4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
24 * Daniel Vetter <daniel@ffwll.ch>
26 * Derived from Xorg ddx, xf86-video-intel, src/i830_video.c
33 #include "intel_drv.h"
35 /* Limits for overlay size. According to intel doc, the real limits are:
36 * Y width: 4095, UV width (planar): 2047, Y height: 2047,
37 * UV width (planar): * 1023. But the xorg thinks 2048 for height and width. Use
38 * the mininum of both. */
39 #define IMAGE_MAX_WIDTH 2048
40 #define IMAGE_MAX_HEIGHT 2046 /* 2 * 1023 */
41 /* on 830 and 845 these large limits result in the card hanging */
42 #define IMAGE_MAX_WIDTH_LEGACY 1024
43 #define IMAGE_MAX_HEIGHT_LEGACY 1088
45 /* overlay register definitions */
47 #define OCMD_TILED_SURFACE (0x1<<19)
48 #define OCMD_MIRROR_MASK (0x3<<17)
49 #define OCMD_MIRROR_MODE (0x3<<17)
50 #define OCMD_MIRROR_HORIZONTAL (0x1<<17)
51 #define OCMD_MIRROR_VERTICAL (0x2<<17)
52 #define OCMD_MIRROR_BOTH (0x3<<17)
53 #define OCMD_BYTEORDER_MASK (0x3<<14) /* zero for YUYV or FOURCC YUY2 */
54 #define OCMD_UV_SWAP (0x1<<14) /* YVYU */
55 #define OCMD_Y_SWAP (0x2<<14) /* UYVY or FOURCC UYVY */
56 #define OCMD_Y_AND_UV_SWAP (0x3<<14) /* VYUY */
57 #define OCMD_SOURCE_FORMAT_MASK (0xf<<10)
58 #define OCMD_RGB_888 (0x1<<10) /* not in i965 Intel docs */
59 #define OCMD_RGB_555 (0x2<<10) /* not in i965 Intel docs */
60 #define OCMD_RGB_565 (0x3<<10) /* not in i965 Intel docs */
61 #define OCMD_YUV_422_PACKED (0x8<<10)
62 #define OCMD_YUV_411_PACKED (0x9<<10) /* not in i965 Intel docs */
63 #define OCMD_YUV_420_PLANAR (0xc<<10)
64 #define OCMD_YUV_422_PLANAR (0xd<<10)
65 #define OCMD_YUV_410_PLANAR (0xe<<10) /* also 411 */
66 #define OCMD_TVSYNCFLIP_PARITY (0x1<<9)
67 #define OCMD_TVSYNCFLIP_ENABLE (0x1<<7)
68 #define OCMD_BUF_TYPE_MASK (Ox1<<5)
69 #define OCMD_BUF_TYPE_FRAME (0x0<<5)
70 #define OCMD_BUF_TYPE_FIELD (0x1<<5)
71 #define OCMD_TEST_MODE (0x1<<4)
72 #define OCMD_BUFFER_SELECT (0x3<<2)
73 #define OCMD_BUFFER0 (0x0<<2)
74 #define OCMD_BUFFER1 (0x1<<2)
75 #define OCMD_FIELD_SELECT (0x1<<2)
76 #define OCMD_FIELD0 (0x0<<1)
77 #define OCMD_FIELD1 (0x1<<1)
78 #define OCMD_ENABLE (0x1<<0)
80 /* OCONFIG register */
81 #define OCONF_PIPE_MASK (0x1<<18)
82 #define OCONF_PIPE_A (0x0<<18)
83 #define OCONF_PIPE_B (0x1<<18)
84 #define OCONF_GAMMA2_ENABLE (0x1<<16)
85 #define OCONF_CSC_MODE_BT601 (0x0<<5)
86 #define OCONF_CSC_MODE_BT709 (0x1<<5)
87 #define OCONF_CSC_BYPASS (0x1<<4)
88 #define OCONF_CC_OUT_8BIT (0x1<<3)
89 #define OCONF_TEST_MODE (0x1<<2)
90 #define OCONF_THREE_LINE_BUFFER (0x1<<0)
91 #define OCONF_TWO_LINE_BUFFER (0x0<<0)
93 /* DCLRKM (dst-key) register */
94 #define DST_KEY_ENABLE (0x1<<31)
95 #define CLK_RGB24_MASK 0x0
96 #define CLK_RGB16_MASK 0x070307
97 #define CLK_RGB15_MASK 0x070707
98 #define CLK_RGB8I_MASK 0xffffff
100 #define RGB16_TO_COLORKEY(c) \
101 (((c & 0xF800) << 8) | ((c & 0x07E0) << 5) | ((c & 0x001F) << 3))
102 #define RGB15_TO_COLORKEY(c) \
103 (((c & 0x7c00) << 9) | ((c & 0x03E0) << 6) | ((c & 0x001F) << 3))
105 /* overlay flip addr flag */
106 #define OFC_UPDATE 0x1
108 /* polyphase filter coefficients */
109 #define N_HORIZ_Y_TAPS 5
110 #define N_VERT_Y_TAPS 3
111 #define N_HORIZ_UV_TAPS 3
112 #define N_VERT_UV_TAPS 3
116 /* memory bufferd overlay registers */
117 struct overlay_registers {
145 u32 RESERVED1; /* 0x6C */
158 u32 FASTHSCALE; /* 0xA0 */
159 u32 UVSCALEV; /* 0xA4 */
160 u32 RESERVEDC[(0x200 - 0xA8) / 4]; /* 0xA8 - 0x1FC */
161 u16 Y_VCOEFS[N_VERT_Y_TAPS * N_PHASES]; /* 0x200 */
162 u16 RESERVEDD[0x100 / 2 - N_VERT_Y_TAPS * N_PHASES];
163 u16 Y_HCOEFS[N_HORIZ_Y_TAPS * N_PHASES]; /* 0x300 */
164 u16 RESERVEDE[0x200 / 2 - N_HORIZ_Y_TAPS * N_PHASES];
165 u16 UV_VCOEFS[N_VERT_UV_TAPS * N_PHASES]; /* 0x500 */
