xref: /linux/drivers/gpu/drm/tegra/dc.c (revision d87c25e8f4051f813762da6a182c57f246b17441)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (C) 2012 Avionic Design GmbH
4  * Copyright (C) 2012 NVIDIA CORPORATION.  All rights reserved.
5  */
6 
7 #include <linux/clk.h>
8 #include <linux/debugfs.h>
9 #include <linux/delay.h>
10 #include <linux/iommu.h>
11 #include <linux/interconnect.h>
12 #include <linux/module.h>
13 #include <linux/of_device.h>
14 #include <linux/pm_domain.h>
15 #include <linux/pm_opp.h>
16 #include <linux/pm_runtime.h>
17 #include <linux/reset.h>
18 
19 #include <soc/tegra/common.h>
20 #include <soc/tegra/pmc.h>
21 
22 #include <drm/drm_atomic.h>
23 #include <drm/drm_atomic_helper.h>
24 #include <drm/drm_debugfs.h>
25 #include <drm/drm_fourcc.h>
26 #include <drm/drm_plane_helper.h>
27 #include <drm/drm_vblank.h>
28 
29 #include "dc.h"
30 #include "drm.h"
31 #include "gem.h"
32 #include "hub.h"
33 #include "plane.h"
34 
35 static void tegra_crtc_atomic_destroy_state(struct drm_crtc *crtc,
36 					    struct drm_crtc_state *state);
37 
38 static void tegra_dc_stats_reset(struct tegra_dc_stats *stats)
39 {
40 	stats->frames = 0;
41 	stats->vblank = 0;
42 	stats->underflow = 0;
43 	stats->overflow = 0;
44 }
45 
46 /* Reads the active copy of a register. */
47 static u32 tegra_dc_readl_active(struct tegra_dc *dc, unsigned long offset)
48 {
49 	u32 value;
50 
51 	tegra_dc_writel(dc, READ_MUX, DC_CMD_STATE_ACCESS);
52 	value = tegra_dc_readl(dc, offset);
53 	tegra_dc_writel(dc, 0, DC_CMD_STATE_ACCESS);
54 
55 	return value;
56 }
57 
58 static inline unsigned int tegra_plane_offset(struct tegra_plane *plane,
59 					      unsigned int offset)
60 {
61 	if (offset >= 0x500 && offset <= 0x638) {
62 		offset = 0x000 + (offset - 0x500);
63 		return plane->offset + offset;
64 	}
65 
66 	if (offset >= 0x700 && offset <= 0x719) {
67 		offset = 0x180 + (offset - 0x700);
68 		return plane->offset + offset;
69 	}
70 
71 	if (offset >= 0x800 && offset <= 0x839) {
72 		offset = 0x1c0 + (offset - 0x800);
73 		return plane->offset + offset;
74 	}
75 
76 	dev_WARN(plane->dc->dev, "invalid offset: %x\n", offset);
77 
78 	return plane->offset + offset;
79 }
80 
81 static inline u32 tegra_plane_readl(struct tegra_plane *plane,
82 				    unsigned int offset)
83 {
84 	return tegra_dc_readl(plane->dc, tegra_plane_offset(plane, offset));
85 }
86 
87 static inline void tegra_plane_writel(struct tegra_plane *plane, u32 value,
88 				      unsigned int offset)
89 {
90 	tegra_dc_writel(plane->dc, value, tegra_plane_offset(plane, offset));
91 }
92 
93 bool tegra_dc_has_output(struct tegra_dc *dc, struct device *dev)
94 {
95 	struct device_node *np = dc->dev->of_node;
96 	struct of_phandle_iterator it;
97 	int err;
98 
99 	of_for_each_phandle(&it, err, np, "nvidia,outputs", NULL, 0)
100 		if (it.node == dev->of_node)
101 			return true;
102 
103 	return false;
104 }
105 
106 /*
107  * Double-buffered registers have two copies: ASSEMBLY and ACTIVE. When the
108  * *_ACT_REQ bits are set the ASSEMBLY copy is latched into the ACTIVE copy.
109  * Latching happens mmediately if the display controller is in STOP mode or
110  * on the next frame boundary otherwise.
111  *
112  * Triple-buffered registers have three copies: ASSEMBLY, ARM and ACTIVE. The
113  * ASSEMBLY copy is latched into the ARM copy immediately after *_UPDATE bits
114  * are written. When the *_ACT_REQ bits are written, the ARM copy is latched
115  * into the ACTIVE copy, either immediately if the display controller is in
116  * STOP mode, or at the next frame boundary otherwise.
117  */
118 void tegra_dc_commit(struct tegra_dc *dc)
119 {
120 	tegra_dc_writel(dc, GENERAL_ACT_REQ << 8, DC_CMD_STATE_CONTROL);
121 	tegra_dc_writel(dc, GENERAL_ACT_REQ, DC_CMD_STATE_CONTROL);
122 }
123 
124 static inline u32 compute_dda_inc(unsigned int in, unsigned int out, bool v,
125 				  unsigned int bpp)
126 {
127 	fixed20_12 outf = dfixed_init(out);
128 	fixed20_12 inf = dfixed_init(in);
129 	u32 dda_inc;
130 	int max;
131 
132 	if (v)
133 		max = 15;
134 	else {
135 		switch (bpp) {
136 		case 2:
137 			max = 8;
138 			break;
139 
140 		default:
141 			WARN_ON_ONCE(1);
142 			fallthrough;
143 		case 4:
144 			max = 4;
145 			break;
146 		}
147 	}
148 
149 	outf.full = max_t(u32, outf.full - dfixed_const(1), dfixed_const(1));
150 	inf.full -= dfixed_const(1);
151 
152 	dda_inc = dfixed_div(inf, outf);
153 	dda_inc = min_t(u32, dda_inc, dfixed_const(max));
154 
155 	return dda_inc;
156 }
157 
158 static inline u32 compute_initial_dda(unsigned int in)
159 {
160 	fixed20_12 inf = dfixed_init(in);
161 	return dfixed_frac(inf);
162 }
163 
164 static void tegra_plane_setup_blending_legacy(struct tegra_plane *plane)
165 {
166 	u32 background[3] = {
167 		BLEND_WEIGHT1(0) | BLEND_WEIGHT0(0) | BLEND_COLOR_KEY_NONE,
168 		BLEND_WEIGHT1(0) | BLEND_WEIGHT0(0) | BLEND_COLOR_KEY_NONE,
169 		BLEND_WEIGHT1(0) | BLEND_WEIGHT0(0) | BLEND_COLOR_KEY_NONE,
170 	};
171 	u32 foreground = BLEND_WEIGHT1(255) | BLEND_WEIGHT0(255) |
172 			 BLEND_COLOR_KEY_NONE;
173 	u32 blendnokey = BLEND_WEIGHT1(255) | BLEND_WEIGHT0(255);
174 	struct tegra_plane_state *state;
175 	u32 blending[2];
176 	unsigned int i;
177 
178 	/* disable blending for non-overlapping case */
179 	tegra_plane_writel(plane, blendnokey, DC_WIN_BLEND_NOKEY);
180 	tegra_plane_writel(plane, foreground, DC_WIN_BLEND_1WIN);
181 
182 	state = to_tegra_plane_state(plane->base.state);
183 
184 	if (state->opaque) {
185 		/*
186 		 * Since custom fix-weight blending isn't utilized and weight
187 		 * of top window is set to max, we can enforce dependent
188 		 * blending which in this case results in transparent bottom
189 		 * window if top window is opaque and if top window enables
190 		 * alpha blending, then bottom window is getting alpha value
191 		 * of 1 minus the sum of alpha components of the overlapping
192 		 * plane.
193 		 */
194 		background[0] |= BLEND_CONTROL_DEPENDENT;
195 		background[1] |= BLEND_CONTROL_DEPENDENT;
196 
197 		/*
198 		 * The region where three windows overlap is the intersection
199 		 * of the two regions where two windows overlap. It contributes
200 		 * to the area if all of the windows on top of it have an alpha
201 		 * component.
202 		 */
203 		switch (state->base.normalized_zpos) {
204 		case 0:
205 			if (state->blending[0].alpha &&
206 			    state->blending[1].alpha)
207 				background[2] |= BLEND_CONTROL_DEPENDENT;
208 			break;
209 
210 		case 1:
211 			background[2] |= BLEND_CONTROL_DEPENDENT;
212 			break;
213 		}
214 	} else {
215 		/*
216 		 * Enable alpha blending if pixel format has an alpha
217 		 * component.
218 		 */
219 		foreground |= BLEND_CONTROL_ALPHA;
220 
221 		/*
222 		 * If any of the windows on top of this window is opaque, it
223 		 * will completely conceal this window within that area. If
224 		 * top window has an alpha component, it is blended over the
225 		 * bottom window.
226 		 */
227 		for (i = 0; i < 2; i++) {
228 			if (state->blending[i].alpha &&
229 			    state->blending[i].top)
230 				background[i] |= BLEND_CONTROL_DEPENDENT;
231 		}
232 
233 		switch (state->base.normalized_zpos) {
234 		case 0:
235 			if (state->blending[0].alpha &&
236 			    state->blending[1].alpha)
237 				background[2] |= BLEND_CONTROL_DEPENDENT;
238 			break;
239 
240 		case 1:
241 			/*
242 			 * When both middle and topmost windows have an alpha,
243 			 * these windows a mixed together and then the result
244 			 * is blended over the bottom window.
245 			 */
246 			if (state->blending[0].alpha &&
247 			    state->blending[0].top)
248 				background[2] |= BLEND_CONTROL_ALPHA;
249 
250 			if (state->blending[1].alpha &&
251 			    state->blending[1].top)
252 				background[2] |= BLEND_CONTROL_ALPHA;
253 			break;
254 		}
255 	}
256 
257 	switch (state->base.normalized_zpos) {
258 	case 0:
259 		tegra_plane_writel(plane, background[0], DC_WIN_BLEND_2WIN_X);
260 		tegra_plane_writel(plane, background[1], DC_WIN_BLEND_2WIN_Y);
261 		tegra_plane_writel(plane, background[2], DC_WIN_BLEND_3WIN_XY);
262 		break;
263 
264 	case 1:
265 		/*
266 		 * If window B / C is topmost, then X / Y registers are
267 		 * matching the order of blending[...] state indices,
268 		 * otherwise a swap is required.
269 		 */
270 		if (!state->blending[0].top && state->blending[1].top) {
271 			blending[0] = foreground;
272 			blending[1] = background[1];
273 		} else {
274 			blending[0] = background[0];
275 			blending[1] = foreground;
276 		}
277 
278 		tegra_plane_writel(plane, blending[0], DC_WIN_BLEND_2WIN_X);
279 		tegra_plane_writel(plane, blending[1], DC_WIN_BLEND_2WIN_Y);
280 		tegra_plane_writel(plane, background[2], DC_WIN_BLEND_3WIN_XY);
281 		break;
282 
283 	case 2:
284 		tegra_plane_writel(plane, foreground, DC_WIN_BLEND_2WIN_X);
285 		tegra_plane_writel(plane, foreground, DC_WIN_BLEND_2WIN_Y);
286 		tegra_plane_writel(plane, foreground, DC_WIN_BLEND_3WIN_XY);
287 		break;
288 	}
289 }
290 
291 static void tegra_plane_setup_blending(struct tegra_plane *plane,
292 				       const struct tegra_dc_window *window)
293 {
294 	u32 value;
295 
296 	value = BLEND_FACTOR_DST_ALPHA_ZERO | BLEND_FACTOR_SRC_ALPHA_K2 |
297 		BLEND_FACTOR_DST_COLOR_NEG_K1_TIMES_SRC |
298 		BLEND_FACTOR_SRC_COLOR_K1_TIMES_SRC;
299 	tegra_plane_writel(plane, value, DC_WIN_BLEND_MATCH_SELECT);
300 
301 	value = BLEND_FACTOR_DST_ALPHA_ZERO | BLEND_FACTOR_SRC_ALPHA_K2 |
302 		BLEND_FACTOR_DST_COLOR_NEG_K1_TIMES_SRC |
303 		BLEND_FACTOR_SRC_COLOR_K1_TIMES_SRC;
304 	tegra_plane_writel(plane, value, DC_WIN_BLEND_NOMATCH_SELECT);
305 
306 	value = K2(255) | K1(255) | WINDOW_LAYER_DEPTH(255 - window->zpos);
307 	tegra_plane_writel(plane, value, DC_WIN_BLEND_LAYER_CONTROL);
308 }
309 
310 static bool
311 tegra_plane_use_horizontal_filtering(struct tegra_plane *plane,
312 				     const struct tegra_dc_window *window)
313 {
314 	struct tegra_dc *dc = plane->dc;
315 
316 	if (window->src.w == window->dst.w)
317 		return false;
318 
319 	if (plane->index == 0 && dc->soc->has_win_a_without_filters)
320 		return false;
321 
322 	return true;
323 }
324 
325 static bool
326 tegra_plane_use_vertical_filtering(struct tegra_plane *plane,
327 				   const struct tegra_dc_window *window)
328 {
329 	struct tegra_dc *dc = plane->dc;
330 
331 	if (window->src.h == window->dst.h)
332 		return false;
333 
334 	if (plane->index == 0 && dc->soc->has_win_a_without_filters)
335 		return false;
336 
337 	if (plane->index == 2 && dc->soc->has_win_c_without_vert_filter)
338 		return false;
339 
340 	return true;
341 }
342 
343 static void tegra_dc_setup_window(struct tegra_plane *plane,
344 				  const struct tegra_dc_window *window)
345 {
346 	unsigned h_offset, v_offset, h_size, v_size, h_dda, v_dda, bpp;
347 	struct tegra_dc *dc = plane->dc;
348 	bool yuv, planar;
349 	u32 value;
350 
351 	/*
352 	 * For YUV planar modes, the number of bytes per pixel takes into
353 	 * account only the luma component and therefore is 1.
354 	 */
355 	yuv = tegra_plane_format_is_yuv(window->format, &planar, NULL);
356 	if (!yuv)
357 		bpp = window->bits_per_pixel / 8;
358 	else
359 		bpp = planar ? 1 : 2;
360 
361 	tegra_plane_writel(plane, window->format, DC_WIN_COLOR_DEPTH);
362 	tegra_plane_writel(plane, window->swap, DC_WIN_BYTE_SWAP);
363 
364 	value = V_POSITION(window->dst.y) | H_POSITION(window->dst.x);
365 	tegra_plane_writel(plane, value, DC_WIN_POSITION);
366 
367 	value = V_SIZE(window->dst.h) | H_SIZE(window->dst.w);
368 	tegra_plane_writel(plane, value, DC_WIN_SIZE);
369 
370 	h_offset = window->src.x * bpp;
371 	v_offset = window->src.y;
372 	h_size = window->src.w * bpp;
373 	v_size = window->src.h;
374 
375 	if (window->reflect_x)
376 		h_offset += (window->src.w - 1) * bpp;
377 
378 	if (window->reflect_y)
379 		v_offset += window->src.h - 1;
380 
381 	value = V_PRESCALED_SIZE(v_size) | H_PRESCALED_SIZE(h_size);
382 	tegra_plane_writel(plane, value, DC_WIN_PRESCALED_SIZE);
383 
384 	/*
385 	 * For DDA computations the number of bytes per pixel for YUV planar
386 	 * modes needs to take into account all Y, U and V components.
387 	 */
388 	if (yuv && planar)
389 		bpp = 2;
390 
391 	h_dda = compute_dda_inc(window->src.w, window->dst.w, false, bpp);
392 	v_dda = compute_dda_inc(window->src.h, window->dst.h, true, bpp);
393 
394 	value = V_DDA_INC(v_dda) | H_DDA_INC(h_dda);
395 	tegra_plane_writel(plane, value, DC_WIN_DDA_INC);
396 
397 	h_dda = compute_initial_dda(window->src.x);
398 	v_dda = compute_initial_dda(window->src.y);
399 
400 	tegra_plane_writel(plane, h_dda, DC_WIN_H_INITIAL_DDA);
401 	tegra_plane_writel(plane, v_dda, DC_WIN_V_INITIAL_DDA);
402 
403 	tegra_plane_writel(plane, 0, DC_WIN_UV_BUF_STRIDE);
404 	tegra_plane_writel(plane, 0, DC_WIN_BUF_STRIDE);
405 
406 	tegra_plane_writel(plane, window->base[0], DC_WINBUF_START_ADDR);
407 
408 	if (yuv && planar) {
409 		tegra_plane_writel(plane, window->base[1], DC_WINBUF_START_ADDR_U);
410 		tegra_plane_writel(plane, window->base[2], DC_WINBUF_START_ADDR_V);
411 		value = window->stride[1] << 16 | window->stride[0];
412 		tegra_plane_writel(plane, value, DC_WIN_LINE_STRIDE);
413 	} else {
414 		tegra_plane_writel(plane, window->stride[0], DC_WIN_LINE_STRIDE);
415 	}
416 
417 	tegra_plane_writel(plane, h_offset, DC_WINBUF_ADDR_H_OFFSET);
418 	tegra_plane_writel(plane, v_offset, DC_WINBUF_ADDR_V_OFFSET);
419 
420 	if (dc->soc->supports_block_linear) {
421 		unsigned long height = window->tiling.value;
422 
423 		switch (window->tiling.mode) {
424 		case TEGRA_BO_TILING_MODE_PITCH:
425 			value = DC_WINBUF_SURFACE_KIND_PITCH;
426 			break;
427 
428 		case TEGRA_BO_TILING_MODE_TILED:
429 			value = DC_WINBUF_SURFACE_KIND_TILED;
430 			break;
431 
432 		case TEGRA_BO_TILING_MODE_BLOCK:
433 			value = DC_WINBUF_SURFACE_KIND_BLOCK_HEIGHT(height) |
434 				DC_WINBUF_SURFACE_KIND_BLOCK;
435 			break;
436 		}
437 
438 		tegra_plane_writel(plane, value, DC_WINBUF_SURFACE_KIND);
439 	} else {
440 		switch (window->tiling.mode) {
441 		case TEGRA_BO_TILING_MODE_PITCH:
442 			value = DC_WIN_BUFFER_ADDR_MODE_LINEAR_UV |
443 				DC_WIN_BUFFER_ADDR_MODE_LINEAR;
444 			break;
445 
446 		case TEGRA_BO_TILING_MODE_TILED:
447 			value = DC_WIN_BUFFER_ADDR_MODE_TILE_UV |
448 				DC_WIN_BUFFER_ADDR_MODE_TILE;
449 			break;
450 
451 		case TEGRA_BO_TILING_MODE_BLOCK:
452 			/*
453 			 * No need to handle this here because ->atomic_check
454 			 * will already have filtered it out.
