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);
226 overlay->hw_wedged = NEEDS_WAIT_FOR_FLIP;
231 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_ON);
232 OUT_RING(overlay->flip_addr | OFC_UPDATE);
233 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
237 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
238 if (overlay->last_flip_req == 0)
241 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
245 overlay->hw_wedged = 0;
246 overlay->last_flip_req = 0;
250 /* overlay needs to be enabled in OCMD reg */
251 static void intel_overlay_continue(struct intel_overlay *overlay,
252 bool load_polyphase_filter)
254 struct drm_device *dev = overlay->dev;
255 drm_i915_private_t *dev_priv = dev->dev_private;
256 u32 flip_addr = overlay->flip_addr;
260 BUG_ON(!overlay->active);
262 if (load_polyphase_filter)
263 flip_addr |= OFC_UPDATE;
265 /* check for underruns */
266 tmp = I915_READ(DOVSTA);
268 DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp);
273 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
277 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
280 static int intel_overlay_wait_flip(struct intel_overlay *overlay)
282 struct drm_device *dev = overlay->dev;
283 drm_i915_private_t *dev_priv = dev->dev_private;
288 if (overlay->last_flip_req != 0) {
289 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
291 overlay->last_flip_req = 0;
293 tmp = I915_READ(ISR);
295 if (!(tmp & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT))
300 /* synchronous slowpath */
301 overlay->hw_wedged = RELEASE_OLD_VID;
304 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
308 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
309 if (overlay->last_flip_req == 0)
312 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
316 overlay->hw_wedged = 0;
317 overlay->last_flip_req = 0;
321 /* overlay needs to be disabled in OCMD reg */
322 static int intel_overlay_off(struct intel_overlay *overlay)
324 u32 flip_addr = overlay->flip_addr;
325 struct drm_device *dev = overlay->dev;
326 drm_i915_private_t *dev_priv = dev->dev_private;
330 BUG_ON(!overlay->active);
332 /* According to intel docs the overlay hw may hang (when switching
333 * off) without loading the filter coeffs. It is however unclear whether
334 * this applies to the disabling of the overlay or to the switching off
335 * of the hw. Do it in both cases */
336 flip_addr |= OFC_UPDATE;
338 /* wait for overlay to go idle */
339 overlay->hw_wedged = SWITCH_OFF_STAGE_1;
344 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
346 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
350 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
351 if (overlay->last_flip_req == 0)
354 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
358 /* turn overlay off */
359 overlay->hw_wedged = SWITCH_OFF_STAGE_2;
364 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
366 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
370 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
371 if (overlay->last_flip_req == 0)
374 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
378 overlay->hw_wedged = 0;
379 overlay->last_flip_req = 0;
383 static void intel_overlay_off_tail(struct intel_overlay *overlay)
385 struct drm_gem_object *obj;
387 /* never have the overlay hw on without showing a frame */
388 BUG_ON(!overlay->vid_bo);
389 obj = overlay->vid_bo->obj;
391 i915_gem_object_unpin(obj);
392 drm_gem_object_unreference(obj);
393 overlay->vid_bo = NULL;
395 overlay->crtc->overlay = NULL;
396 overlay->crtc = NULL;
400 /* recover from an interruption due to a signal
401 * We have to be careful not to repeat work forever an make forward progess. */
402 int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
405 struct drm_device *dev = overlay->dev;
406 drm_i915_private_t *dev_priv = dev->dev_private;
407 struct drm_gem_object *obj;
412 if (overlay->hw_wedged == HW_WEDGED)
415 if (overlay->last_flip_req == 0) {
416 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
417 if (overlay->last_flip_req == 0)
421 ret = i915_do_wait_request(dev, overlay->last_flip_req, interruptible);
425 switch (overlay->hw_wedged) {
426 case RELEASE_OLD_VID:
427 obj = overlay->old_vid_bo->obj;
428 i915_gem_object_unpin(obj);
429 drm_gem_object_unreference(obj);
430 overlay->old_vid_bo = NULL;
432 case SWITCH_OFF_STAGE_1:
433 flip_addr = overlay->flip_addr;
434 flip_addr |= OFC_UPDATE;
436 overlay->hw_wedged = SWITCH_OFF_STAGE_2;
441 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
443 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
447 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
448 if (overlay->last_flip_req == 0)
451 ret = i915_do_wait_request(dev, overlay->last_flip_req,
456 case SWITCH_OFF_STAGE_2:
457 intel_overlay_off_tail(overlay);
460 BUG_ON(overlay->hw_wedged != NEEDS_WAIT_FOR_FLIP);
463 overlay->hw_wedged = 0;
464 overlay->last_flip_req = 0;
468 /* Wait for pending overlay flip and release old frame.
