xref: /linux/drivers/gpu/drm/i915/display/intel_overlay.c (revision e3f33adfa3a3be16ef59ed849fbbd10e966e98b0)
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 	atomic_inc(&display->restore.pending_fb_pin);
485 
486 	vma = intel_parent_overlay_pin_fb(display, obj, &offset);
487 	if (IS_ERR(vma)) {
488 		ret = PTR_ERR(vma);
489 		goto out_pin_section;
490 	}
491 
492 	if (!intel_parent_overlay_is_active(display)) {
493 		const struct intel_crtc_state *crtc_state =
494 			overlay->crtc->config;
495 		u32 oconfig = 0;
496 
497 		if (crtc_state->gamma_enable &&
498 		    crtc_state->gamma_mode == GAMMA_MODE_MODE_8BIT)
499 			oconfig |= OCONF_CC_OUT_8BIT;
500 		if (crtc_state->gamma_enable)
501 			oconfig |= OCONF_GAMMA2_ENABLE;
502 		if (DISPLAY_VER(display) == 4)
503 			oconfig |= OCONF_CSC_MODE_BT709;
504 		oconfig |= pipe == 0 ?
505 			OCONF_PIPE_A : OCONF_PIPE_B;
506 		iowrite32(oconfig, &regs->OCONFIG);
507 
508 		ret = intel_parent_overlay_on(display, INTEL_FRONTBUFFER_OVERLAY(pipe));
509 		if (ret != 0)
510 			goto out_unpin;
511 	}
512 
513 	iowrite32(params->dst_y << 16 | params->dst_x, &regs->DWINPOS);
514 	iowrite32(params->dst_height << 16 | params->dst_width, &regs->DWINSZ);
515 
516 	if (params->flags & I915_OVERLAY_YUV_PACKED)
517 		tmp_width = packed_width_bytes(params->flags,
518 					       params->src_width);
519 	else
520 		tmp_width = params->src_width;
521 
522 	swidth = params->src_width;
523 	swidthsw = calc_swidthsw(display, params->offset_Y, tmp_width);
524 	sheight = params->src_height;
525 	iowrite32(offset + params->offset_Y, &regs->OBUF_0Y);
526 	ostride = params->stride_Y;
527 
528 	if (params->flags & I915_OVERLAY_YUV_PLANAR) {
529 		int uv_hscale = uv_hsubsampling(params->flags);
530 		int uv_vscale = uv_vsubsampling(params->flags);
531 		u32 tmp_U, tmp_V;
532 
533 		swidth |= (params->src_width / uv_hscale) << 16;
534 		sheight |= (params->src_height / uv_vscale) << 16;
535 
536 		tmp_U = calc_swidthsw(display, params->offset_U,
537 				      params->src_width / uv_hscale);
538 		tmp_V = calc_swidthsw(display, params->offset_V,
539 				      params->src_width / uv_hscale);
540 		swidthsw |= max(tmp_U, tmp_V) << 16;
541 
542 		iowrite32(offset + params->offset_U,
543 			  &regs->OBUF_0U);
544 		iowrite32(offset + params->offset_V,
545 			  &regs->OBUF_0V);
546 
547 		ostride |= params->stride_UV << 16;
548 	}
549 
550 	iowrite32(swidth, &regs->SWIDTH);
551 	iowrite32(swidthsw, &regs->SWIDTHSW);
552 	iowrite32(sheight, &regs->SHEIGHT);
553 	iowrite32(ostride, &regs->OSTRIDE);
554 
555 	scale_changed = update_scaling_factors(overlay, regs, params);
556 
557 	update_colorkey(overlay, regs);
558 
559 	iowrite32(overlay_cmd_reg(params), &regs->OCMD);
560 
561 	/* check for underruns */
562 	tmp = intel_de_read(display, DOVSTA);
563 	if (tmp & (1 << 17))
564 		drm_dbg(display->drm, "overlay underrun, DOVSTA: %x\n", tmp);
565 
566 	ret = intel_parent_overlay_continue(display, vma, scale_changed);
567 	if (ret)
568 		goto out_unpin;
569 
570 	return 0;
571 
572 out_unpin:
573 	intel_parent_overlay_unpin_fb(display, vma);
574 out_pin_section:
575 	atomic_dec(&display->restore.pending_fb_pin);
576 
577 	return ret;
578 }
579 
580 int intel_overlay_switch_off(struct intel_overlay *overlay)
581 {
582 	struct intel_display *display = overlay->display;
583 	int ret;
584 
585 	drm_WARN_ON(display->drm,
586 		    !drm_modeset_is_locked(&display->drm->mode_config.connection_mutex));
587 
