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_IRONLAKE(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 if (OVERLAY_NONPHYSICAL(overlay->dev))
203 io_mapping_unmap_atomic(overlay->virt_addr);
205 overlay->virt_addr = NULL;
210 /* overlay needs to be disable in OCMD reg */
211 static int intel_overlay_on(struct intel_overlay *overlay)
213 struct drm_device *dev = overlay->dev;
214 drm_i915_private_t *dev_priv = dev->dev_private;
218 BUG_ON(overlay->active);
221 overlay->hw_wedged = NEEDS_WAIT_FOR_FLIP;
226 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_ON);
227 OUT_RING(overlay->flip_addr | OFC_UPDATE);
228 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
232 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
233 if (overlay->last_flip_req == 0)
236 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
240 overlay->hw_wedged = 0;
241 overlay->last_flip_req = 0;
245 /* overlay needs to be enabled in OCMD reg */
246 static void intel_overlay_continue(struct intel_overlay *overlay,
247 bool load_polyphase_filter)
249 struct drm_device *dev = overlay->dev;
250 drm_i915_private_t *dev_priv = dev->dev_private;
251 u32 flip_addr = overlay->flip_addr;
255 BUG_ON(!overlay->active);
257 if (load_polyphase_filter)
258 flip_addr |= OFC_UPDATE;
260 /* check for underruns */
261 tmp = I915_READ(DOVSTA);
263 DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp);
268 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
272 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
275 static int intel_overlay_wait_flip(struct intel_overlay *overlay)
277 struct drm_device *dev = overlay->dev;
278 drm_i915_private_t *dev_priv = dev->dev_private;
283 if (overlay->last_flip_req != 0) {
284 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
286 overlay->last_flip_req = 0;
288 tmp = I915_READ(ISR);
290 if (!(tmp & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT))
295 /* synchronous slowpath */
296 overlay->hw_wedged = RELEASE_OLD_VID;
299 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
303 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
304 if (overlay->last_flip_req == 0)
307 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
311 overlay->hw_wedged = 0;
312 overlay->last_flip_req = 0;
316 /* overlay needs to be disabled in OCMD reg */
317 static int intel_overlay_off(struct intel_overlay *overlay)
319 u32 flip_addr = overlay->flip_addr;
320 struct drm_device *dev = overlay->dev;
321 drm_i915_private_t *dev_priv = dev->dev_private;
325 BUG_ON(!overlay->active);
327 /* According to intel docs the overlay hw may hang (when switching
328 * off) without loading the filter coeffs. It is however unclear whether
329 * this applies to the disabling of the overlay or to the switching off
330 * of the hw. Do it in both cases */
331 flip_addr |= OFC_UPDATE;
333 /* wait for overlay to go idle */
334 overlay->hw_wedged = SWITCH_OFF_STAGE_1;
339 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
341 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
345 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
346 if (overlay->last_flip_req == 0)
349 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
353 /* turn overlay off */
354 overlay->hw_wedged = SWITCH_OFF_STAGE_2;
359 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
361 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
365 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
366 if (overlay->last_flip_req == 0)
369 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
373 overlay->hw_wedged = 0;
374 overlay->last_flip_req = 0;
378 static void intel_overlay_off_tail(struct intel_overlay *overlay)
380 struct drm_gem_object *obj;
382 /* never have the overlay hw on without showing a frame */
383 BUG_ON(!overlay->vid_bo);
384 obj = overlay->vid_bo->obj;
386 i915_gem_object_unpin(obj);
387 drm_gem_object_unreference(obj);
388 overlay->vid_bo = NULL;
390 overlay->crtc->overlay = NULL;
391 overlay->crtc = NULL;
395 /* recover from an interruption due to a signal
396 * We have to be careful not to repeat work forever an make forward progess. */
397 int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
400 struct drm_device *dev = overlay->dev;
401 drm_i915_private_t *dev_priv = dev->dev_private;
402 struct drm_gem_object *obj;
407 if (overlay->hw_wedged == HW_WEDGED)
410 if (overlay->last_flip_req == 0) {
411 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
412 if (overlay->last_flip_req == 0)
416 ret = i915_do_wait_request(dev, overlay->last_flip_req, interruptible);
420 switch (overlay->hw_wedged) {
421 case RELEASE_OLD_VID:
422 obj = overlay->old_vid_bo->obj;
423 i915_gem_object_unpin(obj);
424 drm_gem_object_unreference(obj);
425 overlay->old_vid_bo = NULL;
427 case SWITCH_OFF_STAGE_1:
428 flip_addr = overlay->flip_addr;
429 flip_addr |= OFC_UPDATE;
431 overlay->hw_wedged = SWITCH_OFF_STAGE_2;
436 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
438 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
442 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
443 if (overlay->last_flip_req == 0)
446 ret = i915_do_wait_request(dev, overlay->last_flip_req,
451 case SWITCH_OFF_STAGE_2:
452 intel_overlay_off_tail(overlay);
455 BUG_ON(overlay->hw_wedged != NEEDS_WAIT_FOR_FLIP);
458 overlay->hw_wedged = 0;
459 overlay->last_flip_req = 0;
463 /* Wait for pending overlay flip and release old frame.