455 			 */
456 			break;
457 		}
458 
459 		tegra_plane_writel(plane, value, DC_WIN_BUFFER_ADDR_MODE);
460 	}
461 
462 	value = WIN_ENABLE;
463 
464 	if (yuv) {
465 		/* setup default colorspace conversion coefficients */
466 		tegra_plane_writel(plane, 0x00f0, DC_WIN_CSC_YOF);
467 		tegra_plane_writel(plane, 0x012a, DC_WIN_CSC_KYRGB);
468 		tegra_plane_writel(plane, 0x0000, DC_WIN_CSC_KUR);
469 		tegra_plane_writel(plane, 0x0198, DC_WIN_CSC_KVR);
470 		tegra_plane_writel(plane, 0x039b, DC_WIN_CSC_KUG);
471 		tegra_plane_writel(plane, 0x032f, DC_WIN_CSC_KVG);
472 		tegra_plane_writel(plane, 0x0204, DC_WIN_CSC_KUB);
473 		tegra_plane_writel(plane, 0x0000, DC_WIN_CSC_KVB);
474 
475 		value |= CSC_ENABLE;
476 	} else if (window->bits_per_pixel < 24) {
477 		value |= COLOR_EXPAND;
478 	}
479 
480 	if (window->reflect_x)
481 		value |= H_DIRECTION;
482 
483 	if (window->reflect_y)
484 		value |= V_DIRECTION;
485 
486 	if (tegra_plane_use_horizontal_filtering(plane, window)) {
487 		/*
488 		 * Enable horizontal 6-tap filter and set filtering
489 		 * coefficients to the default values defined in TRM.
490 		 */
491 		tegra_plane_writel(plane, 0x00008000, DC_WIN_H_FILTER_P(0));
492 		tegra_plane_writel(plane, 0x3e087ce1, DC_WIN_H_FILTER_P(1));
493 		tegra_plane_writel(plane, 0x3b117ac1, DC_WIN_H_FILTER_P(2));
494 		tegra_plane_writel(plane, 0x591b73aa, DC_WIN_H_FILTER_P(3));
495 		tegra_plane_writel(plane, 0x57256d9a, DC_WIN_H_FILTER_P(4));
496 		tegra_plane_writel(plane, 0x552f668b, DC_WIN_H_FILTER_P(5));
497 		tegra_plane_writel(plane, 0x73385e8b, DC_WIN_H_FILTER_P(6));
498 		tegra_plane_writel(plane, 0x72435583, DC_WIN_H_FILTER_P(7));
499 		tegra_plane_writel(plane, 0x714c4c8b, DC_WIN_H_FILTER_P(8));
500 		tegra_plane_writel(plane, 0x70554393, DC_WIN_H_FILTER_P(9));
501 		tegra_plane_writel(plane, 0x715e389b, DC_WIN_H_FILTER_P(10));
502 		tegra_plane_writel(plane, 0x71662faa, DC_WIN_H_FILTER_P(11));
503 		tegra_plane_writel(plane, 0x536d25ba, DC_WIN_H_FILTER_P(12));
504 		tegra_plane_writel(plane, 0x55731bca, DC_WIN_H_FILTER_P(13));
505 		tegra_plane_writel(plane, 0x387a11d9, DC_WIN_H_FILTER_P(14));
506 		tegra_plane_writel(plane, 0x3c7c08f1, DC_WIN_H_FILTER_P(15));
507 
508 		value |= H_FILTER;
509 	}
510 
511 	if (tegra_plane_use_vertical_filtering(plane, window)) {
512 		unsigned int i, k;
513 
514 		/*
515 		 * Enable vertical 2-tap filter and set filtering
516 		 * coefficients to the default values defined in TRM.
517 		 */
518 		for (i = 0, k = 128; i < 16; i++, k -= 8)
519 			tegra_plane_writel(plane, k, DC_WIN_V_FILTER_P(i));
520 
521 		value |= V_FILTER;
522 	}
523 
524 	tegra_plane_writel(plane, value, DC_WIN_WIN_OPTIONS);
525 
526 	if (dc->soc->has_legacy_blending)
527 		tegra_plane_setup_blending_legacy(plane);
528 	else
529 		tegra_plane_setup_blending(plane, window);
530 }
531 
532 static const u32 tegra20_primary_formats[] = {
533 	DRM_FORMAT_ARGB4444,
534 	DRM_FORMAT_ARGB1555,
535 	DRM_FORMAT_RGB565,
536 	DRM_FORMAT_RGBA5551,
537 	DRM_FORMAT_ABGR8888,
538 	DRM_FORMAT_ARGB8888,
539 	/* non-native formats */
540 	DRM_FORMAT_XRGB1555,
541 	DRM_FORMAT_RGBX5551,
542 	DRM_FORMAT_XBGR8888,
543 	DRM_FORMAT_XRGB8888,
544 };
545 
546 static const u64 tegra20_modifiers[] = {
547 	DRM_FORMAT_MOD_LINEAR,
548 	DRM_FORMAT_MOD_NVIDIA_TEGRA_TILED,
549 	DRM_FORMAT_MOD_INVALID
550 };
551 
552 static const u32 tegra114_primary_formats[] = {
553 	DRM_FORMAT_ARGB4444,
554 	DRM_FORMAT_ARGB1555,
555 	DRM_FORMAT_RGB565,
556 	DRM_FORMAT_RGBA5551,
557 	DRM_FORMAT_ABGR8888,
558 	DRM_FORMAT_ARGB8888,
559 	/* new on Tegra114 */
560 	DRM_FORMAT_ABGR4444,
561 	DRM_FORMAT_ABGR1555,
562 	DRM_FORMAT_BGRA5551,
563 	DRM_FORMAT_XRGB1555,
564 	DRM_FORMAT_RGBX5551,
565 	DRM_FORMAT_XBGR1555,
566 	DRM_FORMAT_BGRX5551,
567 	DRM_FORMAT_BGR565,
568 	DRM_FORMAT_BGRA8888,
569 	DRM_FORMAT_RGBA8888,
570 	DRM_FORMAT_XRGB8888,
571 	DRM_FORMAT_XBGR8888,
572 };
573 
574 static const u32 tegra124_primary_formats[] = {
575 	DRM_FORMAT_ARGB4444,
576 	DRM_FORMAT_ARGB1555,
577 	DRM_FORMAT_RGB565,
578 	DRM_FORMAT_RGBA5551,
579 	DRM_FORMAT_ABGR8888,
580 	DRM_FORMAT_ARGB8888,
581 	/* new on Tegra114 */
582 	DRM_FORMAT_ABGR4444,
583 	DRM_FORMAT_ABGR1555,
584 	DRM_FORMAT_BGRA5551,
585 	DRM_FORMAT_XRGB1555,
586 	DRM_FORMAT_RGBX5551,
587 	DRM_FORMAT_XBGR1555,
588 	DRM_FORMAT_BGRX5551,
589 	DRM_FORMAT_BGR565,
590 	DRM_FORMAT_BGRA8888,
591 	DRM_FORMAT_RGBA8888,
592 	DRM_FORMAT_XRGB8888,
593 	DRM_FORMAT_XBGR8888,
594 	/* new on Tegra124 */
595 	DRM_FORMAT_RGBX8888,
596 	DRM_FORMAT_BGRX8888,
597 };
598 
599 static const u64 tegra124_modifiers[] = {
600 	DRM_FORMAT_MOD_LINEAR,
601 	DRM_FORMAT_MOD_NVIDIA_16BX2_BLOCK(0),
602 	DRM_FORMAT_MOD_NVIDIA_16BX2_BLOCK(1),
603 	DRM_FORMAT_MOD_NVIDIA_16BX2_BLOCK(2),
604 	DRM_FORMAT_MOD_NVIDIA_16BX2_BLOCK(3),
605 	DRM_FORMAT_MOD_NVIDIA_16BX2_BLOCK(4),
606 	DRM_FORMAT_MOD_NVIDIA_16BX2_BLOCK(5),
607 	DRM_FORMAT_MOD_INVALID
608 };
609 
610 static int tegra_plane_atomic_check(struct drm_plane *plane,
611 				    struct drm_atomic_state *state)
612 {
613 	struct drm_plane_state *new_plane_state = drm_atomic_get_new_plane_state(state,
614 										 plane);
615 	struct tegra_plane_state *plane_state = to_tegra_plane_state(new_plane_state);
616 	unsigned int supported_rotation = DRM_MODE_ROTATE_0 |
617 					  DRM_MODE_REFLECT_X |
618 					  DRM_MODE_REFLECT_Y;
619 	unsigned int rotation = new_plane_state->rotation;
620 	struct tegra_bo_tiling *tiling = &plane_state->tiling;
621 	struct tegra_plane *tegra = to_tegra_plane(plane);
622 	struct tegra_dc *dc = to_tegra_dc(new_plane_state->crtc);
623 	int err;
624 
625 	plane_state->peak_memory_bandwidth = 0;
626 	plane_state->avg_memory_bandwidth = 0;
627 
628 	/* no need for further checks if the plane is being disabled */
629 	if (!new_plane_state->crtc) {
630 		plane_state->total_peak_memory_bandwidth = 0;
631 		return 0;
632 	}
633 
634 	err = tegra_plane_format(new_plane_state->fb->format->format,
635 				 &plane_state->format,
636 				 &plane_state->swap);
637 	if (err < 0)
638 		return err;
639 
640 	/*
641 	 * Tegra20 and Tegra30 are special cases here because they support
642 	 * only variants of specific formats with an alpha component, but not
643 	 * the corresponding opaque formats. However, the opaque formats can
644 	 * be emulated by disabling alpha blending for the plane.
645 	 */
646 	if (dc->soc->has_legacy_blending) {
647 		err = tegra_plane_setup_legacy_state(tegra, plane_state);
648 		if (err < 0)
649 			return err;
650 	}
651 
652 	err = tegra_fb_get_tiling(new_plane_state->fb, tiling);
653 	if (err < 0)
654 		return err;
655 
656 	if (tiling->mode == TEGRA_BO_TILING_MODE_BLOCK &&
657 	    !dc->soc->supports_block_linear) {
658 		DRM_ERROR("hardware doesn't support block linear mode\n");
659 		return -EINVAL;
660 	}
661 
662 	/*
663 	 * Older userspace used custom BO flag in order to specify the Y
664 	 * reflection, while modern userspace uses the generic DRM rotation
665 	 * property in order to achieve the same result.  The legacy BO flag
666 	 * duplicates the DRM rotation property when both are set.
667 	 */
668 	if (tegra_fb_is_bottom_up(new_plane_state->fb))
669 		rotation |= DRM_MODE_REFLECT_Y;
670 
671 	rotation = drm_rotation_simplify(rotation, supported_rotation);
672 
673 	if (rotation & DRM_MODE_REFLECT_X)
674 		plane_state->reflect_x = true;
675 	else
676 		plane_state->reflect_x = false;
677 
678 	if (rotation & DRM_MODE_REFLECT_Y)
679 		plane_state->reflect_y = true;
680 	else
681 		plane_state->reflect_y = false;
682 
683 	/*
684 	 * Tegra doesn't support different strides for U and V planes so we
685 	 * error out if the user tries to display a framebuffer with such a
686 	 * configuration.
687 	 */
688 	if (new_plane_state->fb->format->num_planes > 2) {
689 		if (new_plane_state->fb->pitches[2] != new_plane_state->fb->pitches[1]) {
690 			DRM_ERROR("unsupported UV-plane configuration\n");
691 			return -EINVAL;
692 		}
693 	}
694 
695 	err = tegra_plane_state_add(tegra, new_plane_state);
696 	if (err < 0)
697 		return err;
698 
699 	return 0;
700 }
701 
702 static void tegra_plane_atomic_disable(struct drm_plane *plane,
703 				       struct drm_atomic_state *state)
704 {
705 	struct drm_plane_state *old_state = drm_atomic_get_old_plane_state(state,
706 									   plane);
707 	struct tegra_plane *p = to_tegra_plane(plane);
708 	u32 value;
709 
710 	/* rien ne va plus */
711 	if (!old_state || !old_state->crtc)
712 		return;
713 
714 	value = tegra_plane_readl(p, DC_WIN_WIN_OPTIONS);
715 	value &= ~WIN_ENABLE;
716 	tegra_plane_writel(p, value, DC_WIN_WIN_OPTIONS);
717 }
718 
719 static void tegra_plane_atomic_update(struct drm_plane *plane,
720 				      struct drm_atomic_state *state)
721 {
722 	struct drm_plane_state *new_state = drm_atomic_get_new_plane_state(state,
723 									   plane);
724 	struct tegra_plane_state *tegra_plane_state = to_tegra_plane_state(new_state);
725 	struct drm_framebuffer *fb = new_state->fb;
726 	struct tegra_plane *p = to_tegra_plane(plane);
727 	struct tegra_dc_window window;
728 	unsigned int i;
729 
730 	/* rien ne va plus */
731 	if (!new_state->crtc || !new_state->fb)
732 		return;
733 
734 	if (!new_state->visible)
735 		return tegra_plane_atomic_disable(plane, state);
736 
737 	memset(&window, 0, sizeof(window));
738 	window.src.x = new_state->src.x1 >> 16;
739 	window.src.y = new_state->src.y1 >> 16;
740 	window.src.w = drm_rect_width(&new_state->src) >> 16;
741 	window.src.h = drm_rect_height(&new_state->src) >> 16;
742 	window.dst.x = new_state->dst.x1;
743 	window.dst.y = new_state->dst.y1;
744 	window.dst.w = drm_rect_width(&new_state->dst);
745 	window.dst.h = drm_rect_height(&new_state->dst);
746 	window.bits_per_pixel = fb->format->cpp[0] * 8;
747 	window.reflect_x = tegra_plane_state->reflect_x;
748 	window.reflect_y = tegra_plane_state->reflect_y;
749 
750 	/* copy from state */
751 	window.zpos = new_state->normalized_zpos;
752 	window.tiling = tegra_plane_state->tiling;
753 	window.format = tegra_plane_state->format;
754 	window.swap = tegra_plane_state->swap;
755 
756 	for (i = 0; i < fb->format->num_planes; i++) {
757 		window.base[i] = tegra_plane_state->iova[i] + fb->offsets[i];
758 
759 		/*
760 		 * Tegra uses a shared stride for UV planes. Framebuffers are
761 		 * already checked for this in the tegra_plane_atomic_check()
762 		 * function, so it's safe to ignore the V-plane pitch here.
763 		 */
764 		if (i < 2)
765 			window.stride[i] = fb->pitches[i];
766 	}
767 
768 	tegra_dc_setup_window(p, &window);
769 }
770 
771 static const struct drm_plane_helper_funcs tegra_plane_helper_funcs = {
772 	.prepare_fb = tegra_plane_prepare_fb,
773 	.cleanup_fb = tegra_plane_cleanup_fb,
774 	.atomic_check = tegra_plane_atomic_check,
775 	.atomic_disable = tegra_plane_atomic_disable,
776 	.atomic_update = tegra_plane_atomic_update,
777 };
778 
779 static unsigned long tegra_plane_get_possible_crtcs(struct drm_device *drm)
780 {
781 	/*
782 	 * Ideally this would use drm_crtc_mask(), but that would require the
783 	 * CRTC to already be in the mode_config's list of CRTCs. However, it
784 	 * will only be added to that list in the drm_crtc_init_with_planes()
785 	 * (in tegra_dc_init()), which in turn requires registration of these
786 	 * planes. So we have ourselves a nice little chicken and egg problem
787 	 * here.
788 	 *
789 	 * We work around this by manually creating the mask from the number
790 	 * of CRTCs that have been registered, and should therefore always be
791 	 * the same as drm_crtc_index() after registration.
792 	 */
793 	return 1 << drm->mode_config.num_crtc;
794 }
795 
796 static struct drm_plane *tegra_primary_plane_create(struct drm_device *drm,
797 						    struct tegra_dc *dc)
798 {
799 	unsigned long possible_crtcs = tegra_plane_get_possible_crtcs(drm);
800 	enum drm_plane_type type = DRM_PLANE_TYPE_PRIMARY;
801 	struct tegra_plane *plane;
802 	unsigned int num_formats;
803 	const u64 *modifiers;
804 	const u32 *formats;
805 	int err;
806 
807 	plane = kzalloc(sizeof(*plane), GFP_KERNEL);
808 	if (!plane)
809 		return ERR_PTR(-ENOMEM);
810 
811 	/* Always use window A as primary window */
812 	plane->offset = 0xa00;
813 	plane->index = 0;
814 	plane->dc = dc;
815 
816 	num_formats = dc->soc->num_primary_formats;
817 	formats = dc->soc->primary_formats;
818 	modifiers = dc->soc->modifiers;
819 
820 	err = tegra_plane_interconnect_init(plane);
821 	if (err) {
822 		kfree(plane);
823 		return ERR_PTR(err);
824 	}
825 
826 	err = drm_universal_plane_init(drm, &plane->base, possible_crtcs,
827 				       &tegra_plane_funcs, formats,
828 				       num_formats, modifiers, type, NULL);
829 	if (err < 0) {
830 		kfree(plane);
831 		return ERR_PTR(err);
832 	}
833 
834 	drm_plane_helper_add(&plane->base, &tegra_plane_helper_funcs);
835 	drm_plane_create_zpos_property(&plane->base, plane->index, 0, 255);
836 
837 	err = drm_plane_create_rotation_property(&plane->base,
838 						 DRM_MODE_ROTATE_0,
839 						 DRM_MODE_ROTATE_0 |
840 						 DRM_MODE_ROTATE_180 |
841 						 DRM_MODE_REFLECT_X |
842 						 DRM_MODE_REFLECT_Y);
843 	if (err < 0)
844 		dev_err(dc->dev, "failed to create rotation property: %d\n",
845 			err);
846 
847 	return &plane->base;
848 }
849 
850 static const u32 tegra_legacy_cursor_plane_formats[] = {
851 	DRM_FORMAT_RGBA8888,
852 };
853 
854 static const u32 tegra_cursor_plane_formats[] = {
855 	DRM_FORMAT_ARGB8888,
856 };
857 
858 static int tegra_cursor_atomic_check(struct drm_plane *plane,
859 				     struct drm_atomic_state *state)
860 {
861 	struct drm_plane_state *new_plane_state = drm_atomic_get_new_plane_state(state,
862 										 plane);
863 	struct tegra_plane_state *plane_state = to_tegra_plane_state(new_plane_state);
864 	struct tegra_plane *tegra = to_tegra_plane(plane);
865 	int err;
866 
867 	plane_state->peak_memory_bandwidth = 0;
868 	plane_state->avg_memory_bandwidth = 0;
869 
870 	/* no need for further checks if the plane is being disabled */
871 	if (!new_plane_state->crtc) {
872 		plane_state->total_peak_memory_bandwidth = 0;
873 		return 0;
874 	}
875 
876 	/* scaling not supported for cursor */
877 	if ((new_plane_state->src_w >> 16 != new_plane_state->crtc_w) ||
878 	    (new_plane_state->src_h >> 16 != new_plane_state->crtc_h))
879 		return -EINVAL;
880 
881 	/* only square cursors supported */
882 	if (new_plane_state->src_w != new_plane_state->src_h)
883 		return -EINVAL;
884 
885 	if (new_plane_state->crtc_w != 32 && new_plane_state->crtc_w != 64 &&
886 	    new_plane_state->crtc_w != 128 && new_plane_state->crtc_w != 256)
887 		return -EINVAL;
888 
889 	err = tegra_plane_state_add(tegra, new_plane_state);
890 	if (err < 0)
891 		return err;
892 
893 	return 0;
894 }
895 
896 static void __tegra_cursor_atomic_update(struct drm_plane *plane,
897 					 struct drm_plane_state *new_state)
898 {
899 	struct tegra_plane_state *tegra_plane_state = to_tegra_plane_state(new_state);
900 	struct tegra_dc *dc = to_tegra_dc(new_state->crtc);
901 	struct tegra_drm *tegra = plane->dev->dev_private;
902 #ifdef CONFIG_ARCH_DMA_ADDR_T_64BIT
903 	u64 dma_mask = *dc->dev->dma_mask;
904 #endif
905 	unsigned int x, y;
906 	u32 value = 0;
907 
908 	/* rien ne va plus */
909 	if (!new_state->crtc || !new_state->fb)
910 		return;
911 
912 	/*
913 	 * Legacy display supports hardware clipping of the cursor, but
914 	 * nvdisplay relies on software to clip the cursor to the screen.