588 	ret = intel_parent_overlay_recover_from_interrupt(display);
589 	if (ret != 0)
590 		return ret;
591 
592 	if (!intel_parent_overlay_is_active(display))
593 		return 0;
594 
595 	ret = intel_parent_overlay_release_old_vid(display);
596 	if (ret != 0)
597 		return ret;
598 
599 	iowrite32(0, &overlay->regs->OCMD);
600 
601 	overlay->crtc->overlay = NULL;
602 	overlay->crtc = NULL;
603 
604 	return intel_parent_overlay_off(display);
605 }
606 
607 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
608 					  struct intel_crtc *crtc)
609 {
610 	if (!crtc->active)
611 		return -EINVAL;
612 
613 	/* can't use the overlay with double wide pipe */
614 	if (crtc->config->double_wide)
615 		return -EINVAL;
616 
617 	return 0;
618 }
619 
620 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
621 {
622 	struct intel_display *display = overlay->display;
623 	u32 ratio;
624 
625 	/* XXX: This is not the same logic as in the xorg driver, but more in
626 	 * line with the intel documentation for the i965
627 	 */
628 	if (DISPLAY_VER(display) >= 4) {
629 		u32 tmp = intel_de_read(display, PFIT_PGM_RATIOS(display));
630 
631 		/* on i965 use the PGM reg to read out the autoscaler values */
632 		ratio = REG_FIELD_GET(PFIT_VERT_SCALE_MASK_965, tmp);
633 	} else {
634 		u32 tmp;
635 
636 		if (intel_de_read(display, PFIT_CONTROL(display)) & PFIT_VERT_AUTO_SCALE)
637 			tmp = intel_de_read(display, PFIT_AUTO_RATIOS(display));
638 		else
639 			tmp = intel_de_read(display, PFIT_PGM_RATIOS(display));
640 
641 		ratio = REG_FIELD_GET(PFIT_VERT_SCALE_MASK, tmp);
642 	}
643 
644 	overlay->pfit_vscale_ratio = ratio;
645 }
646 
647 static int check_overlay_dst(struct intel_overlay *overlay,
648 			     struct drm_intel_overlay_put_image *rec)
649 {
650 	const struct intel_crtc_state *crtc_state =
651 		overlay->crtc->config;
652 	struct drm_rect req, clipped;
653 
654 	drm_rect_init(&req, rec->dst_x, rec->dst_y,
655 		      rec->dst_width, rec->dst_height);
656 
657 	clipped = req;
658 
659 	if (!drm_rect_intersect(&clipped, &crtc_state->pipe_src))
660 		return -EINVAL;
661 
662 	if (!drm_rect_equals(&clipped, &req))
663 		return -EINVAL;
664 
665 	return 0;
666 }
667 
668 static int check_overlay_scaling(struct drm_intel_overlay_put_image *rec)
669 {
670 	u32 tmp;
671 
672 	/* downscaling limit is 8.0 */
673 	tmp = ((rec->src_scan_height << 16) / rec->dst_height) >> 16;
674 	if (tmp > 7)
675 		return -EINVAL;
676 
677 	tmp = ((rec->src_scan_width << 16) / rec->dst_width) >> 16;
678 	if (tmp > 7)
679 		return -EINVAL;
680 
681 	return 0;
682 }
683 
684 static int check_overlay_src(struct intel_display *display,
685 			     struct drm_intel_overlay_put_image *rec,
686 			     struct drm_gem_object *obj)
687 {
688 	int uv_hscale = uv_hsubsampling(rec->flags);
689 	int uv_vscale = uv_vsubsampling(rec->flags);
690 	u32 stride_mask;
691 	int depth;
692 	u32 tmp;
693 
694 	/* check src dimensions */
695 	if (display->platform.i845g || display->platform.i830) {
696 		if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY ||
697 		    rec->src_width  > IMAGE_MAX_WIDTH_LEGACY)
698 			return -EINVAL;
699 	} else {
700 		if (rec->src_height > IMAGE_MAX_HEIGHT ||
701 		    rec->src_width  > IMAGE_MAX_WIDTH)
702 			return -EINVAL;
703 	}
704 
705 	/* better safe than sorry, use 4 as the maximal subsampling ratio */
706 	if (rec->src_height < N_VERT_Y_TAPS*4 ||
707 	    rec->src_width  < N_HORIZ_Y_TAPS*4)
708 		return -EINVAL;
709 
710 	/* check alignment constraints */
