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 <linux/seq_file.h>
35 #include "intel_drv.h"
37 /* Limits for overlay size. According to intel doc, the real limits are:
38 * Y width: 4095, UV width (planar): 2047, Y height: 2047,
39 * UV width (planar): * 1023. But the xorg thinks 2048 for height and width. Use
40 * the mininum of both. */
41 #define IMAGE_MAX_WIDTH 2048
42 #define IMAGE_MAX_HEIGHT 2046 /* 2 * 1023 */
43 /* on 830 and 845 these large limits result in the card hanging */
44 #define IMAGE_MAX_WIDTH_LEGACY 1024
45 #define IMAGE_MAX_HEIGHT_LEGACY 1088
47 /* overlay register definitions */
49 #define OCMD_TILED_SURFACE (0x1<<19)
50 #define OCMD_MIRROR_MASK (0x3<<17)
51 #define OCMD_MIRROR_MODE (0x3<<17)
52 #define OCMD_MIRROR_HORIZONTAL (0x1<<17)
53 #define OCMD_MIRROR_VERTICAL (0x2<<17)
54 #define OCMD_MIRROR_BOTH (0x3<<17)
55 #define OCMD_BYTEORDER_MASK (0x3<<14) /* zero for YUYV or FOURCC YUY2 */
56 #define OCMD_UV_SWAP (0x1<<14) /* YVYU */
57 #define OCMD_Y_SWAP (0x2<<14) /* UYVY or FOURCC UYVY */
58 #define OCMD_Y_AND_UV_SWAP (0x3<<14) /* VYUY */
59 #define OCMD_SOURCE_FORMAT_MASK (0xf<<10)
60 #define OCMD_RGB_888 (0x1<<10) /* not in i965 Intel docs */
61 #define OCMD_RGB_555 (0x2<<10) /* not in i965 Intel docs */
62 #define OCMD_RGB_565 (0x3<<10) /* not in i965 Intel docs */
63 #define OCMD_YUV_422_PACKED (0x8<<10)
64 #define OCMD_YUV_411_PACKED (0x9<<10) /* not in i965 Intel docs */
65 #define OCMD_YUV_420_PLANAR (0xc<<10)
66 #define OCMD_YUV_422_PLANAR (0xd<<10)
67 #define OCMD_YUV_410_PLANAR (0xe<<10) /* also 411 */
68 #define OCMD_TVSYNCFLIP_PARITY (0x1<<9)
69 #define OCMD_TVSYNCFLIP_ENABLE (0x1<<7)
70 #define OCMD_BUF_TYPE_MASK (0x1<<5)
71 #define OCMD_BUF_TYPE_FRAME (0x0<<5)
72 #define OCMD_BUF_TYPE_FIELD (0x1<<5)
73 #define OCMD_TEST_MODE (0x1<<4)
74 #define OCMD_BUFFER_SELECT (0x3<<2)
75 #define OCMD_BUFFER0 (0x0<<2)
76 #define OCMD_BUFFER1 (0x1<<2)
77 #define OCMD_FIELD_SELECT (0x1<<2)
78 #define OCMD_FIELD0 (0x0<<1)
79 #define OCMD_FIELD1 (0x1<<1)
80 #define OCMD_ENABLE (0x1<<0)
82 /* OCONFIG register */
83 #define OCONF_PIPE_MASK (0x1<<18)
84 #define OCONF_PIPE_A (0x0<<18)
85 #define OCONF_PIPE_B (0x1<<18)
86 #define OCONF_GAMMA2_ENABLE (0x1<<16)
87 #define OCONF_CSC_MODE_BT601 (0x0<<5)
88 #define OCONF_CSC_MODE_BT709 (0x1<<5)
89 #define OCONF_CSC_BYPASS (0x1<<4)
90 #define OCONF_CC_OUT_8BIT (0x1<<3)
91 #define OCONF_TEST_MODE (0x1<<2)
92 #define OCONF_THREE_LINE_BUFFER (0x1<<0)
93 #define OCONF_TWO_LINE_BUFFER (0x0<<0)
95 /* DCLRKM (dst-key) register */
96 #define DST_KEY_ENABLE (0x1<<31)
97 #define CLK_RGB24_MASK 0x0
98 #define CLK_RGB16_MASK 0x070307
99 #define CLK_RGB15_MASK 0x070707
100 #define CLK_RGB8I_MASK 0xffffff
102 #define RGB16_TO_COLORKEY(c) \
103 (((c & 0xF800) << 8) | ((c & 0x07E0) << 5) | ((c & 0x001F) << 3))
104 #define RGB15_TO_COLORKEY(c) \
105 (((c & 0x7c00) << 9) | ((c & 0x03E0) << 6) | ((c & 0x001F) << 3))
107 /* overlay flip addr flag */
108 #define OFC_UPDATE 0x1
110 /* polyphase filter coefficients */
111 #define N_HORIZ_Y_TAPS 5
112 #define N_VERT_Y_TAPS 3
113 #define N_HORIZ_UV_TAPS 3
114 #define N_VERT_UV_TAPS 3
118 /* memory bufferd overlay registers */
119 struct overlay_registers {
147 u32 RESERVED1; /* 0x6C */
160 u32 FASTHSCALE; /* 0xA0 */
161 u32 UVSCALEV; /* 0xA4 */
162 u32 RESERVEDC[(0x200 - 0xA8) / 4]; /* 0xA8 - 0x1FC */
163 u16 Y_VCOEFS[N_VERT_Y_TAPS * N_PHASES]; /* 0x200 */
164 u16 RESERVEDD[0x100 / 2 - N_VERT_Y_TAPS * N_PHASES];
165 u16 Y_HCOEFS[N_HORIZ_Y_TAPS * N_PHASES]; /* 0x300 */
166 u16 RESERVEDE[0x200 / 2 - N_HORIZ_Y_TAPS * N_PHASES];
167 u16 UV_VCOEFS[N_VERT_UV_TAPS * N_PHASES]; /* 0x500 */
168 u16 RESERVEDF[0x100 / 2 - N_VERT_UV_TAPS * N_PHASES];
169 u16 UV_HCOEFS[N_HORIZ_UV_TAPS * N_PHASES]; /* 0x600 */
170 u16 RESERVEDG[0x100 / 2 - N_HORIZ_UV_TAPS * N_PHASES];
173 struct intel_overlay {
174 struct drm_device *dev;
175 struct intel_crtc *crtc;
176 struct drm_i915_gem_object *vid_bo;
177 struct drm_i915_gem_object *old_vid_bo;
180 u32 pfit_vscale_ratio; /* shifted-point number, (1<<12) == 1.0 */
182 u32 brightness, contrast, saturation;
183 u32 old_xscale, old_yscale;
184 /* register access */
186 struct drm_i915_gem_object *reg_bo;
188 uint32_t last_flip_req;
189 void (*flip_tail)(struct intel_overlay *);
192 static struct overlay_registers *
193 intel_overlay_map_regs(struct intel_overlay *overlay)
195 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
196 struct overlay_registers *regs;
