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,
217 void (*tail)(struct intel_overlay *))
219 struct drm_device *dev = overlay->dev;
220 drm_i915_private_t *dev_priv = dev->dev_private;
223 BUG_ON(overlay->last_flip_req);
224 ret = i915_add_request(LP_RING(dev_priv), NULL, request);
229 overlay->last_flip_req = request->seqno;
230 overlay->flip_tail = tail;
231 ret = i915_wait_request(LP_RING(dev_priv),
232 overlay->last_flip_req, true);
236 overlay->last_flip_req = 0;
240 /* Workaround for i830 bug where pipe a must be enable to change control regs */
242 i830_activate_pipe_a(struct drm_device *dev)
244 drm_i915_private_t *dev_priv = dev->dev_private;
245 struct intel_crtc *crtc;
246 struct drm_crtc_helper_funcs *crtc_funcs;
247 struct drm_display_mode vesa_640x480 = {
248 DRM_MODE("640x480", DRM_MODE_TYPE_DRIVER, 25175, 640, 656,
249 752, 800, 0, 480, 489, 492, 525, 0,
250 DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC)
253 crtc = to_intel_crtc(dev_priv->pipe_to_crtc_mapping[0]);
254 if (crtc->dpms_mode == DRM_MODE_DPMS_ON)
257 /* most i8xx have pipe a forced on, so don't trust dpms mode */
258 if (I915_READ(_PIPEACONF) & PIPECONF_ENABLE)
261 crtc_funcs = crtc->base.helper_private;
262 if (crtc_funcs->dpms == NULL)
265 DRM_DEBUG_DRIVER("Enabling pipe A in order to enable overlay\n");
267 mode = drm_mode_duplicate(dev, &vesa_640x480);
268 drm_mode_set_crtcinfo(mode, CRTC_INTERLACE_HALVE_V);
269 if(!drm_crtc_helper_set_mode(&crtc->base, mode,
270 crtc->base.x, crtc->base.y,
274 crtc_funcs->dpms(&crtc->base, DRM_MODE_DPMS_ON);
279 i830_deactivate_pipe_a(struct drm_device *dev)
281 drm_i915_private_t *dev_priv = dev->dev_private;
282 struct drm_crtc *crtc = dev_priv->pipe_to_crtc_mapping[0];
283 struct drm_crtc_helper_funcs *crtc_funcs = crtc->helper_private;
285 crtc_funcs->dpms(crtc, DRM_MODE_DPMS_OFF);
288 /* overlay needs to be disable in OCMD reg */
289 static int intel_overlay_on(struct intel_overlay *overlay)
291 struct drm_device *dev = overlay->dev;
292 struct drm_i915_private *dev_priv = dev->dev_private;
293 struct drm_i915_gem_request *request;
294 int pipe_a_quirk = 0;
297 BUG_ON(overlay->active);
301 pipe_a_quirk = i830_activate_pipe_a(dev);
302 if (pipe_a_quirk < 0)
306 request = kzalloc(sizeof(*request), GFP_KERNEL);
307 if (request == NULL) {
312 ret = BEGIN_LP_RING(4);
318 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_ON);
319 OUT_RING(overlay->flip_addr | OFC_UPDATE);
320 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
324 ret = intel_overlay_do_wait_request(overlay, request, true, NULL);
327 i830_deactivate_pipe_a(dev);
332 /* overlay needs to be enabled in OCMD reg */
333 static int intel_overlay_continue(struct intel_overlay *overlay,
334 bool load_polyphase_filter)
336 struct drm_device *dev = overlay->dev;
337 drm_i915_private_t *dev_priv = dev->dev_private;
338 struct drm_i915_gem_request *request;
339 u32 flip_addr = overlay->flip_addr;
343 BUG_ON(!overlay->active);
345 request = kzalloc(sizeof(*request), GFP_KERNEL);
349 if (load_polyphase_filter)
350 flip_addr |= OFC_UPDATE;
352 /* check for underruns */
353 tmp = I915_READ(DOVSTA);
355 DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp);
357 ret = BEGIN_LP_RING(2);
362 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
366 ret = i915_add_request(LP_RING(dev_priv), NULL, request);
372 overlay->last_flip_req = request->seqno;
376 static void intel_overlay_release_old_vid_tail(struct intel_overlay *overlay)
378 struct drm_i915_gem_object *obj = overlay->old_vid_bo;
380 i915_gem_object_unpin(obj);
381 drm_gem_object_unreference(&obj->base);
383 overlay->old_vid_bo = NULL;
386 static void intel_overlay_off_tail(struct intel_overlay *overlay)
388 struct drm_i915_gem_object *obj = overlay->vid_bo;
390 /* never have the overlay hw on without showing a frame */
391 BUG_ON(!overlay->vid_bo);
393 i915_gem_object_unpin(obj);
394 drm_gem_object_unreference(&obj->base);
395 overlay->vid_bo = NULL;
397 overlay->crtc->overlay = NULL;
398 overlay->crtc = NULL;
402 /* overlay needs to be disabled in OCMD reg */
403 static int intel_overlay_off(struct intel_overlay *overlay,
406 struct drm_device *dev = overlay->dev;
407 struct drm_i915_private *dev_priv = dev->dev_private;
408 u32 flip_addr = overlay->flip_addr;
409 struct drm_i915_gem_request *request;
