1 // SPDX-License-Identifier: GPL-2.0+
3 * Copyright (C) 2018 BayLibre, SAS
5 * Copyright (C) 2015 Amlogic, Inc. All rights reserved.
8 #include <linux/kernel.h>
9 #include <linux/module.h>
10 #include <linux/mutex.h>
11 #include <linux/bitfield.h>
12 #include <linux/platform_device.h>
14 #include <drm/drm_atomic.h>
15 #include <drm/drm_atomic_helper.h>
16 #include <drm/drm_plane_helper.h>
17 #include <drm/drm_gem_cma_helper.h>
18 #include <drm/drm_fb_cma_helper.h>
19 #include <drm/drm_gem_framebuffer_helper.h>
20 #include <drm/drm_rect.h>
22 #include "meson_overlay.h"
23 #include "meson_vpp.h"
24 #include "meson_viu.h"
25 #include "meson_registers.h"
28 #define VD_URGENT_CHROMA BIT(28)
29 #define VD_URGENT_LUMA BIT(27)
30 #define VD_HOLD_LINES(lines) FIELD_PREP(GENMASK(24, 19), lines)
31 #define VD_DEMUX_MODE_RGB BIT(16)
32 #define VD_BYTES_PER_PIXEL(val) FIELD_PREP(GENMASK(15, 14), val)
33 #define VD_CHRO_RPT_LASTL_CTRL BIT(6)
34 #define VD_LITTLE_ENDIAN BIT(4)
35 #define VD_SEPARATE_EN BIT(1)
36 #define VD_ENABLE BIT(0)
39 #define CANVAS_ADDR2(addr) FIELD_PREP(GENMASK(23, 16), addr)
40 #define CANVAS_ADDR1(addr) FIELD_PREP(GENMASK(15, 8), addr)
41 #define CANVAS_ADDR0(addr) FIELD_PREP(GENMASK(7, 0), addr)
43 /* VD1_IF0_LUMA_X0 VD1_IF0_CHROMA_X0 */
44 #define VD_X_START(value) FIELD_PREP(GENMASK(14, 0), value)
45 #define VD_X_END(value) FIELD_PREP(GENMASK(30, 16), value)
47 /* VD1_IF0_LUMA_Y0 VD1_IF0_CHROMA_Y0 */
48 #define VD_Y_START(value) FIELD_PREP(GENMASK(12, 0), value)
49 #define VD_Y_END(value) FIELD_PREP(GENMASK(28, 16), value)
51 /* VD1_IF0_GEN_REG2 */
52 #define VD_COLOR_MAP(value) FIELD_PREP(GENMASK(1, 0), value)
54 /* VIU_VD1_FMT_CTRL */
55 #define VD_HORZ_Y_C_RATIO(value) FIELD_PREP(GENMASK(22, 21), value)
56 #define VD_HORZ_FMT_EN BIT(20)
57 #define VD_VERT_RPT_LINE0 BIT(16)
58 #define VD_VERT_INITIAL_PHASE(value) FIELD_PREP(GENMASK(11, 8), value)
59 #define VD_VERT_PHASE_STEP(value) FIELD_PREP(GENMASK(7, 1), value)
60 #define VD_VERT_FMT_EN BIT(0)
62 /* VPP_POSTBLEND_VD1_H_START_END */
63 #define VD_H_END(value) FIELD_PREP(GENMASK(11, 0), value)
64 #define VD_H_START(value) FIELD_PREP(GENMASK(27, 16), value)
66 /* VPP_POSTBLEND_VD1_V_START_END */
67 #define VD_V_END(value) FIELD_PREP(GENMASK(11, 0), value)
68 #define VD_V_START(value) FIELD_PREP(GENMASK(27, 16), value)
70 /* VPP_BLEND_VD2_V_START_END */
71 #define VD2_V_END(value) FIELD_PREP(GENMASK(11, 0), value)
72 #define VD2_V_START(value) FIELD_PREP(GENMASK(27, 16), value)
75 #define VD_V_WIDTH(value) FIELD_PREP(GENMASK(11, 0), value)
76 #define VD_H_WIDTH(value) FIELD_PREP(GENMASK(27, 16), value)
78 /* VPP_HSC_REGION12_STARTP VPP_HSC_REGION34_STARTP */
79 #define VD_REGION24_START(value) FIELD_PREP(GENMASK(11, 0), value)
80 #define VD_REGION13_END(value) FIELD_PREP(GENMASK(27, 16), value)
82 struct meson_overlay {
83 struct drm_plane base;
84 struct meson_drm *priv;
86 #define to_meson_overlay(x) container_of(x, struct meson_overlay, base)
88 #define FRAC_16_16(mult, div) (((mult) << 16) / (div))
90 static int meson_overlay_atomic_check(struct drm_plane *plane,
91 struct drm_plane_state *state)
93 struct drm_crtc_state *crtc_state;
98 crtc_state = drm_atomic_get_crtc_state(state->state, state->crtc);
99 if (IS_ERR(crtc_state))