166 u16 RESERVEDF[0x100 / 2 - N_VERT_UV_TAPS * N_PHASES];
167 u16 UV_HCOEFS[N_HORIZ_UV_TAPS * N_PHASES]; /* 0x600 */
168 u16 RESERVEDG[0x100 / 2 - N_HORIZ_UV_TAPS * N_PHASES];
171 /* overlay flip addr flag */
172 #define OFC_UPDATE 0x1
174 #define OVERLAY_NONPHYSICAL(dev) (IS_G33(dev) || IS_I965G(dev))
175 #define OVERLAY_EXISTS(dev) (!IS_G4X(dev) && !IS_IGDNG(dev))
178 static struct overlay_registers *intel_overlay_map_regs_atomic(struct intel_overlay *overlay)
180 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
181 struct overlay_registers *regs;
183 /* no recursive mappings */
184 BUG_ON(overlay->virt_addr);
186 if (OVERLAY_NONPHYSICAL(overlay->dev)) {
187 regs = io_mapping_map_atomic_wc(dev_priv->mm.gtt_mapping,
188 overlay->reg_bo->gtt_offset);
191 DRM_ERROR("failed to map overlay regs in GTT\n");
195 regs = overlay->reg_bo->phys_obj->handle->vaddr;
197 return overlay->virt_addr = regs;
200 static void intel_overlay_unmap_regs_atomic(struct intel_overlay *overlay)
202 struct drm_device *dev = overlay->dev;
203 drm_i915_private_t *dev_priv = dev->dev_private;
205 if (OVERLAY_NONPHYSICAL(overlay->dev))
206 io_mapping_unmap_atomic(overlay->virt_addr);
208 overlay->virt_addr = NULL;
210 I915_READ(OVADD); /* flush wc cashes */
215 /* overlay needs to be disable in OCMD reg */
216 static int intel_overlay_on(struct intel_overlay *overlay)
218 struct drm_device *dev = overlay->dev;
219 drm_i915_private_t *dev_priv = dev->dev_private;
223 BUG_ON(overlay->active);
228 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_ON);
229 OUT_RING(overlay->flip_addr | OFC_UPDATE);
230 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
234 ret = i915_lp_ring_sync(dev);
236 DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
237 overlay->hw_wedged = 1;
246 /* overlay needs to be enabled in OCMD reg */
247 static void intel_overlay_continue(struct intel_overlay *overlay,
248 bool load_polyphase_filter)
250 struct drm_device *dev = overlay->dev;
251 drm_i915_private_t *dev_priv = dev->dev_private;
252 u32 flip_addr = overlay->flip_addr;
256 BUG_ON(!overlay->active);
258 if (load_polyphase_filter)
259 flip_addr |= OFC_UPDATE;
261 /* check for underruns */
262 tmp = I915_READ(DOVSTA);
264 DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp);
269 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
273 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
276 static int intel_overlay_wait_flip(struct intel_overlay *overlay)
278 struct drm_device *dev = overlay->dev;
279 drm_i915_private_t *dev_priv = dev->dev_private;
284 if (overlay->last_flip_req != 0) {
285 ret = i915_do_wait_request(dev, overlay->last_flip_req, 0);
290 overlay->last_flip_req = 0;
292 tmp = I915_READ(ISR);
294 if (!(tmp & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT))
298 /* synchronous slowpath */
300 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
304 /* run in lockstep with the hw for easier testing */
305 ret = i915_lp_ring_sync(dev);
307 DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
308 overlay->hw_wedged = 1;
314 /* overlay needs to be disabled in OCMD reg */
315 static int intel_overlay_off(struct intel_overlay *overlay)
317 u32 flip_addr = overlay->flip_addr;
318 struct drm_device *dev = overlay->dev;
319 drm_i915_private_t *dev_priv = dev->dev_private;
323 BUG_ON(!overlay->active);
325 /* According to intel docs the overlay hw may hang (when switching
326 * off) without loading the filter coeffs. It is however unclear whether
327 * this applies to the disabling of the overlay or to the switching off
328 * of the hw. Do it in both cases */
329 flip_addr |= OFC_UPDATE;
331 /* wait for overlay to go idle */