198 if (OVERLAY_NEEDS_PHYSICAL(overlay->dev))
199 regs = overlay->reg_bo->phys_obj->handle->vaddr;
201 regs = io_mapping_map_wc(dev_priv->mm.gtt_mapping,
202 overlay->reg_bo->gtt_offset);
207 static void intel_overlay_unmap_regs(struct intel_overlay *overlay,
208 struct overlay_registers *regs)
210 if (!OVERLAY_NEEDS_PHYSICAL(overlay->dev))
211 io_mapping_unmap(regs);
214 static int intel_overlay_do_wait_request(struct intel_overlay *overlay,
215 struct drm_i915_gem_request *request,
216 void (*tail)(struct intel_overlay *))
218 struct drm_device *dev = overlay->dev;
219 drm_i915_private_t *dev_priv = dev->dev_private;
222 BUG_ON(overlay->last_flip_req);
223 ret = i915_add_request(LP_RING(dev_priv), NULL, request);
228 overlay->last_flip_req = request->seqno;
229 overlay->flip_tail = tail;
230 ret = i915_wait_request(LP_RING(dev_priv), overlay->last_flip_req);
234 overlay->last_flip_req = 0;
238 /* Workaround for i830 bug where pipe a must be enable to change control regs */
240 i830_activate_pipe_a(struct drm_device *dev)
242 drm_i915_private_t *dev_priv = dev->dev_private;
243 struct intel_crtc *crtc;
244 struct drm_crtc_helper_funcs *crtc_funcs;
245 struct drm_display_mode vesa_640x480 = {
246 DRM_MODE("640x480", DRM_MODE_TYPE_DRIVER, 25175, 640, 656,
247 752, 800, 0, 480, 489, 492, 525, 0,
248 DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC)
251 crtc = to_intel_crtc(dev_priv->pipe_to_crtc_mapping[0]);
252 if (crtc->dpms_mode == DRM_MODE_DPMS_ON)
255 /* most i8xx have pipe a forced on, so don't trust dpms mode */
256 if (I915_READ(_PIPEACONF) & PIPECONF_ENABLE)
259 crtc_funcs = crtc->base.helper_private;
260 if (crtc_funcs->dpms == NULL)
263 DRM_DEBUG_DRIVER("Enabling pipe A in order to enable overlay\n");
265 mode = drm_mode_duplicate(dev, &vesa_640x480);
266 drm_mode_set_crtcinfo(mode, CRTC_INTERLACE_HALVE_V);
267 if (!drm_crtc_helper_set_mode(&crtc->base, mode,
268 crtc->base.x, crtc->base.y,
272 crtc_funcs->dpms(&crtc->base, DRM_MODE_DPMS_ON);
277 i830_deactivate_pipe_a(struct drm_device *dev)
279 drm_i915_private_t *dev_priv = dev->dev_private;
280 struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[0];
281 struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
283 crtc_funcs->dpms(crtc, DRM_MODE_DPMS_OFF);
286 /* overlay needs to be disable in OCMD reg */
287 static int intel_overlay_on(struct intel_overlay *overlay)
289 struct drm_device *dev = overlay->dev;
290 struct drm_i915_private *dev_priv = dev->dev_private;
291 struct drm_i915_gem_request *request;
292 int pipe_a_quirk = 0;
295 BUG_ON(overlay->active);
299 pipe_a_quirk = i830_activate_pipe_a(dev);
300 if (pipe_a_quirk < 0)
304 request = kzalloc(sizeof(*request), GFP_KERNEL);
305 if (request == NULL) {
310 ret = BEGIN_LP_RING(4);
316 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_ON);
317 OUT_RING(overlay->flip_addr | OFC_UPDATE);
318 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
322 ret = intel_overlay_do_wait_request(overlay, request, NULL);
325 i830_deactivate_pipe_a(dev);
330 /* overlay needs to be enabled in OCMD reg */
331 static int intel_overlay_continue(struct intel_overlay *overlay,
332 bool load_polyphase_filter)
334 struct drm_device *dev = overlay->dev;
335 drm_i915_private_t *dev_priv = dev->dev_private;
336 struct drm_i915_gem_request *request;
337 u32 flip_addr = overlay->flip_addr;
341 BUG_ON(!overlay->active);
343 request = kzalloc(sizeof(*request), GFP_KERNEL);
347 if (load_polyphase_filter)
348 flip_addr |= OFC_UPDATE;
350 /* check for underruns */
351 tmp = I915_READ(DOVSTA);
353 DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp);
355 ret = BEGIN_LP_RING(2);
360 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
364 ret = i915_add_request(LP_RING(dev_priv), NULL, request);
370 overlay->last_flip_req = request->seqno;
374 static void intel_overlay_release_old_vid_tail(struct intel_overlay *overlay)
376 struct drm_i915_gem_object *obj = overlay->old_vid_bo;
378 i915_gem_object_unpin(obj);
379 drm_gem_object_unreference(&obj->base);
381 overlay->old_vid_bo = NULL;
384 static void intel_overlay_off_tail(struct intel_overlay *overlay)
386 struct drm_i915_gem_object *obj = overlay->vid_bo;
388 /* never have the overlay hw on without showing a frame */
389 BUG_ON(!overlay->vid_bo);
391 i915_gem_object_unpin(obj);
392 drm_gem_object_unreference(&obj->base);
393 overlay->vid_bo = NULL;
395 overlay->crtc->overlay = NULL;
396 overlay->crtc = NULL;
400 /* overlay needs to be disabled in OCMD reg */
