale_UV = 0;
350 	  }*/
351 
352 	if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
353 		scale_changed = true;
354 	overlay->old_xscale = xscale;
355 	overlay->old_yscale = yscale;
356 
357 	iowrite32(((yscale & FRACT_MASK) << 20) |
358 		  ((xscale >> FP_SHIFT)  << 16) |
359 		  ((xscale & FRACT_MASK) << 3),
360 		 &regs->YRGBSCALE);
361 
362 	iowrite32(((yscale_UV & FRACT_MASK) << 20) |
363 		  ((xscale_UV >> FP_SHIFT)  << 16) |
364 		  ((xscale_UV & FRACT_MASK) << 3),
365 		 &regs->UVSCALE);
366 
367 	iowrite32((((yscale    >> FP_SHIFT) << 16) |
368 		   ((yscale_UV >> FP_SHIFT) << 0)),
369 		 &regs->UVSCALEV);
370 
371 	if (scale_changed)
372 		update_polyphase_filter(regs);
373 
374 	return scale_changed;
375 }
376 
377 static void update_colorkey(struct intel_overlay *overlay,
378 			    struct overlay_registers __iomem *regs)
379 {
380 	const struct intel_plane_state *state =
381 		to_intel_plane_state(overlay->crtc->base.primary->state);
382 	u32 key = overlay->color_key;
383 	u32 format = 0;
384 	u32 flags = 0;
385 
386 	if (overlay->color_key_enabled)
387 		flags |= DST_KEY_ENABLE;
388 
389 	if (state->uapi.visible)
390 		format = state->hw.fb->format->format;
391 
392 	switch (format) {
393 	case DRM_FORMAT_C8:
394 		key = RGB8I_TO_COLORKEY(key);
395 		flags |= CLK_RGB24_MASK;
396 		break;
397 	case DRM_FORMAT_XRGB1555:
398 		key = RGB15_TO_COLORKEY(key);
399 		flags |= CLK_RGB15_MASK;
400 		break;
401 	case DRM_FORMAT_RGB565:
402 		key = RGB16_TO_COLORKEY(key);
403 		flags |= CLK_RGB16_MASK;
404 		break;
405 	case DRM_FORMAT_XRGB2101010:
406 	case DRM_FORMAT_XBGR2101010:
407 		key = RGB30_TO_COLORKEY(key);
408 		flags |= CLK_RGB24_MASK;
409 		break;
410 	default:
411 		flags |= CLK_RGB24_MASK;
412 		break;
413 	}
414 
415 	iowrite32(key, &regs->DCLRKV);
416 	iowrite32(flags, &regs->DCLRKM);
417 }
418 
419 static u32 overlay_cmd_reg(struct drm_intel_overlay_put_image *params)
420 {
421 	u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
422 
423 	if (params->flags & I915_OVERLAY_YUV_PLANAR) {
424 		switch (params->flags & I915_OVERLAY_DEPTH_MASK) {
425 		case I915_OVERLAY_YUV422:
426 			cmd |= OCMD_YUV_422_PLANAR;
427 			break;
428 		case I915_OVERLAY_YUV420:
429 			cmd |= OCMD_YUV_420_PLANAR;
430 			break;
431 		case I915_OVERLAY_YUV411:
432 		case I915_OVERLAY_YUV410:
433 			cmd |= OCMD_YUV_410_PLANAR;
434 			break;
435 		}
436 	} else { /* YUV packed */
437 		switch (params->flags & I915_OVERLAY_DEPTH_MASK) {
438 		case I915_OVERLAY_YUV422:
439 			cmd |= OCMD_YUV_422_PACKED;
440 			break;
441 		case I915_OVERLAY_YUV411:
442 			cmd |= OCMD_YUV_411_PACKED;
443 			break;
444 		}
445 
446 		switch (params->flags & I915_OVERLAY_SWAP_MASK) {
447 		case I915_OVERLAY_NO_SWAP:
448 			break;
449 		case I915_OVERLAY_UV_SWAP:
450 			cmd |= OCMD_UV_SWAP;
451 			break;
452 		case I915_OVERLAY_Y_SWAP:
453 			cmd |= OCMD_Y_SWAP;
454 			break;
455 		case I915_OVERLAY_Y_AND_UV_SWAP:
456 			cmd |= OCMD_Y_AND_UV_SWAP;
457 			break;
458 		}
459 	}
460 
461 	return cmd;
462 }
463 
464 static int intel_overlay_do_put_image(struct intel_overlay *overlay,
465 				      struct drm_gem_object *obj,
466 				      struct drm_intel_overlay_put_image *params)
467 {
468 	struct intel_display *display = overlay->display;
469 	struct overlay_registers __iomem *regs = overlay->regs;
470 	u32 swidth, swidthsw, sheight, ostride;
471 	enum pipe pipe = overlay->crtc->pipe;
472 	bool scale_changed = false;
473 	struct i915_vma *vma;