915 	 */
916 	if (!dc->soc->has_nvdisplay)
917 		value |= CURSOR_CLIP_DISPLAY;
918 
919 	switch (new_state->crtc_w) {
920 	case 32:
921 		value |= CURSOR_SIZE_32x32;
922 		break;
923 
924 	case 64:
925 		value |= CURSOR_SIZE_64x64;
926 		break;
927 
928 	case 128:
929 		value |= CURSOR_SIZE_128x128;
930 		break;
931 
932 	case 256:
933 		value |= CURSOR_SIZE_256x256;
934 		break;
935 
936 	default:
937 		WARN(1, "cursor size %ux%u not supported\n",
938 		     new_state->crtc_w, new_state->crtc_h);
939 		return;
940 	}
941 
942 	value |= (tegra_plane_state->iova[0] >> 10) & 0x3fffff;
943 	tegra_dc_writel(dc, value, DC_DISP_CURSOR_START_ADDR);
944 
945 #ifdef CONFIG_ARCH_DMA_ADDR_T_64BIT
946 	value = (tegra_plane_state->iova[0] >> 32) & (dma_mask >> 32);
947 	tegra_dc_writel(dc, value, DC_DISP_CURSOR_START_ADDR_HI);
948 #endif
949 
950 	/* enable cursor and set blend mode */
951 	value = tegra_dc_readl(dc, DC_DISP_DISP_WIN_OPTIONS);
952 	value |= CURSOR_ENABLE;
953 	tegra_dc_writel(dc, value, DC_DISP_DISP_WIN_OPTIONS);
954 
955 	value = tegra_dc_readl(dc, DC_DISP_BLEND_CURSOR_CONTROL);
956 	value &= ~CURSOR_DST_BLEND_MASK;
957 	value &= ~CURSOR_SRC_BLEND_MASK;
958 
959 	if (dc->soc->has_nvdisplay)
960 		value &= ~CURSOR_COMPOSITION_MODE_XOR;
961 	else
962 		value |= CURSOR_MODE_NORMAL;
963 
964 	value |= CURSOR_DST_BLEND_NEG_K1_TIMES_SRC;
965 	value |= CURSOR_SRC_BLEND_K1_TIMES_SRC;
966 	value |= CURSOR_ALPHA;
967 	tegra_dc_writel(dc, value, DC_DISP_BLEND_CURSOR_CONTROL);
968 
969 	/* nvdisplay relies on software for clipping */
970 	if (dc->soc->has_nvdisplay) {
971 		struct drm_rect src;
972 
973 		x = new_state->dst.x1;
974 		y = new_state->dst.y1;
975 
976 		drm_rect_fp_to_int(&src, &new_state->src);
977 
978 		value = (src.y1 & tegra->vmask) << 16 | (src.x1 & tegra->hmask);
979 		tegra_dc_writel(dc, value, DC_DISP_PCALC_HEAD_SET_CROPPED_POINT_IN_CURSOR);
980 
981 		value = (drm_rect_height(&src) & tegra->vmask) << 16 |
982 			(drm_rect_width(&src) & tegra->hmask);
983 		tegra_dc_writel(dc, value, DC_DISP_PCALC_HEAD_SET_CROPPED_SIZE_IN_CURSOR);
984 	} else {
985 		x = new_state->crtc_x;
986 		y = new_state->crtc_y;
987 	}
988 
989 	/* position the cursor */
990 	value = ((y & tegra->vmask) << 16) | (x & tegra->hmask);
991 	tegra_dc_writel(dc, value, DC_DISP_CURSOR_POSITION);
992 }
993 
994 static void tegra_cursor_atomic_update(struct drm_plane *plane,
995 				       struct drm_atomic_state *state)
996 {
997 	struct drm_plane_state *new_state = drm_atomic_get_new_plane_state(state, plane);
998 
999 	__tegra_cursor_atomic_update(plane, new_state);
1000 }
1001 
1002 static void tegra_cursor_atomic_disable(struct drm_plane *plane,
1003 					struct drm_atomic_state *state)
1004 {
1005 	struct drm_plane_state *old_state = drm_atomic_get_old_plane_state(state,
1006 									   plane);
1007 	struct tegra_dc *dc;
1008 	u32 value;
1009 
1010 	/* rien ne va plus */
1011 	if (!old_state || !old_state->crtc)
1012 		return;
1013 
1014 	dc = to_tegra_dc(old_state->crtc);
1015 
1016 	value = tegra_dc_readl(dc, DC_DISP_DISP_WIN_OPTIONS);
1017 	value &= ~CURSOR_ENABLE;
1018 	tegra_dc_writel(dc, value, DC_DISP_DISP_WIN_OPTIONS);
1019 }
1020 
1021 static int tegra_cursor_atomic_async_check(struct drm_plane *plane, struct drm_atomic_state *state)
1022 {
1023 	struct drm_plane_state *new_state = drm_atomic_get_new_plane_state(state, plane);
1024 	struct drm_crtc_state *crtc_state;
1025 	int min_scale, max_scale;
1026 	int err;
1027 
1028 	crtc_state = drm_atomic_get_existing_crtc_state(state, new_state->crtc);
1029 	if (WARN_ON(!crtc_state))
1030 		return -EINVAL;
1031 
1032 	if (!crtc_state->active)
1033 		return -EINVAL;
1034 
1035 	if (plane->state->crtc != new_state->crtc ||
1036 	    plane->state->src_w != new_state->src_w ||
1037 	    plane->state->src_h != new_state->src_h ||
1038 	    plane->state->crtc_w != new_state->crtc_w ||
1039 	    plane->state->crtc_h != new_state->crtc_h ||
1040 	    plane->state->fb != new_state->fb ||
1041 	    plane->state->fb == NULL)
1042 		return -EINVAL;
1043 
1044 	min_scale = (1 << 16) / 8;
1045 	max_scale = (8 << 16) / 1;
1046 
1047 	err = drm_atomic_helper_check_plane_state(new_state, crtc_state, min_scale, max_scale,
1048 						  true, true);
1049 	if (err < 0)
1050 		return err;
1051 
1052 	if (new_state->visible != plane->state->visible)
1053 		return -EINVAL;
1054 
1055 	return 0;
1056 }
1057 
1058 static void tegra_cursor_atomic_async_update(struct drm_plane *plane,
1059 					     struct drm_atomic_state *state)
1060 {
1061 	struct drm_plane_state *new_state = drm_atomic_get_new_plane_state(state, plane);
1062 	struct tegra_dc *dc = to_tegra_dc(new_state->crtc);
1063 
1064 	plane->state->src_x = new_state->src_x;
1065 	plane->state->src_y = new_state->src_y;
1066 	plane->state->crtc_x = new_state->crtc_x;
1067 	plane->state->crtc_y = new_state->crtc_y;
1068 
1069 	if (new_state->visible) {
1070 		struct tegra_plane *p = to_tegra_plane(plane);
1071 		u32 value;
1072 
1073 		__tegra_cursor_atomic_update(plane, new_state);
1074 
1075 		value = (WIN_A_ACT_REQ << p->index) << 8 | GENERAL_UPDATE;
1076 		tegra_dc_writel(dc, value, DC_CMD_STATE_CONTROL);
1077 		(void)tegra_dc_readl(dc, DC_CMD_STATE_CONTROL);
1078 
1079 		value = (WIN_A_ACT_REQ << p->index) | GENERAL_ACT_REQ;
1080 		tegra_dc_writel(dc, value, DC_CMD_STATE_CONTROL);
1081 		(void)tegra_dc_readl(dc, DC_CMD_STATE_CONTROL);
1082 	}
1083 }
1084 
1085 static const struct drm_plane_helper_funcs tegra_cursor_plane_helper_funcs = {
1086 	.prepare_fb = tegra_plane_prepare_fb,
1087 	.cleanup_fb = tegra_plane_cleanup_fb,
1088 	.atomic_check = tegra_cursor_atomic_check,
1089 	.atomic_update = tegra_cursor_atomic_update,
1090 	.atomic_disable = tegra_cursor_atomic_disable,
1091 	.atomic_async_check = tegra_cursor_atomic_async_check,
1092 	.atomic_async_update = tegra_cursor_atomic_async_update,
1093 };
1094 
1095 static const uint64_t linear_modifiers[] = {
1096 	DRM_FORMAT_MOD_LINEAR,
1097 	DRM_FORMAT_MOD_INVALID
1098 };
1099 
1100 static struct drm_plane *tegra_dc_cursor_plane_create(struct drm_device *drm,
1101 						      struct tegra_dc *dc)
1102 {
1103 	unsigned long possible_crtcs = tegra_plane_get_possible_crtcs(drm);
1104 	struct tegra_plane *plane;
1105 	unsigned int num_formats;
1106 	const u32 *formats;
1107 	int err;
1108 
1109 	plane = kzalloc(sizeof(*plane), GFP_KERNEL);
1110 	if (!plane)
1111 		return ERR_PTR(-ENOMEM);
1112 
1113 	/*
1114 	 * This index is kind of fake. The cursor isn't a regular plane, but
1115 	 * its update and activation request bits in DC_CMD_STATE_CONTROL do
1116 	 * use the same programming. Setting this fake index here allows the
1117 	 * code in tegra_add_plane_state() to do the right thing without the
1118 	 * need to special-casing the cursor plane.
1119 	 */
1120 	plane->index = 6;
1121 	plane->dc = dc;
1122 
1123 	if (!dc->soc->has_nvdisplay) {
1124 		num_formats = ARRAY_SIZE(tegra_legacy_cursor_plane_formats);
1125 		formats = tegra_legacy_cursor_plane_formats;
1126 
1127 		err = tegra_plane_interconnect_init(plane);
1128 		if (err) {
1129 			kfree(plane);
1130 			return ERR_PTR(err);
1131 		}
1132 	} else {
1133 		num_formats = ARRAY_SIZE(tegra_cursor_plane_formats);
1134 		formats = tegra_cursor_plane_formats;
1135 	}
1136 
1137 	err = drm_universal_plane_init(drm, &plane->base, possible_crtcs,
1138 				       &tegra_plane_funcs, formats,
1139 				       num_formats, linear_modifiers,
1140 				       DRM_PLANE_TYPE_CURSOR, NULL);
1141 	if (err < 0) {
1142 		kfree(plane);
1143 		return ERR_PTR(err);
1144 	}
1145 
1146 	drm_plane_helper_add(&plane->base, &tegra_cursor_plane_helper_funcs);
1147 	drm_plane_create_zpos_immutable_property(&plane->base, 255);
1148 
1149 	return &plane->base;
1150 }
1151 
1152 static const u32 tegra20_overlay_formats[] = {
1153 	DRM_FORMAT_ARGB4444,
1154 	DRM_FORMAT_ARGB1555,
1155 	DRM_FORMAT_RGB565,
1156 	DRM_FORMAT_RGBA5551,
1157 	DRM_FORMAT_ABGR8888,
1158 	DRM_FORMAT_ARGB8888,
1159 	/* non-native formats */
1160 	DRM_FORMAT_XRGB1555,
1161 	DRM_FORMAT_RGBX5551,
1162 	DRM_FORMAT_XBGR8888,
1163 	DRM_FORMAT_XRGB8888,
1164 	/* planar formats */
1165 	DRM_FORMAT_UYVY,
1166 	DRM_FORMAT_YUYV,
1167 	DRM_FORMAT_YUV420,
1168 	DRM_FORMAT_YUV422,
1169 };
1170 
1171 static const u32 tegra114_overlay_formats[] = {
1172 	DRM_FORMAT_ARGB4444,
1173 	DRM_FORMAT_ARGB1555,
1174 	DRM_FORMAT_RGB565,
1175 	DRM_FORMAT_RGBA5551,
1176 	DRM_FORMAT_ABGR8888,
1177 	DRM_FORMAT_ARGB8888,
1178 	/* new on Tegra114 */
1179 	DRM_FORMAT_ABGR4444,
1180 	DRM_FORMAT_ABGR1555,
1181 	DRM_FORMAT_BGRA5551,
1182 	DRM_FORMAT_XRGB1555,
1183 	DRM_FORMAT_RGBX5551,
1184 	DRM_FORMAT_XBGR1555,
1185 	DRM_FORMAT_BGRX5551,
1186 	DRM_FORMAT_BGR565,
1187 	DRM_FORMAT_BGRA8888,
1188 	DRM_FORMAT_RGBA8888,
1189 	DRM_FORMAT_XRGB8888,
1190 	DRM_FORMAT_XBGR8888,
1191 	/* planar formats */
1192 	DRM_FORMAT_UYVY,
1193 	DRM_FORMAT_YUYV,
1194 	DRM_FORMAT_YUV420,
1195 	DRM_FORMAT_YUV422,
1196 };
1197 
1198 static const u32 tegra124_overlay_formats[] = {
1199 	DRM_FORMAT_ARGB4444,
1200 	DRM_FORMAT_ARGB1555,
1201 	DRM_FORMAT_RGB565,
1202 	DRM_FORMAT_RGBA5551,
1203 	DRM_FORMAT_ABGR8888,
1204 	DRM_FORMAT_ARGB8888,
1205 	/* new on Tegra114 */
1206 	DRM_FORMAT_ABGR4444,
1207 	DRM_FORMAT_ABGR1555,
1208 	DRM_FORMAT_BGRA5551,
1209 	DRM_FORMAT_XRGB1555,
1210 	DRM_FORMAT_RGBX5551,
1211 	DRM_FORMAT_XBGR1555,
1212 	DRM_FORMAT_BGRX5551,
1213 	DRM_FORMAT_BGR565,
1214 	DRM_FORMAT_BGRA8888,
1215 	DRM_FORMAT_RGBA8888,
1216 	DRM_FORMAT_XRGB8888,
1217 	DRM_FORMAT_XBGR8888,
1218 	/* new on Tegra124 */
1219 	DRM_FORMAT_RGBX8888,
1220 	DRM_FORMAT_BGRX8888,
1221 	/* planar formats */
1222 	DRM_FORMAT_UYVY,
1223 	DRM_FORMAT_YUYV,
1224 	DRM_FORMAT_YUV420,
1225 	DRM_FORMAT_YUV422,
1226 };
1227 
1228 static struct drm_plane *tegra_dc_overlay_plane_create(struct drm_device *drm,
1229 						       struct tegra_dc *dc,
1230 						       unsigned int index,
1231 						       bool cursor)
1232 {
1233 	unsigned long possible_crtcs = tegra_plane_get_possible_crtcs(drm);
1234 	struct tegra_plane *plane;
1235 	unsigned int num_formats;
1236 	enum drm_plane_type type;
1237 	const u32 *formats;
1238 	int err;
1239 
1240 	plane = kzalloc(sizeof(*plane), GFP_KERNEL);
1241 	if (!plane)
1242 		return ERR_PTR(-ENOMEM);
1243 
1244 	plane->offset = 0xa00 + 0x200 * index;
1245 	plane->index = index;
1246 	plane->dc = dc;
1247 
1248 	num_formats = dc->soc->num_overlay_formats;
1249 	formats = dc->soc->overlay_formats;
1250 
1251 	err = tegra_plane_interconnect_init(plane);
1252 	if (err) {
1253 		kfree(plane);
1254 		return ERR_PTR(err);
1255 	}
1256 
1257 	if (!cursor)
1258 		type = DRM_PLANE_TYPE_OVERLAY;
1259 	else
1260 		type = DRM_PLANE_TYPE_CURSOR;
1261 
1262 	err = drm_universal_plane_init(drm, &plane->base, possible_crtcs,
1263 				       &tegra_plane_funcs, formats,
1264 				       num_formats, linear_modifiers,
1265 				       type, NULL);
1266 	if (err < 0) {
1267 		kfree(plane);
1268 		return ERR_PTR(err);
1269 	}
1270 
1271 	drm_plane_helper_add(&plane->base, &tegra_plane_helper_funcs);
1272 	drm_plane_create_zpos_property(&plane->base, plane->index, 0, 255);
1273 
1274 	err = drm_plane_create_rotation_property(&plane->base,
1275 						 DRM_MODE_ROTATE_0,
1276 						 DRM_MODE_ROTATE_0 |
1277 						 DRM_MODE_ROTATE_180 |
1278 						 DRM_MODE_REFLECT_X |
1279 						 DRM_MODE_REFLECT_Y);
1280 	if (err < 0)
1281 		dev_err(dc->dev, "failed to create rotation property: %d\n",
1282 			err);
1283 
1284 	return &plane->base;
1285 }
1286 
1287 static struct drm_plane *tegra_dc_add_shared_planes(struct drm_device *drm,
1288 						    struct tegra_dc *dc)
1289 {
1290 	struct drm_plane *plane, *primary = NULL;
1291 	unsigned int i, j;
1292 
1293 	for (i = 0; i < dc->soc->num_wgrps; i++) {
1294 		const struct tegra_windowgroup_soc *wgrp = &dc->soc->wgrps[i];
1295 
1296 		if (wgrp->dc == dc->pipe) {
1297 			for (j = 0; j < wgrp->num_windows; j++) {
1298 				unsigned int index = wgrp->windows[j];
1299 
1300 				plane = tegra_shared_plane_create(drm, dc,
1301 								  wgrp->index,
1302 								  index);
1303 				if (IS_ERR(plane))
1304 					return plane;
1305 
1306 				/*
1307 				 * Choose the first shared plane owned by this
1308 				 * head as the primary plane.
