4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
24 * Daniel Vetter <daniel@ffwll.ch>
26 * Derived from Xorg ddx, xf86-video-intel, src/i830_video.c
33 #include "intel_drv.h"
35 /* Limits for overlay size. According to intel doc, the real limits are:
36 * Y width: 4095, UV width (planar): 2047, Y height: 2047,
37 * UV width (planar): * 1023. But the xorg thinks 2048 for height and width. Use
38 * the mininum of both. */
39 #define IMAGE_MAX_WIDTH 2048
40 #define IMAGE_MAX_HEIGHT 2046 /* 2 * 1023 */
41 /* on 830 and 845 these large limits result in the card hanging */
42 #define IMAGE_MAX_WIDTH_LEGACY 1024
43 #define IMAGE_MAX_HEIGHT_LEGACY 1088
45 /* overlay register definitions */
47 #define OCMD_TILED_SURFACE (0x1<<19)
48 #define OCMD_MIRROR_MASK (0x3<<17)
49 #define OCMD_MIRROR_MODE (0x3<<17)
50 #define OCMD_MIRROR_HORIZONTAL (0x1<<17)
51 #define OCMD_MIRROR_VERTICAL (0x2<<17)
52 #define OCMD_MIRROR_BOTH (0x3<<17)
53 #define OCMD_BYTEORDER_MASK (0x3<<14) /* zero for YUYV or FOURCC YUY2 */
54 #define OCMD_UV_SWAP (0x1<<14) /* YVYU */
55 #define OCMD_Y_SWAP (0x2<<14) /* UYVY or FOURCC UYVY */
56 #define OCMD_Y_AND_UV_SWAP (0x3<<14) /* VYUY */
57 #define OCMD_SOURCE_FORMAT_MASK (0xf<<10)
58 #define OCMD_RGB_888 (0x1<<10) /* not in i965 Intel docs */
59 #define OCMD_RGB_555 (0x2<<10) /* not in i965 Intel docs */
60 #define OCMD_RGB_565 (0x3<<10) /* not in i965 Intel docs */
61 #define OCMD_YUV_422_PACKED (0x8<<10)
62 #define OCMD_YUV_411_PACKED (0x9<<10) /* not in i965 Intel docs */
63 #define OCMD_YUV_420_PLANAR (0xc<<10)
64 #define OCMD_YUV_422_PLANAR (0xd<<10)
65 #define OCMD_YUV_410_PLANAR (0xe<<10) /* also 411 */
66 #define OCMD_TVSYNCFLIP_PARITY (0x1<<9)
67 #define OCMD_TVSYNCFLIP_ENABLE (0x1<<7)
68 #define OCMD_BUF_TYPE_MASK (Ox1<<5)
69 #define OCMD_BUF_TYPE_FRAME (0x0<<5)
70 #define OCMD_BUF_TYPE_FIELD (0x1<<5)
71 #define OCMD_TEST_MODE (0x1<<4)
72 #define OCMD_BUFFER_SELECT (0x3<<2)
73 #define OCMD_BUFFER0 (0x0<<2)
74 #define OCMD_BUFFER1 (0x1<<2)
75 #define OCMD_FIELD_SELECT (0x1<<2)
76 #define OCMD_FIELD0 (0x0<<1)
77 #define OCMD_FIELD1 (0x1<<1)
78 #define OCMD_ENABLE (0x1<<0)
80 /* OCONFIG register */
81 #define OCONF_PIPE_MASK (0x1<<18)
82 #define OCONF_PIPE_A (0x0<<18)
83 #define OCONF_PIPE_B (0x1<<18)
84 #define OCONF_GAMMA2_ENABLE (0x1<<16)
85 #define OCONF_CSC_MODE_BT601 (0x0<<5)
86 #define OCONF_CSC_MODE_BT709 (0x1<<5)
87 #define OCONF_CSC_BYPASS (0x1<<4)
88 #define OCONF_CC_OUT_8BIT (0x1<<3)
89 #define OCONF_TEST_MODE (0x1<<2)
90 #define OCONF_THREE_LINE_BUFFER (0x1<<0)
91 #define OCONF_TWO_LINE_BUFFER (0x0<<0)
93 /* DCLRKM (dst-key) register */
94 #define DST_KEY_ENABLE (0x1<<31)
95 #define CLK_RGB24_MASK 0x0
96 #define CLK_RGB16_MASK 0x070307
97 #define CLK_RGB15_MASK 0x070707
98 #define CLK_RGB8I_MASK 0xffffff
100 #define RGB16_TO_COLORKEY(c) \
101 (((c & 0xF800) << 8) | ((c & 0x07E0) << 5) | ((c & 0x001F) << 3))
102 #define RGB15_TO_COLORKEY(c) \
103 (((c & 0x7c00) << 9) | ((c & 0x03E0) << 6) | ((c & 0x001F) << 3))
105 /* overlay flip addr flag */
106 #define OFC_UPDATE 0x1
108 /* polyphase filter coefficients */
109 #define N_HORIZ_Y_TAPS 5
110 #define N_VERT_Y_TAPS 3
111 #define N_HORIZ_UV_TAPS 3
112 #define N_VERT_UV_TAPS 3
116 /* memory bufferd overlay registers */
117 struct overlay_registers
{
145 u32 RESERVED1
; /* 0x6C */
158 u32 FASTHSCALE
; /* 0xA0 */
159 u32 UVSCALEV
; /* 0xA4 */
160 u32 RESERVEDC
[(0x200 - 0xA8) / 4]; /* 0xA8 - 0x1FC */
161 u16 Y_VCOEFS
[N_VERT_Y_TAPS
* N_PHASES
]; /* 0x200 */
162 u16 RESERVEDD
[0x100 / 2 - N_VERT_Y_TAPS
* N_PHASES
];
163 u16 Y_HCOEFS
[N_HORIZ_Y_TAPS
* N_PHASES
]; /* 0x300 */
164 u16 RESERVEDE
[0x200 / 2 - N_HORIZ_Y_TAPS
* N_PHASES
];
165 u16 UV_VCOEFS
[N_VERT_UV_TAPS
* N_PHASES
]; /* 0x500 */