412 BUG_ON(!overlay->active);
414 request = kzalloc(sizeof(*request), GFP_KERNEL);
418 /* According to intel docs the overlay hw may hang (when switching
419 * off) without loading the filter coeffs. It is however unclear whether
420 * this applies to the disabling of the overlay or to the switching off
421 * of the hw. Do it in both cases */
422 flip_addr |= OFC_UPDATE;
424 ret = BEGIN_LP_RING(6);
429 /* wait for overlay to go idle */
430 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
432 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
433 /* turn overlay off */
434 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
436 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
439 return intel_overlay_do_wait_request(overlay, request, interruptible,
440 intel_overlay_off_tail);
443 /* recover from an interruption due to a signal
444 * We have to be careful not to repeat work forever an make forward progess. */
445 static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
448 struct drm_device *dev = overlay->dev;
449 drm_i915_private_t *dev_priv = dev->dev_private;
452 if (overlay->last_flip_req == 0)
455 ret = i915_wait_request(LP_RING(dev_priv),
456 overlay->last_flip_req, interruptible);
460 if (overlay->flip_tail)
461 overlay->flip_tail(overlay);
463 overlay->last_flip_req = 0;
467 /* Wait for pending overlay flip and release old frame.
468 * Needs to be called before the overlay register are changed
469 * via intel_overlay_(un)map_regs
471 static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
473 struct drm_device *dev = overlay->dev;
474 drm_i915_private_t *dev_priv = dev->dev_private;
477 /* Only wait if there is actually an old frame to release to
478 * guarantee forward progress.
480 if (!overlay->old_vid_bo)
483 if (I915_READ(ISR) & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT) {
484 struct drm_i915_gem_request *request;
486 /* synchronous slowpath */
487 request = kzalloc(sizeof(*request), GFP_KERNEL);
491 ret = BEGIN_LP_RING(2);
497 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
501 ret = intel_overlay_do_wait_request(overlay, request, true,
502 intel_overlay_release_old_vid_tail);
507 intel_overlay_release_old_vid_tail(overlay);
511 struct put_image_params {
528 static int packed_depth_bytes(u32 format)
530 switch (format & I915_OVERLAY_DEPTH_MASK) {
531 case I915_OVERLAY_YUV422:
533 case I915_OVERLAY_YUV411:
534 /* return 6; not implemented */
540 static int packed_width_bytes(u32 format, short width)
542 switch (format & I915_OVERLAY_DEPTH_MASK) {
543 case I915_OVERLAY_YUV422:
550 static int uv_hsubsampling(u32 format)
552 switch (format & I915_OVERLAY_DEPTH_MASK) {
553 case I915_OVERLAY_YUV422:
554 case I915_OVERLAY_YUV420:
556 case I915_OVERLAY_YUV411:
557 case I915_OVERLAY_YUV410:
564 static int uv_vsubsampling(u32 format)
566 switch (format & I915_OVERLAY_DEPTH_MASK) {
567 case I915_OVERLAY_YUV420:
568 case I915_OVERLAY_YUV410:
570 case I915_OVERLAY_YUV422:
571 case I915_OVERLAY_YUV411:
578 static u32 calc_swidthsw(struct drm_device *dev, u32 offset, u32 width)
580 u32 mask, shift, ret;
588 ret = ((offset + width + mask) >> shift) - (offset >> shift);
595 static const u16 y_static_hcoeffs[N_HORIZ_Y_TAPS * N_PHASES] = {
596 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
597 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
598 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
599 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
600 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
601 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
602 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
603 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
604 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
605 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
606 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
607 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
608 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
609 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
610 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
611 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
612 0xb000, 0x3000, 0x0800, 0x3000, 0xb000
615 static const u16 uv_static_hcoeffs[N_HORIZ_UV_TAPS * N_PHASES] = {
616 0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
