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 
31 #include "gem/i915_gem_pm.h"
32 #include "gt/intel_gpu_commands.h"
33 #include "gt/intel_ring.h"
34 
35 #include "i915_drv.h"
36 #include "i915_reg.h"
37 #include "intel_de.h"
38 #include "intel_display_types.h"
39 #include "intel_frontbuffer.h"
40 #include "intel_overlay.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 mininum of both.  */
46 #define IMAGE_MAX_WIDTH		2048
47 #define IMAGE_MAX_HEIGHT	2046 /* 2 * 1023 */
48 /* on 830 and 845 these large limits result in the card hanging */
49 #define IMAGE_MAX_WIDTH_LEGACY	1024
50 #define IMAGE_MAX_HEIGHT_LEGACY	1088
51 
52 /* overlay register definitions */
53 /* OCMD register */
54 #define OCMD_TILED_SURFACE	(0x1<<19)
55 #define OCMD_MIRROR_MASK	(0x3<<17)
56 #define OCMD_MIRROR_MODE	(0x3<<17)
57 #define OCMD_MIRROR_HORIZONTAL	(0x1<<17)
58 #define OCMD_MIRROR_VERTICAL	(0x2<<17)
59 #define OCMD_MIRROR_BOTH	(0x3<<17)
60 #define OCMD_BYTEORDER_MASK	(0x3<<14) /* zero for YUYV or FOURCC YUY2 */
61 #define OCMD_UV_SWAP		(0x1<<14) /* YVYU */
62 #define OCMD_Y_SWAP		(0x2<<14) /* UYVY or FOURCC UYVY */
63 #define OCMD_Y_AND_UV_SWAP	(0x3<<14) /* VYUY */
64 #define OCMD_SOURCE_FORMAT_MASK (0xf<<10)
65 #define OCMD_RGB_888		(0x1<<10) /* not in i965 Intel docs */
66 #define OCMD_RGB_555		(0x2<<10) /* not in i965 Intel docs */
67 #define OCMD_RGB_565		(0x3<<10) /* not in i965 Intel docs */
68 #define OCMD_YUV_422_PACKED	(0x8<<10)
69 #define OCMD_YUV_411_PACKED	(0x9<<10) /* not in i965 Intel docs */
70 #define OCMD_YUV_420_PLANAR	(0xc<<10)
71 #define OCMD_YUV_422_PLANAR	(0xd<<10)
72 #define OCMD_YUV_410_PLANAR	(0xe<<10) /* also 411 */
73 #define OCMD_TVSYNCFLIP_PARITY	(0x1<<9)
74 #define OCMD_TVSYNCFLIP_ENABLE	(0x1<<7)
75 #define OCMD_BUF_TYPE_MASK	(0x1<<5)
76 #define OCMD_BUF_TYPE_FRAME	(0x0<<5)
77 #define OCMD_BUF_TYPE_FIELD	(0x1<<5)
78 #define OCMD_TEST_MODE		(0x1<<4)
79 #define OCMD_BUFFER_SELECT	(0x3<<2)
80 #define OCMD_BUFFER0		(0x0<<2)
81 #define OCMD_BUFFER1		(0x1<<2)
82 #define OCMD_FIELD_SELECT	(0x1<<2)
83 #define OCMD_FIELD0		(0x0<<1)
84 #define OCMD_FIELD1		(0x1<<1)
85 #define OCMD_ENABLE		(0x1<<0)
86 
87 /* OCONFIG register */
88 #define OCONF_PIPE_MASK		(0x1<<18)
89 #define OCONF_PIPE_A		(0x0<<18)
90 #define OCONF_PIPE_B		(0x1<<18)
91 #define OCONF_GAMMA2_ENABLE	(0x1<<16)
92 #define OCONF_CSC_MODE_BT601	(0x0<<5)
93 #define OCONF_CSC_MODE_BT709	(0x1<<5)
94 #define OCONF_CSC_BYPASS	(0x1<<4)
95 #define OCONF_CC_OUT_8BIT	(0x1<<3)
96 #define OCONF_TEST_MODE		(0x1<<2)
97 #define OCONF_THREE_LINE_BUFFER	(0x1<<0)
98 #define OCONF_TWO_LINE_BUFFER	(0x0<<0)
99 
100 /* DCLRKM (dst-key) register */
101 #define DST_KEY_ENABLE		(0x1<<31)
102 #define CLK_RGB24_MASK		0x0
103 #define CLK_RGB16_MASK		0x070307
104 #define CLK_RGB15_MASK		0x070707
105 
106 #define RGB30_TO_COLORKEY(c) \
107 	((((c) & 0x3fc00000) >> 6) | (((c) & 0x000ff000) >> 4) | (((c) & 0x000003fc) >> 2))
108 #define RGB16_TO_COLORKEY(c) \
109 	((((c) & 0xf800) << 8) | (((c) & 0x07e0) << 5) | (((c) & 0x001f) << 3))
110 #define RGB15_TO_COLORKEY(c) \
111 	((((c) & 0x7c00) << 9) | (((c) & 0x03e0) << 6) | (((c) & 0x001f) << 3))
112 #define RGB8I_TO_COLORKEY(c) \
113 	((((c) & 0xff) << 16) | (((c) & 0xff) << 8) | (((c) & 0xff) << 0))
114 
115 /* overlay flip addr flag */
116 #define OFC_UPDATE		0x1
117 
118 /* polyphase filter coefficients */
119 #define N_HORIZ_Y_TAPS          5
120 #define N_VERT_Y_TAPS           3
121 #define N_HORIZ_UV_TAPS         3
122 #define N_VERT_UV_TAPS          3
123 #define N_PHASES                17
124 #define MAX_TAPS                5
125 
126 /* memory bufferd overlay registers */
127 struct overlay_registers {
128 	u32 OBUF_0Y;
129 	u32 OBUF_1Y;
130 	u32 OBUF_0U;
131 	u32 OBUF_0V;
132 	u32 OBUF_1U;
133 	u32 OBUF_1V;
134 	u32 OSTRIDE;
135 	u32 YRGB_VPH;
136 	u32 UV_VPH;
137 	u32 HORZ_PH;
138 	u32 INIT_PHS;
139 	u32 DWINPOS;
140 	u32 DWINSZ;
141 	u32 SWIDTH;
142 	u32 SWIDTHSW;
143 	u32 SHEIGHT;
144 	u32 YRGBSCALE;
145 	u32 UVSCALE;
146 	u32 OCLRC0;
147 	u32 OCLRC1;
148 	u32 DCLRKV;
149 	u32 DCLRKM;
150 	u32 SCLRKVH;
151 	u32 SCLRKVL;
152 	u32 SCLRKEN;
153 	u32 OCONFIG;
154 	u32 OCMD;
155 	u32 RESERVED1; /* 0x6C */
156 	u32 OSTART_0Y;
157 	u32 OSTART_1Y;
158 	u32 OSTART_0U;
159 	u32 OSTART_0V;
160 	u32 OSTART_1U;
161 	u32 OSTART_1V;
162 	u32 OTILEOFF_0Y;
163 	u32 OTILEOFF_1Y;
164 	u32 OTILEOFF_0U;
165 	u32 OTILEOFF_0V;
166 	u32 OTILEOFF_1U;
167 	u32 OTILEOFF_1V;
168 	u32 FASTHSCALE; /* 0xA0 */
169 	u32 UVSCALEV; /* 0xA4 */
170 	u32 RESERVEDC[(0x200 - 0xA8) / 4]; /* 0xA8 - 0x1FC */
171 	u16 Y_VCOEFS[N_VERT_Y_TAPS * N_PHASES]; /* 0x200 */
172 	u16 RESERVEDD[0x100 / 2 - N_VERT_Y_TAPS * N_PHASES];
173 	u16 Y_HCOEFS[N_HORIZ_Y_TAPS * N_PHASES]; /* 0x300 */
174 	u16 RESERVEDE[0x200 / 2 - N_HORIZ_Y_TAPS * N_PHASES];
175 	u16 UV_VCOEFS[N_VERT_UV_TAPS * N_PHASES]; /* 0x500 */
176 	u16 RESERVEDF[0x100 / 2 - N_VERT_UV_TAPS * N_PHASES];
177 	u16 UV_HCOEFS[N_HORIZ_UV_TAPS * N_PHASES]; /* 0x600 */
178 	u16 RESERVEDG[0x100 / 2 - N_HORIZ_UV_TAPS * N_PHASES];
179 };
180 
181 struct intel_overlay {
182 	struct drm_i915_private *i915;
183 	struct intel_context *context;
184 	struct intel_crtc *crtc;
185 	struct i915_vma *vma;
186 	struct i915_vma *old_vma;
187 	struct intel_frontbuffer *frontbuffer;
188 	bool active;
189 	bool pfit_active;
190 	u32 pfit_vscale_ratio; /* shifted-point number, (1<<12) == 1.0 */
191 	u32 color_key:24;
192 	u32 color_key_enabled:1;
193 	u32 brightness, contrast, saturation;
194 	u32 old_xscale, old_yscale;
195 	/* register access */
196 	struct drm_i915_gem_object *reg_bo;