166 u16 RESERVEDF
[0x100 / 2 - N_VERT_UV_TAPS
* N_PHASES
];
167 u16 UV_HCOEFS
[N_HORIZ_UV_TAPS
* N_PHASES
]; /* 0x600 */
168 u16 RESERVEDG
[0x100 / 2 - N_HORIZ_UV_TAPS
* N_PHASES
];
171 /* overlay flip addr flag */
172 #define OFC_UPDATE 0x1
174 #define OVERLAY_NONPHYSICAL(dev) (IS_G33(dev) || IS_I965G(dev))
175 #define OVERLAY_EXISTS(dev) (!IS_G4X(dev) && !IS_IGDNG(dev))
178 static struct overlay_registers
*intel_overlay_map_regs_atomic(struct intel_overlay
*overlay
)
180 drm_i915_private_t
*dev_priv
= overlay
->dev
->dev_private
;
181 struct overlay_registers
*regs
;
183 /* no recursive mappings */
184 BUG_ON(overlay
->virt_addr
);
186 if (OVERLAY_NONPHYSICAL(overlay
->dev
)) {
187 regs
= io_mapping_map_atomic_wc(dev_priv
->mm
.gtt_mapping
,
188 overlay
->reg_bo
->gtt_offset
);
191 DRM_ERROR("failed to map overlay regs in GTT\n");
195 regs
= overlay
->reg_bo
->phys_obj
->handle
->vaddr
;
197 return overlay
->virt_addr
= regs
;
200 static void intel_overlay_unmap_regs_atomic(struct intel_overlay
*overlay
)
202 struct drm_device
*dev
= overlay
->dev
;
203 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
205 if (OVERLAY_NONPHYSICAL(overlay
->dev
))
206 io_mapping_unmap_atomic(overlay
->virt_addr
);
208 overlay
->virt_addr
= NULL
;
210 I915_READ(OVADD
); /* flush wc cashes */
215 /* overlay needs to be disable in OCMD reg */
216 static int intel_overlay_on(struct intel_overlay
*overlay
)
218 struct drm_device
*dev
= overlay
->dev
;
219 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
223 BUG_ON(overlay
->active
);
228 OUT_RING(MI_OVERLAY_FLIP
| MI_OVERLAY_ON
);
229 OUT_RING(overlay
->flip_addr
| OFC_UPDATE
);
230 OUT_RING(MI_WAIT_FOR_EVENT
| MI_WAIT_FOR_OVERLAY_FLIP
);
234 ret
= i915_lp_ring_sync(dev
);
236 DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
237 overlay
->hw_wedged
= 1;
246 /* overlay needs to be enabled in OCMD reg */
247 static void intel_overlay_continue(struct intel_overlay
*overlay
,
248 bool load_polyphase_filter
)
250 struct drm_device
*dev
= overlay
->dev
;
251 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
252 u32 flip_addr
= overlay
->flip_addr
;
257 BUG_ON(!overlay
->active
);
259 if (load_polyphase_filter
)
260 flip_addr
|= OFC_UPDATE
;
262 /* check for underruns */
263 tmp
= I915_READ(DOVSTA
);
265 DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp
);
270 OUT_RING(MI_OVERLAY_FLIP
| MI_OVERLAY_CONTINUE
);
272 OUT_RING(MI_WAIT_FOR_EVENT
| MI_WAIT_FOR_OVERLAY_FLIP
);
276 /* run in lockstep with the hw for easier testing */
277 ret
= i915_lp_ring_sync(dev
);
279 DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
280 overlay
->hw_wedged
= 1;
284 static int intel_overlay_wait_flip(struct intel_overlay
*overlay
)
286 /* don't overcomplicate things for now with asynchronous operations
287 * see comment above */
291 /* overlay needs to be disabled in OCMD reg */
292 static int intel_overlay_off(struct intel_overlay
*overlay
)
294 u32 flip_addr
= overlay
->flip_addr
;
295 struct drm_device
*dev
= overlay
->dev
;
296 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
300 BUG_ON(!overlay
->active
);
302 /* According to intel docs the overlay hw may hang (when switching
303 * off) without loading the filter coeffs. It is however unclear whether
304 * this applies to the disabling of the overlay or to the switching off
305 * of the hw. Do it in both cases */
306 flip_addr
|= OFC_UPDATE
;
308 /* wait for overlay to go idle */
312 OUT_RING(MI_OVERLAY_FLIP
| MI_OVERLAY_CONTINUE
);
314 OUT_RING(MI_WAIT_FOR_EVENT
| MI_WAIT_FOR_OVERLAY_FLIP
);
318 ret
= i915_lp_ring_sync(dev
);
320 DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
321 overlay
->hw_wedged
= 1;
325 /* turn overlay off */
326 /* this is not done in userspace!
