/*- * Copyright (c) 2015 Michal Meloun * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "tegra_drm_if.h" #include "tegra_dc_if.h" #define WR4(_sc, _r, _v) bus_write_4((_sc)->mem_res, 4 * (_r), (_v)) #define RD4(_sc, _r) bus_read_4((_sc)->mem_res, 4 * (_r)) #define LOCK(_sc) mtx_lock(&(_sc)->mtx) #define UNLOCK(_sc) mtx_unlock(&(_sc)->mtx) #define SLEEP(_sc, timeout) \ mtx_sleep(sc, &sc->mtx, 0, "tegra_dc_wait", timeout); #define LOCK_INIT(_sc) \ mtx_init(&_sc->mtx, device_get_nameunit(_sc->dev), "tegra_dc", MTX_DEF) #define LOCK_DESTROY(_sc) mtx_destroy(&_sc->mtx) #define ASSERT_LOCKED(_sc) mtx_assert(&_sc->mtx, MA_OWNED) #define ASSERT_UNLOCKED(_sc) mtx_assert(&_sc->mtx, MA_NOTOWNED) #define SYNCPT_VBLANK0 26 #define SYNCPT_VBLANK1 27 #define DC_MAX_PLANES 2 /* Maximum planes */ /* DRM Formats supported by DC */ /* XXXX expand me */ static uint32_t dc_plane_formats[] = { DRM_FORMAT_XBGR8888, DRM_FORMAT_XRGB8888, DRM_FORMAT_RGB565, DRM_FORMAT_UYVY, DRM_FORMAT_YUYV, DRM_FORMAT_YUV420, DRM_FORMAT_YUV422, }; /* Complete description of one window (plane) */ struct dc_window { /* Source (in framebuffer) rectangle, in pixels */ u_int src_x; u_int src_y; u_int src_w; u_int src_h; /* Destination (on display) rectangle, in pixels */ u_int dst_x; u_int dst_y; u_int dst_w; u_int dst_h; /* Parsed pixel format */ u_int bits_per_pixel; bool is_yuv; /* any YUV mode */ bool is_yuv_planar; /* planar YUV mode */ uint32_t color_mode; /* DC_WIN_COLOR_DEPTH */ uint32_t swap; /* DC_WIN_BYTE_SWAP */ uint32_t surface_kind; /* DC_WINBUF_SURFACE_KIND */ uint32_t block_height; /* DC_WINBUF_SURFACE_KIND */ /* Parsed flipping, rotation is not supported for pitched modes */ bool flip_x; /* inverted X-axis */ bool flip_y; /* inverted Y-axis */ bool transpose_xy; /* swap X and Y-axis */ /* Color planes base addresses and strides */ bus_size_t base[3]; uint32_t stride[3]; /* stride[2] isn't used by HW */ }; struct dc_softc { device_t dev; struct resource *mem_res; struct resource *irq_res; void *irq_ih; struct mtx mtx; clk_t clk_parent; clk_t clk_dc; hwreset_t hwreset_dc; int pitch_align; struct tegra_crtc tegra_crtc; struct drm_pending_vblank_event *event; struct drm_gem_object *cursor_gem; }; static struct ofw_compat_data compat_data[] = { {"nvidia,tegra124-dc", 1}, {NULL, 0}, }; /* Convert standard drm pixel format to tegra windows parameters. */ static int dc_parse_drm_format(struct tegra_fb *fb, struct dc_window *win) { struct tegra_bo *bo; uint32_t cm; uint32_t sw; bool is_yuv, is_yuv_planar; int nplanes, i; switch (fb->drm_fb.pixel_format) { case DRM_FORMAT_XBGR8888: sw = BYTE_SWAP(NOSWAP); cm = WIN_COLOR_DEPTH_R8G8B8A8; is_yuv = false; is_yuv_planar = false; break; case DRM_FORMAT_XRGB8888: sw = BYTE_SWAP(NOSWAP); cm = WIN_COLOR_DEPTH_B8G8R8A8; is_yuv = false; is_yuv_planar = false; break; case DRM_FORMAT_RGB565: sw = BYTE_SWAP(NOSWAP); cm = WIN_COLOR_DEPTH_B5G6R5; is_yuv = false; is_yuv_planar = false; break; case DRM_FORMAT_UYVY: sw = BYTE_SWAP(NOSWAP); cm = WIN_COLOR_DEPTH_YCbCr422; is_yuv = true; is_yuv_planar = false; break; case DRM_FORMAT_YUYV: sw = BYTE_SWAP(SWAP2); cm = WIN_COLOR_DEPTH_YCbCr422; is_yuv = true; is_yuv_planar = false; break; case DRM_FORMAT_YUV420: sw = BYTE_SWAP(NOSWAP); cm = WIN_COLOR_DEPTH_YCbCr420P; is_yuv = true; is_yuv_planar = true; break; case DRM_FORMAT_YUV422: sw = BYTE_SWAP(NOSWAP); cm = WIN_COLOR_DEPTH_YCbCr422P; is_yuv = true; is_yuv_planar = true; break; default: /* Unsupported format */ return (-EINVAL); } /* Basic check of arguments. */ switch (fb->rotation) { case 0: case 180: break; case 90: /* Rotation is supported only */ case 270: /* for block linear surfaces */ if (!fb->block_linear) return (-EINVAL); break; default: return (-EINVAL); } /* XXX Add more checks (sizes, scaling...) */ if (win == NULL) return (0); win->surface_kind = fb->block_linear ? SURFACE_KIND_BL_16B2: SURFACE_KIND_PITCH; win->block_height = fb->block_height; switch (fb->rotation) { case 0: /* (0,0,0) */ win->transpose_xy = false; win->flip_x = false; win->flip_y = false; break; case 90: /* (1,0,1) */ win->transpose_xy = true; win->flip_x = false; win->flip_y = true; break; case 180: /* (0,1,1) */ win->transpose_xy = false; win->flip_x = true; win->flip_y = true; break; case 270: /* (1,1,0) */ win->transpose_xy = true; win->flip_x = true; win->flip_y = false; break; } win->flip_x ^= fb->flip_x; win->flip_y ^= fb->flip_y; win->color_mode = cm; win->swap = sw; win->bits_per_pixel = fb->drm_fb.bits_per_pixel; win->is_yuv = is_yuv; win->is_yuv_planar = is_yuv_planar; nplanes = drm_format_num_planes(fb->drm_fb.pixel_format); for (i = 0; i < nplanes; i++) { bo = fb->planes[i]; win->base[i] = bo->pbase + fb->drm_fb.offsets[i]; win->stride[i] = fb->drm_fb.pitches[i]; } return (0); } /* * Scaling functions. * * It's unclear if we want/must program the fractional portion * (aka bias) of init_dda registers, mainly when mirrored axis * modes are used. * For now, we use 1.0 as recommended by TRM. */ static inline uint32_t dc_scaling_init(uint32_t start) { return (1 << 12); } static inline uint32_t dc_scaling_incr(uint32_t src, uint32_t dst, uint32_t maxscale) { uint32_t val; val = (src - 1) << 12 ; /* 4.12 fixed float */ val /= (dst - 1); if (val > (maxscale << 12)) val = maxscale << 12; return val; } /* ------------------------------------------------------------------- * * HW Access. * */ /* * Setup pixel clock. * Minimal frequency is pixel clock, but output is free to select * any higher. */ static int dc_setup_clk(struct dc_softc *sc, struct drm_crtc *crtc, struct drm_display_mode *mode, uint32_t *div) { uint64_t pclk, freq; struct tegra_drm_encoder *output; struct drm_encoder *encoder; long rv; pclk = mode->clock * 1000; /* Find attached encoder */ output = NULL; list_for_each_entry(encoder, &crtc->dev->mode_config.encoder_list, head) { if (encoder->crtc == crtc) { output = container_of(encoder, struct tegra_drm_encoder, encoder); break; } } if (output == NULL) return (-ENODEV); if (output->setup_clock == NULL) panic("Output have not setup_clock function.\n"); rv = output->setup_clock(output, sc->clk_dc, pclk); if (rv != 0) { device_printf(sc->dev, "Cannot setup pixel clock: %llu\n", pclk); return (rv); } rv = clk_get_freq(sc->clk_dc, &freq); *div = (freq * 2 / pclk) - 2; DRM_DEBUG_KMS("frequency: %llu, DC divider: %u\n", freq, *div); return 0; } static void dc_setup_window(struct dc_softc *sc, unsigned int index, struct dc_window *win) { uint32_t h_offset, v_offset, h_size, v_size, bpp; uint32_t h_init_dda, v_init_dda, h_incr_dda, v_incr_dda; uint32_t val; #ifdef DMR_DEBUG_WINDOW printf("%s window: %d\n", __func__, index); printf(" src: x: %d, y: %d, w: %d, h: %d\n", win->src_x, win->src_y, win->src_w, win->src_h); printf(" dst: x: %d, y: %d, w: %d, h: %d\n", win->dst_x, win->dst_y, win->dst_w, win->dst_h); printf(" bpp: %d, color_mode: %d, swap: %d\n", win->bits_per_pixel, win->color_mode, win->swap); #endif if (win->is_yuv) bpp = win->is_yuv_planar ? 1 : 2; else bpp = (win->bits_per_pixel + 7) / 8; if (!win->transpose_xy) { h_size = win->src_w * bpp; v_size = win->src_h; } else { h_size = win->src_h * bpp; v_size = win->src_w; } h_offset = win->src_x * bpp; v_offset = win->src_y; if (win->flip_x) { h_offset += win->src_w * bpp - 1; } if (win->flip_y) v_offset += win->src_h - 1; /* Adjust offsets for planar yuv modes */ if (win->is_yuv_planar) { h_offset &= ~1; if (win->flip_x ) h_offset |= 1; v_offset &= ~1; if (win->flip_y ) v_offset |= 1; } /* Setup scaling. */ if (!win->transpose_xy) { h_init_dda = dc_scaling_init(win->src_x); v_init_dda = dc_scaling_init(win->src_y); h_incr_dda = dc_scaling_incr(win->src_w, win->dst_w, 4); v_incr_dda = dc_scaling_incr(win->src_h, win->dst_h, 15); } else { h_init_dda = dc_scaling_init(win->src_y); v_init_dda = dc_scaling_init(win->src_x); h_incr_dda = dc_scaling_incr(win->src_h, win->dst_h, 4); v_incr_dda = dc_scaling_incr(win->src_w, win->dst_w, 15); } #ifdef DMR_DEBUG_WINDOW printf("\n"); printf(" bpp: %d, size: h: %d v: %d, offset: h:%d v: %d\n", bpp, h_size, v_size, h_offset, v_offset); printf(" init_dda: h: %d v: %d, incr_dda: h: %d v: %d\n", h_init_dda, v_init_dda, h_incr_dda, v_incr_dda); #endif LOCK(sc); /* Select target window */ val = WINDOW_A_SELECT << index; WR4(sc, DC_CMD_DISPLAY_WINDOW_HEADER, val); /* Sizes */ WR4(sc, DC_WIN_POSITION, WIN_POSITION(win->dst_x, win->dst_y)); WR4(sc, DC_WIN_SIZE, WIN_SIZE(win->dst_w, win->dst_h)); WR4(sc, DC_WIN_PRESCALED_SIZE, WIN_PRESCALED_SIZE(h_size, v_size)); /* DDA */ WR4(sc, DC_WIN_DDA_INCREMENT, WIN_DDA_INCREMENT(h_incr_dda, v_incr_dda)); WR4(sc, DC_WIN_H_INITIAL_DDA, h_init_dda); WR4(sc, DC_WIN_V_INITIAL_DDA, v_init_dda); /* Color planes base addresses and strides */ WR4(sc, DC_WINBUF_START_ADDR, win->base[0]); if (win->is_yuv_planar) { WR4(sc, DC_WINBUF_START_ADDR_U, win->base[1]); WR4(sc, DC_WINBUF_START_ADDR_V, win->base[2]); WR4(sc, DC_WIN_LINE_STRIDE, win->stride[1] << 16 | win->stride[0]); } else { WR4(sc, DC_WIN_LINE_STRIDE, win->stride[0]); } /* Offsets for rotation and axis flip */ WR4(sc, DC_WINBUF_ADDR_H_OFFSET, h_offset); WR4(sc, DC_WINBUF_ADDR_V_OFFSET, v_offset); /* Color format */ WR4(sc, DC_WIN_COLOR_DEPTH, win->color_mode); WR4(sc, DC_WIN_BYTE_SWAP, win->swap); /* Tiling */ val = win->surface_kind; if (win->surface_kind == SURFACE_KIND_BL_16B2) val |= SURFACE_KIND_BLOCK_HEIGHT(win->block_height); WR4(sc, DC_WINBUF_SURFACE_KIND, val); /* Color space coefs for YUV modes */ if (win->is_yuv) { WR4(sc, DC_WINC_CSC_YOF, 0x00f0); WR4(sc, DC_WINC_CSC_KYRGB, 0x012a); WR4(sc, DC_WINC_CSC_KUR, 0x0000); WR4(sc, DC_WINC_CSC_KVR, 0x0198); WR4(sc, DC_WINC_CSC_KUG, 0x039b); WR4(sc, DC_WINC_CSC_KVG, 0x032f); WR4(sc, DC_WINC_CSC_KUB, 0x0204); WR4(sc, DC_WINC_CSC_KVB, 0x0000); } val = WIN_ENABLE; if (win->is_yuv) val |= CSC_ENABLE; else if (win->bits_per_pixel < 24) val |= COLOR_EXPAND; if (win->flip_y) val |= V_DIRECTION; if (win->flip_x) val |= H_DIRECTION; if (win->transpose_xy) val |= SCAN_COLUMN; WR4(sc, DC_WINC_WIN_OPTIONS, val); #ifdef DMR_DEBUG_WINDOW /* Set underflow debug mode -> highlight missing pixels. */ WR4(sc, DC_WINBUF_UFLOW_CTRL, UFLOW_CTR_ENABLE); WR4(sc, DC_WINBUF_UFLOW_DBG_PIXEL, 0xFFFF0000); #endif UNLOCK(sc); } /* ------------------------------------------------------------------- * * Plane functions. * */ static int dc_plane_update(struct drm_plane *drm_plane, struct drm_crtc *drm_crtc, struct drm_framebuffer *drm_fb, int crtc_x, int crtc_y, unsigned int crtc_w, unsigned int crtc_h, uint32_t src_x, uint32_t src_y, uint32_t src_w, uint32_t src_h) { struct tegra_plane *plane; struct tegra_crtc *crtc; struct tegra_fb *fb; struct dc_softc *sc; struct dc_window win; int rv; plane = container_of(drm_plane, struct tegra_plane, drm_plane); fb = container_of(drm_fb, struct tegra_fb, drm_fb); crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); memset(&win, 0, sizeof(win)); win.src_x = src_x >> 16; win.src_y = src_y >> 16; win.src_w = src_w >> 16; win.src_h = src_h >> 16; win.dst_x = crtc_x; win.dst_y = crtc_y; win.dst_w = crtc_w; win.dst_h = crtc_h; rv = dc_parse_drm_format(fb, &win); if (rv != 0) { DRM_WARNING("unsupported pixel format %d\n", fb->drm_fb.pixel_format); return (rv); } dc_setup_window(sc, plane->index, &win); WR4(sc, DC_CMD_STATE_CONTROL, WIN_A_UPDATE << plane->index); WR4(sc, DC_CMD_STATE_CONTROL, WIN_A_ACT_REQ << plane->index); return (0); } static int dc_plane_disable(struct drm_plane *drm_plane) { struct tegra_plane *plane; struct tegra_crtc *crtc; struct dc_softc *sc; uint32_t val, idx; if (drm_plane->crtc == NULL) return (0); plane = container_of(drm_plane, struct tegra_plane, drm_plane); crtc = container_of(drm_plane->crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); idx = plane->index; LOCK(sc); WR4(sc, DC_CMD_DISPLAY_WINDOW_HEADER, WINDOW_A_SELECT << idx); val = RD4(sc, DC_WINC_WIN_OPTIONS); val &= ~WIN_ENABLE; WR4(sc, DC_WINC_WIN_OPTIONS, val); UNLOCK(sc); WR4(sc, DC_CMD_STATE_CONTROL, WIN_A_UPDATE << idx); WR4(sc, DC_CMD_STATE_CONTROL, WIN_A_ACT_REQ << idx); return (0); } static void dc_plane_destroy(struct drm_plane *plane) { dc_plane_disable(plane); drm_plane_cleanup(plane); free(plane, DRM_MEM_KMS); } static const struct drm_plane_funcs dc_plane_funcs = { .update_plane = dc_plane_update, .disable_plane = dc_plane_disable, .destroy = dc_plane_destroy, }; /* ------------------------------------------------------------------- * * CRTC helper functions. * */ static void dc_crtc_dpms(struct drm_crtc *crtc, int mode) { /* Empty function */ } static bool dc_crtc_mode_fixup(struct drm_crtc *crtc, const struct drm_display_mode *mode, struct drm_display_mode *adjusted) { return (true); } static int dc_set_base(struct dc_softc *sc, int x, int y, struct tegra_fb *fb) { struct dc_window win; int rv; memset(&win, 0, sizeof(win)); win.src_x = x; win.src_y = y; win.src_w = fb->drm_fb.width; win.src_h = fb->drm_fb.height; win.dst_x = x; win.dst_y = y; win.dst_w = fb->drm_fb.width; win.dst_h = fb->drm_fb.height; rv = dc_parse_drm_format(fb, &win); if (rv != 0) { DRM_WARNING("unsupported pixel format %d\n", fb->drm_fb.pixel_format); return (rv); } dc_setup_window(sc, 0, &win); return (0); } static int dc_crtc_mode_set(struct drm_crtc *drm_crtc, struct drm_display_mode *mode, struct drm_display_mode *adjusted, int x, int y, struct drm_framebuffer *old_fb) { struct dc_softc *sc; struct tegra_crtc *crtc; struct tegra_fb *fb; struct dc_window win; uint32_t div, h_ref_to_sync, v_ref_to_sync; int rv; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); fb = container_of(drm_crtc->fb, struct tegra_fb, drm_fb); h_ref_to_sync = 1; v_ref_to_sync = 1; /* Setup timing */ rv = dc_setup_clk(sc, drm_crtc, mode, &div); if (rv != 0) { device_printf(sc->dev, "Cannot set pixel clock\n"); return (rv); } /* Timing */ WR4(sc, DC_DISP_DISP_TIMING_OPTIONS, 0); WR4(sc, DC_DISP_REF_TO_SYNC, (v_ref_to_sync << 16) | h_ref_to_sync); WR4(sc, DC_DISP_SYNC_WIDTH, ((mode->vsync_end - mode->vsync_start) << 16) | ((mode->hsync_end - mode->hsync_start) << 0)); WR4(sc, DC_DISP_BACK_PORCH, ((mode->vtotal - mode->vsync_end) << 16) | ((mode->htotal - mode->hsync_end) << 0)); WR4(sc, DC_DISP_FRONT_PORCH, ((mode->vsync_start - mode->vdisplay) << 16) | ((mode->hsync_start - mode->hdisplay) << 0)); WR4(sc, DC_DISP_DISP_ACTIVE, (mode->vdisplay << 16) | mode->hdisplay); WR4(sc, DC_DISP_DISP_INTERFACE_CONTROL, DISP_DATA_FORMAT(DF1P1C)); WR4(sc,DC_DISP_DISP_CLOCK_CONTROL, SHIFT_CLK_DIVIDER(div) | PIXEL_CLK_DIVIDER(PCD1)); memset(&win, 0, sizeof(win)); win.src_x = x; win.src_y = y; win.src_w = mode->hdisplay; win.src_h = mode->vdisplay; win.dst_x = x; win.dst_y = y; win.dst_w = mode->hdisplay; win.dst_h = mode->vdisplay; rv = dc_parse_drm_format(fb, &win); if (rv != 0) { DRM_WARNING("unsupported pixel format %d\n", drm_crtc->fb->pixel_format); return (rv); } dc_setup_window(sc, 0, &win); return (0); } static int dc_crtc_mode_set_base(struct drm_crtc *drm_crtc, int x, int y, struct drm_framebuffer *old_fb) { struct dc_softc *sc; struct tegra_crtc *crtc; struct tegra_fb *fb; int rv; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); fb = container_of(drm_crtc->fb, struct tegra_fb, drm_fb); sc = device_get_softc(crtc->dev); rv = dc_set_base(sc, x, y, fb); /* Commit */ WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_UPDATE | WIN_A_UPDATE); WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_ACT_REQ | WIN_A_ACT_REQ); return (rv); } static void dc_crtc_prepare(struct drm_crtc *drm_crtc) { struct dc_softc *sc; struct tegra_crtc *crtc; uint32_t val; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); WR4(sc, DC_CMD_GENERAL_INCR_SYNCPT_CNTRL, SYNCPT_CNTRL_NO_STALL); /* XXX