197 	struct overlay_registers __iomem *regs;
198 	u32 flip_addr;
199 	/* flip handling */
200 	struct i915_active last_flip;
201 	void (*flip_complete)(struct intel_overlay *ovl);
202 };
203 
204 static void i830_overlay_clock_gating(struct drm_i915_private *dev_priv,
205 				      bool enable)
206 {
207 	struct pci_dev *pdev = to_pci_dev(dev_priv->drm.dev);
208 	u8 val;
209 
210 	/* WA_OVERLAY_CLKGATE:alm */
211 	if (enable)
212 		intel_de_write(dev_priv, DSPCLK_GATE_D, 0);
213 	else
214 		intel_de_write(dev_priv, DSPCLK_GATE_D,
215 			       OVRUNIT_CLOCK_GATE_DISABLE);
216 
217 	/* WA_DISABLE_L2CACHE_CLOCK_GATING:alm */
218 	pci_bus_read_config_byte(pdev->bus,
219 				 PCI_DEVFN(0, 0), I830_CLOCK_GATE, &val);
220 	if (enable)
221 		val &= ~I830_L2_CACHE_CLOCK_GATE_DISABLE;
222 	else
223 		val |= I830_L2_CACHE_CLOCK_GATE_DISABLE;
224 	pci_bus_write_config_byte(pdev->bus,
225 				  PCI_DEVFN(0, 0), I830_CLOCK_GATE, val);
226 }
227 
228 static struct i915_request *
229 alloc_request(struct intel_overlay *overlay, void (*fn)(struct intel_overlay *))
230 {
231 	struct i915_request *rq;
232 	int err;
233 
234 	overlay->flip_complete = fn;
235 
236 	rq = i915_request_create(overlay->context);
237 	if (IS_ERR(rq))
238 		return rq;
239 
240 	err = i915_active_add_request(&overlay->last_flip, rq);
241 	if (err) {
242 		i915_request_add(rq);
243 		return ERR_PTR(err);
244 	}
245 
246 	return rq;
247 }
248 
249 /* overlay needs to be disable in OCMD reg */
250 static int intel_overlay_on(struct intel_overlay *overlay)
251 {
252 	struct drm_i915_private *dev_priv = overlay->i915;
253 	struct i915_request *rq;
254 	u32 *cs;
255 
256 	drm_WARN_ON(&dev_priv->drm, overlay->active);
257 
258 	rq = alloc_request(overlay, NULL);
259 	if (IS_ERR(rq))
260 		return PTR_ERR(rq);
261 
262 	cs = intel_ring_begin(rq, 4);
263 	if (IS_ERR(cs)) {
264 		i915_request_add(rq);
265 		return PTR_ERR(cs);
266 	}
267 
268 	overlay->active = true;
269 
270 	if (IS_I830(dev_priv))
271 		i830_overlay_clock_gating(dev_priv, false);
272 
273 	*cs++ = MI_OVERLAY_FLIP | MI_OVERLAY_ON;
274 	*cs++ = overlay->flip_addr | OFC_UPDATE;
275 	*cs++ = MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP;
276 	*cs++ = MI_NOOP;
277 	intel_ring_advance(rq, cs);
278 
279 	i915_request_add(rq);
280 
281 	return i915_active_wait(&overlay->last_flip);
282 }
283 
284 static void intel_overlay_flip_prepare(struct intel_overlay *overlay,
285 				       struct i915_vma *vma)
286 {
287 	enum pipe pipe = overlay->crtc->pipe;
288 	struct intel_frontbuffer *frontbuffer = NULL;
289 
290 	drm_WARN_ON(&overlay->i915->drm, overlay->old_vma);
291 
292 	if (vma)
293 		frontbuffer = intel_frontbuffer_get(vma->obj);
294 
295 	intel_frontbuffer_track(overlay->frontbuffer, frontbuffer,
296 				INTEL_FRONTBUFFER_OVERLAY(pipe));
297 
298 	if (overlay->frontbuffer)
299 		intel_frontbuffer_put(overlay->frontbuffer);
300 	overlay->frontbuffer = frontbuffer;
301 
302 	intel_frontbuffer_flip_prepare(overlay->i915,
303 				       INTEL_FRONTBUFFER_OVERLAY(pipe));
304 
305 	overlay->old_vma = overlay->vma;
306 	if (vma)
307 		overlay->vma = i915_vma_get(vma);
308 	else
309 		overlay->vma = NULL;
310 }
311 
312 /* overlay needs to be enabled in OCMD reg */
313 static int intel_overlay_continue(struct intel_overlay *overlay,
314 				  struct i915_vma *vma,
315 				  bool load_polyphase_filter)
316 {
317 	struct drm_i915_private *dev_priv = overlay->i915;
318 	struct i915_request *rq;
319 	u32 flip_addr = overlay->flip_addr;
320 	u32 tmp, *cs;
321 
322 	drm_WARN_ON(&dev_priv->drm, !overlay->active);
323 
324 	if (load_polyphase_filter)
325 		flip_addr |= OFC_UPDATE;
326 
327 	/* check for underruns */
328 	tmp = intel_de_read(dev_priv, DOVSTA);
329 	if (tmp & (1 << 17))
330 		drm_dbg(&dev_priv->drm, "overlay underrun, DOVSTA: %x\n", tmp);
331 
332 	rq = alloc_request(overlay, NULL);
333 	if (IS_ERR(rq))
334 		return PTR_ERR(rq);
335 
336 	cs = intel_ring_begin(rq, 2);
337 	if (IS_ERR(cs)) {
338 		i915_request_add(rq);
339 		return PTR_ERR(cs);
340 	}
341 
342 	*cs++ = MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE;
343 	*cs++ = flip_addr;
344 	intel_ring_advance(rq, cs);
345 
346 	intel_overlay_flip_prepare(overlay, vma);
347 	i915_request_add(rq);
348 
349 	return 0;
350 }
351 
352 static void intel_overlay_release_old_vma(struct intel_overlay *overlay)
353 {
354 	struct i915_vma *vma;
355 
356 	vma = fetch_and_zero(&overlay->old_vma);
357 	if (drm_WARN_ON(&overlay->i915->drm, !vma))
358 		return;
359 
360 	intel_frontbuffer_flip_complete(overlay->i915,
361 					INTEL_FRONTBUFFER_OVERLAY(overlay->crtc->pipe));
362 
363 	i915_vma_unpin(vma);
364 	i915_vma_put(vma);
365 }
366 
367 static void
368 intel_overlay_release_old_vid_tail(struct intel_overlay *overlay)
369 {
370 	intel_overlay_release_old_vma(overlay);
371 }
372 
373 static void intel_overlay_off_tail(struct intel_overlay *overlay)
374 {
375 	struct drm_i915_private *dev_priv = overlay->i915;
376 
377 	intel_overlay_release_old_vma(overlay);
378 
379 	overlay->crtc->overlay = NULL;
380 	overlay->crtc = NULL;
381 	overlay->active = false;
382 
383 	if (IS_I830(dev_priv))
384 		i830_overlay_clock_gating(dev_priv, true);
385 }
386 
387 static void intel_overlay_last_flip_retire(struct i915_active *active)
388 {
389 	struct intel_overlay *overlay =
390 		container_of(active, typeof(*overlay), last_flip);
391 
392 	if (overlay->flip_complete)
393 		overlay->flip_complete(overlay);
394 }
395 
396 /* overlay needs to be disabled in OCMD reg */