330 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
332 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
336 ret = i915_lp_ring_sync(dev);
338 DRM_ERROR("intel overlay: ring sync failed, hw likely wedged\n");
339 overlay->hw_wedged = 1;
348 /* wait for pending overlay flip and release old frame */
349 static int intel_overlay_release_old_vid(struct intel_overlay
*overlay
)
352 struct drm_gem_object
*obj
;
354 ret
= intel_overlay_wait_flip(overlay
);
358 if (!overlay
->old_vid_bo
)
361 obj
= overlay
->old_vid_bo
->obj
;
362 i915_gem_object_unpin(obj
);
363 drm_gem_object_unreference(obj
);
364 overlay
->old_vid_bo
= NULL
;
369 struct put_image_params
{
386 static int packed_depth_bytes(u32 format
)
388 switch (format
& I915_OVERLAY_DEPTH_MASK
) {
389 case I915_OVERLAY_YUV422
:
391 case I915_OVERLAY_YUV411
:
392 /* return 6; not implemented */
398 static int packed_width_bytes(u32 format
, short width
)
400 switch (format
& I915_OVERLAY_DEPTH_MASK
) {
401 case I915_OVERLAY_YUV422
:
408 static int uv_hsubsampling(u32 format
)
410 switch (format
& I915_OVERLAY_DEPTH_MASK
) {
411 case I915_OVERLAY_YUV422
:
412 case I915_OVERLAY_YUV420
:
414 case I915_OVERLAY_YUV411
:
415 case I915_OVERLAY_YUV410
:
422 static int uv_vsubsampling(u32 format
)
424 switch (format
& I915_OVERLAY_DEPTH_MASK
) {
425 case I915_OVERLAY_YUV420
:
426 case I915_OVERLAY_YUV410
:
428 case I915_OVERLAY_YUV422
:
429 case I915_OVERLAY_YUV411
:
436 static u32
calc_swidthsw(struct drm_device
*dev
, u32 offset
, u32 width
)
438 u32 mask
, shift
, ret
;
446 ret
= ((offset
+ width
+ mask
) >> shift
) - (offset
>> shift
);
453 static const u16 y_static_hcoeffs
[N_HORIZ_Y_TAPS
* N_PHASES
] = {
454 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
455 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
456 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
457 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
458 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
459 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
460 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
461 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
462 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
463 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
464 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
465 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
466 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
467 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
468 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
469 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
470 0xb000, 0x3000, 0x0800, 0x3000, 0xb000};
471 static const u16 uv_static_hcoeffs
[N_HORIZ_UV_TAPS
* N_PHASES
] = {
472 0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
473 0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
474 0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
475 0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
476 0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
477 0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
478 0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
479 0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
480 0x3000, 0x0800, 0x3000};
482 static void update_polyphase_filter(struct overlay_registers
*regs
)
484 memcpy(regs
->Y_HCOEFS
, y_static_hcoeffs
, sizeof(y_static_hcoeffs
));
485 memcpy(regs
->UV_HCOEFS
, uv_static_hcoeffs
, sizeof(uv_static_hcoeffs
));
488 static bool update_scaling_factors(struct intel_overlay
*overlay
,
489 struct overlay_registers
*regs
,
490 struct put_image_params
*params
)
492 /* fixed point with a 12 bit shift */