1309 				 */
1310 				if (!primary) {
1311 					plane->type = DRM_PLANE_TYPE_PRIMARY;
1312 					primary = plane;
1313 				}
1314 			}
1315 		}
1316 	}
1317 
1318 	return primary;
1319 }
1320 
1321 static struct drm_plane *tegra_dc_add_planes(struct drm_device *drm,
1322 					     struct tegra_dc *dc)
1323 {
1324 	struct drm_plane *planes[2], *primary;
1325 	unsigned int planes_num;
1326 	unsigned int i;
1327 	int err;
1328 
1329 	primary = tegra_primary_plane_create(drm, dc);
1330 	if (IS_ERR(primary))
1331 		return primary;
1332 
1333 	if (dc->soc->supports_cursor)
1334 		planes_num = 2;
1335 	else
1336 		planes_num = 1;
1337 
1338 	for (i = 0; i < planes_num; i++) {
1339 		planes[i] = tegra_dc_overlay_plane_create(drm, dc, 1 + i,
1340 							  false);
1341 		if (IS_ERR(planes[i])) {
1342 			err = PTR_ERR(planes[i]);
1343 
1344 			while (i--)
1345 				planes[i]->funcs->destroy(planes[i]);
1346 
1347 			primary->funcs->destroy(primary);
1348 			return ERR_PTR(err);
1349 		}
1350 	}
1351 
1352 	return primary;
1353 }
1354 
1355 static void tegra_dc_destroy(struct drm_crtc *crtc)
1356 {
1357 	drm_crtc_cleanup(crtc);
1358 }
1359 
1360 static void tegra_crtc_reset(struct drm_crtc *crtc)
1361 {
1362 	struct tegra_dc_state *state = kzalloc(sizeof(*state), GFP_KERNEL);
1363 
1364 	if (crtc->state)
1365 		tegra_crtc_atomic_destroy_state(crtc, crtc->state);
1366 
1367 	__drm_atomic_helper_crtc_reset(crtc, &state->base);
1368 }
1369 
1370 static struct drm_crtc_state *
1371 tegra_crtc_atomic_duplicate_state(struct drm_crtc *crtc)
1372 {
1373 	struct tegra_dc_state *state = to_dc_state(crtc->state);
1374 	struct tegra_dc_state *copy;
1375 
1376 	copy = kmalloc(sizeof(*copy), GFP_KERNEL);
1377 	if (!copy)
1378 		return NULL;
1379 
1380 	__drm_atomic_helper_crtc_duplicate_state(crtc, &copy->base);
1381 	copy->clk = state->clk;
1382 	copy->pclk = state->pclk;
1383 	copy->div = state->div;
1384 	copy->planes = state->planes;
1385 
1386 	return &copy->base;
1387 }
1388 
1389 static void tegra_crtc_atomic_destroy_state(struct drm_crtc *crtc,
1390 					    struct drm_crtc_state *state)
1391 {
1392 	__drm_atomic_helper_crtc_destroy_state(state);
1393 	kfree(state);
1394 }
1395 
1396 #define DEBUGFS_REG32(_name) { .name = #_name, .offset = _name }
1397 
1398 static const struct debugfs_reg32 tegra_dc_regs[] = {
1399 	DEBUGFS_REG32(DC_CMD_GENERAL_INCR_SYNCPT),
1400 	DEBUGFS_REG32(DC_CMD_GENERAL_INCR_SYNCPT_CNTRL),
1401 	DEBUGFS_REG32(DC_CMD_GENERAL_INCR_SYNCPT_ERROR),
1402 	DEBUGFS_REG32(DC_CMD_WIN_A_INCR_SYNCPT),
1403 	DEBUGFS_REG32(DC_CMD_WIN_A_INCR_SYNCPT_CNTRL),
1404 	DEBUGFS_REG32(DC_CMD_WIN_A_INCR_SYNCPT_ERROR),
1405 	DEBUGFS_REG32(DC_CMD_WIN_B_INCR_SYNCPT),
1406 	DEBUGFS_REG32(DC_CMD_WIN_B_INCR_SYNCPT_CNTRL),
1407 	DEBUGFS_REG32(DC_CMD_WIN_B_INCR_SYNCPT_ERROR),
1408 	DEBUGFS_REG32(DC_CMD_WIN_C_INCR_SYNCPT),
1409 	DEBUGFS_REG32(DC_CMD_WIN_C_INCR_SYNCPT_CNTRL),
1410 	DEBUGFS_REG32(DC_CMD_WIN_C_INCR_SYNCPT_ERROR),
1411 	DEBUGFS_REG32(DC_CMD_CONT_SYNCPT_VSYNC),
1412 	DEBUGFS_REG32(DC_CMD_DISPLAY_COMMAND_OPTION0),
1413 	DEBUGFS_REG32(DC_CMD_DISPLAY_COMMAND),
1414 	DEBUGFS_REG32(DC_CMD_SIGNAL_RAISE),
1415 	DEBUGFS_REG32(DC_CMD_DISPLAY_POWER_CONTROL),
1416 	DEBUGFS_REG32(DC_CMD_INT_STATUS),
1417 	DEBUGFS_REG32(DC_CMD_INT_MASK),
1418 	DEBUGFS_REG32(DC_CMD_INT_ENABLE),
1419 	DEBUGFS_REG32(DC_CMD_INT_TYPE),
1420 	DEBUGFS_REG32(DC_CMD_INT_POLARITY),
1421 	DEBUGFS_REG32(DC_CMD_SIGNAL_RAISE1),
1422 	DEBUGFS_REG32(DC_CMD_SIGNAL_RAISE2),
1423 	DEBUGFS_REG32(DC_CMD_SIGNAL_RAISE3),
1424 	DEBUGFS_REG32(DC_CMD_STATE_ACCESS),
1425 	DEBUGFS_REG32(DC_CMD_STATE_CONTROL),
1426 	DEBUGFS_REG32(DC_CMD_DISPLAY_WINDOW_HEADER),
1427 	DEBUGFS_REG32(DC_CMD_REG_ACT_CONTROL),
1428 	DEBUGFS_REG32(DC_COM_CRC_CONTROL),
1429 	DEBUGFS_REG32(DC_COM_CRC_CHECKSUM),
1430 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_ENABLE(0)),
1431 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_ENABLE(1)),
1432 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_ENABLE(2)),
1433 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_ENABLE(3)),
1434 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_POLARITY(0)),
1435 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_POLARITY(1)),
1436 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_POLARITY(2)),
1437 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_POLARITY(3)),
1438 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_DATA(0)),
1439 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_DATA(1)),
1440 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_DATA(2)),
1441 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_DATA(3)),
1442 	DEBUGFS_REG32(DC_COM_PIN_INPUT_ENABLE(0)),
1443 	DEBUGFS_REG32(DC_COM_PIN_INPUT_ENABLE(1)),
1444 	DEBUGFS_REG32(DC_COM_PIN_INPUT_ENABLE(2)),
1445 	DEBUGFS_REG32(DC_COM_PIN_INPUT_ENABLE(3)),
1446 	DEBUGFS_REG32(DC_COM_PIN_INPUT_DATA(0)),
1447 	DEBUGFS_REG32(DC_COM_PIN_INPUT_DATA(1)),
1448 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_SELECT(0)),
1449 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_SELECT(1)),
1450 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_SELECT(2)),
1451 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_SELECT(3)),
1452 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_SELECT(4)),
1453 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_SELECT(5)),
1454 	DEBUGFS_REG32(DC_COM_PIN_OUTPUT_SELECT(6)),
1455 	DEBUGFS_REG32(DC_COM_PIN_MISC_CONTROL),
1456 	DEBUGFS_REG32(DC_COM_PIN_PM0_CONTROL),
1457 	DEBUGFS_REG32(DC_COM_PIN_PM0_DUTY_CYCLE),
1458 	DEBUGFS_REG32(DC_COM_PIN_PM1_CONTROL),
1459 	DEBUGFS_REG32(DC_COM_PIN_PM1_DUTY_CYCLE),
1460 	DEBUGFS_REG32(DC_COM_SPI_CONTROL),
1461 	DEBUGFS_REG32(DC_COM_SPI_START_BYTE),
1462 	DEBUGFS_REG32(DC_COM_HSPI_WRITE_DATA_AB),
1463 	DEBUGFS_REG32(DC_COM_HSPI_WRITE_DATA_CD),
1464 	DEBUGFS_REG32(DC_COM_HSPI_CS_DC),
1465 	DEBUGFS_REG32(DC_COM_SCRATCH_REGISTER_A),
1466 	DEBUGFS_REG32(DC_COM_SCRATCH_REGISTER_B),
1467 	DEBUGFS_REG32(DC_COM_GPIO_CTRL),
1468 	DEBUGFS_REG32(DC_COM_GPIO_DEBOUNCE_COUNTER),
1469 	DEBUGFS_REG32(DC_COM_CRC_CHECKSUM_LATCHED),
1470 	DEBUGFS_REG32(DC_DISP_DISP_SIGNAL_OPTIONS0),
1471 	DEBUGFS_REG32(DC_DISP_DISP_SIGNAL_OPTIONS1),
1472 	DEBUGFS_REG32(DC_DISP_DISP_WIN_OPTIONS),
1473 	DEBUGFS_REG32(DC_DISP_DISP_MEM_HIGH_PRIORITY),
1474 	DEBUGFS_REG32(DC_DISP_DISP_MEM_HIGH_PRIORITY_TIMER),
1475 	DEBUGFS_REG32(DC_DISP_DISP_TIMING_OPTIONS),
1476 	DEBUGFS_REG32(DC_DISP_REF_TO_SYNC),
1477 	DEBUGFS_REG32(DC_DISP_SYNC_WIDTH),
1478 	DEBUGFS_REG32(DC_DISP_BACK_PORCH),
1479 	DEBUGFS_REG32(DC_DISP_ACTIVE),
1480 	DEBUGFS_REG32(DC_DISP_FRONT_PORCH),
1481 	DEBUGFS_REG32(DC_DISP_H_PULSE0_CONTROL),
1482 	DEBUGFS_REG32(DC_DISP_H_PULSE0_POSITION_A),
1483 	DEBUGFS_REG32(DC_DISP_H_PULSE0_POSITION_B),
1484 	DEBUGFS_REG32(DC_DISP_H_PULSE0_POSITION_C),
1485 	DEBUGFS_REG32(DC_DISP_H_PULSE0_POSITION_D),
1486 	DEBUGFS_REG32(DC_DISP_H_PULSE1_CONTROL),
1487 	DEBUGFS_REG32(DC_DISP_H_PULSE1_POSITION_A),
1488 	DEBUGFS_REG32(DC_DISP_H_PULSE1_POSITION_B),
1489 	DEBUGFS_REG32(DC_DISP_H_PULSE1_POSITION_C),
1490 	DEBUGFS_REG32(DC_DISP_H_PULSE1_POSITION_D),
1491 	DEBUGFS_REG32(DC_DISP_H_PULSE2_CONTROL),
1492 	DEBUGFS_REG32(DC_DISP_H_PULSE2_POSITION_A),
1493 	DEBUGFS_REG32(DC_DISP_H_PULSE2_POSITION_B),
1494 	DEBUGFS_REG32(DC_DISP_H_PULSE2_POSITION_C),
1495 	DEBUGFS_REG32(DC_DISP_H_PULSE2_POSITION_D),
1496 	DEBUGFS_REG32(DC_DISP_V_PULSE0_CONTROL),
1497 	DEBUGFS_REG32(DC_DISP_V_PULSE0_POSITION_A),
1498 	DEBUGFS_REG32(DC_DISP_V_PULSE0_POSITION_B),
1499 	DEBUGFS_REG32(DC_DISP_V_PULSE0_POSITION_C),
1500 	DEBUGFS_REG32(DC_DISP_V_PULSE1_CONTROL),
1501 	DEBUGFS_REG32(DC_DISP_V_PULSE1_POSITION_A),
1502 	DEBUGFS_REG32(DC_DISP_V_PULSE1_POSITION_B),
1503 	DEBUGFS_REG32(DC_DISP_V_PULSE1_POSITION_C),
1504 	DEBUGFS_REG32(DC_DISP_V_PULSE2_CONTROL),
1505 	DEBUGFS_REG32(DC_DISP_V_PULSE2_POSITION_A),
1506 	DEBUGFS_REG32(DC_DISP_V_PULSE3_CONTROL),
1507 	DEBUGFS_REG32(DC_DISP_V_PULSE3_POSITION_A),
1508 	DEBUGFS_REG32(DC_DISP_M0_CONTROL),
1509 	DEBUGFS_REG32(DC_DISP_M1_CONTROL),
1510 	DEBUGFS_REG32(DC_DISP_DI_CONTROL),
1511 	DEBUGFS_REG32(DC_DISP_PP_CONTROL),
1512 	DEBUGFS_REG32(DC_DISP_PP_SELECT_A),
1513 	DEBUGFS_REG32(DC_DISP_PP_SELECT_B),
1514 	DEBUGFS_REG32(DC_DISP_PP_SELECT_C),
1515 	DEBUGFS_REG32(DC_DISP_PP_SELECT_D),
1516 	DEBUGFS_REG32(DC_DISP_DISP_CLOCK_CONTROL),
1517 	DEBUGFS_REG32(DC_DISP_DISP_INTERFACE_CONTROL),
1518 	DEBUGFS_REG32(DC_DISP_DISP_COLOR_CONTROL),
1519 	DEBUGFS_REG32(DC_DISP_SHIFT_CLOCK_OPTIONS),
1520 	DEBUGFS_REG32(DC_DISP_DATA_ENABLE_OPTIONS),
1521 	DEBUGFS_REG32(DC_DISP_SERIAL_INTERFACE_OPTIONS),
1522 	DEBUGFS_REG32(DC_DISP_LCD_SPI_OPTIONS),
1523 	DEBUGFS_REG32(DC_DISP_BORDER_COLOR),
1524 	DEBUGFS_REG32(DC_DISP_COLOR_KEY0_LOWER),
1525 	DEBUGFS_REG32(DC_DISP_COLOR_KEY0_UPPER),
1526 	DEBUGFS_REG32(DC_DISP_COLOR_KEY1_LOWER),
1527 	DEBUGFS_REG32(DC_DISP_COLOR_KEY1_UPPER),
1528 	DEBUGFS_REG32(DC_DISP_CURSOR_FOREGROUND),
1529 	DEBUGFS_REG32(DC_DISP_CURSOR_BACKGROUND),
1530 	DEBUGFS_REG32(DC_DISP_CURSOR_START_ADDR),
1531 	DEBUGFS_REG32(DC_DISP_CURSOR_START_ADDR_NS),
1532 	DEBUGFS_REG32(DC_DISP_CURSOR_POSITION),
1533 	DEBUGFS_REG32(DC_DISP_CURSOR_POSITION_NS),
1534 	DEBUGFS_REG32(DC_DISP_INIT_SEQ_CONTROL),
1535 	DEBUGFS_REG32(DC_DISP_SPI_INIT_SEQ_DATA_A),
1536 	DEBUGFS_REG32(DC_DISP_SPI_INIT_SEQ_DATA_B),
1537 	DEBUGFS_REG32(DC_DISP_SPI_INIT_SEQ_DATA_C),
1538 	DEBUGFS_REG32(DC_DISP_SPI_INIT_SEQ_DATA_D),
1539 	DEBUGFS_REG32(DC_DISP_DC_MCCIF_FIFOCTRL),
1540 	DEBUGFS_REG32(DC_DISP_MCCIF_DISPLAY0A_HYST),
1541 	DEBUGFS_REG32(DC_DISP_MCCIF_DISPLAY0B_HYST),
1542 	DEBUGFS_REG32(DC_DISP_MCCIF_DISPLAY1A_HYST),
1543 	DEBUGFS_REG32(DC_DISP_MCCIF_DISPLAY1B_HYST),
1544 	DEBUGFS_REG32(DC_DISP_DAC_CRT_CTRL),
1545 	DEBUGFS_REG32(DC_DISP_DISP_MISC_CONTROL),
1546 	DEBUGFS_REG32(DC_DISP_SD_CONTROL),
1547 	DEBUGFS_REG32(DC_DISP_SD_CSC_COEFF),
1548 	DEBUGFS_REG32(DC_DISP_SD_LUT(0)),
1549 	DEBUGFS_REG32(DC_DISP_SD_LUT(1)),
1550 	DEBUGFS_REG32(DC_DISP_SD_LUT(2)),
1551 	DEBUGFS_REG32(DC_DISP_SD_LUT(3)),
1552 	DEBUGFS_REG32(DC_DISP_SD_LUT(4)),
1553 	DEBUGFS_REG32(DC_DISP_SD_LUT(5)),
1554 	DEBUGFS_REG32(DC_DISP_SD_LUT(6)),
1555 	DEBUGFS_REG32(DC_DISP_SD_LUT(7)),
1556 	DEBUGFS_REG32(DC_DISP_SD_LUT(8)),
1557 	DEBUGFS_REG32(DC_DISP_SD_FLICKER_CONTROL),
1558 	DEBUGFS_REG32(DC_DISP_DC_PIXEL_COUNT),
1559 	DEBUGFS_REG32(DC_DISP_SD_HISTOGRAM(0)),
1560 	DEBUGFS_REG32(DC_DISP_SD_HISTOGRAM(1)),
1561 	DEBUGFS_REG32(DC_DISP_SD_HISTOGRAM(2)),
1562 	DEBUGFS_REG32(DC_DISP_SD_HISTOGRAM(3)),
1563 	DEBUGFS_REG32(DC_DISP_SD_HISTOGRAM(4)),
1564 	DEBUGFS_REG32(DC_DISP_SD_HISTOGRAM(5)),
1565 	DEBUGFS_REG32(DC_DISP_SD_HISTOGRAM(6)),
1566 	DEBUGFS_REG32(DC_DISP_SD_HISTOGRAM(7)),
1567 	DEBUGFS_REG32(DC_DISP_SD_BL_TF(0)),
1568 	DEBUGFS_REG32(DC_DISP_SD_BL_TF(1)),
1569 	DEBUGFS_REG32(DC_DISP_SD_BL_TF(2)),
1570 	DEBUGFS_REG32(DC_DISP_SD_BL_TF(3)),
1571 	DEBUGFS_REG32(DC_DISP_SD_BL_CONTROL),
1572 	DEBUGFS_REG32(DC_DISP_SD_HW_K_VALUES),
1573 	DEBUGFS_REG32(DC_DISP_SD_MAN_K_VALUES),
1574 	DEBUGFS_REG32(DC_DISP_CURSOR_START_ADDR_HI),