397 static int intel_overlay_off(struct intel_overlay *overlay)
398 {
399 	struct i915_request *rq;
400 	u32 *cs, flip_addr = overlay->flip_addr;
401 
402 	drm_WARN_ON(&overlay->i915->drm, !overlay->active);
403 
404 	/* According to intel docs the overlay hw may hang (when switching
405 	 * off) without loading the filter coeffs. It is however unclear whether
406 	 * this applies to the disabling of the overlay or to the switching off
407 	 * of the hw. Do it in both cases */
408 	flip_addr |= OFC_UPDATE;
409 
410 	rq = alloc_request(overlay, intel_overlay_off_tail);
411 	if (IS_ERR(rq))
412 		return PTR_ERR(rq);
413 
414 	cs = intel_ring_begin(rq, 6);
415 	if (IS_ERR(cs)) {
416 		i915_request_add(rq);
417 		return PTR_ERR(cs);
418 	}
419 
420 	/* wait for overlay to go idle */
421 	*cs++ = MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE;
422 	*cs++ = flip_addr;
423 	*cs++ = MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP;
424 
425 	/* turn overlay off */
426 	*cs++ = MI_OVERLAY_FLIP | MI_OVERLAY_OFF;
427 	*cs++ = flip_addr;
428 	*cs++ = MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP;
429 
430 	intel_ring_advance(rq, cs);
431 
432 	intel_overlay_flip_prepare(overlay, NULL);
433 	i915_request_add(rq);
434 
435 	return i915_active_wait(&overlay->last_flip);
436 }
437 
438 /* recover from an interruption due to a signal
439  * We have to be careful not to repeat work forever an make forward progess. */
440 static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay)
441 {
442 	return i915_active_wait(&overlay->last_flip);
443 }
444 
445 /* Wait for pending overlay flip and release old frame.
446  * Needs to be called before the overlay register are changed
447  * via intel_overlay_(un)map_regs
448  */
449 static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
450 {
451 	struct drm_i915_private *dev_priv = overlay->i915;
452 	struct i915_request *rq;
453 	u32 *cs;
454 
455 	/*
456 	 * Only wait if there is actually an old frame to release to
457 	 * guarantee forward progress.
458 	 */
459 	if (!overlay->old_vma)
460 		return 0;
461 
462 	if (!(intel_de_read(dev_priv, GEN2_ISR) & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT)) {
463 		intel_overlay_release_old_vid_tail(overlay);
464 		return 0;
465 	}
466 
467 	rq = alloc_request(overlay, intel_overlay_release_old_vid_tail);
468 	if (IS_ERR(rq))
469 		return PTR_ERR(rq);
470 
471 	cs = intel_ring_begin(rq, 2);
472 	if (IS_ERR(cs)) {
473 		i915_request_add(rq);
474 		return PTR_ERR(cs);
475 	}
476 
477 	*cs++ = MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP;
478 	*cs++ = MI_NOOP;
479 	intel_ring_advance(rq, cs);
480 
481 	i915_request_add(rq);
482 
483 	return i915_active_wait(&overlay->last_flip);
484 }
485 
486 void intel_overlay_reset(struct drm_i915_private *dev_priv)
487 {
488 	struct intel_overlay *overlay = dev_priv->overlay;
489 
490 	if (!overlay)
491 		return;
492 
493 	overlay->old_xscale = 0;
494 	overlay->old_yscale = 0;
495 	overlay->crtc = NULL;
496 	overlay->active = false;
497 }
498 
499 static int packed_depth_bytes(u32 format)
500 {
501 	switch (format & I915_OVERLAY_DEPTH_MASK) {
502 	case I915_OVERLAY_YUV422:
503 		return 4;
504 	case I915_OVERLAY_YUV411:
505 		/* return 6; not implemented */
506 	default:
507 		return -EINVAL;
508 	}
509 }
510 
511 static int packed_width_bytes(u32 format, short width)
512 {
513 	switch (format & I915_OVERLAY_DEPTH_MASK) {
514 	case I915_OVERLAY_YUV422:
515 		return width << 1;
516 	default:
517 		return -EINVAL;
518 	}
519 }
520 
521 static int uv_hsubsampling(u32 format)
522 {
523 	switch (format & I915_OVERLAY_DEPTH_MASK) {
524 	case I915_OVERLAY_YUV422:
525 	case I915_OVERLAY_YUV420:
526 		return 2;
527 	case I915_OVERLAY_YUV411:
528 	case I915_OVERLAY_YUV410:
529 		return 4;
530 	default:
531 		return -EINVAL;
532 	}
533 }
534 
535 static int uv_vsubsampling(u32 format)
536 {
537 	switch (format & I915_OVERLAY_DEPTH_MASK) {
538 	case I915_OVERLAY_YUV420:
539 	case I915_OVERLAY_YUV410:
540 		return 2;
541 	case I915_OVERLAY_YUV422:
542 	case I915_OVERLAY_YUV411:
543 		return 1;
544 	default:
545 		return -EINVAL;
546 	}
547 }
548 
549 static u32 calc_swidthsw(struct drm_i915_private *dev_priv, u32 offset, u32 width)
550 {
551 	u32 sw;
552 
553 	if (DISPLAY_VER(dev_priv) == 2)
554 		sw = ALIGN((offset & 31) + width, 32);
555 	else
556 		sw = ALIGN((offset & 63) + width, 64);
557 
558 	if (sw == 0)
559 		return 0;
560 
561 	return (sw - 32) >> 3;
562 }
563 
564 static const u16 y_static_hcoeffs[N_PHASES][N_HORIZ_Y_TAPS] = {
565 	[ 0] = { 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0, },
566 	[ 1] = { 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440, },
567 	[ 2] = { 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0, },
568 	[ 3] = { 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380, },
569 	[ 4] = { 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320, },
570 	[ 5] = { 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0, },
571 	[ 6] = { 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260, },
572 	[ 7] = { 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200, },
573 	[ 8] = { 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0, },
574 	[ 9] = { 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160, },
575 	[10] = { 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120, },
576 	[11] = { 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0, },
577 	[12] = { 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0, },
578 	[13] = { 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060, },