493 u32 xscale
, yscale
, xscale_UV
, yscale_UV
;
495 #define FRACT_MASK 0xfff
496 bool scale_changed
= false;
497 int uv_hscale
= uv_hsubsampling(params
->format
);
498 int uv_vscale
= uv_vsubsampling(params
->format
);
500 if (params
->dst_w
> 1)
501 xscale
= ((params
->src_scan_w
- 1) << FP_SHIFT
)
504 xscale
= 1 << FP_SHIFT
;
506 if (params
->dst_h
> 1)
507 yscale
= ((params
->src_scan_h
- 1) << FP_SHIFT
)
510 yscale
= 1 << FP_SHIFT
;
512 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
513 xscale_UV
= xscale
/uv_hscale
;
514 yscale_UV
= yscale
/uv_vscale
;
515 /* make the Y scale to UV scale ratio an exact multiply */
516 xscale
= xscale_UV
* uv_hscale
;
517 yscale
= yscale_UV
* uv_vscale
;
523 if (xscale
!= overlay
->old_xscale
|| yscale
!= overlay
->old_yscale
)
524 scale_changed
= true;
525 overlay
->old_xscale
= xscale
;
526 overlay
->old_yscale
= yscale
;
528 regs
->YRGBSCALE
= ((yscale
& FRACT_MASK
) << 20)
529 | ((xscale
>> FP_SHIFT
) << 16)
530 | ((xscale
& FRACT_MASK
) << 3);
531 regs
->UVSCALE
= ((yscale_UV
& FRACT_MASK
) << 20)
532 | ((xscale_UV
>> FP_SHIFT
) << 16)
533 | ((xscale_UV
& FRACT_MASK
) << 3);
534 regs
->UVSCALEV
= ((yscale
>> FP_SHIFT
) << 16)
535 | ((yscale_UV
>> FP_SHIFT
) << 0);
538 update_polyphase_filter(regs
);
540 return scale_changed
;
543 static void update_colorkey(struct intel_overlay
*overlay
,
544 struct overlay_registers
*regs
)
546 u32 key
= overlay
->color_key
;
547 switch (overlay
->crtc
->base
.fb
->bits_per_pixel
) {
550 regs
->DCLRKM
= CLK_RGB8I_MASK
| DST_KEY_ENABLE
;
552 if (overlay
->crtc
->base
.fb
->depth
== 15) {
553 regs
->DCLRKV
= RGB15_TO_COLORKEY(key
);
554 regs
->DCLRKM
= CLK_RGB15_MASK
| DST_KEY_ENABLE
;
556 regs
->DCLRKV
= RGB16_TO_COLORKEY(key
);
557 regs
->DCLRKM
= CLK_RGB16_MASK
| DST_KEY_ENABLE
;
562 regs
->DCLRKM
= CLK_RGB24_MASK
| DST_KEY_ENABLE
;
566 static u32
overlay_cmd_reg(struct put_image_params
*params
)
568 u32 cmd
= OCMD_ENABLE
| OCMD_BUF_TYPE_FRAME
| OCMD_BUFFER0
;
570 if (params
->format
& I915_OVERLAY_YUV_PLANAR
) {
571 switch (params
->format
& I915_OVERLAY_DEPTH_MASK
) {
572 case I915_OVERLAY_YUV422
:
573 cmd
|= OCMD_YUV_422_PLANAR
;
575 case I915_OVERLAY_YUV420
:
576 cmd
|= OCMD_YUV_420_PLANAR
;
578 case I915_OVERLAY_YUV411
:
579 case I915_OVERLAY_YUV410
:
580 cmd
|= OCMD_YUV_410_PLANAR
;
583 } else { /* YUV packed */
584 switch (params
->format
& I915_OVERLAY_DEPTH_MASK
) {
585 case I915_OVERLAY_YUV422
:
586 cmd
|= OCMD_YUV_422_PACKED
;
588 case I915_OVERLAY_YUV411
:
589 cmd
|= OCMD_YUV_411_PACKED
;
593 switch (params
->format
& I915_OVERLAY_SWAP_MASK
) {
594 case I915_OVERLAY_NO_SWAP
:
596 case I915_OVERLAY_UV_SWAP
:
599 case I915_OVERLAY_Y_SWAP
:
602 case I915_OVERLAY_Y_AND_UV_SWAP
:
603 cmd
|= OCMD_Y_AND_UV_SWAP
;
611 int intel_overlay_do_put_image(struct intel_overlay
*overlay
,
612 struct drm_gem_object
*new_bo
,
613 struct put_image_params
*params
)
616 struct overlay_registers
*regs
;
617 bool scale_changed
= false;
618 struct drm_i915_gem_object
*bo_priv
= new_bo
->driver_private
;
619 struct drm_device
*dev
= overlay
->dev
;
621 BUG_ON(!mutex_is_locked(&dev
->struct_mutex
));
622 BUG_ON(!mutex_is_locked(&dev
->mode_config
.mutex
));
625 if (overlay
->hw_wedged
)
628 ret
= intel_overlay_release_old_vid(overlay
);
632 ret
= i915_gem_object_pin(new_bo
, PAGE_SIZE
);
636 ret
= i915_gem_object_set_to_gtt_domain(new_bo
, 0);
640 if (!overlay
->active
) {
641 regs
= intel_overlay_map_regs_atomic(overlay
);
646 regs
->OCONFIG
= OCONF_CC_OUT_8BIT
;
647 if (IS_I965GM(overlay
->dev
))
648 regs
->OCONFIG
|= OCONF_CSC_MODE_BT709
;
649 regs
->OCONFIG
|= overlay
->crtc
->pipe
== 0 ?
