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 overlay->last_flip_req =
225 i915_add_request(dev, NULL, request, &dev_priv->render_ring);
226 if (overlay->last_flip_req == 0)
229 overlay->flip_tail = tail;
230 ret = i915_do_wait_request(dev,
231 overlay->last_flip_req, true,
232 &dev_priv->render_ring);
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_gem_request *request;
293 int pipe_a_quirk = 0;
296 BUG_ON(overlay->active);
300 pipe_a_quirk = i830_activate_pipe_a(dev);
301 if (pipe_a_quirk < 0)
305 request = kzalloc(sizeof(*request), GFP_KERNEL);
306 if (request == NULL) {
312 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_ON);
313 OUT_RING(overlay->flip_addr | OFC_UPDATE);
314 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
318 ret = intel_overlay_do_wait_request(overlay, request, true, NULL);
321 i830_deactivate_pipe_a(dev);
326 /* overlay needs to be enabled in OCMD reg */
327 static int intel_overlay_continue(struct intel_overlay *overlay,
328 bool load_polyphase_filter)
330 struct drm_device *dev = overlay->dev;
331 drm_i915_private_t *dev_priv = dev->dev_private;
332 struct drm_i915_gem_request *request;
333 u32 flip_addr = overlay->flip_addr;
336 BUG_ON(!overlay->active);
338 request = kzalloc(sizeof(*request), GFP_KERNEL);
342 if (load_polyphase_filter)
343 flip_addr |= OFC_UPDATE;
345 /* check for underruns */
346 tmp = I915_READ(DOVSTA);
348 DRM_DEBUG("overlay underrun, DOVSTA: %x\n", tmp);
351 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
355 overlay->last_flip_req =
356 i915_add_request(dev, NULL, request, &dev_priv->render_ring);
360 static void intel_overlay_release_old_vid_tail(struct intel_overlay *overlay)
362 struct drm_gem_object *obj = &overlay->old_vid_bo->base;
364 i915_gem_object_unpin(obj);
365 drm_gem_object_unreference(obj);
367 overlay->old_vid_bo = NULL;
370 static void intel_overlay_off_tail(struct intel_overlay *overlay)
372 struct drm_gem_object *obj;
374 /* never have the overlay hw on without showing a frame */
375 BUG_ON(!overlay->vid_bo);
376 obj = &overlay->vid_bo->base;
378 i915_gem_object_unpin(obj);
379 drm_gem_object_unreference(obj);
380 overlay->vid_bo = NULL;
382 overlay->crtc->overlay = NULL;
383 overlay->crtc = NULL;
387 /* overlay needs to be disabled in OCMD reg */
388 static int intel_overlay_off(struct intel_overlay *overlay,
391 struct drm_device *dev = overlay->dev;
392 u32 flip_addr = overlay->flip_addr;
393 struct drm_i915_gem_request *request;
395 BUG_ON(!overlay->active);
397 request = kzalloc(sizeof(*request), GFP_KERNEL);
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;
408 /* wait for overlay to go idle */
409 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_CONTINUE);
411 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
412 /* turn overlay off */
413 OUT_RING(MI_OVERLAY_FLIP | MI_OVERLAY_OFF);
415 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
418 return intel_overlay_do_wait_request(overlay, request, interruptible,
419 intel_overlay_off_tail);
422 /* recover from an interruption due to a signal
423 * We have to be careful not to repeat work forever an make forward progess. */
424 static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
427 struct drm_device *dev = overlay->dev;
428 drm_i915_private_t *dev_priv = dev->dev_private;
431 if (overlay->last_flip_req == 0)
434 ret = i915_do_wait_request(dev, overlay->last_flip_req,
435 interruptible, &dev_priv->render_ring);
439 if (overlay->flip_tail)
440 overlay->flip_tail(overlay);
442 overlay->last_flip_req = 0;
446 /* Wait for pending overlay flip and release old frame.
447 * Needs to be called before the overlay register are changed
448 * via intel_overlay_(un)map_regs
450 static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
452 struct drm_device *dev = overlay->dev;
453 drm_i915_private_t *dev_priv = dev->dev_private;
456 /* Only wait if there is actually an old frame to release to
457 * guarantee forward progress.