579 	[14] = { 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040, },
580 	[15] = { 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020, },
581 	[16] = { 0xb000, 0x3000, 0x0800, 0x3000, 0xb000, },
582 };
583 
584 static const u16 uv_static_hcoeffs[N_PHASES][N_HORIZ_UV_TAPS] = {
585 	[ 0] = { 0x3000, 0x1800, 0x1800, },
586 	[ 1] = { 0xb000, 0x18d0, 0x2e60, },
587 	[ 2] = { 0xb000, 0x1990, 0x2ce0, },
588 	[ 3] = { 0xb020, 0x1a68, 0x2b40, },
589 	[ 4] = { 0xb040, 0x1b20, 0x29e0, },
590 	[ 5] = { 0xb060, 0x1bd8, 0x2880, },
591 	[ 6] = { 0xb080, 0x1c88, 0x3e60, },
592 	[ 7] = { 0xb0a0, 0x1d28, 0x3c00, },
593 	[ 8] = { 0xb0c0, 0x1db8, 0x39e0, },
594 	[ 9] = { 0xb0e0, 0x1e40, 0x37e0, },
595 	[10] = { 0xb100, 0x1eb8, 0x3620, },
596 	[11] = { 0xb100, 0x1f18, 0x34a0, },
597 	[12] = { 0xb100, 0x1f68, 0x3360, },
598 	[13] = { 0xb0e0, 0x1fa8, 0x3240, },
599 	[14] = { 0xb0c0, 0x1fe0, 0x3140, },
600 	[15] = { 0xb060, 0x1ff0, 0x30a0, },
601 	[16] = { 0x3000, 0x0800, 0x3000, },
602 };
603 
604 static void update_polyphase_filter(struct overlay_registers __iomem *regs)
605 {
606 	memcpy_toio(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
607 	memcpy_toio(regs->UV_HCOEFS, uv_static_hcoeffs,
608 		    sizeof(uv_static_hcoeffs));
609 }
610 
611 static bool update_scaling_factors(struct intel_overlay *overlay,
612 				   struct overlay_registers __iomem *regs,
613 				   struct drm_intel_overlay_put_image *params)
614 {
615 	/* fixed point with a 12 bit shift */
616 	u32 xscale, yscale, xscale_UV, yscale_UV;
617 #define FP_SHIFT 12
618 #define FRACT_MASK 0xfff
619 	bool scale_changed = false;
620 	int uv_hscale = uv_hsubsampling(params->flags);
621 	int uv_vscale = uv_vsubsampling(params->flags);
622 
623 	if (params->dst_width > 1)
624 		xscale = ((params->src_scan_width - 1) << FP_SHIFT) /
625 			params->dst_width;
626 	else
627 		xscale = 1 << FP_SHIFT;
628 
629 	if (params->dst_height > 1)
630 		yscale = ((params->src_scan_height - 1) << FP_SHIFT) /
631 			params->dst_height;
632 	else
633 		yscale = 1 << FP_SHIFT;
634 
635 	/*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
636 	xscale_UV = xscale/uv_hscale;
637 	yscale_UV = yscale/uv_vscale;
638 	/* make the Y scale to UV scale ratio an exact multiply */
639 	xscale = xscale_UV * uv_hscale;
640 	yscale = yscale_UV * uv_vscale;
641 	/*} else {
642 	  xscale_UV = 0;
643 	  yscale_UV = 0;
644 	  }*/
645 
646 	if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
647 		scale_changed = true;
648 	overlay->old_xscale = xscale;
649 	overlay->old_yscale = yscale;
650 
651 	iowrite32(((yscale & FRACT_MASK) << 20) |
652 		  ((xscale >> FP_SHIFT)  << 16) |
653 		  ((xscale & FRACT_MASK) << 3),
654 		 &regs->YRGBSCALE);
655 
656 	iowrite32(((yscale_UV & FRACT_MASK) << 20) |
657 		  ((xscale_UV >> FP_SHIFT)  << 16) |
658 		  ((xscale_UV & FRACT_MASK) << 3),
659 		 &regs->UVSCALE);
660 
661 	iowrite32((((yscale    >> FP_SHIFT) << 16) |
662 		   ((yscale_UV >> FP_SHIFT) << 0)),
663 		 &regs->UVSCALEV);
664 
665 	if (scale_changed)
666 		update_polyphase_filter(regs);
667 
668 	return scale_changed;
669 }
670 
671 static void update_colorkey(struct intel_overlay *overlay,
672 			    struct overlay_registers __iomem *regs)
673 {
674 	const struct intel_plane_state *state =
675 		to_intel_plane_state(overlay->crtc->base.primary->state);
676 	u32 key = overlay->color_key;
677 	u32 format = 0;
678 	u32 flags = 0;
679 
680 	if (overlay->color_key_enabled)
681 		flags |= DST_KEY_ENABLE;
682 
683 	if (state->uapi.visible)
684 		format = state->hw.fb->format->format;
685 
686 	switch (format) {
687 	case DRM_FORMAT_C8:
688 		key = RGB8I_TO_COLORKEY(key);
689 		flags |= CLK_RGB24_MASK;
690 		break;
691 	case DRM_FORMAT_XRGB1555:
692 		key = RGB15_TO_COLORKEY(key);
693 		flags |= CLK_RGB15_MASK;
694 		break;
695 	case DRM_FORMAT_RGB565:
696 		key = RGB16_TO_COLORKEY(key);
697 		flags |= CLK_RGB16_MASK;
698 		break;
699 	case DRM_FORMAT_XRGB2101010:
700 	case DRM_FORMAT_XBGR2101010:
701 		key = RGB30_TO_COLORKEY(key);
702 		flags |= CLK_RGB24_MASK;
703 		break;
704 	default:
705 		flags |= CLK_RGB24_MASK;
706 		break;
707 	}
708 
709 	iowrite32(key, &regs->DCLRKV);
710 	iowrite32(flags, &regs->DCLRKM);
711 }
712 
713 static u32 overlay_cmd_reg(struct drm_intel_overlay_put_image *params)
714 {
715 	u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
716 
717 	if (params->flags & I915_OVERLAY_YUV_PLANAR) {
718 		switch (params->flags & I915_OVERLAY_DEPTH_MASK) {
719 		case I915_OVERLAY_YUV422:
720 			cmd |= OCMD_YUV_422_PLANAR;
721 			break;
722 		case I915_OVERLAY_YUV420:
723 			cmd |= OCMD_YUV_420_PLANAR;
724 			break;
725 		case I915_OVERLAY_YUV411:
726 		case I915_OVERLAY_YUV410:
727 			cmd |= OCMD_YUV_410_PLANAR;
728 			break;
729 		}
730 	} else { /* YUV packed */
731 		switch (params->flags & I915_OVERLAY_DEPTH_MASK) {
732 		case I915_OVERLAY_YUV422:
733 			cmd |= OCMD_YUV_422_PACKED;
734 			break;
735 		case I915_OVERLAY_YUV411:
736 			cmd |= OCMD_YUV_411_PACKED;
737 			break;
738 		}
739 
740 		switch (params->flags & I915_OVERLAY_SWAP_MASK) {
741 		case I915_OVERLAY_NO_SWAP:
742 			break;
743 		case I915_OVERLAY_UV_SWAP:
744 			cmd |= OCMD_UV_SWAP;
745 			break;
746 		case I915_OVERLAY_Y_SWAP:
747 			cmd |= OCMD_Y_SWAP;
748 			break;
749 		case I915_OVERLAY_Y_AND_UV_SWAP:
750 			cmd |= OCMD_Y_AND_UV_SWAP;
751 			break;
752 		}
753 	}
754 
755 	return cmd;
756 }
757 
758 static struct i915_vma *intel_overlay_pin_fb(struct drm_i915_gem_object *new_bo)
759 {
760 	struct i915_gem_ww_ctx ww;
