// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (C) Rockchip Electronics Co., Ltd.
 *
 * Author: Huang Lee <Putin.li@rock-chips.com>
 */

#include "rga3_reg_info.h"
#include "rga_dma_buf.h"
#include "rga_iommu.h"
#include "rga_common.h"
#include "rga_debugger.h"
#include "rga_hw_config.h"

#define FACTOR_MAX ((int)(2 << 15))

static void RGA3_set_reg_win0_info(u8 *base, struct rga3_req *msg)
{
	u32 *bRGA3_WIN0_RD_CTRL;
	u32 *bRGA3_WIN0_Y_BASE, *bRGA3_WIN0_U_BASE, *bRGA3_WIN0_V_BASE;
	u32 *bRGA3_WIN0_VIR_STRIDE;
	u32 *bRGA3_WIN0_UV_VIR_STRIDE;
	u32 *bRGA3_WIN0_SRC_SIZE;
	u32 *bRGA3_WIN0_ACT_OFF;
	u32 *bRGA3_WIN0_ACT_SIZE;
	u32 *bRGA3_WIN0_DST_SIZE;

	u32 *bRGA3_WIN0_SCL_FAC;
	/* Not used yet. */
	// u32 *bRGA3_WIN0_FBC_OFF;

	u32 sw = 0, sh = 0;
	u32 dw = 0, dh = 0;
	u32 param_x = 0, param_y = 0;
	u8 x_up = 0, y_up = 0, x_by = 0, y_by = 0;

	u32 reg = 0;

	u8 win_format = 0;
	u8 win_yc_swp = 0;

	/* rb swap on RGB, uv swap on YUV */
	u8 win_pix_swp = 0;

	/*
	 * 1: Semi planar, for yuv 4:2:x
	 * 2: Interleaved (yuyv), for yuv422 8bit only ，RGB
	 */
	u8 win_interleaved = 1;

	/* enable r2y or y2r */
	u8 win_r2y = 0;
	u8 win_y2r = 0;

	u8 rotate_mode = 0;
	u8 xmirror = 0;
	u8 ymirror = 0;

	u8 pixel_width = 1;
	u8 yuv10 = 0;

	u32 stride = 0;
	u32 uv_stride = 0;

	bRGA3_WIN0_RD_CTRL = (u32 *) (base + RGA3_WIN0_RD_CTRL_OFFSET);

	bRGA3_WIN0_Y_BASE = (u32 *) (base + RGA3_WIN0_Y_BASE_OFFSET);
	bRGA3_WIN0_U_BASE = (u32 *) (base + RGA3_WIN0_U_BASE_OFFSET);
	bRGA3_WIN0_V_BASE = (u32 *) (base + RGA3_WIN0_V_BASE_OFFSET);

	bRGA3_WIN0_VIR_STRIDE = (u32 *) (base + RGA3_WIN0_VIR_STRIDE_OFFSET);
	bRGA3_WIN0_UV_VIR_STRIDE =
		(u32 *) (base + RGA3_WIN0_UV_VIR_STRIDE_OFFSET);

	/* Not used yet. */
	// bRGA3_WIN0_FBC_OFF = (u32 *) (base + RGA3_WIN0_FBC_OFF_OFFSET);
	bRGA3_WIN0_ACT_OFF = (u32 *) (base + RGA3_WIN0_ACT_OFF_OFFSET);
	bRGA3_WIN0_SRC_SIZE = (u32 *) (base + RGA3_WIN0_SRC_SIZE_OFFSET);
	bRGA3_WIN0_ACT_SIZE = (u32 *) (base + RGA3_WIN0_ACT_SIZE_OFFSET);
	bRGA3_WIN0_DST_SIZE = (u32 *) (base + RGA3_WIN0_DST_SIZE_OFFSET);

	bRGA3_WIN0_SCL_FAC = (u32 *) (base + RGA3_WIN0_SCL_FAC_OFFSET);

	if (msg->win0.rotate_mode != 0) {
		rotate_mode = msg->rotate_mode & RGA3_ROT_BIT_ROT_90 ? 1 : 0;
		xmirror = msg->rotate_mode & RGA3_ROT_BIT_X_MIRROR ? 1 : 0;
		ymirror = msg->rotate_mode & RGA3_ROT_BIT_Y_MIRROR ? 1 : 0;
	}

	/* scale */
	dw = msg->win0.dst_act_w;
	dh = msg->win0.dst_act_h;

	if (rotate_mode) {
		sh = msg->win0.src_act_w;
		sw = msg->win0.src_act_h;
	} else {
		sw = msg->win0.src_act_w;
		sh = msg->win0.src_act_h;
	}

	if (sw > dw) {
		x_up = 0;
		x_by = 0;
	} else if (sw < dw) {
		x_up = 1;
		x_by = 0;
	} else {
		x_up = 0;
		x_by = 1;
	}

	if (sh > dh) {
		y_up = 0;
		y_by = 0;
	} else if (sh < dh) {
		y_up = 1;
		y_by = 0;
	} else {
		y_up = 0;
		y_by = 1;
	}

	if (x_by == 1 && x_up == 0)
		param_x = 0;
	else if (x_up == 1 && x_by == 0) {
		param_x = FACTOR_MAX * (sw - 1) / (dw - 1);
		/* even multiples of 128 require a scaling factor -1 */
		if ((FACTOR_MAX * (sw - 1)) % (dw - 1) == 0)
			param_x = param_x - 1;
	} else
		param_x = FACTOR_MAX * (dw - 1) / (sw - 1) + 1;

	if (y_by == 1 && y_up == 0)
		param_y = 0;
	else if (y_up == 1 && y_by == 0) {
		param_y = FACTOR_MAX * (sh - 1) / (dh - 1);
		/* even multiples of 128 require a scaling factor -1 */
		if ((FACTOR_MAX * (sh - 1)) % (dh - 1) == 0)
			param_y = param_y - 1;
	} else
		param_y = FACTOR_MAX * (dh - 1) / (sh - 1) + 1;

	switch (msg->win0.format) {
	case RGA_FORMAT_RGBA_8888:
		win_format = 0x8;
		pixel_width = 4;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_BGRA_8888:
		win_format = 0x6;
		pixel_width = 4;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_ARGB_8888:
		win_format = 0x9;
		pixel_width = 4;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_ABGR_8888:
		win_format = 0x7;
		pixel_width = 4;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_RGB_888:
		win_format = 0x5;
		pixel_width = 3;
		win_interleaved = 2;
		win_pix_swp = 1;
		break;
	case RGA_FORMAT_BGR_888:
		win_format = 0x5;
		pixel_width = 3;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_RGB_565:
		win_format = 0x4;
		pixel_width = 2;
		win_interleaved = 2;
		win_pix_swp = 1;
		break;
	case RGA_FORMAT_BGR_565:
		win_format = 0x4;
		pixel_width = 2;
		win_interleaved = 2;
		break;

	case RGA_FORMAT_YVYU_422:
		win_format = 0x1;
		pixel_width = 2;
		win_pix_swp = 1;
		win_yc_swp = 1;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_VYUY_422:
		win_format = 0x1;
		pixel_width = 2;
		win_pix_swp = 1;
		win_yc_swp = 0;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_YUYV_422:
		win_format = 0x1;
		pixel_width = 2;
		win_pix_swp = 0;
		win_yc_swp = 1;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_UYVY_422:
		win_format = 0x1;
		pixel_width = 2;
		win_pix_swp = 0;
		win_yc_swp = 0;
		win_interleaved = 2;
		break;

	case RGA_FORMAT_YCbCr_422_SP:
		win_format = 0x1;
		break;
	case RGA_FORMAT_YCbCr_420_SP:
		win_format = 0x0;
		break;
	case RGA_FORMAT_YCrCb_422_SP:
		win_format = 0x1;
		win_pix_swp = 1;
		break;
	case RGA_FORMAT_YCrCb_420_SP:
		win_format = 0x0;
		win_pix_swp = 1;
		break;

	case RGA_FORMAT_YCbCr_420_SP_10B:
		win_format = 0x2;
		yuv10 = 1;
		break;
	case RGA_FORMAT_YCrCb_420_SP_10B:
		win_format = 0x2;
		yuv10 = 1;
		win_pix_swp = 1;
		break;
	case RGA_FORMAT_YCbCr_422_SP_10B:
		win_format = 0x3;
		yuv10 = 1;
		break;
	case RGA_FORMAT_YCrCb_422_SP_10B:
		win_format = 0x3;
		yuv10 = 1;
		win_pix_swp = 1;
		break;
	};

	if (rga_is_rgb_format(msg->win0.format) &&
	    rga_is_yuv_format(msg->wr.format))
		win_r2y = 1;
	if (rga_is_yuv_format(msg->win0.format) &&
	    rga_is_rgb_format(msg->wr.format))
		win_y2r = 1;

	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_R2Y_EN)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_R2Y_EN(win_r2y)));
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_Y2R_EN)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_Y2R_EN(win_y2r)));

	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_PIC_FORMAT)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_PIC_FORMAT(win_format)));
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_PIX_SWAP)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_PIX_SWAP(win_pix_swp)));
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_YC_SWAP)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_YC_SWAP(win_yc_swp)));
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_RD_FORMAT)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_RD_FORMAT(win_interleaved)));

	if (win_r2y == 1) {
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_CSC_MODE)) |
			(s_RGA3_WIN0_RD_CTRL_SW_WIN0_CSC_MODE(msg->win0.r2y_mode)));
	} else if (win_y2r == 1) {
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_CSC_MODE)) |
			(s_RGA3_WIN0_RD_CTRL_SW_WIN0_CSC_MODE(msg->win0.y2r_mode)));
	}

	/* rotate & mirror */
	if (msg->win0.rotate_mode == 1) {
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_ROT)) |
			 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_ROT(rotate_mode)));
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_XMIRROR)) |
			 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_XMIRROR(xmirror)));
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_YMIRROR)) |
			 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_YMIRROR(ymirror)));
	}