464 * Needs to be called before the overlay register are changed
465 * via intel_overlay_(un)map_regs_atomic */
466 static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
469 struct drm_gem_object *obj;
471 /* only wait if there is actually an old frame to release to
472 * guarantee forward progress */
473 if (!overlay->old_vid_bo)
476 ret = intel_overlay_wait_flip(overlay);
480 obj = overlay->old_vid_bo->obj;
481 i915_gem_object_unpin(obj);
482 drm_gem_object_unreference(obj);
483 overlay->old_vid_bo = NULL;
488 struct put_image_params {
505 static int packed_depth_bytes(u32 format)
507 switch (format & I915_OVERLAY_DEPTH_MASK) {
508 case I915_OVERLAY_YUV422:
510 case I915_OVERLAY_YUV411:
511 /* return 6; not implemented */
517 static int packed_width_bytes(u32 format, short width)
519 switch (format & I915_OVERLAY_DEPTH_MASK) {
520 case I915_OVERLAY_YUV422:
527 static int uv_hsubsampling(u32 format)
529 switch (format & I915_OVERLAY_DEPTH_MASK) {
530 case I915_OVERLAY_YUV422:
531 case I915_OVERLAY_YUV420:
533 case I915_OVERLAY_YUV411:
534 case I915_OVERLAY_YUV410:
541 static int uv_vsubsampling(u32 format)
543 switch (format & I915_OVERLAY_DEPTH_MASK) {
544 case I915_OVERLAY_YUV420:
545 case I915_OVERLAY_YUV410:
547 case I915_OVERLAY_YUV422:
548 case I915_OVERLAY_YUV411:
555 static u32 calc_swidthsw(struct drm_device *dev, u32 offset, u32 width)
557 u32 mask, shift, ret;
565 ret = ((offset + width + mask) >> shift) - (offset >> shift);
572 static const u16 y_static_hcoeffs[N_HORIZ_Y_TAPS * N_PHASES] = {
573 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
574 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
575 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
576 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
577 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
578 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
579 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
580 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
581 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
582 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
583 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
584 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
585 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
586 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
587 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
588 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
589 0xb000, 0x3000, 0x0800, 0x3000, 0xb000};
590 static const u16 uv_static_hcoeffs[N_HORIZ_UV_TAPS * N_PHASES] = {
591 0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
592 0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
593 0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
594 0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
595 0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
596 0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
597 0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
598 0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
599 0x3000, 0x0800, 0x3000};
601 static void update_polyphase_filter(struct overlay_registers *regs)
603 memcpy(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
604 memcpy(regs->UV_HCOEFS, uv_static_hcoeffs, sizeof(uv_static_hcoeffs));
607 static bool update_scaling_factors(struct intel_overlay *overlay,
608 struct overlay_registers *regs,
609 struct put_image_params *params)
611 /* fixed point with a 12 bit shift */