459 if (!overlay->old_vid_bo)
462 if (I915_READ(ISR) & I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT) {
463 struct drm_i915_gem_request *request;
465 /* synchronous slowpath */
466 request = kzalloc(sizeof(*request), GFP_KERNEL);
471 OUT_RING(MI_WAIT_FOR_EVENT | MI_WAIT_FOR_OVERLAY_FLIP);
475 ret = intel_overlay_do_wait_request(overlay, request, true,
476 intel_overlay_release_old_vid_tail);
481 intel_overlay_release_old_vid_tail(overlay);
485 struct put_image_params {
502 static int packed_depth_bytes(u32 format)
504 switch (format & I915_OVERLAY_DEPTH_MASK) {
505 case I915_OVERLAY_YUV422:
507 case I915_OVERLAY_YUV411:
508 /* return 6; not implemented */
514 static int packed_width_bytes(u32 format, short width)
516 switch (format & I915_OVERLAY_DEPTH_MASK) {
517 case I915_OVERLAY_YUV422:
524 static int uv_hsubsampling(u32 format)
526 switch (format & I915_OVERLAY_DEPTH_MASK) {
527 case I915_OVERLAY_YUV422:
528 case I915_OVERLAY_YUV420:
530 case I915_OVERLAY_YUV411:
531 case I915_OVERLAY_YUV410:
538 static int uv_vsubsampling(u32 format)
540 switch (format & I915_OVERLAY_DEPTH_MASK) {
541 case I915_OVERLAY_YUV420:
542 case I915_OVERLAY_YUV410:
544 case I915_OVERLAY_YUV422:
545 case I915_OVERLAY_YUV411:
552 static u32 calc_swidthsw(struct drm_device *dev, u32 offset, u32 width)
554 u32 mask, shift, ret;
562 ret = ((offset + width + mask) >> shift) - (offset >> shift);
569 static const u16 y_static_hcoeffs[N_HORIZ_Y_TAPS * N_PHASES] = {
570 0x3000, 0xb4a0, 0x1930, 0x1920, 0xb4a0,
571 0x3000, 0xb500, 0x19d0, 0x1880, 0xb440,
572 0x3000, 0xb540, 0x1a88, 0x2f80, 0xb3e0,
573 0x3000, 0xb580, 0x1b30, 0x2e20, 0xb380,
574 0x3000, 0xb5c0, 0x1bd8, 0x2cc0, 0xb320,
575 0x3020, 0xb5e0, 0x1c60, 0x2b80, 0xb2c0,
576 0x3020, 0xb5e0, 0x1cf8, 0x2a20, 0xb260,
577 0x3020, 0xb5e0, 0x1d80, 0x28e0, 0xb200,
578 0x3020, 0xb5c0, 0x1e08, 0x3f40, 0xb1c0,
579 0x3020, 0xb580, 0x1e78, 0x3ce0, 0xb160,
580 0x3040, 0xb520, 0x1ed8, 0x3aa0, 0xb120,
581 0x3040, 0xb4a0, 0x1f30, 0x3880, 0xb0e0,
582 0x3040, 0xb400, 0x1f78, 0x3680, 0xb0a0,
583 0x3020, 0xb340, 0x1fb8, 0x34a0, 0xb060,
584 0x3020, 0xb240, 0x1fe0, 0x32e0, 0xb040,
585 0x3020, 0xb140, 0x1ff8, 0x3160, 0xb020,
586 0xb000, 0x3000, 0x0800, 0x3000, 0xb000
589 static const u16 uv_static_hcoeffs[N_HORIZ_UV_TAPS * N_PHASES] = {
590 0x3000, 0x1800, 0x1800, 0xb000, 0x18d0, 0x2e60,
591 0xb000, 0x1990, 0x2ce0, 0xb020, 0x1a68, 0x2b40,
592 0xb040, 0x1b20, 0x29e0, 0xb060, 0x1bd8, 0x2880,
593 0xb080, 0x1c88, 0x3e60, 0xb0a0, 0x1d28, 0x3c00,
594 0xb0c0, 0x1db8, 0x39e0, 0xb0e0, 0x1e40, 0x37e0,
595 0xb100, 0x1eb8, 0x3620, 0xb100, 0x1f18, 0x34a0,
596 0xb100, 0x1f68, 0x3360, 0xb0e0, 0x1fa8, 0x3240,
597 0xb0c0, 0x1fe0, 0x3140, 0xb060, 0x1ff0, 0x30a0,
598 0x3000, 0x0800, 0x3000
601 static void update_polyphase_filter(struct overlay_registers *regs)
603 memcpy(regs->Y_HCOEFS, y_static_hcoeffs, sizeof(y_static_hcoeffs));
604 memcpy(regs->UV_HCOEFS, uv_static_hcoeffs, sizeof(uv_static_hcoeffs));
607 static bool update_scaling_factors(struct intel_overlay *overlay,
608 struct overlay_registers *regs,
609 struct put_image_params *params)
611 /* fixed point with a 12 bit shift */
612 u32 xscale, yscale, xscale_UV, yscale_UV;
614 #define FRACT_MASK 0xfff
615 bool scale_changed = false;
616 int uv_hscale = uv_hsubsampling(params->format);