401 static int intel_overlay_off(struct intel_overlay *overlay)
403 struct drm_device *dev = overlay->dev;
404 struct drm_i915_private *dev_priv = dev->dev_private;
405 u32 flip_addr = overlay->flip_addr;
406 struct drm_i915_gem_request *request;
409 BUG_ON(!overlay->active);
411 request = kzalloc(sizeof(*request), GFP_KERNEL);
415 /* According to intel docs the overlay hw may hang (when switching
416 * off) without loading the filter coeffs. It is however unclear whether
417 * this applies to the disabling of the overlay or to the switching off
418 * of the hw. Do it in both cases */
419 flip_addr |= OFC_UPDATE;
421 ret = BEGIN_LP_RING(6);
426 /* wait for overlay to go idle */
427 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
429 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
430 /* turn overlay off */
432 /* Workaround: Don't disable the overlay fully, since otherwise
433 * it dies on the next OVERLAY_ON cmd. */
438 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
440 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
444 return intel_overlay_do_wait_request(overlay, request,
445 intel_overlay_off_tail);
448 /* recover from an interruption due to a signal
449 * We have to be careful not to repeat work forever an make forward progess. */
450 static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay)
452 struct drm_device *dev = overlay->dev;
453 drm_i915_private_t *dev_priv = dev->dev_private;
456 if (overlay->last_flip_req == 0)
459 ret = i915_wait_request(LP_RING(dev_priv), overlay->last_flip_req);
463 if (overlay->flip_tail)
464 overlay->flip_tail(overlay);
466 overlay->last_flip_req = 0;
470 /* Wait for pending overlay flip and release old frame.
471 * Needs to be called before the overlay register are changed
472 * via intel_overlay_(un)map_regs
474 static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
476 struct drm_device *dev = overlay->dev;
477 drm_i915_private_t *dev_priv = dev->dev_private;
480 /* Only wait if there is actually an old frame to release to
481 * guarantee forward progress.
483 if (!overlay->old_vid_bo)
486 if (I915_READ(ISR) & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT) {
487 struct drm_i915_gem_request *request;
489 /* synchronous slowpath */
490 request = kzalloc(sizeof(*request), GFP_KERNEL);
494 ret = BEGIN_LP_RING(2);
500 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
504 ret = intel_overlay_do_wait_request(overlay, request,
505 intel_overlay_release_old_vid_tail);
510 intel_overlay_release_old_vid_tail(overlay);
514 struct put_image_params {
531 static int packed_depth_bytes(u32 format)
533 switch (format & I915_OVERLAY_DEPTH_MASK) {
534 case I915_OVERLAY_YUV422:
536 case I915_OVERLAY_YUV411:
537 /* return 6; not implemented */
543 static int packed_width_bytes(u32 format, short width)
545 switch (format & I915_OVERLAY_DEPTH_MASK) {
546 case I915_OVERLAY_YUV422:
553 static int uv_hsubsampling(u32 format)
555 switch (format & I915_OVERLAY_DEPTH_MASK) {
556 case I915_OVERLAY_YUV422:
557 case I915_OVERLAY_YUV420:
559 case I915_OVERLAY_YUV411:
560 case I915_OVERLAY_YUV410:
567 static int uv_vsubsampling(u32 format)
569 switch (format & I915_OVERLAY_DEPTH_MASK) {
570 case I915_OVERLAY_YUV420:
571 case I915_OVERLAY_YUV410:
573 case I915_OVERLAY_YUV422:
574 case I915_OVERLAY_YUV411:
581 static u32 calc_swidthsw(struct drm_device *dev, u32 offset, u32 width)
583 u32 mask, shift, ret;
591 ret = ((offset + width + mask) >> shift) - (offset >> shift);
598 static const u16 y_static_hcoeffs[N_HORIZ_Y_TAPS * N_PHASES] = {
599 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
600 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
601 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
602 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
603 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
604 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
605 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
606 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
607 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
608 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
609 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
610 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
611 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
612 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
613 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