612 u32 xscale, yscale, xscale_UV, yscale_UV;
614 #define FRACT_MASK 0xfff
615 bool scale_changed = false;
616 int uv_hscale = uv_hsubsampling(params->format);
617 int uv_vscale = uv_vsubsampling(params->format);
619 if (params->dst_w > 1)
620 xscale = ((params->src_scan_w - 1) << FP_SHIFT)
623 xscale = 1 << FP_SHIFT;
625 if (params->dst_h > 1)
626 yscale = ((params->src_scan_h - 1) << FP_SHIFT)
629 yscale = 1 << FP_SHIFT;
631 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
632 xscale_UV = xscale/uv_hscale;
633 yscale_UV = yscale/uv_vscale;
634 /* make the Y scale to UV scale ratio an exact multiply */
635 xscale = xscale_UV * uv_hscale;
636 yscale = yscale_UV * uv_vscale;
642 if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
643 scale_changed = true;
644 overlay->old_xscale = xscale;
645 overlay->old_yscale = yscale;
647 regs->YRGBSCALE = ((yscale & FRACT_MASK) << 20)
648 | ((xscale >> FP_SHIFT) << 16)
649 | ((xscale & FRACT_MASK) << 3);
650 regs->UVSCALE = ((yscale_UV & FRACT_MASK) << 20)
651 | ((xscale_UV >> FP_SHIFT) << 16)
652 | ((xscale_UV & FRACT_MASK) << 3);
653 regs->UVSCALEV = ((yscale >> FP_SHIFT) << 16)
654 | ((yscale_UV >> FP_SHIFT) << 0);
657 update_polyphase_filter(regs);
659 return scale_changed;
662 static void update_colorkey(struct intel_overlay *overlay,
663 struct overlay_registers *regs)
665 u32 key = overlay->color_key;
666 switch (overlay->crtc->base.fb->bits_per_pixel) {
669 regs->DCLRKM = CLK_RGB8I_MASK | DST_KEY_ENABLE;
671 if (overlay->crtc->base.fb->depth == 15) {
672 regs->DCLRKV = RGB15_TO_COLORKEY(key);
673 regs->DCLRKM = CLK_RGB15_MASK | DST_KEY_ENABLE;
675 regs->DCLRKV = RGB16_TO_COLORKEY(key);
676 regs->DCLRKM = CLK_RGB16_MASK | DST_KEY_ENABLE;
681 regs->DCLRKM = CLK_RGB24_MASK | DST_KEY_ENABLE;
685 static u32 overlay_cmd_reg(struct put_image_params *params)
687 u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
689 if (params->format & I915_OVERLAY_YUV_PLANAR) {
690 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
691 case I915_OVERLAY_YUV422:
692 cmd |= OCMD_YUV_422_PLANAR;
694 case I915_OVERLAY_YUV420:
695 cmd |= OCMD_YUV_420_PLANAR;
697 case I915_OVERLAY_YUV411:
698 case I915_OVERLAY_YUV410:
699 cmd |= OCMD_YUV_410_PLANAR;
702 } else { /* YUV packed */
703 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
704 case I915_OVERLAY_YUV422:
705 cmd |= OCMD_YUV_422_PACKED;
707 case I915_OVERLAY_YUV411:
708 cmd |= OCMD_YUV_411_PACKED;
712 switch (params->format & I915_OVERLAY_SWAP_MASK) {
713 case I915_OVERLAY_NO_SWAP:
715 case I915_OVERLAY_UV_SWAP:
718 case I915_OVERLAY_Y_SWAP:
721 case I915_OVERLAY_Y_AND_UV_SWAP:
722 cmd |= OCMD_Y_AND_UV_SWAP;
730 int intel_overlay_do_put_image(struct intel_overlay *overlay,
731 struct drm_gem_object *new_bo,
732 struct put_image_params *params)
735 struct overlay_registers *regs;
736 bool scale_changed = false;
737 struct drm_i915_gem_object *bo_priv = new_bo->driver_private;
738 struct drm_device *dev = overlay->dev;
740 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
741 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
744 ret = intel_overlay_release_old_vid(overlay);
748 ret = i915_gem_object_pin(new_bo, PAGE_SIZE);
752 ret = i915_gem_object_set_to_gtt_domain(new_bo, 0);
756 if (!overlay->active) {
757 regs = intel_overlay_map_regs_atomic(overlay);
762 regs->OCONFIG = OCONF_CC_OUT_8BIT;
763 if (IS_I965GM(overlay->dev))
764 regs->OCONFIG |= OCONF_CSC_MODE_BT709;
765 regs->OCONFIG |= overlay->crtc->pipe == 0 ?