711 	switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
712 	case I915_OVERLAY_RGB:
713 		/* not implemented */
714 		return -EINVAL;
715 
716 	case I915_OVERLAY_YUV_PACKED:
717 		if (uv_vscale != 1)
718 			return -EINVAL;
719 
720 		depth = packed_depth_bytes(rec->flags);
721 		if (depth < 0)
722 			return depth;
723 
724 		/* ignore UV planes */
725 		rec->stride_UV = 0;
726 		rec->offset_U = 0;
727 		rec->offset_V = 0;
728 		/* check pixel alignment */
729 		if (rec->offset_Y % depth)
730 			return -EINVAL;
731 		break;
732 
733 	case I915_OVERLAY_YUV_PLANAR:
734 		if (uv_vscale < 0 || uv_hscale < 0)
735 			return -EINVAL;
736 		/* no offset restrictions for planar formats */
737 		break;
738 
739 	default:
740 		return -EINVAL;
741 	}
742 
743 	if (rec->src_width % uv_hscale)
744 		return -EINVAL;
745 
746 	/* stride checking */
747 	if (display->platform.i830 || display->platform.i845g)
748 		stride_mask = 255;
749 	else
750 		stride_mask = 63;
751 
752 	if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
753 		return -EINVAL;
754 	if (DISPLAY_VER(display) == 4 && rec->stride_Y < 512)
755 		return -EINVAL;
756 
757 	tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
758 		4096 : 8192;
759 	if (rec->stride_Y > tmp || rec->stride_UV > 2*1024)
760 		return -EINVAL;
761 
762 	/* check buffer dimensions */
763 	switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
764 	case I915_OVERLAY_RGB:
765 	case I915_OVERLAY_YUV_PACKED:
766 		/* always 4 Y values per depth pixels */
767 		if (packed_width_bytes(rec->flags, rec->src_width) > rec->stride_Y)
768 			return -EINVAL;
769 
770 		tmp = rec->stride_Y*rec->src_height;
771 		if (rec->offset_Y + tmp > obj->size)
772 			return -EINVAL;
773 		break;
774 
775 	case I915_OVERLAY_YUV_PLANAR:
776 		if (rec->src_width > rec->stride_Y)
777 			return -EINVAL;
778 		if (rec->src_width/uv_hscale > rec->stride_UV)
779 			return -EINVAL;
780 
781 		tmp = rec->stride_Y * rec->src_height;
782 		if (rec->offset_Y + tmp > obj->size)
783 			return -EINVAL;
784 
785 		tmp = rec->stride_UV * (rec->src_height / uv_vscale);
786 		if (rec->offset_U + tmp > obj->size ||
787 		    rec->offset_V + tmp > obj->size)
788 			return -EINVAL;
789 		break;
790 	}
791 
792 	return 0;
793 }
794 
795 int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
796 				  struct drm_file *file_priv)
797 {
798 	struct intel_display *display = to_intel_display(dev);
799 	struct drm_intel_overlay_put_image *params = data;
800 	struct intel_overlay *overlay;
801 	struct drm_crtc *drmmode_crtc;
802 	struct drm_gem_object *obj;
803 	struct intel_crtc *crtc;
804 	int ret;
805 
806 	overlay = display->overlay;
807 	if (!overlay) {
808 		drm_dbg(display->drm, "userspace bug: no overlay\n");
809 		return -ENODEV;
810 	}
811 
812 	if (!(params->flags & I915_OVERLAY_ENABLE)) {
813 		drm_modeset_lock_all(dev);
814 		ret = intel_overlay_switch_off(overlay);
815 		drm_modeset_unlock_all(dev);
816 
817 		return ret;
818 	}
819 
820 	drmmode_crtc = drm_crtc_find(dev, file_priv, params->crtc_id);
821 	if (!drmmode_crtc)
822 		return -ENOENT;
823 	crtc = to_intel_crtc(drmmode_crtc);
824 
825 	obj = intel_parent_overlay_obj_lookup(display, file_priv, params->bo_handle);
826 	if (IS_ERR(obj))
827 		return PTR_ERR(obj);
828 
829 	drm_modeset_lock_all(dev);
830 
831 	ret = intel_parent_overlay_recover_from_interrupt(display);
832 	if (ret != 0)
833 		goto out_unlock;
834 
835 	if (overlay->crtc != crtc) {