617 int uv_vscale = uv_vsubsampling(params->format);
619 if (params->dst_w > 1)
620 xscale = ((params->src_scan_w - 1) << FP_SHIFT)
623 xscale = 1 << FP_SHIFT;
625 if (params->dst_h > 1)
626 yscale = ((params->src_scan_h - 1) << FP_SHIFT)
629 yscale = 1 << FP_SHIFT;
631 /*if (params->format & I915_OVERLAY_YUV_PLANAR) {*/
632 xscale_UV = xscale/uv_hscale;
633 yscale_UV = yscale/uv_vscale;
634 /* make the Y scale to UV scale ratio an exact multiply */
635 xscale = xscale_UV * uv_hscale;
636 yscale = yscale_UV * uv_vscale;
642 if (xscale != overlay->old_xscale || yscale != overlay->old_yscale)
643 scale_changed = true;
644 overlay->old_xscale = xscale;
645 overlay->old_yscale = yscale;
647 regs->YRGBSCALE = (((yscale & FRACT_MASK) << 20) |
648 ((xscale >> FP_SHIFT) << 16) |
649 ((xscale & FRACT_MASK) << 3));
651 regs->UVSCALE = (((yscale_UV & FRACT_MASK) << 20) |
652 ((xscale_UV >> FP_SHIFT) << 16) |
653 ((xscale_UV & FRACT_MASK) << 3));
655 regs->UVSCALEV = ((((yscale >> FP_SHIFT) << 16) |
656 ((yscale_UV >> FP_SHIFT) << 0)));
659 update_polyphase_filter(regs);
661 return scale_changed;
664 static void update_colorkey(struct intel_overlay *overlay,
665 struct overlay_registers *regs)
667 u32 key = overlay->color_key;
669 switch (overlay->crtc->base.fb->bits_per_pixel) {
672 regs->DCLRKM = CLK_RGB8I_MASK | DST_KEY_ENABLE;
676 if (overlay->crtc->base.fb->depth == 15) {
677 regs->DCLRKV = RGB15_TO_COLORKEY(key);
678 regs->DCLRKM = CLK_RGB15_MASK | DST_KEY_ENABLE;
680 regs->DCLRKV = RGB16_TO_COLORKEY(key);
681 regs->DCLRKM = CLK_RGB16_MASK | DST_KEY_ENABLE;
688 regs->DCLRKM = CLK_RGB24_MASK | DST_KEY_ENABLE;
693 static u32 overlay_cmd_reg(struct put_image_params *params)
695 u32 cmd = OCMD_ENABLE | OCMD_BUF_TYPE_FRAME | OCMD_BUFFER0;
697 if (params->format & I915_OVERLAY_YUV_PLANAR) {
698 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
699 case I915_OVERLAY_YUV422:
700 cmd |= OCMD_YUV_422_PLANAR;
702 case I915_OVERLAY_YUV420:
703 cmd |= OCMD_YUV_420_PLANAR;
705 case I915_OVERLAY_YUV411:
706 case I915_OVERLAY_YUV410:
707 cmd |= OCMD_YUV_410_PLANAR;
710 } else { /* YUV packed */
711 switch (params->format & I915_OVERLAY_DEPTH_MASK) {
712 case I915_OVERLAY_YUV422:
713 cmd |= OCMD_YUV_422_PACKED;
715 case I915_OVERLAY_YUV411:
716 cmd |= OCMD_YUV_411_PACKED;
720 switch (params->format & I915_OVERLAY_SWAP_MASK) {
721 case I915_OVERLAY_NO_SWAP:
723 case I915_OVERLAY_UV_SWAP:
726 case I915_OVERLAY_Y_SWAP:
729 case I915_OVERLAY_Y_AND_UV_SWAP:
730 cmd |= OCMD_Y_AND_UV_SWAP;
738 static int intel_overlay_do_put_image(struct intel_overlay *overlay,
739 struct drm_gem_object *new_bo,
740 struct put_image_params *params)
743 struct overlay_registers *regs;
744 bool scale_changed = false;
745 struct drm_i915_gem_object *bo_priv = to_intel_bo(new_bo);
746 struct drm_device *dev = overlay->dev;
748 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
749 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
752 ret = intel_overlay_release_old_vid(overlay);
756 ret = i915_gem_object_pin(new_bo, PAGE_SIZE);
760 ret = i915_gem_object_set_to_gtt_domain(new_bo, 0);
764 if (!overlay->active) {
765 regs = intel_overlay_map_regs(overlay);
770 regs->OCONFIG = OCONF_CC_OUT_8BIT;
771 if (IS_I965GM(overlay->dev))
772 regs->OCONFIG |= OCONF_CSC_MODE_BT709;
773 regs->OCONFIG |= overlay->crtc->pipe == 0 ?
