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
29 #include <drm/drm_fourcc.h>
30 #include <drm/i915_drm.h>
32 #include "gem/i915_gem_pm.h"
33 #include "gt/intel_ring.h"
37 #include "intel_display_types.h"
38 #include "intel_frontbuffer.h"
39 #include "intel_overlay.h"
41 /* Limits for overlay size. According to intel doc, the real limits are:
42 * Y width: 4095, UV width (planar): 2047, Y height: 2047,
43 * UV width (planar): * 1023. But the xorg thinks 2048 for height and width. Use
44 * the mininum of both. */
45 #define IMAGE_MAX_WIDTH 2048
46 #define IMAGE_MAX_HEIGHT 2046 /* 2 * 1023 */
47 /* on 830 and 845 these large limits result in the card hanging */
48 #define IMAGE_MAX_WIDTH_LEGACY 1024
49 #define IMAGE_MAX_HEIGHT_LEGACY 1088
51 /* overlay register definitions */
53 #define OCMD_TILED_SURFACE (0x1<<19)
54 #define OCMD_MIRROR_MASK (0x3<<17)
55 #define OCMD_MIRROR_MODE (0x3<<17)
56 #define OCMD_MIRROR_HORIZONTAL (0x1<<17)
57 #define OCMD_MIRROR_VERTICAL (0x2<<17)
58 #define OCMD_MIRROR_BOTH (0x3<<17)
59 #define OCMD_BYTEORDER_MASK (0x3<<14) /* zero for YUYV or FOURCC YUY2 */
60 #define OCMD_UV_SWAP (0x1<<14) /* YVYU */
61 #define OCMD_Y_SWAP (0x2<<14) /* UYVY or FOURCC UYVY */
62 #define OCMD_Y_AND_UV_SWAP (0x3<<14) /* VYUY */
63 #define OCMD_SOURCE_FORMAT_MASK (0xf<<10)
64 #define OCMD_RGB_888 (0x1<<10) /* not in i965 Intel docs */
65 #define OCMD_RGB_555 (0x2<<10) /* not in i965 Intel docs */
66 #define OCMD_RGB_565 (0x3<<10) /* not in i965 Intel docs */
67 #define OCMD_YUV_422_PACKED (0x8<<10)
68 #define OCMD_YUV_411_PACKED (0x9<<10) /* not in i965 Intel docs */
69 #define OCMD_YUV_420_PLANAR (0xc<<10)
70 #define OCMD_YUV_422_PLANAR (0xd<<10)
71 #define OCMD_YUV_410_PLANAR (0xe<<10) /* also 411 */
72 #define OCMD_TVSYNCFLIP_PARITY (0x1<<9)
73 #define OCMD_TVSYNCFLIP_ENABLE (0x1<<7)
74 #define OCMD_BUF_TYPE_MASK (0x1<<5)
75 #define OCMD_BUF_TYPE_FRAME (0x0<<5)
76 #define OCMD_BUF_TYPE_FIELD (0x1<<5)
77 #define OCMD_TEST_MODE (0x1<<4)
78 #define OCMD_BUFFER_SELECT (0x3<<2)
79 #define OCMD_BUFFER0 (0x0<<2)
80 #define OCMD_BUFFER1 (0x1<<2)
81 #define OCMD_FIELD_SELECT (0x1<<2)
82 #define OCMD_FIELD0 (0x0<<1)
83 #define OCMD_FIELD1 (0x1<<1)
84 #define OCMD_ENABLE (0x1<<0)
86 /* OCONFIG register */
87 #define OCONF_PIPE_MASK (0x1<<18)
88 #define OCONF_PIPE_A (0x0<<18)
89 #define OCONF_PIPE_B (0x1<<18)
90 #define OCONF_GAMMA2_ENABLE (0x1<<16)
91 #define OCONF_CSC_MODE_BT601 (0x0<<5)
92 #define OCONF_CSC_MODE_BT709 (0x1<<5)
93 #define OCONF_CSC_BYPASS (0x1<<4)
94 #define OCONF_CC_OUT_8BIT (0x1<<3)
95 #define OCONF_TEST_MODE (0x1<<2)
96 #define OCONF_THREE_LINE_BUFFER (0x1<<0)
97 #define OCONF_TWO_LINE_BUFFER (0x0<<0)
99 /* DCLRKM (dst-key) register */
100 #define DST_KEY_ENABLE (0x1<<31)
101 #define CLK_RGB24_MASK 0x0
102 #define CLK_RGB16_MASK 0x070307
103 #define CLK_RGB15_MASK 0x070707
104 #define CLK_RGB8I_MASK 0xffffff
106 #define RGB16_TO_COLORKEY(c) \
107 (((c & 0xF800) << 8) | ((c & 0x07E0) << 5) | ((c & 0x001F) << 3))
108 #define RGB15_TO_COLORKEY(c) \
109 (((c & 0x7c00) << 9) | ((c & 0x03E0) << 6) | ((c & 0x001F) << 3))
111 /* overlay flip addr flag */
112 #define OFC_UPDATE 0x1
114 /* polyphase filter coefficients */
115 #define N_HORIZ_Y_TAPS 5
116 #define N_VERT_Y_TAPS 3
117 #define N_HORIZ_UV_TAPS 3
118 #define N_VERT_UV_TAPS 3
122 /* memory bufferd overlay registers */
123 struct overlay_registers
{
151 u32 RESERVED1
; /* 0x6C */
164 u32 FASTHSCALE
; /* 0xA0 */
165 u32 UVSCALEV
; /* 0xA4 */
166 u32 RESERVEDC
[(0x200 - 0xA8) / 4]; /* 0xA8 - 0x1FC */
167 u16 Y_VCOEFS
[N_VERT_Y_TAPS
* N_PHASES
]; /* 0x200 */
168 u16 RESERVEDD
[0x100 / 2 - N_VERT_Y_TAPS
* N_PHASES
];
169 u16 Y_HCOEFS
[N_HORIZ_Y_TAPS
* N_PHASES
]; /* 0x300 */
170 u16 RESERVEDE
[0x200 / 2 - N_HORIZ_Y_TAPS
* N_PHASES
];
171 u16 UV_VCOEFS
[N_VERT_UV_TAPS
* N_PHASES
]; /* 0x500 */
172 u16 RESERVEDF
[0x100 / 2 - N_VERT_UV_TAPS
* N_PHASES
];
173 u16 UV_HCOEFS
[N_HORIZ_UV_TAPS
* N_PHASES
]; /* 0x600 */
174 u16 RESERVEDG
