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, 0);
293 overlay->last_flip_req = 0;
295 tmp = I915_READ(ISR);
297 if (!(tmp & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT))
301 /* synchronous slowpath */
302 overlay->hw_wedged = RELEASE_OLD_VID;
305 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
309 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
310 if (overlay->last_flip_req == 0)
313 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
317 overlay->hw_wedged = 0;
318 overlay->last_flip_req = 0;
322 /* overlay needs to be disabled in OCMD reg */
323 static int intel_overlay_off(struct intel_overlay *overlay)
325 u32 flip_addr = overlay->flip_addr;
326 struct drm_device *dev = overlay->dev;
327 drm_i915_private_t *dev_priv = dev->dev_private;
331 BUG_ON(!overlay->active);
333 /* According to intel docs the overlay hw may hang (when switching
334 * off) without loading the filter coeffs. It is however unclear whether
335 * this applies to the disabling of the overlay or to the switching off
336 * of the hw. Do it in both cases */
337 flip_addr |= OFC_UPDATE;
339 /* wait for overlay to go idle */
340 overlay->hw_wedged = SWITCH_OFF_STAGE_1;
345 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
347 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
351 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
352 if (overlay->last_flip_req == 0)
355 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
359 /* turn overlay off */
360 overlay->hw_wedged = SWITCH_OFF_STAGE_2;
365 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
367 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
371 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
372 if (overlay->last_flip_req == 0)
375 ret = i915_do_wait_request(dev, overlay->last_flip_req, 1);
380 overlay->hw_wedged = 0;
381 overlay->last_flip_req = 0;
385 /* recover from an interruption due to a signal
386 * We have to be careful not to repeat work forever an make forward progess. */
387 int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
390 struct drm_device *dev = overlay->dev;
391 drm_i915_private_t *dev_priv = dev->dev_private;
392 struct drm_gem_object *obj;
397 if (overlay->hw_wedged == HW_WEDGED)
400 if (overlay->last_flip_req == 0) {
401 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
402 if (overlay->last_flip_req == 0)
406 ret = i915_do_wait_request(dev, overlay->last_flip_req, interruptible);
410 switch (overlay->hw_wedged) {
411 case RELEASE_OLD_VID:
412 obj = overlay->old_vid_bo->obj;
413 i915_gem_object_unpin(obj);
414 drm_gem_object_unreference(obj);
415 overlay->old_vid_bo = NULL;
417 case SWITCH_OFF_STAGE_1:
418 flip_addr = overlay->flip_addr;
419 flip_addr |= OFC_UPDATE;
421 overlay->hw_wedged = SWITCH_OFF_STAGE_2;
426 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
428 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
432 overlay->last_flip_req = i915_add_request(dev, NULL, 0);
433 if (overlay->last_flip_req == 0)
436 ret = i915_do_wait_request(dev, overlay->last_flip_req,
441 case SWITCH_OFF_STAGE_2:
442 printk("switch off 2\n");
444 BUG_ON(!overlay->vid_bo);
445 obj = overlay->vid_bo->obj;
447 i915_gem_object_unpin(obj);
448 drm_gem_object_unreference(obj);
449 overlay->vid_bo = NULL;
451 overlay->crtc->overlay = NULL;
452 overlay->crtc = NULL;
457 BUG_ON(overlay->hw_wedged != NEEDS_WAIT_FOR_FLIP);
460 overlay->hw_wedged = 0;
461 overlay->last_flip_req = 0;
465 /* Wait for pending overlay flip and release old frame.