	/* scale */
	*bRGA3_WIN0_SCL_FAC = param_x | param_y << 16;

	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_HOR_BY)) |
			(s_RGA3_WIN0_RD_CTRL_SW_WIN0_HOR_BY(x_by)));
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_HOR_UP)) |
			(s_RGA3_WIN0_RD_CTRL_SW_WIN0_HOR_UP(x_up)));
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_VER_BY)) |
			(s_RGA3_WIN0_RD_CTRL_SW_WIN0_VER_BY(y_by)));
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_VER_UP)) |
			(s_RGA3_WIN0_RD_CTRL_SW_WIN0_VER_UP(y_up)));

	/* rd_mode */
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_RD_MODE)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_RD_MODE(msg->win0.rd_mode)));
	/* win0 enable */
	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_ENABLE)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_ENABLE(msg->win0.enable)));

	reg =
		((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_YUV10B_COMPACT)) |
		 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_YUV10B_COMPACT(1)));

	/* Only on raster mode, yuv 10bit can change to compact or set endian */
	if (msg->win0.rd_mode == 0 && yuv10 == 1) {
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_YUV10B_COMPACT)) |
			 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_YUV10B_COMPACT
			 (msg->win0.is_10b_compact)));
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_ENDIAN_MODE)) |
			 (s_RGA3_WIN0_RD_CTRL_SW_WIN0_ENDIAN_MODE
			 (msg->win0.is_10b_endian)));
	}

	*bRGA3_WIN0_RD_CTRL = reg;

	switch (msg->win0.rd_mode) {
	case 0: /* raster */
		stride = (((msg->win0.vir_w * pixel_width) + 15) & ~15) >> 2;
		if (rga_is_yuv420_semi_planar_format(msg->win0.format))
			uv_stride = ((msg->win0.vir_w + 15) & ~15) >> 2;
		else
			uv_stride = stride;
		break;

	case 1: /* fbc */
		stride = ((msg->win0.vir_w + 15) & ~15) >> 2;
		if (rga_is_yuv420_semi_planar_format(msg->win0.format))
			uv_stride = ((msg->win0.vir_w + 15) & ~15) >> 2;
		else
			uv_stride = stride;
		break;

	case 2: /* tile 8*8 */
		/*
		 * tile 8*8 mode 8 lines of data are read/written at one time,
		 * so stride needs * 8. YUV420 only has 4 lines of UV data, so
		 * it needs to >>1.
		 */
		stride = (((msg->win0.vir_w * pixel_width * 8) + 15) & ~15) >> 2;
		if (rga_is_yuv420_semi_planar_format(msg->win0.format))
			uv_stride = ((((msg->win0.vir_w * 8) + 15) & ~15) >> 1) >> 2;
		else
			uv_stride = stride;
		break;
	}

	*bRGA3_WIN0_Y_BASE = (u32) msg->win0.yrgb_addr;
	*bRGA3_WIN0_U_BASE = (u32) msg->win0.uv_addr;
	*bRGA3_WIN0_V_BASE = (u32) msg->win0.v_addr;

	*bRGA3_WIN0_VIR_STRIDE = stride;
	*bRGA3_WIN0_UV_VIR_STRIDE = uv_stride;

	*bRGA3_WIN0_ACT_OFF = msg->win0.x_offset | (msg->win0.y_offset << 16);
	/* fbcd offset */
	/*
	 *	*bRGA3_WIN0_FBC_OFF = msg->win0.fbc_x_offset |
	 *		 (msg->win0.fbc_y_offset << 16);
	 */

	/* do not use win0 src size except fbcd */
	/* in FBCD, src_width needs to be aligned at 16 */
	*bRGA3_WIN0_SRC_SIZE = ALIGN(msg->win0.src_act_w + msg->win0.x_offset, 16) |
			       (ALIGN(msg->win0.y_offset + msg->win0.src_act_h, 16) << 16);
	*bRGA3_WIN0_ACT_SIZE =
		msg->win0.src_act_w | (msg->win0.src_act_h << 16);
	*bRGA3_WIN0_DST_SIZE =
		msg->win0.dst_act_w | (msg->win0.dst_act_h << 16);
}

static void RGA3_set_reg_win1_info(u8 *base, struct rga3_req *msg)
{
	u32 *bRGA3_WIN1_RD_CTRL;
	u32 *bRGA3_WIN1_Y_BASE, *bRGA3_WIN1_U_BASE, *bRGA3_WIN1_V_BASE;
	u32 *bRGA3_WIN1_VIR_STRIDE;
	u32 *bRGA3_WIN1_UV_VIR_STRIDE;
	u32 *bRGA3_WIN1_SRC_SIZE;
	u32 *bRGA3_WIN1_ACT_OFF;
	u32 *bRGA3_WIN1_ACT_SIZE;
	u32 *bRGA3_WIN1_DST_SIZE;

	u32 *bRGA3_WIN1_SCL_FAC;
	/* Not used yet. */
	// u32 *bRGA3_WIN1_FBC_OFF;

	u32 sw = 0, sh = 0;
	u32 dw = 0, dh = 0;
	u32 param_x = 0, param_y = 0;
	u8 x_up = 0, y_up = 0, x_by = 0, y_by = 0;

	u32 reg = 0;

	u8 win_format = 0;
	u8 win_yc_swp = 0;

	/* rb swap on RGB, uv swap on YUV */
	u8 win_pix_swp = 0;

	/*
	 * 1: Semi planar, for yuv 4:2:x
	 * 2: Interleaved (yuyv), for yuv422 8bit only ，RGB
	 */
	u8 win_interleaved = 1;

	u8 pixel_width = 1;
	u8 yuv10 = 0;

	/* enable r2y or y2r */
	u8 win_r2y = 0;
	u8 win_y2r = 0;

	u8 rotate_mode = 0;
	u8 xmirror = 0;
	u8 ymirror = 0;

	u32 stride = 0;
	u32 uv_stride = 0;

	bRGA3_WIN1_RD_CTRL = (u32 *) (base + RGA3_WIN1_RD_CTRL_OFFSET);

	bRGA3_WIN1_Y_BASE = (u32 *) (base + RGA3_WIN1_Y_BASE_OFFSET);
	bRGA3_WIN1_U_BASE = (u32 *) (base + RGA3_WIN1_U_BASE_OFFSET);
	bRGA3_WIN1_V_BASE = (u32 *) (base + RGA3_WIN1_V_BASE_OFFSET);

	bRGA3_WIN1_VIR_STRIDE = (u32 *) (base + RGA3_WIN1_VIR_STRIDE_OFFSET);
	bRGA3_WIN1_UV_VIR_STRIDE =
		(u32 *) (base + RGA3_WIN1_UV_VIR_STRIDE_OFFSET);

	/* Not used yet. */
	// bRGA3_WIN1_FBC_OFF = (u32 *) (base + RGA3_WIN1_FBC_OFF_OFFSET);
	bRGA3_WIN1_ACT_OFF = (u32 *) (base + RGA3_WIN1_ACT_OFF_OFFSET);
	bRGA3_WIN1_SRC_SIZE = (u32 *) (base + RGA3_WIN1_SRC_SIZE_OFFSET);
	bRGA3_WIN1_ACT_SIZE = (u32 *) (base + RGA3_WIN1_ACT_SIZE_OFFSET);
	bRGA3_WIN1_DST_SIZE = (u32 *) (base + RGA3_WIN1_DST_SIZE_OFFSET);

	bRGA3_WIN1_SCL_FAC = (u32 *) (base + RGA3_WIN1_SCL_FAC_OFFSET);

	if (msg->win1.rotate_mode != 0) {
		rotate_mode = msg->rotate_mode & RGA3_ROT_BIT_ROT_90 ? 1 : 0;
		xmirror = msg->rotate_mode & RGA3_ROT_BIT_X_MIRROR ? 1 : 0;
		ymirror = msg->rotate_mode & RGA3_ROT_BIT_Y_MIRROR ? 1 : 0;
	}

	/* scale */
	dw = msg->win1.dst_act_w;
	dh = msg->win1.dst_act_h;

	if (rotate_mode) {
		sh = msg->win1.src_act_w;
		sw = msg->win1.src_act_h;
	} else {
		sw = msg->win1.src_act_w;
		sh = msg->win1.src_act_h;
	}

	if (sw > dw) {
		x_up = 0;
		x_by = 0;
	} else if (sw < dw) {
		x_up = 1;
		x_by = 0;
	} else {
		x_up = 0;
		x_by = 1;
	}

	if (sh > dh) {
		y_up = 0;
		y_by = 0;
	} else if (sh < dh) {
		y_up = 1;
		y_by = 0;
	} else {
		y_up = 0;
		y_by = 1;
	}

	if (x_by == 1)
		param_x = 0;
	else if (x_up == 1) {
		param_x = (FACTOR_MAX * (sw - 1)) / (dw - 1);
		/* even multiples of 128 require a scaling factor -1 */
		if ((FACTOR_MAX * (sw - 1)) % (dw - 1) == 0)
			param_x = param_x - 1;
	} else
		param_x = (FACTOR_MAX * (dw - 1)) / (sw - 1) + 1;

	if (y_by == 1)
		param_y = 0;
	else if (y_up == 1) {
		param_y = (FACTOR_MAX * (sh - 1)) / (dh - 1);
		/* even multiples of 128 require a scaling factor -1 */
		if ((FACTOR_MAX * (sh - 1)) % (dh - 1) == 0)
			param_y = param_y - 1;
	} else
		param_y = (FACTOR_MAX * (dh - 1)) / (sh - 1) + 1;