[0x100 / 2 - N_HORIZ_UV_TAPS
* N_PHASES
];
177 struct intel_overlay
{
178 struct drm_i915_private
*i915
;
179 struct intel_context
*context
;
180 struct intel_crtc
*crtc
;
181 struct i915_vma
*vma
;
182 struct i915_vma
*old_vma
;
185 u32 pfit_vscale_ratio
; /* shifted-point number, (1<<12) == 1.0 */
187 u32 color_key_enabled
:1;
188 u32 brightness
, contrast
, saturation
;
189 u32 old_xscale
, old_yscale
;
190 /* register access */
191 struct drm_i915_gem_object
*reg_bo
;
192 struct overlay_registers __iomem
*regs
;
195 struct i915_active last_flip
;
196 void (*flip_complete
)(struct intel_overlay
*ovl
);
199 static void i830_overlay_clock_gating(struct drm_i915_private
*dev_priv
,
202 struct pci_dev
*pdev
= dev_priv
->drm
.pdev
;
205 /* WA_OVERLAY_CLKGATE:alm */
207 I915_WRITE(DSPCLK_GATE_D
, 0);
209 I915_WRITE(DSPCLK_GATE_D
, OVRUNIT_CLOCK_GATE_DISABLE
);
211 /* WA_DISABLE_L2CACHE_CLOCK_GATING:alm */
212 pci_bus_read_config_byte(pdev
->bus
,
213 PCI_DEVFN(0, 0), I830_CLOCK_GATE
, &val
);
215 val
&= ~I830_L2_CACHE_CLOCK_GATE_DISABLE
;
217 val
|= I830_L2_CACHE_CLOCK_GATE_DISABLE
;
218 pci_bus_write_config_byte(pdev
->bus
,
219 PCI_DEVFN(0, 0), I830_CLOCK_GATE
, val
);
222 static struct i915_request
*
223 alloc_request(struct intel_overlay
*overlay
, void (*fn
)(struct intel_overlay
*))
225 struct i915_request
*rq
;
228 overlay
->flip_complete
= fn
;
230 rq
= i915_request_create(overlay
->context
);
234 err
= i915_active_add_request(&overlay
->last_flip
, rq
);
236 i915_request_add(rq
);
243 /* overlay needs to be disable in OCMD reg */
244 static int intel_overlay_on(struct intel_overlay
*overlay
)
246 struct drm_i915_private
*dev_priv
= overlay
->i915
;
247 struct i915_request
*rq
;
250 WARN_ON(overlay
->active
);
252 rq
= alloc_request(overlay
, NULL
);
256 cs
= intel_ring_begin(rq
, 4);
258 i915_request_add(rq
);
262 overlay
->active
= true;
264 if (IS_I830(dev_priv
))
265 i830_overlay_clock_gating(dev_priv
, false);
267 *cs
++ = MI_OVERLAY_FLIP
| MI_OVERLAY_ON
;
268 *cs
++ = overlay
->flip_addr
| OFC_UPDATE
;
269 *cs
++ = MI_WAIT_FOR_EVENT
| MI_WAIT_FOR_OVERLAY_FLIP
;
271 intel_ring_advance(rq
, cs
);
273 i915_request_add(rq
);
275 return i915_active_wait(&overlay
->last_flip
);
278 static void intel_overlay_flip_prepare(struct intel_overlay
*overlay
,
279 struct i915_vma
*vma
)
281 enum pipe pipe
= overlay
->crtc
->pipe
;
282 struct intel_frontbuffer
*from
= NULL
, *to
= NULL
;
284 WARN_ON(overlay
->old_vma
);
287 from
= intel_frontbuffer_get(overlay
->vma
->obj
);
289 to
= intel_frontbuffer_get(vma
->obj
);
291 intel_frontbuffer_track(from
, to
, INTEL_FRONTBUFFER_OVERLAY(pipe
));
294 intel_frontbuffer_put(to
);
296 intel_frontbuffer_put(from
);
298 intel_frontbuffer_flip_prepare(overlay
->i915
,
299 INTEL_FRONTBUFFER_OVERLAY(pipe
));
301 overlay
->old_vma
= overlay
->vma
;
303 overlay
->vma
= i915_vma_get(vma
);
308 /* overlay needs to be enabled in OCMD reg */
309 static int intel_overlay_continue(struct intel_overlay
*overlay
,
310 struct i915_vma
*vma
,
311 bool load_polyphase_filter
)
313 struct drm_i915_private
*dev_priv
= overlay
->i915
;
314 struct i915_request
*rq
;
315 u32 flip_addr
= overlay
->flip_addr
;
318 WARN_ON(!overlay
->active
);
320 if (load_polyphase_filter
)
321 flip_addr
|= OFC_UPDATE
;
323 /* check for underruns */
324 tmp
= I915_READ(DOVSTA
);
326 DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp
);
328 rq
= alloc_request(overlay
, NULL
);
332 cs
= intel_ring_begin(rq
, 2);
334 i915_request_add(rq
);
338 *cs
++ = MI_OVERLAY_FLIP
| MI_OVERLAY_CONTINUE
;
340 intel_ring_advance(rq
, cs
);
342 intel_overlay_flip_prepare(overlay
, vma
);
343 i915_request_add(rq
);
348 static void intel_overlay_release_old_vma(struct intel_overlay
*overlay
)
350 struct i915_vma
*vma
;
352 vma
= fetch_and_zero(&overlay
->old_vma
);
356 intel_frontbuffer_flip_complete(overlay
->i915
,
357 INTEL_FRONTBUFFER_OVERLAY(overlay
->crtc
->pipe
));
359 i915_gem_object_unpin_from_display_plane(vma