466 * Needs to be called before the overlay register are changed
467 * via intel_overlay_(un)map_regs_atomic */
468 static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
471 struct drm_gem_object *obj;
473 /* only wait if there is actually an old frame to release to
474 * guarantee forward progress */
475 if (!overlay->old_vid_bo)
478 ret = intel_overlay_wait_flip(overlay);
482 obj = overlay->old_vid_bo->obj;
483 i915_gem_object_unpin(obj);
484 drm_gem_object_unreference(obj);
485 overlay->old_vid_bo = NULL;
490 struct put_image_params {
507 static int packed_depth_bytes(u32 format)
509 switch (format & I915_OVERLAY_DEPTH_MASK) {
510 case I915_OVERLAY_YUV422:
512 case I915_OVERLAY_YUV411:
513 /* return 6; not implemented */
519 static int packed_width_bytes(u32 format, short width)
521 switch (format & I915_OVERLAY_DEPTH_MASK) {
522 case I915_OVERLAY_YUV422:
529 static int uv_hsubsampling(u32 format)
531 switch (format & I915_OVERLAY_DEPTH_MASK) {
532 case I915_OVERLAY_YUV422:
533 case I915_OVERLAY_YUV420:
535 case I915_OVERLAY_YUV411:
536 case I915_OVERLAY_YUV410:
543 static int uv_vsubsampling(u32 format)
545 switch (format & I915_OVERLAY_DEPTH_MASK) {
546 case I915_OVERLAY_YUV420:
547 case I915_OVERLAY_YUV410:
549 case I915_OVERLAY_YUV422:
550 case I915_OVERLAY_YUV411:
557 static u32 calc_swidthsw(struct drm_device *dev, u32 offset, u32 width)
559 u32 mask, shift, ret;
567 ret = ((offset + width + mask) >> shift) - (offset >> shift);
574 static const u16 y_static_hcoeffs[N_HORIZ_Y_TAPS * N_PHASES] = {
575 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
576 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
577 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
578 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
579 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
580 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
581 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
582 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
583 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
584 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
585 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
586 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
587 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
588 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
589 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
590 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
591 0xb000, 0x3000, 0x0800, 0x3000, 0xb000};
592 static const u16 uv_static_hcoeffs[N_HORIZ_UV_TAPS * N_PHASES] = {
593 0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
594 0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
595 0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
596 0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
597 0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
598 0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
599 0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
600 0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
601 0x3000, 0x0800, 0x3000};
603 static void update_polyphase_filter(struct overlay_registers *regs)
605 memcpy(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
606 memcpy(regs->UV_HCOEFS, uv_static_hcoeffs, sizeof(uv_static_hcoeffs));
609 static bool update_scaling_factors(struct intel_overlay *overlay,
610 struct overlay_registers *regs,
611 struct put_image_params *params)
613 /* fixed point with a 12 bit shift */
614 u32 xscale, yscale, xscale_UV, yscale_UV;