allocate syncpoint from host1x */ WR4(sc, DC_CMD_CONT_SYNCPT_VSYNC, SYNCPT_VSYNC_ENABLE | (sc->tegra_crtc.nvidia_head == 0 ? SYNCPT_VBLANK0: SYNCPT_VBLANK1)); WR4(sc, DC_CMD_DISPLAY_POWER_CONTROL, PW0_ENABLE | PW1_ENABLE | PW2_ENABLE | PW3_ENABLE | PW4_ENABLE | PM0_ENABLE | PM1_ENABLE); val = RD4(sc, DC_CMD_DISPLAY_COMMAND); val |= DISPLAY_CTRL_MODE(CTRL_MODE_C_DISPLAY); WR4(sc, DC_CMD_DISPLAY_COMMAND, val); WR4(sc, DC_CMD_INT_MASK, WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT | WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT); WR4(sc, DC_CMD_INT_ENABLE, VBLANK_INT | WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT | WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT); } static void dc_crtc_commit(struct drm_crtc *drm_crtc) { struct dc_softc *sc; struct tegra_crtc *crtc; uint32_t val; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_UPDATE | WIN_A_UPDATE); val = RD4(sc, DC_CMD_INT_MASK); val |= FRAME_END_INT; WR4(sc, DC_CMD_INT_MASK, val); val = RD4(sc, DC_CMD_INT_ENABLE); val |= FRAME_END_INT; WR4(sc, DC_CMD_INT_ENABLE, val); WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_ACT_REQ | WIN_A_ACT_REQ); } static void dc_crtc_load_lut(struct drm_crtc *crtc) { /* empty function */ } static const struct drm_crtc_helper_funcs dc_crtc_helper_funcs = { .dpms = dc_crtc_dpms, .mode_fixup = dc_crtc_mode_fixup, .mode_set = dc_crtc_mode_set, .mode_set_base = dc_crtc_mode_set_base, .prepare = dc_crtc_prepare, .commit = dc_crtc_commit, .load_lut = dc_crtc_load_lut, }; static int drm_crtc_index(struct drm_crtc *crtc) { int idx; struct drm_crtc *tmp; idx = 0; list_for_each_entry(tmp, &crtc->dev->mode_config.crtc_list, head) { if (tmp == crtc) return (idx); idx++; } panic("Cannot find CRTC"); } /* ------------------------------------------------------------------- * * Exported functions (mainly vsync related). * * XXX revisit this -> convert to bus methods? */ int tegra_dc_get_pipe(struct drm_crtc *drm_crtc) { struct tegra_crtc *crtc; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); return (crtc->nvidia_head); } void tegra_dc_enable_vblank(struct drm_crtc *drm_crtc) { struct dc_softc *sc; struct tegra_crtc *crtc; uint32_t val; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); LOCK(sc); val = RD4(sc, DC_CMD_INT_MASK); val |= VBLANK_INT; WR4(sc, DC_CMD_INT_MASK, val); UNLOCK(sc); } void tegra_dc_disable_vblank(struct drm_crtc *drm_crtc) { struct dc_softc *sc; struct tegra_crtc *crtc; uint32_t val; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); LOCK(sc); val = RD4(sc, DC_CMD_INT_MASK); val &= ~VBLANK_INT; WR4(sc, DC_CMD_INT_MASK, val); UNLOCK(sc); } static void dc_finish_page_flip(struct dc_softc *sc) { struct drm_crtc *drm_crtc; struct drm_device *drm; struct tegra_fb *fb; struct tegra_bo *bo; uint32_t base; int idx; drm_crtc = &sc->tegra_crtc.drm_crtc; drm = drm_crtc->dev; fb = container_of(drm_crtc->fb, struct tegra_fb, drm_fb); mtx_lock(&drm->event_lock); if (sc->event == NULL) { mtx_unlock(&drm->event_lock); return; } LOCK(sc); /* Read active copy of WINBUF_START_ADDR */ WR4(sc, DC_CMD_DISPLAY_WINDOW_HEADER, WINDOW_A_SELECT); WR4(sc, DC_CMD_STATE_ACCESS, READ_MUX); base = RD4(sc, DC_WINBUF_START_ADDR); WR4(sc, DC_CMD_STATE_ACCESS, 0); UNLOCK(sc); /* Is already active */ bo = tegra_fb_get_plane(fb, 0); if (base == (bo->pbase + fb->drm_fb.offsets[0])) { idx = drm_crtc_index(drm_crtc); drm_send_vblank_event(drm, idx, sc->event); drm_vblank_put(drm, idx); sc->event = NULL; } mtx_unlock(&drm->event_lock); } void tegra_dc_cancel_page_flip(struct drm_crtc *drm_crtc, struct drm_file *file) { struct dc_softc *sc; struct tegra_crtc *crtc; struct drm_device *drm; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); drm = drm_crtc->dev; mtx_lock(&drm->event_lock); if ((sc->event != NULL) && (sc->event->base.file_priv == file)) { sc->event->base.destroy(&sc->event->base); drm_vblank_put(drm, drm_crtc_index(drm_crtc)); sc->event = NULL; } mtx_unlock(&drm->event_lock); } /* ------------------------------------------------------------------- * * CRTC functions. * */ static int dc_page_flip(struct drm_crtc *drm_crtc, struct drm_framebuffer *drm_fb, struct drm_pending_vblank_event *event) { struct dc_softc *sc; struct tegra_crtc *crtc; struct tegra_fb *fb; struct drm_device *drm; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); fb = container_of(drm_crtc->fb, struct tegra_fb, drm_fb); drm = drm_crtc->dev; if (sc->event != NULL) return (-EBUSY); if (event != NULL) { event->pipe = sc->tegra_crtc.nvidia_head; sc->event = event; drm_vblank_get(drm, event->pipe); } dc_set_base(sc, drm_crtc->x, drm_crtc->y, fb); drm_crtc->fb = drm_fb; /* Commit */ WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_UPDATE | WIN_A_UPDATE); return (0); } static int dc_cursor_set(struct drm_crtc *drm_crtc, struct drm_file *file, uint32_t handle, uint32_t width, uint32_t height) { struct dc_softc *sc; struct tegra_crtc *crtc; struct drm_gem_object *gem; struct tegra_bo *bo; int i; uint32_t val, *src, *dst; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); if (width != height) return (-EINVAL); switch (width) { case 32: val = CURSOR_SIZE(C32x32); break; case 64: val = CURSOR_SIZE(C64x64); break; case 128: val = CURSOR_SIZE(C128x128); break; case 256: val = CURSOR_SIZE(C256x256); break; default: return (-EINVAL); } bo = NULL; gem = NULL; if (handle != 0) { gem = drm_gem_object_lookup(drm_crtc->dev, file, handle); if (gem == NULL) return (-ENOENT); bo = container_of(gem, struct tegra_bo, gem_obj); } if (sc->cursor_gem != NULL) { drm_gem_object_unreference(sc->cursor_gem); } sc->cursor_gem = gem; if (bo != NULL) { /* * Copy cursor into cache and convert it from ARGB to RGBA. * XXXX - this is broken by design - client can write to BO at * any time. We can dedicate other window for cursor or switch * to sw cursor in worst case. */ src = (uint32_t *)bo->vbase; dst = (uint32_t *)crtc->cursor_vbase; for (i = 0; i < width * height; i++) dst[i] = (src[i] << 8) | (src[i] >> 24); val |= CURSOR_CLIP(CC_DISPLAY); val |= CURSOR_START_ADDR(crtc->cursor_pbase); WR4(sc, DC_DISP_CURSOR_START_ADDR, val); val = RD4(sc, DC_DISP_BLEND_CURSOR_CONTROL); val &= ~CURSOR_DST_BLEND_FACTOR_SELECT(~0); val &= ~CURSOR_SRC_BLEND_FACTOR_SELECT(~0); val |= CURSOR_MODE_SELECT; val |= CURSOR_DST_BLEND_FACTOR_SELECT(DST_NEG_K1_TIMES_SRC); val |= CURSOR_SRC_BLEND_FACTOR_SELECT(SRC_BLEND_K1_TIMES_SRC); val |= CURSOR_ALPHA(~0); WR4(sc, DC_DISP_BLEND_CURSOR_CONTROL, val); val = RD4(sc, DC_DISP_DISP_WIN_OPTIONS); val |= CURSOR_ENABLE; WR4(sc, DC_DISP_DISP_WIN_OPTIONS, val); } else { val = RD4(sc, DC_DISP_DISP_WIN_OPTIONS); val &= ~CURSOR_ENABLE; WR4(sc, DC_DISP_DISP_WIN_OPTIONS, val); } /* XXX This fixes cursor underflow issues, but why ? */ WR4(sc, DC_DISP_CURSOR_UNDERFLOW_CTRL, CURSOR_UFLOW_CYA); WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_UPDATE | CURSOR_UPDATE ); WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_ACT_REQ | CURSOR_ACT_REQ); return (0); } static int dc_cursor_move(struct drm_crtc *drm_crtc, int x, int y) { struct dc_softc *sc; struct tegra_crtc *crtc; crtc = container_of(drm_crtc, struct tegra_crtc, drm_crtc); sc = device_get_softc(crtc->dev); WR4(sc, DC_DISP_CURSOR_POSITION, CURSOR_POSITION(x, y)); WR4(sc, DC_CMD_STATE_CONTROL, CURSOR_UPDATE); WR4(sc, DC_CMD_STATE_CONTROL, CURSOR_ACT_REQ); return (0); } static void dc_destroy(struct drm_crtc *crtc) { drm_crtc_cleanup(crtc); memset(crtc, 0, sizeof(*crtc)); } static const struct drm_crtc_funcs dc_crtc_funcs = { .page_flip = dc_page_flip, .cursor_set = dc_cursor_set, .cursor_move = dc_cursor_move, .set_config = drm_crtc_helper_set_config, .destroy = dc_destroy, }; /* ------------------------------------------------------------------- * * Bus and infrastructure. * */ static int dc_init_planes(struct dc_softc *sc, struct