1575 	DEBUGFS_REG32(DC_DISP_BLEND_CURSOR_CONTROL),
1576 	DEBUGFS_REG32(DC_WIN_WIN_OPTIONS),
1577 	DEBUGFS_REG32(DC_WIN_BYTE_SWAP),
1578 	DEBUGFS_REG32(DC_WIN_BUFFER_CONTROL),
1579 	DEBUGFS_REG32(DC_WIN_COLOR_DEPTH),
1580 	DEBUGFS_REG32(DC_WIN_POSITION),
1581 	DEBUGFS_REG32(DC_WIN_SIZE),
1582 	DEBUGFS_REG32(DC_WIN_PRESCALED_SIZE),
1583 	DEBUGFS_REG32(DC_WIN_H_INITIAL_DDA),
1584 	DEBUGFS_REG32(DC_WIN_V_INITIAL_DDA),
1585 	DEBUGFS_REG32(DC_WIN_DDA_INC),
1586 	DEBUGFS_REG32(DC_WIN_LINE_STRIDE),
1587 	DEBUGFS_REG32(DC_WIN_BUF_STRIDE),
1588 	DEBUGFS_REG32(DC_WIN_UV_BUF_STRIDE),
1589 	DEBUGFS_REG32(DC_WIN_BUFFER_ADDR_MODE),
1590 	DEBUGFS_REG32(DC_WIN_DV_CONTROL),
1591 	DEBUGFS_REG32(DC_WIN_BLEND_NOKEY),
1592 	DEBUGFS_REG32(DC_WIN_BLEND_1WIN),
1593 	DEBUGFS_REG32(DC_WIN_BLEND_2WIN_X),
1594 	DEBUGFS_REG32(DC_WIN_BLEND_2WIN_Y),
1595 	DEBUGFS_REG32(DC_WIN_BLEND_3WIN_XY),
1596 	DEBUGFS_REG32(DC_WIN_HP_FETCH_CONTROL),
1597 	DEBUGFS_REG32(DC_WINBUF_START_ADDR),
1598 	DEBUGFS_REG32(DC_WINBUF_START_ADDR_NS),
1599 	DEBUGFS_REG32(DC_WINBUF_START_ADDR_U),
1600 	DEBUGFS_REG32(DC_WINBUF_START_ADDR_U_NS),
1601 	DEBUGFS_REG32(DC_WINBUF_START_ADDR_V),
1602 	DEBUGFS_REG32(DC_WINBUF_START_ADDR_V_NS),
1603 	DEBUGFS_REG32(DC_WINBUF_ADDR_H_OFFSET),
1604 	DEBUGFS_REG32(DC_WINBUF_ADDR_H_OFFSET_NS),
1605 	DEBUGFS_REG32(DC_WINBUF_ADDR_V_OFFSET),
1606 	DEBUGFS_REG32(DC_WINBUF_ADDR_V_OFFSET_NS),
1607 	DEBUGFS_REG32(DC_WINBUF_UFLOW_STATUS),
1608 	DEBUGFS_REG32(DC_WINBUF_AD_UFLOW_STATUS),
1609 	DEBUGFS_REG32(DC_WINBUF_BD_UFLOW_STATUS),
1610 	DEBUGFS_REG32(DC_WINBUF_CD_UFLOW_STATUS),
1611 };
1612 
1613 static int tegra_dc_show_regs(struct seq_file *s, void *data)
1614 {
1615 	struct drm_info_node *node = s->private;
1616 	struct tegra_dc *dc = node->info_ent->data;
1617 	unsigned int i;
1618 	int err = 0;
1619 
1620 	drm_modeset_lock(&dc->base.mutex, NULL);
1621 
1622 	if (!dc->base.state->active) {
1623 		err = -EBUSY;
1624 		goto unlock;
1625 	}
1626 
1627 	for (i = 0; i < ARRAY_SIZE(tegra_dc_regs); i++) {
1628 		unsigned int offset = tegra_dc_regs[i].offset;
1629 
1630 		seq_printf(s, "%-40s %#05x %08x\n", tegra_dc_regs[i].name,
1631 			   offset, tegra_dc_readl(dc, offset));
1632 	}
1633 
1634 unlock:
1635 	drm_modeset_unlock(&dc->base.mutex);
1636 	return err;
1637 }
1638 
1639 static int tegra_dc_show_crc(struct seq_file *s, void *data)
1640 {
1641 	struct drm_info_node *node = s->private;
1642 	struct tegra_dc *dc = node->info_ent->data;
1643 	int err = 0;
1644 	u32 value;
1645 
1646 	drm_modeset_lock(&dc->base.mutex, NULL);
1647 
1648 	if (!dc->base.state->active) {
1649 		err = -EBUSY;
1650 		goto unlock;
1651 	}
1652 
1653 	value = DC_COM_CRC_CONTROL_ACTIVE_DATA | DC_COM_CRC_CONTROL_ENABLE;
1654 	tegra_dc_writel(dc, value, DC_COM_CRC_CONTROL);
1655 	tegra_dc_commit(dc);
1656 
1657 	drm_crtc_wait_one_vblank(&dc->base);
1658 	drm_crtc_wait_one_vblank(&dc->base);
1659 
1660 	value = tegra_dc_readl(dc, DC_COM_CRC_CHECKSUM);
1661 	seq_printf(s, "%08x\n", value);
1662 
1663 	tegra_dc_writel(dc, 0, DC_COM_CRC_CONTROL);
1664 
1665 unlock:
1666 	drm_modeset_unlock(&dc->base.mutex);
1667 	return err;
1668 }
1669 
1670 static int tegra_dc_show_stats(struct seq_file *s, void *data)
1671 {
1672 	struct drm_info_node *node = s->private;
1673 	struct tegra_dc *dc = node->info_ent->data;
1674 
1675 	seq_printf(s, "frames: %lu\n", dc->stats.frames);
1676 	seq_printf(s, "vblank: %lu\n", dc->stats.vblank);
1677 	seq_printf(s, "underflow: %lu\n", dc->stats.underflow);
1678 	seq_printf(s, "overflow: %lu\n", dc->stats.overflow);
1679 
1680 	seq_printf(s, "frames total: %lu\n", dc->stats.frames_total);
1681 	seq_printf(s, "vblank total: %lu\n", dc->stats.vblank_total);
1682 	seq_printf(s, "underflow total: %lu\n", dc->stats.underflow_total);
1683 	seq_printf(s, "overflow total: %lu\n", dc->stats.overflow_total);
1684 
1685 	return 0;
1686 }
1687 
1688 static struct drm_info_list debugfs_files[] = {
1689 	{ "regs", tegra_dc_show_regs, 0, NULL },
1690 	{ "crc", tegra_dc_show_crc, 0, NULL },
1691 	{ "stats", tegra_dc_show_stats, 0, NULL },
1692 };
1693 
1694 static int tegra_dc_late_register(struct drm_crtc *crtc)
1695 {
1696 	unsigned int i, count = ARRAY_SIZE(debugfs_files);
1697 	struct drm_minor *minor = crtc->dev->primary;
1698 	struct dentry *root;
1699 	struct tegra_dc *dc = to_tegra_dc(crtc);
1700 
1701 #ifdef CONFIG_DEBUG_FS
1702 	root = crtc->debugfs_entry;
1703 #else
1704 	root = NULL;
1705 #endif
1706 
1707 	dc->debugfs_files = kmemdup(debugfs_files, sizeof(debugfs_files),
1708 				    GFP_KERNEL);
1709 	if (!dc->debugfs_files)
1710 		return -ENOMEM;
1711 
1712 	for (i = 0; i < count; i++)
1713 		dc->debugfs_files[i].data = dc;
1714 
1715 	drm_debugfs_create_files(dc->debugfs_files, count, root, minor);
1716 
1717 	return 0;
1718 }
1719 
1720 static void tegra_dc_early_unregister(struct drm_crtc *crtc)
1721 {
1722 	unsigned int count = ARRAY_SIZE(debugfs_files);
1723 	struct drm_minor *minor = crtc->dev->primary;
1724 	struct tegra_dc *dc = to_tegra_dc(crtc);
1725 
1726 	drm_debugfs_remove_files(dc->debugfs_files, count, minor);
1727 	kfree(dc->debugfs_files);
1728 	dc->debugfs_files = NULL;
1729 }
1730 
1731 static u32 tegra_dc_get_vblank_counter(struct drm_crtc *crtc)
1732 {
1733 	struct tegra_dc *dc = to_tegra_dc(crtc);
1734 
1735 	/* XXX vblank syncpoints don't work with nvdisplay yet */
1736 	if (dc->syncpt && !dc->soc->has_nvdisplay)
1737 		return host1x_syncpt_read(dc->syncpt);
1738 
1739 	/* fallback to software emulated VBLANK counter */
1740 	return (u32)drm_crtc_vblank_count(&dc->base);
1741 }
1742 
1743 static int tegra_dc_enable_vblank(struct drm_crtc *crtc)
1744 {
1745 	struct tegra_dc *dc = to_tegra_dc(crtc);
1746 	u32 value;
1747 
1748 	value = tegra_dc_readl(dc, DC_CMD_INT_MASK);
1749 	value |= VBLANK_INT;
1750 	tegra_dc_writel(dc, value, DC_CMD_INT_MASK);
1751 
1752 	return 0;
1753 }
1754 
1755 static void tegra_dc_disable_vblank(struct drm_crtc *crtc)
1756 {
1757 	struct tegra_dc *dc = to_tegra_dc(crtc);
1758 	u32 value;
1759 
1760 	value = tegra_dc_readl(dc, DC_CMD_INT_MASK);
1761 	value &= ~VBLANK_INT;
1762 	tegra_dc_writel(dc, value, DC_CMD_INT_MASK);
1763 }
1764 
1765 static const struct drm_crtc_funcs tegra_crtc_funcs = {
1766 	.page_flip = drm_atomic_helper_page_flip,
1767 	.set_config = drm_atomic_helper_set_config,
1768 	.destroy = tegra_dc_destroy,
1769 	.reset = tegra_crtc_reset,
1770 	.atomic_duplicate_state = tegra_crtc_atomic_duplicate_state,
1771 	.atomic_destroy_state = tegra_crtc_atomic_destroy_state,
1772 	.late_register = tegra_dc_late_register,
1773 	.early_unregister = tegra_dc_early_unregister,
1774 	.get_vblank_counter = tegra_dc_get_vblank_counter,
1775 	.enable_vblank = tegra_dc_enable_vblank,
1776 	.disable_vblank = tegra_dc_disable_vblank,
1777 };
1778 
1779 static int tegra_dc_set_timings(struct tegra_dc *dc,
1780 				struct drm_display_mode *mode)
1781 {
1782 	unsigned int h_ref_to_sync = 1;
1783 	unsigned int v_ref_to_sync = 1;
1784 	unsigned long value;
1785 
1786 	if (!dc->soc->has_nvdisplay) {
1787 		tegra_dc_writel(dc, 0x0, DC_DISP_DISP_TIMING_OPTIONS);
1788 
1789 		value = (v_ref_to_sync << 16) | h_ref_to_sync;
1790 		tegra_dc_writel(dc, value, DC_DISP_REF_TO_SYNC);
1791 	}
1792 
1793 	value = ((mode->vsync_end - mode->vsync_start) << 16) |
1794 		((mode->hsync_end - mode->hsync_start) <<  0);
1795 	tegra_dc_writel(dc, value, DC_DISP_SYNC_WIDTH);
1796 
1797 	value = ((mode->vtotal - mode->vsync_end) << 16) |
1798 		((mode->htotal - mode->hsync_end) <<  0);
1799 	tegra_dc_writel(dc, value, DC_DISP_BACK_PORCH);
1800 
1801 	value = ((mode->vsync_start - mode->vdisplay) << 16) |
1802 		((mode->hsync_start - mode->hdisplay) <<  0);
1803 	tegra_dc_writel(dc, value, DC_DISP_FRONT_PORCH);
1804 
1805 	value = (mode->vdisplay << 16) | mode->hdisplay;
1806 	tegra_dc_writel(dc, value, DC_DISP_ACTIVE);
1807 
1808 	return 0;
1809 }
1810 
1811 /**
1812  * tegra_dc_state_setup_clock - check clock settings and store them in atomic
1813  *     state
1814  * @dc: display controller
1815  * @crtc_state: CRTC atomic state
1816  * @clk: parent clock for display controller
1817  * @pclk: pixel clock
1818  * @div: shift clock divider
1819  *
1820  * Returns:
1821  * 0 on success or a negative error-code on failure.
1822  */
1823 int tegra_dc_state_setup_clock(struct tegra_dc *dc,
1824 			       struct drm_crtc_state *crtc_state,
1825 			       struct clk *clk, unsigned long pclk,
1826 			       unsigned int div)
1827 {
1828 	struct tegra_dc_state *state = to_dc_state(crtc_state);
1829 
1830 	if (!clk_has_parent(dc->clk, clk))
1831 		return -EINVAL;
1832 
1833 	state->clk = clk;
1834 	state->pclk = pclk;
1835 	state->div = div;
1836 
1837 	return 0;
1838 }
1839 
1840 static void tegra_dc_update_voltage_state(struct tegra_dc *dc,
1841 					  struct tegra_dc_state *state)
1842 {
1843 	unsigned long rate, pstate;
1844 	struct dev_pm_opp *opp;
1845 	int err;
1846 
1847 	if (!dc->has_opp_table)
1848 		return;
1849 
1850 	/* calculate actual pixel clock rate which depends on internal divider */
1851 	rate = DIV_ROUND_UP(clk_get_rate(dc->clk) * 2, state->div + 2);
1852 
1853 	/* find suitable OPP for the rate */
1854 	opp = dev_pm_opp_find_freq_ceil(dc->dev, &rate);
1855 
1856 	/*
1857 	 * Very high resolution modes may results in a clock rate that is
1858 	 * above the characterized maximum. In this case it's okay to fall
1859 	 * back to the characterized maximum.
1860 	 */
1861 	if (opp == ERR_PTR(-ERANGE))
1862 		opp = dev_pm_opp_find_freq_floor(dc->dev, &rate);
1863 
1864 	if (IS_ERR(opp)) {
1865 		dev_err(dc->dev, "failed to find OPP for %luHz: %pe\n",
1866 			rate, opp);
1867 		return;
1868 	}
1869 
1870 	pstate = dev_pm_opp_get_required_pstate(opp, 0);
1871 	dev_pm_opp_put(opp);
1872 
1873 	/*
1874 	 * The minimum core voltage depends on the pixel clock rate (which
1875 	 * depends on internal clock divider of the CRTC) and not on the
1876 	 * rate of the display controller clock. This is why we're not using
1877 	 * dev_pm_opp_set_rate() API and instead controlling the power domain
1878 	 * directly.
1879 	 */
1880 	err = dev_pm_genpd_set_performance_state(dc->dev, pstate);
1881 	if (err)
1882 		dev_err(dc->dev, "failed to set power domain state to %lu: %d\n",
1883 			pstate, err);
1884 }
1885 
1886 static void tegra_dc_set_clock_rate(struct tegra_dc *dc,
1887 				    struct tegra_dc_state *state)
1888 {
1889 	int err;
1890 
1891 	err = clk_set_parent(dc->clk, state->clk);
1892 	if (err < 0)
1893 		dev_err(dc->dev, "failed to set parent clock: %d\n", err);
1894 
1895 	/*
1896 	 * Outputs may not want to change the parent clock rate. This is only
1897 	 * relevant to Tegra20 where only a single display PLL is available.
1898 	 * Since that PLL would typically be used for HDMI, an internal LVDS
1899 	 * panel would need to be driven by some other clock such as PLL_P
1900 	 * which is shared with other peripherals. Changing the clock rate
1901 	 * should therefore be avoided.