469 * Needs to be called before the overlay register are changed
470 * via intel_overlay_(un)map_regs_atomic */
471 static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
474 struct drm_gem_object *obj;
476 /* only wait if there is actually an old frame to release to
477 * guarantee forward progress */
478 if (!overlay->old_vid_bo)
481 ret = intel_overlay_wait_flip(overlay);
485 obj = overlay->old_vid_bo->obj;
486 i915_gem_object_unpin(obj);
487 drm_gem_object_unreference(obj);
488 overlay->old_vid_bo = NULL;
493 struct put_image_params {
510 static int packed_depth_bytes(u32 format)
512 switch (format & I915_OVERLAY_DEPTH_MASK) {
513 case I915_OVERLAY_YUV422:
515 case I915_OVERLAY_YUV411:
516 /* return 6; not implemented */
522 static int packed_width_bytes(u32 format, short width)
524 switch (format & I915_OVERLAY_DEPTH_MASK) {
525 case I915_OVERLAY_YUV422:
532 static int uv_hsubsampling(u32 format)
534 switch (format & I915_OVERLAY_DEPTH_MASK) {
535 case I915_OVERLAY_YUV422:
536 case I915_OVERLAY_YUV420:
538 case I915_OVERLAY_YUV411:
539 case I915_OVERLAY_YUV410:
546 static int uv_vsubsampling(u32 format)
548 switch (format & I915_OVERLAY_DEPTH_MASK) {
549 case I915_OVERLAY_YUV420:
550 case I915_OVERLAY_YUV410:
552 case I915_OVERLAY_YUV422:
553 case I915_OVERLAY_YUV411:
560 static u32 calc_swidthsw(struct drm_device *dev, u32 offset, u32 width)
562 u32 mask, shift, ret;
570 ret = ((offset + width + mask) >> shift) - (offset >> shift);
577 static const u16 y_static_hcoeffs[N_HORIZ_Y_TAPS * N_PHASES] = {
578 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
579 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
580 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
581 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
582 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
583 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
584 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
585 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
586 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
587 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
588 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
589 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
590 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
591 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
592 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
593 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
594 0xb000, 0x3000, 0x0800, 0x3000, 0xb000};
595 static const u16 uv_static_hcoeffs[N_HORIZ_UV_TAPS * N_PHASES] = {
596 0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
597 0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
598 0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
599 0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
600 0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
601 0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
602 0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
603 0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
604 0x3000, 0x0800, 0x3000};
606 static void update_polyphase_filter(struct overlay_registers *regs)
608 memcpy(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
609 memcpy(regs->UV_HCOEFS, uv_static_hcoeffs, sizeof(uv_static_hcoeffs));
612 static bool update_scaling_factors(struct intel_overlay *overlay,
613 struct overlay_registers *regs,
614 struct put_image_params *params)
616 /* fixed point with a 12 bit shift */