774 OCONF_PIPE_A : OCONF_PIPE_B;
775 intel_overlay_unmap_regs(overlay, regs);
777 ret = intel_overlay_on(overlay);
782 regs = intel_overlay_map_regs(overlay);
788 regs->DWINPOS = (params->dst_y << 16) | params->dst_x;
789 regs->DWINSZ = (params->dst_h << 16) | params->dst_w;
791 if (params->format & I915_OVERLAY_YUV_PACKED)
792 tmp_width = packed_width_bytes(params->format, params->src_w);
794 tmp_width = params->src_w;
796 regs->SWIDTH = params->src_w;
797 regs->SWIDTHSW = calc_swidthsw(overlay->dev,
798 params->offset_Y, tmp_width);
799 regs->SHEIGHT = params->src_h;
800 regs->OBUF_0Y = bo_priv->gtt_offset + params-> offset_Y;
801 regs->OSTRIDE = params->stride_Y;
803 if (params->format & I915_OVERLAY_YUV_PLANAR) {
804 int uv_hscale = uv_hsubsampling(params->format);
805 int uv_vscale = uv_vsubsampling(params->format);
807 regs->SWIDTH |= (params->src_w/uv_hscale) << 16;
808 tmp_U = calc_swidthsw(overlay->dev, params->offset_U,
809 params->src_w/uv_hscale);
810 tmp_V = calc_swidthsw(overlay->dev, params->offset_V,
811 params->src_w/uv_hscale);
812 regs->SWIDTHSW |= max_t(u32, tmp_U, tmp_V) << 16;
813 regs->SHEIGHT |= (params->src_h/uv_vscale) << 16;
814 regs->OBUF_0U = bo_priv->gtt_offset + params->offset_U;
815 regs->OBUF_0V = bo_priv->gtt_offset + params->offset_V;
816 regs->OSTRIDE |= params->stride_UV << 16;
819 scale_changed = update_scaling_factors(overlay, regs, params);
821 update_colorkey(overlay, regs);
823 regs->OCMD = overlay_cmd_reg(params);
825 intel_overlay_unmap_regs(overlay, regs);
827 ret = intel_overlay_continue(overlay, scale_changed);
831 overlay->old_vid_bo = overlay->vid_bo;
832 overlay->vid_bo = to_intel_bo(new_bo);
837 i915_gem_object_unpin(new_bo);
841 int intel_overlay_switch_off(struct intel_overlay *overlay,
844 struct overlay_registers *regs;
845 struct drm_device *dev = overlay->dev;
848 BUG_ON(!mutex_is_locked(&dev->struct_mutex));
849 BUG_ON(!mutex_is_locked(&dev->mode_config.mutex));
851 ret = intel_overlay_recover_from_interrupt(overlay, interruptible);
855 if (!overlay->active)
858 ret = intel_overlay_release_old_vid(overlay);
862 regs = intel_overlay_map_regs(overlay);
864 intel_overlay_unmap_regs(overlay, regs);
866 ret = intel_overlay_off(overlay, interruptible);
870 intel_overlay_off_tail(overlay);
874 static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
875 struct intel_crtc *crtc)
877 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
882 /* can't use the overlay with double wide pipe */
883 if (!IS_I965G(overlay->dev) &&
884 (I915_READ(PIPECONF(crtc->pipe)) & (PIPECONF_DOUBLE_WIDE | PIPECONF_ENABLE)) != PIPECONF_ENABLE)
890 static void update_pfit_vscale_ratio(struct intel_overlay *overlay)
892 struct drm_device *dev = overlay->dev;
893 drm_i915_private_t *dev_priv = dev->dev_private;
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 (!IS_I965G(dev)) {
901 if (pfit_control & VERT_AUTO_SCALE)
902 ratio = I915_READ(PFIT_AUTO_RATIOS);
904 ratio = I915_READ(PFIT_PGM_RATIOS);
905 ratio >>= PFIT_VERT_SCALE_SHIFT;
906 } else { /* on i965 use the PGM reg to read out the autoscaler values */
907 ratio = I915_READ(PFIT_PGM_RATIOS) >> PFIT_VERT_SCALE_SHIFT_965;
910 overlay->pfit_vscale_ratio = ratio;
913 static int check_overlay_dst(struct intel_overlay *overlay,
914 struct drm_intel_overlay_put_image *rec)