761 	struct i915_vma *vma;
762 	int ret;
763 
764 	i915_gem_ww_ctx_init(&ww, true);
765 retry:
766 	ret = i915_gem_object_lock(new_bo, &ww);
767 	if (!ret) {
768 		vma = i915_gem_object_pin_to_display_plane(new_bo, &ww, 0,
769 							   NULL, PIN_MAPPABLE);
770 		ret = PTR_ERR_OR_ZERO(vma);
771 	}
772 	if (ret == -EDEADLK) {
773 		ret = i915_gem_ww_ctx_backoff(&ww);
774 		if (!ret)
775 			goto retry;
776 	}
777 	i915_gem_ww_ctx_fini(&ww);
778 	if (ret)
779 		return ERR_PTR(ret);
780 
781 	return vma;
782 }
783 
784 static int intel_overlay_do_put_image(struct intel_overlay *overlay,
785 				      struct drm_i915_gem_object *new_bo,
786 				      struct drm_intel_overlay_put_image *params)
787 {
788 	struct overlay_registers __iomem *regs = overlay->regs;
789 	struct drm_i915_private *dev_priv = overlay->i915;
790 	u32 swidth, swidthsw, sheight, ostride;
791 	enum pipe pipe = overlay->crtc->pipe;
792 	bool scale_changed = false;
793 	struct i915_vma *vma;
794 	int ret, tmp_width;
795 
796 	drm_WARN_ON(&dev_priv->drm,
797 		    !drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
798 
799 	ret = intel_overlay_release_old_vid(overlay);
800 	if (ret != 0)
801 		return ret;
802 
803 	atomic_inc(&dev_priv->gpu_error.pending_fb_pin);
804 
805 	vma = intel_overlay_pin_fb(new_bo);
806 	if (IS_ERR(vma)) {
807 		ret = PTR_ERR(vma);
808 		goto out_pin_section;
809 	}
810 
811 	i915_gem_object_flush_frontbuffer(new_bo, ORIGIN_DIRTYFB);
812 
813 	if (!overlay->active) {
814 		const struct intel_crtc_state *crtc_state =
815 			overlay->crtc->config;
816 		u32 oconfig = 0;
817 
818 		if (crtc_state->gamma_enable &&
819 		    crtc_state->gamma_mode == GAMMA_MODE_MODE_8BIT)
820 			oconfig |= OCONF_CC_OUT_8BIT;
821 		if (crtc_state->gamma_enable)
822 			oconfig |= OCONF_GAMMA2_ENABLE;
823 		if (DISPLAY_VER(dev_priv) == 4)
824 			oconfig |= OCONF_CSC_MODE_BT709;
825 		oconfig |= pipe == 0 ?
826 			OCONF_PIPE_A : OCONF_PIPE_B;
827 		iowrite32(oconfig, &regs->OCONFIG);
828 
829 		ret = intel_overlay_on(overlay);
830 		if (ret != 0)
831 			goto out_unpin;
832 	}
833 
834 	iowrite32(params->dst_y << 16 | params->dst_x, &regs->DWINPOS);
835 	iowrite32(params->dst_height << 16 | params->dst_width, &regs->DWINSZ);
836 
837 	if (params->flags & I915_OVERLAY_YUV_PACKED)
838 		tmp_width = packed_width_bytes(params->flags,
839 					       params->src_width);
840 	else
841 		tmp_width = params->src_width;
842 
843 	swidth = params->src_width;
844 	swidthsw = calc_swidthsw(dev_priv, params->offset_Y, tmp_width);
845 	sheight = params->src_height;
846 	iowrite32(i915_ggtt_offset(vma) + params->offset_Y, &regs->OBUF_0Y);
847 	ostride = params->stride_Y;
848 
849 	if (params->flags & I915_OVERLAY_YUV_PLANAR) {
850 		int uv_hscale = uv_hsubsampling(params->flags);
851 		int uv_vscale = uv_vsubsampling(params->flags);
852 		u32 tmp_U, tmp_V;
853 
854 		swidth |= (params->src_width / uv_hscale) << 16;
855 		sheight |= (params->src_height / uv_vscale) << 16;
856 
857 		tmp_U = calc_swidthsw(dev_priv, params->offset_U,
858 				      params->src_width / uv_hscale);
859 		tmp_V = calc_swidthsw(dev_priv, params->offset_V,
860 				      params->src_width / uv_hscale);
861 		swidthsw |= max(tmp_U, tmp_V) << 16;
862 
863 		iowrite32(i915_ggtt_offset(vma) + params->offset_U,
864 			  &regs->OBUF_0U);
865 		iowrite32(i915_ggtt_offset(vma) + params->offset_V,
866 			  &regs->OBUF_0V);
867 
868 		ostride |= params->stride_UV << 16;
869 	}
870 
871 	iowrite32(swidth, &regs->SWIDTH);
872 	iowrite32(swidthsw, &regs->SWIDTHSW);
873 	iowrite32(sheight, &regs->SHEIGHT);
874 	iowrite32(ostride, &regs->OSTRIDE);
875 
876 	scale_changed = update_scaling_factors(overlay, regs, params);
877 
878 	update_colorkey(overlay, regs);
879 
880 	iowrite32(overlay_cmd_reg(params), &regs->OCMD);
881 
882 	ret = intel_overlay_continue(overlay, vma, scale_changed);
883 	if (ret)
884 		goto out_unpin;
885 
886 	return 0;
887 
888 out_unpin:
889 	i915_vma_unpin(vma);
890 out_pin_section:
891 	atomic_dec(&dev_priv->gpu_error.pending_fb_pin);
892 
893 	return ret;
894 }
895 
896 int intel_overlay_switch_off(struct intel_overlay *overlay)
897 {
898 	struct drm_i915_private *dev_priv = overlay->i915;
899 	int ret;
900 
901 	drm_WARN_ON(&dev_priv->drm,
902 		    !drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
903 
904 	ret = intel_overlay_recover_from_interrupt(overlay);
905 	if (ret != 0)
906 		return ret;
907 
908 	if (!overlay->active)
909 		return 0;
910 
911 	ret = intel_overlay_release_old_vid(overlay);
912 	if (ret != 0)
913 		return ret;
914 
915 	iowrite32(0, &overlay->regs->OCMD);
916 
917 	return intel_overlay_off(overlay);
918 }
919 
920 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
921 					  struct intel_crtc *crtc)
922 {
923 	if (!crtc->active)
924 		return -EINVAL;
925 
926 	/* can't use the overlay with double wide pipe */
927 	if (crtc->config->double_wide)
928 		return -EINVAL;
929 
930 	return 0;
931 }
932 
933 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
934 {
935 	struct drm_i915_private *dev_priv = overlay->i915;
936 	u32 pfit_control = intel_de_read(dev_priv, PFIT_CONTROL);
937 	u32 ratio;
938 
939 	/* XXX: This is not the same logic as in the xorg driver, but more in
940 	 * line with the intel documentation for the i965
941 	 */
942 	if (DISPLAY_VER(dev_priv) >= 4) {
943 		/* on i965 use the PGM reg to read out the autoscaler values */
944 		ratio = intel_de_read(dev_priv, PFIT_PGM_RATIOS) >> PFIT_VERT_SCALE_SHIFT_965;
945 	} else {
946 		if (pfit_control & VERT_AUTO_SCALE)