616 #define FRACT_MASK 0xfff
617 bool scale_changed = false;
618 int uv_hscale = uv_hsubsampling(params->format);
619 int uv_vscale = uv_vsubsampling(params->format);
621 if (params->dst_w > 1)
622 xscale = ((params->src_scan_w - 1) << FP_SHIFT)
625 xscale = 1 << FP_SHIFT;
627 if (params->dst_h > 1)
628 yscale = ((params->src_scan_h - 1) << FP_SHIFT)
631 yscale = 1 << FP_SHIFT;
633 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
634 xscale_UV = xscale/uv_hscale;
635 yscale_UV = yscale/uv_vscale;
636 /* make the Y scale to UV scale ratio an exact multiply */
637 xscale = xscale_UV * uv_hscale;
638 yscale = yscale_UV * uv_vscale;
644 if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
645 scale_changed = true;
646 overlay->old_xscale = xscale;
647 overlay->old_yscale = yscale;
649 regs->YRGBSCALE = ((yscale & FRACT_MASK) << 20)
650 | ((xscale >> FP_SHIFT) << 16)
651 | ((xscale & FRACT_MASK) << 3);
652 regs->UVSCALE = ((yscale_UV & FRACT_MASK) << 20)
653 | ((xscale_UV >> FP_SHIFT) << 16)
654 | ((xscale_UV & FRACT_MASK) << 3);
655 regs->UVSCALEV = ((yscale >> FP_SHIFT) << 16)
656 | ((yscale_UV >> FP_SHIFT) << 0);
659 update_polyphase_filter(regs);
661 return scale_changed;
664 static void update_colorkey(struct intel_overlay *overlay,
665 struct overlay_registers *regs)
667 u32 key = overlay->color_key;
668 switch (overlay->crtc->base.fb->bits_per_pixel) {
671 regs->DCLRKM = CLK_RGB8I_MASK | DST_KEY_ENABLE;
673 if (overlay->crtc->base.fb->depth == 15) {
674 regs->DCLRKV = RGB15_TO_COLORKEY(key);
675 regs->DCLRKM = CLK_RGB15_MASK | DST_KEY_ENABLE;
677 regs->DCLRKV = RGB16_TO_COLORKEY(key);
678 regs->DCLRKM = CLK_RGB16_MASK | DST_KEY_ENABLE;
683 regs->DCLRKM = CLK_RGB24_MASK | DST_KEY_ENABLE;
687 static u32 overlay_cmd_reg(struct put_image_params *params)
689 u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
691 if (params->format & I915_OVERLAY_YUV_PLANAR) {
692 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
693 case I915_OVERLAY_YUV422:
694 cmd |= OCMD_YUV_422_PLANAR;
696 case I915_OVERLAY_YUV420:
697 cmd |= OCMD_YUV_420_PLANAR;
699 case I915_OVERLAY_YUV411:
700 case I915_OVERLAY_YUV410:
701 cmd |= OCMD_YUV_410_PLANAR;
704 } else { /* YUV packed */
705 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
706 case I915_OVERLAY_YUV422:
707 cmd |= OCMD_YUV_422_PACKED;
709 case I915_OVERLAY_YUV411:
710 cmd |= OCMD_YUV_411_PACKED;
714 switch (params->format & I915_OVERLAY_SWAP_MASK) {
715 case I915_OVERLAY_NO_SWAP:
717 case I915_OVERLAY_UV_SWAP:
720 case I915_OVERLAY_Y_SWAP:
723 case I915_OVERLAY_Y_AND_UV_SWAP:
724 cmd |= OCMD_Y_AND_UV_SWAP;
732 int intel_overlay_do_put_image(struct intel_overlay *overlay,
733 struct drm_gem_object *new_bo,
734 struct put_image_params *params)
737 struct overlay_registers *regs;
738 bool scale_changed = false;
739 struct drm_i915_gem_object *bo_priv = new_bo->driver_private;
740 struct drm_device *dev = overlay->dev;
742 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
743 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
746 ret = intel_overlay_release_old_vid(overlay);
750 ret = i915_gem_object_pin(new_bo, PAGE_SIZE);
754 ret = i915_gem_object_set_to_gtt_domain(new_bo, 0);
758 if (!overlay->active) {
759 regs = intel_overlay_map_regs_atomic(overlay);
764 regs->OCONFIG = OCONF_CC_OUT_8BIT;
765 if (IS_I965GM(overlay->dev))
766 regs->OCONFIG |= OCONF_CSC_MODE_BT709;
767 regs->OCONFIG |= overlay->crtc->pipe == 0 ?
768 OCONF_PIPE_A : OCONF_PIPE_B;
769 intel_overlay_unmap_regs_atomic(overlay);
771 ret = intel_overlay_on(overlay);