617 u32 xscale, yscale, xscale_UV, yscale_UV;
619 #define FRACT_MASK 0xfff
620 bool scale_changed = false;
621 int uv_hscale = uv_hsubsampling(params->format);
622 int uv_vscale = uv_vsubsampling(params->format);
624 if (params->dst_w > 1)
625 xscale = ((params->src_scan_w - 1) << FP_SHIFT)
628 xscale = 1 << FP_SHIFT;
630 if (params->dst_h > 1)
631 yscale = ((params->src_scan_h - 1) << FP_SHIFT)
634 yscale = 1 << FP_SHIFT;
636 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
637 xscale_UV = xscale/uv_hscale;
638 yscale_UV = yscale/uv_vscale;
639 /* make the Y scale to UV scale ratio an exact multiply */
640 xscale = xscale_UV * uv_hscale;
641 yscale = yscale_UV * uv_vscale;
647 if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
648 scale_changed = true;
649 overlay->old_xscale = xscale;
650 overlay->old_yscale = yscale;
652 regs->YRGBSCALE = ((yscale & FRACT_MASK) << 20)
653 | ((xscale >> FP_SHIFT) << 16)
654 | ((xscale & FRACT_MASK) << 3);
655 regs->UVSCALE = ((yscale_UV & FRACT_MASK) << 20)
656 | ((xscale_UV >> FP_SHIFT) << 16)
657 | ((xscale_UV & FRACT_MASK) << 3);
658 regs->UVSCALEV = ((yscale >> FP_SHIFT) << 16)
659 | ((yscale_UV >> FP_SHIFT) << 0);
662 update_polyphase_filter(regs);
664 return scale_changed;
667 static void update_colorkey(struct intel_overlay *overlay,
668 struct overlay_registers *regs)
670 u32 key = overlay->color_key;
671 switch (overlay->crtc->base.fb->bits_per_pixel) {
674 regs->DCLRKM = CLK_RGB8I_MASK | DST_KEY_ENABLE;
676 if (overlay->crtc->base.fb->depth == 15) {
677 regs->DCLRKV = RGB15_TO_COLORKEY(key);
678 regs->DCLRKM = CLK_RGB15_MASK | DST_KEY_ENABLE;
680 regs->DCLRKV = RGB16_TO_COLORKEY(key);
681 regs->DCLRKM = CLK_RGB16_MASK | DST_KEY_ENABLE;
686 regs->DCLRKM = CLK_RGB24_MASK | DST_KEY_ENABLE;
690 static u32 overlay_cmd_reg(struct put_image_params *params)
692 u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
694 if (params->format & I915_OVERLAY_YUV_PLANAR) {
695 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
696 case I915_OVERLAY_YUV422:
697 cmd |= OCMD_YUV_422_PLANAR;
699 case I915_OVERLAY_YUV420:
700 cmd |= OCMD_YUV_420_PLANAR;
702 case I915_OVERLAY_YUV411:
703 case I915_OVERLAY_YUV410:
704 cmd |= OCMD_YUV_410_PLANAR;
707 } else { /* YUV packed */
708 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
709 case I915_OVERLAY_YUV422:
710 cmd |= OCMD_YUV_422_PACKED;
712 case I915_OVERLAY_YUV411:
713 cmd |= OCMD_YUV_411_PACKED;
717 switch (params->format & I915_OVERLAY_SWAP_MASK) {
718 case I915_OVERLAY_NO_SWAP:
720 case I915_OVERLAY_UV_SWAP:
723 case I915_OVERLAY_Y_SWAP:
726 case I915_OVERLAY_Y_AND_UV_SWAP:
727 cmd |= OCMD_Y_AND_UV_SWAP;
735 int intel_overlay_do_put_image(struct intel_overlay *overlay,
736 struct drm_gem_object *new_bo,
737 struct put_image_params *params)
740 struct overlay_registers *regs;
741 bool scale_changed = false;
742 struct drm_i915_gem_object *bo_priv = new_bo->driver_private;
743 struct drm_device *dev = overlay->dev;
745 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
746 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
749 ret = intel_overlay_release_old_vid(overlay);
753 ret = i915_gem_object_pin(new_bo, PAGE_SIZE);
757 ret = i915_gem_object_set_to_gtt_domain(new_bo, 0);
761 if (!overlay->active) {
762 regs = intel_overlay_map_regs_atomic(overlay);
767 regs->OCONFIG = OCONF_CC_OUT_8BIT;
768 if (IS_I965GM(overlay->dev))
769 regs->OCONFIG |= OCONF_CSC_MODE_BT709;
770 regs->OCONFIG |= overlay->crtc->pipe == 0 ?