947 			ratio = intel_de_read(dev_priv, PFIT_AUTO_RATIOS);
948 		else
949 			ratio = intel_de_read(dev_priv, PFIT_PGM_RATIOS);
950 		ratio >>= PFIT_VERT_SCALE_SHIFT;
951 	}
952 
953 	overlay->pfit_vscale_ratio = ratio;
954 }
955 
956 static int check_overlay_dst(struct intel_overlay *overlay,
957 			     struct drm_intel_overlay_put_image *rec)
958 {
959 	const struct intel_crtc_state *pipe_config =
960 		overlay->crtc->config;
961 
962 	if (rec->dst_x < pipe_config->pipe_src_w &&
963 	    rec->dst_x + rec->dst_width <= pipe_config->pipe_src_w &&
964 	    rec->dst_y < pipe_config->pipe_src_h &&
965 	    rec->dst_y + rec->dst_height <= pipe_config->pipe_src_h)
966 		return 0;
967 	else
968 		return -EINVAL;
969 }
970 
971 static int check_overlay_scaling(struct drm_intel_overlay_put_image *rec)
972 {
973 	u32 tmp;
974 
975 	/* downscaling limit is 8.0 */
976 	tmp = ((rec->src_scan_height << 16) / rec->dst_height) >> 16;
977 	if (tmp > 7)
978 		return -EINVAL;
979 
980 	tmp = ((rec->src_scan_width << 16) / rec->dst_width) >> 16;
981 	if (tmp > 7)
982 		return -EINVAL;
983 
984 	return 0;
985 }
986 
987 static int check_overlay_src(struct drm_i915_private *dev_priv,
988 			     struct drm_intel_overlay_put_image *rec,
989 			     struct drm_i915_gem_object *new_bo)
990 {
991 	int uv_hscale = uv_hsubsampling(rec->flags);
992 	int uv_vscale = uv_vsubsampling(rec->flags);
993 	u32 stride_mask;
994 	int depth;
995 	u32 tmp;
996 
997 	/* check src dimensions */
998 	if (IS_I845G(dev_priv) || IS_I830(dev_priv)) {
999 		if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY ||
1000 		    rec->src_width  > IMAGE_MAX_WIDTH_LEGACY)
1001 			return -EINVAL;
1002 	} else {
1003 		if (rec->src_height > IMAGE_MAX_HEIGHT ||
1004 		    rec->src_width  > IMAGE_MAX_WIDTH)
1005 			return -EINVAL;
1006 	}
1007 
1008 	/* better safe than sorry, use 4 as the maximal subsampling ratio */
1009 	if (rec->src_height < N_VERT_Y_TAPS*4 ||
1010 	    rec->src_width  < N_HORIZ_Y_TAPS*4)
1011 		return -EINVAL;
1012 
1013 	/* check alignment constraints */
1014 	switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1015 	case I915_OVERLAY_RGB:
1016 		/* not implemented */
1017 		return -EINVAL;
1018 
1019 	case I915_OVERLAY_YUV_PACKED:
1020 		if (uv_vscale != 1)
1021 			return -EINVAL;
1022 
1023 		depth = packed_depth_bytes(rec->flags);
1024 		if (depth < 0)
1025 			return depth;
1026 
1027 		/* ignore UV planes */
1028 		rec->stride_UV = 0;
1029 		rec->offset_U = 0;
1030 		rec->offset_V = 0;
1031 		/* check pixel alignment */
1032 		if (rec->offset_Y % depth)
1033 			return -EINVAL;
1034 		break;
1035 
1036 	case I915_OVERLAY_YUV_PLANAR:
1037 		if (uv_vscale < 0 || uv_hscale < 0)
1038 			return -EINVAL;
1039 		/* no offset restrictions for planar formats */
1040 		break;
1041 
1042 	default:
1043 		return -EINVAL;
1044 	}
1045 
1046 	if (rec->src_width % uv_hscale)
1047 		return -EINVAL;
1048 
1049 	/* stride checking */
1050 	if (IS_I830(dev_priv) || IS_I845G(dev_priv))
1051 		stride_mask = 255;
1052 	else
1053 		stride_mask = 63;
1054 
1055 	if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
1056 		return -EINVAL;
1057 	if (DISPLAY_VER(dev_priv) == 4 && rec->stride_Y < 512)
1058 		return -EINVAL;
1059 
1060 	tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
1061 		4096 : 8192;
1062 	if (rec->stride_Y > tmp || rec->stride_UV > 2*1024)
1063 		return -EINVAL;
1064 
1065 	/* check buffer dimensions */
1066 	switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1067 	case I915_OVERLAY_RGB:
1068 	case I915_OVERLAY_YUV_PACKED:
1069 		/* always 4 Y values per depth pixels */
1070 		if (packed_width_bytes(rec->flags, rec->src_width) > rec->stride_Y)
1071 			return -EINVAL;
1072 
1073 		tmp = rec->stride_Y*rec->src_height;
1074 		if (rec->offset_Y + tmp > new_bo->base.size)
1075 			return -EINVAL;
1076 		break;
1077 
1078 	case I915_OVERLAY_YUV_PLANAR:
1079 		if (rec->src_width > rec->stride_Y)
1080 			return -EINVAL;
1081 		if (rec->src_width/uv_hscale > rec->stride_UV)
1082 			return -EINVAL;
1083 
1084 		tmp = rec->stride_Y * rec->src_height;
1085 		if (rec->offset_Y + tmp > new_bo->base.size)
1086 			return -EINVAL;
1087 
1088 		tmp = rec->stride_UV * (rec->src_height / uv_vscale);
1089 		if (rec->offset_U + tmp > new_bo->base.size ||
1090 		    rec->offset_V + tmp > new_bo->base.size)
1091 			return -EINVAL;
1092 		break;
1093 	}
1094 
1095 	return 0;
1096 }
1097 
1098 int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
1099 				  struct drm_file *file_priv)
1100 {
1101 	struct drm_intel_overlay_put_image *params = data;
1102 	struct drm_i915_private *dev_priv = to_i915(dev);
1103 	struct intel_overlay *overlay;
1104 	struct drm_crtc *drmmode_crtc;
1105 	struct intel_crtc *crtc;
1106 	struct drm_i915_gem_object *new_bo;
1107 	int ret;
1108 
1109 	overlay = dev_priv->overlay;
1110 	if (!overlay) {
1111 		drm_dbg(&dev_priv->drm, "userspace bug: no overlay\n");
1112 		return -ENODEV;
1113 	}
1114 
1115 	if (!(params->flags & I915_OVERLAY_ENABLE)) {
1116 		drm_modeset_lock_all(dev);
1117 		ret = intel_overlay_switch_off(overlay);
1118 		drm_modeset_unlock_all(dev);
1119 
1120 		return ret;
1121 	}
1122 
1123 	drmmode_crtc = drm_crtc_find(dev, file_priv, params->crtc_id);
1124 	if (!drmmode_crtc)
1125 		return -ENOENT;
1126 	crtc = to_intel_crtc(drmmode_crtc);
1127 
1128 	new_bo = i915_gem_object_lookup(file_priv, params->bo_handle);
1129 	if (!new_bo)
1130 		return -ENOENT;
1131 
1132 	drm_modeset_lock_all(dev);
1133 
1134 	if (i915_gem_object_is_tiled(new_bo)) {