617 0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
618 0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
619 0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
620 0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
621 0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
622 0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
623 0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
624 0x3000, 0x0800, 0x3000
627 static void update_polyphase_filter(struct overlay_registers *regs)
629 memcpy(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
630 memcpy(regs->UV_HCOEFS, uv_static_hcoeffs, sizeof(uv_static_hcoeffs));
633 static bool update_scaling_factors(struct intel_overlay *overlay,
634 struct overlay_registers *regs,
635 struct put_image_params *params)
637 /* fixed point with a 12 bit shift */
638 u32 xscale, yscale, xscale_UV, yscale_UV;
640 #define FRACT_MASK 0xfff
641 bool scale_changed = false;
642 int uv_hscale = uv_hsubsampling(params->format);
643 int uv_vscale = uv_vsubsampling(params->format);
645 if (params->dst_w > 1)
646 xscale = ((params->src_scan_w - 1) << FP_SHIFT)
649 xscale = 1 << FP_SHIFT;
651 if (params->dst_h > 1)
652 yscale = ((params->src_scan_h - 1) << FP_SHIFT)
655 yscale = 1 << FP_SHIFT;
657 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
658 xscale_UV = xscale/uv_hscale;
659 yscale_UV = yscale/uv_vscale;
660 /* make the Y scale to UV scale ratio an exact multiply */
661 xscale = xscale_UV * uv_hscale;
662 yscale = yscale_UV * uv_vscale;
668 if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
669 scale_changed = true;
670 overlay->old_xscale = xscale;
671 overlay->old_yscale = yscale;
673 regs->YRGBSCALE = (((yscale & FRACT_MASK) << 20) |
674 ((xscale >> FP_SHIFT) << 16) |
675 ((xscale & FRACT_MASK) << 3));
677 regs->UVSCALE = (((yscale_UV & FRACT_MASK) << 20) |
678 ((xscale_UV >> FP_SHIFT) << 16) |
679 ((xscale_UV & FRACT_MASK) << 3));
681 regs->UVSCALEV = ((((yscale >> FP_SHIFT) << 16) |
682 ((yscale_UV >> FP_SHIFT) << 0)));
685 update_polyphase_filter(regs);
687 return scale_changed;
690 static void update_colorkey(struct intel_overlay *overlay,
691 struct overlay_registers *regs)
693 u32 key = overlay->color_key;
695 switch (overlay->crtc->base.fb->bits_per_pixel) {
698 regs->DCLRKM = CLK_RGB8I_MASK | DST_KEY_ENABLE;
702 if (overlay->crtc->base.fb->depth == 15) {
703 regs->DCLRKV = RGB15_TO_COLORKEY(key);
704 regs->DCLRKM = CLK_RGB15_MASK | DST_KEY_ENABLE;
706 regs->DCLRKV = RGB16_TO_COLORKEY(key);
707 regs->DCLRKM = CLK_RGB16_MASK | DST_KEY_ENABLE;
714 regs->DCLRKM = CLK_RGB24_MASK | DST_KEY_ENABLE;
719 static u32 overlay_cmd_reg(struct put_image_params *params)
721 u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
723 if (params->format & I915_OVERLAY_YUV_PLANAR) {
724 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
725 case I915_OVERLAY_YUV422:
726 cmd |= OCMD_YUV_422_PLANAR;
728 case I915_OVERLAY_YUV420:
729 cmd |= OCMD_YUV_420_PLANAR;
731 case I915_OVERLAY_YUV411:
732 case I915_OVERLAY_YUV410:
733 cmd |= OCMD_YUV_410_PLANAR;
736 } else { /* YUV packed */
737 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
738 case I915_OVERLAY_YUV422:
739 cmd |= OCMD_YUV_422_PACKED;
741 case I915_OVERLAY_YUV411:
742 cmd |= OCMD_YUV_411_PACKED;
746 switch (params->format & I915_OVERLAY_SWAP_MASK) {
747 case I915_OVERLAY_NO_SWAP:
749 case I915_OVERLAY_UV_SWAP:
752 case I915_OVERLAY_Y_SWAP:
755 case I915_OVERLAY_Y_AND_UV_SWAP:
756 cmd |= OCMD_Y_AND_UV_SWAP;
764 static int intel_overlay_do_put_image(struct intel_overlay *overlay,
765 struct drm_i915_gem_object *new_bo,
766 struct put_image_params *params)
769 struct overlay_registers *regs;
770 bool scale_changed = false;
771 struct drm_device *dev = overlay->dev;
773 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
774 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
777 ret = intel_overlay_release_old_vid(overlay);
781 ret = i915_gem_object_pin(new_bo, PAGE_SIZE, true);
785 ret = i915_gem_object_set_to_gtt_domain(new_bo, 0);
789 ret = i915_gem_object_put_fence(new_bo);
793 if (!overlay->active) {
794 regs = intel_overlay_map_regs(overlay);
799 regs->OCONFIG = OCONF_CC_OUT_8BIT;
800 if (IS_GEN4(overlay->dev))
801 regs->OCONFIG |= OCONF_CSC_MODE_BT709;
802 regs->OCONFIG |= overlay->crtc->pipe == 0 ?