	switch (msg->win1.format) {
	case RGA_FORMAT_RGBA_8888:
		win_format = 0x8;
		pixel_width = 4;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_BGRA_8888:
		win_format = 0x6;
		pixel_width = 4;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_ARGB_8888:
		win_format = 0x9;
		pixel_width = 4;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_ABGR_8888:
		win_format = 0x7;
		pixel_width = 4;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_RGB_888:
		win_format = 0x5;
		pixel_width = 3;
		win_interleaved = 2;
		win_pix_swp = 1;
		break;
	case RGA_FORMAT_BGR_888:
		win_format = 0x5;
		pixel_width = 3;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_RGB_565:
		win_format = 0x4;
		pixel_width = 2;
		win_interleaved = 2;
		win_pix_swp = 1;
		break;
	case RGA_FORMAT_BGR_565:
		win_format = 0x4;
		pixel_width = 2;
		win_interleaved = 2;
		break;

	case RGA_FORMAT_YVYU_422:
		win_format = 0x1;
		pixel_width = 2;
		win_pix_swp = 1;
		win_yc_swp = 1;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_VYUY_422:
		win_format = 0x1;
		pixel_width = 2;
		win_pix_swp = 1;
		win_yc_swp = 0;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_YUYV_422:
		win_format = 0x1;
		pixel_width = 2;
		win_pix_swp = 0;
		win_yc_swp = 1;
		win_interleaved = 2;
		break;
	case RGA_FORMAT_UYVY_422:
		win_format = 0x1;
		pixel_width = 2;
		win_pix_swp = 0;
		win_yc_swp = 0;
		win_interleaved = 2;
		break;

	case RGA_FORMAT_YCbCr_422_SP:
		win_format = 0x1;
		break;
	case RGA_FORMAT_YCbCr_420_SP:
		win_format = 0x0;
		break;
	case RGA_FORMAT_YCrCb_422_SP:
		win_format = 0x1;
		win_pix_swp = 1;
		break;
	case RGA_FORMAT_YCrCb_420_SP:
		win_format = 0x0;
		win_pix_swp = 1;
		break;

	case RGA_FORMAT_YCbCr_420_SP_10B:
		win_format = 0x2;
		yuv10 = 1;
		break;
	case RGA_FORMAT_YCrCb_420_SP_10B:
		win_format = 0x2;
		win_pix_swp = 1;
		yuv10 = 1;
		break;
	case RGA_FORMAT_YCbCr_422_SP_10B:
		win_format = 0x3;
		yuv10 = 1;
		break;
	case RGA_FORMAT_YCrCb_422_SP_10B:
		win_format = 0x3;
		win_pix_swp = 1;
		yuv10 = 1;
		break;
	};

	if (rga_is_rgb_format(msg->win1.format) &&
	    rga_is_yuv_format(msg->wr.format))
		win_r2y = 1;
	if (rga_is_yuv_format(msg->win1.format) &&
	    rga_is_rgb_format(msg->wr.format))
		win_y2r = 1;

	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_R2Y_EN)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_R2Y_EN(win_r2y)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_Y2R_EN)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_Y2R_EN(win_y2r)));

	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_PIC_FORMAT)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_PIC_FORMAT(win_format)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_PIX_SWAP)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_PIX_SWAP(win_pix_swp)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_YC_SWAP)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_YC_SWAP(win_yc_swp)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_RD_FORMAT)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_RD_FORMAT(win_interleaved)));

	if (win_r2y == 1) {
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_CSC_MODE)) |
			(s_RGA3_WIN0_RD_CTRL_SW_WIN0_CSC_MODE(msg->win1.r2y_mode)));
	} else if (win_y2r == 1) {
		reg =
			((reg & (~m_RGA3_WIN0_RD_CTRL_SW_WIN0_CSC_MODE)) |
			(s_RGA3_WIN0_RD_CTRL_SW_WIN0_CSC_MODE(msg->win1.y2r_mode)));
	}

	/* rotate & mirror */
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_ROT)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_ROT(rotate_mode)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_XMIRROR)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_XMIRROR(xmirror)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_YMIRROR)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_YMIRROR(ymirror)));
	//warning: TRM not complete
	/* scale */
	*bRGA3_WIN1_SCL_FAC = param_x | param_y << 16;

	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_HOR_BY)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_HOR_BY(x_by)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_HOR_UP)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_HOR_UP(x_up)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_VER_BY)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_VER_BY(y_by)));
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_VER_UP)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_VER_UP(y_up)));

	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_YUV10B_COMPACT)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_YUV10B_COMPACT(1)));

	/* Only on roster mode, yuv 10bit can change to compact or set endian */
	if (msg->win1.rd_mode == 0 && yuv10 == 1) {
		reg =
			((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_YUV10B_COMPACT)) |
			 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_YUV10B_COMPACT
			 (msg->win1.is_10b_compact)));
		reg =
			((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_ENDIAN_MODE)) |
			 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_ENDIAN_MODE
			 (msg->win1.is_10b_endian)));
	}

	/* rd_mode */
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_RD_MODE)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_RD_MODE(msg->win1.rd_mode)));
	/* win1 enable */
	reg =
		((reg & (~m_RGA3_WIN1_RD_CTRL_SW_WIN1_ENABLE)) |
		 (s_RGA3_WIN1_RD_CTRL_SW_WIN1_ENABLE(msg->win1.enable)));

	*bRGA3_WIN1_RD_CTRL = reg;

	switch (msg->win1.rd_mode) {
	case 0: /* raster */
		stride = (((msg->win1.vir_w * pixel_width) + 15) & ~15) >> 2;
		if (rga_is_yuv420_semi_planar_format(msg->win1.format))
			uv_stride = ((msg->win1.vir_w + 15) & ~15) >> 2;
		else
			uv_stride = stride;
		break;

	case 1: /* fbc */
		stride = ((msg->win1.vir_w + 15) & ~15) >> 2;
		if (rga_is_yuv420_semi_planar_format(msg->win1.format))
			uv_stride = ((msg->win1.vir_w + 15) & ~15) >> 2;
		else
			uv_stride = stride;
		break;

	case 2: /* tile 8*8 */
		stride = (((msg->win1.vir_w * pixel_width * 8) + 15) & ~15) >> 2;
		if (rga_is_yuv420_semi_planar_format(msg->win1.format))
			uv_stride = ((((msg->win1.vir_w * 8) + 15) & ~15) >> 1) >> 2;
		else
			uv_stride = stride;
		break;
	}

	*bRGA3_WIN1_Y_BASE = (u32) msg->win1.yrgb_addr;
	*bRGA3_WIN1_U_BASE = (u32) msg->win1.uv_addr;
	*bRGA3_WIN1_V_BASE = (u32) msg->win1.v_addr;

	*bRGA3_WIN1_VIR_STRIDE = stride;
	*bRGA3_WIN1_UV_VIR_STRIDE = uv_stride;

	*bRGA3_WIN1_ACT_OFF = msg->win1.x_offset | (msg->win1.y_offset << 16);
	/* fbcd offset */
	/*
	 *		 *bRGA3_WIN1_FBC_OFF = msg->win1.fbc_x_offset |
	 *			(msg->win1.fbc_y_offset << 16);
	 */

	/* do not use win1 src size except fbcd */
	*bRGA3_WIN1_SRC_SIZE = (msg->win1.src_act_w +
		msg->win1.x_offset) | ((msg->win1.src_act_h +
		msg->win1.y_offset) << 16);
	*bRGA3_WIN1_ACT_SIZE =
		msg->win1.src_act_w | (msg->win1.src_act_h << 16);
	*bRGA3_WIN1_DST_SIZE =
		msg->win1.dst_act_w | (msg->win1.dst_act_h << 16);
}

static void RGA3_set_reg_wr_info(u8 *base, struct rga3_req *msg)
{
	u32 *bRGA3_WR_RD_CTRL;
	u32 *bRGA3_WR_Y_BASE, *bRGA3_WR_U_BASE, *bRGA3_WR_V_BASE;
	u32 *bRGA3_WR_VIR_STRIDE;
	u32 *bRGA3_WR_PL_VIR_STRIDE;
	u32 *bRGA3_WR_FBCD_CTRL;

	u32 reg = 0;
	u32 fbcd_reg = 0;

	u8 wr_format = 0;
	u8 wr_yc_swp = 0;

	/* rb swap on RGB, uv swap on YUV */
	u8 wr_pix_swp = 0;

	u8 pixel_width = 1;
	u8 yuv10 = 0;

	/*
	 * 1: Semi planar, for yuv 4:2:x
	 * 2: Interleaved (yuyv), for yuv422 8bit only ，RGB
	 */
	u8 wr_interleaved = 1;

	u32 stride = 0;
	u32 uv_stride = 0;

	u32 vir_h = 0;

	bRGA3_WR_RD_CTRL = (u32 *) (base + RGA3_WR_CTRL_OFFSET);
	bRGA3_WR_FBCD_CTRL = (u32 *) (base + RGA3_WR_FBCE_CTRL_OFFSET);

	bRGA3_WR_Y_BASE = (u32 *) (base + RGA3_WR_Y_BASE_OFFSET);
	bRGA3_WR_U_BASE = (u32 *) (base + RGA3_WR_U_BASE_OFFSET);
	bRGA3_WR_V_BASE = (u32 *) (base + RGA3_WR_V_BASE_OFFSET);

	bRGA3_WR_VIR_STRIDE = (u32 *) (base + RGA3_WR_VIR_STRIDE_OFFSET);
	bRGA3_WR_PL_VIR_STRIDE =
		(u32 *) (base + RGA3_WR_PL_VIR_STRIDE_OFFSET);