1902 	 */
1903 	if (state->pclk > 0) {
1904 		err = clk_set_rate(state->clk, state->pclk);
1905 		if (err < 0)
1906 			dev_err(dc->dev,
1907 				"failed to set clock rate to %lu Hz\n",
1908 				state->pclk);
1909 
1910 		err = clk_set_rate(dc->clk, state->pclk);
1911 		if (err < 0)
1912 			dev_err(dc->dev, "failed to set clock %pC to %lu Hz: %d\n",
1913 				dc->clk, state->pclk, err);
1914 	}
1915 
1916 	DRM_DEBUG_KMS("rate: %lu, div: %u\n", clk_get_rate(dc->clk),
1917 		      state->div);
1918 	DRM_DEBUG_KMS("pclk: %lu\n", state->pclk);
1919 
1920 	tegra_dc_update_voltage_state(dc, state);
1921 }
1922 
1923 static void tegra_dc_stop(struct tegra_dc *dc)
1924 {
1925 	u32 value;
1926 
1927 	/* stop the display controller */
1928 	value = tegra_dc_readl(dc, DC_CMD_DISPLAY_COMMAND);
1929 	value &= ~DISP_CTRL_MODE_MASK;
1930 	tegra_dc_writel(dc, value, DC_CMD_DISPLAY_COMMAND);
1931 
1932 	tegra_dc_commit(dc);
1933 }
1934 
1935 static bool tegra_dc_idle(struct tegra_dc *dc)
1936 {
1937 	u32 value;
1938 
1939 	value = tegra_dc_readl_active(dc, DC_CMD_DISPLAY_COMMAND);
1940 
1941 	return (value & DISP_CTRL_MODE_MASK) == 0;
1942 }
1943 
1944 static int tegra_dc_wait_idle(struct tegra_dc *dc, unsigned long timeout)
1945 {
1946 	timeout = jiffies + msecs_to_jiffies(timeout);
1947 
1948 	while (time_before(jiffies, timeout)) {
1949 		if (tegra_dc_idle(dc))
1950 			return 0;
1951 
1952 		usleep_range(1000, 2000);
1953 	}
1954 
1955 	dev_dbg(dc->dev, "timeout waiting for DC to become idle\n");
1956 	return -ETIMEDOUT;
1957 }
1958 
1959 static void
1960 tegra_crtc_update_memory_bandwidth(struct drm_crtc *crtc,
1961 				   struct drm_atomic_state *state,
1962 				   bool prepare_bandwidth_transition)
1963 {
1964 	const struct tegra_plane_state *old_tegra_state, *new_tegra_state;
1965 	u32 i, new_avg_bw, old_avg_bw, new_peak_bw, old_peak_bw;
1966 	const struct drm_plane_state *old_plane_state;
1967 	const struct drm_crtc_state *old_crtc_state;
1968 	struct tegra_dc_window window, old_window;
1969 	struct tegra_dc *dc = to_tegra_dc(crtc);
1970 	struct tegra_plane *tegra;
1971 	struct drm_plane *plane;
1972 
1973 	if (dc->soc->has_nvdisplay)
1974 		return;
1975 
1976 	old_crtc_state = drm_atomic_get_old_crtc_state(state, crtc);
1977 
1978 	if (!crtc->state->active) {
1979 		if (!old_crtc_state->active)
1980 			return;
1981 
1982 		/*
1983 		 * When CRTC is disabled on DPMS, the state of attached planes
1984 		 * is kept unchanged. Hence we need to enforce removal of the
1985 		 * bandwidths from the ICC paths.
1986 		 */
1987 		drm_atomic_crtc_for_each_plane(plane, crtc) {
1988 			tegra = to_tegra_plane(plane);
1989 
1990 			icc_set_bw(tegra->icc_mem, 0, 0);
1991 			icc_set_bw(tegra->icc_mem_vfilter, 0, 0);
1992 		}
1993 
1994 		return;
1995 	}
1996 
1997 	for_each_old_plane_in_state(old_crtc_state->state, plane,
1998 				    old_plane_state, i) {
1999 		old_tegra_state = to_const_tegra_plane_state(old_plane_state);
2000 		new_tegra_state = to_const_tegra_plane_state(plane->state);
2001 		tegra = to_tegra_plane(plane);
2002 
2003 		/*
2004 		 * We're iterating over the global atomic state and it contains
2005 		 * planes from another CRTC, hence we need to filter out the
2006 		 * planes unrelated to this CRTC.
2007 		 */
2008 		if (tegra->dc != dc)
2009 			continue;
2010 
2011 		new_avg_bw = new_tegra_state->avg_memory_bandwidth;
2012 		old_avg_bw = old_tegra_state->avg_memory_bandwidth;
2013 
2014 		new_peak_bw = new_tegra_state->total_peak_memory_bandwidth;
2015 		old_peak_bw = old_tegra_state->total_peak_memory_bandwidth;
2016 
2017 		/*
2018 		 * See the comment related to !crtc->state->active above,
2019 		 * which explains why bandwidths need to be updated when
2020 		 * CRTC is turning ON.
2021 		 */
2022 		if (new_avg_bw == old_avg_bw && new_peak_bw == old_peak_bw &&
2023 		    old_crtc_state->active)
2024 			continue;
2025 
2026 		window.src.h = drm_rect_height(&plane->state->src) >> 16;
2027 		window.dst.h = drm_rect_height(&plane->state->dst);
2028 
2029 		old_window.src.h = drm_rect_height(&old_plane_state->src) >> 16;
2030 		old_window.dst.h = drm_rect_height(&old_plane_state->dst);
2031 
2032 		/*
2033 		 * During the preparation phase (atomic_begin), the memory
2034 		 * freq should go high before the DC changes are committed
2035 		 * if bandwidth requirement goes up, otherwise memory freq
2036 		 * should to stay high if BW requirement goes down.  The
2037 		 * opposite applies to the completion phase (post_commit).
2038 		 */
2039 		if (prepare_bandwidth_transition) {
2040 			new_avg_bw = max(old_avg_bw, new_avg_bw);
2041 			new_peak_bw = max(old_peak_bw, new_peak_bw);
2042 
2043 			if (tegra_plane_use_vertical_filtering(tegra, &old_window))
2044 				window = old_window;
2045 		}
2046 
2047 		icc_set_bw(tegra->icc_mem, new_avg_bw, new_peak_bw);
2048 
2049 		if (tegra_plane_use_vertical_filtering(tegra, &window))
2050 			icc_set_bw(tegra->icc_mem_vfilter, new_avg_bw, new_peak_bw);
2051 		else
2052 			icc_set_bw(tegra->icc_mem_vfilter, 0, 0);
2053 	}
2054 }
2055 
2056 static void tegra_crtc_atomic_disable(struct drm_crtc *crtc,
2057 				      struct drm_atomic_state *state)
2058 {
2059 	struct tegra_dc *dc = to_tegra_dc(crtc);
2060 	u32 value;
2061 	int err;
2062 
2063 	if (!tegra_dc_idle(dc)) {
2064 		tegra_dc_stop(dc);
2065 
2066 		/*
2067 		 * Ignore the return value, there isn't anything useful to do
2068 		 * in case this fails.
2069 		 */
2070 		tegra_dc_wait_idle(dc, 100);
2071 	}
2072 
2073 	/*
2074 	 * This should really be part of the RGB encoder driver, but clearing
2075 	 * these bits has the side-effect of stopping the display controller.
2076 	 * When that happens no VBLANK interrupts will be raised. At the same
2077 	 * time the encoder is disabled before the display controller, so the
2078 	 * above code is always going to timeout waiting for the controller
2079 	 * to go idle.
2080 	 *
2081 	 * Given the close coupling between the RGB encoder and the display
2082 	 * controller doing it here is still kind of okay. None of the other
2083 	 * encoder drivers require these bits to be cleared.
2084 	 *
2085 	 * XXX: Perhaps given that the display controller is switched off at
2086 	 * this point anyway maybe clearing these bits isn't even useful for
2087 	 * the RGB encoder?
2088 	 */
2089 	if (dc->rgb) {
2090 		value = tegra_dc_readl(dc, DC_CMD_DISPLAY_POWER_CONTROL);
2091 		value &= ~(PW0_ENABLE | PW1_ENABLE | PW2_ENABLE | PW3_ENABLE |
2092 			   PW4_ENABLE | PM0_ENABLE | PM1_ENABLE);
2093 		tegra_dc_writel(dc, value, DC_CMD_DISPLAY_POWER_CONTROL);
2094 	}
2095 
2096 	tegra_dc_stats_reset(&dc->stats);
2097 	drm_crtc_vblank_off(crtc);
2098 
2099 	spin_lock_irq(&crtc->dev->event_lock);
2100 
2101 	if (crtc->state->event) {
2102 		drm_crtc_send_vblank_event(crtc, crtc->state->event);
2103 		crtc->state->event = NULL;
2104 	}
2105 
2106 	spin_unlock_irq(&crtc->dev->event_lock);
2107 
2108 	err = host1x_client_suspend(&dc->client);
2109 	if (err < 0)
2110 		dev_err(dc->dev, "failed to suspend: %d\n", err);
2111 
2112 	if (dc->has_opp_table) {
2113 		err = dev_pm_genpd_set_performance_state(dc->dev, 0);
2114 		if (err)
2115 			dev_err(dc->dev,
2116 				"failed to clear power domain state: %d\n", err);
2117 	}
2118 }
2119 
2120 static void tegra_crtc_atomic_enable(struct drm_crtc *crtc,
2121 				     struct drm_atomic_state *state)
2122 {
2123 	struct drm_display_mode *mode = &crtc->state->adjusted_mode;
2124 	struct tegra_dc_state *crtc_state = to_dc_state(crtc->state);
2125 	struct tegra_dc *dc = to_tegra_dc(crtc);
2126 	u32 value;
2127 	int err;
2128 
2129 	/* apply PLL changes */
2130 	tegra_dc_set_clock_rate(dc, crtc_state);
2131 
2132 	err = host1x_client_resume(&dc->client);
2133 	if (err < 0) {
2134 		dev_err(dc->dev, "failed to resume: %d\n", err);
2135 		return;
2136 	}
2137 
2138 	/* initialize display controller */
2139 	if (dc->syncpt) {
2140 		u32 syncpt = host1x_syncpt_id(dc->syncpt), enable;
2141 
2142 		if (dc->soc->has_nvdisplay)
2143 			enable = 1 << 31;
2144 		else
2145 			enable = 1 << 8;
2146 
2147 		value = SYNCPT_CNTRL_NO_STALL;
2148 		tegra_dc_writel(dc, value, DC_CMD_GENERAL_INCR_SYNCPT_CNTRL);
2149 
2150 		value = enable | syncpt;
2151 		tegra_dc_writel(dc, value, DC_CMD_CONT_SYNCPT_VSYNC);
2152 	}
2153 
2154 	if (dc->soc->has_nvdisplay) {
2155 		value = DSC_TO_UF_INT | DSC_BBUF_UF_INT | DSC_RBUF_UF_INT |
2156 			DSC_OBUF_UF_INT;
2157 		tegra_dc_writel(dc, value, DC_CMD_INT_TYPE);
2158 
2159 		value = DSC_TO_UF_INT | DSC_BBUF_UF_INT | DSC_RBUF_UF_INT |
2160 			DSC_OBUF_UF_INT | SD3_BUCKET_WALK_DONE_INT |
2161 			HEAD_UF_INT | MSF_INT | REG_TMOUT_INT |
2162 			REGION_CRC_INT | V_PULSE2_INT | V_PULSE3_INT |
2163 			VBLANK_INT | FRAME_END_INT;
2164 		tegra_dc_writel(dc, value, DC_CMD_INT_POLARITY);
2165 
2166 		value = SD3_BUCKET_WALK_DONE_INT | HEAD_UF_INT | VBLANK_INT |
2167 			FRAME_END_INT;
2168 		tegra_dc_writel(dc, value, DC_CMD_INT_ENABLE);
2169 
2170 		value = HEAD_UF_INT | REG_TMOUT_INT | FRAME_END_INT;
2171 		tegra_dc_writel(dc, value, DC_CMD_INT_MASK);
2172 
2173 		tegra_dc_writel(dc, READ_MUX, DC_CMD_STATE_ACCESS);
2174 	} else {
2175 		value = WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT |
2176 			WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT;
2177 		tegra_dc_writel(dc, value, DC_CMD_INT_TYPE);
2178 
2179 		value = WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT |
2180 			WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT;
2181 		tegra_dc_writel(dc, value, DC_CMD_INT_POLARITY);
2182 
2183 		/* initialize timer */
2184 		value = CURSOR_THRESHOLD(0) | WINDOW_A_THRESHOLD(0x20) |
2185 			WINDOW_B_THRESHOLD(0x20) | WINDOW_C_THRESHOLD(0x20);
2186 		tegra_dc_writel(dc, value, DC_DISP_DISP_MEM_HIGH_PRIORITY);
2187 
2188 		value = CURSOR_THRESHOLD(0) | WINDOW_A_THRESHOLD(1) |
2189 			WINDOW_B_THRESHOLD(1) | WINDOW_C_THRESHOLD(1);
2190 		tegra_dc_writel(dc, value, DC_DISP_DISP_MEM_HIGH_PRIORITY_TIMER);
2191 
2192 		value = VBLANK_INT | WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT |
2193 			WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT;
2194 		tegra_dc_writel(dc, value, DC_CMD_INT_ENABLE);
2195 
2196 		value = WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT |
2197 			WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT;
2198 		tegra_dc_writel(dc, value, DC_CMD_INT_MASK);
2199 	}
2200 
2201 	if (dc->soc->supports_background_color)
2202 		tegra_dc_writel(dc, 0, DC_DISP_BLEND_BACKGROUND_COLOR);
2203 	else
2204 		tegra_dc_writel(dc, 0, DC_DISP_BORDER_COLOR);
2205 
2206 	/* apply pixel clock changes */
2207 	if (!dc->soc->has_nvdisplay) {
2208 		value = SHIFT_CLK_DIVIDER(crtc_state->div) | PIXEL_CLK_DIVIDER_PCD1;
2209 		tegra_dc_writel(dc, value, DC_DISP_DISP_CLOCK_CONTROL);
2210 	}
2211 
2212 	/* program display mode */
2213 	tegra_dc_set_timings(dc, mode);
2214 
2215 	/* interlacing isn't supported yet, so disable it */
2216 	if (dc->soc->supports_interlacing) {
2217 		value = tegra_dc_readl(dc, DC_DISP_INTERLACE_CONTROL);
2218 		value &= ~INTERLACE_ENABLE;
2219 		tegra_dc_writel(dc, value, DC_DISP_INTERLACE_CONTROL);
2220 	}
2221 
2222 	value = tegra_dc_readl(dc, DC_CMD_DISPLAY_COMMAND);
2223 	value &= ~DISP_CTRL_MODE_MASK;
2224 	value |= DISP_CTRL_MODE_C_DISPLAY;
2225 	tegra_dc_writel(dc, value, DC_CMD_DISPLAY_COMMAND);
2226 
2227 	if (!dc->soc->has_nvdisplay) {
2228 		value = tegra_dc_readl(dc, DC_CMD_DISPLAY_POWER_CONTROL);
2229 		value |= PW0_ENABLE | PW1_ENABLE | PW2_ENABLE | PW3_ENABLE |
2230 			 PW4_ENABLE | PM0_ENABLE | PM1_ENABLE;
2231 		tegra_dc_writel(dc, value, DC_CMD_DISPLAY_POWER_CONTROL);
2232 	}
2233 
2234 	/* enable underflow reporting and display red for missing pixels */
2235 	if (dc->soc->has_nvdisplay) {
2236 		value = UNDERFLOW_MODE_RED | UNDERFLOW_REPORT_ENABLE;
2237 		tegra_dc_writel(dc, value, DC_COM_RG_UNDERFLOW);
2238 	}
2239 
2240 	if (dc->rgb) {
2241 		/* XXX: parameterize? */
2242 		value = SC0_H_QUALIFIER_NONE | SC1_H_QUALIFIER_NONE;
2243 		tegra_dc_writel(dc, value, DC_DISP_SHIFT_CLOCK_OPTIONS);
2244 	}
2245 
2246 	tegra_dc_commit(dc);
2247 
2248 	drm_crtc_vblank_on(crtc);
2249 }
2250 
2251 static void tegra_crtc_atomic_begin(struct drm_crtc *crtc,
2252 				    struct drm_atomic_state *state)
2253 {
2254 	unsigned long flags;
2255 
2256 	tegra_crtc_update_memory_bandwidth(crtc, state, true);
2257 
2258 	if (crtc->state->event) {
2259 		spin_lock_irqsave(&crtc->dev->event_lock, flags);
2260 
2261 		if (drm_crtc_vblank_get(crtc) != 0)
2262 			drm_crtc_send_vblank_event(crtc, crtc->state->event);
2263 		else
2264 			drm_crtc_arm_vblank_event(crtc, crtc->state->event);
2265 
2266 		spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
2267 
2268 		crtc->state->event = NULL;
2269 	}
2270 }
2271 
2272 static void tegra_crtc_atomic_flush(struct drm_crtc *crtc,
2273 				    struct drm_atomic_state *state)
2274 {
2275 	struct drm_crtc_state *crtc_state = drm_atomic_get_new_crtc_state(state,
2276 									  crtc);
2277 	struct tegra_dc_state *dc_state = to_dc_state(crtc_state);
2278 	struct tegra_dc *dc = to_tegra_dc(crtc);
2279 	u32 value;
2280 
2281 	value = dc_state->planes << 8 | GENERAL_UPDATE;
2282 	tegra_dc_writel(dc, value, DC_CMD_STATE_CONTROL);
2283 	value = tegra_dc_readl(dc, DC_CMD_STATE_CONTROL);
2284 
2285 	value = dc_state->planes | GENERAL_ACT_REQ;
2286 	tegra_dc_writel(dc, value, DC_CMD_STATE_CONTROL);
2287 	value = tegra_dc_readl(dc, DC_CMD_STATE_CONTROL);
2288 }
2289 
2290 static bool tegra_plane_is_cursor(const struct drm_plane_state *state)
2291 {
2292 	const struct tegra_dc_soc_info *soc = to_tegra_dc(state->crtc)->soc;
2293 	const struct drm_format_info *fmt = state->fb->format;
2294 	unsigned int src_w = drm_rect_width(&state->src) >> 16;
2295 	unsigned int dst_w = drm_rect_width(&state->dst);
2296 
2297 	if (state->plane->type != DRM_PLANE_TYPE_CURSOR)
2298 		return false;
2299 
2300 	if (soc->supports_cursor)
2301 		return true;
2302 
2303 	if (src_w != dst_w || fmt->num_planes != 1 || src_w * fmt->cpp[0] > 256)
2304 		return false;
2305 
2306 	return true;
2307 }
2308 
2309 static unsigned long
2310 tegra_plane_overlap_mask(struct drm_crtc_state *state,
2311 			 const struct drm_plane_state *plane_state)
2312 {
2313 	const struct drm_plane_state *other_state;
2314 	const struct tegra_plane *tegra;
2315 	unsigned long overlap_mask = 0;
2316 	struct drm_plane *plane;
2317 	struct drm_rect rect;
2318 
2319 	if (!plane_state->visible || !plane_state->fb)
2320 		return 0;
2321 
2322 	/*
2323 	 * Data-prefetch FIFO will easily help to overcome temporal memory
2324 	 * pressure if other plane overlaps with the cursor plane.
2325 	 */
2326 	if (tegra_plane_is_cursor(plane_state))
2327 		return 0;
2328 
2329 	drm_atomic_crtc_state_for_each_plane_state(plane, other_state, state) {
2330 		rect = plane_state->dst;
2331 
2332 		tegra = to_tegra_plane(other_state->plane);
2333 
2334 		if (!other_state->visible || !other_state->fb)
2335 			continue;
2336 
2337 		/*
2338 		 * Ignore cursor plane overlaps because it's not practical to
2339 		 * assume that it contributes to the bandwidth in overlapping
2340 		 * area if window width is small.