916 struct drm_display_mode *mode = &overlay->crtc->base.mode;
918 if (rec->dst_x < mode->crtc_hdisplay &&
919 rec->dst_x + rec->dst_width <= mode->crtc_hdisplay &&
920 rec->dst_y < mode->crtc_vdisplay &&
921 rec->dst_y + rec->dst_height <= mode->crtc_vdisplay)
927 static int check_overlay_scaling(struct put_image_params *rec)
931 /* downscaling limit is 8.0 */
932 tmp = ((rec->src_scan_h << 16) / rec->dst_h) >> 16;
935 tmp = ((rec->src_scan_w << 16) / rec->dst_w) >> 16;
942 static int check_overlay_src(struct drm_device *dev,
943 struct drm_intel_overlay_put_image *rec,
944 struct drm_gem_object *new_bo)
946 int uv_hscale = uv_hsubsampling(rec->flags);
947 int uv_vscale = uv_vsubsampling(rec->flags);
948 u32 stride_mask, depth, tmp;
950 /* check src dimensions */
951 if (IS_845G(dev) || IS_I830(dev)) {
952 if (rec->src_height > IMAGE_MAX_HEIGHT_LEGACY ||
953 rec->src_width > IMAGE_MAX_WIDTH_LEGACY)
956 if (rec->src_height > IMAGE_MAX_HEIGHT ||
957 rec->src_width > IMAGE_MAX_WIDTH)
961 /* better safe than sorry, use 4 as the maximal subsampling ratio */
962 if (rec->src_height < N_VERT_Y_TAPS*4 ||
963 rec->src_width < N_HORIZ_Y_TAPS*4)
966 /* check alignment constraints */
967 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
968 case I915_OVERLAY_RGB:
969 /* not implemented */
972 case I915_OVERLAY_YUV_PACKED:
976 depth = packed_depth_bytes(rec->flags);
980 /* ignore UV planes */
984 /* check pixel alignment */
985 if (rec->offset_Y % depth)
989 case I915_OVERLAY_YUV_PLANAR:
990 if (uv_vscale < 0 || uv_hscale < 0)
992 /* no offset restrictions for planar formats */
999 if (rec->src_width % uv_hscale)
1002 /* stride checking */
1003 if (IS_I830(dev) || IS_845G(dev))
1008 if (rec->stride_Y & stride_mask || rec->stride_UV & stride_mask)
1010 if (IS_I965G(dev) && rec->stride_Y < 512)
1013 tmp = (rec->flags & I915_OVERLAY_TYPE_MASK) == I915_OVERLAY_YUV_PLANAR ?
1015 if (rec->stride_Y > tmp || rec->stride_UV > 2*1024)
1018 /* check buffer dimensions */
1019 switch (rec->flags & I915_OVERLAY_TYPE_MASK) {
1020 case I915_OVERLAY_RGB:
1021 case I915_OVERLAY_YUV_PACKED:
1022 /* always 4 Y values per depth pixels */
1023 if (packed_width_bytes(rec->flags, rec->src_width) > rec->stride_Y)
1026 tmp = rec->stride_Y*rec->src_height;
1027 if (rec->offset_Y + tmp > new_bo->size)
1031 case I915_OVERLAY_YUV_PLANAR:
1032 if (rec->src_width > rec->stride_Y)
1034 if (rec->src_width/uv_hscale > rec->stride_UV)
1037 tmp = rec->stride_Y * rec->src_height;
1038 if (rec->offset_Y + tmp > new_bo->size)
1041 tmp = rec->stride_UV * (rec->src_height / uv_vscale);
1042 if (rec->offset_U + tmp > new_bo->size ||
1043 rec->offset_V + tmp > new_bo->size)
1052 * Return the pipe currently connected to the panel fitter,
1053 * or -1 if the panel fitter is not present or not in use
1055 static int intel_panel_fitter_pipe(struct drm_device *dev)
1057 struct drm_i915_private *dev_priv = dev->dev_private;
1060 /* i830 doesn't have a panel fitter */
1064 pfit_control = I915_READ(PFIT_CONTROL);
1066 /* See if the panel fitter is in use */
1067 if ((pfit_control & PFIT_ENABLE) == 0)
1070 /* 965 can place panel fitter on either pipe */
1072 return (pfit_control >> 29) & 0x3;
1074 /* older chips can only use pipe 1 */
1078 int intel_overlay_put_image(struct drm_device *dev, void *data,
1079 struct drm_file *file_priv)