836 		ret = intel_overlay_switch_off(overlay);
837 		if (ret != 0)
838 			goto out_unlock;
839 
840 		ret = check_overlay_possible_on_crtc(overlay, crtc);
841 		if (ret != 0)
842 			goto out_unlock;
843 
844 		overlay->crtc = crtc;
845 		crtc->overlay = overlay;
846 
847 		/* line too wide, i.e. one-line-mode */
848 		if (drm_rect_width(&crtc->config->pipe_src) > 1024 &&
849 		    crtc->config->gmch_pfit.control & PFIT_ENABLE) {
850 			overlay->pfit_active = true;
851 			update_pfit_vscale_ratio(overlay);
852 		} else
853 			overlay->pfit_active = false;
854 	}
855 
856 	ret = check_overlay_dst(overlay, params);
857 	if (ret != 0)
858 		goto out_unlock;
859 
860 	if (overlay->pfit_active) {
861 		params->dst_y = (((u32)params->dst_y << 12) /
862 				 overlay->pfit_vscale_ratio);
863 		/* shifting right rounds downwards, so add 1 */
864 		params->dst_height = (((u32)params->dst_height << 12) /
865 				 overlay->pfit_vscale_ratio) + 1;
866 	}
867 
868 	if (params->src_scan_height > params->src_height ||
869 	    params->src_scan_width > params->src_width) {
870 		ret = -EINVAL;
871 		goto out_unlock;
872 	}
873 
874 	ret = check_overlay_src(display, params, obj);
875 	if (ret != 0)
876 		goto out_unlock;
877 
878 	/* Check scaling after src size to prevent a divide-by-zero. */
879 	ret = check_overlay_scaling(params);
880 	if (ret != 0)
881 		goto out_unlock;
882 
883 	ret = intel_overlay_do_put_image(overlay, obj, params);
884 	if (ret != 0)
885 		goto out_unlock;
886 
887 	drm_modeset_unlock_all(dev);
888 	drm_gem_object_put(obj);
889 
890 	return 0;
891 
892 out_unlock:
893 	drm_modeset_unlock_all(dev);
894 	drm_gem_object_put(obj);
895 
896 	return ret;
897 }
898 
899 static void update_reg_attrs(struct intel_overlay *overlay,
900 			     struct overlay_registers __iomem *regs)
901 {
902 	iowrite32((overlay->contrast << 18) | (overlay->brightness & 0xff),
903 		  &regs->OCLRC0);
904 	iowrite32(overlay->saturation, &regs->OCLRC1);
905 }
906 
907 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
908 {
909 	int i;
910 
911 	if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
912 		return false;
913 
914 	for (i = 0; i < 3; i++) {
915 		if (((gamma1 >> i*8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
916 			return false;
917 	}
918 
919 	return true;
920 }
921 
922 static bool check_gamma5_errata(u32 gamma5)
923 {
924 	int i;
925 
926 	for (i = 0; i < 3; i++) {
927 		if (((gamma5 >> i*8) & 0xff) == 0x80)
928 			return false;
929 	}
930 
931 	return true;
932 }
933 
934 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
935 {
936 	if (!check_gamma_bounds(0, attrs->gamma0) ||
937 	    !check_gamma_bounds(attrs->gamma0, attrs->gamma1) ||
938 	    !check_gamma_bounds(attrs->gamma1, attrs->gamma2) ||
939 	    !check_gamma_bounds(attrs->gamma2, attrs->gamma3) ||
940 	    !check_gamma_bounds(attrs->gamma3, attrs->gamma4) ||
941 	    !check_gamma_bounds(attrs->gamma4, attrs->gamma5) ||
942 	    !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
943 		return -EINVAL;
944 
945 	if (!check_gamma5_errata(attrs->gamma5))
946 		return -EINVAL;
947 
948 	return 0;
949 }
950 
951 int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
952 			      struct drm_file *file_priv)
953 {
954 	struct intel_display *display = to_intel_display(dev);
955 	struct drm_intel_overlay_attrs *attrs = data;
956 	struct intel_overlay *overlay;
957 	int ret;
958 
959 	overlay = display->overlay;
960 	if (!overlay) {