2341 		 */
2342 		if (tegra_plane_is_cursor(other_state))
2343 			continue;
2344 
2345 		if (drm_rect_intersect(&rect, &other_state->dst))
2346 			overlap_mask |= BIT(tegra->index);
2347 	}
2348 
2349 	return overlap_mask;
2350 }
2351 
2352 static int tegra_crtc_calculate_memory_bandwidth(struct drm_crtc *crtc,
2353 						 struct drm_atomic_state *state)
2354 {
2355 	ulong overlap_mask[TEGRA_DC_LEGACY_PLANES_NUM] = {}, mask;
2356 	u32 plane_peak_bw[TEGRA_DC_LEGACY_PLANES_NUM] = {};
2357 	bool all_planes_overlap_simultaneously = true;
2358 	const struct tegra_plane_state *tegra_state;
2359 	const struct drm_plane_state *plane_state;
2360 	struct tegra_dc *dc = to_tegra_dc(crtc);
2361 	const struct drm_crtc_state *old_state;
2362 	struct drm_crtc_state *new_state;
2363 	struct tegra_plane *tegra;
2364 	struct drm_plane *plane;
2365 
2366 	/*
2367 	 * The nv-display uses shared planes.  The algorithm below assumes
2368 	 * maximum 3 planes per-CRTC, this assumption isn't applicable to
2369 	 * the nv-display.  Note that T124 support has additional windows,
2370 	 * but currently they aren't supported by the driver.
2371 	 */
2372 	if (dc->soc->has_nvdisplay)
2373 		return 0;
2374 
2375 	new_state = drm_atomic_get_new_crtc_state(state, crtc);
2376 	old_state = drm_atomic_get_old_crtc_state(state, crtc);
2377 
2378 	/*
2379 	 * For overlapping planes pixel's data is fetched for each plane at
2380 	 * the same time, hence bandwidths are accumulated in this case.
2381 	 * This needs to be taken into account for calculating total bandwidth
2382 	 * consumed by all planes.
2383 	 *
2384 	 * Here we get the overlapping state of each plane, which is a
2385 	 * bitmask of plane indices telling with what planes there is an
2386 	 * overlap. Note that bitmask[plane] includes BIT(plane) in order
2387 	 * to make further code nicer and simpler.
2388 	 */
2389 	drm_atomic_crtc_state_for_each_plane_state(plane, plane_state, new_state) {
2390 		tegra_state = to_const_tegra_plane_state(plane_state);
2391 		tegra = to_tegra_plane(plane);
2392 
2393 		if (WARN_ON_ONCE(tegra->index >= TEGRA_DC_LEGACY_PLANES_NUM))
2394 			return -EINVAL;
2395 
2396 		plane_peak_bw[tegra->index] = tegra_state->peak_memory_bandwidth;
2397 		mask = tegra_plane_overlap_mask(new_state, plane_state);
2398 		overlap_mask[tegra->index] = mask;
2399 
2400 		if (hweight_long(mask) != 3)
2401 			all_planes_overlap_simultaneously = false;
2402 	}
2403 
2404 	/*
2405 	 * Then we calculate maximum bandwidth of each plane state.
2406 	 * The bandwidth includes the plane BW + BW of the "simultaneously"
2407 	 * overlapping planes, where "simultaneously" means areas where DC
2408 	 * fetches from the planes simultaneously during of scan-out process.
2409 	 *
2410 	 * For example, if plane A overlaps with planes B and C, but B and C
2411 	 * don't overlap, then the peak bandwidth will be either in area where
2412 	 * A-and-B or A-and-C planes overlap.
2413 	 *
2414 	 * The plane_peak_bw[] contains peak memory bandwidth values of
2415 	 * each plane, this information is needed by interconnect provider
2416 	 * in order to set up latency allowance based on the peak BW, see
2417 	 * tegra_crtc_update_memory_bandwidth().
2418 	 */
2419 	drm_atomic_crtc_state_for_each_plane_state(plane, plane_state, new_state) {
2420 		u32 i, old_peak_bw, new_peak_bw, overlap_bw = 0;
2421 
2422 		/*
2423 		 * Note that plane's atomic check doesn't touch the
2424 		 * total_peak_memory_bandwidth of enabled plane, hence the
2425 		 * current state contains the old bandwidth state from the
2426 		 * previous CRTC commit.
2427 		 */
2428 		tegra_state = to_const_tegra_plane_state(plane_state);
2429 		tegra = to_tegra_plane(plane);
2430 
2431 		for_each_set_bit(i, &overlap_mask[tegra->index], 3) {
2432 			if (i == tegra->index)
2433 				continue;
2434 
2435 			if (all_planes_overlap_simultaneously)
2436 				overlap_bw += plane_peak_bw[i];
2437 			else
2438 				overlap_bw = max(overlap_bw, plane_peak_bw[i]);
2439 		}
2440 
2441 		new_peak_bw = plane_peak_bw[tegra->index] + overlap_bw;
2442 		old_peak_bw = tegra_state->total_peak_memory_bandwidth;
2443 
2444 		/*
2445 		 * If plane's peak bandwidth changed (for example plane isn't
2446 		 * overlapped anymore) and plane isn't in the atomic state,
2447 		 * then add plane to the state in order to have the bandwidth
2448 		 * updated.
2449 		 */
2450 		if (old_peak_bw != new_peak_bw) {
2451 			struct tegra_plane_state *new_tegra_state;
2452 			struct drm_plane_state *new_plane_state;
2453 
2454 			new_plane_state = drm_atomic_get_plane_state(state, plane);
2455 			if (IS_ERR(new_plane_state))
2456 				return PTR_ERR(new_plane_state);
2457 
2458 			new_tegra_state = to_tegra_plane_state(new_plane_state);
2459 			new_tegra_state->total_peak_memory_bandwidth = new_peak_bw;
2460 		}
2461 	}
2462 
2463 	return 0;
2464 }
2465 
2466 static int tegra_crtc_atomic_check(struct drm_crtc *crtc,
2467 				   struct drm_atomic_state *state)
2468 {
2469 	int err;
2470 
2471 	err = tegra_crtc_calculate_memory_bandwidth(crtc, state);
2472 	if (err)
2473 		return err;
2474 
2475 	return 0;
2476 }
2477 
2478 void tegra_crtc_atomic_post_commit(struct drm_crtc *crtc,
2479 				   struct drm_atomic_state *state)
2480 {
2481 	/*
2482 	 * Display bandwidth is allowed to go down only once hardware state
2483 	 * is known to be armed, i.e. state was committed and VBLANK event
2484 	 * received.
2485 	 */
2486 	tegra_crtc_update_memory_bandwidth(crtc, state, false);
2487 }
2488 
2489 static const struct drm_crtc_helper_funcs tegra_crtc_helper_funcs = {
2490 	.atomic_check = tegra_crtc_atomic_check,
2491 	.atomic_begin = tegra_crtc_atomic_begin,
2492 	.atomic_flush = tegra_crtc_atomic_flush,
2493 	.atomic_enable = tegra_crtc_atomic_enable,
2494 	.atomic_disable = tegra_crtc_atomic_disable,
2495 };
2496 
2497 static irqreturn_t tegra_dc_irq(int irq, void *data)
2498 {
2499 	struct tegra_dc *dc = data;
2500 	unsigned long status;
2501 
2502 	status = tegra_dc_readl(dc, DC_CMD_INT_STATUS);
2503 	tegra_dc_writel(dc, status, DC_CMD_INT_STATUS);
2504 
2505 	if (status & FRAME_END_INT) {
2506 		/*
2507 		dev_dbg(dc->dev, "%s(): frame end\n", __func__);
2508 		*/
2509 		dc->stats.frames_total++;
2510 		dc->stats.frames++;
2511 	}
2512 
2513 	if (status & VBLANK_INT) {
2514 		/*
2515 		dev_dbg(dc->dev, "%s(): vertical blank\n", __func__);
2516 		*/
2517 		drm_crtc_handle_vblank(&dc->base);
2518 		dc->stats.vblank_total++;
2519 		dc->stats.vblank++;
2520 	}
2521 
2522 	if (status & (WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT)) {
2523 		/*
2524 		dev_dbg(dc->dev, "%s(): underflow\n", __func__);
2525 		*/
2526 		dc->stats.underflow_total++;
2527 		dc->stats.underflow++;
2528 	}
2529 
2530 	if (status & (WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT)) {
2531 		/*
2532 		dev_dbg(dc->dev, "%s(): overflow\n", __func__);
2533 		*/
2534 		dc->stats.overflow_total++;
2535 		dc->stats.overflow++;
2536 	}
2537 
2538 	if (status & HEAD_UF_INT) {
2539 		dev_dbg_ratelimited(dc->dev, "%s(): head underflow\n", __func__);
2540 		dc->stats.underflow_total++;
2541 		dc->stats.underflow++;
2542 	}
2543 
2544 	return IRQ_HANDLED;
2545 }
2546 
2547 static bool tegra_dc_has_window_groups(struct tegra_dc *dc)
2548 {
2549 	unsigned int i;
2550 
2551 	if (!dc->soc->wgrps)
2552 		return true;
2553 
2554 	for (i = 0; i < dc->soc->num_wgrps; i++) {
2555 		const struct tegra_windowgroup_soc *wgrp = &dc->soc->wgrps[i];
2556 
2557 		if (wgrp->dc == dc->pipe && wgrp->num_windows > 0)
2558 			return true;
2559 	}
2560 
2561 	return false;
2562 }
2563 
2564 static int tegra_dc_early_init(struct host1x_client *client)
2565 {
2566 	struct drm_device *drm = dev_get_drvdata(client->host);
2567 	struct tegra_drm *tegra = drm->dev_private;
2568 
2569 	tegra->num_crtcs++;
2570 
2571 	return 0;
2572 }
2573 
2574 static int tegra_dc_init(struct host1x_client *client)
2575 {
2576 	struct drm_device *drm = dev_get_drvdata(client->host);
2577 	unsigned long flags = HOST1X_SYNCPT_CLIENT_MANAGED;
2578 	struct tegra_dc *dc = host1x_client_to_dc(client);
2579 	struct tegra_drm *tegra = drm->dev_private;
2580 	struct drm_plane *primary = NULL;
2581 	struct drm_plane *cursor = NULL;
2582 	int err;
2583 
2584 	/*
2585 	 * DC has been reset by now, so VBLANK syncpoint can be released
2586 	 * for general use.
2587 	 */
2588 	host1x_syncpt_release_vblank_reservation(client, 26 + dc->pipe);
2589 
2590 	/*
2591 	 * XXX do not register DCs with no window groups because we cannot
2592 	 * assign a primary plane to them, which in turn will cause KMS to
2593 	 * crash.
2594 	 */
2595 	if (!tegra_dc_has_window_groups(dc))
2596 		return 0;
2597 
2598 	/*
2599 	 * Set the display hub as the host1x client parent for the display
2600 	 * controller. This is needed for the runtime reference counting that
2601 	 * ensures the display hub is always powered when any of the display
2602 	 * controllers are.
2603 	 */
2604 	if (dc->soc->has_nvdisplay)
2605 		client->parent = &tegra->hub->client;
2606 
2607 	dc->syncpt = host1x_syncpt_request(client, flags);
2608 	if (!dc->syncpt)
2609 		dev_warn(dc->dev, "failed to allocate syncpoint\n");
2610 
2611 	err = host1x_client_iommu_attach(client);
2612 	if (err < 0 && err != -ENODEV) {
2613 		dev_err(client->dev, "failed to attach to domain: %d\n", err);
2614 		return err;
2615 	}
2616 
2617 	if (dc->soc->wgrps)
2618 		primary = tegra_dc_add_shared_planes(drm, dc);
2619 	else
2620 		primary = tegra_dc_add_planes(drm, dc);
2621 
2622 	if (IS_ERR(primary)) {
2623 		err = PTR_ERR(primary);
2624 		goto cleanup;
2625 	}
2626 
2627 	if (dc->soc->supports_cursor) {
2628 		cursor = tegra_dc_cursor_plane_create(drm, dc);
2629 		if (IS_ERR(cursor)) {
2630 			err = PTR_ERR(cursor);
2631 			goto cleanup;
2632 		}
2633 	} else {
2634 		/* dedicate one overlay to mouse cursor */
2635 		cursor = tegra_dc_overlay_plane_create(drm, dc, 2, true);
2636 		if (IS_ERR(cursor)) {
2637 			err = PTR_ERR(cursor);
2638 			goto cleanup;
2639 		}
2640 	}
2641 
2642 	err = drm_crtc_init_with_planes(drm, &dc->base, primary, cursor,
2643 					&tegra_crtc_funcs, NULL);
2644 	if (err < 0)
2645 		goto cleanup;
2646 
2647 	drm_crtc_helper_add(&dc->base, &tegra_crtc_helper_funcs);
2648 
2649 	/*
2650 	 * Keep track of the minimum pitch alignment across all display
2651 	 * controllers.
2652 	 */
2653 	if (dc->soc->pitch_align > tegra->pitch_align)
2654 		tegra->pitch_align = dc->soc->pitch_align;
2655 
2656 	/* track maximum resolution */
2657 	if (dc->soc->has_nvdisplay)
2658 		drm->mode_config.max_width = drm->mode_config.max_height = 16384;
2659 	else
2660 		drm->mode_config.max_width = drm->mode_config.max_height = 4096;
2661 
2662 	err = tegra_dc_rgb_init(drm, dc);
2663 	if (err < 0 && err != -ENODEV) {
2664 		dev_err(dc->dev, "failed to initialize RGB output: %d\n", err);
2665 		goto cleanup;
2666 	}
2667 
2668 	err = devm_request_irq(dc->dev, dc->irq, tegra_dc_irq, 0,
2669 			       dev_name(dc->dev), dc);
2670 	if (err < 0) {
2671 		dev_err(dc->dev, "failed to request IRQ#%u: %d\n", dc->irq,
2672 			err);
2673 		goto cleanup;
2674 	}
2675 
2676 	/*
2677 	 * Inherit the DMA parameters (such as maximum segment size) from the
2678 	 * parent host1x device.