803 OCONF_PIPE_A : OCONF_PIPE_B;
804 intel_overlay_unmap_regs(overlay, regs);
806 ret = intel_overlay_on(overlay);
811 regs = intel_overlay_map_regs(overlay);
817 regs->DWINPOS = (params->dst_y << 16) | params->dst_x;
818 regs->DWINSZ = (params->dst_h << 16) | params->dst_w;
820 if (params->format & I915_OVERLAY_YUV_PACKED)
821 tmp_width = packed_width_bytes(params->format, params->src_w);
823 tmp_width = params->src_w;
825 regs->SWIDTH = params->src_w;
826 regs->SWIDTHSW = calc_swidthsw(overlay->dev,
827 params->offset_Y, tmp_width);
828 regs->SHEIGHT = params->src_h;
829 regs->OBUF_0Y = new_bo->gtt_offset + params-> offset_Y;
830 regs->OSTRIDE = params->stride_Y;
832 if (params->format & I915_OVERLAY_YUV_PLANAR) {
833 int uv_hscale = uv_hsubsampling(params->format);
834 int uv_vscale = uv_vsubsampling(params->format);
836 regs->SWIDTH |= (params->src_w/uv_hscale) << 16;
837 tmp_U = calc_swidthsw(overlay->dev, params->offset_U,
838 params->src_w/uv_hscale);
839 tmp_V = calc_swidthsw(overlay->dev, params->offset_V,
840 params->src_w/uv_hscale);
841 regs->SWIDTHSW |= max_t(u32, tmp_U, tmp_V) << 16;
842 regs->SHEIGHT |= (params->src_h/uv_vscale) << 16;
843 regs->OBUF_0U = new_bo->gtt_offset + params->offset_U;
844 regs->OBUF_0V = new_bo->gtt_offset + params->offset_V;
845 regs->OSTRIDE |= params->stride_UV << 16;
848 scale_changed = update_scaling_factors(overlay, regs, params);
850 update_colorkey(overlay, regs);
852 regs->OCMD = overlay_cmd_reg(params);
854 intel_overlay_unmap_regs(overlay, regs);
856 ret = intel_overlay_continue(overlay, scale_changed);
860 overlay->old_vid_bo = overlay->vid_bo;
861 overlay->vid_bo = new_bo;
866 i915_gem_object_unpin(new_bo);
870 int intel_overlay_switch_off(struct intel_overlay *overlay,
873 struct overlay_registers *regs;
874 struct drm_device *dev = overlay->dev;
877 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
878 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
880 ret = intel_overlay_recover_from_interrupt(overlay, interruptible);
884 if (!overlay->active)
887 ret = intel_overlay_release_old_vid(overlay);
891 regs = intel_overlay_map_regs(overlay);
893 intel_overlay_unmap_regs(overlay, regs);
895 ret = intel_overlay_off(overlay, interruptible);
899 intel_overlay_off_tail(overlay);
903 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
904 struct intel_crtc *crtc)
906 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
911 /* can't use the overlay with double wide pipe */
912 if (INTEL_INFO(overlay->dev)->gen < 4 &&
913 (I915_READ(PIPECONF(crtc->pipe)) & (PIPECONF_DOUBLE_WIDE | PIPECONF_ENABLE)) != PIPECONF_ENABLE)
919 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
921 struct drm_device *dev = overlay->dev;
922 drm_i915_private_t *dev_priv = dev->dev_private;
923 u32 pfit_control = I915_READ(PFIT_CONTROL);
926 /* XXX: This is not the same logic as in the xorg driver, but more in
927 * line with the intel documentation for the i965
929 if (INTEL_INFO(dev)->gen >= 4) {
930 /* on i965 use the PGM reg to read out the autoscaler values */
931 ratio = I915_READ(PFIT_PGM_RATIOS) >> PFIT_VERT_SCALE_SHIFT_965;
933 if (pfit_control & VERT_AUTO_SCALE)
934 ratio = I915_READ(PFIT_AUTO_RATIOS);
936 ratio = I915_READ(PFIT_PGM_RATIOS);
937 ratio >>= PFIT_VERT_SCALE_SHIFT;
940 overlay->pfit_vscale_ratio = ratio;
943 static int check_overlay_dst(struct intel_overlay *overlay,