474 	int ret, tmp_width;
475 	u32 tmp, offset;
476 
477 	drm_WARN_ON(display->drm,
478 		    !drm_modeset_is_locked(&display->drm->mode_config.connection_mutex));
479 
480 	ret = intel_parent_overlay_release_old_vid(display);
481 	if (ret != 0)
482 		return ret;
483 
484 	vma = intel_parent_overlay_pin_fb(display, obj, &offset);
485 	if (IS_ERR(vma))
486 		return PTR_ERR(vma);
487 
488 	if (!intel_parent_overlay_is_active(display)) {
489 		const struct intel_crtc_state *crtc_state =
490 			overlay->crtc->config;
491 		u32 oconfig = 0;
492 
493 		if (crtc_state->gamma_enable &&
494 		    crtc_state->gamma_mode == GAMMA_MODE_MODE_8BIT)
495 			oconfig |= OCONF_CC_OUT_8BIT;
496 		if (crtc_state->gamma_enable)
497 			oconfig |= OCONF_GAMMA2_ENABLE;
498 		if (DISPLAY_VER(display) == 4)
499 			oconfig |= OCONF_CSC_MODE_BT709;
500 		oconfig |= pipe == 0 ?
501 			OCONF_PIPE_A : OCONF_PIPE_B;
502 		iowrite32(oconfig, &regs->OCONFIG);
503 
504 		ret = intel_parent_overlay_on(display, INTEL_FRONTBUFFER_OVERLAY(pipe));
505 		if (ret != 0)
506 			goto out_unpin;
507 	}
508 
509 	iowrite32(params->dst_y << 16 | params->dst_x, &regs->DWINPOS);
510 	iowrite32(params->dst_height << 16 | params->dst_width, &regs->DWINSZ);
511 
512 	if (params->flags & I915_OVERLAY_YUV_PACKED)
513 		tmp_width = packed_width_bytes(params->flags,
514 					       params->src_width);
515 	else
516 		tmp_width = params->src_width;
517 
518 	swidth = params->src_width;
519 	swidthsw = calc_swidthsw(display, params->offset_Y, tmp_width);
520 	sheight = params->src_height;
521 	iowrite32(offset + params->offset_Y, &regs->OBUF_0Y);
522 	ostride = params->stride_Y;
523 
524 	if (params->flags & I915_OVERLAY_YUV_PLANAR) {
525 		int uv_hscale = uv_hsubsampling(params->flags);
526 		int uv_vscale = uv_vsubsampling(params->flags);
527 		u32 tmp_U, tmp_V;
528 
529 		swidth |= (params->src_width / uv_hscale) << 16;
530 		sheight |= (params->src_height / uv_vscale) << 16;
531 
532 		tmp_U = calc_swidthsw(display, params->offset_U,
533 				      params->src_width / uv_hscale);
534 		tmp_V = calc_swidthsw(display, params->offset_V,
535 				      params->src_width / uv_hscale);
536 		swidthsw |= max(tmp_U, tmp_V) << 16;
537 
538 		iowrite32(offset + params->offset_U,
539 			  &regs->OBUF_0U);
540 		iowrite32(offset + params->offset_V,
541 			  &regs->OBUF_0V);
542 
543 		ostride |= params->stride_UV << 16;
544 	}
545 
546 	iowrite32(swidth, &regs->SWIDTH);
547 	iowrite32(swidthsw, &regs->SWIDTHSW);
548 	iowrite32(sheight, &regs->SHEIGHT);
549 	iowrite32(ostride, &regs->OSTRIDE);
550 
551 	scale_changed = update_scaling_factors(overlay, regs, params);
552 
553 	update_colorkey(overlay, regs);
554 
555 	iowrite32(overlay_cmd_reg(params), &regs->OCMD);
556 
557 	/* check for underruns */
558 	tmp = intel_de_read(display, DOVSTA);
559 	if (tmp & (1 << 17))
560 		drm_dbg(display->drm, "overlay underrun, DOVSTA: %x\n", tmp);
561 
562 	ret = intel_parent_overlay_continue(display, vma, scale_changed);
563 	if (ret)
564 		goto out_unpin;
565 
566 	return 0;
567 
568 out_unpin:
569 	intel_parent_overlay_unpin_fb(display, vma);
570 
571 	return ret;
572 }
573 
574 int intel_overlay_switch_off(struct intel_overlay *overlay)
575 {
576 	struct intel_display *display = overlay->display;
577 	int ret;
578 
579 	drm_WARN_ON(display->drm,
580 		    !drm_modeset_is_locked(&display->drm->mode_config.connection_mutex));
581 
582 	ret = intel_parent_overlay_recover_from_interrupt(display);