335 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
337 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
341 ret = i915_lp_ring_sync(dev);
343 DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
344 overlay->hw_wedged = 1;
348 /* turn overlay off */
352 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
354 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
358 ret = i915_lp_ring_sync(dev);
360 DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
361 overlay->hw_wedged = 1;
370 /* Wait for pending overlay flip and release old frame.
371 * Needs to be called before the overlay register are changed
372 * via intel_overlay_(un)map_regs_atomic */
373 static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
376 struct drm_gem_object *obj;
378 ret = intel_overlay_wait_flip(overlay);
382 if (!overlay->old_vid_bo)
385 obj = overlay->old_vid_bo->obj;
386 i915_gem_object_unpin(obj);
387 drm_gem_object_unreference(obj);
388 overlay->old_vid_bo = NULL;
393 struct put_image_params {
410 static int packed_depth_bytes(u32 format)
412 switch (format & I915_OVERLAY_DEPTH_MASK) {
413 case I915_OVERLAY_YUV422:
415 case I915_OVERLAY_YUV411:
416 /* return 6; not implemented */
422 static int packed_width_bytes(u32 format, short width)
424 switch (format & I915_OVERLAY_DEPTH_MASK) {
425 case I915_OVERLAY_YUV422:
432 static int uv_hsubsampling(u32 format)
434 switch (format & I915_OVERLAY_DEPTH_MASK) {
435 case I915_OVERLAY_YUV422:
436 case I915_OVERLAY_YUV420:
438 case I915_OVERLAY_YUV411:
439 case I915_OVERLAY_YUV410:
446 static int uv_vsubsampling(u32 format)
448 switch (format & I915_OVERLAY_DEPTH_MASK) {
449 case I915_OVERLAY_YUV420:
450 case I915_OVERLAY_YUV410:
452 case I915_OVERLAY_YUV422:
453 case I915_OVERLAY_YUV411:
460 static u32 calc_swidthsw(struct drm_device *dev, u32 offset, u32 width)
462 u32 mask, shift, ret;
470 ret = ((offset + width + mask) >> shift) - (offset >> shift);
477 static const u16 y_static_hcoeffs[N_HORIZ_Y_TAPS * N_PHASES] = {
478 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
479 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
480 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
481 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
482 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
483 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
484 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
485 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
486 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
487 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
488 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
489 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
490 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
491 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
492 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
493 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
494 0xb000, 0x3000, 0x0800, 0x3000, 0xb000};
495 static const u16 uv_static_hcoeffs[N_HORIZ_UV_TAPS * N_PHASES] = {
496 0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
497 0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
498 0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
499 0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
500 0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
501 0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
502 0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
503 0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
504 0x3000, 0x0800, 0x3000};
506 static void update_polyphase_filter(struct overlay_registers *regs)
508 memcpy(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
509 memcpy(regs->UV_HCOEFS, uv_static_hcoeffs, sizeof(uv_static_hcoeffs));