tegra_drm *drm) { int i, rv; struct tegra_plane *plane; rv = 0; for (i = 0; i < DC_MAX_PLANES; i++) { plane = malloc(sizeof(*plane), DRM_MEM_KMS, M_WAITOK | M_ZERO); plane->index = i + 1; rv = drm_plane_init(&drm->drm_dev, &plane->drm_plane, 1 << sc->tegra_crtc.nvidia_head, &dc_plane_funcs, dc_plane_formats, nitems(dc_plane_formats), false); if (rv != 0) { free(plane, DRM_MEM_KMS); return (rv); } } return 0; } static void dc_display_enable(device_t dev, bool enable) { struct dc_softc *sc; uint32_t val; sc = device_get_softc(dev); /* Set display mode */ val = enable ? CTRL_MODE_C_DISPLAY: CTRL_MODE_STOP; WR4(sc, DC_CMD_DISPLAY_COMMAND, DISPLAY_CTRL_MODE(val)); /* and commit it*/ WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_UPDATE); WR4(sc, DC_CMD_STATE_CONTROL, GENERAL_ACT_REQ); } static void dc_hdmi_enable(device_t dev, bool enable) { struct dc_softc *sc; uint32_t val; sc = device_get_softc(dev); val = RD4(sc, DC_DISP_DISP_WIN_OPTIONS); if (enable) val |= HDMI_ENABLE; else val &= ~HDMI_ENABLE; WR4(sc, DC_DISP_DISP_WIN_OPTIONS, val); } static void dc_setup_timing(device_t dev, int h_pulse_start) { struct dc_softc *sc; sc = device_get_softc(dev); /* Setup display timing */ WR4(sc, DC_DISP_DISP_TIMING_OPTIONS, VSYNC_H_POSITION(1)); WR4(sc, DC_DISP_DISP_COLOR_CONTROL, DITHER_CONTROL(DITHER_DISABLE) | BASE_COLOR_SIZE(SIZE_BASE888)); WR4(sc, DC_DISP_DISP_SIGNAL_OPTIONS0, H_PULSE2_ENABLE); WR4(sc, DC_DISP_H_PULSE2_CONTROL, PULSE_CONTROL_QUAL(QUAL_VACTIVE) | PULSE_CONTROL_LAST(LAST_END_A)); WR4(sc, DC_DISP_H_PULSE2_POSITION_A, PULSE_START(h_pulse_start) | PULSE_END(h_pulse_start + 8)); } static void dc_intr(void *arg) { struct dc_softc *sc; uint32_t status; sc = arg; /* Confirm interrupt */ status = RD4(sc, DC_CMD_INT_STATUS); WR4(sc, DC_CMD_INT_STATUS, status); if (status & VBLANK_INT) { drm_handle_vblank(sc->tegra_crtc.drm_crtc.dev, sc->tegra_crtc.nvidia_head); dc_finish_page_flip(sc); } } static int dc_init_client(device_t dev, device_t host1x, struct tegra_drm *drm) { struct dc_softc *sc; int rv; sc = device_get_softc(dev); if (drm->pitch_align < sc->pitch_align) drm->pitch_align = sc->pitch_align; drm_crtc_init(&drm->drm_dev, &sc->tegra_crtc.drm_crtc, &dc_crtc_funcs); drm_mode_crtc_set_gamma_size(&sc->tegra_crtc.drm_crtc, 256); drm_crtc_helper_add(&sc->tegra_crtc.drm_crtc, &dc_crtc_helper_funcs); rv = dc_init_planes(sc, drm); if (rv!= 0){ device_printf(dev, "Cannot init planes\n"); return (rv); } WR4(sc, DC_CMD_INT_TYPE, WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT | WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT); WR4(sc, DC_CMD_INT_POLARITY, WIN_A_UF_INT | WIN_B_UF_INT | WIN_C_UF_INT | WIN_A_OF_INT | WIN_B_OF_INT | WIN_C_OF_INT); WR4(sc, DC_CMD_INT_ENABLE, 0); WR4(sc, DC_CMD_INT_MASK, 0); rv = bus_setup_intr(dev, sc->irq_res, INTR_TYPE_MISC | INTR_MPSAFE, NULL, dc_intr, sc, &sc->irq_ih); if (rv != 0) { device_printf(dev, "Cannot register interrupt handler\n"); return (rv); } /* allocate memory for cursor cache */ sc->tegra_crtc.cursor_vbase = kmem_alloc_contig(256 * 256 * 4, M_WAITOK | M_ZERO, 0, -1UL, PAGE_SIZE, 0, VM_MEMATTR_WRITE_COMBINING); sc->tegra_crtc.cursor_pbase = vtophys(sc->tegra_crtc.cursor_vbase); return (0); } static int dc_exit_client(device_t dev, device_t host1x, struct tegra_drm *drm) { struct dc_softc *sc; sc = device_get_softc(dev); if (sc->irq_ih != NULL) bus_teardown_intr(dev, sc->irq_res, sc->irq_ih); sc->irq_ih = NULL; return (0); } static int get_fdt_resources(struct dc_softc *sc, phandle_t node) { int rv; rv = hwreset_get_by_ofw_name(sc->dev, 0, "dc", &sc->hwreset_dc); if (rv != 0) { device_printf(sc->dev, "Cannot get 'dc' reset\n"); return (rv); } rv = clk_get_by_ofw_name(sc->dev, 0, "parent", &sc->clk_parent); if (rv != 0) { device_printf(sc->dev, "Cannot get 'parent' clock\n"); return (rv); } rv = clk_get_by_ofw_name(sc->dev, 0, "dc", &sc->clk_dc); if (rv != 0) { device_printf(sc->dev, "Cannot