	switch (msg->wr.format) {
	case RGA_FORMAT_RGBA_8888:
		wr_format = 0x6;
		pixel_width = 4;
		wr_interleaved = 2;
		wr_pix_swp = 1;
		break;
	case RGA_FORMAT_BGRA_8888:
		wr_format = 0x6;
		pixel_width = 4;
		wr_interleaved = 2;
		break;
	case RGA_FORMAT_RGB_888:
		wr_format = 0x5;
		pixel_width = 3;
		wr_interleaved = 2;
		wr_pix_swp = 1;
		break;
	case RGA_FORMAT_BGR_888:
		wr_format = 0x5;
		pixel_width = 3;
		wr_interleaved = 2;
		break;
	case RGA_FORMAT_RGB_565:
		wr_format = 0x4;
		pixel_width = 2;
		wr_interleaved = 2;
		wr_pix_swp = 1;
		break;
	case RGA_FORMAT_BGR_565:
		wr_format = 0x4;
		pixel_width = 2;
		wr_interleaved = 2;
		break;

	case RGA_FORMAT_YVYU_422:
		wr_format = 0x1;
		pixel_width = 2;
		wr_pix_swp = 1;
		wr_yc_swp = 1;
		wr_interleaved = 2;
		break;
	case RGA_FORMAT_VYUY_422:
		wr_format = 0x1;
		pixel_width = 2;
		wr_pix_swp = 1;
		wr_yc_swp = 0;
		wr_interleaved = 2;
		break;
	case RGA_FORMAT_YUYV_422:
		wr_format = 0x1;
		pixel_width = 2;
		wr_pix_swp = 0;
		wr_yc_swp = 1;
		wr_interleaved = 2;
		break;
	case RGA_FORMAT_UYVY_422:
		wr_format = 0x1;
		pixel_width = 2;
		wr_pix_swp = 0;
		wr_yc_swp = 0;
		wr_interleaved = 2;
		break;

	case RGA_FORMAT_YCbCr_422_SP:
		wr_format = 0x1;
		break;
	case RGA_FORMAT_YCbCr_420_SP:
		wr_format = 0x0;
		break;
	case RGA_FORMAT_YCrCb_422_SP:
		wr_format = 0x1;
		wr_pix_swp = 1;
		break;
	case RGA_FORMAT_YCrCb_420_SP:
		wr_format = 0x0;
		wr_pix_swp = 1;
		break;

	case RGA_FORMAT_YCbCr_420_SP_10B:
		wr_format = 0x2;
		yuv10 = 1;
		break;
	case RGA_FORMAT_YCrCb_420_SP_10B:
		wr_format = 0x2;
		wr_pix_swp = 1;
		yuv10 = 1;
		break;
	case RGA_FORMAT_YCbCr_422_SP_10B:
		wr_format = 0x3;
		yuv10 = 1;
		break;
	case RGA_FORMAT_YCrCb_422_SP_10B:
		wr_format = 0x3;
		wr_pix_swp = 1;
		yuv10 = 1;
		break;
	};

	reg =
		((reg & (~m_RGA3_WR_CTRL_SW_WR_PIC_FORMAT)) |
		 (s_RGA3_WR_CTRL_SW_WR_PIC_FORMAT(wr_format)));
	reg =
		((reg & (~m_RGA3_WR_CTRL_SW_WR_PIX_SWAP)) |
		 (s_RGA3_WR_CTRL_SW_WR_PIX_SWAP(wr_pix_swp)));
	reg =
		((reg & (~m_RGA3_WR_CTRL_SW_WR_YC_SWAP)) |
		 (s_RGA3_WR_CTRL_SW_WR_YC_SWAP(wr_yc_swp)));
	reg =
		((reg & (~m_RGA3_WR_CTRL_SW_WR_FORMAT)) |
		 (s_RGA3_WR_CTRL_SW_WR_FORMAT(wr_interleaved)));
	reg =
		((reg & (~m_RGA3_WR_CTRL_SW_WR_FBCE_SPARSE_EN)) |
		 (s_RGA3_WR_CTRL_SW_WR_FBCE_SPARSE_EN(1)));

	reg =
		((reg & (~m_RGA3_WR_CTRL_SW_OUTSTANDING_MAX)) |
		 (s_RGA3_WR_CTRL_SW_OUTSTANDING_MAX(0xf)));

	reg =
		((reg & (~m_RGA3_WR_CTRL_SW_WR_YUV10B_COMPACT)) |
		 (s_RGA3_WR_CTRL_SW_WR_YUV10B_COMPACT(1)));

	/* Only on roster mode, yuv 10bit can change to compact or set endian */
	if (msg->wr.rd_mode == 0 && yuv10 == 1) {
		reg =
			((reg & (~m_RGA3_WR_CTRL_SW_WR_YUV10B_COMPACT)) |
			 (s_RGA3_WR_CTRL_SW_WR_YUV10B_COMPACT
			 (msg->wr.is_10b_compact)));
		reg =
			((reg & (~m_RGA3_WR_CTRL_SW_WR_ENDIAN_MODE)) |
			 (s_RGA3_WR_CTRL_SW_WR_ENDIAN_MODE
			 (msg->wr.is_10b_endian)));
	}

	/* rd_mode */
	reg =
		((reg & (~m_RGA3_WR_CTRL_SW_WR_MODE)) |
		 (s_RGA3_WR_CTRL_SW_WR_MODE(msg->wr.rd_mode)));

	fbcd_reg = ((fbcd_reg & (~m_RGA3_WR_FBCE_CTRL_SW_WR_FBCE_HOFF_DISS)) |
		 (s_RGA3_WR_FBCE_CTRL_SW_WR_FBCE_HOFF_DISS(0)));

	*bRGA3_WR_RD_CTRL = reg;
	*bRGA3_WR_FBCD_CTRL = fbcd_reg;

	switch (msg->wr.rd_mode) {
	case 0: /* raster */
		stride = (((msg->wr.vir_w * pixel_width) + 15) & ~15) >> 2;
		uv_stride = ((msg->wr.vir_w + 15) & ~15) >> 2;

		*bRGA3_WR_U_BASE = (u32) msg->wr.uv_addr;

		break;

	case 1: /* fbc */
		stride = ((msg->wr.vir_w + 15) & ~15) >> 2;
		/* need to calculate fbcd header size */
		vir_h = ((msg->wr.vir_h + 15) & ~15);

		/* RGBA8888 */
		if (wr_format == 0x6)
			uv_stride = ((msg->wr.vir_w + 15) & ~15);
		/* RGB888 */
		else if (wr_format == 0x5)
			uv_stride = (((msg->wr.vir_w + 15) & ~15) >> 2) * 3;
		/* RGB565, yuv422 8bit, yuv420 10bit */
		else if (wr_format == 0x4 || wr_format == 0x1 || wr_format == 0x2)
			uv_stride = ((msg->wr.vir_w + 15) & ~15) >> 1;
		/* yuv420 8bit */
		else if (wr_format == 0x0)
			uv_stride = (((msg->wr.vir_w + 15) & ~15) >> 3) * 3;
		/* yuv422 10bit */
		else if (wr_format == 0x3)
			uv_stride = (((msg->wr.vir_w + 15) & ~15) >> 3) * 5;

		*bRGA3_WR_U_BASE = (u32) (msg->wr.uv_addr + ((stride * vir_h)>>2));

		break;

	case 2: /* tile 8*8 */
		stride = (((msg->wr.vir_w * pixel_width * 8) + 15) & ~15) >> 2;
		if (rga_is_yuv420_semi_planar_format(msg->win0.format))
			uv_stride = ((((msg->wr.vir_w * 8) + 15) & ~15) >> 1) >> 2;
		else
			uv_stride = stride;

		*bRGA3_WR_U_BASE = (u32) msg->wr.uv_addr;
		break;
	}

	*bRGA3_WR_Y_BASE = (u32) msg->wr.yrgb_addr;
	*bRGA3_WR_V_BASE = (u32) msg->wr.v_addr;

	*bRGA3_WR_VIR_STRIDE = stride;
	*bRGA3_WR_PL_VIR_STRIDE = uv_stride;
}

static void RGA3_set_reg_overlap_info(u8 *base, struct rga3_req *msg)
{
	u32 *bRGA_OVERLAP_TOP_CTRL;
	u32 *bRGA_OVERLAP_BOT_CTRL;
	u32 *bRGA_OVERLAP_TOP_ALPHA;
	u32 *bRGA_OVERLAP_BOT_ALPHA;
	u32 *bRGA_OVERLAP_TOP_KEY_MIN;
	u32 *bRGA_OVERLAP_TOP_KEY_MAX;

	u32 *bRGA_OVERLAP_CTRL;
	u32 *bRGA3_OVLP_OFF;

	u32 reg;
	union rga3_color_ctrl top_color_ctrl, bottom_color_ctrl;
	union rga3_alpha_ctrl top_alpha_ctrl, bottom_alpha_ctrl;
	struct rga_alpha_config *config;

	bRGA_OVERLAP_TOP_CTRL = (u32 *) (base + RGA3_OVLP_TOP_CTRL_OFFSET);
	bRGA_OVERLAP_BOT_CTRL = (u32 *) (base + RGA3_OVLP_BOT_CTRL_OFFSET);
	bRGA_OVERLAP_TOP_ALPHA = (u32 *) (base + RGA3_OVLP_TOP_ALPHA_OFFSET);
	bRGA_OVERLAP_BOT_ALPHA = (u32 *) (base + RGA3_OVLP_BOT_ALPHA_OFFSET);

	bRGA_OVERLAP_CTRL = (u32 *) (base + RGA3_OVLP_CTRL_OFFSET);
	bRGA3_OVLP_OFF = (u32 *) (base + RGA3_OVLP_OFF_OFFSET);

	/* Alpha blend */
	/*bot -> win0(dst), top -> win1(src). */
	top_color_ctrl.value = 0;
	bottom_color_ctrl.value = 0;
	top_alpha_ctrl.value = 0;
	bottom_alpha_ctrl.value = 0;
	config = &msg->alpha_config;

	if (config->fg_pixel_alpha_en)
		top_color_ctrl.bits.blend_mode =
			config->fg_global_alpha_en ? RGA_ALPHA_PER_PIXEL_GLOBAL :
			RGA_ALPHA_PER_PIXEL;
	else
		top_color_ctrl.bits.blend_mode = RGA_ALPHA_GLOBAL;

	if (config->bg_pixel_alpha_en)
		bottom_color_ctrl.bits.blend_mode =
			config->bg_global_alpha_en ? RGA_ALPHA_PER_PIXEL_GLOBAL :
			RGA_ALPHA_PER_PIXEL;
	else
		bottom_color_ctrl.bits.blend_mode = RGA_ALPHA_GLOBAL;