512 static bool update_scaling_factors(struct intel_overlay *overlay,
513 struct overlay_registers *regs,
514 struct put_image_params *params)
516 /* fixed point with a 12 bit shift */
517 u32 xscale, yscale, xscale_UV, yscale_UV;
519 #define FRACT_MASK 0xfff
520 bool scale_changed = false;
521 int uv_hscale = uv_hsubsampling(params->format);
522 int uv_vscale = uv_vsubsampling(params->format);
524 if (params->dst_w > 1)
525 xscale = ((params->src_scan_w - 1) << FP_SHIFT)
528 xscale = 1 << FP_SHIFT;
530 if (params->dst_h > 1)
531 yscale = ((params->src_scan_h - 1) << FP_SHIFT)
534 yscale = 1 << FP_SHIFT;
536 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
537 xscale_UV = xscale/uv_hscale;
538 yscale_UV = yscale/uv_vscale;
539 /* make the Y scale to UV scale ratio an exact multiply */
540 xscale = xscale_UV * uv_hscale;
541 yscale = yscale_UV * uv_vscale;
547 if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
548 scale_changed = true;
549 overlay->old_xscale = xscale;
550 overlay->old_yscale = yscale;
552 regs->YRGBSCALE = ((yscale & FRACT_MASK) << 20)
553 | ((xscale >> FP_SHIFT) << 16)
554 | ((xscale & FRACT_MASK) << 3);
555 regs->UVSCALE = ((yscale_UV & FRACT_MASK) << 20)
556 | ((xscale_UV >> FP_SHIFT) << 16)
557 | ((xscale_UV & FRACT_MASK) << 3);
558 regs->UVSCALEV = ((yscale >> FP_SHIFT) << 16)
559 | ((yscale_UV >> FP_SHIFT) << 0);
562 update_polyphase_filter(regs);
564 return scale_changed;
567 static void update_colorkey(struct intel_overlay *overlay,
568 struct overlay_registers *regs)
570 u32 key = overlay->color_key;
571 switch (overlay->crtc->base.fb->bits_per_pixel) {
574 regs->DCLRKM = CLK_RGB8I_MASK | DST_KEY_ENABLE;
576 if (overlay->crtc->base.fb->depth == 15) {
577 regs->DCLRKV = RGB15_TO_COLORKEY(key);
578 regs->DCLRKM = CLK_RGB15_MASK | DST_KEY_ENABLE;
580 regs->DCLRKV = RGB16_TO_COLORKEY(key);
581 regs->DCLRKM = CLK_RGB16_MASK | DST_KEY_ENABLE;
586 regs->DCLRKM = CLK_RGB24_MASK | DST_KEY_ENABLE;
590 static u32 overlay_cmd_reg(struct put_image_params *params)
592 u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
594 if (params->format & I915_OVERLAY_YUV_PLANAR) {
595 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
596 case I915_OVERLAY_YUV422:
597 cmd |= OCMD_YUV_422_PLANAR;
599 case I915_OVERLAY_YUV420:
600 cmd |= OCMD_YUV_420_PLANAR;
602 case I915_OVERLAY_YUV411:
603 case I915_OVERLAY_YUV410:
604 cmd |= OCMD_YUV_410_PLANAR;
607 } else { /* YUV packed */
608 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
609 case I915_OVERLAY_YUV422:
610 cmd |= OCMD_YUV_422_PACKED;
612 case I915_OVERLAY_YUV411:
613 cmd |= OCMD_YUV_411_PACKED;
617 switch (params->format & I915_OVERLAY_SWAP_MASK) {
618 case I915_OVERLAY_NO_SWAP:
620 case I915_OVERLAY_UV_SWAP:
623 case I915_OVERLAY_Y_SWAP:
626 case I915_OVERLAY_Y_AND_UV_SWAP:
627 cmd |= OCMD_Y_AND_UV_SWAP;
635 int intel_overlay_do_put_image(struct intel_overlay *overlay,
636 struct drm_gem_object *new_bo,
637 struct put_image_params *params)
640 struct overlay_registers *regs;
641 bool scale_changed = false;
642 struct drm_i915_gem_object *bo_priv = new_bo->driver_private;
643 struct drm_device *dev = overlay->dev;
645 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
646 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
649 if (overlay->hw_wedged)
652 ret = intel_overlay_release_old_vid(overlay);
656 ret = i915_gem_object_pin(new_bo, PAGE_SIZE);
660 ret = i915_gem_object_set_to_gtt_domain(new_bo, 0);
664 if (!overlay->active) {
665 regs = intel_overlay_map_regs_atomic(overlay);
670 regs->OCONFIG = OCONF_CC_OUT_8BIT;
671 if (IS_I965GM(overlay->dev))
672 regs->OCONFIG |= OCONF_CSC_MODE_BT709;
673 regs->OCONFIG |= overlay->crtc->pipe == 0 ?