100 return PTR_ERR(crtc_state);
102 return drm_atomic_helper_check_plane_state(state, crtc_state,
108 /* Takes a fixed 16.16 number and converts it to integer. */
109 static inline int64_t fixed16_to_int(int64_t value)
114 static const uint8_t skip_tab[6] = {
115 0x24, 0x04, 0x68, 0x48, 0x28, 0x08,
118 static void meson_overlay_get_vertical_phase(unsigned int ratio_y, int *phase,
119 int *repeat, bool interlace)
125 if (!interlace && ratio_y > (1 << 18))
126 offset_out = (1 * ratio_y) >> 10;
128 while ((offset_in + (4 << 8)) <= offset_out) {
133 *phase = (offset_out - offset_in) >> 2;
138 *phase = *phase & 0xff;
143 *repeat = skip_tab[repeat_skip];
146 static void meson_overlay_setup_scaler_params(struct meson_drm *priv,
147 struct drm_plane *plane,
150 struct drm_crtc_state *crtc_state = priv->crtc->state;
151 int video_top, video_left, video_width, video_height;
152 struct drm_plane_state *state = plane->state;
153 unsigned int vd_start_lines, vd_end_lines;
154 unsigned int hd_start_lines, hd_end_lines;
155 unsigned int crtc_height, crtc_width;
156 unsigned int vsc_startp, vsc_endp;
157 unsigned int hsc_startp, hsc_endp;
158 unsigned int crop_top, crop_left;
159 int vphase, vphase_repeat_skip;
160 unsigned int ratio_x, ratio_y;
161 int temp_height, temp_width;
162 unsigned int w_in, h_in;
163 int temp, start, end;
166 DRM_ERROR("Invalid crtc_state\n");
170 crtc_height = crtc_state->mode.vdisplay;
171 crtc_width = crtc_state->mode.hdisplay;
173 w_in = fixed16_to_int(state->src_w);
174 h_in = fixed16_to_int(state->src_h);
175 crop_top = fixed16_to_int(state->src_x);
176 crop_left = fixed16_to_int(state->src_x);
178 video_top = state->crtc_y;
179 video_left = state->crtc_x;
180 video_width = state->crtc_w;
181 video_height = state->crtc_h;
183 DRM_DEBUG("crtc_width %d crtc_height %d interlace %d\n",
184 crtc_width, crtc_height, interlace_mode);
185 DRM_DEBUG("w_in %d h_in %d crop_top %d crop_left %d\n",
186 w_in, h_in, crop_top, crop_left);
187 DRM_DEBUG("video top %d left %d width %d height %d\n",
188 video_top, video_left, video_width, video_height);
190 ratio_x = (w_in << 18) / video_width;
191 ratio_y = (h_in << 18) / video_height;
193 if (ratio_x * video_width < (w_in << 18))
196 DRM_DEBUG("ratio x 0x%x y 0x%x\n", ratio_x, ratio_y);
198 meson_overlay_get_vertical_phase(ratio_y, &vphase, &vphase_repeat_skip,
201 DRM_DEBUG("vphase 0x%x skip %d\n", vphase, vphase_repeat_skip);
205 start = video_top + video_height / 2 - ((h_in << 17) / ratio_y);
206 end = (h_in << 18) / ratio_y + start - 1;
208 if (video_top < 0 && start < 0)
209 vd_start_lines = (-(start) * ratio_y) >> 18;
210 else if (start < video_top)
211 vd_start_lines = ((video_top - start) * ratio_y) >> 18;
216 temp_height = min_t(unsigned int,
217 video_top + video_height - 1,
220 temp_height = min_t(unsigned int,
221 video_top + video_height - 1,
222 crtc_height - 1) - video_top + 1;
224 temp = vd_start_lines + (temp_height * ratio_y >> 18);
225 vd_end_lines = (temp <= (h_in - 1)) ? temp : (h_in - 1);
227 vd_start_lines += crop_left;
228 vd_end_lines += crop_left;
231 * TOFIX: Input frames are handled and scaled like progressive frames,
232 * proper handling of interlaced field input frames need to be figured
233 * out using the proper framebuffer flags set by userspace.