650 OCONF_PIPE_A
: OCONF_PIPE_B
;
651 intel_overlay_unmap_regs_atomic(overlay
);
653 ret
= intel_overlay_on(overlay
);
658 regs
= intel_overlay_map_regs_atomic(overlay
);
664 regs
->DWINPOS
= (params
->dst_y
<< 16) | params
->dst_x
;
665 regs
->DWINSZ
= (params
->dst_h
<< 16) | params
->dst_w
;
667 if (params
->format
& I915_OVERLAY_YUV_PACKED
)
668 tmp_width
= packed_width_bytes(params
->format
, params
->src_w
);
670 tmp_width
= params
->src_w
;
672 regs
->SWIDTH
= params
->src_w
;
673 regs
->SWIDTHSW
= calc_swidthsw(overlay
->dev
,
674 params
->offset_Y
, tmp_width
);
675 regs
->SHEIGHT
= params
->src_h
;
676 regs
->OBUF_0Y
= bo_priv
->gtt_offset
+ params
-> offset_Y
;
677 regs
->OSTRIDE
= params
->stride_Y
;
679 if (params
->format
& I915_OVERLAY_YUV_PLANAR
) {
680 int uv_hscale
= uv_hsubsampling(params
->format
);
681 int uv_vscale
= uv_vsubsampling(params
->format
);
683 regs
->SWIDTH
|= (params
->src_w
/uv_hscale
) << 16;
684 tmp_U
= calc_swidthsw(overlay
->dev
, params
->offset_U
,
685 params
->src_w
/uv_hscale
);
686 tmp_V
= calc_swidthsw(overlay
->dev
, params
->offset_V
,
687 params
->src_w
/uv_hscale
);
688 regs
->SWIDTHSW
|= max_t(u32
, tmp_U
, tmp_V
) << 16;
689 regs
->SHEIGHT
|= (params
->src_h
/uv_vscale
) << 16;
690 regs
->OBUF_0U
= bo_priv
->gtt_offset
+ params
->offset_U
;
691 regs
->OBUF_0V
= bo_priv
->gtt_offset
+ params
->offset_V
;
692 regs
->OSTRIDE
|= params
->stride_UV
<< 16;
695 scale_changed
= update_scaling_factors(overlay
, regs
, params
);
697 update_colorkey(overlay
, regs
);
699 regs
->OCMD
= overlay_cmd_reg(params
);
701 intel_overlay_unmap_regs_atomic(overlay
);
703 intel_overlay_continue(overlay
, scale_changed
);
705 overlay
->old_vid_bo
= overlay
->vid_bo
;
706 overlay
->vid_bo
= new_bo
->driver_private
;
711 i915_gem_object_unpin(new_bo
);
715 int intel_overlay_switch_off(struct intel_overlay
*overlay
)
718 struct overlay_registers
*regs
;
719 struct drm_gem_object
*obj
;
720 struct drm_device
*dev
= overlay
->dev
;
722 BUG_ON(!mutex_is_locked(&dev
->struct_mutex
));
723 BUG_ON(!mutex_is_locked(&dev
->mode_config
.mutex
));
725 if (!overlay
->active
)
728 if (overlay
->hw_wedged
)
731 ret
= intel_overlay_release_old_vid(overlay
);
735 regs
= intel_overlay_map_regs_atomic(overlay
);
737 intel_overlay_unmap_regs_atomic(overlay
);
739 ret
= intel_overlay_off(overlay
);
740 /* never have the overlay hw on without showing a frame */
741 BUG_ON(!overlay
->vid_bo
);
742 obj
= overlay
->vid_bo
->obj
;
744 i915_gem_object_unpin(obj
);
745 drm_gem_object_unreference(obj
);
746 overlay
->vid_bo
= NULL
;
748 overlay
->crtc
->overlay
= NULL
;
749 overlay
->crtc
= NULL
;
754 static int check_overlay_possible_on_crtc(struct intel_overlay
*overlay
,
755 struct intel_crtc
*crtc
)
757 drm_i915_private_t
*dev_priv
= overlay
->dev
->dev_private
;
759 int pipeconf_reg
= (crtc
->pipe
== 0) ? PIPEACONF
: PIPEBCONF
;
761 if (!crtc
->base
.enabled
|| crtc
->dpms_mode
!= DRM_MODE_DPMS_ON
)
764 pipeconf
= I915_READ(pipeconf_reg
);
766 /* can't use the overlay with double wide pipe */
767 if (!IS_I965G(overlay
->dev
) && pipeconf
& PIPEACONF_DOUBLE_WIDE
)
773 static void update_pfit_vscale_ratio(struct intel_overlay
*overlay
)
775 struct drm_device
*dev
= overlay
->dev
;
776 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
778 u32 pfit_control
= I915_READ(PFIT_CONTROL
);
780 /* XXX: This is not the same logic as in the xorg driver, but more in
781 * line with the intel documentation for the i965 */
782 if (!IS_I965G(dev
) && (pfit_control
& VERT_AUTO_SCALE
)) {
783 ratio
= I915_READ(PFIT_AUTO_RATIOS
) >> PFIT_VERT_SCALE_SHIFT
;
784 } else { /* on i965 use the PGM reg to read out the autoscaler values */
785 ratio