944 struct drm_intel_overlay_put_image *rec)
946 struct drm_display_mode *mode = &overlay->crtc->base.mode;
948 if (rec->dst_x < mode->crtc_hdisplay &&
949 rec->dst_x + rec->dst_width <= mode->crtc_hdisplay &&
950 rec->dst_y < mode->crtc_vdisplay &&
951 rec->dst_y + rec->dst_height <= mode->crtc_vdisplay)
957 static int check_overlay_scaling(struct put_image_params *rec)
961 /* downscaling limit is 8.0 */
962 tmp = ((rec->src_scan_h << 16) / rec->dst_h) >> 16;
965 tmp = ((rec->src_scan_w << 16) / rec->dst_w) >> 16;
972 static int check_overlay_src(struct drm_device *dev,
973 struct drm_intel_overlay_put_image *rec,
974 struct drm_i915_gem_object *new_bo)
976 int uv_hscale = uv_hsubsampling(rec->flags);
977 int uv_vscale = uv_vsubsampling(rec->flags);
982 /* check src dimensions */
983 if (IS_845G(dev) || IS_I830(dev)) {
984 if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY ||
985 rec->src_width > IMAGE_MAX_WIDTH_LEGACY)
988 if (rec->src_height > IMAGE_MAX_HEIGHT ||
989 rec->src_width > IMAGE_MAX_WIDTH)
993 /* better safe than sorry, use 4 as the maximal subsampling ratio */
994 if (rec->src_height < N_VERT_Y_TAPS*4 ||
995 rec->src_width < N_HORIZ_Y_TAPS*4)
998 /* check alignment constraints */
999 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1000 case I915_OVERLAY_RGB:
1001 /* not implemented */
1004 case I915_OVERLAY_YUV_PACKED:
1008 depth = packed_depth_bytes(rec->flags);
1012 /* ignore UV planes */
1016 /* check pixel alignment */
1017 if (rec->offset_Y % depth)
1021 case I915_OVERLAY_YUV_PLANAR:
1022 if (uv_vscale < 0 || uv_hscale < 0)
1024 /* no offset restrictions for planar formats */
1031 if (rec->src_width % uv_hscale)
1034 /* stride checking */
1035 if (IS_I830(dev) || IS_845G(dev))
1040 if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
1042 if (IS_GEN4(dev) && rec->stride_Y < 512)
1045 tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
1047 if (rec->stride_Y > tmp || rec->stride_UV > 2*1024)
1050 /* check buffer dimensions */
1051 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1052 case I915_OVERLAY_RGB:
1053 case I915_OVERLAY_YUV_PACKED:
1054 /* always 4 Y values per depth pixels */
1055 if (packed_width_bytes(rec->flags, rec->src_width) > rec->stride_Y)
1058 tmp = rec->stride_Y*rec->src_height;
1059 if (rec->offset_Y + tmp > new_bo->base.size)
1063 case I915_OVERLAY_YUV_PLANAR:
1064 if (rec->src_width > rec->stride_Y)
1066 if (rec->src_width/uv_hscale > rec->stride_UV)
1069 tmp = rec->stride_Y * rec->src_height;
1070 if (rec->offset_Y + tmp > new_bo->base.size)
1073 tmp = rec->stride_UV * (rec->src_height / uv_vscale);
1074 if (rec->offset_U + tmp > new_bo->base.size ||
1075 rec->offset_V + tmp > new_bo->base.size)
1084 * Return the pipe currently connected to the panel fitter,
1085 * or -1 if the panel fitter is not present or not in use
1087 static int intel_panel_fitter_pipe(struct drm_device *dev)
1089 struct drm_i915_private *dev_priv = dev->dev_private;
1092 /* i830 doesn't have a panel fitter */
1096 pfit_control = I915_READ(PFIT_CONTROL);
1098 /* See if the panel fitter is in use */
1099 if ((pfit_control & PFIT_ENABLE) == 0)
1102 /* 965 can place panel fitter on either pipe */
1104 return (pfit_control >> 29) & 0x3;
1106 /* older chips can only use pipe 1 */
1110 int intel_overlay_put_image(struct drm_device *dev, void *data,
1111 struct drm_file *file_priv)