771 OCONF_PIPE_A : OCONF_PIPE_B;
772 intel_overlay_unmap_regs_atomic(overlay);
774 ret = intel_overlay_on(overlay);
779 regs = intel_overlay_map_regs_atomic(overlay);
785 regs->DWINPOS = (params->dst_y << 16) | params->dst_x;
786 regs->DWINSZ = (params->dst_h << 16) | params->dst_w;
788 if (params->format & I915_OVERLAY_YUV_PACKED)
789 tmp_width = packed_width_bytes(params->format, params->src_w);
791 tmp_width = params->src_w;
793 regs->SWIDTH = params->src_w;
794 regs->SWIDTHSW = calc_swidthsw(overlay->dev,
795 params->offset_Y, tmp_width);
796 regs->SHEIGHT = params->src_h;
797 regs->OBUF_0Y = bo_priv->gtt_offset + params-> offset_Y;
798 regs->OSTRIDE = params->stride_Y;
800 if (params->format & I915_OVERLAY_YUV_PLANAR) {
801 int uv_hscale = uv_hsubsampling(params->format);
802 int uv_vscale = uv_vsubsampling(params->format);
804 regs->SWIDTH |= (params->src_w/uv_hscale) << 16;
805 tmp_U = calc_swidthsw(overlay->dev, params->offset_U,
806 params->src_w/uv_hscale);
807 tmp_V = calc_swidthsw(overlay->dev, params->offset_V,
808 params->src_w/uv_hscale);
809 regs->SWIDTHSW |= max_t(u32, tmp_U, tmp_V) << 16;
810 regs->SHEIGHT |= (params->src_h/uv_vscale) << 16;
811 regs->OBUF_0U = bo_priv->gtt_offset + params->offset_U;
812 regs->OBUF_0V = bo_priv->gtt_offset + params->offset_V;
813 regs->OSTRIDE |= params->stride_UV << 16;
816 scale_changed = update_scaling_factors(overlay, regs, params);
818 update_colorkey(overlay, regs);
820 regs->OCMD = overlay_cmd_reg(params);
822 intel_overlay_unmap_regs_atomic(overlay);
824 intel_overlay_continue(overlay, scale_changed);
826 overlay->old_vid_bo = overlay->vid_bo;
827 overlay->vid_bo = new_bo->driver_private;
832 i915_gem_object_unpin(new_bo);
836 int intel_overlay_switch_off(struct intel_overlay *overlay)
839 struct overlay_registers *regs;
840 struct drm_device *dev = overlay->dev;
842 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
843 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
845 if (!overlay->active)
848 if (overlay->hw_wedged)
851 ret = intel_overlay_release_old_vid(overlay);
855 regs = intel_overlay_map_regs_atomic(overlay);
857 intel_overlay_unmap_regs_atomic(overlay);
859 ret = intel_overlay_off(overlay);
863 intel_overlay_off_tail(overlay);
868 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
869 struct intel_crtc *crtc)
871 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
873 int pipeconf_reg = (crtc->pipe == 0) ? PIPEACONF : PIPEBCONF;
875 if (!crtc->base.enabled || crtc->dpms_mode != DRM_MODE_DPMS_ON)
878 pipeconf = I915_READ(pipeconf_reg);
880 /* can't use the overlay with double wide pipe */
881 if (!IS_I965G(overlay->dev) && pipeconf & PIPEACONF_DOUBLE_WIDE)
887 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
889 struct drm_device *dev = overlay->dev;
890 drm_i915_private_t *dev_priv = dev->dev_private;
892 u32 pfit_control = I915_READ(PFIT_CONTROL);
894 /* XXX: This is not the same logic as in the xorg driver, but more in
895 * line with the intel documentation for the i965 */
896 if (!IS_I965G(dev) && (pfit_control & VERT_AUTO_SCALE)) {
897 ratio = I915_READ(PFIT_AUTO_RATIOS) >> PFIT_VERT_SCALE_SHIFT;
898 } else { /* on i965 use the PGM reg to read out the autoscaler values */
899 ratio = I915_READ(PFIT_PGM_RATIOS);
901 ratio >>= PFIT_VERT_SCALE_SHIFT_965;
903 ratio >>= PFIT_VERT_SCALE_SHIFT;
906 overlay->pfit_vscale_ratio = ratio;
909 static int check_overlay_dst(struct intel_overlay *overlay,
910 struct drm_intel_overlay_put_image *rec)