1081 struct drm_intel_overlay_put_image *put_image_rec = data;
1082 drm_i915_private_t *dev_priv = dev->dev_private;
1083 struct intel_overlay *overlay;
1084 struct drm_mode_object *drmmode_obj;
1085 struct intel_crtc *crtc;
1086 struct drm_gem_object *new_bo;
1087 struct put_image_params *params;
1091 DRM_ERROR("called with no initialization\n");
1095 overlay = dev_priv->overlay;
1097 DRM_DEBUG("userspace bug: no overlay\n");
1101 if (!(put_image_rec->flags & I915_OVERLAY_ENABLE)) {
1102 mutex_lock(&dev->mode_config.mutex);
1103 mutex_lock(&dev->struct_mutex);
1105 ret = intel_overlay_switch_off(overlay, true);
1107 mutex_unlock(&dev->struct_mutex);
1108 mutex_unlock(&dev->mode_config.mutex);
1113 params = kmalloc(sizeof(struct put_image_params), GFP_KERNEL);
1117 drmmode_obj = drm_mode_object_find(dev, put_image_rec->crtc_id,
1118 DRM_MODE_OBJECT_CRTC);
1123 crtc = to_intel_crtc(obj_to_crtc(drmmode_obj));
1125 new_bo = drm_gem_object_lookup(dev, file_priv,
1126 put_image_rec->bo_handle);
1132 mutex_lock(&dev->mode_config.mutex);
1133 mutex_lock(&dev->struct_mutex);
1135 ret = intel_overlay_recover_from_interrupt(overlay, true);
1139 if (overlay->crtc != crtc) {
1140 struct drm_display_mode *mode = &crtc->base.mode;
1141 ret = intel_overlay_switch_off(overlay, true);
1145 ret = check_overlay_possible_on_crtc(overlay, crtc);
1149 overlay->crtc = crtc;
1150 crtc->overlay = overlay;
1152 /* line too wide, i.e. one-line-mode */
1153 if (mode->hdisplay > 1024 &&
1154 intel_panel_fitter_pipe(dev) == crtc->pipe) {
1155 overlay->pfit_active = 1;
1156 update_pfit_vscale_ratio(overlay);
1158 overlay->pfit_active = 0;
1161 ret = check_overlay_dst(overlay, put_image_rec);
1165 if (overlay->pfit_active) {
1166 params->dst_y = ((((u32)put_image_rec->dst_y) << 12) /
1167 overlay->pfit_vscale_ratio);
1168 /* shifting right rounds downwards, so add 1 */
1169 params->dst_h = ((((u32)put_image_rec->dst_height) << 12) /
1170 overlay->pfit_vscale_ratio) + 1;
1172 params->dst_y = put_image_rec->dst_y;
1173 params->dst_h = put_image_rec->dst_height;
1175 params->dst_x = put_image_rec->dst_x;
1176 params->dst_w = put_image_rec->dst_width;
1178 params->src_w = put_image_rec->src_width;
1179 params->src_h = put_image_rec->src_height;
1180 params->src_scan_w = put_image_rec->src_scan_width;
1181 params->src_scan_h = put_image_rec->src_scan_height;
1182 if (params->src_scan_h > params->src_h ||
1183 params->src_scan_w > params->src_w) {
1188 ret = check_overlay_src(dev, put_image_rec, new_bo);
1191 params->format = put_image_rec->flags & ~I915_OVERLAY_FLAGS_MASK;
1192 params->stride_Y = put_image_rec->stride_Y;
1193 params->stride_UV = put_image_rec->stride_UV;
1194 params->offset_Y = put_image_rec->offset_Y;
1195 params->offset_U = put_image_rec->offset_U;
1196 params->offset_V = put_image_rec->offset_V;
1198 /* Check scaling after src size to prevent a divide-by-zero. */
1199 ret = check_overlay_scaling(params);
1203 ret = intel_overlay_do_put_image(overlay, new_bo, params);
1207 mutex_unlock(&dev->struct_mutex);
1208 mutex_unlock(&dev->mode_config.mutex);
1215 mutex_unlock(&dev->struct_mutex);
1216 mutex_unlock(&dev->mode_config.mutex);
1217 drm_gem_object_unreference_unlocked(new_bo);
1224 static void update_reg_attrs(struct intel_overlay *overlay,
1225 struct overlay_registers *regs)