1113 struct drm_intel_overlay_put_image *put_image_rec = data;
1114 drm_i915_private_t *dev_priv = dev->dev_private;
1115 struct intel_overlay *overlay;
1116 struct drm_mode_object *drmmode_obj;
1117 struct intel_crtc *crtc;
1118 struct drm_i915_gem_object *new_bo;
1119 struct put_image_params *params;
1123 DRM_ERROR("called with no initialization\n");
1127 overlay = dev_priv->overlay;
1129 DRM_DEBUG("userspace bug: no overlay\n");
1133 if (!(put_image_rec->flags & I915_OVERLAY_ENABLE)) {
1134 mutex_lock(&dev->mode_config.mutex);
1135 mutex_lock(&dev->struct_mutex);
1137 ret = intel_overlay_switch_off(overlay, true);
1139 mutex_unlock(&dev->struct_mutex);
1140 mutex_unlock(&dev->mode_config.mutex);
1145 params = kmalloc(sizeof(struct put_image_params), GFP_KERNEL);
1149 drmmode_obj = drm_mode_object_find(dev, put_image_rec->crtc_id,
1150 DRM_MODE_OBJECT_CRTC);
1155 crtc = to_intel_crtc(obj_to_crtc(drmmode_obj));
1157 new_bo = to_intel_bo(drm_gem_object_lookup(dev, file_priv,
1158 put_image_rec->bo_handle));
1164 mutex_lock(&dev->mode_config.mutex);
1165 mutex_lock(&dev->struct_mutex);
1167 if (new_bo->tiling_mode) {
1168 DRM_ERROR("buffer used for overlay image can not be tiled\n");
1173 ret = intel_overlay_recover_from_interrupt(overlay, true);
1177 if (overlay->crtc != crtc) {
1178 struct drm_display_mode *mode = &crtc->base.mode;
1179 ret = intel_overlay_switch_off(overlay, true);
1183 ret = check_overlay_possible_on_crtc(overlay, crtc);
1187 overlay->crtc = crtc;
1188 crtc->overlay = overlay;
1190 /* line too wide, i.e. one-line-mode */
1191 if (mode->hdisplay > 1024 &&
1192 intel_panel_fitter_pipe(dev) == crtc->pipe) {
1193 overlay->pfit_active = 1;
1194 update_pfit_vscale_ratio(overlay);
1196 overlay->pfit_active = 0;
1199 ret = check_overlay_dst(overlay, put_image_rec);
1203 if (overlay->pfit_active) {
1204 params->dst_y = ((((u32)put_image_rec->dst_y) << 12) /
1205 overlay->pfit_vscale_ratio);
1206 /* shifting right rounds downwards, so add 1 */
1207 params->dst_h = ((((u32)put_image_rec->dst_height) << 12) /
1208 overlay->pfit_vscale_ratio) + 1;
1210 params->dst_y = put_image_rec->dst_y;
1211 params->dst_h = put_image_rec->dst_height;
1213 params->dst_x = put_image_rec->dst_x;
1214 params->dst_w = put_image_rec->dst_width;
1216 params->src_w = put_image_rec->src_width;
1217 params->src_h = put_image_rec->src_height;
1218 params->src_scan_w = put_image_rec->src_scan_width;
1219 params->src_scan_h = put_image_rec->src_scan_height;
1220 if (params->src_scan_h > params->src_h ||
1221 params->src_scan_w > params->src_w) {
1226 ret = check_overlay_src(dev, put_image_rec, new_bo);
1229 params->format = put_image_rec->flags & ~I915_OVERLAY_FLAGS_MASK;
1230 params->stride_Y = put_image_rec->stride_Y;
1231 params->stride_UV = put_image_rec->stride_UV;
1232 params->offset_Y = put_image_rec->offset_Y;
1233 params->offset_U = put_image_rec->offset_U;
1234 params->offset_V = put_image_rec->offset_V;
1236 /* Check scaling after src size to prevent a divide-by-zero. */
1237 ret = check_overlay_scaling(params);
1241 ret = intel_overlay_do_put_image(overlay, new_bo, params);
1245 mutex_unlock(&dev->struct_mutex);
1246 mutex_unlock(&dev->mode_config.mutex);
1253 mutex_unlock(&dev->struct_mutex);
1254 mutex_unlock(&dev->mode_config.mutex);
1255 drm_gem_object_unreference_unlocked(&new_bo->base);
1262 static void update_reg_attrs(struct intel_overlay *overlay,