);
364 intel_overlay_release_old_vid_tail(struct intel_overlay
*overlay
)
366 intel_overlay_release_old_vma(overlay
);
369 static void intel_overlay_off_tail(struct intel_overlay
*overlay
)
371 struct drm_i915_private
*dev_priv
= overlay
->i915
;
373 intel_overlay_release_old_vma(overlay
);
375 overlay
->crtc
->overlay
= NULL
;
376 overlay
->crtc
= NULL
;
377 overlay
->active
= false;
379 if (IS_I830(dev_priv
))
380 i830_overlay_clock_gating(dev_priv
, true);
384 intel_overlay_last_flip_retire(struct i915_active
*active
)
386 struct intel_overlay
*overlay
=
387 container_of(active
, typeof(*overlay
), last_flip
);
389 if (overlay
->flip_complete
)
390 overlay
->flip_complete(overlay
);
393 /* overlay needs to be disabled in OCMD reg */
394 static int intel_overlay_off(struct intel_overlay
*overlay
)
396 struct i915_request
*rq
;
397 u32
*cs
, flip_addr
= overlay
->flip_addr
;
399 WARN_ON(!overlay
->active
);
401 /* According to intel docs the overlay hw may hang (when switching
402 * off) without loading the filter coeffs. It is however unclear whether
403 * this applies to the disabling of the overlay or to the switching off
404 * of the hw. Do it in both cases */
405 flip_addr
|= OFC_UPDATE
;
407 rq
= alloc_request(overlay
, intel_overlay_off_tail
);
411 cs
= intel_ring_begin(rq
, 6);
413 i915_request_add(rq
);
417 /* wait for overlay to go idle */
418 *cs
++ = MI_OVERLAY_FLIP
| MI_OVERLAY_CONTINUE
;
420 *cs
++ = MI_WAIT_FOR_EVENT
| MI_WAIT_FOR_OVERLAY_FLIP
;
422 /* turn overlay off */
423 *cs
++ = MI_OVERLAY_FLIP
| MI_OVERLAY_OFF
;
425 *cs
++ = MI_WAIT_FOR_EVENT
| MI_WAIT_FOR_OVERLAY_FLIP
;
427 intel_ring_advance(rq
, cs
);
429 intel_overlay_flip_prepare(overlay
, NULL
);
430 i915_request_add(rq
);
432 return i915_active_wait(&overlay
->last_flip
);
435 /* recover from an interruption due to a signal
436 * We have to be careful not to repeat work forever an make forward progess. */
437 static int intel_overlay_recover_from_interrupt(struct intel_overlay
*overlay
)
439 return i915_active_wait(&overlay
->last_flip
);
442 /* Wait for pending overlay flip and release old frame.
443 * Needs to be called before the overlay register are changed
444 * via intel_overlay_(un)map_regs
446 static int intel_overlay_release_old_vid(struct intel_overlay
*overlay
)
448 struct drm_i915_private
*dev_priv
= overlay
->i915
;
449 struct i915_request
*rq
;
453 * Only wait if there is actually an old frame to release to
454 * guarantee forward progress.
456 if (!overlay
->old_vma
)
459 if (!(I915_READ(GEN2_ISR
) & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT
)) {
460 intel_overlay_release_old_vid_tail(overlay
);
464 rq
= alloc_request(overlay
, intel_overlay_release_old_vid_tail
);
468 cs
= intel_ring_begin(rq
, 2);
470 i915_request_add(rq
);
474 *cs
++ = MI_WAIT_FOR_EVENT
| MI_WAIT_FOR_OVERLAY_FLIP
;
476 intel_ring_advance(rq
, cs
);
478 i915_request_add(rq
);
480 return i915_active_wait(&overlay
->last_flip
);
483 void intel_overlay_reset(struct drm_i915_private
*dev_priv
)
485 struct intel_overlay
*overlay
= dev_priv
->overlay
;
490 overlay
->old_xscale
= 0;
491 overlay
->old_yscale
= 0;
492 overlay
->crtc
= NULL
;
493 overlay
->active
= false;
496 static int packed_depth_bytes(u32 format
)
498 switch (format
& I915_OVERLAY_DEPTH_MASK
) {
499 case I915_OVERLAY_YUV422
:
501 case I915_OVERLAY_YUV411
:
502 /* return 6; not implemented */
508 static int packed_width_bytes(u32 format
, short width
)
510 switch (format
& I915_OVERLAY_DEPTH_MASK
) {
511 case I915_OVERLAY_YUV422
:
518 static int uv_hsubsampling(u32 format
)
520 switch (format
& I915_OVERLAY_DEPTH_MASK
) {
521 case I915_OVERLAY_YUV422
:
522 case I915_OVERLAY_YUV420
:
524 case I915_OVERLAY_YUV411
:
525 case I915_OVERLAY_YUV410
:
532 static int uv_vsubsampling(u32 format
)
534 switch (format
& I915_OVERLAY_DEPTH_MASK
) {
535 case I915_OVERLAY_YUV420
:
536 case I915_OVERLAY_YUV410
:
538 case I915_OVERLAY_YUV422
:
539 case I915_OVERLAY_YUV411