614 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
615 0xb000, 0x3000, 0x0800, 0x3000, 0xb000
618 static const u16 uv_static_hcoeffs[N_HORIZ_UV_TAPS * N_PHASES] = {
619 0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
620 0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
621 0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
622 0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
623 0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
624 0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
625 0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
626 0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
627 0x3000, 0x0800, 0x3000
630 static void update_polyphase_filter(struct overlay_registers *regs)
632 memcpy(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
633 memcpy(regs->UV_HCOEFS, uv_static_hcoeffs, sizeof(uv_static_hcoeffs));
636 static bool update_scaling_factors(struct intel_overlay *overlay,
637 struct overlay_registers *regs,
638 struct put_image_params *params)
640 /* fixed point with a 12 bit shift */
641 u32 xscale, yscale, xscale_UV, yscale_UV;
643 #define FRACT_MASK 0xfff
644 bool scale_changed = false;
645 int uv_hscale = uv_hsubsampling(params->format);
646 int uv_vscale = uv_vsubsampling(params->format);
648 if (params->dst_w > 1)
649 xscale = ((params->src_scan_w - 1) << FP_SHIFT)
652 xscale = 1 << FP_SHIFT;
654 if (params->dst_h > 1)
655 yscale = ((params->src_scan_h - 1) << FP_SHIFT)
658 yscale = 1 << FP_SHIFT;
660 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
661 xscale_UV = xscale/uv_hscale;
662 yscale_UV = yscale/uv_vscale;
663 /* make the Y scale to UV scale ratio an exact multiply */
664 xscale = xscale_UV * uv_hscale;
665 yscale = yscale_UV * uv_vscale;
671 if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
672 scale_changed = true;
673 overlay->old_xscale = xscale;
674 overlay->old_yscale = yscale;
676 regs->YRGBSCALE = (((yscale & FRACT_MASK) << 20) |
677 ((xscale >> FP_SHIFT) << 16) |
678 ((xscale & FRACT_MASK) << 3));
680 regs->UVSCALE = (((yscale_UV & FRACT_MASK) << 20) |
681 ((xscale_UV >> FP_SHIFT) << 16) |
682 ((xscale_UV & FRACT_MASK) << 3));
684 regs->UVSCALEV = ((((yscale >> FP_SHIFT) << 16) |
685 ((yscale_UV >> FP_SHIFT) << 0)));
688 update_polyphase_filter(regs);
690 return scale_changed;
693 static void update_colorkey(struct intel_overlay *overlay,
694 struct overlay_registers *regs)
696 u32 key = overlay->color_key;
698 switch (overlay->crtc->base.fb->bits_per_pixel) {
701 regs->DCLRKM = CLK_RGB8I_MASK | DST_KEY_ENABLE;
705 if (overlay->crtc->base.fb->depth == 15) {
706 regs->DCLRKV = RGB15_TO_COLORKEY(key);
707 regs->DCLRKM = CLK_RGB15_MASK | DST_KEY_ENABLE;
709 regs->DCLRKV = RGB16_TO_COLORKEY(key);
710 regs->DCLRKM = CLK_RGB16_MASK | DST_KEY_ENABLE;
717 regs->DCLRKM = CLK_RGB24_MASK | DST_KEY_ENABLE;
722 static u32 overlay_cmd_reg(struct put_image_params *params)
724 u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
726 if (params->format & I915_OVERLAY_YUV_PLANAR) {
727 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
728 case I915_OVERLAY_YUV422:
729 cmd |= OCMD_YUV_422_PLANAR;
731 case I915_OVERLAY_YUV420:
732 cmd |= OCMD_YUV_420_PLANAR;
734 case I915_OVERLAY_YUV411:
735 case I915_OVERLAY_YUV410:
736 cmd |= OCMD_YUV_410_PLANAR;
739 } else { /* YUV packed */
740 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
741 case I915_OVERLAY_YUV422:
742 cmd |= OCMD_YUV_422_PACKED;
744 case I915_OVERLAY_YUV411:
745 cmd |= OCMD_YUV_411_PACKED;
749 switch (params->format & I915_OVERLAY_SWAP_MASK) {
750 case I915_OVERLAY_NO_SWAP:
752 case I915_OVERLAY_UV_SWAP:
755 case I915_OVERLAY_Y_SWAP:
758 case I915_OVERLAY_Y_AND_UV_SWAP:
759 cmd |= OCMD_Y_AND_UV_SWAP;
767 static int intel_overlay_do_put_image(struct intel_overlay *overlay,
768 struct drm_i915_gem_object *new_bo,
769 struct put_image_params *params)
772 struct overlay_registers *regs;
773 bool scale_changed = false;
774 struct drm_device *dev = overlay->dev;
776 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
777 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
780 ret = intel_overlay_release_old_vid(overlay);
784 ret = i915_gem_object_pin_to_display_plane(new_bo, 0, NULL);
788 ret = i915_gem_object_put_fence(new_bo);
792 if (!overlay->active) {
793 regs = intel_overlay_map_regs(overlay);
798 regs->OCONFIG = OCONF_CC_OUT_8BIT;
799 if (IS_GEN4(overlay->dev))
800 regs->OCONFIG |= OCONF_CSC_MODE_BT709;
801 regs->OCONFIG |= overlay->crtc->pipe == 0 ?