766 OCONF_PIPE_A : OCONF_PIPE_B;
767 intel_overlay_unmap_regs_atomic(overlay);
769 ret = intel_overlay_on(overlay);
774 regs = intel_overlay_map_regs_atomic(overlay);
780 regs->DWINPOS = (params->dst_y << 16) | params->dst_x;
781 regs->DWINSZ = (params->dst_h << 16) | params->dst_w;
783 if (params->format & I915_OVERLAY_YUV_PACKED)
784 tmp_width = packed_width_bytes(params->format, params->src_w);
786 tmp_width = params->src_w;
788 regs->SWIDTH = params->src_w;
789 regs->SWIDTHSW = calc_swidthsw(overlay->dev,
790 params->offset_Y, tmp_width);
791 regs->SHEIGHT = params->src_h;
792 regs->OBUF_0Y = bo_priv->gtt_offset + params-> offset_Y;
793 regs->OSTRIDE = params->stride_Y;
795 if (params->format & I915_OVERLAY_YUV_PLANAR) {
796 int uv_hscale = uv_hsubsampling(params->format);
797 int uv_vscale = uv_vsubsampling(params->format);
799 regs->SWIDTH |= (params->src_w/uv_hscale) << 16;
800 tmp_U = calc_swidthsw(overlay->dev, params->offset_U,
801 params->src_w/uv_hscale);
802 tmp_V = calc_swidthsw(overlay->dev, params->offset_V,
803 params->src_w/uv_hscale);
804 regs->SWIDTHSW |= max_t(u32, tmp_U, tmp_V) << 16;
805 regs->SHEIGHT |= (params->src_h/uv_vscale) << 16;
806 regs->OBUF_0U = bo_priv->gtt_offset + params->offset_U;
807 regs->OBUF_0V = bo_priv->gtt_offset + params->offset_V;
808 regs->OSTRIDE |= params->stride_UV << 16;
811 scale_changed = update_scaling_factors(overlay, regs, params);
813 update_colorkey(overlay, regs);
815 regs->OCMD = overlay_cmd_reg(params);
817 intel_overlay_unmap_regs_atomic(overlay);
819 intel_overlay_continue(overlay, scale_changed);
821 overlay->old_vid_bo = overlay->vid_bo;
822 overlay->vid_bo = new_bo->driver_private;
827 i915_gem_object_unpin(new_bo);
831 int intel_overlay_switch_off(struct intel_overlay *overlay)
834 struct overlay_registers *regs;
835 struct drm_device *dev = overlay->dev;
837 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
838 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
840 if (overlay->hw_wedged) {
841 ret = intel_overlay_recover_from_interrupt(overlay, 1);
846 if (!overlay->active)
849 ret = intel_overlay_release_old_vid(overlay);
853 regs = intel_overlay_map_regs_atomic(overlay);
855 intel_overlay_unmap_regs_atomic(overlay);
857 ret = intel_overlay_off(overlay);
861 intel_overlay_off_tail(overlay);
866 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
867 struct intel_crtc *crtc)
869 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
871 int pipeconf_reg = (crtc->pipe == 0) ? PIPEACONF : PIPEBCONF;
873 if (!crtc->base.enabled || crtc->dpms_mode != DRM_MODE_DPMS_ON)
876 pipeconf = I915_READ(pipeconf_reg);
878 /* can't use the overlay with double wide pipe */
879 if (!IS_I965G(overlay->dev) && pipeconf & PIPEACONF_DOUBLE_WIDE)
885 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
887 struct drm_device *dev = overlay->dev;
888 drm_i915_private_t *dev_priv = dev->dev_private;
890 u32 pfit_control = I915_READ(PFIT_CONTROL);
892 /* XXX: This is not the same logic as in the xorg driver, but more in
893 * line with the intel documentation for the i965 */
894 if (!IS_I965G(dev) && (pfit_control & VERT_AUTO_SCALE)) {
895 ratio = I915_READ(PFIT_AUTO_RATIOS) >> PFIT_VERT_SCALE_SHIFT;
896 } else { /* on i965 use the PGM reg to read out the autoscaler values */
897 ratio = I915_READ(PFIT_PGM_RATIOS);
899 ratio >>= PFIT_VERT_SCALE_SHIFT_965;
901 ratio >>= PFIT_VERT_SCALE_SHIFT;
904 overlay->pfit_vscale_ratio = ratio;
907 static int check_overlay_dst(struct intel_overlay *overlay,
908 struct drm_intel_overlay_put_image *rec)