:
546 static u32
calc_swidthsw(struct drm_i915_private
*dev_priv
, u32 offset
, u32 width
)
550 if (IS_GEN(dev_priv
, 2))
551 sw
= ALIGN((offset
& 31) + width
, 32);
553 sw
= ALIGN((offset
& 63) + width
, 64);
558 return (sw
- 32) >> 3;
561 static const u16 y_static_hcoeffs
[N_PHASES
][N_HORIZ_Y_TAPS
] = {
562 [ 0] = { 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0, },
563 [ 1] = { 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440, },
564 [ 2] = { 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0, },
565 [ 3] = { 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380, },
566 [ 4] = { 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320, },
567 [ 5] = { 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0, },
568 [ 6] = { 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260, },
569 [ 7] = { 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200, },
570 [ 8] = { 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0, },
571 [ 9] = { 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160, },
572 [10] = { 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120, },
573 [11] = { 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0, },
574 [12] = { 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0, },
575 [13] = { 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060, },
576 [14] = { 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040, },
577 [15] = { 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020, },
578 [16] = { 0xb000, 0x3000, 0x0800, 0x3000, 0xb000, },
581 static const u16 uv_static_hcoeffs
[N_PHASES
][N_HORIZ_UV_TAPS
] = {
582 [ 0] = { 0x3000, 0x1800, 0x1800, },
583 [ 1] = { 0xb000, 0x18d0, 0x2e60, },
584 [ 2] = { 0xb000, 0x1990, 0x2ce0, },
585 [ 3] = { 0xb020, 0x1a68, 0x2b40, },
586 [ 4] = { 0xb040, 0x1b20, 0x29e0, },
587 [ 5] = { 0xb060, 0x1bd8, 0x2880, },
588 [ 6] = { 0xb080, 0x1c88, 0x3e60, },
589 [ 7] = { 0xb0a0, 0x1d28, 0x3c00, },
590 [ 8] = { 0xb0c0, 0x1db8, 0x39e0, },
591 [ 9] = { 0xb0e0, 0x1e40, 0x37e0, },
592 [10] = { 0xb100, 0x1eb8, 0x3620, },
593 [11] = { 0xb100, 0x1f18, 0x34a0, },
594 [12] = { 0xb100, 0x1f68, 0x3360, },
595 [13] = { 0xb0e0, 0x1fa8, 0x3240, },
596 [14] = { 0xb0c0, 0x1fe0, 0x3140, },
597 [15] = { 0xb060, 0x1ff0, 0x30a0, },
598 [16] = { 0x3000, 0x0800, 0x3000, },
601 static void update_polyphase_filter(struct overlay_registers __iomem
*regs
)
603 memcpy_toio(regs
->Y_HCOEFS
, y_static_hcoeffs
, sizeof(y_static_hcoeffs
));
604 memcpy_toio(regs
->UV_HCOEFS
, uv_static_hcoeffs
,
605 sizeof(uv_static_hcoeffs
));
608 static bool update_scaling_factors(struct intel_overlay
*overlay
,
609 struct overlay_registers __iomem
*regs
,
610 struct drm_intel_overlay_put_image
*params
)
612 /* fixed point with a 12 bit shift */
613 u32 xscale
, yscale
, xscale_UV
, yscale_UV
;
615 #define FRACT_MASK 0xfff
616 bool scale_changed
= false;
617 int uv_hscale
= uv_hsubsampling(params
->flags
);
618 int uv_vscale
= uv_vsubsampling(params
->flags
);
620 if (params
->dst_width
> 1)
621 xscale
= ((params
->src_scan_width
- 1) << FP_SHIFT
) /
624 xscale
= 1 << FP_SHIFT
;
626 if (params
->dst_height
> 1)
627 yscale
= ((params
->src_scan_height
- 1) << FP_SHIFT
) /
630 yscale
= 1 << FP_SHIFT
;
632 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
633 xscale_UV
= xscale
/uv_hscale
;
634 yscale_UV
= yscale
/uv_vscale
;
635 /* make the Y scale to UV scale ratio an exact multiply */
636 xscale
= xscale_UV
* uv_hscale
;
637 yscale
= yscale_UV
* uv_vscale
;
643 if (xscale
!= overlay
->old_xscale
|| yscale
!= overlay
->old_yscale
)
644 scale_changed
= true;
645 overlay
->old_xscale
= xscale
;
646 overlay
->old_yscale
= yscale
;
648 iowrite32(((yscale
& FRACT_MASK
) << 20) |
649 ((xscale
>> FP_SHIFT
) << 16) |
650 ((xscale
& FRACT_MASK
) << 3),
653 iowrite32(((yscale_UV
& FRACT_MASK
) << 20) |
654 ((xscale_UV
>> FP_SHIFT
) << 16) |
655 ((xscale_UV
& FRACT_MASK
) << 3),
658 iowrite32((((yscale
>> FP_SHIFT
) << 16) |
659 ((yscale_UV