802 OCONF_PIPE_A : OCONF_PIPE_B;
803 intel_overlay_unmap_regs(overlay, regs);
805 ret = intel_overlay_on(overlay);
810 regs = intel_overlay_map_regs(overlay);
816 regs->DWINPOS = (params->dst_y << 16) | params->dst_x;
817 regs->DWINSZ = (params->dst_h << 16) | params->dst_w;
819 if (params->format & I915_OVERLAY_YUV_PACKED)
820 tmp_width = packed_width_bytes(params->format, params->src_w);
822 tmp_width = params->src_w;
824 regs->SWIDTH = params->src_w;
825 regs->SWIDTHSW = calc_swidthsw(overlay->dev,
826 params->offset_Y, tmp_width);
827 regs->SHEIGHT = params->src_h;
828 regs->OBUF_0Y = new_bo->gtt_offset + params->offset_Y;
829 regs->OSTRIDE = params->stride_Y;
831 if (params->format & I915_OVERLAY_YUV_PLANAR) {
832 int uv_hscale = uv_hsubsampling(params->format);
833 int uv_vscale = uv_vsubsampling(params->format);
835 regs->SWIDTH |= (params->src_w/uv_hscale) << 16;
836 tmp_U = calc_swidthsw(overlay->dev, params->offset_U,
837 params->src_w/uv_hscale);
838 tmp_V = calc_swidthsw(overlay->dev, params->offset_V,
839 params->src_w/uv_hscale);
840 regs->SWIDTHSW |= max_t(u32, tmp_U, tmp_V) << 16;
841 regs->SHEIGHT |= (params->src_h/uv_vscale) << 16;
842 regs->OBUF_0U = new_bo->gtt_offset + params->offset_U;
843 regs->OBUF_0V = new_bo->gtt_offset + params->offset_V;
844 regs->OSTRIDE |= params->stride_UV << 16;
847 scale_changed = update_scaling_factors(overlay, regs, params);
849 update_colorkey(overlay, regs);
851 regs->OCMD = overlay_cmd_reg(params);
853 intel_overlay_unmap_regs(overlay, regs);
855 ret = intel_overlay_continue(overlay, scale_changed);
859 overlay->old_vid_bo = overlay->vid_bo;
860 overlay->vid_bo = new_bo;
865 i915_gem_object_unpin(new_bo);
869 int intel_overlay_switch_off(struct intel_overlay *overlay)
871 struct overlay_registers *regs;
872 struct drm_device *dev = overlay->dev;
875 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
876 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
878 ret = intel_overlay_recover_from_interrupt(overlay);
882 if (!overlay->active)
885 ret = intel_overlay_release_old_vid(overlay);
889 regs = intel_overlay_map_regs(overlay);
891 intel_overlay_unmap_regs(overlay, regs);
893 ret = intel_overlay_off(overlay);
897 intel_overlay_off_tail(overlay);
901 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
902 struct intel_crtc *crtc)
904 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
909 /* can't use the overlay with double wide pipe */
910 if (INTEL_INFO(overlay->dev)->gen < 4 &&
911 (I915_READ(PIPECONF(crtc->pipe)) & (PIPECONF_DOUBLE_WIDE | PIPECONF_ENABLE)) != PIPECONF_ENABLE)
917 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
919 struct drm_device *dev = overlay->dev;
920 drm_i915_private_t *dev_priv = dev->dev_private;
921 u32 pfit_control = I915_READ(PFIT_CONTROL);
924 /* XXX: This is not the same logic as in the xorg driver, but more in
925 * line with the intel documentation for the i965
927 if (INTEL_INFO(dev)->gen >= 4) {
928 /* on i965 use the PGM reg to read out the autoscaler values */
929 ratio = I915_READ(PFIT_PGM_RATIOS) >> PFIT_VERT_SCALE_SHIFT_965;
931 if (pfit_control & VERT_AUTO_SCALE)
932 ratio = I915_READ(PFIT_AUTO_RATIOS);
934 ratio = I915_READ(PFIT_PGM_RATIOS);
935 ratio >>= PFIT_VERT_SCALE_SHIFT;
938 overlay->pfit_vscale_ratio = ratio;
941 static int check_overlay_dst(struct intel_overlay *overlay,
942 struct drm_intel_overlay_put_image *rec)
944 struct drm_display_mode *mode = &overlay->crtc->base.mode;
946 if (rec->dst_x < mode->crtc_hdisplay &&
947 rec->dst_x + rec->dst_width <= mode->crtc_hdisplay &&
948 rec->dst_y < mode->crtc_vdisplay &&
949 rec->dst_y + rec->dst_height <= mode->crtc_vdisplay)
955 static int check_overlay_scaling(struct put_image_params *rec)
959 /* downscaling limit is 8.0 */
960 tmp = ((rec->src_scan_h << 16) / rec->dst_h) >> 16;
963 tmp = ((rec->src_scan_w << 16) / rec->dst_w) >> 16;
970 static int check_overlay_src(struct drm_device *dev,
971 struct drm_intel_overlay_put_image *rec,
972 struct drm_i915_gem_object *new_bo)
974 int uv_hscale = uv_hsubsampling(rec->flags);
975 int uv_vscale = uv_vsubsampling(rec->flags);
980 /* check src dimensions */
981 if (IS_845G(dev) || IS_I830(dev)) {
982 if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY ||