910 struct drm_display_mode *mode = &overlay->crtc->base.mode;
912 if ((rec->dst_x < mode->crtc_hdisplay)
913 && (rec->dst_x + rec->dst_width
914 <= mode->crtc_hdisplay)
915 && (rec->dst_y < mode->crtc_vdisplay)
916 && (rec->dst_y + rec->dst_height
917 <= mode->crtc_vdisplay))
923 static int check_overlay_scaling(struct put_image_params *rec)
927 /* downscaling limit is 8.0 */
928 tmp = ((rec->src_scan_h << 16) / rec->dst_h) >> 16;
931 tmp = ((rec->src_scan_w << 16) / rec->dst_w) >> 16;
938 static int check_overlay_src(struct drm_device *dev,
939 struct drm_intel_overlay_put_image *rec,
940 struct drm_gem_object *new_bo)
944 int uv_hscale = uv_hsubsampling(rec->flags);
945 int uv_vscale = uv_vsubsampling(rec->flags);
948 /* check src dimensions */
949 if (IS_845G(dev) || IS_I830(dev)) {
950 if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY
951 || rec->src_width > IMAGE_MAX_WIDTH_LEGACY)
954 if (rec->src_height > IMAGE_MAX_HEIGHT
955 || rec->src_width > IMAGE_MAX_WIDTH)
958 /* better safe than sorry, use 4 as the maximal subsampling ratio */
959 if (rec->src_height < N_VERT_Y_TAPS*4
960 || rec->src_width < N_HORIZ_Y_TAPS*4)
963 /* check alingment constrains */
964 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
965 case I915_OVERLAY_RGB:
966 /* not implemented */
968 case I915_OVERLAY_YUV_PACKED:
969 depth = packed_depth_bytes(rec->flags);
974 /* ignore UV planes */
978 /* check pixel alignment */
979 if (rec->offset_Y % depth)
982 case I915_OVERLAY_YUV_PLANAR:
983 if (uv_vscale < 0 || uv_hscale < 0)
985 /* no offset restrictions for planar formats */
991 if (rec->src_width % uv_hscale)
994 /* stride checking */
997 if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
999 if (IS_I965G(dev) && rec->stride_Y < 512)
1002 tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
1004 if (rec->stride_Y > tmp*1024 || rec->stride_UV > 2*1024)
1007 /* check buffer dimensions */
1008 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1009 case I915_OVERLAY_RGB:
1010 case I915_OVERLAY_YUV_PACKED:
1011 /* always 4 Y values per depth pixels */
1012 if (packed_width_bytes(rec->flags, rec->src_width)
1016 tmp = rec->stride_Y*rec->src_height;
1017 if (rec->offset_Y + tmp > new_bo->size)
1020 case I915_OVERLAY_YUV_PLANAR:
1021 if (rec->src_width > rec->stride_Y)
1023 if (rec->src_width/uv_hscale > rec->stride_UV)
1026 tmp = rec->stride_Y*rec->src_height;
1027 if (rec->offset_Y + tmp > new_bo->size)
1029 tmp = rec->stride_UV*rec->src_height;
1031 if (rec->offset_U + tmp > new_bo->size
1032 || rec->offset_V + tmp > new_bo->size)
1040 int intel_overlay_put_image(struct drm_device *dev, void *data,
1041 struct drm_file *file_priv)
1043 struct drm_intel_overlay_put_image *put_image_rec = data;
1044 drm_i915_private_t *dev_priv = dev->dev_private;
1045 struct intel_overlay *overlay;
1046 struct drm_mode_object *drmmode_obj;
1047 struct intel_crtc *crtc;
1048 struct drm_gem_object *new_bo;
1049 struct put_image_params *params;
1053 DRM_ERROR("called with no initialization\n");
1057 overlay = dev_priv->overlay;
1059 DRM_DEBUG("userspace bug: no overlay\n");
1063 if (!(put_image_rec->flags & I915_OVERLAY_ENABLE)) {
1064 mutex_lock(&dev->mode_config.mutex);
1065 mutex_lock(&dev->struct_mutex);
1067 ret = intel_overlay_switch_off(overlay);
1069 mutex_unlock(&dev->struct_mutex);
1070 mutex_unlock(&dev->mode_config.mutex);
1075 params = kmalloc(sizeof(struct put_image_params), GFP_KERNEL);
1079 drmmode_obj = drm_mode_object_find(dev, put_image_rec->crtc_id,