1227 regs->OCLRC0 = (overlay->contrast << 18) | (overlay->brightness & 0xff);
1228 regs->OCLRC1 = overlay->saturation;
1231 static bool check_gamma_bounds(u32 gamma1, u32 gamma2)
1235 if (gamma1 & 0xff000000 || gamma2 & 0xff000000)
1238 for (i = 0; i < 3; i++) {
1239 if (((gamma1 >> i*8) & 0xff) >= ((gamma2 >> i*8) & 0xff))
1246 static bool check_gamma5_errata(u32 gamma5)
1250 for (i = 0; i < 3; i++) {
1251 if (((gamma5 >> i*8) & 0xff) == 0x80)
1258 static int check_gamma(struct drm_intel_overlay_attrs *attrs)
1260 if (!check_gamma_bounds(0, attrs->gamma0) ||
1261 !check_gamma_bounds(attrs->gamma0, attrs->gamma1) ||
1262 !check_gamma_bounds(attrs->gamma1, attrs->gamma2) ||
1263 !check_gamma_bounds(attrs->gamma2, attrs->gamma3) ||
1264 !check_gamma_bounds(attrs->gamma3, attrs->gamma4) ||
1265 !check_gamma_bounds(attrs->gamma4, attrs->gamma5) ||
1266 !check_gamma_bounds(attrs->gamma5, 0x00ffffff))
1269 if (!check_gamma5_errata(attrs->gamma5))
1275 int intel_overlay_attrs(struct drm_device *dev, void *data,
1276 struct drm_file *file_priv)
1278 struct drm_intel_overlay_attrs *attrs = data;
1279 drm_i915_private_t *dev_priv = dev->dev_private;
1280 struct intel_overlay *overlay;
1281 struct overlay_registers *regs;
1285 DRM_ERROR("called with no initialization\n");
1289 overlay = dev_priv->overlay;
1291 DRM_DEBUG("userspace bug: no overlay\n");
1295 mutex_lock(&dev->mode_config.mutex);
1296 mutex_lock(&dev->struct_mutex);
1299 if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
1300 attrs->color_key = overlay->color_key;
1301 attrs->brightness = overlay->brightness;
1302 attrs->contrast = overlay->contrast;
1303 attrs->saturation = overlay->saturation;
1306 attrs->gamma0 = I915_READ(OGAMC0);
1307 attrs->gamma1 = I915_READ(OGAMC1);
1308 attrs->gamma2 = I915_READ(OGAMC2);
1309 attrs->gamma3 = I915_READ(OGAMC3);
1310 attrs->gamma4 = I915_READ(OGAMC4);
1311 attrs->gamma5 = I915_READ(OGAMC5);
1314 if (attrs->brightness < -128 || attrs->brightness > 127)
1316 if (attrs->contrast > 255)
1318 if (attrs->saturation > 1023)
1321 overlay->color_key = attrs->color_key;
1322 overlay->brightness = attrs->brightness;
1323 overlay->contrast = attrs->contrast;
1324 overlay->saturation = attrs->saturation;
1326 regs = intel_overlay_map_regs(overlay);
1332 update_reg_attrs(overlay, regs);
1334 intel_overlay_unmap_regs(overlay, regs);
1336 if (attrs->flags & I915_OVERLAY_UPDATE_GAMMA) {
1340 if (overlay->active) {
1345 ret = check_gamma(attrs);
1349 I915_WRITE(OGAMC0, attrs->gamma0);
1350 I915_WRITE(OGAMC1, attrs->gamma1);
1351 I915_WRITE(OGAMC2, attrs->gamma2);
1352 I915_WRITE(OGAMC3, attrs->gamma3);
1353 I915_WRITE(OGAMC4, attrs->gamma4);
1354 I915_WRITE(OGAMC5, attrs->gamma5);
1360 mutex_unlock(&dev->struct_mutex);
1361 mutex_unlock(&dev->mode_config.mutex);
1366 void intel_setup_overlay(struct drm_device *dev)
1368 drm_i915_private_t *dev_priv = dev->dev_private;
1369 struct intel_overlay *overlay;
1370 struct drm_gem_object *reg_bo;
1371 struct overlay_registers *regs;
1374 if (!HAS_OVERLAY(dev))
1377 overlay = kzalloc(sizeof(struct intel_overlay), GFP_KERNEL);
1382 reg_bo = i915_gem_alloc_object(dev, PAGE_SIZE);
1385 overlay->reg_bo = to_intel_bo(reg_bo);
1387 if (OVERLAY_NEEDS_PHYSICAL(dev)) {
1388 ret = i915_gem_attach_phys_object(dev, reg_bo,
1389 I915_GEM_PHYS_OVERLAY_REGS,