235 if (interlace_mode) {
240 vsc_startp = max_t(int, start,
241 max_t(int, 0, video_top));
242 vsc_endp = min_t(int, end,
243 min_t(int, crtc_height - 1,
244 video_top + video_height - 1));
246 DRM_DEBUG("vsc startp %d endp %d start_lines %d end_lines %d\n",
247 vsc_startp, vsc_endp, vd_start_lines, vd_end_lines);
251 start = video_left + video_width / 2 - ((w_in << 17) / ratio_x);
252 end = (w_in << 18) / ratio_x + start - 1;
254 if (video_left < 0 && start < 0)
255 hd_start_lines = (-(start) * ratio_x) >> 18;
256 else if (start < video_left)
257 hd_start_lines = ((video_left - start) * ratio_x) >> 18;
262 temp_width = min_t(unsigned int,
263 video_left + video_width - 1,
266 temp_width = min_t(unsigned int,
267 video_left + video_width - 1,
268 crtc_width - 1) - video_left + 1;
270 temp = hd_start_lines + (temp_width * ratio_x >> 18);
271 hd_end_lines = (temp <= (w_in - 1)) ? temp : (w_in - 1);
273 priv->viu.vpp_line_in_length = hd_end_lines - hd_start_lines + 1;
274 hsc_startp = max_t(int, start, max_t(int, 0, video_left));
275 hsc_endp = min_t(int, end, min_t(int, crtc_width - 1,
276 video_left + video_width - 1));
278 hd_start_lines += crop_top;
279 hd_end_lines += crop_top;
281 DRM_DEBUG("hsc startp %d endp %d start_lines %d end_lines %d\n",
282 hsc_startp, hsc_endp, hd_start_lines, hd_end_lines);
284 priv->viu.vpp_vsc_start_phase_step = ratio_y << 6;
286 priv->viu.vpp_vsc_ini_phase = vphase << 8;
287 priv->viu.vpp_vsc_phase_ctrl = (1 << 13) | (4 << 8) |
290 priv->viu.vd1_if0_luma_x0 = VD_X_START(hd_start_lines) |
291 VD_X_END(hd_end_lines);
292 priv->viu.vd1_if0_chroma_x0 = VD_X_START(hd_start_lines >> 1) |
293 VD_X_END(hd_end_lines >> 1);
295 priv->viu.viu_vd1_fmt_w =
296 VD_H_WIDTH(hd_end_lines - hd_start_lines + 1) |
297 VD_V_WIDTH(hd_end_lines/2 - hd_start_lines/2 + 1);
299 priv->viu.vd1_if0_luma_y0 = VD_Y_START(vd_start_lines) |
300 VD_Y_END(vd_end_lines);
302 priv->viu.vd1_if0_chroma_y0 = VD_Y_START(vd_start_lines >> 1) |
303 VD_Y_END(vd_end_lines >> 1);
305 priv->viu.vpp_pic_in_height = h_in;
307 priv->viu.vpp_postblend_vd1_h_start_end = VD_H_START(hsc_startp) |
309 priv->viu.vpp_blend_vd2_h_start_end = VD_H_START(hd_start_lines) |
310 VD_H_END(hd_end_lines);
311 priv->viu.vpp_hsc_region12_startp = VD_REGION13_END(0) |
312 VD_REGION24_START(hsc_startp);
313 priv->viu.vpp_hsc_region34_startp =
314 VD_REGION13_END(hsc_startp) |
315 VD_REGION24_START(hsc_endp - hsc_startp);