1135 		drm_dbg_kms(&dev_priv->drm,
1136 			    "buffer used for overlay image can not be tiled\n");
1137 		ret = -EINVAL;
1138 		goto out_unlock;
1139 	}
1140 
1141 	ret = intel_overlay_recover_from_interrupt(overlay);
1142 	if (ret != 0)
1143 		goto out_unlock;
1144 
1145 	if (overlay->crtc != crtc) {
1146 		ret = intel_overlay_switch_off(overlay);
1147 		if (ret != 0)
1148 			goto out_unlock;
1149 
1150 		ret = check_overlay_possible_on_crtc(overlay, crtc);
1151 		if (ret != 0)
1152 			goto out_unlock;
1153 
1154 		overlay->crtc = crtc;
1155 		crtc->overlay = overlay;
1156 
1157 		/* line too wide, i.e. one-line-mode */
1158 		if (crtc->config->pipe_src_w > 1024 &&
1159 		    crtc->config->gmch_pfit.control & PFIT_ENABLE) {
1160 			overlay->pfit_active = true;
1161 			update_pfit_vscale_ratio(overlay);
1162 		} else
1163 			overlay->pfit_active = false;
1164 	}
1165 
1166 	ret = check_overlay_dst(overlay, params);
1167 	if (ret != 0)
1168 		goto out_unlock;
1169 
1170 	if (overlay->pfit_active) {
1171 		params->dst_y = (((u32)params->dst_y << 12) /
1172 				 overlay->pfit_vscale_ratio);
1173 		/* shifting right rounds downwards, so add 1 */
1174 		params->dst_height = (((u32)params->dst_height << 12) /
1175 				 overlay->pfit_vscale_ratio) + 1;
1176 	}
1177 
1178 	if (params->src_scan_height > params->src_height ||
1179 	    params->src_scan_width > params->src_width) {
1180 		ret = -EINVAL;
1181 		goto out_unlock;
1182 	}
1183 
1184 	ret = check_overlay_src(dev_priv, params, new_bo);
1185 	if (ret != 0)
1186 		goto out_unlock;
1187 
1188 	/* Check scaling after src size to prevent a divide-by-zero. */
1189 	ret = check_overlay_scaling(params);
1190 	if (ret != 0)
1191 		goto out_unlock;
1192 
1193 	ret = intel_overlay_do_put_image(overlay, new_bo, params);
1194 	if (ret != 0)
1195 		goto out_unlock;
1196 
1197 	drm_modeset_unlock_all(dev);
1198 	i915_gem_object_put(new_bo);
1199 
1200 	return 0;
1201 
1202 out_unlock:
1203 	drm_modeset_unlock_all(dev);
1204 	i915_gem_object_put(new_bo);
1205 
1206 	return ret;
1207 }
1208 
1209 static void update_reg_attrs(struct intel_overlay *overlay,
1210 			     struct overlay_registers __iomem *regs)
1211 {
1212 	iowrite32((overlay->contrast << 18) | (overlay->brightness & 0xff),
1213 		  &regs->OCLRC0);
1214 	iowrite32(overlay->saturation, &regs->OCLRC1);
1215 }
1216 
1217 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
1218 {
1219 	int i;
1220 
1221 	if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
1222 		return false;
1223 
1224 	for (i = 0; i < 3; i++) {
1225 		if (((gamma1 >> i*8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
1226 			return false;
1227 	}
1228 
1229 	return true;
1230 }
1231 
1232 static bool check_gamma5_errata(u32 gamma5)
1233 {
1234 	int i;
1235 
1236 	for (i = 0; i < 3; i++) {
1237 		if (((gamma5 >> i*8) & 0xff) == 0x80)
1238 			return false;
1239 	}
1240 
1241 	return true;
1242 }
1243 
1244 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
1245 {
1246 	if (!check_gamma_bounds(0, attrs->gamma0) ||
1247 	    !check_gamma_bounds(attrs->gamma0, attrs->gamma1) ||
1248 	    !check_gamma_bounds(attrs->gamma1, attrs->gamma2) ||
1249 	    !check_gamma_bounds(attrs->gamma2, attrs->gamma3) ||
1250 	    !check_gamma_bounds(attrs->gamma3, attrs->gamma4) ||
1251 	    !check_gamma_bounds(attrs->gamma4, attrs->gamma5) ||
1252 	    !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
1253 		return -EINVAL;
1254 
1255 	if (!check_gamma5_errata(attrs->gamma5))
1256 		return -EINVAL;
1257 
1258 	return 0;
1259 }
1260 
1261 int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
1262 			      struct drm_file *file_priv)
1263 {
1264 	struct drm_intel_overlay_attrs *attrs = data;
1265 	struct drm_i915_private *dev_priv = to_i915(dev);
1266 	struct intel_overlay *overlay;
1267 	int ret;
1268 
1269 	overlay = dev_priv->overlay;
1270 	if (!overlay) {
1271 		drm_dbg(&dev_priv->drm, "userspace bug: no overlay\n");
1272 		return -ENODEV;
1273 	}
1274 
1275 	drm_modeset_lock_all(dev);
1276 
1277 	ret = -EINVAL;
1278 	if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
1279 		attrs->color_key  = overlay->color_key;
1280 		attrs->brightness = overlay->brightness;
1281 		attrs->contrast   = overlay->contrast;
1282 		attrs->saturation = overlay->saturation;
1283 
1284 		if (DISPLAY_VER(dev_priv) != 2) {
1285 			attrs->gamma0 = intel_de_read(dev_priv, OGAMC0);
1286 			attrs->gamma1 = intel_de_read(dev_priv, OGAMC1);
1287 			attrs->gamma2 = intel_de_read(dev_priv, OGAMC2);
1288 			attrs->gamma3 = intel_de_read(dev_priv, OGAMC3);
1289 			attrs->gamma4 = intel_de_read(dev_priv, OGAMC4);
1290 			attrs->gamma5 = intel_de_read(dev_priv, OGAMC5);
1291 		}
1292 	} else {
1293 		if (attrs->brightness < -128 || attrs->brightness > 127)
1294 			goto out_unlock;
1295 		if (attrs->contrast > 255)
1296 			goto out_unlock;
1297 		if (attrs->saturation > 1023)
1298 			goto out_unlock;
1299 
1300 		overlay->color_key  = attrs->color_key;
1301 		overlay->brightness = attrs->brightness;
1302 		overlay->contrast   = attrs->contrast;
1303 		overlay->saturation = attrs->saturation;
1304 
1305 		update_reg_attrs(overlay, overlay->regs);
1306 
1307 		if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
1308 			if (DISPLAY_VER(dev_priv) == 2)
1309 				goto out_unlock;
1310 
1311 			if (overlay->active) {
1312 				ret = -EBUSY;
1313 				goto out_unlock;
1314 			}
1315 
1316 			ret = check_gamma(attrs);
1317 			if (ret)
1318 				goto out_unlock;
1319 
1320 			intel_de_write(dev_priv, OGAMC0, attrs->gamma0);