1263 struct overlay_registers *regs)
1265 regs->OCLRC0 = (overlay->contrast << 18) | (overlay->brightness & 0xff);
1266 regs->OCLRC1 = overlay->saturation;
1269 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
1273 if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
1276 for (i = 0; i < 3; i++) {
1277 if (((gamma1 >> i*8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
1284 static bool check_gamma5_errata(u32 gamma5)
1288 for (i = 0; i < 3; i++) {
1289 if (((gamma5 >> i*8) & 0xff) == 0x80)
1296 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
1298 if (!check_gamma_bounds(0, attrs->gamma0) ||
1299 !check_gamma_bounds(attrs->gamma0, attrs->gamma1) ||
1300 !check_gamma_bounds(attrs->gamma1, attrs->gamma2) ||
1301 !check_gamma_bounds(attrs->gamma2, attrs->gamma3) ||
1302 !check_gamma_bounds(attrs->gamma3, attrs->gamma4) ||
1303 !check_gamma_bounds(attrs->gamma4, attrs->gamma5) ||
1304 !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
1307 if (!check_gamma5_errata(attrs->gamma5))
1313 int intel_overlay_attrs(struct drm_device *dev, void *data,
1314 struct drm_file *file_priv)
1316 struct drm_intel_overlay_attrs *attrs = data;
1317 drm_i915_private_t *dev_priv = dev->dev_private;
1318 struct intel_overlay *overlay;
1319 struct overlay_registers *regs;
1323 DRM_ERROR("called with no initialization\n");
1327 overlay = dev_priv->overlay;
1329 DRM_DEBUG("userspace bug: no overlay\n");
1333 mutex_lock(&dev->mode_config.mutex);
1334 mutex_lock(&dev->struct_mutex);
1337 if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
1338 attrs->color_key = overlay->color_key;
1339 attrs->brightness = overlay->brightness;
1340 attrs->contrast = overlay->contrast;
1341 attrs->saturation = overlay->saturation;
1343 if (!IS_GEN2(dev)) {
1344 attrs->gamma0 = I915_READ(OGAMC0);
1345 attrs->gamma1 = I915_READ(OGAMC1);
1346 attrs->gamma2 = I915_READ(OGAMC2);
1347 attrs->gamma3 = I915_READ(OGAMC3);
1348 attrs->gamma4 = I915_READ(OGAMC4);
1349 attrs->gamma5 = I915_READ(OGAMC5);
1352 if (attrs->brightness < -128 || attrs->brightness > 127)
1354 if (attrs->contrast > 255)
1356 if (attrs->saturation > 1023)
1359 overlay->color_key = attrs->color_key;
1360 overlay->brightness = attrs->brightness;
1361 overlay->contrast = attrs->contrast;
1362 overlay->saturation = attrs->saturation;
1364 regs = intel_overlay_map_regs(overlay);
1370 update_reg_attrs(overlay, regs);
1372 intel_overlay_unmap_regs(overlay, regs);
1374 if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
1378 if (overlay->active) {
1383 ret = check_gamma(attrs);
1387 I915_WRITE(OGAMC0, attrs->gamma0);
1388 I915_WRITE(OGAMC1, attrs->gamma1);
1389 I915_WRITE(OGAMC2, attrs->gamma2);
1390 I915_WRITE(OGAMC3, attrs->gamma3);
1391 I915_WRITE(OGAMC4, attrs->gamma4);
1392 I915_WRITE(OGAMC5, attrs->gamma5);
1398 mutex_unlock(&dev->struct_mutex);
1399 mutex_unlock(&dev->mode_config.mutex);
1404 void intel_setup_overlay(struct drm_device *dev)
1406 drm_i915_private_t *dev_priv = dev->dev_private;
1407 struct intel_overlay *overlay;
1408 struct drm_i915_gem_object *reg_bo;
1409 struct overlay_registers *regs;
1412 if (!HAS_OVERLAY(dev))
1415 overlay = kzalloc(sizeof(struct intel_overlay), GFP_KERNEL);
1420 reg_bo = i915_gem_alloc_object(dev, PAGE_SIZE);
1423 overlay->reg_bo = reg_bo;
1425 if (OVERLAY_NEEDS_PHYSICAL(dev)) {