1080 DRM_MODE_OBJECT_CRTC);
1083 crtc = to_intel_crtc(obj_to_crtc(drmmode_obj));
1085 new_bo = drm_gem_object_lookup(dev, file_priv,
1086 put_image_rec->bo_handle);
1090 mutex_lock(&dev->mode_config.mutex);
1091 mutex_lock(&dev->struct_mutex);
1093 if (overlay->hw_wedged) {
1094 ret = intel_overlay_recover_from_interrupt(overlay, 1);
1099 if (overlay->crtc != crtc) {
1100 struct drm_display_mode *mode = &crtc->base.mode;
1101 ret = intel_overlay_switch_off(overlay);
1105 ret = check_overlay_possible_on_crtc(overlay, crtc);
1109 overlay->crtc = crtc;
1110 crtc->overlay = overlay;
1112 if (intel_panel_fitter_pipe(dev) == crtc->pipe
1113 /* and line to wide, i.e. one-line-mode */
1114 && mode->hdisplay > 1024) {
1115 overlay->pfit_active = 1;
1116 update_pfit_vscale_ratio(overlay);
1118 overlay->pfit_active = 0;
1121 ret = check_overlay_dst(overlay, put_image_rec);
1125 if (overlay->pfit_active) {
1126 params->dst_y = ((((u32)put_image_rec->dst_y) << 12) /
1127 overlay->pfit_vscale_ratio);
1128 /* shifting right rounds downwards, so add 1 */
1129 params->dst_h = ((((u32)put_image_rec->dst_height) << 12) /
1130 overlay->pfit_vscale_ratio) + 1;
1132 params->dst_y = put_image_rec->dst_y;
1133 params->dst_h = put_image_rec->dst_height;
1135 params->dst_x = put_image_rec->dst_x;
1136 params->dst_w = put_image_rec->dst_width;
1138 params->src_w = put_image_rec->src_width;
1139 params->src_h = put_image_rec->src_height;
1140 params->src_scan_w = put_image_rec->src_scan_width;
1141 params->src_scan_h = put_image_rec->src_scan_height;
1142 if (params->src_scan_h > params->src_h
1143 || params->src_scan_w > params->src_w) {
1148 ret = check_overlay_src(dev, put_image_rec, new_bo);
1151 params->format = put_image_rec->flags & ~I915_OVERLAY_FLAGS_MASK;
1152 params->stride_Y = put_image_rec->stride_Y;
1153 params->stride_UV = put_image_rec->stride_UV;
1154 params->offset_Y = put_image_rec->offset_Y;
1155 params->offset_U = put_image_rec->offset_U;
1156 params->offset_V = put_image_rec->offset_V;
1158 /* Check scaling after src size to prevent a divide-by-zero. */
1159 ret = check_overlay_scaling(params);
1163 ret = intel_overlay_do_put_image(overlay, new_bo, params);
1167 mutex_unlock(&dev->struct_mutex);
1168 mutex_unlock(&dev->mode_config.mutex);
1175 mutex_unlock(&dev->struct_mutex);
1176 mutex_unlock(&dev->mode_config.mutex);
1177 drm_gem_object_unreference(new_bo);
1183 static void update_reg_attrs(struct intel_overlay *overlay,
1184 struct overlay_registers *regs)
1186 regs->OCLRC0 = (overlay->contrast << 18) | (overlay->brightness & 0xff);
1187 regs->OCLRC1 = overlay->saturation;
1190 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
1194 if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
1197 for (i = 0; i < 3; i++) {
1198 if (((gamma1 >> i * 8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
1205 static bool check_gamma5_errata(u32 gamma5)
1209 for (i = 0; i < 3; i++) {
1210 if (((gamma5 >> i*8) & 0xff) == 0x80)
1217 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
1219 if (!check_gamma_bounds(0, attrs->gamma0)
1220 || !check_gamma_bounds(attrs->gamma0, attrs->gamma1)
1221 || !check_gamma_bounds(attrs->gamma1, attrs->gamma2)
1222 || !check_gamma_bounds(attrs->gamma2, attrs->gamma3)
1223 || !check_gamma_bounds(attrs->gamma3, attrs->gamma4)
1224 || !check_gamma_bounds(attrs->gamma4, attrs->gamma5)
1225 || !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
1227 if (!check_gamma5_errata(attrs->gamma5))
1232 int intel_overlay_attrs(struct drm_device *dev, void *data,