776 regs = intel_overlay_map_regs_atomic(overlay);
782 regs->DWINPOS = (params->dst_y << 16) | params->dst_x;
783 regs->DWINSZ = (params->dst_h << 16) | params->dst_w;
785 if (params->format & I915_OVERLAY_YUV_PACKED)
786 tmp_width = packed_width_bytes(params->format, params->src_w);
788 tmp_width = params->src_w;
790 regs->SWIDTH = params->src_w;
791 regs->SWIDTHSW = calc_swidthsw(overlay->dev,
792 params->offset_Y, tmp_width);
793 regs->SHEIGHT = params->src_h;
794 regs->OBUF_0Y = bo_priv->gtt_offset + params-> offset_Y;
795 regs->OSTRIDE = params->stride_Y;
797 if (params->format & I915_OVERLAY_YUV_PLANAR) {
798 int uv_hscale = uv_hsubsampling(params->format);
799 int uv_vscale = uv_vsubsampling(params->format);
801 regs->SWIDTH |= (params->src_w/uv_hscale) << 16;
802 tmp_U = calc_swidthsw(overlay->dev, params->offset_U,
803 params->src_w/uv_hscale);
804 tmp_V = calc_swidthsw(overlay->dev, params->offset_V,
805 params->src_w/uv_hscale);
806 regs->SWIDTHSW |= max_t(u32, tmp_U, tmp_V) << 16;
807 regs->SHEIGHT |= (params->src_h/uv_vscale) << 16;
808 regs->OBUF_0U = bo_priv->gtt_offset + params->offset_U;
809 regs->OBUF_0V = bo_priv->gtt_offset + params->offset_V;
810 regs->OSTRIDE |= params->stride_UV << 16;
813 scale_changed = update_scaling_factors(overlay, regs, params);
815 update_colorkey(overlay, regs);
817 regs->OCMD = overlay_cmd_reg(params);
819 intel_overlay_unmap_regs_atomic(overlay);
821 intel_overlay_continue(overlay, scale_changed);
823 overlay->old_vid_bo = overlay->vid_bo;
824 overlay->vid_bo = new_bo->driver_private;
829 i915_gem_object_unpin(new_bo);
833 int intel_overlay_switch_off(struct intel_overlay *overlay)
836 struct overlay_registers *regs;
837 struct drm_gem_object *obj;
838 struct drm_device *dev = overlay->dev;
840 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
841 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
843 if (!overlay->active)
846 if (overlay->hw_wedged)
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 /* never have the overlay hw on without showing a frame */
862 BUG_ON(!overlay->vid_bo);
863 obj = overlay->vid_bo->obj;
865 i915_gem_object_unpin(obj);
866 drm_gem_object_unreference(obj);
867 overlay->vid_bo = NULL;
869 overlay->crtc->overlay = NULL;
870 overlay->crtc = NULL;
875 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
876 struct intel_crtc *crtc)
878 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
880 int pipeconf_reg = (crtc->pipe == 0) ? PIPEACONF : PIPEBCONF;
882 if (!crtc->base.enabled || crtc->dpms_mode != DRM_MODE_DPMS_ON)
885 pipeconf = I915_READ(pipeconf_reg);
887 /* can't use the overlay with double wide pipe */
888 if (!IS_I965G(overlay->dev) && pipeconf & PIPEACONF_DOUBLE_WIDE)
894 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
896 struct drm_device *dev = overlay->dev;
897 drm_i915_private_t *dev_priv = dev->dev_private;
899 u32 pfit_control = I915_READ(PFIT_CONTROL);
901 /* XXX: This is not the same logic as in the xorg driver, but more in
902 * line with the intel documentation for the i965 */
903 if (!IS_I965G(dev) && (pfit_control & VERT_AUTO_SCALE)) {
904 ratio = I915_READ(PFIT_AUTO_RATIOS) >> PFIT_VERT_SCALE_SHIFT;
905 } else { /* on i965 use the PGM reg to read out the autoscaler values */
906 ratio = I915_READ(PFIT_PGM_RATIOS);
908 ratio >>= PFIT_VERT_SCALE_SHIFT_965;
910 ratio >>= PFIT_VERT_SCALE_SHIFT;
913 overlay->pfit_vscale_ratio = ratio;
916 static int check_overlay_dst(struct intel_overlay *overlay,
917 struct drm_intel_overlay_put_image *rec)