316 priv->viu.vpp_hsc_region4_endp = hsc_endp - hsc_startp;
317 priv->viu.vpp_hsc_start_phase_step = ratio_x << 6;
318 priv->viu.vpp_hsc_region1_phase_slope = 0;
319 priv->viu.vpp_hsc_region3_phase_slope = 0;
320 priv->viu.vpp_hsc_phase_ctrl = (1 << 21) | (4 << 16);
322 priv->viu.vpp_line_in_length = hd_end_lines - hd_start_lines + 1;
323 priv->viu.vpp_preblend_h_size = hd_end_lines - hd_start_lines + 1;
325 priv->viu.vpp_postblend_vd1_v_start_end = VD_V_START(vsc_startp) |
327 priv->viu.vpp_blend_vd2_v_start_end =
328 VD2_V_START((vd_end_lines + 1) >> 1) |
329 VD2_V_END(vd_end_lines);
331 priv->viu.vpp_vsc_region12_startp = 0;
332 priv->viu.vpp_vsc_region34_startp =
333 VD_REGION13_END(vsc_endp - vsc_startp) |
334 VD_REGION24_START(vsc_endp - vsc_startp);
335 priv->viu.vpp_vsc_region4_endp = vsc_endp - vsc_startp;
336 priv->viu.vpp_vsc_start_phase_step = ratio_y << 6;
339 static void meson_overlay_atomic_update(struct drm_plane *plane,
340 struct drm_plane_state *old_state)
342 struct meson_overlay *meson_overlay = to_meson_overlay(plane);
343 struct drm_plane_state *state = plane->state;
344 struct drm_framebuffer *fb = state->fb;
345 struct meson_drm *priv = meson_overlay->priv;
346 struct drm_gem_cma_object *gem;
350 DRM_DEBUG_DRIVER("\n");
352 interlace_mode = state->crtc->mode.flags & DRM_MODE_FLAG_INTERLACE;
354 spin_lock_irqsave(&priv->drm->event_lock, flags);
356 priv->viu.vd1_if0_gen_reg = VD_URGENT_CHROMA |
359 VD_CHRO_RPT_LASTL_CTRL |
362 /* Setup scaler params */
363 meson_overlay_setup_scaler_params(priv, plane, interlace_mode);
365 priv->viu.vd1_if0_repeat_loop = 0;
366 priv->viu.vd1_if0_luma0_rpt_pat = interlace_mode ? 8 : 0;
367 priv->viu.vd1_if0_chroma0_rpt_pat = interlace_mode ? 8 : 0;
368 priv->viu.vd1_range_map_y = 0;
369 priv->viu.vd1_range_map_cb = 0;
370 priv->viu.vd1_range_map_cr = 0;
372 /* Default values for RGB888/YUV444 */
373 priv->viu.vd1_if0_gen_reg2 = 0;
374 priv->viu.viu_vd1_fmt_ctrl = 0;
376 switch (fb->format->format) {
377 /* TOFIX DRM_FORMAT_RGB888 should be supported */
378 case DRM_FORMAT_YUYV:
379 priv->viu.vd1_if0_gen_reg |= VD_BYTES_PER_PIXEL(1);
380 priv->viu.vd1_if0_canvas0 =
381 CANVAS_ADDR2(priv->canvas_id_vd1_0) |
382 CANVAS_ADDR1(priv->canvas_id_vd1_0) |
383 CANVAS_ADDR0(priv->canvas_id_vd1_0);
384 priv->viu.viu_vd1_fmt_ctrl = VD_HORZ_Y_C_RATIO(1) | /* /2 */
387 VD_VERT_INITIAL_PHASE(12) |
388 VD_VERT_PHASE_STEP(16) | /* /2 */
391 case DRM_FORMAT_NV12:
392 case DRM_FORMAT_NV21:
393 priv->viu.vd1_if0_gen_reg |= VD_SEPARATE_EN;
394 priv->viu.vd1_if0_canvas0 =
395 CANVAS_ADDR2(priv->canvas_id_vd1_1) |
396 CANVAS_ADDR1(priv->canvas_id_vd1_1) |
397 CANVAS_ADDR0(priv->canvas_id_vd1_0);
398 if (fb->format->format == DRM_FORMAT_NV12)
399 priv->viu.vd1_if0_gen_reg2 = VD_COLOR_MAP(1);
401 priv->viu.vd1_if0_gen_reg2 = VD_COLOR_MAP(2);
402 priv->viu.viu_vd1_fmt_ctrl = VD_HORZ_Y_C_RATIO(1) | /* /2 */
405 VD_VERT_INITIAL_PHASE(12) |
406 VD_VERT_PHASE_STEP(8) | /* /4 */
409 case DRM_FORMAT_YUV444:
410 case DRM_FORMAT_YUV422:
411 case DRM_FORMAT_YUV420:
412 case DRM_FORMAT_YUV411:
413 case DRM_FORMAT_YUV410:
414 priv->viu.vd1_if0_gen_reg |= VD_SEPARATE_EN;
415 priv->viu.vd1_if0_canvas0 =
416 CANVAS_ADDR2(priv->canvas_id_vd1_2) |
417 CANVAS_ADDR1(priv->canvas_id_vd1_1) |
418 CANVAS_ADDR0(priv->canvas_id_vd1_0);
419 switch (fb->format->format) {
420 case DRM_FORMAT_YUV422:
421 priv->viu.viu_vd1_fmt_ctrl =
422 VD_HORZ_Y_C_RATIO(1) | /* /2 */
425 VD_VERT_INITIAL_PHASE(12) |
426 VD_VERT_PHASE_STEP(16) | /* /2 */
429 case DRM_FORMAT_YUV420:
430 priv->viu.viu_vd1_fmt_ctrl =
431 VD_HORZ_Y_C_RATIO(1) | /* /2 */
434 VD_VERT_INITIAL_PHASE(12) |
435 VD_VERT_PHASE_STEP(8) | /* /4 */
438 case DRM_FORMAT_YUV411:
439 priv->viu.viu_vd1_fmt_ctrl =
440 VD_HORZ_Y_C_RATIO(2) | /* /4 */
443 VD_VERT_INITIAL_PHASE(12) |
444 VD_VERT_PHASE_STEP(16) | /* /2 */
447 case DRM_FORMAT_YUV410:
448 priv->viu.viu_vd1_fmt_ctrl =
449 VD_HORZ_Y_C_RATIO(2) | /* /4 */
452 VD_VERT_INITIAL_PHASE(12) |
453 VD_VERT_PHASE_STEP(8) | /* /4 */
460 /* Update Canvas with buffer address */
461 priv->viu.vd1_planes = drm_format_num_planes(fb->format->format);
463 switch (priv->viu.vd1_planes) {
465 gem = drm_fb_cma_get_gem_obj(fb, 2);
466 priv->viu.vd1_addr2 = gem->paddr + fb->offsets[2];
467 priv->viu.vd1_stride2 = fb->pitches[2];
468 priv->viu.vd1_height2 =
469 drm_format_plane_height(fb->height,
470 fb->format->format, 2);
471 DRM_DEBUG("plane 2 addr 0x%x stride %d height %d\n",
473 priv->viu.vd1_stride2,
474 priv->viu.vd1_height2);
477 gem = drm_fb_cma_get_gem_obj(fb, 1);
478 priv->viu.vd1_addr1 = gem->paddr + fb->offsets[1];
479 priv->viu.vd1_stride1 = fb->pitches[1];
480 priv->viu.vd1_height1 =
481 drm_format_plane_height(fb->height,
482 fb->format->format, 1);
483 DRM_DEBUG("plane 1 addr 0x%x stride %d height %d\n",
485 priv->viu.vd1_stride1,
486 priv->viu.vd1_height1);
489 gem = drm_fb_cma_get_gem_obj(fb, 0);