1426 ret = i915_gem_attach_phys_object(dev, reg_bo,
1427 I915_GEM_PHYS_OVERLAY_REGS,
1430 DRM_ERROR("failed to attach phys overlay regs\n");
1433 overlay->flip_addr = reg_bo->phys_obj->handle->busaddr;
1435 ret = i915_gem_object_pin(reg_bo, PAGE_SIZE, true);
1437 DRM_ERROR("failed to pin overlay register bo\n");
1440 overlay->flip_addr = reg_bo->gtt_offset;
1442 ret = i915_gem_object_set_to_gtt_domain(reg_bo, true);
1444 DRM_ERROR("failed to move overlay register bo into the GTT\n");
1449 /* init all values */
1450 overlay->color_key = 0x0101fe;
1451 overlay->brightness = -19;
1452 overlay->contrast = 75;
1453 overlay->saturation = 146;
1455 regs = intel_overlay_map_regs(overlay);
1459 memset(regs, 0, sizeof(struct overlay_registers));
1460 update_polyphase_filter(regs);
1461 update_reg_attrs(overlay, regs);
1463 intel_overlay_unmap_regs(overlay, regs);
1465 dev_priv->overlay = overlay;
1466 DRM_INFO("initialized overlay support\n");
1470 i915_gem_object_unpin(reg_bo);
1472 drm_gem_object_unreference(®_bo->base);
1478 void intel_cleanup_overlay(struct drm_device *dev)
1480 drm_i915_private_t *dev_priv = dev->dev_private;
1482 if (!dev_priv->overlay)
1485 /* The bo's should be free'd by the generic code already.
1486 * Furthermore modesetting teardown happens beforehand so the
1487 * hardware should be off already */
1488 BUG_ON(dev_priv->overlay->active);
1490 drm_gem_object_unreference_unlocked(&dev_priv->overlay->reg_bo->base);
1491 kfree(dev_priv->overlay);
1494 #ifdef CONFIG_DEBUG_FS
1495 #include <linux/seq_file.h>
1497 struct intel_overlay_error_state {
1498 struct overlay_registers regs;
1504 static struct overlay_registers *
1505 intel_overlay_map_regs_atomic(struct intel_overlay *overlay)
1507 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
1508 struct overlay_registers *regs;
1510 if (OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1511 regs = overlay->reg_bo->phys_obj->handle->vaddr;
1513 regs = io_mapping_map_atomic_wc(dev_priv->mm.gtt_mapping,
1514 overlay->reg_bo->gtt_offset);
1519 static void intel_overlay_unmap_regs_atomic(struct intel_overlay *overlay,
1520 struct overlay_registers *regs)
1522 if (!OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1523 io_mapping_unmap_atomic(regs);
1527 struct intel_overlay_error_state *
1528 intel_overlay_capture_error_state(struct drm_device *dev)
1530 drm_i915_private_t *dev_priv = dev->dev_private;
1531 struct intel_overlay *overlay = dev_priv->overlay;
1532 struct intel_overlay_error_state *error;
1533 struct overlay_registers __iomem *regs;
1535 if (!overlay || !overlay->active)
1538 error = kmalloc(sizeof(*error), GFP_ATOMIC);
1542 error->dovsta = I915_READ(DOVSTA);
1543 error->isr = I915_READ(ISR);
1544 if (OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1545 error->base = (long) overlay->reg_bo->phys_obj->handle->vaddr;
1547 error->base = (long) overlay->reg_bo->gtt_offset;
1549 regs = intel_overlay_map_regs_atomic(overlay);
1553 memcpy_fromio(&error->regs, regs, sizeof(struct overlay_registers));
1554 intel_overlay_unmap_regs_atomic(overlay, regs);
1564 intel_overlay_print_error_state(struct seq_file *m, struct intel_overlay_error_state *error)
1566 seq_printf(m, "Overlay, status: 0x%08x, interrupt: 0x%08x\n",
1567 error->dovsta, error->isr);
1568 seq_printf(m, " Register file at 0x%08lx:\n",
1571 #define P(x) seq_printf(m, " " #x ": 0x%08x\n", error->regs.x)