674 OCONF_PIPE_A : OCONF_PIPE_B;
675 intel_overlay_unmap_regs_atomic(overlay);
677 ret = intel_overlay_on(overlay);
682 regs = intel_overlay_map_regs_atomic(overlay);
688 regs->DWINPOS = (params->dst_y << 16) | params->dst_x;
689 regs->DWINSZ = (params->dst_h << 16) | params->dst_w;
691 if (params->format & I915_OVERLAY_YUV_PACKED)
692 tmp_width = packed_width_bytes(params->format, params->src_w);
694 tmp_width = params->src_w;
696 regs->SWIDTH = params->src_w;
697 regs->SWIDTHSW = calc_swidthsw(overlay->dev,
698 params->offset_Y, tmp_width);
699 regs->SHEIGHT = params->src_h;
700 regs->OBUF_0Y = bo_priv->gtt_offset + params-> offset_Y;
701 regs->OSTRIDE = params->stride_Y;
703 if (params->format & I915_OVERLAY_YUV_PLANAR) {
704 int uv_hscale = uv_hsubsampling(params->format);
705 int uv_vscale = uv_vsubsampling(params->format);
707 regs->SWIDTH |= (params->src_w/uv_hscale) << 16;
708 tmp_U = calc_swidthsw(overlay->dev, params->offset_U,
709 params->src_w/uv_hscale);
710 tmp_V = calc_swidthsw(overlay->dev, params->offset_V,
711 params->src_w/uv_hscale);
712 regs->SWIDTHSW |= max_t(u32, tmp_U, tmp_V) << 16;
713 regs->SHEIGHT |= (params->src_h/uv_vscale) << 16;
714 regs->OBUF_0U = bo_priv->gtt_offset + params->offset_U;
715 regs->OBUF_0V = bo_priv->gtt_offset + params->offset_V;
716 regs->OSTRIDE |= params->stride_UV << 16;
719 scale_changed = update_scaling_factors(overlay, regs, params);
721 update_colorkey(overlay, regs);
723 regs->OCMD = overlay_cmd_reg(params);
725 intel_overlay_unmap_regs_atomic(overlay);
727 intel_overlay_continue(overlay, scale_changed);
729 overlay->old_vid_bo = overlay->vid_bo;
730 overlay->vid_bo = new_bo->driver_private;
735 i915_gem_object_unpin(new_bo);
739 int intel_overlay_switch_off(struct intel_overlay *overlay)
742 struct overlay_registers *regs;
743 struct drm_gem_object *obj;
744 struct drm_device *dev = overlay->dev;
746 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
747 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
749 if (!overlay->active)
752 if (overlay->hw_wedged)
755 ret = intel_overlay_release_old_vid(overlay);
759 regs = intel_overlay_map_regs_atomic(overlay);
761 intel_overlay_unmap_regs_atomic(overlay);
763 ret = intel_overlay_off(overlay);
764 /* never have the overlay hw on without showing a frame */
765 BUG_ON(!overlay->vid_bo);
766 obj = overlay->vid_bo->obj;
768 i915_gem_object_unpin(obj);
769 drm_gem_object_unreference(obj);
770 overlay->vid_bo = NULL;
772 overlay->crtc->overlay = NULL;
773 overlay->crtc = NULL;
778 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
779 struct intel_crtc *crtc)
781 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
783 int pipeconf_reg = (crtc->pipe == 0) ? PIPEACONF : PIPEBCONF;
785 if (!crtc->base.enabled || crtc->dpms_mode != DRM_MODE_DPMS_ON)
788 pipeconf = I915_READ(pipeconf_reg);
790 /* can't use the overlay with double wide pipe */
791 if (!IS_I965G(overlay->dev) && pipeconf & PIPEACONF_DOUBLE_WIDE)
797 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
799 struct drm_device *dev = overlay->dev;
800 drm_i915_private_t *dev_priv = dev->dev_private;
802 u32 pfit_control = I915_READ(PFIT_CONTROL);
804 /* XXX: This is not the same logic as in the xorg driver, but more in
805 * line with the intel documentation for the i965 */
806 if (!IS_I965G(dev) && (pfit_control & VERT_AUTO_SCALE)) {
807 ratio = I915_READ(PFIT_AUTO_RATIOS) >> PFIT_VERT_SCALE_SHIFT;
808 } else { /* on i965 use the PGM reg to read out the autoscaler values */
809 ratio = I915_READ(PFIT_PGM_RATIOS);