583 	if (ret != 0)
584 		return ret;
585 
586 	if (!intel_parent_overlay_is_active(display))
587 		return 0;
588 
589 	ret = intel_parent_overlay_release_old_vid(display);
590 	if (ret != 0)
591 		return ret;
592 
593 	iowrite32(0, &overlay->regs->OCMD);
594 
595 	overlay->crtc->overlay = NULL;
596 	overlay->crtc = NULL;
597 
598 	return intel_parent_overlay_off(display);
599 }
600 
601 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
602 					  struct intel_crtc *crtc)
603 {
604 	if (!crtc->active)
605 		return -EINVAL;
606 
607 	/* can't use the overlay with double wide pipe */
608 	if (crtc->config->double_wide)
609 		return -EINVAL;
610 
611 	return 0;
612 }
613 
614 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
615 {
616 	struct intel_display *display = overlay->display;
617 	u32 ratio;
618 
619 	/* XXX: This is not the same logic as in the xorg driver, but more in
620 	 * line with the intel documentation for the i965
621 	 */
622 	if (DISPLAY_VER(display) >= 4) {
623 		u32 tmp = intel_de_read(display, PFIT_PGM_RATIOS(display));
624 
625 		/* on i965 use the PGM reg to read out the autoscaler values */
626 		ratio = REG_FIELD_GET(PFIT_VERT_SCALE_MASK_965, tmp);
627 	} else {
628 		u32 tmp;
629 
630 		if (intel_de_read(display, PFIT_CONTROL(display)) & PFIT_VERT_AUTO_SCALE)
631 			tmp = intel_de_read(display, PFIT_AUTO_RATIOS(display));
632 		else
633 			tmp = intel_de_read(display, PFIT_PGM_RATIOS(display));
634 
635 		ratio = REG_FIELD_GET(PFIT_VERT_SCALE_MASK, tmp);
636 	}
637 
638 	overlay->pfit_vscale_ratio = ratio;
639 }
640 
641 static int check_overlay_dst(struct intel_overlay *overlay,
642 			     struct drm_intel_overlay_put_image *rec)
643 {
644 	const struct intel_crtc_state *crtc_state =
645 		overlay->crtc->config;
646 	struct drm_rect req, clipped;
647 
648 	drm_rect_init(&req, rec->dst_x, rec->dst_y,
649 		      rec->dst_width, rec->dst_height);
650 
651 	clipped = req;
652 
653 	if (!drm_rect_intersect(&clipped, &crtc_state->pipe_src))
654 		return -EINVAL;
655 
656 	if (!drm_rect_equals(&clipped, &req))
657 		return -EINVAL;
658 
659 	return 0;
660 }
661 
662 static int check_overlay_scaling(struct drm_intel_overlay_put_image *rec)
663 {
664 	u32 tmp;
665 
666 	/* downscaling limit is 8.0 */
667 	tmp = ((rec->src_scan_height << 16) / rec->dst_height) >> 16;
668 	if (tmp > 7)
669 		return -EINVAL;
670 
671 	tmp = ((rec->src_scan_width << 16) / rec->dst_width) >> 16;
672 	if (tmp > 7)
673 		return -EINVAL;
674 
675 	return 0;
676 }
677 
678 static int check_overlay_src(struct intel_display *display,
679 			     struct drm_intel_overlay_put_image *rec,
680 			     struct drm_gem_object *obj)
681 {
682 	int uv_hscale = uv_hsubsampling(rec->flags);
683 	int uv_vscale = uv_vsubsampling(rec->flags);
684 	u32 stride_mask;
685 	int depth;
686 	u32 tmp;
687 
688 	/* check src dimensions */
689 	if (display->platform.i845g || display->platform.i830) {
690 		if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY ||
691 		    rec->src_width  > IMAGE_MAX_WIDTH_LEGACY)
692 			return -EINVAL;
693 	} else {
694 		if (rec->src_height > IMAGE_MAX_HEIGHT ||
695 		    rec->src_width  > IMAGE_MAX_WIDTH)
696 			return -EINVAL;
697 	}
698 
699 	/* better safe than sorry, use 4 as the maximal subsampling ratio */
700 	if (rec->src_height < N_VERT_Y_TAPS*4 ||
701 	    rec->src_width  < N_HORIZ_Y_TAPS*4)
702 		return -EINVAL;
703 
704 	/* check alignment constraints */