1392 DRM_ERROR("failed to attach phys overlay regs\n");
1395 overlay->flip_addr = overlay->reg_bo->phys_obj->handle->busaddr;
1397 ret = i915_gem_object_pin(reg_bo, PAGE_SIZE);
1399 DRM_ERROR("failed to pin overlay register bo\n");
1402 overlay->flip_addr = overlay->reg_bo->gtt_offset;
1404 ret = i915_gem_object_set_to_gtt_domain(reg_bo, true);
1406 DRM_ERROR("failed to move overlay register bo into the GTT\n");
1411 /* init all values */
1412 overlay->color_key = 0x0101fe;
1413 overlay->brightness = -19;
1414 overlay->contrast = 75;
1415 overlay->saturation = 146;
1417 regs = intel_overlay_map_regs(overlay);
1421 memset(regs, 0, sizeof(struct overlay_registers));
1422 update_polyphase_filter(regs);
1423 update_reg_attrs(overlay, regs);
1425 intel_overlay_unmap_regs(overlay, regs);
1427 dev_priv->overlay = overlay;
1428 DRM_INFO("initialized overlay support\n");
1432 i915_gem_object_unpin(reg_bo);
1434 drm_gem_object_unreference(reg_bo);
1440 void intel_cleanup_overlay(struct drm_device *dev)
1442 drm_i915_private_t *dev_priv = dev->dev_private;
1444 if (!dev_priv->overlay)
1447 /* The bo's should be free'd by the generic code already.
1448 * Furthermore modesetting teardown happens beforehand so the
1449 * hardware should be off already */
1450 BUG_ON(dev_priv->overlay->active);
1452 drm_gem_object_unreference_unlocked(&dev_priv->overlay->reg_bo->base);
1453 kfree(dev_priv->overlay);
1456 #ifdef CONFIG_DEBUG_FS
1457 #include <linux/seq_file.h>
1459 struct intel_overlay_error_state {
1460 struct overlay_registers regs;
1466 static struct overlay_registers *
1467 intel_overlay_map_regs_atomic(struct intel_overlay *overlay,
1470 drm_i915_private_t *dev_priv = overlay->dev->dev_private;
1471 struct overlay_registers *regs;
1473 if (OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1474 regs = overlay->reg_bo->phys_obj->handle->vaddr;
1476 regs = io_mapping_map_atomic_wc(dev_priv->mm.gtt_mapping,
1477 overlay->reg_bo->gtt_offset,
1483 static void intel_overlay_unmap_regs_atomic(struct intel_overlay *overlay,
1485 struct overlay_registers *regs)
1487 if (!OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1488 io_mapping_unmap_atomic(regs, slot);
1492 struct intel_overlay_error_state *
1493 intel_overlay_capture_error_state(struct drm_device *dev)
1495 drm_i915_private_t *dev_priv = dev->dev_private;
1496 struct intel_overlay *overlay = dev_priv->overlay;
1497 struct intel_overlay_error_state *error;
1498 struct overlay_registers __iomem *regs;
1500 if (!overlay || !overlay->active)
1503 error = kmalloc(sizeof(*error), GFP_ATOMIC);
1507 error->dovsta = I915_READ(DOVSTA);
1508 error->isr = I915_READ(ISR);
1509 if (OVERLAY_NEEDS_PHYSICAL(overlay->dev))
1510 error->base = (long) overlay->reg_bo->phys_obj->handle->vaddr;
1512 error->base = (long) overlay->reg_bo->gtt_offset;
1514 regs = intel_overlay_map_regs_atomic(overlay, KM_IRQ0);
1518 memcpy_fromio(&error->regs, regs, sizeof(struct overlay_registers));
1519 intel_overlay_unmap_regs_atomic(overlay, KM_IRQ0, regs);
1529 intel_overlay_print_error_state(struct seq_file *m, struct intel_overlay_error_state *error)
1531 seq_printf(m, "Overlay, status: 0x%08x, interrupt: 0x%08x\n",
1532 error->dovsta, error->isr);
1533 seq_printf(m, " Register file at 0x%08lx:\n",
1536 #define P(x) seq_printf(m, " " #x ": 0x%08x\n", error->regs.x)