983 rec->src_width > IMAGE_MAX_WIDTH_LEGACY)
986 if (rec->src_height > IMAGE_MAX_HEIGHT ||
987 rec->src_width > IMAGE_MAX_WIDTH)
991 /* better safe than sorry, use 4 as the maximal subsampling ratio */
992 if (rec->src_height < N_VERT_Y_TAPS*4 ||
993 rec->src_width < N_HORIZ_Y_TAPS*4)
996 /* check alignment constraints */
997 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
998 case I915_OVERLAY_RGB:
999 /* not implemented */
1002 case I915_OVERLAY_YUV_PACKED:
1006 depth = packed_depth_bytes(rec->flags);
1010 /* ignore UV planes */
1014 /* check pixel alignment */
1015 if (rec->offset_Y % depth)
1019 case I915_OVERLAY_YUV_PLANAR:
1020 if (uv_vscale < 0 || uv_hscale < 0)
1022 /* no offset restrictions for planar formats */
1029 if (rec->src_width % uv_hscale)
1032 /* stride checking */
1033 if (IS_I830(dev) || IS_845G(dev))
1038 if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
1040 if (IS_GEN4(dev) && rec->stride_Y < 512)
1043 tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
1045 if (rec->stride_Y > tmp || rec->stride_UV > 2*1024)
1048 /* check buffer dimensions */
1049 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1050 case I915_OVERLAY_RGB:
1051 case I915_OVERLAY_YUV_PACKED:
1052 /* always 4 Y values per depth pixels */
1053 if (packed_width_bytes(rec->flags, rec->src_width) > rec->stride_Y)
1056 tmp = rec->stride_Y*rec->src_height;
1057 if (rec->offset_Y + tmp > new_bo->base.size)
1061 case I915_OVERLAY_YUV_PLANAR:
1062 if (rec->src_width > rec->stride_Y)
1064 if (rec->src_width/uv_hscale > rec->stride_UV)
1067 tmp = rec->stride_Y * rec->src_height;
1068 if (rec->offset_Y + tmp > new_bo->base.size)
1071 tmp = rec->stride_UV * (rec->src_height / uv_vscale);
1072 if (rec->offset_U + tmp > new_bo->base.size ||
1073 rec->offset_V + tmp > new_bo->base.size)
1082 * Return the pipe currently connected to the panel fitter,
1083 * or -1 if the panel fitter is not present or not in use
1085 static int intel_panel_fitter_pipe(struct drm_device *dev)
1087 struct drm_i915_private *dev_priv = dev->dev_private;
1090 /* i830 doesn't have a panel fitter */
1094 pfit_control = I915_READ(PFIT_CONTROL);
1096 /* See if the panel fitter is in use */
1097 if ((pfit_control & PFIT_ENABLE) == 0)
1100 /* 965 can place panel fitter on either pipe */
1102 return (pfit_control >> 29) & 0x3;
1104 /* older chips can only use pipe 1 */
1108 int intel_overlay_put_image(struct drm_device *dev, void *data,
1109 struct drm_file *file_priv)
1111 struct drm_intel_overlay_put_image *put_image_rec = data;
1112 drm_i915_private_t *dev_priv = dev->dev_private;
1113 struct intel_overlay *overlay;
1114 struct drm_mode_object *drmmode_obj;
1115 struct intel_crtc *crtc;
1116 struct drm_i915_gem_object *new_bo;
1117 struct put_image_params *params;
1121 DRM_ERROR("called with no initialization\n");
1125 overlay = dev_priv->overlay;
1127 DRM_DEBUG("userspace bug: no overlay\n");
1131 if (!(put_image_rec->flags & I915_OVERLAY_ENABLE)) {
1132 mutex_lock(&dev->mode_config.mutex);
1133 mutex_lock(&dev->struct_mutex);
1135 ret = intel_overlay_switch_off(overlay);
1137 mutex_unlock(&dev->struct_mutex);
1138 mutex_unlock(&dev->mode_config.mutex);
1143 params = kmalloc(sizeof(struct put_image_params), GFP_KERNEL);
1147 drmmode_obj = drm_mode_object_find(dev, put_image_rec->crtc_id,
1148 DRM_MODE_OBJECT_CRTC);
1153 crtc = to_intel_crtc(obj_to_crtc(drmmode_obj));
1155 new_bo = to_intel_bo(drm_gem_object_lookup(dev, file_priv,
1156 put_image_rec->bo_handle));
1157 if (&new_bo->base == NULL) {
1162 mutex_lock(&dev->mode_config.mutex);
1163 mutex_lock(&dev->struct_mutex);
1165 if (new_bo->tiling_mode) {
1166 DRM_ERROR("buffer used for overlay image can not be tiled\n");
1171 ret = intel_overlay_recover_from_interrupt(overlay);
1175 if (overlay->crtc != crtc) {
1176 struct drm_display_mode *mode = &crtc->base.mode;
1177 ret = intel_overlay_switch_off(overlay);
1181 ret = check_overlay_possible_on_crtc(overlay, crtc);
1185 overlay->crtc = crtc;
1186 crtc->overlay = overlay;
1188 /* line too wide, i.e. one-line-mode */
1189 if (mode->hdisplay > 1024 &&
1190 intel_panel_fitter_pipe(dev) == crtc->pipe) {
1191 overlay->pfit_active = 1;
1192 update_pfit_vscale_ratio(overlay);
1194 overlay->pfit_active = 0;
1197 ret = check_overlay_dst(overlay, put_image_rec);