	/*
	 * Since the hardware uses 256 as 1, the original alpha value needs to
	 * be + (alpha >> 7).
	 */
	top_color_ctrl.bits.alpha_cal_mode = RGA_ALPHA_SATURATION;
	bottom_color_ctrl.bits.alpha_cal_mode = RGA_ALPHA_SATURATION;

	top_color_ctrl.bits.global_alpha = config->fg_global_alpha_value;
	bottom_color_ctrl.bits.global_alpha = config->bg_global_alpha_value;

	/* porter duff alpha enable */
	switch (config->mode) {
	case RGA_ALPHA_BLEND_SRC:
		/*
		 * SRC mode:
		 *	Sf = 1, Df = 0；
		 *	[Rc,Ra] = [Sc,Sa]；
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_ONE;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_ZERO;

		break;

	case RGA_ALPHA_BLEND_DST:
		/*
		 * SRC mode:
		 *	Sf = 0, Df = 1；
		 *	[Rc,Ra] = [Dc,Da]；
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_ZERO;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_ONE;

		break;

	case RGA_ALPHA_BLEND_SRC_OVER:
		/*
		 * SRC-OVER mode:
		 *	Sf = 1, Df = (1 - Sa)
		 *	[Rc,Ra] = [ Sc + (1 - Sa) * Dc, Sa + (1 - Sa) * Da ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_ONE;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE_INVERSE;

		break;

	case RGA_ALPHA_BLEND_DST_OVER:
		/*
		 * DST-OVER mode:
		 *	Sf = (1 - Da) , Df = 1
		 *	[Rc,Ra] = [ Sc * (1 - Da) + Dc, Sa * (1 - Da) + Da ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE_INVERSE;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_ONE;

		break;

	case RGA_ALPHA_BLEND_SRC_IN:
		/*
		 * SRC-IN mode:
		 *	Sf = Da , Df = 0
		 *	[Rc,Ra] = [ Sc * Da, Sa * Da ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_ZERO;

		break;

	case RGA_ALPHA_BLEND_DST_IN:
		/*
		 * DST-IN mode:
		 *	Sf = 0 , Df = Sa
		 *	[Rc,Ra] = [ Dc * Sa, Da * Sa ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_ZERO;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE;

		break;

	case RGA_ALPHA_BLEND_SRC_OUT:
		/*
		 * SRC-OUT mode:
		 *	Sf = (1 - Da) , Df = 0
		 *	[Rc,Ra] = [ Sc * (1 - Da), Sa * (1 - Da) ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE_INVERSE;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_ZERO;

		break;

	case RGA_ALPHA_BLEND_DST_OUT:
		/*
		 * DST-OUT mode:
		 *	Sf = 0 , Df = (1 - Sa)
		 *	[Rc,Ra] = [ Dc * (1 - Sa), Da * (1 - Sa) ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_ZERO;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE_INVERSE;

		break;

	case RGA_ALPHA_BLEND_SRC_ATOP:
		/*
		 * SRC-ATOP mode:
		 *	Sf = Da , Df = (1 - Sa)
		 *	[Rc,Ra] = [ Sc * Da + Dc * (1 - Sa), Sa * Da + Da * (1 - Sa) ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE_INVERSE;

		break;

	case RGA_ALPHA_BLEND_DST_ATOP:
		/*
		 * DST-ATOP mode:
		 *	Sf = (1 - Da) , Df = Sa
		 *	[Rc,Ra] = [ Sc * (1 - Da) + Dc * Sa, Sa * (1 - Da) + Da * Sa ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE_INVERSE;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE;

		break;

	case RGA_ALPHA_BLEND_XOR:
		/*
		 * DST-XOR mode:
		 *	Sf = (1 - Da) , Df = (1 - Sa)
		 *	[Rc,Ra] = [ Sc * (1 - Da) + Dc * (1 - Sa), Sa * (1 - Da) + Da * (1 - Sa) ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE_INVERSE;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_OPPOSITE_INVERSE;

		break;

	case RGA_ALPHA_BLEND_CLEAR:
		/*
		 * DST-CLEAR mode:
		 *	Sf = 0 , Df = 0
		 *	[Rc,Ra] = [ 0, 0 ]
		 */
		top_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		top_color_ctrl.bits.factor_mode = RGA_ALPHA_ZERO;

		bottom_color_ctrl.bits.alpha_mode = RGA_ALPHA_STRAIGHT;
		bottom_color_ctrl.bits.factor_mode = RGA_ALPHA_ZERO;

		break;

	default:
		break;
	}

	if (!config->enable && msg->abb_alpha_pass) {
		/*
		 * enabled by default bot_blend_m1 && bot_alpha_cal_m1 for src channel(win0)
		 * In ABB mode, the number will be fetched according to 16*16, so it needs to
		 * be enabled top_blend_m1 && top_alpha_cal_m1 for dst channel(wr).
		 */
		top_color_ctrl.bits.color_mode = RGA_ALPHA_PRE_MULTIPLIED;

		top_alpha_ctrl.bits.blend_mode = RGA_ALPHA_PER_PIXEL;
		top_alpha_ctrl.bits.alpha_cal_mode = RGA_ALPHA_NO_SATURATION;

		bottom_color_ctrl.bits.color_mode = RGA_ALPHA_PRE_MULTIPLIED;

		bottom_alpha_ctrl.bits.blend_mode = RGA_ALPHA_PER_PIXEL;
		bottom_alpha_ctrl.bits.alpha_cal_mode = RGA_ALPHA_NO_SATURATION;
	} else {
		top_color_ctrl.bits.color_mode =
			config->fg_pre_multiplied ?
				RGA_ALPHA_PRE_MULTIPLIED : RGA_ALPHA_NO_PRE_MULTIPLIED;

		top_alpha_ctrl.bits.blend_mode = top_color_ctrl.bits.blend_mode;
		top_alpha_ctrl.bits.alpha_cal_mode = top_color_ctrl.bits.alpha_cal_mode;
		top_alpha_ctrl.bits.alpha_mode = top_color_ctrl.bits.alpha_mode;
		top_alpha_ctrl.bits.factor_mode = top_color_ctrl.bits.factor_mode;

		bottom_color_ctrl.bits.color_mode =
			config->bg_pre_multiplied ?
				RGA_ALPHA_PRE_MULTIPLIED : RGA_ALPHA_NO_PRE_MULTIPLIED;

		bottom_alpha_ctrl.bits.blend_mode = bottom_color_ctrl.bits.blend_mode;
		bottom_alpha_ctrl.bits.alpha_cal_mode = bottom_color_ctrl.bits.alpha_cal_mode;
		bottom_alpha_ctrl.bits.alpha_mode = bottom_color_ctrl.bits.alpha_mode;
		bottom_alpha_ctrl.bits.factor_mode = bottom_color_ctrl.bits.factor_mode;
	}

	*bRGA_OVERLAP_TOP_CTRL = top_color_ctrl.value;
	*bRGA_OVERLAP_BOT_CTRL = bottom_color_ctrl.value;
	*bRGA_OVERLAP_TOP_ALPHA = top_alpha_ctrl.value;
	*bRGA_OVERLAP_BOT_ALPHA = bottom_alpha_ctrl.value;

	/* set RGA_OVERLAP_CTRL */
	reg = 0;
	/* color key */
	bRGA_OVERLAP_TOP_KEY_MIN =
		(u32 *) (base + RGA3_OVLP_TOP_KEY_MIN_OFFSET);
	bRGA_OVERLAP_TOP_KEY_MAX =
		(u32 *) (base + RGA3_OVLP_TOP_KEY_MAX_OFFSET);

	/*
	 * YG : value		 (0:9)
	 * UB : value >> 10	 (10:19)
	 * VG : value >> 20	 (20:29)
	 */
	if (msg->color_key_min > 0 || msg->color_key_max > 0) {
		*bRGA_OVERLAP_TOP_KEY_MIN = msg->color_key_min;
		*bRGA_OVERLAP_TOP_KEY_MAX = msg->color_key_max;
		reg = ((reg & (~m_RGA3_OVLP_CTRL_SW_TOP_KEY_EN)) |
			 (s_RGA3_OVLP_CTRL_SW_TOP_KEY_EN(1)));
	}

	/* 1: ABB mode, 0: ABC mode， ABB cannot support fbc in&out */
	if (msg->win0.yrgb_addr == msg->wr.yrgb_addr)
		reg = ((reg & (~m_RGA3_OVLP_CTRL_SW_OVLP_MODE)) |
			(s_RGA3_OVLP_CTRL_SW_OVLP_MODE(1)));

	/* 1: yuv field, 0: rgb field */
	if (rga_is_yuv_format(msg->wr.format))
		reg = ((reg & (~m_RGA3_OVLP_CTRL_SW_OVLP_FIELD)) |
			 (s_RGA3_OVLP_CTRL_SW_OVLP_FIELD(1)));

	/*
	 * warning: if m1 & m0 need config split，need to redesign
	 * this judge, which consider RGBA8888 format
	 */
	reg = ((reg & (~m_RGA3_OVLP_CTRL_SW_TOP_ALPHA_EN)) |
	       (s_RGA3_OVLP_CTRL_SW_TOP_ALPHA_EN(config->enable)));

	*bRGA_OVERLAP_CTRL = reg;