= I915_READ(PFIT_PGM_RATIOS
);
787 ratio
>>= PFIT_VERT_SCALE_SHIFT_965
;
789 ratio
>>= PFIT_VERT_SCALE_SHIFT
;
792 overlay
->pfit_vscale_ratio
= ratio
;
795 static int check_overlay_dst(struct intel_overlay
*overlay
,
796 struct drm_intel_overlay_put_image
*rec
)
798 struct drm_display_mode
*mode
= &overlay
->crtc
->base
.mode
;
800 if ((rec
->dst_x
< mode
->crtc_hdisplay
)
801 && (rec
->dst_x
+ rec
->dst_width
802 <= mode
->crtc_hdisplay
)
803 && (rec
->dst_y
< mode
->crtc_vdisplay
)
804 && (rec
->dst_y
+ rec
->dst_height
805 <= mode
->crtc_vdisplay
))
811 static int check_overlay_scaling(struct put_image_params
*rec
)
815 /* downscaling limit is 8.0 */
816 tmp
= ((rec
->src_scan_h
<< 16) / rec
->dst_h
) >> 16;
819 tmp
= ((rec
->src_scan_w
<< 16) / rec
->dst_w
) >> 16;
826 static int check_overlay_src(struct drm_device
*dev
,
827 struct drm_intel_overlay_put_image
*rec
,
828 struct drm_gem_object
*new_bo
)
832 int uv_hscale
= uv_hsubsampling(rec
->flags
);
833 int uv_vscale
= uv_vsubsampling(rec
->flags
);
836 /* check src dimensions */
837 if (IS_845G(dev
) || IS_I830(dev
)) {
838 if (rec
->src_height
> IMAGE_MAX_HEIGHT_LEGACY
839 || rec
->src_width
> IMAGE_MAX_WIDTH_LEGACY
)
842 if (rec
->src_height
> IMAGE_MAX_HEIGHT
843 || rec
->src_width
> IMAGE_MAX_WIDTH
)
846 /* better safe than sorry, use 4 as the maximal subsampling ratio */
847 if (rec
->src_height
< N_VERT_Y_TAPS
*4
848 || rec
->src_width
< N_HORIZ_Y_TAPS
*4)
851 /* check alingment constrains */
852 switch (rec
->flags
& I915_OVERLAY_TYPE_MASK
) {
853 case I915_OVERLAY_RGB
:
854 /* not implemented */
856 case I915_OVERLAY_YUV_PACKED
:
857 depth
= packed_depth_bytes(rec
->flags
);
862 /* ignore UV planes */
866 /* check pixel alignment */
867 if (rec
->offset_Y
% depth
)
870 case I915_OVERLAY_YUV_PLANAR
:
871 if (uv_vscale
< 0 || uv_hscale
< 0)
873 /* no offset restrictions for planar formats */
879 if (rec
->src_width
% uv_hscale
)
882 /* stride checking */
885 if (rec
->stride_Y
& stride_mask
|| rec
->stride_UV
& stride_mask
)
887 if (IS_I965G(dev
) && rec
->stride_Y
< 512)
890 tmp
= (rec
->flags
& I915_OVERLAY_TYPE_MASK
) == I915_OVERLAY_YUV_PLANAR
?
892 if (rec
->stride_Y
> tmp
*1024 || rec
->stride_UV
> 2*1024)
895 /* check buffer dimensions */
896 switch (rec
->flags
& I915_OVERLAY_TYPE_MASK
) {
897 case I915_OVERLAY_RGB
:
898 case I915_OVERLAY_YUV_PACKED
:
899 /* always 4 Y values per depth pixels */
900 if (packed_width_bytes(rec
->flags
, rec
->src_width
)
904 tmp
= rec
->stride_Y
*rec
->src_height
;
905 if (rec
->offset_Y
+ tmp
> new_bo
->size
)
908 case I915_OVERLAY_YUV_PLANAR
:
909 if (rec
->src_width
> rec
->stride_Y
)
911 if (rec
->src_width
/uv_hscale
> rec
->stride_UV
)
914 tmp
= rec
->stride_Y
*rec
->src_height
;
915 if (rec
->offset_Y
+ tmp
> new_bo
->size
)
917 tmp
= rec
->stride_UV
*rec
->src_height
;
919 if (rec
->offset_U
+ tmp
> new_bo
->size
920 || rec
->offset_V
+ tmp
> new_bo
->size
)
928 int intel_overlay_put_image(struct drm_device
*dev
, void *data
,
929 struct drm_file
*file_priv
)
931 struct drm_intel_overlay_put_image
*put_image_rec
= data
;
932 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
933 struct intel_overlay
*overlay
;
934 struct drm_mode_object
*drmmode_obj
;
935 struct intel_crtc
*crtc
;
936 struct drm_gem_object
*new_bo
;
937 struct put_image_params
*params
;
941 DRM_ERROR("called with no initialization\n");
945 overlay
= dev_priv
->overlay
;
947 DRM_DEBUG("userspace bug: no overlay\n");
951 if (!(put_image_rec
->flags
& I915_OVERLAY_ENABLE
)) {
952 mutex_lock(&dev
->mode_config
.mutex
);
953 mutex_lock(&dev
->struct_mutex
);
955 ret
= intel_overlay_switch_off(overlay
);
957 mutex_unlock(&dev
->struct_mutex
);
958 mutex_unlock(&dev