1201 if (overlay->pfit_active) {
1202 params->dst_y = ((((u32)put_image_rec->dst_y) << 12) /
1203 overlay->pfit_vscale_ratio);
1204 /* shifting right rounds downwards, so add 1 */
1205 params->dst_h = ((((u32)put_image_rec->dst_height) << 12) /
1206 overlay->pfit_vscale_ratio) + 1;
1208 params->dst_y = put_image_rec->dst_y;
1209 params->dst_h = put_image_rec->dst_height;
1211 params->dst_x = put_image_rec->dst_x;
1212 params->dst_w = put_image_rec->dst_width;
1214 params->src_w = put_image_rec->src_width;
1215 params->src_h = put_image_rec->src_height;
1216 params->src_scan_w = put_image_rec->src_scan_width;
1217 params->src_scan_h = put_image_rec->src_scan_height;
1218 if (params->src_scan_h > params->src_h ||
1219 params->src_scan_w > params->src_w) {
1224 ret = check_overlay_src(dev, put_image_rec, new_bo);
1227 params->format = put_image_rec->flags & ~I915_OVERLAY_FLAGS_MASK;
1228 params->stride_Y = put_image_rec->stride_Y;
1229 params->stride_UV = put_image_rec->stride_UV;
1230 params->offset_Y = put_image_rec->offset_Y;
1231 params->offset_U = put_image_rec->offset_U;
1232 params->offset_V = put_image_rec->offset_V;
1234 /* Check scaling after src size to prevent a divide-by-zero. */
1235 ret = check_overlay_scaling(params);
1239 ret = intel_overlay_do_put_image(overlay, new_bo, params);
1243 mutex_unlock(&dev->struct_mutex);
1244 mutex_unlock(&dev->mode_config.mutex);
1251 mutex_unlock(&dev->struct_mutex);
1252 mutex_unlock(&dev->mode_config.mutex);
1253 drm_gem_object_unreference_unlocked(&new_bo->base);
1260 static void update_reg_attrs(struct intel_overlay *overlay,
1261 struct overlay_registers *regs)
1263 regs->OCLRC0 = (overlay->contrast << 18) | (overlay->brightness & 0xff);
1264 regs->OCLRC1 = overlay->saturation;
1267 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
1271 if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
1274 for (i = 0; i < 3; i++) {
1275 if (((gamma1 >> i*8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
1282 static bool check_gamma5_errata(u32 gamma5)
1286 for (i = 0; i < 3; i++) {
1287 if (((gamma5 >> i*8) & 0xff) == 0x80)
1294 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
1296 if (!check_gamma_bounds(0, attrs->gamma0) ||
1297 !check_gamma_bounds(attrs->gamma0, attrs->gamma1) ||
1298 !check_gamma_bounds(attrs->gamma1, attrs->gamma2) ||
1299 !check_gamma_bounds(attrs->gamma2, attrs->gamma3) ||
1300 !check_gamma_bounds(attrs->gamma3, attrs->gamma4) ||
1301 !check_gamma_bounds(attrs->gamma4, attrs->gamma5) ||
1302 !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
1305 if (!check_gamma5_errata(attrs->gamma5))
1311 int intel_overlay_attrs(struct drm_device *dev, void *data,
1312 struct drm_file *file_priv)
1314 struct drm_intel_overlay_attrs *attrs = data;
1315 drm_i915_private_t *dev_priv = dev->dev_private;
1316 struct intel_overlay *overlay;
1317 struct overlay_registers *regs;
1321 DRM_ERROR("called with no initialization\n");
1325 overlay = dev_priv->overlay;
1327 DRM_DEBUG("userspace bug: no overlay\n");
1331 mutex_lock(&dev->mode_config.mutex);
1332 mutex_lock(&dev->struct_mutex);
1335 if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
1336 attrs->color_key = overlay->color_key;
1337 attrs->brightness = overlay->brightness;
1338 attrs->contrast = overlay->contrast;
1339 attrs->saturation = overlay->saturation;
1341 if (!IS_GEN2(dev)) {
1342 attrs->gamma0 = I915_READ(OGAMC0);
1343 attrs->gamma1 = I915_READ(OGAMC1);
1344 attrs->gamma2 = I915_READ(OGAMC2);
1345 attrs->gamma3 = I915_READ(OGAMC3);
1346 attrs->gamma4 = I915_READ(OGAMC4);
1347 attrs->gamma5 = I915_READ(OGAMC5);
1350 if (attrs->brightness < -128 || attrs->brightness > 127)
1352 if (attrs->contrast > 255)
1354 if (attrs->saturation > 1023)
1357 overlay->color_key = attrs->color_key;
1358 overlay->brightness = attrs->brightness;
1359 overlay->contrast = attrs->contrast;
1360 overlay->saturation = attrs->saturation;
1362 regs = intel_overlay_map_regs(overlay);
1368 update_reg_attrs(overlay, regs);
1370 intel_overlay_unmap_regs(overlay, regs);
1372 if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
1376 if (overlay->active) {
1381 ret = check_gamma(attrs);
1385 I915_WRITE(OGAMC0, attrs->gamma0);
1386 I915_WRITE(OGAMC1, attrs->gamma1);