919 struct drm_display_mode *mode = &overlay->crtc->base.mode;
921 if ((rec->dst_x < mode->crtc_hdisplay)
922 && (rec->dst_x + rec->dst_width
923 <= mode->crtc_hdisplay)
924 && (rec->dst_y < mode->crtc_vdisplay)
925 && (rec->dst_y + rec->dst_height
926 <= mode->crtc_vdisplay))
932 static int check_overlay_scaling(struct put_image_params *rec)
936 /* downscaling limit is 8.0 */
937 tmp = ((rec->src_scan_h << 16) / rec->dst_h) >> 16;
940 tmp = ((rec->src_scan_w << 16) / rec->dst_w) >> 16;
947 static int check_overlay_src(struct drm_device *dev,
948 struct drm_intel_overlay_put_image *rec,
949 struct drm_gem_object *new_bo)
953 int uv_hscale = uv_hsubsampling(rec->flags);
954 int uv_vscale = uv_vsubsampling(rec->flags);
957 /* check src dimensions */
958 if (IS_845G(dev) || IS_I830(dev)) {
959 if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY
960 || rec->src_width > IMAGE_MAX_WIDTH_LEGACY)
963 if (rec->src_height > IMAGE_MAX_HEIGHT
964 || rec->src_width > IMAGE_MAX_WIDTH)
967 /* better safe than sorry, use 4 as the maximal subsampling ratio */
968 if (rec->src_height < N_VERT_Y_TAPS*4
969 || rec->src_width < N_HORIZ_Y_TAPS*4)
972 /* check alingment constrains */
973 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
974 case I915_OVERLAY_RGB:
975 /* not implemented */
977 case I915_OVERLAY_YUV_PACKED:
978 depth = packed_depth_bytes(rec->flags);
983 /* ignore UV planes */
987 /* check pixel alignment */
988 if (rec->offset_Y % depth)
991 case I915_OVERLAY_YUV_PLANAR:
992 if (uv_vscale < 0 || uv_hscale < 0)
994 /* no offset restrictions for planar formats */
1000 if (rec->src_width % uv_hscale)
1003 /* stride checking */
1006 if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
1008 if (IS_I965G(dev) && rec->stride_Y < 512)
1011 tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
1013 if (rec->stride_Y > tmp*1024 || rec->stride_UV > 2*1024)
1016 /* check buffer dimensions */
1017 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1018 case I915_OVERLAY_RGB:
1019 case I915_OVERLAY_YUV_PACKED:
1020 /* always 4 Y values per depth pixels */
1021 if (packed_width_bytes(rec->flags, rec->src_width)
1025 tmp = rec->stride_Y*rec->src_height;
1026 if (rec->offset_Y + tmp > new_bo->size)
1029 case I915_OVERLAY_YUV_PLANAR:
1030 if (rec->src_width > rec->stride_Y)
1032 if (rec->src_width/uv_hscale > rec->stride_UV)
1035 tmp = rec->stride_Y*rec->src_height;
1036 if (rec->offset_Y + tmp > new_bo->size)
1038 tmp = rec->stride_UV*rec->src_height;
1040 if (rec->offset_U + tmp > new_bo->size
1041 || rec->offset_V + tmp > new_bo->size)
1049 int intel_overlay_put_image(struct drm_device *dev, void *data,
1050 struct drm_file *file_priv)
1052 struct drm_intel_overlay_put_image *put_image_rec = data;
1053 drm_i915_private_t *dev_priv = dev->dev_private;
1054 struct intel_overlay *overlay;
1055 struct drm_mode_object *drmmode_obj;
1056 struct intel_crtc *crtc;
1057 struct drm_gem_object *new_bo;
1058 struct put_image_params *params;
1062 DRM_ERROR("called with no initialization\n");
1066 overlay = dev_priv->overlay;
1068 DRM_DEBUG("userspace bug: no overlay\n");
1072 if (!(put_image_rec->flags & I915_OVERLAY_ENABLE)) {
1073 mutex_lock(&dev->mode_config.mutex);
1074 mutex_lock(&dev->struct_mutex);
1076 ret = intel_overlay_switch_off(overlay);
1078 mutex_unlock(&dev->struct_mutex);
1079 mutex_unlock(&dev->mode_config.mutex);
1084 params = kmalloc(sizeof(struct put_image_params), GFP_KERNEL);
1088 drmmode_obj = drm_mode_object_find(dev, put_image_rec->crtc_id,
1089 DRM_MODE_OBJECT_CRTC);