>> FP_SHIFT
) << 0)),
663 update_polyphase_filter(regs
);
665 return scale_changed
;
668 static void update_colorkey(struct intel_overlay
*overlay
,
669 struct overlay_registers __iomem
*regs
)
671 const struct intel_plane_state
*state
=
672 to_intel_plane_state(overlay
->crtc
->base
.primary
->state
);
673 u32 key
= overlay
->color_key
;
677 if (overlay
->color_key_enabled
)
678 flags
|= DST_KEY_ENABLE
;
680 if (state
->uapi
.visible
)
681 format
= state
->hw
.fb
->format
->format
;
686 flags
|= CLK_RGB8I_MASK
;
688 case DRM_FORMAT_XRGB1555
:
689 key
= RGB15_TO_COLORKEY(key
);
690 flags
|= CLK_RGB15_MASK
;
692 case DRM_FORMAT_RGB565
:
693 key
= RGB16_TO_COLORKEY(key
);
694 flags
|= CLK_RGB16_MASK
;
697 flags
|= CLK_RGB24_MASK
;
701 iowrite32(key
, ®s
->DCLRKV
);
702 iowrite32(flags
, ®s
->DCLRKM
);
705 static u32
overlay_cmd_reg(struct drm_intel_overlay_put_image
*params
)
707 u32 cmd
= OCMD_ENABLE
| OCMD_BUF_TYPE_FRAME
| OCMD_BUFFER0
;
709 if (params
->flags
& I915_OVERLAY_YUV_PLANAR
) {
710 switch (params
->flags
& I915_OVERLAY_DEPTH_MASK
) {
711 case I915_OVERLAY_YUV422
:
712 cmd
|= OCMD_YUV_422_PLANAR
;
714 case I915_OVERLAY_YUV420
:
715 cmd
|= OCMD_YUV_420_PLANAR
;
717 case I915_OVERLAY_YUV411
:
718 case I915_OVERLAY_YUV410
:
719 cmd
|= OCMD_YUV_410_PLANAR
;
722 } else { /* YUV packed */
723 switch (params
->flags
& I915_OVERLAY_DEPTH_MASK
) {
724 case I915_OVERLAY_YUV422
:
725 cmd
|= OCMD_YUV_422_PACKED
;
727 case I915_OVERLAY_YUV411
:
728 cmd
|= OCMD_YUV_411_PACKED
;
732 switch (params
->flags
& I915_OVERLAY_SWAP_MASK
) {
733 case I915_OVERLAY_NO_SWAP
:
735 case I915_OVERLAY_UV_SWAP
:
738 case I915_OVERLAY_Y_SWAP
:
741 case I915_OVERLAY_Y_AND_UV_SWAP
:
742 cmd
|= OCMD_Y_AND_UV_SWAP
;
750 static int intel_overlay_do_put_image(struct intel_overlay
*overlay
,
751 struct drm_i915_gem_object
*new_bo
,
752 struct drm_intel_overlay_put_image
*params
)
754 struct overlay_registers __iomem
*regs
= overlay
->regs
;
755 struct drm_i915_private
*dev_priv
= overlay
->i915
;
756 u32 swidth
, swidthsw
, sheight
, ostride
;
757 enum pipe pipe
= overlay
->crtc
->pipe
;
758 bool scale_changed
= false;
759 struct i915_vma
*vma
;
762 WARN_ON(!drm_modeset_is_locked(&dev_priv
->drm
.mode_config
.connection_mutex
));
764 ret
= intel_overlay_release_old_vid(overlay
);
768 atomic_inc(&dev_priv
->gpu_error
.pending_fb_pin
);
770 vma
= i915_gem_object_pin_to_display_plane(new_bo
,
771 0, NULL
, PIN_MAPPABLE
);
774 goto out_pin_section
;
776 i915_gem_object_flush_frontbuffer(new_bo
, ORIGIN_DIRTYFB
);
778 if (!overlay
->active
) {
781 oconfig
= OCONF_CC_OUT_8BIT
;
782 if (IS_GEN(dev_priv
, 4))
783 oconfig
|= OCONF_CSC_MODE_BT709
;
784 oconfig
|= pipe
== 0 ?
785 OCONF_PIPE_A
: OCONF_PIPE_B
;
786 iowrite32(oconfig
, ®s
->OCONFIG
);
788 ret
= intel_overlay_on(overlay
);
793 iowrite32(params
->dst_y
<< 16 | params
->dst_x
, ®s
->DWINPOS
);
794 iowrite32(params
->dst_height
<< 16 | params
->dst_width
, ®s
->DWINSZ
);
796 if (params
->flags
& I915_OVERLAY_YUV_PACKED
)
797 tmp_width
= packed_width_bytes(params
->flags
,
800 tmp_width
= params
->src_width
;
802 swidth
= params
->src_width
;
803 swidthsw
= calc_swidthsw(dev_priv
, params
->offset_Y
, tmp_width
);
804 sheight
= params
->src_height
;
805 iowrite32(i915_ggtt_offset(vma
) + params
->offset_Y
, ®s
->OBUF_0Y
);
806 ostride
= params
->stride_Y
;
808 if (params
->flags
& I915_OVERLAY_YUV_PLANAR
) {
809 int uv_hscale
= uv_hsubsampling(params
->flags
);
810 int uv_vscale
= uv_vsubsampling(params
->flags
);
813 swidth
|= (params
->src_width
/ uv_hscale
) << 16;
814 sheight
|= (params
->src_height
/ uv_vscale
) << 16;
816 tmp_U
= calc_swidthsw(dev_priv
, params
->offset_U
,
817 params
->src_width
/ uv_hscale
);
818 tmp_V
= calc_swidthsw(dev_priv
, params
->offset_V
,
819 params
->src_width
/ uv_hscale
);
820 swidthsw
|= max(tmp_U
, tmp_V
) << 16;
822 iowrite32(i915_ggtt_offset(vma
) + params
->offset_U
,
824 iowrite32(i915_ggtt_offset(vma
) + params
->offset_V
,
827 ostride