961 		drm_dbg(display->drm, "userspace bug: no overlay\n");
962 		return -ENODEV;
963 	}
964 
965 	drm_modeset_lock_all(dev);
966 
967 	ret = -EINVAL;
968 	if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
969 		attrs->color_key  = overlay->color_key;
970 		attrs->brightness = overlay->brightness;
971 		attrs->contrast   = overlay->contrast;
972 		attrs->saturation = overlay->saturation;
973 
974 		if (DISPLAY_VER(display) != 2) {
975 			attrs->gamma0 = intel_de_read(display, OGAMC0);
976 			attrs->gamma1 = intel_de_read(display, OGAMC1);
977 			attrs->gamma2 = intel_de_read(display, OGAMC2);
978 			attrs->gamma3 = intel_de_read(display, OGAMC3);
979 			attrs->gamma4 = intel_de_read(display, OGAMC4);
980 			attrs->gamma5 = intel_de_read(display, OGAMC5);
981 		}
982 	} else {
983 		if (attrs->brightness < -128 || attrs->brightness > 127)
984 			goto out_unlock;
985 		if (attrs->contrast > 255)
986 			goto out_unlock;
987 		if (attrs->saturation > 1023)
988 			goto out_unlock;
989 
990 		overlay->color_key  = attrs->color_key;
991 		overlay->brightness = attrs->brightness;
992 		overlay->contrast   = attrs->contrast;
993 		overlay->saturation = attrs->saturation;
994 
995 		update_reg_attrs(overlay, overlay->regs);
996 
997 		if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
998 			if (DISPLAY_VER(display) == 2)
999 				goto out_unlock;
1000 
1001 			if (intel_parent_overlay_is_active(display)) {
1002 				ret = -EBUSY;
1003 				goto out_unlock;
1004 			}
1005 
1006 			ret = check_gamma(attrs);
1007 			if (ret)
1008 				goto out_unlock;
1009 
1010 			intel_de_write(display, OGAMC0, attrs->gamma0);
1011 			intel_de_write(display, OGAMC1, attrs->gamma1);
1012 			intel_de_write(display, OGAMC2, attrs->gamma2);
1013 			intel_de_write(display, OGAMC3, attrs->gamma3);
1014 			intel_de_write(display, OGAMC4, attrs->gamma4);
1015 			intel_de_write(display, OGAMC5, attrs->gamma5);
1016 		}
1017 	}
1018 	overlay->color_key_enabled = (attrs->flags & I915_OVERLAY_DISABLE_DEST_COLORKEY) == 0;
1019 
1020 	ret = 0;
1021 out_unlock:
1022 	drm_modeset_unlock_all(dev);
1023 
1024 	return ret;
1025 }
1026 
1027 void intel_overlay_setup(struct intel_display *display)
1028 {
1029 	struct intel_overlay *overlay;
1030 	void __iomem *regs;
1031 
1032 	if (!HAS_OVERLAY(display))
1033 		return;
1034 
1035 	overlay = kzalloc_obj(*overlay);
1036 	if (!overlay)
1037 		return;
1038 
1039 	regs = intel_parent_overlay_setup(display,
1040 					  OVERLAY_NEEDS_PHYSICAL(display));
1041 	if (IS_ERR(regs))
1042 		goto out_free;
1043 
1044 	overlay->display = display;
1045 	overlay->regs = regs;
1046 	overlay->color_key = 0x0101fe;
1047 	overlay->color_key_enabled = true;
1048 	overlay->brightness = -19;
1049 	overlay->contrast = 75;
1050 	overlay->saturation = 146;
1051 
1052 	memset_io(overlay->regs, 0, sizeof(struct overlay_registers));
1053 	update_polyphase_filter(overlay->regs);
1054 	update_reg_attrs(overlay, overlay->regs);
1055 
1056 	display->overlay = overlay;
1057 	drm_info(display->drm, "Initialized overlay support.\n");
1058 	return;
1059 
1060 out_free:
1061 	kfree(overlay);
1062 }
1063 
1064 bool intel_overlay_available(struct intel_display *display)
1065 {
1066 	return display->overlay;
1067 }
1068 
1069 void intel_overlay_cleanup(struct intel_display *display)
1070 {
1071 	if (!display->overlay)
1072 		return;
1073 
1074 	intel_parent_overlay_cleanup(display);
1075 
1076 	kfree(display->overlay);
1077 	display->overlay = NULL;
1078 }
1079