1092 crtc = to_intel_crtc(obj_to_crtc(drmmode_obj));
1094 new_bo = drm_gem_object_lookup(dev, file_priv,
1095 put_image_rec->bo_handle);
1099 mutex_lock(&dev->mode_config.mutex);
1100 mutex_lock(&dev->struct_mutex);
1102 if (overlay->hw_wedged) {
1103 ret = intel_overlay_recover_from_interrupt(overlay, 1);
1108 if (overlay->crtc != crtc) {
1109 struct drm_display_mode *mode = &crtc->base.mode;
1110 ret = intel_overlay_switch_off(overlay);
1114 ret = check_overlay_possible_on_crtc(overlay, crtc);
1118 overlay->crtc = crtc;
1119 crtc->overlay = overlay;
1121 if (intel_panel_fitter_pipe(dev) == crtc->pipe
1122 /* and line to wide, i.e. one-line-mode */
1123 && mode->hdisplay > 1024) {
1124 overlay->pfit_active = 1;
1125 update_pfit_vscale_ratio(overlay);
1127 overlay->pfit_active = 0;
1130 ret = check_overlay_dst(overlay, put_image_rec);
1134 if (overlay->pfit_active) {
1135 params->dst_y = ((((u32)put_image_rec->dst_y) << 12) /
1136 overlay->pfit_vscale_ratio);
1137 /* shifting right rounds downwards, so add 1 */
1138 params->dst_h = ((((u32)put_image_rec->dst_height) << 12) /
1139 overlay->pfit_vscale_ratio) + 1;
1141 params->dst_y = put_image_rec->dst_y;
1142 params->dst_h = put_image_rec->dst_height;
1144 params->dst_x = put_image_rec->dst_x;
1145 params->dst_w = put_image_rec->dst_width;
1147 params->src_w = put_image_rec->src_width;
1148 params->src_h = put_image_rec->src_height;
1149 params->src_scan_w = put_image_rec->src_scan_width;
1150 params->src_scan_h = put_image_rec->src_scan_height;
1151 if (params->src_scan_h > params->src_h
1152 || params->src_scan_w > params->src_w) {
1157 ret = check_overlay_src(dev, put_image_rec, new_bo);
1160 params->format = put_image_rec->flags & ~I915_OVERLAY_FLAGS_MASK;
1161 params->stride_Y = put_image_rec->stride_Y;
1162 params->stride_UV = put_image_rec->stride_UV;
1163 params->offset_Y = put_image_rec->offset_Y;
1164 params->offset_U = put_image_rec->offset_U;
1165 params->offset_V = put_image_rec->offset_V;
1167 /* Check scaling after src size to prevent a divide-by-zero. */
1168 ret = check_overlay_scaling(params);
1172 ret = intel_overlay_do_put_image(overlay, new_bo, params);
1176 mutex_unlock(&dev->struct_mutex);
1177 mutex_unlock(&dev->mode_config.mutex);
1184 mutex_unlock(&dev->struct_mutex);
1185 mutex_unlock(&dev->mode_config.mutex);
1186 drm_gem_object_unreference(new_bo);
1192 static void update_reg_attrs(struct intel_overlay *overlay,
1193 struct overlay_registers *regs)
1195 regs->OCLRC0 = (overlay->contrast << 18) | (overlay->brightness & 0xff);
1196 regs->OCLRC1 = overlay->saturation;
1199 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
1203 if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
1206 for (i = 0; i < 3; i++) {
1207 if (((gamma1 >> i * 8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
1214 static bool check_gamma5_errata(u32 gamma5)
1218 for (i = 0; i < 3; i++) {
1219 if (((gamma5 >> i*8) & 0xff) == 0x80)
1226 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
1228 if (!check_gamma_bounds(0, attrs->gamma0)
1229 || !check_gamma_bounds(attrs->gamma0, attrs->gamma1)
1230 || !check_gamma_bounds(attrs->gamma1, attrs->gamma2)
1231 || !check_gamma_bounds(attrs->gamma2, attrs->gamma3)
1232 || !check_gamma_bounds(attrs->gamma3, attrs->gamma4)
1233 || !check_gamma_bounds(attrs->gamma4, attrs->gamma5)
1234 || !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
1236 if (!check_gamma5_errata(attrs->gamma5))
1241 int intel_overlay_attrs(struct drm_device *dev, void *data,