1321 			intel_de_write(dev_priv, OGAMC1, attrs->gamma1);
1322 			intel_de_write(dev_priv, OGAMC2, attrs->gamma2);
1323 			intel_de_write(dev_priv, OGAMC3, attrs->gamma3);
1324 			intel_de_write(dev_priv, OGAMC4, attrs->gamma4);
1325 			intel_de_write(dev_priv, OGAMC5, attrs->gamma5);
1326 		}
1327 	}
1328 	overlay->color_key_enabled = (attrs->flags & I915_OVERLAY_DISABLE_DEST_COLORKEY) == 0;
1329 
1330 	ret = 0;
1331 out_unlock:
1332 	drm_modeset_unlock_all(dev);
1333 
1334 	return ret;
1335 }
1336 
1337 static int get_registers(struct intel_overlay *overlay, bool use_phys)
1338 {
1339 	struct drm_i915_private *i915 = overlay->i915;
1340 	struct drm_i915_gem_object *obj;
1341 	struct i915_vma *vma;
1342 	int err;
1343 
1344 	obj = i915_gem_object_create_stolen(i915, PAGE_SIZE);
1345 	if (IS_ERR(obj))
1346 		obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
1347 	if (IS_ERR(obj))
1348 		return PTR_ERR(obj);
1349 
1350 	vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE);
1351 	if (IS_ERR(vma)) {
1352 		err = PTR_ERR(vma);
1353 		goto err_put_bo;
1354 	}
1355 
1356 	if (use_phys)
1357 		overlay->flip_addr = sg_dma_address(obj->mm.pages->sgl);
1358 	else
1359 		overlay->flip_addr = i915_ggtt_offset(vma);
1360 	overlay->regs = i915_vma_pin_iomap(vma);
1361 	i915_vma_unpin(vma);
1362 
1363 	if (IS_ERR(overlay->regs)) {
1364 		err = PTR_ERR(overlay->regs);
1365 		goto err_put_bo;
1366 	}
1367 
1368 	overlay->reg_bo = obj;
1369 	return 0;
1370 
1371 err_put_bo:
1372 	i915_gem_object_put(obj);
1373 	return err;
1374 }
1375 
1376 void intel_overlay_setup(struct drm_i915_private *dev_priv)
1377 {
1378 	struct intel_overlay *overlay;
1379 	struct intel_engine_cs *engine;
1380 	int ret;
1381 
1382 	if (!HAS_OVERLAY(dev_priv))
1383 		return;
1384 
1385 	engine = to_gt(dev_priv)->engine[RCS0];
1386 	if (!engine || !engine->kernel_context)
1387 		return;
1388 
1389 	overlay = kzalloc(sizeof(*overlay), GFP_KERNEL);
1390 	if (!overlay)
1391 		return;
1392 
1393 	overlay->i915 = dev_priv;
1394 	overlay->context = engine->kernel_context;
1395 	GEM_BUG_ON(!overlay->context);
1396 
1397 	overlay->color_key = 0x0101fe;
1398 	overlay->color_key_enabled = true;
1399 	overlay->brightness = -19;
1400 	overlay->contrast = 75;
1401 	overlay->saturation = 146;
1402 
1403 	i915_active_init(&overlay->last_flip,
1404 			 NULL, intel_overlay_last_flip_retire, 0);
1405 
1406 	ret = get_registers(overlay, OVERLAY_NEEDS_PHYSICAL(dev_priv));
1407 	if (ret)
1408 		goto out_free;
1409 
1410 	memset_io(overlay->regs, 0, sizeof(struct overlay_registers));
1411 	update_polyphase_filter(overlay->regs);
1412 	update_reg_attrs(overlay, overlay->regs);
1413 
1414 	dev_priv->overlay = overlay;
1415 	drm_info(&dev_priv->drm, "Initialized overlay support.\n");
1416 	return;
1417 
1418 out_free:
1419 	kfree(overlay);
1420 }
1421 
1422 void intel_overlay_cleanup(struct drm_i915_private *dev_priv)
1423 {
1424 	struct intel_overlay *overlay;
1425 
1426 	overlay = fetch_and_zero(&dev_priv->overlay);
1427 	if (!overlay)
1428 		return;
1429 
1430 	/*
1431 	 * The bo's should be free'd by the generic code already.
1432 	 * Furthermore modesetting teardown happens beforehand so the
1433 	 * hardware should be off already.
1434 	 */
1435 	drm_WARN_ON(&dev_priv->drm, overlay->active);
1436 
1437 	i915_gem_object_put(overlay->reg_bo);
1438 	i915_active_fini(&overlay->last_flip);
1439 
1440 	kfree(overlay);
1441 }
1442 
1443 #if IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR)
1444 
1445 struct intel_overlay_error_state {
1446 	struct overlay_registers regs;
1447 	unsigned long base;
1448 	u32 dovsta;
1449 	u32 isr;
1450 };
1451 
1452 struct intel_overlay_error_state *
1453 intel_overlay_capture_error_state(struct drm_i915_private *dev_priv)
1454 {
1455 	struct intel_overlay *overlay = dev_priv->overlay;
1456 	struct intel_overlay_error_state *error;
1457 
1458 	if (!overlay || !overlay->active)
1459 		return NULL;
1460 
1461 	error = kmalloc(sizeof(*error), GFP_ATOMIC);
1462 	if (error == NULL)
1463 		return NULL;
1464 
1465 	error->dovsta = intel_de_read(dev_priv, DOVSTA);
1466 	error->isr = intel_de_read(dev_priv, GEN2_ISR);
1467 	error->base = overlay->flip_addr;
1468 
1469 	memcpy_fromio(&error->regs, overlay->regs, sizeof(error->regs));
1470 
1471 	return error;
1472 }
1473 
1474 void
1475 intel_overlay_print_error_state(struct drm_i915_error_state_buf *m,
1476 				struct intel_overlay_error_state *error)
1477 {
1478 	i915_error_printf(m, "Overlay, status: 0x%08x, interrupt: 0x%08x\n",
1479 			  error->dovsta, error->isr);
1480 	i915_error_printf(m, "  Register file at 0x%08lx:\n",
1481 			  error->base);
1482 
1483 #define P(x) i915_error_printf(m, "    " #x ":	0x%08x\n", error->regs.x)
1484 	P(OBUF_0Y);
1485 	P(OBUF_1Y);
1486 	P(OBUF_0U);
1487 	P(OBUF_0V);
1488 	P(OBUF_1U);
1489 	P(OBUF_1V);
1490 	P(OSTRIDE);
1491 	P(YRGB_VPH);
1492 	P(UV_VPH);
1493 	P(HORZ_PH);
1494 	P(INIT_PHS);
1495 	P(DWINPOS);
1496 	P(DWINSZ);
1497 	P(SWIDTH);
1498 	P(SWIDTHSW);
1499 	P(SHEIGHT);
1500 	P(YRGBSCALE);
1501 	P(UVSCALE);
1502 	P(OCLRC0);
1503 	P(OCLRC1);
1504 	P(DCLRKV);
1505 	P(DCLRKM);
1506 	P(SCLRKVH);
1507 	P(SCLRKVL);
1508 	P(SCLRKEN);
1509 	P(OCONFIG);
1510 	P(OCMD);
1511 	P(OSTART_0Y);
1512 	P(OSTART_1Y);
1513 	P(OSTART_0U);
1514 	P(OSTART_0V);
1515 	P(OSTART_1U);
1516 	P(OSTART_1V);
1517 	P(OTILEOFF_0Y);
1518 	P(OTILEOFF_1Y);
1519 	P(OTILEOFF_0U);
1520 	P(OTILEOFF_0V);
1521 	P(OTILEOFF_1U);
1522 	P(OTILEOFF_1V);
1523 	P(FASTHSCALE);
1524 	P(UVSCALEV);
1525 #undef P
1526 }
1527 
1528 #endif
1529