811 ratio >>= PFIT_VERT_SCALE_SHIFT_965;
813 ratio >>= PFIT_VERT_SCALE_SHIFT;
816 overlay->pfit_vscale_ratio = ratio;
819 static int check_overlay_dst(struct intel_overlay *overlay,
820 struct drm_intel_overlay_put_image *rec)
822 struct drm_display_mode *mode = &overlay->crtc->base.mode;
824 if ((rec->dst_x < mode->crtc_hdisplay)
825 && (rec->dst_x + rec->dst_width
826 <= mode->crtc_hdisplay)
827 && (rec->dst_y < mode->crtc_vdisplay)
828 && (rec->dst_y + rec->dst_height
829 <= mode->crtc_vdisplay))
835 static int check_overlay_scaling(struct put_image_params *rec)
839 /* downscaling limit is 8.0 */
840 tmp = ((rec->src_scan_h << 16) / rec->dst_h) >> 16;
843 tmp = ((rec->src_scan_w << 16) / rec->dst_w) >> 16;
850 static int check_overlay_src(struct drm_device *dev,
851 struct drm_intel_overlay_put_image *rec,
852 struct drm_gem_object *new_bo)
856 int uv_hscale = uv_hsubsampling(rec->flags);
857 int uv_vscale = uv_vsubsampling(rec->flags);
860 /* check src dimensions */
861 if (IS_845G(dev) || IS_I830(dev)) {
862 if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY
863 || rec->src_width > IMAGE_MAX_WIDTH_LEGACY)
866 if (rec->src_height > IMAGE_MAX_HEIGHT
867 || rec->src_width > IMAGE_MAX_WIDTH)
870 /* better safe than sorry, use 4 as the maximal subsampling ratio */
871 if (rec->src_height < N_VERT_Y_TAPS*4
872 || rec->src_width < N_HORIZ_Y_TAPS*4)
875 /* check alingment constrains */
876 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
877 case I915_OVERLAY_RGB:
878 /* not implemented */
880 case I915_OVERLAY_YUV_PACKED:
881 depth = packed_depth_bytes(rec->flags);
886 /* ignore UV planes */
890 /* check pixel alignment */
891 if (rec->offset_Y % depth)
894 case I915_OVERLAY_YUV_PLANAR:
895 if (uv_vscale < 0 || uv_hscale < 0)
897 /* no offset restrictions for planar formats */
903 if (rec->src_width % uv_hscale)
906 /* stride checking */
909 if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
911 if (IS_I965G(dev) && rec->stride_Y < 512)
914 tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
916 if (rec->stride_Y > tmp*1024 || rec->stride_UV > 2*1024)
919 /* check buffer dimensions */
920 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
921 case I915_OVERLAY_RGB:
922 case I915_OVERLAY_YUV_PACKED:
923 /* always 4 Y values per depth pixels */
924 if (packed_width_bytes(rec->flags, rec->src_width)
928 tmp = rec->stride_Y*rec->src_height;
929 if (rec->offset_Y + tmp > new_bo->size)
932 case I915_OVERLAY_YUV_PLANAR:
933 if (rec->src_width > rec->stride_Y)
935 if (rec->src_width/uv_hscale > rec->stride_UV)
938 tmp = rec->stride_Y*rec->src_height;
939 if (rec->offset_Y + tmp > new_bo->size)
941 tmp = rec->stride_UV*rec->src_height;
943 if (rec->offset_U + tmp > new_bo->size
944 || rec->offset_V + tmp > new_bo->size)
952 int intel_overlay_put_image(struct drm_device *dev, void *data,
953 struct drm_file *file_priv)
955 struct drm_intel_overlay_put_image *put_image_rec = data;
956 drm_i915_private_t *dev_priv = dev->dev_private;
957 struct intel_overlay *overlay;
958 struct drm_mode_object *drmmode_obj;
959 struct intel_crtc *crtc;
960 struct drm_gem_object *new_bo;
961 struct put_image_params *params;
965 DRM_ERROR("called with no initialization\n");
969 overlay = dev_priv->overlay;
971 DRM_DEBUG("userspace bug: no overlay\n");
975 if (!(put_image_rec->flags & I915_OVERLAY_ENABLE)) {
976 mutex_lock(&dev->mode_config.mutex);
977 mutex_lock(&dev->struct_mutex);
979 ret = intel_overlay_switch_off(overlay);
981 mutex_unlock(&dev->struct_mutex);
982 mutex_unlock(&dev->mode_config.mutex);