705 	switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
706 	case I915_OVERLAY_RGB:
707 		/* not implemented */
708 		return -EINVAL;
709 
710 	case I915_OVERLAY_YUV_PACKED:
711 		if (uv_vscale != 1)
712 			return -EINVAL;
713 
714 		depth = packed_depth_bytes(rec->flags);
715 		if (depth < 0)
716 			return depth;
717 
718 		/* ignore UV planes */
719 		rec->stride_UV = 0;
720 		rec->offset_U = 0;
721 		rec->offset_V = 0;
722 		/* check pixel alignment */
723 		if (rec->offset_Y % depth)
724 			return -EINVAL;
725 		break;
726 
727 	case I915_OVERLAY_YUV_PLANAR:
728 		if (uv_vscale < 0 || uv_hscale < 0)
729 			return -EINVAL;
730 		/* no offset restrictions for planar formats */
731 		break;
732 
733 	default:
734 		return -EINVAL;
735 	}
736 
737 	if (rec->src_width % uv_hscale)
738 		return -EINVAL;
739 
740 	/* stride checking */
741 	if (display->platform.i830 || display->platform.i845g)
742 		stride_mask = 255;
743 	else
744 		stride_mask = 63;
745 
746 	if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
747 		return -EINVAL;
748 	if (DISPLAY_VER(display) == 4 && rec->stride_Y < 512)
749 		return -EINVAL;
750 
751 	tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
752 		4096 : 8192;
753 	if (rec->stride_Y > tmp || rec->stride_UV > 2*1024)
754 		return -EINVAL;
755 
756 	/* check buffer dimensions */
757 	switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
758 	case I915_OVERLAY_RGB:
759 	case I915_OVERLAY_YUV_PACKED:
760 		/* always 4 Y values per depth pixels */
761 		if (packed_width_bytes(rec->flags, rec->src_width) > rec->stride_Y)
762 			return -EINVAL;
763 
764 		tmp = rec->stride_Y*rec->src_height;
765 		if (rec->offset_Y + tmp > obj->size)
766 			return -EINVAL;
767 		break;
768 
769 	case I915_OVERLAY_YUV_PLANAR:
770 		if (rec->src_width > rec->stride_Y)
771 			return -EINVAL;
772 		if (rec->src_width/uv_hscale > rec->stride_UV)
773 			return -EINVAL;
774 
775 		tmp = rec->stride_Y * rec->src_height;
776 		if (rec->offset_Y + tmp > obj->size)
777 			return -EINVAL;
778 
779 		tmp = rec->stride_UV * (rec->src_height / uv_vscale);
780 		if (rec->offset_U + tmp > obj->size ||
781 		    rec->offset_V + tmp > obj->size)
782 			return -EINVAL;
783 		break;
784 	}
785 
786 	return 0;
787 }
788 
789 int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
790 				  struct drm_file *file_priv)
791 {
792 	struct intel_display *display = to_intel_display(dev);
793 	struct drm_intel_overlay_put_image *params = data;
794 	struct intel_overlay *overlay;
795 	struct drm_crtc *drmmode_crtc;
796 	struct drm_gem_object *obj;
797 	struct intel_crtc *crtc;
798 	int ret;
799 
800 	overlay = display->overlay;
801 	if (!overlay) {
802 		drm_dbg(display->drm, "userspace bug: no overlay\n");
803 		return -ENODEV;
804 	}
805 
806 	if (!(params->flags & I915_OVERLAY_ENABLE)) {
807 		drm_modeset_lock_all(dev);
808 		ret = intel_overlay_switch_off(overlay);
809 		drm_modeset_unlock_all(dev);
810 
811 		return ret;
812 	}
813 
814 	drmmode_crtc = drm_crtc_find(dev, file_priv, params->crtc_id);
815 	if (!drmmode_crtc)
816 		return -ENOENT;
817 	crtc = to_intel_crtc(drmmode_crtc);
818 
819 	obj = intel_parent_overlay_obj_lookup(display, file_priv, params->bo_handle);
820 	if (IS_ERR(obj))
821 		return PTR_ERR(obj);
822 
823 	drm_modeset_lock_all(dev);
824 
825 	ret = intel_parent_overlay_recover_from_interrupt(display);
826 	if (ret != 0)
827 		goto out_unlock;
828 
829 	if (overlay->crtc != crtc) {