|= params
->stride_UV
<< 16;
830 iowrite32(swidth
, ®s
->SWIDTH
);
831 iowrite32(swidthsw
, ®s
->SWIDTHSW
);
832 iowrite32(sheight
, ®s
->SHEIGHT
);
833 iowrite32(ostride
, ®s
->OSTRIDE
);
835 scale_changed
= update_scaling_factors(overlay
, regs
, params
);
837 update_colorkey(overlay
, regs
);
839 iowrite32(overlay_cmd_reg(params
), ®s
->OCMD
);
841 ret
= intel_overlay_continue(overlay
, vma
, scale_changed
);
848 i915_gem_object_unpin_from_display_plane(vma
);
850 atomic_dec(&dev_priv
->gpu_error
.pending_fb_pin
);
855 int intel_overlay_switch_off(struct intel_overlay
*overlay
)
857 struct drm_i915_private
*dev_priv
= overlay
->i915
;
860 WARN_ON(!drm_modeset_is_locked(&dev_priv
->drm
.mode_config
.connection_mutex
));
862 ret
= intel_overlay_recover_from_interrupt(overlay
);
866 if (!overlay
->active
)
869 ret
= intel_overlay_release_old_vid(overlay
);
873 iowrite32(0, &overlay
->regs
->OCMD
);
875 return intel_overlay_off(overlay
);
878 static int check_overlay_possible_on_crtc(struct intel_overlay
*overlay
,
879 struct intel_crtc
*crtc
)
884 /* can't use the overlay with double wide pipe */
885 if (crtc
->config
->double_wide
)
891 static void update_pfit_vscale_ratio(struct intel_overlay
*overlay
)
893 struct drm_i915_private
*dev_priv
= overlay
->i915
;
894 u32 pfit_control
= I915_READ(PFIT_CONTROL
);
897 /* XXX: This is not the same logic as in the xorg driver, but more in
898 * line with the intel documentation for the i965
900 if (INTEL_GEN(dev_priv
) >= 4) {
901 /* on i965 use the PGM reg to read out the autoscaler values */
902 ratio
= I915_READ(PFIT_PGM_RATIOS
) >> PFIT_VERT_SCALE_SHIFT_965
;
904 if (pfit_control
& VERT_AUTO_SCALE
)
905 ratio
= I915_READ(PFIT_AUTO_RATIOS
);
907 ratio
= I915_READ(PFIT_PGM_RATIOS
);
908 ratio
>>= PFIT_VERT_SCALE_SHIFT
;
911 overlay
->pfit_vscale_ratio
= ratio
;
914 static int check_overlay_dst(struct intel_overlay
*overlay
,
915 struct drm_intel_overlay_put_image
*rec
)
917 const struct intel_crtc_state
*pipe_config
=
918 overlay
->crtc
->config
;
920 if (rec
->dst_x
< pipe_config
->pipe_src_w
&&
921 rec
->dst_x
+ rec
->dst_width
<= pipe_config
->pipe_src_w
&&
922 rec
->dst_y
< pipe_config
->pipe_src_h
&&
923 rec
->dst_y
+ rec
->dst_height
<= pipe_config
->pipe_src_h
)
929 static int check_overlay_scaling(struct drm_intel_overlay_put_image
*rec
)
933 /* downscaling limit is 8.0 */
934 tmp
= ((rec
->src_scan_height
<< 16) / rec
->dst_height
) >> 16;
938 tmp
= ((rec
->src_scan_width
<< 16) / rec
->dst_width
) >> 16;
945 static int check_overlay_src(struct drm_i915_private
*dev_priv
,
946 struct drm_intel_overlay_put_image
*rec
,
947 struct drm_i915_gem_object
*new_bo
)
949 int uv_hscale
= uv_hsubsampling(rec
->flags
);
950 int uv_vscale
= uv_vsubsampling(rec
->flags
);
955 /* check src dimensions */
956 if (IS_I845G(dev_priv
) || IS_I830(dev_priv
)) {
957 if (rec
->src_height
> IMAGE_MAX_HEIGHT_LEGACY
||
958 rec
->src_width
> IMAGE_MAX_WIDTH_LEGACY
)
961 if (rec
->src_height
> IMAGE_MAX_HEIGHT
||
962 rec
->src_width
> IMAGE_MAX_WIDTH
)
966 /* better safe than sorry, use 4 as the maximal subsampling ratio */
967 if (rec
->src_height
< N_VERT_Y_TAPS
*4 ||
968 rec
->src_width
< N_HORIZ_Y_TAPS
*4)
971 /* check alignment constraints */
972 switch (rec
->flags
& I915_OVERLAY_TYPE_MASK
) {
973 case I915_OVERLAY_RGB
:
974 /* not implemented */
977 case I915_OVERLAY_YUV_PACKED
:
981 depth
= packed_depth_bytes(rec
->flags
);
985 /* ignore UV planes */
989 /* check pixel alignment */
990 if (rec
->offset_Y
% depth
)
994 case I915_OVERLAY_YUV_PLANAR
:
995 if (uv_vscale
< 0 || uv_hscale
< 0)
997 /* no offset restrictions for planar formats */
1004 if (rec
->src_width
% uv_hscale
)
1007 /* stride checking */
1008 if (IS_I830(dev_priv
) || IS_I845G(dev_priv
))
1013 if (rec
->stride_Y
& stride_mask
|| rec
->stride_UV
& stride_mask
)
1015 if (IS_GEN(dev_priv
, 4) && rec
->stride_Y
< 512)
1018 tmp
= (rec
->flags
& I915_OVERLAY_TYPE_MASK
) == I915_OVERLAY_YUV_PLANAR
?