987 params = kmalloc(sizeof(struct put_image_params), GFP_KERNEL);
991 drmmode_obj = drm_mode_object_find(dev, put_image_rec->crtc_id,
992 DRM_MODE_OBJECT_CRTC);
995 crtc = to_intel_crtc(obj_to_crtc(drmmode_obj));
997 new_bo = drm_gem_object_lookup(dev, file_priv,
998 put_image_rec->bo_handle);
1002 mutex_lock(&dev->mode_config.mutex);
1003 mutex_lock(&dev->struct_mutex);
1005 if (overlay->crtc != crtc) {
1006 struct drm_display_mode *mode = &crtc->base.mode;
1007 ret = intel_overlay_switch_off(overlay);
1011 ret = check_overlay_possible_on_crtc(overlay, crtc);
1015 overlay->crtc = crtc;
1016 crtc->overlay = overlay;
1018 if (intel_panel_fitter_pipe(dev) == crtc->pipe
1019 /* and line to wide, i.e. one-line-mode */
1020 && mode->hdisplay > 1024) {
1021 overlay->pfit_active = 1;
1022 update_pfit_vscale_ratio(overlay);
1024 overlay->pfit_active = 0;
1027 ret = check_overlay_dst(overlay, put_image_rec);
1031 if (overlay->pfit_active) {
1032 params->dst_y = ((((u32)put_image_rec->dst_y) << 12) /
1033 overlay->pfit_vscale_ratio);
1034 /* shifting right rounds downwards, so add 1 */
1035 params->dst_h = ((((u32)put_image_rec->dst_height) << 12) /
1036 overlay->pfit_vscale_ratio) + 1;
1038 params->dst_y = put_image_rec->dst_y;
1039 params->dst_h = put_image_rec->dst_height;
1041 params->dst_x = put_image_rec->dst_x;
1042 params->dst_w = put_image_rec->dst_width;
1044 params->src_w = put_image_rec->src_width;
1045 params->src_h = put_image_rec->src_height;
1046 params->src_scan_w = put_image_rec->src_scan_width;
1047 params->src_scan_h = put_image_rec->src_scan_height;
1048 if (params->src_scan_h > params->src_h
1049 || params->src_scan_w > params->src_w) {
1054 ret = check_overlay_src(dev, put_image_rec, new_bo);
1057 params->format = put_image_rec->flags & ~I915_OVERLAY_FLAGS_MASK;
1058 params->stride_Y = put_image_rec->stride_Y;
1059 params->stride_UV = put_image_rec->stride_UV;
1060 params->offset_Y = put_image_rec->offset_Y;
1061 params->offset_U = put_image_rec->offset_U;
1062 params->offset_V = put_image_rec->offset_V;
1064 /* Check scaling after src size to prevent a divide-by-zero. */
1065 ret = check_overlay_scaling(params);
1069 ret = intel_overlay_do_put_image(overlay, new_bo, params);
1073 mutex_unlock(&dev->struct_mutex);
1074 mutex_unlock(&dev->mode_config.mutex);
1081 mutex_unlock(&dev->struct_mutex);
1082 mutex_unlock(&dev->mode_config.mutex);
1083 drm_gem_object_unreference(new_bo);
1089 static void update_reg_attrs(struct intel_overlay *overlay,
1090 struct overlay_registers *regs)
1092 regs->OCLRC0 = (overlay->contrast << 18) | (overlay->brightness & 0xff);
1093 regs->OCLRC1 = overlay->saturation;
1096 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
1100 if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
1103 for (i = 0; i < 3; i++) {
1104 if (((gamma1 >> i * 8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
1111 static bool check_gamma5_errata(u32 gamma5)
1115 for (i = 0; i < 3; i++) {
1116 if (((gamma5 >> i*8) & 0xff) == 0x80)
1123 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
1125 if (!check_gamma_bounds(0, attrs->gamma0)
1126 || !check_gamma_bounds(attrs->gamma0, attrs->gamma1)
1127 || !check_gamma_bounds(attrs->gamma1, attrs->gamma2)
1128 || !check_gamma_bounds(attrs->gamma2, attrs->gamma3)
1129 || !check_gamma_bounds(attrs->gamma3, attrs->gamma4)
1130 || !check_gamma_bounds(attrs->gamma4, attrs->gamma5)
1131 || !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
1133 if (!check_gamma5_errata(attrs->gamma5))
1138 int intel_overlay_attrs(struct drm_device *dev, void *data,