	*bRGA3_OVLP_OFF = msg->wr.x_offset | (msg->wr.y_offset << 16);
}

static int rga3_gen_reg_info(u8 *base, struct rga3_req *msg)
{
	switch (msg->render_mode) {
	case BITBLT_MODE:
		RGA3_set_reg_win0_info(base, msg);
		RGA3_set_reg_win1_info(base, msg);
		RGA3_set_reg_overlap_info(base, msg);
		RGA3_set_reg_wr_info(base, msg);
		break;
	default:
		rga_err("error msg render mode %d\n", msg->render_mode);
		break;
	}

	return 0;
}

static void addr_copy(struct rga_win_info_t *win, struct rga_img_info_t *img)
{
	win->yrgb_addr = img->yrgb_addr;
	win->uv_addr = img->uv_addr;
	win->v_addr = img->v_addr;
	win->enable = 1;
}

static void set_win_info(struct rga_win_info_t *win, struct rga_img_info_t *img)
{
	win->x_offset = img->x_offset;
	win->y_offset = img->y_offset;
	win->src_act_w = img->act_w;
	win->src_act_h = img->act_h;
	win->vir_w = img->vir_w;
	win->vir_h = img->vir_h;
	if (img->rd_mode == RGA_RASTER_MODE)
		win->rd_mode = 0;
	else if (img->rd_mode == RGA_FBC_MODE)
		win->rd_mode = 1;
	else if (img->rd_mode == RGA_TILE_MODE)
		win->rd_mode = 2;

	switch (img->compact_mode) {
	case RGA_10BIT_INCOMPACT:
		win->is_10b_compact = 0;
		break;
	case RGA_10BIT_COMPACT:
	default:
		win->is_10b_compact = 1;
		break;
	}

	win->is_10b_endian = img->is_10b_endian;
}

static void set_wr_info(struct rga_req *req_rga, struct rga3_req *req)
{
	/* The output w/h are bound to the dst_act_w/h of win0. */
	req->wr.dst_act_w = req->win0.dst_act_w;
	req->wr.dst_act_h = req->win0.dst_act_h;

	/* Some configurations need to be all equal to the output w/h. */
	req->wr.vir_w = req_rga->dst.vir_w;
	req->wr.vir_h = req_rga->dst.vir_h;

	if (req_rga->dst.rd_mode == RGA_RASTER_MODE)
		req->wr.rd_mode = 0;
	else if (req_rga->dst.rd_mode == RGA_FBC_MODE)
		req->wr.rd_mode = 1;
	else if (req_rga->dst.rd_mode == RGA_TILE_MODE)
		req->wr.rd_mode = 2;

	switch (req_rga->dst.compact_mode) {
	case RGA_10BIT_INCOMPACT:
		req->wr.is_10b_compact = 0;
		break;
	case RGA_10BIT_COMPACT:
	default:
		req->wr.is_10b_compact = 1;
		break;
	}

	req->wr.is_10b_endian = req_rga->dst.is_10b_endian;
}

/* TODO: common part */
static void rga_cmd_to_rga3_cmd(struct rga_req *req_rga, struct rga3_req *req)
{
	struct rga_img_info_t tmp;

	req->render_mode = BITBLT_MODE;

	/* rotate & mirror */
	switch (req_rga->rotate_mode & 0x0f) {
	case 0x1:
		if (req_rga->sina == 65536 && req_rga->cosa == 0) {
			/* rot-90 */
			req->rotate_mode = RGA3_ROT_BIT_ROT_90;
		} else if (req_rga->sina == 0 && req_rga->cosa == -65536) {
			/* rot-180 = X-mirror + Y-mirror */
			req->rotate_mode = RGA3_ROT_BIT_X_MIRROR | RGA3_ROT_BIT_Y_MIRROR;
		} else if (req_rga->sina == -65536 && req_rga->cosa == 0) {
			/* rot-270 or -90 = rot-90 + X-mirror + Y-mirror */
			req->rotate_mode = RGA3_ROT_BIT_X_MIRROR | RGA3_ROT_BIT_Y_MIRROR |
					   RGA3_ROT_BIT_ROT_90;
		} else if (req_rga->sina == 0 && req_rga->cosa == 65536) {
			/* bypass */
			req->rotate_mode = 0;
		}
		break;
	case 0x2:
		/* X-mirror */
		req->rotate_mode = RGA3_ROT_BIT_X_MIRROR;
		break;
	case 0x3:
		/* Y-mirror */
		req->rotate_mode = RGA3_ROT_BIT_Y_MIRROR;
		break;
	case 0x4:
		/* X-mirror + Y-mirror */
		req->rotate_mode = RGA3_ROT_BIT_X_MIRROR | RGA3_ROT_BIT_Y_MIRROR;
		break;
	default:
		req->rotate_mode = 0;
		break;
	}

	/* The upper four bits are only allowed to configure the mirror. */
	switch ((req_rga->rotate_mode & 0xf0) >> 4) {
	case 2:
		/* X-mirror */
		req->rotate_mode ^= RGA3_ROT_BIT_X_MIRROR;
		break;
	case 3:
		/* Y-mirror */
		req->rotate_mode ^= RGA3_ROT_BIT_Y_MIRROR;
		break;
	case 0x4:
		/* X-mirror + Y-mirror */
		req->rotate_mode ^= RGA3_ROT_BIT_X_MIRROR | RGA3_ROT_BIT_Y_MIRROR;
		break;
	}

	req->win0_a_global_val = req_rga->alpha_global_value;
	req->win1_a_global_val = req_rga->alpha_global_value;

	/* fixup yuv/rgb convert to rgba missing alpha channel */
	if (!(req_rga->alpha_rop_flag & 1)) {
		if (!rga_is_alpha_format(req_rga->src.format) &&
		    rga_is_alpha_format(req_rga->dst.format)) {
			req->alpha_config.fg_global_alpha_value = 0xff;
			req->alpha_config.bg_global_alpha_value = 0xff;
		}
	}

	/* simple win can not support dst offset */
	if ((!((req_rga->alpha_rop_flag) & 1)) &&
	    (req_rga->dst.x_offset == 0 && req_rga->dst.y_offset == 0) &&
	    (req_rga->src.yrgb_addr != req_rga->dst.yrgb_addr)) {
		/*
		 * ABB mode Layer binding:
		 *     src => win0
		 *     dst => wr
		 */

		/*
		 * enabled by default bot_blend_m1 && bot_alpha_cal_m1 for src channel(win0)
		 * In ABB mode, the number will be fetched according to 16*16, so it needs to
		 * be enabled top_blend_m1 && top_alpha_cal_m1 for dst channel(wr).
		 */
		if (rga_is_alpha_format(req_rga->src.format))
			req->abb_alpha_pass = true;

		set_win_info(&req->win0, &req_rga->src);

		/* enable win0 rotate */
		req->win0.rotate_mode = 1;

		/* set win dst size */
		req->win0.dst_act_w = req_rga->dst.act_w;
		req->win0.dst_act_h = req_rga->dst.act_h;

		addr_copy(&req->win0, &req_rga->src);
		addr_copy(&req->wr, &req_rga->dst);

		req->win0.format = req_rga->src.format;
		req->wr.format = req_rga->dst.format;
	} else {
		/*
		 * ABC mode Layer binding:
		 *     src => win1
		 *     src1/dst => win0
		 *     dst => wr
		 */

		/*
		 * enabled by default top_blend_m1 && top_alpha_cal_m1 for src channel(win1)
		 * In ABB mode, the number will be fetched according to 16*16, so it needs to
		 * be enabled bot_blend_m1 && bot_alpha_cal_m1 for src1/dst channel(win0).
		 */
		if (rga_is_alpha_format(req_rga->src.format))
			req->abb_alpha_pass = true;

		if (req_rga->pat.yrgb_addr != 0) {
			if (req_rga->src.yrgb_addr == req_rga->dst.yrgb_addr) {
				/* Convert ABC mode to ABB mode. */
				memcpy(&req_rga->src, &req_rga->pat, sizeof(req_rga->src));
				memset(&req_rga->pat, 0x0, sizeof(req_rga->pat));
				req_rga->bsfilter_flag = 0;

				rga_swap_pd_mode(req_rga);
			} else if ((req_rga->dst.x_offset + req_rga->src.act_w >
				    req_rga->pat.act_w) ||
				   (req_rga->dst.y_offset + req_rga->src.act_h >
				    req_rga->pat.act_h)) {
				/* wr_offset + win1.act_size need > win0.act_size */
				memcpy(&tmp, &req_rga->src, sizeof(tmp));
				memcpy(&req_rga->src, &req_rga->pat, sizeof(req_rga->src));
				memcpy(&req_rga->pat, &tmp, sizeof(req_rga->pat));

				rga_swap_pd_mode(req_rga);
			}
		}

		set_win_info(&req->win1, &req_rga->src);

		/* enable win1 rotate */
		req->win1.rotate_mode = 1;

		addr_copy(&req->win1, &req_rga->src);
		addr_copy(&req->wr, &req_rga->dst);

		req->win1.format = req_rga->src.format;
		req->wr.format = req_rga->dst.format;

		if (req_rga->pat.yrgb_addr != 0) {
			/* A+B->C mode */
			set_win_info(&req->win0, &req_rga->pat);
			addr_copy(&req->win0, &req_rga->pat);
			req->win0.format = req_rga->pat.format;

			/* set win0 dst size */
			if (req->win0.x_offset || req->win0.y_offset) {
				req->win0.src_act_w = req->win0.src_act_w + req->win0.x_offset;
				req->win0.src_act_h = req->win0.src_act_h + req->win0.y_offset;
				req->win0.dst_act_w = req_rga->dst.act_w + req->win0.x_offset;
				req->win0.dst_act_h = req_rga->dst.act_h + req->win0.y_offset;

				req->win0.x_offset = 0;
				req->win0.y_offset = 0;
			} else {
				req->win0.dst_act_w = req_rga->dst.act_w;
				req->win0.dst_act_h = req_rga->dst.act_h;
			}
			/* set win1 dst size */
			req->win1.dst_act_w = req_rga->dst.act_w;
			req->win1.dst_act_h = req_rga->dst.act_h;
		} else {
			/* A+B->B mode */
			set_win_info(&req->win0, &req_rga->dst);
			addr_copy(&req->win0, &req_rga->dst);
			req->win0.format = req_rga->dst.format;