1020 if (rec
->stride_Y
> tmp
|| rec
->stride_UV
> 2*1024)
1023 /* check buffer dimensions */
1024 switch (rec
->flags
& I915_OVERLAY_TYPE_MASK
) {
1025 case I915_OVERLAY_RGB
:
1026 case I915_OVERLAY_YUV_PACKED
:
1027 /* always 4 Y values per depth pixels */
1028 if (packed_width_bytes(rec
->flags
, rec
->src_width
) > rec
->stride_Y
)
1031 tmp
= rec
->stride_Y
*rec
->src_height
;
1032 if (rec
->offset_Y
+ tmp
> new_bo
->base
.size
)
1036 case I915_OVERLAY_YUV_PLANAR
:
1037 if (rec
->src_width
> rec
->stride_Y
)
1039 if (rec
->src_width
/uv_hscale
> rec
->stride_UV
)
1042 tmp
= rec
->stride_Y
* rec
->src_height
;
1043 if (rec
->offset_Y
+ tmp
> new_bo
->base
.size
)
1046 tmp
= rec
->stride_UV
* (rec
->src_height
/ uv_vscale
);
1047 if (rec
->offset_U
+ tmp
> new_bo
->base
.size
||
1048 rec
->offset_V
+ tmp
> new_bo
->base
.size
)
1056 int intel_overlay_put_image_ioctl(struct drm_device
*dev
, void *data
,
1057 struct drm_file
*file_priv
)
1059 struct drm_intel_overlay_put_image
*params
= data
;
1060 struct drm_i915_private
*dev_priv
= to_i915(dev
);
1061 struct intel_overlay
*overlay
;
1062 struct drm_crtc
*drmmode_crtc
;
1063 struct intel_crtc
*crtc
;
1064 struct drm_i915_gem_object
*new_bo
;
1067 overlay
= dev_priv
->overlay
;
1069 DRM_DEBUG("userspace bug: no overlay\n");
1073 if (!(params
->flags
& I915_OVERLAY_ENABLE
)) {
1074 drm_modeset_lock_all(dev
);
1075 ret
= intel_overlay_switch_off(overlay
);
1076 drm_modeset_unlock_all(dev
);
1081 drmmode_crtc
= drm_crtc_find(dev
, file_priv
, params
->crtc_id
);
1084 crtc
= to_intel_crtc(drmmode_crtc
);
1086 new_bo
= i915_gem_object_lookup(file_priv
, params
->bo_handle
);
1090 drm_modeset_lock_all(dev
);
1092 if (i915_gem_object_is_tiled(new_bo
)) {
1093 DRM_DEBUG_KMS("buffer used for overlay image can not be tiled\n");
1098 ret
= intel_overlay_recover_from_interrupt(overlay
);
1102 if (overlay
->crtc
!= crtc
) {
1103 ret
= intel_overlay_switch_off(overlay
);
1107 ret
= check_overlay_possible_on_crtc(overlay
, crtc
);
1111 overlay
->crtc
= crtc
;
1112 crtc
->overlay
= overlay
;
1114 /* line too wide, i.e. one-line-mode */
1115 if (crtc
->config
->pipe_src_w
> 1024 &&
1116 crtc
->config
->gmch_pfit
.control
& PFIT_ENABLE
) {
1117 overlay
->pfit_active
= true;
1118 update_pfit_vscale_ratio(overlay
);
1120 overlay
->pfit_active
= false;
1123 ret
= check_overlay_dst(overlay
, params
);
1127 if (overlay
->pfit_active
) {
1128 params
->dst_y
= (((u32
)params
->dst_y
<< 12) /
1129 overlay
->pfit_vscale_ratio
);
1130 /* shifting right rounds downwards, so add 1 */
1131 params
->dst_height
= (((u32
)params
->dst_height
<< 12) /
1132 overlay
->pfit_vscale_ratio
) + 1;
1135 if (params
->src_scan_height
> params
->src_height
||
1136 params
->src_scan_width
> params
->src_width
) {
1141 ret
= check_overlay_src(dev_priv
, params
, new_bo
);
1145 /* Check scaling after src size to prevent a divide-by-zero. */
1146 ret
= check_overlay_scaling(params
);
1150 ret
= intel_overlay_do_put_image(overlay
, new_bo
, params
);
1154 drm_modeset_unlock_all(dev
);
1155 i915_gem_object_put(new_bo
);
1160 drm_modeset_unlock_all(dev
);
1161 i915_gem_object_put(new_bo
);
1166 static void update_reg_attrs(struct intel_overlay
*overlay
,
1167 struct overlay_registers __iomem
*regs
)
1169 iowrite32((overlay
->contrast
<< 18) | (overlay
->brightness
& 0xff),
1171 iowrite32(overlay
->saturation
, ®s
->OCLRC1
);
1174 static bool check_gamma_bounds(u32 gamma1
, u32 gamma2
)
1178 if (gamma1
& 0xff000000 || gamma2
& 0xff000000)
1181 for (i
= 0; i
< 3; i
++) {
1182 if (((gamma1
>> i
*8) & 0xff) >= ((gamma2
>> i
*8) & 0xff))
1189 static bool check_gamma5_errata(u32 gamma5
)
1193 for (i
= 0; i
< 3; i
++) {
1194 if (((gamma5
>> i
*8) & 0xff) == 0x80)
1201 static int check_gamma(struct drm_intel_overlay_attrs
*attrs
)
1203 if (!check_gamma_bounds(0, attrs
->gamma0
) ||
1204 !check_gamma_bounds(attrs
->gamma0
, attrs
->gamma1
) ||
1205 !check_gamma_bounds(attrs
->gamma1
, attrs
->gamma2
) ||
1206 !check_gamma_bounds(attrs
->gamma2
, attrs
->gamma3
) ||
1207 !check_gamma_bounds(attrs
->gamma3
, attrs
->gamma4
) ||
1208 !check_gamma_bounds(attrs
->gamma4
, attrs
->gamma5
) ||
1209 !check_gamma_bounds(attrs
->gamma5
, 0x00ffffff))
1212 if (!check_gamma5_errata(attrs
->gamma5
))
1218 int intel_overlay_attrs_ioctl(struct drm_device
*dev
, void *data
,
1219 struct drm_file
*file_priv
)
1221 struct drm_intel_overlay_attrs
*attrs
= data
;
1222 struct drm_i915_private
*dev_priv
= to_i915(dev
);
1223 struct intel_overlay
*overlay
;
1226 overlay
= dev_priv
->overlay
;
1228 DRM_DEBUG("userspace bug: no overlay\n");
1232 drm_modeset_lock_all(dev
);
1235 if (!(attrs
->flags