1139 struct drm_file *file_priv)
1141 struct drm_intel_overlay_attrs *attrs = data;
1142 drm_i915_private_t *dev_priv = dev->dev_private;
1143 struct intel_overlay *overlay;
1144 struct overlay_registers *regs;
1148 DRM_ERROR("called with no initialization\n");
1152 overlay = dev_priv->overlay;
1154 DRM_DEBUG("userspace bug: no overlay\n");
1158 mutex_lock(&dev->mode_config.mutex);
1159 mutex_lock(&dev->struct_mutex);
1161 if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
1162 attrs->color_key = overlay->color_key;
1163 attrs->brightness = overlay->brightness;
1164 attrs->contrast = overlay->contrast;
1165 attrs->saturation = overlay->saturation;
1168 attrs->gamma0 = I915_READ(OGAMC0);
1169 attrs->gamma1 = I915_READ(OGAMC1);
1170 attrs->gamma2 = I915_READ(OGAMC2);
1171 attrs->gamma3 = I915_READ(OGAMC3);
1172 attrs->gamma4 = I915_READ(OGAMC4);
1173 attrs->gamma5 = I915_READ(OGAMC5);
1177 overlay->color_key = attrs->color_key;
1178 if (attrs->brightness >= -128 && attrs->brightness <= 127) {
1179 overlay->brightness = attrs->brightness;
1184 if (attrs->contrast <= 255) {
1185 overlay->contrast = attrs->contrast;
1190 if (attrs->saturation <= 1023) {
1191 overlay->saturation = attrs->saturation;
1197 regs = intel_overlay_map_regs_atomic(overlay);
1203 update_reg_attrs(overlay, regs);
1205 intel_overlay_unmap_regs_atomic(overlay);
1207 if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
1208 if (!IS_I9XX(dev)) {
1213 if (overlay->active) {
1218 ret = check_gamma(attrs);
1222 I915_WRITE(OGAMC0, attrs->gamma0);
1223 I915_WRITE(OGAMC1, attrs->gamma1);
1224 I915_WRITE(OGAMC2, attrs->gamma2);
1225 I915_WRITE(OGAMC3, attrs->gamma3);
1226 I915_WRITE(OGAMC4, attrs->gamma4);
1227 I915_WRITE(OGAMC5, attrs->gamma5);
1233 mutex_unlock(&dev->struct_mutex);
1234 mutex_unlock(&dev->mode_config.mutex);
1239 void intel_setup_overlay(struct drm_device *dev)
1241 drm_i915_private_t *dev_priv = dev->dev_private;
1242 struct intel_overlay *overlay;
1243 struct drm_gem_object *reg_bo;
1244 struct overlay_registers *regs;
1247 if (!OVERLAY_EXISTS(dev))
1250 overlay = kzalloc(sizeof(struct intel_overlay), GFP_KERNEL);
1255 reg_bo = drm_gem_object_alloc(dev, PAGE_SIZE);
1258 overlay->reg_bo = reg_bo->driver_private;
1260 if (OVERLAY_NONPHYSICAL(dev)) {
1261 ret = i915_gem_object_pin(reg_bo, PAGE_SIZE);
1263 DRM_ERROR("failed to pin overlay register bo\n");
1266 overlay->flip_addr = overlay->reg_bo->gtt_offset;
1268 ret = i915_gem_attach_phys_object(dev, reg_bo,
1269 I915_GEM_PHYS_OVERLAY_REGS);
1271 DRM_ERROR("failed to attach phys overlay regs\n");
1274 overlay->flip_addr = overlay->reg_bo->phys_obj->handle->busaddr;
1277 /* init all values */
1278 overlay->color_key = 0x0101fe;
1279 overlay->brightness = -19;
1280 overlay->contrast = 75;
1281 overlay->saturation = 146;
1283 regs = intel_overlay_map_regs_atomic(overlay);
1287 memset(regs, 0, sizeof(struct overlay_registers));
1288 update_polyphase_filter(regs);
1290 update_reg_attrs(overlay, regs);
1292 intel_overlay_unmap_regs_atomic(overlay);
1294 dev_priv->overlay = overlay;
1295 DRM_INFO("initialized overlay support\n");
1299 drm_gem_object_unreference(reg_bo);
1305 void intel_cleanup_overlay(struct drm_device *dev)
1307 drm_i915_private_t *dev_priv = dev->dev_private;
1309 if (dev_priv->overlay) {
1310 /* The bo's should be free'd by the generic code already.
1311 * Furthermore modesetting teardown happens beforehand so the
1312 * hardware should be off already */
1313 BUG_ON(dev_priv->overlay->active);
1315 kfree(dev_priv->overlay);