			/* only win1 && wr support fbcd, win0 default raster */
			req->win0.rd_mode = 0;

			/* set win0 dst size */
			req->win0.dst_act_w = req_rga->dst.act_w;
			req->win0.dst_act_h = req_rga->dst.act_h;
			/* set win1 dst size */
			req->win1.dst_act_w = req_rga->dst.act_w;
			req->win1.dst_act_h = req_rga->dst.act_h;
		}

		/* dst offset need to config overlap offset */
		req->wr.x_offset = req_rga->dst.x_offset;
		req->wr.y_offset = req_rga->dst.y_offset;
	}
	set_wr_info(req_rga, req);

	if (req->rotate_mode & RGA3_ROT_BIT_ROT_90) {
		if (req->win1.yrgb_addr != 0) {
			/* ABB */
			if (req->win0.yrgb_addr == req->wr.yrgb_addr) {
				req->win1.dst_act_w = req_rga->dst.act_h;
				req->win1.dst_act_h = req_rga->dst.act_w;

				/* win0 do not need rotate, but net equal to wr */
				req->win0.dst_act_w = req_rga->dst.act_h;
				req->win0.dst_act_h = req_rga->dst.act_w;
				req->win0.src_act_w = req_rga->dst.act_h;
				req->win0.src_act_h = req_rga->dst.act_w;
			}
		} else {
			req->win0.rotate_mode = 1;
			req->win0.dst_act_w = req_rga->dst.act_h;
			req->win0.dst_act_h = req_rga->dst.act_w;
		}
	}

	/* overlap */
	/* Alpha blend mode */
	if (((req_rga->alpha_rop_flag) & 1)) {
		if ((req_rga->alpha_rop_flag >> 3) & 1) {
			req->alpha_config.enable = true;

			if ((req_rga->alpha_rop_flag >> 9) & 1) {
				req->alpha_config.fg_pre_multiplied = false;
				req->alpha_config.bg_pre_multiplied = false;
			} else {
				req->alpha_config.fg_pre_multiplied = true;
				req->alpha_config.bg_pre_multiplied = true;
			}

			req->alpha_config.fg_pixel_alpha_en = rga_is_alpha_format(req->win1.format);
			req->alpha_config.bg_pixel_alpha_en = rga_is_alpha_format(req->win0.format);

			if (req_rga->feature.global_alpha_en) {
				if (req_rga->fg_global_alpha < 0xff) {
					req->alpha_config.fg_global_alpha_en = true;
					req->alpha_config.fg_global_alpha_value =
						req_rga->fg_global_alpha;
				} else if (!req->alpha_config.fg_pixel_alpha_en) {
					req->alpha_config.fg_global_alpha_en = true;
					req->alpha_config.fg_global_alpha_value = 0xff;
				}

				if (req_rga->bg_global_alpha < 0xff) {
					req->alpha_config.bg_global_alpha_en = true;
					req->alpha_config.bg_global_alpha_value =
						req_rga->bg_global_alpha;
				} else if (!req->alpha_config.bg_pixel_alpha_en) {
					req->alpha_config.bg_global_alpha_en = true;
					req->alpha_config.bg_global_alpha_value = 0xff;
				}
			} else {
				req->alpha_config.fg_global_alpha_value = 0xff;
				req->alpha_config.bg_global_alpha_value = 0xff;
			}

			req->alpha_config.mode = req_rga->PD_mode;
		}
	}

	/* yuv to rgb */
	/* 601 limit */
	if (req_rga->yuv2rgb_mode == 1) {
		req->win0.y2r_mode = 0;
		req->win1.y2r_mode = 0;
	/* 601 full */
	} else if (req_rga->yuv2rgb_mode == 2) {
		req->win0.y2r_mode = 2;
		req->win1.y2r_mode = 2;
	/* 709 limit */
	} else if (req_rga->yuv2rgb_mode == 3) {
		req->win0.y2r_mode = 1;
		req->win1.y2r_mode = 1;
	}

	/* rgb to yuv */
	/* 601 limit */
	if ((req_rga->yuv2rgb_mode >> 2) == 2) {
		req->win0.r2y_mode = 0;
		req->win1.r2y_mode = 0;
	/* 601 full */
	} else if ((req_rga->yuv2rgb_mode >> 2) == 1) {
		req->win0.r2y_mode = 2;
		req->win1.r2y_mode = 2;
	/* 709 limit */
	} else if ((req_rga->yuv2rgb_mode >> 2) == 3) {
		req->win0.r2y_mode = 1;
		req->win1.r2y_mode = 1;
	}

	/* color key: 8bit->10bit */
	req->color_key_min = (req_rga->color_key_min & 0xff) << 22 |
			     ((req_rga->color_key_min >> 8) & 0xff) << 2 |
			     ((req_rga->color_key_min >> 16) & 0xff) << 12;
	req->color_key_max = (req_rga->color_key_max & 0xff) << 22 |
			     ((req_rga->color_key_max >> 8) & 0xff) << 2 |
			     ((req_rga->color_key_max >> 16) & 0xff) << 12;

	if (req_rga->mmu_info.mmu_en && (req_rga->mmu_info.mmu_flag & 1) == 1) {
		req->mmu_info.src0_mmu_flag = 1;
		req->mmu_info.src1_mmu_flag = 1;
		req->mmu_info.dst_mmu_flag = 1;
	}
}

static void rga3_soft_reset(struct rga_scheduler_t *scheduler)
{
	u32 i;
	u32 iommu_dte_addr = 0;

	if (scheduler->data->mmu == RGA_IOMMU)
		iommu_dte_addr = rga_read(RGA_IOMMU_DTE_ADDR, scheduler);

	rga_write(s_RGA3_SYS_CTRL_CCLK_SRESET(1) | s_RGA3_SYS_CTRL_ACLK_SRESET(1),
		  RGA3_SYS_CTRL, scheduler);

	for (i = 0; i < RGA_RESET_TIMEOUT; i++) {
		if (rga_read(RGA3_RO_SRST, scheduler) & m_RGA3_RO_SRST_RO_RST_DONE)
			break;

		udelay(1);
	}

	rga_write(s_RGA3_SYS_CTRL_CCLK_SRESET(0) | s_RGA3_SYS_CTRL_ACLK_SRESET(0),
		  RGA3_SYS_CTRL, scheduler);

	if (scheduler->data->mmu == RGA_IOMMU) {
		rga_write(iommu_dte_addr, RGA_IOMMU_DTE_ADDR, scheduler);
		/* enable iommu */
		rga_write(RGA_IOMMU_CMD_ENABLE_PAGING, RGA_IOMMU_COMMAND, scheduler);
	}

	if (i == RGA_RESET_TIMEOUT)
		rga_err("%s[%#x] soft reset timeout. SYS_CTRL[0x%x], RO_SRST[0x%x]\n",
			rga_get_core_name(scheduler->core), scheduler->core,
			rga_read(RGA3_SYS_CTRL, scheduler),
			rga_read(RGA3_RO_SRST, scheduler));
	else
		rga_log("%s[%#x] soft reset complete.\n",
			rga_get_core_name(scheduler->core), scheduler->core);
}

static int rga3_scale_check(struct rga_job *job, const struct rga3_req *req)
{
	u32 win0_saw, win0_sah, win0_daw, win0_dah;
	u32 win1_saw, win1_sah, win1_daw, win1_dah;

	if (req->rotate_mode & RGA3_ROT_BIT_ROT_90) {
		if (req->win1.yrgb_addr != 0) {
			/* ABB */
			if (req->win0.yrgb_addr == req->wr.yrgb_addr) {
				/* win0 do not need rotate, but net equal to wr */
				win0_saw = req->win0.src_act_h;
				win0_sah = req->win0.src_act_w;
				win0_daw = req->win0.dst_act_h;
				win0_dah = req->win0.dst_act_w;

				win1_saw = req->win1.dst_act_w;
				win1_sah = req->win1.dst_act_h;
				win1_daw = req->win1.dst_act_h;
				win1_dah = req->win1.dst_act_w;
			} else {
				win0_saw = req->win0.src_act_w;
				win0_sah = req->win0.src_act_h;
				win0_daw = req->win0.dst_act_w;
				win0_dah = req->win0.dst_act_h;

				win1_saw = req->win1.src_act_w;
				win1_sah = req->win1.src_act_h;
				win1_daw = req->win1.dst_act_w;
				win1_dah = req->win1.dst_act_h;
			}
		} else {
			win0_saw = req->win0.src_act_w;
			win0_sah = req->win0.src_act_h;
			win0_daw = req->win0.dst_act_h;
			win0_dah = req->win0.dst_act_w;
		}
	} else {
		win0_saw = req->win0.src_act_w;
		win0_sah = req->win0.src_act_h;
		win0_daw = req->win0.dst_act_w;
		win0_dah = req->win0.dst_act_h;

		if (req->win1.yrgb_addr != 0) {
			win1_saw = req->win1.src_act_w;
			win1_sah = req->win1.src_act_h;
			win1_daw = req->win1.dst_act_w;
			win1_dah = req->win1.dst_act_h;
		}
	}

	if (((win0_saw >> 3) > win0_daw) || ((win0_sah >> 3) > win0_dah)) {
		rga_job_log(job, "win0 unsupported to scaling less than 1/8 times. src[%d, %d], dst[%d, %d]\n",
			win0_saw, win0_sah, win0_daw, win0_dah);
		return -EINVAL;
	}
	if (((win0_daw >> 3) > win0_saw) || ((win0_dah >> 3) > win0_sah)) {
		rga_job_log(job, "win0 unsupported to scaling more than 8 times. src[%d, %d], dst[%d, %d]\n",
			win0_saw, win0_sah, win0_daw, win0_dah);
		return -EINVAL;
	}