get 'dc' clock\n"); return (rv); } rv = OF_getencprop(node, "nvidia,head", &sc->tegra_crtc.nvidia_head, sizeof(sc->tegra_crtc.nvidia_head)); if (rv <= 0) { device_printf(sc->dev, "Cannot get 'nvidia,head' property\n"); return (rv); } return (0); } static int enable_fdt_resources(struct dc_softc *sc) { int id, rv; rv = clk_set_parent_by_clk(sc->clk_dc, sc->clk_parent); if (rv != 0) { device_printf(sc->dev, "Cannot set parent for 'dc' clock\n"); return (rv); } id = (sc->tegra_crtc.nvidia_head == 0) ? TEGRA_POWERGATE_DIS: TEGRA_POWERGATE_DISB; rv = tegra_powergate_sequence_power_up(id, sc->clk_dc, sc->hwreset_dc); if (rv != 0) { device_printf(sc->dev, "Cannot enable 'DIS' powergate\n"); return (rv); } return (0); } static int dc_probe(device_t dev) { if (!ofw_bus_status_okay(dev)) return (ENXIO); if (ofw_bus_search_compatible(dev, compat_data)->ocd_data == 0) return (ENXIO); device_set_desc(dev, "Tegra Display Controller"); return (BUS_PROBE_DEFAULT); } static int dc_attach(device_t dev) { struct dc_softc *sc; phandle_t node; int rid, rv; sc = device_get_softc(dev); sc->dev = dev; sc->tegra_crtc.dev = dev; node = ofw_bus_get_node(sc->dev); LOCK_INIT(sc); rid = 0; sc->mem_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid, RF_ACTIVE); if (sc->mem_res == NULL) { device_printf(dev, "Cannot allocate memory resources\n"); goto fail; } rid = 0; sc->irq_res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid, RF_ACTIVE); if (sc->irq_res == NULL) { device_printf(dev, "Cannot allocate IRQ resources\n"); goto fail; } rv = get_fdt_resources(sc, node); if (rv != 0) { device_printf(dev, "Cannot parse FDT resources\n"); goto fail; } rv = enable_fdt_resources(sc); if (rv != 0) { device_printf(dev, "Cannot enable FDT resources\n"); goto fail; } /* * Tegra124 * - 64 for RGB modes * - 128 for YUV planar modes * - 256 for block linear modes */ sc->pitch_align = 256; rv = TEGRA_DRM_REGISTER_CLIENT(device_get_parent(sc->dev), sc->dev); if (rv != 0) { device_printf(dev, "Cannot register DRM device\n"); goto fail; } return (bus_generic_attach(dev)); fail: TEGRA_DRM_DEREGISTER_CLIENT(device_get_parent(sc->dev), sc->dev); if (sc->irq_ih != NULL) bus_teardown_intr(dev, sc->irq_res, sc->irq_ih); if (sc->clk_parent != NULL) clk_release(sc->clk_parent); if (sc->clk_dc != NULL) clk_release(sc->clk_dc); if (sc->hwreset_dc != NULL) hwreset_release(sc->hwreset_dc); if (sc->irq_res != NULL) bus_release_resource(dev, SYS_RES_IRQ, 0, sc->irq_res); if (sc->mem_res != NULL) bus_release_resource(dev, SYS_RES_MEMORY, 0, sc->mem_res); LOCK_DESTROY(sc); return (ENXIO); } static int dc_detach(device_t dev) { struct dc_softc *sc; sc = device_get_softc(dev); TEGRA_DRM_DEREGISTER_CLIENT(device_get_parent(sc->dev), sc->dev); if (sc->irq_ih != NULL) bus_teardown_intr(dev, sc->irq_res, sc->irq_ih); if (sc->clk_parent != NULL) clk_release(sc->clk_parent); if (sc->clk_dc != NULL) clk_release(sc->clk_dc); if (sc->hwreset_dc != NULL) hwreset_release(sc->hwreset_dc); if (sc->irq_res != NULL) bus_release_resource(dev, SYS_RES_IRQ, 0, sc->irq_res); if (sc->mem_res != NULL) bus_release_resource(dev, SYS_RES_MEMORY, 0, sc->mem_res); LOCK_DESTROY(sc); return (bus_generic_detach(dev)); } static device_method_t tegra_dc_methods[] = { /* Device interface */ DEVMETHOD(device_probe, dc_probe), DEVMETHOD(device_attach, dc_attach), DEVMETHOD(device_detach, dc_detach), /* tegra drm interface */ DEVMETHOD(tegra_drm_init_client, dc_init_client), DEVMETHOD(tegra_drm_exit_client, dc_exit_client), /* tegra dc interface */ DEVMETHOD(tegra_dc_display_enable, dc_display_enable), DEVMETHOD(tegra_dc_hdmi_enable, dc_hdmi_enable), DEVMETHOD(tegra_dc_setup_timing, dc_setup_timing), DEVMETHOD_END }; static devclass_t tegra_dc_devclass; DEFINE_CLASS_0(tegra_dc, tegra_dc_driver, tegra_dc_methods, sizeof(struct dc_softc)); DRIVER_MODULE(tegra_dc, host1x, tegra_dc_driver, tegra_dc_devclass, NULL, NULL);