912 struct drm_display_mode *mode = &overlay->crtc->base.mode;
914 if ((rec->dst_x < mode->crtc_hdisplay)
915 && (rec->dst_x + rec->dst_width
916 <= mode->crtc_hdisplay)
917 && (rec->dst_y < mode->crtc_vdisplay)
918 && (rec->dst_y + rec->dst_height
919 <= mode->crtc_vdisplay))
925 static int check_overlay_scaling(struct put_image_params *rec)
929 /* downscaling limit is 8.0 */
930 tmp = ((rec->src_scan_h << 16) / rec->dst_h) >> 16;
933 tmp = ((rec->src_scan_w << 16) / rec->dst_w) >> 16;
940 static int check_overlay_src(struct drm_device *dev,
941 struct drm_intel_overlay_put_image *rec,
942 struct drm_gem_object *new_bo)
946 int uv_hscale = uv_hsubsampling(rec->flags);
947 int uv_vscale = uv_vsubsampling(rec->flags);
950 /* check src dimensions */
951 if (IS_845G(dev) || IS_I830(dev)) {
952 if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY
953 || rec->src_width > IMAGE_MAX_WIDTH_LEGACY)
956 if (rec->src_height > IMAGE_MAX_HEIGHT
957 || rec->src_width > IMAGE_MAX_WIDTH)
960 /* better safe than sorry, use 4 as the maximal subsampling ratio */
961 if (rec->src_height < N_VERT_Y_TAPS*4
962 || rec->src_width < N_HORIZ_Y_TAPS*4)
965 /* check alingment constrains */
966 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
967 case I915_OVERLAY_RGB:
968 /* not implemented */
970 case I915_OVERLAY_YUV_PACKED:
971 depth = packed_depth_bytes(rec->flags);
976 /* ignore UV planes */
980 /* check pixel alignment */
981 if (rec->offset_Y % depth)
984 case I915_OVERLAY_YUV_PLANAR:
985 if (uv_vscale < 0 || uv_hscale < 0)
987 /* no offset restrictions for planar formats */
993 if (rec->src_width % uv_hscale)
996 /* stride checking */
999 if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
1001 if (IS_I965G(dev) && rec->stride_Y < 512)
1004 tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
1006 if (rec->stride_Y > tmp*1024 || rec->stride_UV > 2*1024)
1009 /* check buffer dimensions */
1010 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1011 case I915_OVERLAY_RGB:
1012 case I915_OVERLAY_YUV_PACKED:
1013 /* always 4 Y values per depth pixels */
1014 if (packed_width_bytes(rec->flags, rec->src_width)
1018 tmp = rec->stride_Y*rec->src_height;
1019 if (rec->offset_Y + tmp > new_bo->size)
1022 case I915_OVERLAY_YUV_PLANAR:
1023 if (rec->src_width > rec->stride_Y)
1025 if (rec->src_width/uv_hscale > rec->stride_UV)
1028 tmp = rec->stride_Y*rec->src_height;
1029 if (rec->offset_Y + tmp > new_bo->size)
1031 tmp = rec->stride_UV*rec->src_height;
1033 if (rec->offset_U + tmp > new_bo->size
1034 || rec->offset_V + tmp > new_bo->size)
1042 int intel_overlay_put_image(struct drm_device *dev, void *data,
1043 struct drm_file *file_priv)
1045 struct drm_intel_overlay_put_image *put_image_rec = data;
1046 drm_i915_private_t *dev_priv = dev->dev_private;
1047 struct intel_overlay *overlay;
1048 struct drm_mode_object *drmmode_obj;
1049 struct intel_crtc *crtc;
1050 struct drm_gem_object *new_bo;
1051 struct put_image_params *params;
1055 DRM_ERROR("called with no initialization\n");
1059 overlay = dev_priv->overlay;
1061 DRM_DEBUG("userspace bug: no overlay\n");
1065 if (!(put_image_rec->flags & I915_OVERLAY_ENABLE)) {
1066 mutex_lock(&dev->mode_config.mutex);
1067 mutex_lock(&dev->struct_mutex);
1069 ret = intel_overlay_switch_off(overlay);
1071 mutex_unlock(&dev->struct_mutex);
1072 mutex_unlock(&dev->mode_config.mutex);
1077 params = kmalloc(sizeof(struct put_image_params), GFP_KERNEL);
1081 drmmode_obj = drm_mode_object_find(dev, put_image_rec->crtc_id,
1082 DRM_MODE_OBJECT_CRTC);