490 priv->viu.vd1_addr0 = gem->paddr + fb->offsets[0];
491 priv->viu.vd1_stride0 = fb->pitches[0];
492 priv->viu.vd1_height0 =
493 drm_format_plane_height(fb->height,
494 fb->format->format, 0);
495 DRM_DEBUG("plane 0 addr 0x%x stride %d height %d\n",
497 priv->viu.vd1_stride0,
498 priv->viu.vd1_height0);
501 priv->viu.vd1_enabled = true;
503 spin_unlock_irqrestore(&priv->drm->event_lock, flags);
505 DRM_DEBUG_DRIVER("\n");
508 static void meson_overlay_atomic_disable(struct drm_plane *plane,
509 struct drm_plane_state *old_state)
511 struct meson_overlay *meson_overlay = to_meson_overlay(plane);
512 struct meson_drm *priv = meson_overlay->priv;
514 DRM_DEBUG_DRIVER("\n");
516 priv->viu.vd1_enabled = false;
519 if (meson_vpu_is_compatible(priv, "amlogic,meson-g12a-vpu")) {
520 writel_relaxed(0, priv->io_base + _REG(VD1_BLEND_SRC_CTRL));
521 writel_relaxed(0, priv->io_base + _REG(VD2_BLEND_SRC_CTRL));
522 writel_relaxed(0, priv->io_base + _REG(VD1_IF0_GEN_REG + 0x17b0));
523 writel_relaxed(0, priv->io_base + _REG(VD2_IF0_GEN_REG + 0x17b0));
525 writel_bits_relaxed(VPP_VD1_POSTBLEND | VPP_VD1_PREBLEND, 0,
526 priv->io_base + _REG(VPP_MISC));
530 static const struct drm_plane_helper_funcs meson_overlay_helper_funcs = {
531 .atomic_check = meson_overlay_atomic_check,
532 .atomic_disable = meson_overlay_atomic_disable,
533 .atomic_update = meson_overlay_atomic_update,
534 .prepare_fb = drm_gem_fb_prepare_fb,
537 static const struct drm_plane_funcs meson_overlay_funcs = {
538 .update_plane = drm_atomic_helper_update_plane,
539 .disable_plane = drm_atomic_helper_disable_plane,
540 .destroy = drm_plane_cleanup,
541 .reset = drm_atomic_helper_plane_reset,
542 .atomic_duplicate_state = drm_atomic_helper_plane_duplicate_state,
543 .atomic_destroy_state = drm_atomic_helper_plane_destroy_state,
546 static const uint32_t supported_drm_formats[] = {
557 int meson_overlay_create(struct meson_drm *priv)
559 struct meson_overlay *meson_overlay;
560 struct drm_plane *plane;
562 DRM_DEBUG_DRIVER("\n");
564 meson_overlay = devm_kzalloc(priv->drm->dev, sizeof(*meson_overlay),
569 meson_overlay->priv = priv;
570 plane = &meson_overlay->base;
572 drm_universal_plane_init(priv->drm, plane, 0xFF,
573 &meson_overlay_funcs,
574 supported_drm_formats,
575 ARRAY_SIZE(supported_drm_formats),
577 DRM_PLANE_TYPE_OVERLAY, "meson_overlay_plane");
579 drm_plane_helper_add(plane, &meson_overlay_helper_funcs);
581 priv->overlay_plane = plane;
583 DRM_DEBUG_DRIVER("\n");