& I915_OVERLAY_UPDATE_ATTRS
)) {
1236 attrs
->color_key
= overlay
->color_key
;
1237 attrs
->brightness
= overlay
->brightness
;
1238 attrs
->contrast
= overlay
->contrast
;
1239 attrs
->saturation
= overlay
->saturation
;
1241 if (!IS_GEN(dev_priv
, 2)) {
1242 attrs
->gamma0
= I915_READ(OGAMC0
);
1243 attrs
->gamma1
= I915_READ(OGAMC1
);
1244 attrs
->gamma2
= I915_READ(OGAMC2
);
1245 attrs
->gamma3
= I915_READ(OGAMC3
);
1246 attrs
->gamma4
= I915_READ(OGAMC4
);
1247 attrs
->gamma5
= I915_READ(OGAMC5
);
1250 if (attrs
->brightness
< -128 || attrs
->brightness
> 127)
1252 if (attrs
->contrast
> 255)
1254 if (attrs
->saturation
> 1023)
1257 overlay
->color_key
= attrs
->color_key
;
1258 overlay
->brightness
= attrs
->brightness
;
1259 overlay
->contrast
= attrs
->contrast
;
1260 overlay
->saturation
= attrs
->saturation
;
1262 update_reg_attrs(overlay
, overlay
->regs
);
1264 if (attrs
->flags
& I915_OVERLAY_UPDATE_GAMMA
) {
1265 if (IS_GEN(dev_priv
, 2))
1268 if (overlay
->active
) {
1273 ret
= check_gamma(attrs
);
1277 I915_WRITE(OGAMC0
, attrs
->gamma0
);
1278 I915_WRITE(OGAMC1
, attrs
->gamma1
);
1279 I915_WRITE(OGAMC2
, attrs
->gamma2
);
1280 I915_WRITE(OGAMC3
, attrs
->gamma3
);
1281 I915_WRITE(OGAMC4
, attrs
->gamma4
);
1282 I915_WRITE(OGAMC5
, attrs
->gamma5
);
1285 overlay
->color_key_enabled
= (attrs
->flags
& I915_OVERLAY_DISABLE_DEST_COLORKEY
) == 0;
1289 drm_modeset_unlock_all(dev
);
1294 static int get_registers(struct intel_overlay
*overlay
, bool use_phys
)
1296 struct drm_i915_private
*i915
= overlay
->i915
;
1297 struct drm_i915_gem_object
*obj
;
1298 struct i915_vma
*vma
;
1301 obj
= i915_gem_object_create_stolen(i915
, PAGE_SIZE
);
1303 obj
= i915_gem_object_create_internal(i915
, PAGE_SIZE
);
1305 return PTR_ERR(obj
);
1307 vma
= i915_gem_object_ggtt_pin(obj
, NULL
, 0, 0, PIN_MAPPABLE
);
1314 overlay
->flip_addr
= sg_dma_address(obj
->mm
.pages
->sgl
);
1316 overlay
->flip_addr
= i915_ggtt_offset(vma
);
1317 overlay
->regs
= i915_vma_pin_iomap(vma
);
1318 i915_vma_unpin(vma
);
1320 if (IS_ERR(overlay
->regs
)) {
1321 err
= PTR_ERR(overlay
->regs
);
1325 overlay
->reg_bo
= obj
;
1329 i915_gem_object_put(obj
);
1333 void intel_overlay_setup(struct drm_i915_private
*dev_priv
)
1335 struct intel_overlay
*overlay
;
1336 struct intel_engine_cs
*engine
;
1339 if (!HAS_OVERLAY(dev_priv
))
1342 engine
= dev_priv
->engine
[RCS0
];
1343 if (!engine
|| !engine
->kernel_context
)
1346 overlay
= kzalloc(sizeof(*overlay
), GFP_KERNEL
);
1350 overlay
->i915
= dev_priv
;
1351 overlay
->context
= engine
->kernel_context
;
1352 GEM_BUG_ON(!overlay
->context
);
1354 overlay
->color_key
= 0x0101fe;
1355 overlay
->color_key_enabled
= true;
1356 overlay
->brightness
= -19;
1357 overlay
->contrast
= 75;
1358 overlay
->saturation
= 146;
1360 i915_active_init(&overlay
->last_flip
,
1361 NULL
, intel_overlay_last_flip_retire
);
1363 ret
= get_registers(overlay
, OVERLAY_NEEDS_PHYSICAL(dev_priv
));
1367 memset_io(overlay
->regs
, 0, sizeof(struct overlay_registers
));
1368 update_polyphase_filter(overlay
->regs
);
1369 update_reg_attrs(overlay
, overlay
->regs
);
1371 dev_priv
->overlay
= overlay
;
1372 DRM_INFO("Initialized overlay support.\n");
1379 void intel_overlay_cleanup(struct drm_i915_private
*dev_priv
)
1381 struct intel_overlay
*overlay
;
1383 overlay
= fetch_and_zero(&dev_priv
->overlay
);
1388 * The bo's should be free'd by the generic code already.
1389 * Furthermore modesetting teardown happens beforehand so the
1390 * hardware should be off already.
1392 WARN_ON(overlay
->active
);
1394 i915_gem_object_put(overlay
->reg_bo
);
1395 i915_active_fini(&overlay
->last_flip
);
1400 #if IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR)
1402 struct intel_overlay_error_state
{
1403 struct overlay_registers regs
;
1409 struct intel_overlay_error_state
*
1410 intel_overlay_capture_error_state(struct drm_i915_private
*dev_priv
)
1412 struct intel_overlay
*overlay
= dev_priv
->overlay
;
1413 struct intel_overlay_error_state
*error
;
1415 if (!overlay
|| !overlay
->active
)
1418 error
= kmalloc(sizeof(*error
), GFP_ATOMIC
);
1422 error
->dovsta
= I915_READ(DOVSTA
);
1423 error
->isr
= I915_READ(GEN2_ISR
);
1424 error
->base
= overlay
->flip_addr
;
1426 memcpy_fromio(&error
->regs
, overlay
->regs
, sizeof(error
->regs
));
1432 intel_overlay_print_error_state(struct drm_i915_error_state_buf
*m
,
1433 struct intel_overlay_error_state
*error
)
1435 i915_error_printf(m
, "Overlay, status: 0x%08x, interrupt: 0x%08x\n",
1436 error
->dovsta
, error
->isr
);
1437 i915_error_printf(m
, " Register file at 0x%08lx:\n",
1440 #define P(x) i915_error_printf(m, " " #x ": 0x%08x\n", error->regs.x)