1085 crtc = to_intel_crtc(obj_to_crtc(drmmode_obj));
1087 new_bo = drm_gem_object_lookup(dev, file_priv,
1088 put_image_rec->bo_handle);
1092 mutex_lock(&dev->mode_config.mutex);
1093 mutex_lock(&dev->struct_mutex);
1095 if (overlay->hw_wedged) {
1096 ret = intel_overlay_recover_from_interrupt(overlay, 1);
1101 if (overlay->crtc != crtc) {
1102 struct drm_display_mode *mode = &crtc->base.mode;
1103 ret = intel_overlay_switch_off(overlay);
1107 ret = check_overlay_possible_on_crtc(overlay, crtc);
1111 overlay->crtc = crtc;
1112 crtc->overlay = overlay;
1114 if (intel_panel_fitter_pipe(dev) == crtc->pipe
1115 /* and line to wide, i.e. one-line-mode */
1116 && mode->hdisplay > 1024) {
1117 overlay->pfit_active = 1;
1118 update_pfit_vscale_ratio(overlay);
1120 overlay->pfit_active = 0;
1123 ret = check_overlay_dst(overlay, put_image_rec);
1127 if (overlay->pfit_active) {
1128 params->dst_y = ((((u32)put_image_rec->dst_y) << 12) /
1129 overlay->pfit_vscale_ratio);
1130 /* shifting right rounds downwards, so add 1 */
1131 params->dst_h = ((((u32)put_image_rec->dst_height) << 12) /
1132 overlay->pfit_vscale_ratio) + 1;
1134 params->dst_y = put_image_rec->dst_y;
1135 params->dst_h = put_image_rec->dst_height;
1137 params->dst_x = put_image_rec->dst_x;
1138 params->dst_w = put_image_rec->dst_width;
1140 params->src_w = put_image_rec->src_width;
1141 params->src_h = put_image_rec->src_height;
1142 params->src_scan_w = put_image_rec->src_scan_width;
1143 params->src_scan_h = put_image_rec->src_scan_height;
1144 if (params->src_scan_h > params->src_h
1145 || params->src_scan_w > params->src_w) {
1150 ret = check_overlay_src(dev, put_image_rec, new_bo);
1153 params->format = put_image_rec->flags & ~I915_OVERLAY_FLAGS_MASK;
1154 params->stride_Y = put_image_rec->stride_Y;
1155 params->stride_UV = put_image_rec->stride_UV;
1156 params->offset_Y = put_image_rec->offset_Y;
1157 params->offset_U = put_image_rec->offset_U;
1158 params->offset_V = put_image_rec->offset_V;
1160 /* Check scaling after src size to prevent a divide-by-zero. */
1161 ret = check_overlay_scaling(params);
1165 ret = intel_overlay_do_put_image(overlay, new_bo, params);
1169 mutex_unlock(&dev->struct_mutex);
1170 mutex_unlock(&dev->mode_config.mutex);
1177 mutex_unlock(&dev->struct_mutex);
1178 mutex_unlock(&dev->mode_config.mutex);
1179 drm_gem_object_unreference(new_bo);
1185 static void update_reg_attrs(struct intel_overlay *overlay,
1186 struct overlay_registers *regs)
1188 regs->OCLRC0 = (overlay->contrast << 18) | (overlay->brightness & 0xff);
1189 regs->OCLRC1 = overlay->saturation;
1192 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
1196 if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
1199 for (i = 0; i < 3; i++) {
1200 if (((gamma1 >> i * 8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
1207 static bool check_gamma5_errata(u32 gamma5)
1211 for (i = 0; i < 3; i++) {
1212 if (((gamma5 >> i*8) & 0xff) == 0x80)
1219 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
1221 if (!check_gamma_bounds(0, attrs->gamma0)
1222 || !check_gamma_bounds(attrs->gamma0, attrs->gamma1)
1223 || !check_gamma_bounds(attrs->gamma1, attrs->gamma2)
1224 || !check_gamma_bounds(attrs->gamma2, attrs->gamma3)
1225 || !check_gamma_bounds(attrs->gamma3, attrs->gamma4)
1226 || !check_gamma_bounds(attrs->gamma4, attrs->gamma5)
1227 || !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
1229 if (!check_gamma5_errata(attrs->gamma5))
1234 int intel_overlay_attrs(struct drm_device *dev, void *data,