->mode_config
.mutex
);
963 params
= kmalloc(sizeof(struct put_image_params
), GFP_KERNEL
);
967 drmmode_obj
= drm_mode_object_find(dev
, put_image_rec
->crtc_id
,
968 DRM_MODE_OBJECT_CRTC
);
971 crtc
= to_intel_crtc(obj_to_crtc(drmmode_obj
));
973 new_bo
= drm_gem_object_lookup(dev
, file_priv
,
974 put_image_rec
->bo_handle
);
978 mutex_lock(&dev
->mode_config
.mutex
);
979 mutex_lock(&dev
->struct_mutex
);
981 if (overlay
->crtc
!= crtc
) {
982 struct drm_display_mode
*mode
= &crtc
->base
.mode
;
983 ret
= intel_overlay_switch_off(overlay
);
987 ret
= check_overlay_possible_on_crtc(overlay
, crtc
);
991 overlay
->crtc
= crtc
;
992 crtc
->overlay
= overlay
;
994 if (intel_panel_fitter_pipe(dev
) == crtc
->pipe
995 /* and line to wide, i.e. one-line-mode */
996 && mode
->hdisplay
> 1024) {
997 overlay
->pfit_active
= 1;
998 update_pfit_vscale_ratio(overlay
);
1000 overlay
->pfit_active
= 0;
1003 ret
= check_overlay_dst(overlay
, put_image_rec
);
1007 if (overlay
->pfit_active
) {
1008 params
->dst_y
= ((((u32
)put_image_rec
->dst_y
) << 12) /
1009 overlay
->pfit_vscale_ratio
);
1010 /* shifting right rounds downwards, so add 1 */
1011 params
->dst_h
= ((((u32
)put_image_rec
->dst_height
) << 12) /
1012 overlay
->pfit_vscale_ratio
) + 1;
1014 params
->dst_y
= put_image_rec
->dst_y
;
1015 params
->dst_h
= put_image_rec
->dst_height
;
1017 params
->dst_x
= put_image_rec
->dst_x
;
1018 params
->dst_w
= put_image_rec
->dst_width
;
1020 params
->src_w
= put_image_rec
->src_width
;
1021 params
->src_h
= put_image_rec
->src_height
;
1022 params
->src_scan_w
= put_image_rec
->src_scan_width
;
1023 params
->src_scan_h
= put_image_rec
->src_scan_height
;
1024 if (params
->src_scan_h
> params
->src_h
1025 || params
->src_scan_w
> params
->src_w
) {
1030 ret
= check_overlay_src(dev
, put_image_rec
, new_bo
);
1033 params
->format
= put_image_rec
->flags
& ~I915_OVERLAY_FLAGS_MASK
;
1034 params
->stride_Y
= put_image_rec
->stride_Y
;
1035 params
->stride_UV
= put_image_rec
->stride_UV
;
1036 params
->offset_Y
= put_image_rec
->offset_Y
;
1037 params
->offset_U
= put_image_rec
->offset_U
;
1038 params
->offset_V
= put_image_rec
->offset_V
;
1040 /* Check scaling after src size to prevent a divide-by-zero. */
1041 ret
= check_overlay_scaling(params
);
1045 ret
= intel_overlay_do_put_image(overlay
, new_bo
, params
);
1049 mutex_unlock(&dev
->struct_mutex
);
1050 mutex_unlock(&dev
->mode_config
.mutex
);
1057 mutex_unlock(&dev
->struct_mutex
);
1058 mutex_unlock(&dev
->mode_config
.mutex
);
1059 drm_gem_object_unreference(new_bo
);
1065 static void update_reg_attrs(struct intel_overlay
*overlay
,
1066 struct overlay_registers
*regs
)
1068 regs
->OCLRC0
= (overlay
->contrast
<< 18) | (overlay
->brightness
& 0xff);
1069 regs
->OCLRC1
= overlay
->saturation
;
1072 static bool check_gamma_bounds(u32 gamma1
, u32 gamma2
)
1076 if (gamma1
& 0xff000000 || gamma2
& 0xff000000)
1079 for (i
= 0; i
< 3; i
++) {
1080 if (((gamma1
>> i
* 8) & 0xff) >= ((gamma2
>> i
*8) & 0xff))
1087 static bool check_gamma5_errata(u32 gamma5
)
1091 for (i
= 0; i
< 3; i
++) {
1092 if (((gamma5
>> i
*8) & 0xff) == 0x80)
1099 static int check_gamma(struct drm_intel_overlay_attrs
*attrs
)
1101 if (!check_gamma_bounds(0, attrs
->gamma0
)
1102 || !check_gamma_bounds(attrs
->gamma0
, attrs
->gamma1
)
1103 || !check_gamma_bounds(attrs
->gamma1
, attrs
->gamma2
)
1104 || !check_gamma_bounds(attrs
->gamma2
, attrs
->gamma3
)
1105 || !check_gamma_bounds(attrs
->gamma3
, attrs
->gamma4
)
1106 || !check_gamma_bounds(attrs
->gamma4
, attrs
->gamma5
)
1107 || !check_gamma_bounds(attrs
->gamma5
, 0x00ffffff))
1109 if (!check_gamma5_errata(attrs
->gamma5
))
1114 int intel_overlay_attrs(struct drm_device