1387 I915_WRITE(OGAMC2, attrs->gamma2);
1388 I915_WRITE(OGAMC3, attrs->gamma3);
1389 I915_WRITE(OGAMC4, attrs->gamma4);
1390 I915_WRITE(OGAMC5, attrs->gamma5);
1396 mutex_unlock(&dev->struct_mutex);
1397 mutex_unlock(&dev->mode_config.mutex);
1402 void intel_setup_overlay(struct drm_device *dev)
1404 drm_i915_private_t *dev_priv = dev->dev_private;
1405 struct intel_overlay *overlay;
1406 struct drm_i915_gem_object *reg_bo;
1407 struct overlay_registers *regs;
1410 if (!HAS_OVERLAY(dev))
1413 overlay = kzalloc(sizeof(struct intel_overlay), GFP_KERNEL);
1417 mutex_lock(&dev->struct_mutex);
1418 if (WARN_ON(dev_priv->overlay))
1423 reg_bo = i915_gem_alloc_object(dev, PAGE_SIZE);
1426 overlay->reg_bo = reg_bo;
1428 if (OVERLAY_NEEDS_PHYSICAL(dev)) {
1429 ret = i915_gem_attach_phys_object(dev, reg_bo,
1430 I915_GEM_PHYS_OVERLAY_REGS,
1433 DRM_ERROR("failed to attach phys overlay regs\n");
1436 overlay->flip_addr = reg_bo->phys_obj->handle->busaddr;
1438 ret = i915_gem_object_pin(reg_bo, PAGE_SIZE, true);
1440 DRM_ERROR("failed to pin overlay register bo\n");
1443 overlay->flip_addr = reg_bo->gtt_offset;
1445 ret = i915_gem_object_set_to_gtt_domain(reg_bo, true);
1447 DRM_ERROR("failed to move overlay register bo into the GTT\n");
1452 /* init all values */
1453 overlay->color_key = 0x0101fe;
1454 overlay->brightness = -19;
1455 overlay->contrast = 75;
1456 overlay->saturation = 146;
1458 regs = intel_overlay_map_regs(overlay);
1462 memset(regs, 0, sizeof(struct overlay_registers));
1463 update_polyphase_filter(regs);
1464 update_reg_attrs(overlay, regs);
1466 intel_overlay_unmap_regs(overlay, regs);
1468 dev_priv->overlay = overlay;
1469 mutex_unlock(&dev->struct_mutex);
1470 DRM_INFO("initialized overlay support\n");
1474 if (!OVERLAY_NEEDS_PHYSICAL(dev))
1475 i915_gem_object_unpin(reg_bo);
1477 drm_gem_object_unreference(®_bo->base);
1479 mutex_unlock(&dev->struct_mutex);
1484 void intel_cleanup_overlay(struct drm_device *dev)
1486 drm_i915_private_t *dev_priv = dev->dev_private;
1488 if (!dev_priv->overlay)
1491 /* The bo's should be free'd by the generic code already.
1492 * Furthermore modesetting teardown happens beforehand so the
1493 * hardware should be off already */
1494 BUG_ON(dev_priv->overlay->active);
1496 drm_gem_object_unreference_unlocked(&dev_priv->overlay->reg_bo->base);
1497 kfree(dev_priv->overlay);
1500 #ifdef CONFIG_DEBUG_FS
1501 #include <linux/seq_file.h>
1503 struct intel_overlay_error_state {
1504 struct overlay_registers regs;
1510 static struct overlay_registers *
1511 intel_overlay_map_regs_atomic(struct intel_overlay *overlay)
1513 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
1514 struct overlay_registers *regs;
1516 if (OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1517 regs = overlay->reg_bo->phys_obj->handle->vaddr;
1519 regs = io_mapping_map_atomic_wc(dev_priv->mm.gtt_mapping,
1520 overlay->reg_bo->gtt_offset);
1525 static void intel_overlay_unmap_regs_atomic(struct intel_overlay *overlay,
1526 struct overlay_registers *regs)
1528 if (!OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1529 io_mapping_unmap_atomic(regs);
1533 struct intel_overlay_error_state *
1534 intel_overlay_capture_error_state(struct drm_device *dev)
1536 drm_i915_private_t *dev_priv = dev->dev_private;
1537 struct intel_overlay *overlay = dev_priv->overlay;
1538 struct intel_overlay_error_state *error;
1539 struct overlay_registers __iomem *regs;
1541 if (!overlay || !overlay->active)
1544 error = kmalloc(sizeof(*error), GFP_ATOMIC);
1548 error->dovsta = I915_READ(DOVSTA);
1549 error->isr = I915_READ(ISR);
1550 if (OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1551 error->base = (long) overlay->reg_bo->phys_obj->handle->vaddr;
1553 error->base = (long) overlay->reg_bo->gtt_offset;
1555 regs = intel_overlay_map_regs_atomic(overlay);
1559 memcpy_fromio(&error->regs, regs, sizeof(struct overlay_registers));
1560 intel_overlay_unmap_regs_atomic(overlay, regs);
1570 intel_overlay_print_error_state(struct seq_file *m, struct intel_overlay_error_state *error)
1572 seq_printf(m, "Overlay, status: 0x%08x, interrupt: 0x%08x\n",
1573 error->dovsta, error->isr);
1574 seq_printf(m, " Register file at 0x%08lx:\n",
1577 #define P(x) seq_printf(m, " " #x ": 0x%08x\n", error->regs.x)