1242 struct drm_file *file_priv)
1244 struct drm_intel_overlay_attrs *attrs = data;
1245 drm_i915_private_t *dev_priv = dev->dev_private;
1246 struct intel_overlay *overlay;
1247 struct overlay_registers *regs;
1251 DRM_ERROR("called with no initialization\n");
1255 overlay = dev_priv->overlay;
1257 DRM_DEBUG("userspace bug: no overlay\n");
1261 mutex_lock(&dev->mode_config.mutex);
1262 mutex_lock(&dev->struct_mutex);
1264 if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
1265 attrs->color_key = overlay->color_key;
1266 attrs->brightness = overlay->brightness;
1267 attrs->contrast = overlay->contrast;
1268 attrs->saturation = overlay->saturation;
1271 attrs->gamma0 = I915_READ(OGAMC0);
1272 attrs->gamma1 = I915_READ(OGAMC1);
1273 attrs->gamma2 = I915_READ(OGAMC2);
1274 attrs->gamma3 = I915_READ(OGAMC3);
1275 attrs->gamma4 = I915_READ(OGAMC4);
1276 attrs->gamma5 = I915_READ(OGAMC5);
1280 overlay->color_key = attrs->color_key;
1281 if (attrs->brightness >= -128 && attrs->brightness <= 127) {
1282 overlay->brightness = attrs->brightness;
1287 if (attrs->contrast <= 255) {
1288 overlay->contrast = attrs->contrast;
1293 if (attrs->saturation <= 1023) {
1294 overlay->saturation = attrs->saturation;
1300 regs = intel_overlay_map_regs_atomic(overlay);
1306 update_reg_attrs(overlay, regs);
1308 intel_overlay_unmap_regs_atomic(overlay);
1310 if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
1311 if (!IS_I9XX(dev)) {
1316 if (overlay->active) {
1321 ret = check_gamma(attrs);
1325 I915_WRITE(OGAMC0, attrs->gamma0);
1326 I915_WRITE(OGAMC1, attrs->gamma1);
1327 I915_WRITE(OGAMC2, attrs->gamma2);
1328 I915_WRITE(OGAMC3, attrs->gamma3);
1329 I915_WRITE(OGAMC4, attrs->gamma4);
1330 I915_WRITE(OGAMC5, attrs->gamma5);
1336 mutex_unlock(&dev->struct_mutex);
1337 mutex_unlock(&dev->mode_config.mutex);
1342 void intel_setup_overlay(struct drm_device *dev)
1344 drm_i915_private_t *dev_priv = dev->dev_private;
1345 struct intel_overlay *overlay;
1346 struct drm_gem_object *reg_bo;
1347 struct overlay_registers *regs;
1350 if (!OVERLAY_EXISTS(dev))
1353 overlay = kzalloc(sizeof(struct intel_overlay), GFP_KERNEL);
1358 reg_bo = drm_gem_object_alloc(dev, PAGE_SIZE);
1361 overlay->reg_bo = reg_bo->driver_private;
1363 if (OVERLAY_NONPHYSICAL(dev)) {
1364 ret = i915_gem_object_pin(reg_bo, PAGE_SIZE);
1366 DRM_ERROR("failed to pin overlay register bo\n");
1369 overlay->flip_addr = overlay->reg_bo->gtt_offset;
1371 ret = i915_gem_attach_phys_object(dev, reg_bo,
1372 I915_GEM_PHYS_OVERLAY_REGS);
1374 DRM_ERROR("failed to attach phys overlay regs\n");
1377 overlay->flip_addr = overlay->reg_bo->phys_obj->handle->busaddr;
1380 /* init all values */
1381 overlay->color_key = 0x0101fe;
1382 overlay->brightness = -19;
1383 overlay->contrast = 75;
1384 overlay->saturation = 146;
1386 regs = intel_overlay_map_regs_atomic(overlay);
1390 memset(regs, 0, sizeof(struct overlay_registers));
1391 update_polyphase_filter(regs);
1393 update_reg_attrs(overlay, regs);
1395 intel_overlay_unmap_regs_atomic(overlay);
1397 dev_priv->overlay = overlay;
1398 DRM_INFO("initialized overlay support\n");
1402 drm_gem_object_unreference(reg_bo);
1408 void intel_cleanup_overlay(struct drm_device *dev)
1410 drm_i915_private_t *dev_priv = dev->dev_private;
1412 if (dev_priv->overlay) {
1413 /* The bo's should be free'd by the generic code already.
1414 * Furthermore modesetting teardown happens beforehand so the
1415 * hardware should be off already */
1416 BUG_ON(dev_priv->overlay->active);
1418 kfree(dev_priv->overlay);