*dev
, void *data
,
1115 struct drm_file
*file_priv
)
1117 struct drm_intel_overlay_attrs
*attrs
= data
;
1118 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
1119 struct intel_overlay
*overlay
;
1120 struct overlay_registers
*regs
;
1124 DRM_ERROR("called with no initialization\n");
1128 overlay
= dev_priv
->overlay
;
1130 DRM_DEBUG("userspace bug: no overlay\n");
1134 mutex_lock(&dev
->mode_config
.mutex
);
1135 mutex_lock(&dev
->struct_mutex
);
1137 if (!(attrs
->flags
& I915_OVERLAY_UPDATE_ATTRS
)) {
1138 attrs
->color_key
= overlay
->color_key
;
1139 attrs
->brightness
= overlay
->brightness
;
1140 attrs
->contrast
= overlay
->contrast
;
1141 attrs
->saturation
= overlay
->saturation
;
1144 attrs
->gamma0
= I915_READ(OGAMC0
);
1145 attrs
->gamma1
= I915_READ(OGAMC1
);
1146 attrs
->gamma2
= I915_READ(OGAMC2
);
1147 attrs
->gamma3
= I915_READ(OGAMC3
);
1148 attrs
->gamma4
= I915_READ(OGAMC4
);
1149 attrs
->gamma5
= I915_READ(OGAMC5
);
1153 overlay
->color_key
= attrs
->color_key
;
1154 if (attrs
->brightness
>= -128 && attrs
->brightness
<= 127) {
1155 overlay
->brightness
= attrs
->brightness
;
1160 if (attrs
->contrast
<= 255) {
1161 overlay
->contrast
= attrs
->contrast
;
1166 if (attrs
->saturation
<= 1023) {
1167 overlay
->saturation
= attrs
->saturation
;
1173 regs
= intel_overlay_map_regs_atomic(overlay
);
1179 update_reg_attrs(overlay
, regs
);
1181 intel_overlay_unmap_regs_atomic(overlay
);
1183 if (attrs
->flags
& I915_OVERLAY_UPDATE_GAMMA
) {
1184 if (!IS_I9XX(dev
)) {
1189 if (overlay
->active
) {
1194 ret
= check_gamma(attrs
);
1198 I915_WRITE(OGAMC0
, attrs
->gamma0
);
1199 I915_WRITE(OGAMC1
, attrs
->gamma1
);
1200 I915_WRITE(OGAMC2
, attrs
->gamma2
);
1201 I915_WRITE(OGAMC3
, attrs
->gamma3
);
1202 I915_WRITE(OGAMC4
, attrs
->gamma4
);
1203 I915_WRITE(OGAMC5
, attrs
->gamma5
);
1209 mutex_unlock(&dev
->struct_mutex
);
1210 mutex_unlock(&dev
->mode_config
.mutex
);
1215 void intel_setup_overlay(struct drm_device
*dev
)
1217 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
1218 struct intel_overlay
*overlay
;
1219 struct drm_gem_object
*reg_bo
;
1220 struct overlay_registers
*regs
;
1223 if (!OVERLAY_EXISTS(dev
))
1226 overlay
= kzalloc(sizeof(struct intel_overlay
), GFP_KERNEL
);
1231 reg_bo
= drm_gem_object_alloc(dev
, PAGE_SIZE
);
1234 overlay
->reg_bo
= reg_bo
->driver_private
;
1236 if (OVERLAY_NONPHYSICAL(dev
)) {
1237 ret
= i915_gem_object_pin(reg_bo
, PAGE_SIZE
);
1239 DRM_ERROR("failed to pin overlay register bo\n");
1242 overlay
->flip_addr
= overlay
->reg_bo
->gtt_offset
;
1244 ret
= i915_gem_attach_phys_object(dev
, reg_bo
,
1245 I915_GEM_PHYS_OVERLAY_REGS
);
1247 DRM_ERROR("failed to attach phys overlay regs\n");
1250 overlay
->flip_addr
= overlay
->reg_bo
->phys_obj
->handle
->busaddr
;
1253 /* init all values */
1254 overlay
->color_key
= 0x0101fe;
1255 overlay
->brightness
= -19;
1256 overlay
->contrast
= 75;
1257 overlay
->saturation
= 146;
1259 regs
= intel_overlay_map_regs_atomic(overlay
);
1263 memset(regs
, 0, sizeof(struct overlay_registers
));
1264 update_polyphase_filter(regs
);
1266 update_reg_attrs(overlay
, regs
);
1268 intel_overlay_unmap_regs_atomic(overlay
);
1270 dev_priv
->overlay
= overlay
;
1271 DRM_INFO("initialized overlay support\n");
1275 drm_gem_object_unreference(reg_bo
);
1281 void intel_cleanup_overlay(struct drm_device
*dev
)
1283 drm_i915_private_t
*dev_priv
= dev
->dev_private
;
1285 if (dev_priv
->overlay
) {
1286 /* The bo's should be free'd by the generic code already.
1287 * Furthermore modesetting teardown happens beforehand so the
1288 * hardware should be off already */
1289 BUG_ON(dev_priv
->overlay
->active
);
1291 kfree(dev_priv
->overlay
);