830 		ret = intel_overlay_switch_off(overlay);
831 		if (ret != 0)
832 			goto out_unlock;
833 
834 		ret = check_overlay_possible_on_crtc(overlay, crtc);
835 		if (ret != 0)
836 			goto out_unlock;
837 
838 		overlay->crtc = crtc;
839 		crtc->overlay = overlay;
840 
841 		/* line too wide, i.e. one-line-mode */
842 		if (drm_rect_width(&crtc->config->pipe_src) > 1024 &&
843 		    crtc->config->gmch_pfit.control & PFIT_ENABLE) {
844 			overlay->pfit_active = true;
845 			update_pfit_vscale_ratio(overlay);
846 		} else
847 			overlay->pfit_active = false;
848 	}
849 
850 	ret = check_overlay_dst(overlay, params);
851 	if (ret != 0)
852 		goto out_unlock;
853 
854 	if (overlay->pfit_active) {
855 		params->dst_y = (((u32)params->dst_y << 12) /
856 				 overlay->pfit_vscale_ratio);
857 		/* shifting right rounds downwards, so add 1 */
858 		params->dst_height = (((u32)params->dst_height << 12) /
859 				 overlay->pfit_vscale_ratio) + 1;
860 	}
861 
862 	if (params->src_scan_height > params->src_height ||
863 	    params->src_scan_width > params->src_width) {
864 		ret = -EINVAL;
865 		goto out_unlock;
866 	}
867 
868 	ret = check_overlay_src(display, params, obj);
869 	if (ret != 0)
870 		goto out_unlock;
871 
872 	/* Check scaling after src size to prevent a divide-by-zero. */
873 	ret = check_overlay_scaling(params);
874 	if (ret != 0)
875 		goto out_unlock;
876 
877 	ret = intel_overlay_do_put_image(overlay, obj, params);
878 	if (ret != 0)
879 		goto out_unlock;
880 
881 	drm_modeset_unlock_all(dev);
882 	drm_gem_object_put(obj);
883 
884 	return 0;
885 
886 out_unlock:
887 	drm_modeset_unlock_all(dev);
888 	drm_gem_object_put(obj);
889 
890 	return ret;
891 }
892 
893 static void update_reg_attrs(struct intel_overlay *overlay,
894 			     struct overlay_registers __iomem *regs)
895 {
896 	iowrite32((overlay->contrast << 18) | (overlay->brightness & 0xff),
897 		  &regs->OCLRC0);
898 	iowrite32(overlay->saturation, &regs->OCLRC1);
899 }
900 
901 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
902 {
903 	int i;
904 
905 	if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
906 		return false;
907 
908 	for (i = 0; i < 3; i++) {
909 		if (((gamma1 >> i*8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
910 			return false;
911 	}
912 
913 	return true;
914 }
915 
916 static bool check_gamma5_errata(u32 gamma5)
917 {
918 	int i;
919 
920 	for (i = 0; i < 3; i++) {
921 		if (((gamma5 >> i*8) & 0xff) == 0x80)
922 			return false;
923 	}
924 
925 	return true;
926 }
927 
928 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
929 {
930 	if (!check_gamma_bounds(0, attrs->gamma0) ||
931 	    !check_gamma_bounds(attrs->gamma0, attrs->gamma1) ||
932 	    !check_gamma_bounds(attrs->gamma1, attrs->gamma2) ||
933 	    !check_gamma_bounds(attrs->gamma2, attrs->gamma3) ||
934 	    !check_gamma_bounds(attrs->gamma3, attrs->gamma4) ||
935 	    !check_gamma_bounds(attrs->gamma4, attrs->gamma5) ||
936 	    !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
937 		return -EINVAL;
938 
939 	if (!check_gamma5_errata(attrs->gamma5))
940 		return -EINVAL;
941 
942 	return 0;
943 }
944 
945 int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
946 			      struct drm_file *file_priv)
947 {
948 	struct intel_display *display = to_intel_display(dev);
949 	struct drm_intel_overlay_attrs *attrs = data;
950 	struct intel_overlay *overlay;
951 	int ret;
952 
953 	overlay = display->overlay;
954 	if (!overlay) {