1235 struct drm_file *file_priv)
1237 struct drm_intel_overlay_attrs *attrs = data;
1238 drm_i915_private_t *dev_priv = dev->dev_private;
1239 struct intel_overlay *overlay;
1240 struct overlay_registers *regs;
1244 DRM_ERROR("called with no initialization\n");
1248 overlay = dev_priv->overlay;
1250 DRM_DEBUG("userspace bug: no overlay\n");
1254 mutex_lock(&dev->mode_config.mutex);
1255 mutex_lock(&dev->struct_mutex);
1257 if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
1258 attrs->color_key = overlay->color_key;
1259 attrs->brightness = overlay->brightness;
1260 attrs->contrast = overlay->contrast;
1261 attrs->saturation = overlay->saturation;
1264 attrs->gamma0 = I915_READ(OGAMC0);
1265 attrs->gamma1 = I915_READ(OGAMC1);
1266 attrs->gamma2 = I915_READ(OGAMC2);
1267 attrs->gamma3 = I915_READ(OGAMC3);
1268 attrs->gamma4 = I915_READ(OGAMC4);
1269 attrs->gamma5 = I915_READ(OGAMC5);
1273 overlay->color_key = attrs->color_key;
1274 if (attrs->brightness >= -128 && attrs->brightness <= 127) {
1275 overlay->brightness = attrs->brightness;
1280 if (attrs->contrast <= 255) {
1281 overlay->contrast = attrs->contrast;
1286 if (attrs->saturation <= 1023) {
1287 overlay->saturation = attrs->saturation;
1293 regs = intel_overlay_map_regs_atomic(overlay);
1299 update_reg_attrs(overlay, regs);
1301 intel_overlay_unmap_regs_atomic(overlay);
1303 if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
1304 if (!IS_I9XX(dev)) {
1309 if (overlay->active) {
1314 ret = check_gamma(attrs);
1318 I915_WRITE(OGAMC0, attrs->gamma0);
1319 I915_WRITE(OGAMC1, attrs->gamma1);
1320 I915_WRITE(OGAMC2, attrs->gamma2);
1321 I915_WRITE(OGAMC3, attrs->gamma3);
1322 I915_WRITE(OGAMC4, attrs->gamma4);
1323 I915_WRITE(OGAMC5, attrs->gamma5);
1329 mutex_unlock(&dev->struct_mutex);
1330 mutex_unlock(&dev->mode_config.mutex);
1335 void intel_setup_overlay(struct drm_device *dev)
1337 drm_i915_private_t *dev_priv = dev->dev_private;
1338 struct intel_overlay *overlay;
1339 struct drm_gem_object *reg_bo;
1340 struct overlay_registers *regs;
1343 if (!OVERLAY_EXISTS(dev))
1346 overlay = kzalloc(sizeof(struct intel_overlay), GFP_KERNEL);
1351 reg_bo = drm_gem_object_alloc(dev, PAGE_SIZE);
1354 overlay->reg_bo = reg_bo->driver_private;
1356 if (OVERLAY_NONPHYSICAL(dev)) {
1357 ret = i915_gem_object_pin(reg_bo, PAGE_SIZE);
1359 DRM_ERROR("failed to pin overlay register bo\n");
1362 overlay->flip_addr = overlay->reg_bo->gtt_offset;
1364 ret = i915_gem_attach_phys_object(dev, reg_bo,
1365 I915_GEM_PHYS_OVERLAY_REGS);
1367 DRM_ERROR("failed to attach phys overlay regs\n");
1370 overlay->flip_addr = overlay->reg_bo->phys_obj->handle->busaddr;
1373 /* init all values */
1374 overlay->color_key = 0x0101fe;
1375 overlay->brightness = -19;
1376 overlay->contrast = 75;
1377 overlay->saturation = 146;
1379 regs = intel_overlay_map_regs_atomic(overlay);
1383 memset(regs, 0, sizeof(struct overlay_registers));
1384 update_polyphase_filter(regs);
1386 update_reg_attrs(overlay, regs);
1388 intel_overlay_unmap_regs_atomic(overlay);
1390 dev_priv->overlay = overlay;
1391 DRM_INFO("initialized overlay support\n");
1395 drm_gem_object_unreference(reg_bo);
1401 void intel_cleanup_overlay(struct drm_device *dev)
1403 drm_i915_private_t *dev_priv = dev->dev_private;
1405 if (dev_priv->overlay) {
1406 /* The bo's should be free'd by the generic code already.
1407 * Furthermore modesetting teardown happens beforehand so the
1408 * hardware should be off already */
1409 BUG_ON(dev_priv->overlay->active);
1411 kfree(dev_priv->overlay);