1233 struct drm_file *file_priv)
1235 struct drm_intel_overlay_attrs *attrs = data;
1236 drm_i915_private_t *dev_priv = dev->dev_private;
1237 struct intel_overlay *overlay;
1238 struct overlay_registers *regs;
1242 DRM_ERROR("called with no initialization\n");
1246 overlay = dev_priv->overlay;
1248 DRM_DEBUG("userspace bug: no overlay\n");
1252 mutex_lock(&dev->mode_config.mutex);
1253 mutex_lock(&dev->struct_mutex);
1255 if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
1256 attrs->color_key = overlay->color_key;
1257 attrs->brightness = overlay->brightness;
1258 attrs->contrast = overlay->contrast;
1259 attrs->saturation = overlay->saturation;
1262 attrs->gamma0 = I915_READ(OGAMC0);
1263 attrs->gamma1 = I915_READ(OGAMC1);
1264 attrs->gamma2 = I915_READ(OGAMC2);
1265 attrs->gamma3 = I915_READ(OGAMC3);
1266 attrs->gamma4 = I915_READ(OGAMC4);
1267 attrs->gamma5 = I915_READ(OGAMC5);
1271 overlay->color_key = attrs->color_key;
1272 if (attrs->brightness >= -128 && attrs->brightness <= 127) {
1273 overlay->brightness = attrs->brightness;
1278 if (attrs->contrast <= 255) {
1279 overlay->contrast = attrs->contrast;
1284 if (attrs->saturation <= 1023) {
1285 overlay->saturation = attrs->saturation;
1291 regs = intel_overlay_map_regs_atomic(overlay);
1297 update_reg_attrs(overlay, regs);
1299 intel_overlay_unmap_regs_atomic(overlay);
1301 if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
1302 if (!IS_I9XX(dev)) {
1307 if (overlay->active) {
1312 ret = check_gamma(attrs);
1316 I915_WRITE(OGAMC0, attrs->gamma0);
1317 I915_WRITE(OGAMC1, attrs->gamma1);
1318 I915_WRITE(OGAMC2, attrs->gamma2);
1319 I915_WRITE(OGAMC3, attrs->gamma3);
1320 I915_WRITE(OGAMC4, attrs->gamma4);
1321 I915_WRITE(OGAMC5, attrs->gamma5);
1327 mutex_unlock(&dev->struct_mutex);
1328 mutex_unlock(&dev->mode_config.mutex);
1333 void intel_setup_overlay(struct drm_device *dev)
1335 drm_i915_private_t *dev_priv = dev->dev_private;
1336 struct intel_overlay *overlay;
1337 struct drm_gem_object *reg_bo;
1338 struct overlay_registers *regs;
1341 if (!OVERLAY_EXISTS(dev))
1344 overlay = kzalloc(sizeof(struct intel_overlay), GFP_KERNEL);
1349 reg_bo = drm_gem_object_alloc(dev, PAGE_SIZE);
1352 overlay->reg_bo = reg_bo->driver_private;
1354 if (OVERLAY_NONPHYSICAL(dev)) {
1355 ret = i915_gem_object_pin(reg_bo, PAGE_SIZE);
1357 DRM_ERROR("failed to pin overlay register bo\n");
1360 overlay->flip_addr = overlay->reg_bo->gtt_offset;
1362 ret = i915_gem_attach_phys_object(dev, reg_bo,
1363 I915_GEM_PHYS_OVERLAY_REGS);
1365 DRM_ERROR("failed to attach phys overlay regs\n");
1368 overlay->flip_addr = overlay->reg_bo->phys_obj->handle->busaddr;
1371 /* init all values */
1372 overlay->color_key = 0x0101fe;
1373 overlay->brightness = -19;
1374 overlay->contrast = 75;
1375 overlay->saturation = 146;
1377 regs = intel_overlay_map_regs_atomic(overlay);
1381 memset(regs, 0, sizeof(struct overlay_registers));
1382 update_polyphase_filter(regs);
1384 update_reg_attrs(overlay, regs);
1386 intel_overlay_unmap_regs_atomic(overlay);
1388 dev_priv->overlay = overlay;
1389 DRM_INFO("initialized overlay support\n");
1393 drm_gem_object_unreference(reg_bo);
1399 void intel_cleanup_overlay(struct drm_device *dev)
1401 drm_i915_private_t *dev_priv = dev->dev_private;
1403 if (dev_priv->overlay) {
1404 /* The bo's should be free'd by the generic code already.
1405 * Furthermore modesetting teardown happens beforehand so the
1406 * hardware should be off already */
1407 BUG_ON(dev_priv->overlay->active);
1409 kfree(dev_priv->overlay);