2679 	 */
2680 	client->dev->dma_parms = client->host->dma_parms;
2681 
2682 	return 0;
2683 
2684 cleanup:
2685 	if (!IS_ERR_OR_NULL(cursor))
2686 		drm_plane_cleanup(cursor);
2687 
2688 	if (!IS_ERR(primary))
2689 		drm_plane_cleanup(primary);
2690 
2691 	host1x_client_iommu_detach(client);
2692 	host1x_syncpt_put(dc->syncpt);
2693 
2694 	return err;
2695 }
2696 
2697 static int tegra_dc_exit(struct host1x_client *client)
2698 {
2699 	struct tegra_dc *dc = host1x_client_to_dc(client);
2700 	int err;
2701 
2702 	if (!tegra_dc_has_window_groups(dc))
2703 		return 0;
2704 
2705 	/* avoid a dangling pointer just in case this disappears */
2706 	client->dev->dma_parms = NULL;
2707 
2708 	devm_free_irq(dc->dev, dc->irq, dc);
2709 
2710 	err = tegra_dc_rgb_exit(dc);
2711 	if (err) {
2712 		dev_err(dc->dev, "failed to shutdown RGB output: %d\n", err);
2713 		return err;
2714 	}
2715 
2716 	host1x_client_iommu_detach(client);
2717 	host1x_syncpt_put(dc->syncpt);
2718 
2719 	return 0;
2720 }
2721 
2722 static int tegra_dc_late_exit(struct host1x_client *client)
2723 {
2724 	struct drm_device *drm = dev_get_drvdata(client->host);
2725 	struct tegra_drm *tegra = drm->dev_private;
2726 
2727 	tegra->num_crtcs--;
2728 
2729 	return 0;
2730 }
2731 
2732 static int tegra_dc_runtime_suspend(struct host1x_client *client)
2733 {
2734 	struct tegra_dc *dc = host1x_client_to_dc(client);
2735 	struct device *dev = client->dev;
2736 	int err;
2737 
2738 	err = reset_control_assert(dc->rst);
2739 	if (err < 0) {
2740 		dev_err(dev, "failed to assert reset: %d\n", err);
2741 		return err;
2742 	}
2743 
2744 	if (dc->soc->has_powergate)
2745 		tegra_powergate_power_off(dc->powergate);
2746 
2747 	clk_disable_unprepare(dc->clk);
2748 	pm_runtime_put_sync(dev);
2749 
2750 	return 0;
2751 }
2752 
2753 static int tegra_dc_runtime_resume(struct host1x_client *client)
2754 {
2755 	struct tegra_dc *dc = host1x_client_to_dc(client);
2756 	struct device *dev = client->dev;
2757 	int err;
2758 
2759 	err = pm_runtime_resume_and_get(dev);
2760 	if (err < 0) {
2761 		dev_err(dev, "failed to get runtime PM: %d\n", err);
2762 		return err;
2763 	}
2764 
2765 	if (dc->soc->has_powergate) {
2766 		err = tegra_powergate_sequence_power_up(dc->powergate, dc->clk,
2767 							dc->rst);
2768 		if (err < 0) {
2769 			dev_err(dev, "failed to power partition: %d\n", err);
2770 			goto put_rpm;
2771 		}
2772 	} else {
2773 		err = clk_prepare_enable(dc->clk);
2774 		if (err < 0) {
2775 			dev_err(dev, "failed to enable clock: %d\n", err);
2776 			goto put_rpm;
2777 		}
2778 
2779 		err = reset_control_deassert(dc->rst);
2780 		if (err < 0) {
2781 			dev_err(dev, "failed to deassert reset: %d\n", err);
2782 			goto disable_clk;
2783 		}
2784 	}
2785 
2786 	return 0;
2787 
2788 disable_clk:
2789 	clk_disable_unprepare(dc->clk);
2790 put_rpm:
2791 	pm_runtime_put_sync(dev);
2792 	return err;
2793 }
2794 
2795 static const struct host1x_client_ops dc_client_ops = {
2796 	.early_init = tegra_dc_early_init,
2797 	.init = tegra_dc_init,
2798 	.exit = tegra_dc_exit,
2799 	.late_exit = tegra_dc_late_exit,
2800 	.suspend = tegra_dc_runtime_suspend,
2801 	.resume = tegra_dc_runtime_resume,
2802 };
2803 
2804 static const struct tegra_dc_soc_info tegra20_dc_soc_info = {
2805 	.supports_background_color = false,
2806 	.supports_interlacing = false,
2807 	.supports_cursor = false,
2808 	.supports_block_linear = false,
2809 	.supports_sector_layout = false,
2810 	.has_legacy_blending = true,
2811 	.pitch_align = 8,
2812 	.has_powergate = false,
2813 	.coupled_pm = true,
2814 	.has_nvdisplay = false,
2815 	.num_primary_formats = ARRAY_SIZE(tegra20_primary_formats),
2816 	.primary_formats = tegra20_primary_formats,
2817 	.num_overlay_formats = ARRAY_SIZE(tegra20_overlay_formats),
2818 	.overlay_formats = tegra20_overlay_formats,
2819 	.modifiers = tegra20_modifiers,
2820 	.has_win_a_without_filters = true,
2821 	.has_win_b_vfilter_mem_client = true,
2822 	.has_win_c_without_vert_filter = true,
2823 	.plane_tiled_memory_bandwidth_x2 = false,
2824 	.has_pll_d2_out0 = false,
2825 };
2826 
2827 static const struct tegra_dc_soc_info tegra30_dc_soc_info = {
2828 	.supports_background_color = false,
2829 	.supports_interlacing = false,
2830 	.supports_cursor = false,
2831 	.supports_block_linear = false,
2832 	.supports_sector_layout = false,
2833 	.has_legacy_blending = true,
2834 	.pitch_align = 8,
2835 	.has_powergate = false,
2836 	.coupled_pm = false,
2837 	.has_nvdisplay = false,
2838 	.num_primary_formats = ARRAY_SIZE(tegra20_primary_formats),
2839 	.primary_formats = tegra20_primary_formats,
2840 	.num_overlay_formats = ARRAY_SIZE(tegra20_overlay_formats),
2841 	.overlay_formats = tegra20_overlay_formats,
2842 	.modifiers = tegra20_modifiers,
2843 	.has_win_a_without_filters = false,
2844 	.has_win_b_vfilter_mem_client = true,
2845 	.has_win_c_without_vert_filter = false,
2846 	.plane_tiled_memory_bandwidth_x2 = true,
2847 	.has_pll_d2_out0 = true,
2848 };
2849 
2850 static const struct tegra_dc_soc_info tegra114_dc_soc_info = {
2851 	.supports_background_color = false,
2852 	.supports_interlacing = false,
2853 	.supports_cursor = false,
2854 	.supports_block_linear = false,
2855 	.supports_sector_layout = false,
2856 	.has_legacy_blending = true,
2857 	.pitch_align = 64,
2858 	.has_powergate = true,
2859 	.coupled_pm = false,
2860 	.has_nvdisplay = false,
2861 	.num_primary_formats = ARRAY_SIZE(tegra114_primary_formats),
2862 	.primary_formats = tegra114_primary_formats,
2863 	.num_overlay_formats = ARRAY_SIZE(tegra114_overlay_formats),
2864 	.overlay_formats = tegra114_overlay_formats,
2865 	.modifiers = tegra20_modifiers,
2866 	.has_win_a_without_filters = false,
2867 	.has_win_b_vfilter_mem_client = false,
2868 	.has_win_c_without_vert_filter = false,
2869 	.plane_tiled_memory_bandwidth_x2 = true,
2870 	.has_pll_d2_out0 = true,
2871 };
2872 
2873 static const struct tegra_dc_soc_info tegra124_dc_soc_info = {
2874 	.supports_background_color = true,
2875 	.supports_interlacing = true,
2876 	.supports_cursor = true,
2877 	.supports_block_linear = true,
2878 	.supports_sector_layout = false,
2879 	.has_legacy_blending = false,
2880 	.pitch_align = 64,
2881 	.has_powergate = true,
2882 	.coupled_pm = false,
2883 	.has_nvdisplay = false,
2884 	.num_primary_formats = ARRAY_SIZE(tegra124_primary_formats),
2885 	.primary_formats = tegra124_primary_formats,
2886 	.num_overlay_formats = ARRAY_SIZE(tegra124_overlay_formats),
2887 	.overlay_formats = tegra124_overlay_formats,
2888 	.modifiers = tegra124_modifiers,
2889 	.has_win_a_without_filters = false,
2890 	.has_win_b_vfilter_mem_client = false,
2891 	.has_win_c_without_vert_filter = false,
2892 	.plane_tiled_memory_bandwidth_x2 = false,
2893 	.has_pll_d2_out0 = true,
2894 };
2895 
2896 static const struct tegra_dc_soc_info tegra210_dc_soc_info = {
2897 	.supports_background_color = true,
2898 	.supports_interlacing = true,
2899 	.supports_cursor = true,
2900 	.supports_block_linear = true,
2901 	.supports_sector_layout = false,
2902 	.has_legacy_blending = false,
2903 	.pitch_align = 64,
2904 	.has_powergate = true,
2905 	.coupled_pm = false,
2906 	.has_nvdisplay = false,
2907 	.num_primary_formats = ARRAY_SIZE(tegra114_primary_formats),
2908 	.primary_formats = tegra114_primary_formats,
2909 	.num_overlay_formats = ARRAY_SIZE(tegra114_overlay_formats),
2910 	.overlay_formats = tegra114_overlay_formats,
2911 	.modifiers = tegra124_modifiers,
2912 	.has_win_a_without_filters = false,
2913 	.has_win_b_vfilter_mem_client = false,
2914 	.has_win_c_without_vert_filter = false,
2915 	.plane_tiled_memory_bandwidth_x2 = false,
2916 	.has_pll_d2_out0 = true,
2917 };
2918 
2919 static const struct tegra_windowgroup_soc tegra186_dc_wgrps[] = {
2920 	{
2921 		.index = 0,
2922 		.dc = 0,
2923 		.windows = (const unsigned int[]) { 0 },
2924 		.num_windows = 1,
2925 	}, {
2926 		.index = 1,
2927 		.dc = 1,
2928 		.windows = (const unsigned int[]) { 1 },
2929 		.num_windows = 1,
2930 	}, {
2931 		.index = 2,
2932 		.dc = 1,
2933 		.windows = (const unsigned int[]) { 2 },
2934 		.num_windows = 1,
2935 	}, {
2936 		.index = 3,
2937 		.dc = 2,
2938 		.windows = (const unsigned int[]) { 3 },
2939 		.num_windows = 1,
2940 	}, {
2941 		.index = 4,
2942 		.dc = 2,
2943 		.windows = (const unsigned int[]) { 4 },
2944 		.num_windows = 1,
2945 	}, {
2946 		.index = 5,
2947 		.dc = 2,
2948 		.windows = (const unsigned int[]) { 5 },
2949 		.num_windows = 1,
2950 	},
2951 };
2952 
2953 static const struct tegra_dc_soc_info tegra186_dc_soc_info = {
2954 	.supports_background_color = true,
2955 	.supports_interlacing = true,
2956 	.supports_cursor = true,
2957 	.supports_block_linear = true,
2958 	.supports_sector_layout = false,
2959 	.has_legacy_blending = false,
2960 	.pitch_align = 64,
2961 	.has_powergate = false,
2962 	.coupled_pm = false,
2963 	.has_nvdisplay = true,
2964 	.wgrps = tegra186_dc_wgrps,
2965 	.num_wgrps = ARRAY_SIZE(tegra186_dc_wgrps),
2966 	.plane_tiled_memory_bandwidth_x2 = false,
2967 	.has_pll_d2_out0 = false,
2968 };
2969 
2970 static const struct tegra_windowgroup_soc tegra194_dc_wgrps[] = {
2971 	{
2972 		.index = 0,
2973 		.dc = 0,
2974 		.windows = (const unsigned int[]) { 0 },
2975 		.num_windows = 1,
2976 	}, {
2977 		.index = 1,
2978 		.dc = 1,
2979 		.windows = (const unsigned int[]) { 1 },
2980 		.num_windows = 1,
2981 	}, {
2982 		.index = 2,
2983 		.dc = 1,
2984 		.windows = (const unsigned int[]) { 2 },
2985 		.num_windows = 1,
2986 	}, {
2987 		.index = 3,
2988 		.dc = 2,
2989 		.windows = (const unsigned int[]) { 3 },
2990 		.num_windows = 1,
2991 	}, {
2992 		.index = 4,
2993 		.dc = 2,
2994 		.windows = (const unsigned int[]) { 4 },
2995 		.num_windows = 1,
2996 	}, {
2997 		.index = 5,
2998 		.dc = 2,
2999 		.windows = (const unsigned int[]) { 5 },
3000 		.num_windows = 1,
3001 	},
3002 };
3003 
3004 static const struct tegra_dc_soc_info tegra194_dc_soc_info = {
3005 	.supports_background_color = true,
3006 	.supports_interlacing = true,
3007 	.supports_cursor = true,
3008 	.supports_block_linear = true,
3009 	.supports_sector_layout = true,
3010 	.has_legacy_blending = false,
3011 	.pitch_align = 64,
3012 	.has_powergate = false,
3013 	.coupled_pm = false,
3014 	.has_nvdisplay = true,
3015 	.wgrps = tegra194_dc_wgrps,
3016 	.num_wgrps = ARRAY_SIZE(tegra194_dc_wgrps),
3017 	.plane_tiled_memory_bandwidth_x2 = false,
3018 	.has_pll_d2_out0 = false,
3019 };
3020 
3021 static const struct of_device_id tegra_dc_of_match[] = {
3022 	{
3023 		.compatible = "nvidia,tegra194-dc",
3024 		.data = &tegra194_dc_soc_info,
3025 	}, {
3026 		.compatible = "nvidia,tegra186-dc",
3027 		.data = &tegra186_dc_soc_info,
3028 	}, {
3029 		.compatible = "nvidia,tegra210-dc",
3030 		.data = &tegra210_dc_soc_info,
3031 	}, {
3032 		.compatible = "nvidia,tegra124-dc",
3033 		.data = &tegra124_dc_soc_info,
3034 	}, {
3035 		.compatible = "nvidia,tegra114-dc",
3036 		.data = &tegra114_dc_soc_info,
3037 	}, {
3038 		.compatible = "nvidia,tegra30-dc",
3039 		.data = &tegra30_dc_soc_info,
3040 	}, {
3041 		.compatible = "nvidia,tegra20-dc",
3042 		.data = &tegra20_dc_soc_info,
3043 	}, {
3044 		/* sentinel */
3045 	}
3046 };
3047 MODULE_DEVICE_TABLE(of, tegra_dc_of_match);
3048 
3049 static int tegra_dc_parse_dt(struct tegra_dc *dc)
3050 {
3051 	struct device_node *np;
3052 	u32 value = 0;
3053 	int err;
3054 
3055 	err = of_property_read_u32(dc->dev->of_node, "nvidia,head", &value);
3056 	if (err < 0) {
3057 		dev_err(dc->dev, "missing \"nvidia,head\" property\n");
3058 
3059 		/*
3060 		 * If the nvidia,head property isn't present, try to find the
3061 		 * correct head number by looking up the position of this
3062 		 * display controller's node within the device tree. Assuming
3063 		 * that the nodes are ordered properly in the DTS file and
3064 		 * that the translation into a flattened device tree blob
3065 		 * preserves that ordering this will actually yield the right
3066 		 * head number.
3067 		 *
3068 		 * If those assumptions don't hold, this will still work for
3069 		 * cases where only a single display controller is used.
3070 		 */
3071 		for_each_matching_node(np, tegra_dc_of_match) {
3072 			if (np == dc->dev->of_node) {
3073 				of_node_put(np);
3074 				break;
3075 			}
3076 
3077 			value++;
3078 		}
3079 	}
3080 
3081 	dc->pipe = value;
3082 
3083 	return 0;
3084 }
3085 
3086 static int tegra_dc_match_by_pipe(struct device *dev, const void *data)
3087 {
3088 	struct tegra_dc *dc = dev_get_drvdata(dev);
3089 	unsigned int pipe = (unsigned long)(void *)data;
3090 
3091 	return dc->pipe == pipe;
3092 }
3093 
3094 static int tegra_dc_couple(struct tegra_dc *dc)
3095 {
3096 	/*
3097 	 * On Tegra20, DC1 requires DC0 to be taken out of reset in order to
3098 	 * be enabled, otherwise CPU hangs on writing to CMD_DISPLAY_COMMAND /
3099 	 * POWER_CONTROL registers during CRTC enabling.
3100 	 */
3101 	if (dc->soc->coupled_pm && dc->pipe == 1) {
3102 		struct device *companion;
3103 		struct tegra_dc *parent;
3104 
3105 		companion = driver_find_device(dc->dev->driver, NULL, (const void *)0,
3106 					       tegra_dc_match_by_pipe);
3107 		if (!companion)
3108 			return -EPROBE_DEFER;
3109 
3110 		parent = dev_get_drvdata(companion);
3111 		dc->client.parent = &parent->client;
3112 
3113 		dev_dbg(dc->dev, "coupled to %s\n", dev_name(companion));
3114 	}
3115 
3116 	return 0;
3117 }
3118 
3119 static int tegra_dc_init_opp_table(struct tegra_dc *dc)
3120 {
3121 	struct tegra_core_opp_params opp_params = {};
3122 	int err;
3123 
3124 	err = devm_tegra_core_dev_init_opp_table(dc->dev, &opp_params);
3125 	if (err && err != -ENODEV)
3126 		return err;
3127 
3128 	if (err)
3129 		dc->has_opp_table = false;
3130 	else
3131 		dc->has_opp_table = true;
3132 
3133 	return 0;
3134 }
3135 
3136 static int tegra_dc_probe(struct platform_device *pdev)
3137 {
3138 	u64 dma_mask = dma_get_mask(pdev->dev.parent);
3139 	struct tegra_dc *dc;
3140 	int err;
3141 
3142 	err = dma_coerce_mask_and_coherent(&pdev->dev, dma_mask);
3143 	if (err < 0) {
3144 		dev_err(&pdev->dev, "failed to set DMA mask: %d\n", err);
3145 		return err;
3146 	}
3147 
3148 	dc = devm_kzalloc(&pdev->dev, sizeof(*dc), GFP_KERNEL);
3149 	if (!dc)
3150 		return -ENOMEM;
3151 
3152 	dc->soc = of_device_get_match_data(&pdev->dev);
3153 
3154 	INIT_LIST_HEAD(&dc->list);
3155 	dc->dev = &pdev->dev;
3156 
3157 	err = tegra_dc_parse_dt(dc);
3158 	if (err < 0)
3159 		return err;
3160 
3161 	err = tegra_dc_couple(dc);
3162 	if (err < 0)
3163 		return err;
3164 
3165 	dc->clk = devm_clk_get(&pdev->dev, NULL);
3166 	if (IS_ERR(dc->clk)) {
3167 		dev_err(&pdev->dev, "failed to get clock\n");
3168 		return PTR_ERR(dc->clk);
3169 	}
3170 
3171 	dc->rst = devm_reset_control_get(&pdev->dev, "dc");
3172 	if (IS_ERR(dc->rst)) {
3173 		dev_err(&pdev->dev, "failed to get reset\n");
3174 		return PTR_ERR(dc->rst);
3175 	}
3176 
3177 	/* assert reset and disable clock */
3178 	err = clk_prepare_enable(dc->clk);
3179 	if (err < 0)
3180 		return err;
3181 
3182 	usleep_range(2000, 4000);
3183 
3184 	err = reset_control_assert(dc->rst);
3185 	if (err < 0)
3186 		return err;
3187 
3188 	usleep_range(2000, 4000);
3189 
3190 	clk_disable_unprepare(dc->clk);
3191 
3192 	if (dc->soc->has_powergate) {
3193 		if (dc->pipe == 0)
3194 			dc->powergate = TEGRA_POWERGATE_DIS;
3195 		else
3196 			dc->powergate = TEGRA_POWERGATE_DISB;
3197 
3198 		tegra_powergate_power_off(dc->powergate);
3199 	}
3200 
3201 	err = tegra_dc_init_opp_table(dc);
3202 	if (err < 0)
3203 		return err;
3204 
3205 	dc->regs = devm_platform_ioremap_resource(pdev, 0);
3206 	if (IS_ERR(dc->regs))
3207 		return PTR_ERR(dc->regs);
3208 
3209 	dc->irq = platform_get_irq(pdev, 0);
3210 	if (dc->irq < 0)
3211 		return -ENXIO;
3212 
3213 	err = tegra_dc_rgb_probe(dc);
3214 	if (err < 0 && err != -ENODEV) {
3215 		const char *level = KERN_ERR;
3216 
3217 		if (err == -EPROBE_DEFER)
3218 			level = KERN_DEBUG;
3219 
3220 		dev_printk(level, dc->dev, "failed to probe RGB output: %d\n",
3221 			   err);
3222 		return err;
3223 	}
3224 
3225 	platform_set_drvdata(pdev, dc);
3226 	pm_runtime_enable(&pdev->dev);
3227 
3228 	INIT_LIST_HEAD(&dc->client.list);
3229 	dc->client.ops = &dc_client_ops;
3230 	dc->client.dev = &pdev->dev;
3231 
3232 	err = host1x_client_register(&dc->client);
3233 	if (err < 0) {
3234 		dev_err(&pdev->dev, "failed to register host1x client: %d\n",
3235 			err);
3236 		goto disable_pm;
3237 	}
3238 
3239 	return 0;
3240 
3241 disable_pm:
3242 	pm_runtime_disable(&pdev->dev);
3243 	tegra_dc_rgb_remove(dc);
3244 
3245 	return err;
3246 }
3247 
3248 static int tegra_dc_remove(struct platform_device *pdev)
3249 {
3250 	struct tegra_dc *dc = platform_get_drvdata(pdev);
3251 	int err;
3252 
3253 	err = host1x_client_unregister(&dc->client);
3254 	if (err < 0) {
3255 		dev_err(&pdev->dev, "failed to unregister host1x client: %d\n",
3256 			err);
3257 		return err;
3258 	}
3259 
3260 	err = tegra_dc_rgb_remove(dc);
3261 	if (err < 0) {
3262 		dev_err(&pdev->dev, "failed to remove RGB output: %d\n", err);
3263 		return err;
3264 	}
3265 
3266 	pm_runtime_disable(&pdev->dev);
3267 
3268 	return 0;
3269 }
3270 
3271 struct platform_driver tegra_dc_driver = {
3272 	.driver = {
3273 		.name = "tegra-dc",
3274 		.of_match_table = tegra_dc_of_match,
3275 	},
3276 	.probe = tegra_dc_probe,
3277 	.remove = tegra_dc_remove,
3278 };
3279