955 		drm_dbg(display->drm, "userspace bug: no overlay\n");
956 		return -ENODEV;
957 	}
958 
959 	drm_modeset_lock_all(dev);
960 
961 	ret = -EINVAL;
962 	if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
963 		attrs->color_key  = overlay->color_key;
964 		attrs->brightness = overlay->brightness;
965 		attrs->contrast   = overlay->contrast;
966 		attrs->saturation = overlay->saturation;
967 
968 		if (DISPLAY_VER(display) != 2) {
969 			attrs->gamma0 = intel_de_read(display, OGAMC0);
970 			attrs->gamma1 = intel_de_read(display, OGAMC1);
971 			attrs->gamma2 = intel_de_read(display, OGAMC2);
972 			attrs->gamma3 = intel_de_read(display, OGAMC3);
973 			attrs->gamma4 = intel_de_read(display, OGAMC4);
974 			attrs->gamma5 = intel_de_read(display, OGAMC5);
975 		}
976 	} else {
977 		if (attrs->brightness < -128 || attrs->brightness > 127)
978 			goto out_unlock;
979 		if (attrs->contrast > 255)
980 			goto out_unlock;
981 		if (attrs->saturation > 1023)
982 			goto out_unlock;
983 
984 		overlay->color_key  = attrs->color_key;
985 		overlay->brightness = attrs->brightness;
986 		overlay->contrast   = attrs->contrast;
987 		overlay->saturation = attrs->saturation;
988 
989 		update_reg_attrs(overlay, overlay->regs);
990 
991 		if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
992 			if (DISPLAY_VER(display) == 2)
993 				goto out_unlock;
994 
995 			if (intel_parent_overlay_is_active(display)) {
996 				ret = -EBUSY;
997 				goto out_unlock;
998 			}
999 
1000 			ret = check_gamma(attrs);
1001 			if (ret)
1002 				goto out_unlock;
1003 
1004 			intel_de_write(display, OGAMC0, attrs->gamma0);
1005 			intel_de_write(display, OGAMC1, attrs->gamma1);
1006 			intel_de_write(display, OGAMC2, attrs->gamma2);
1007 			intel_de_write(display, OGAMC3, attrs->gamma3);
1008 			intel_de_write(display, OGAMC4, attrs->gamma4);
1009 			intel_de_write(display, OGAMC5, attrs->gamma5);
1010 		}
1011 	}
1012 	overlay->color_key_enabled = (attrs->flags & I915_OVERLAY_DISABLE_DEST_COLORKEY) == 0;
1013 
1014 	ret = 0;
1015 out_unlock:
1016 	drm_modeset_unlock_all(dev);
1017 
1018 	return ret;
1019 }
1020 
1021 void intel_overlay_setup(struct intel_display *display)
1022 {
1023 	struct intel_overlay *overlay;
1024 	void __iomem *regs;
1025 
1026 	if (!HAS_OVERLAY(display))
1027 		return;
1028 
1029 	overlay = kzalloc_obj(*overlay);
1030 	if (!overlay)
1031 		return;
1032 
1033 	regs = intel_parent_overlay_setup(display,
1034 					  OVERLAY_NEEDS_PHYSICAL(display));
1035 	if (IS_ERR(regs))
1036 		goto out_free;
1037 
1038 	overlay->display = display;
1039 	overlay->regs = regs;
1040 	overlay->color_key = 0x0101fe;
1041 	overlay->color_key_enabled = true;
1042 	overlay->brightness = -19;
1043 	overlay->contrast = 75;
1044 	overlay->saturation = 146;
1045 
1046 	memset_io(overlay->regs, 0, sizeof(struct overlay_registers));
1047 	update_polyphase_filter(overlay->regs);
1048 	update_reg_attrs(overlay, overlay->regs);
1049 
1050 	display->overlay = overlay;
1051 	drm_info(display->drm, "Initialized overlay support.\n");
1052 	return;
1053 
1054 out_free:
1055 	kfree(overlay);
1056 }
1057 
1058 bool intel_overlay_available(struct intel_display *display)
1059 {
1060 	return display->overlay;
1061 }
1062 
1063 void intel_overlay_cleanup(struct intel_display *display)
1064 {
1065 	if (!display->overlay)
1066 		return;
1067 
1068 	intel_parent_overlay_cleanup(display);
1069 
1070 	kfree(display->overlay);
1071 	display->overlay = NULL;
1072 }
1073