	if (req->win1.yrgb_addr != 0) {
		if (((win1_saw >> 3) > win1_daw) || ((win1_sah >> 3) > win1_dah)) {
			rga_job_log(job, "win1 unsupported to scaling less than 1/8 times. src[%d, %d], dst[%d, %d]\n",
				win1_saw, win1_sah, win1_daw, win1_dah);
			return -EINVAL;
		}
		if (((win1_daw >> 3) > win1_saw) || ((win1_dah >> 3) > win1_sah)) {
			rga_job_log(job, "win1 unsupported to scaling more than 8 times. src[%d, %d], dst[%d, %d]\n",
				win1_saw, win1_sah, win1_daw, win1_dah);
			return -EINVAL;
		}
	}

	return 0;
}

static int rga3_check_param(struct rga_job *job, const struct rga_hw_data *data,
			    const struct rga3_req *req)
{
	if (unlikely(rga_hw_out_of_range(&(data->input_range),
					 req->win0.src_act_w, req->win0.src_act_h) ||
		     rga_hw_out_of_range(&(data->input_range),
					 req->win0.dst_act_w, req->win0.dst_act_h) ||
		     rga_hw_out_of_range(&(data->input_range),
					 req->win0.src_act_w + req->win0.x_offset,
					 req->win0.src_act_h + req->win0.y_offset))) {
		rga_job_err(job, "invalid win0, src[w,h] = [%d, %d], dst[w,h] = [%d, %d], off[x,y] = [%d,%d]\n",
			req->win0.src_act_w, req->win0.src_act_h,
			req->win0.dst_act_w, req->win0.dst_act_h,
			req->win0.x_offset, req->win0.y_offset);
		return -EINVAL;
	}

	if (unlikely(req->win0.vir_w * rga_get_pixel_stride_from_format(req->win0.format) >
		     data->max_byte_stride * 8)) {
		rga_job_err(job, "invalid win0 stride, stride = %d, pixel_stride = %d, max_byte_stride = %d\n",
			req->win0.vir_w, rga_get_pixel_stride_from_format(req->win0.format),
			data->max_byte_stride);
		return -EINVAL;
	}

	if (unlikely(rga_hw_out_of_range(&(data->output_range),
					 req->wr.dst_act_w, req->wr.dst_act_h))) {
		rga_job_err(job, "invalid wr, [w,h] = [%d, %d]\n", req->wr.dst_act_w, req->wr.dst_act_h);
		return -EINVAL;
	}

	if (unlikely(req->wr.vir_w * rga_get_pixel_stride_from_format(req->wr.format) >
		     data->max_byte_stride * 8)) {
		rga_job_err(job, "invalid wr stride, stride = %d, pixel_stride = %d, max_byte_stride = %d\n",
			req->wr.vir_w, rga_get_pixel_stride_from_format(req->wr.format),
			data->max_byte_stride);
		return -EINVAL;
	}

	if (req->win1.yrgb_addr != 0) {
		if (unlikely(rga_hw_out_of_range(&(data->input_range),
						 req->win1.src_act_w, req->win1.src_act_h) ||
			     rga_hw_out_of_range(&(data->input_range),
						 req->win1.dst_act_w, req->win1.dst_act_h) ||
			     rga_hw_out_of_range(&(data->input_range),
						 req->win1.src_act_w + req->win1.x_offset,
						 req->win1.src_act_h + req->win1.y_offset))) {
			rga_job_err(job, "invalid win1, src[w,h] = [%d, %d], dst[w,h] = [%d, %d], off[x,y] = [%d,%d]\n",
				req->win1.src_act_w, req->win1.src_act_h,
				req->win1.dst_act_w, req->win1.dst_act_h,
				req->win1.x_offset, req->win1.y_offset);
			return -EINVAL;
		}

		if (unlikely(req->win1.vir_w * rga_get_pixel_stride_from_format(req->win1.format) >
			     data->max_byte_stride * 8)) {
			rga_job_err(job, "invalid win1 stride, stride = %d, pixel_stride = %d, max_byte_stride = %d\n",
				req->win1.vir_w, rga_get_pixel_stride_from_format(req->win1.format),
				data->max_byte_stride);
			return -EINVAL;
		}

		/* warning: rotate mode skip this judge */
		if (req->rotate_mode == 0) {
			/* check win0 dst size > win1 dst size */
			if (unlikely((req->win1.dst_act_w > req->win0.dst_act_w) ||
				     (req->win1.dst_act_h > req->win0.dst_act_h))) {
				rga_job_err(job, "invalid output param win0[w,h] = [%d, %d], win1[w,h] = [%d, %d]\n",
					req->win0.dst_act_w, req->win0.dst_act_h,
					req->win1.dst_act_w, req->win1.dst_act_h);
				return -EINVAL;
			}
		}
	}

	if (rga3_scale_check(job, req) < 0)
		return -EINVAL;

	return 0;
}

static void print_debug_info(struct rga_job *job, struct rga3_req *req)
{
	rga_job_log(job, "render_mode:%s, bitblit_mode=%d, rotate_mode:%x\n",
		rga_get_render_mode_str(req->render_mode), req->bitblt_mode,
		req->rotate_mode);
	rga_job_log(job, "win0: y = %lx uv = %lx v = %lx src_w = %d src_h = %d\n",
		req->win0.yrgb_addr, req->win0.uv_addr, req->win0.v_addr,
		req->win0.src_act_w, req->win0.src_act_h);
	rga_job_log(job, "win0: vw = %d vh = %d xoff = %d yoff = %d format = %s\n",
		req->win0.vir_w, req->win0.vir_h,
		req->win0.x_offset, req->win0.y_offset,
		rga_get_format_name(req->win0.format));
	rga_job_log(job, "win0: dst_w = %d, dst_h = %d, rd_mode = %d\n",
		req->win0.dst_act_w, req->win0.dst_act_h, req->win0.rd_mode);
	rga_job_log(job, "win0: rot_mode = %d, en = %d, compact = %d, endian = %d\n",
		req->win0.rotate_mode, req->win0.enable,
		req->win0.is_10b_compact, req->win0.is_10b_endian);

	if (req->win1.yrgb_addr != 0 || req->win1.uv_addr != 0
		|| req->win1.v_addr != 0) {
		rga_job_log(job, "win1: y = %lx uv = %lx v = %lx src_w = %d src_h = %d\n",
			req->win1.yrgb_addr, req->win1.uv_addr,
			req->win1.v_addr, req->win1.src_act_w,
			req->win1.src_act_h);
		rga_job_log(job, "win1: vw = %d vh = %d xoff = %d yoff = %d format = %s\n",
			req->win1.vir_w, req->win1.vir_h,
			req->win1.x_offset, req->win1.y_offset,
			rga_get_format_name(req->win1.format));
		rga_job_log(job, "win1: dst_w = %d, dst_h = %d, rd_mode = %d\n",
			req->win1.dst_act_w, req->win1.dst_act_h,
			req->win1.rd_mode);
		rga_job_log(job, "win1: rot_mode = %d, en = %d, compact = %d, endian = %d\n",
			req->win1.rotate_mode, req->win1.enable,
			req->win1.is_10b_compact, req->win1.is_10b_endian);
	}

	rga_job_log(job, "wr: y = %lx uv = %lx v = %lx vw = %d vh = %d\n",
		req->wr.yrgb_addr, req->wr.uv_addr, req->wr.v_addr,
		req->wr.vir_w, req->wr.vir_h);
	rga_job_log(job, "wr: ovlp_xoff = %d ovlp_yoff = %d format = %s rdmode = %d\n",
		req->wr.x_offset, req->wr.y_offset,
		rga_get_format_name(req->wr.format), req->wr.rd_mode);

	rga_job_log(job, "mmu: win0 = %.2x win1 = %.2x wr = %.2x\n",
		req->mmu_info.src0_mmu_flag, req->mmu_info.src1_mmu_flag,
		req->mmu_info.dst_mmu_flag);
	rga_job_log(job, "alpha: flag %x mode=%s\n",
		req->alpha_rop_flag, rga_get_blend_mode_str(req->alpha_config.mode));
	rga_job_log(job, "alpha: pre_multi=[%d,%d] pixl=[%d,%d] glb=[%d,%d]\n",
		req->alpha_config.fg_pre_multiplied, req->alpha_config.bg_pre_multiplied,
		req->alpha_config.fg_pixel_alpha_en, req->alpha_config.bg_pixel_alpha_en,
		req->alpha_config.fg_global_alpha_en, req->alpha_config.bg_global_alpha_en);
	rga_job_log(job, "alpha: fg_global_alpha=%x bg_global_alpha=%x\n",
		req->alpha_config.fg_global_alpha_value, req->alpha_config.bg_global_alpha_value);
	rga_job_log(job, "yuv2rgb mode is %x\n", req->yuv2rgb_mode);
}

static int rga3_align_check(struct rga_job *job, struct rga3_req *req)
{
	if (rga_is_yuv10bit_format(req->win0.format))
		if ((req->win0.vir_w % 64) || (req->win0.x_offset % 4) ||
			(req->win0.src_act_w % 4) || (req->win0.y_offset % 4) ||
			(req->win0.src_act_h % 4) || (req->win0.vir_h % 2))
			rga_job_log(job, "yuv10bit err win0 wstride is not align\n");
	if (rga_is_yuv10bit_format(req->win1.format))
		if ((req->win1.vir_w % 64) || (req->win1.x_offset % 4) ||
			(req->win1.src_act_w % 4) || (req->win1.y_offset % 4) ||
			(req->win1.src_act_h % 4) || (req->win1.vir_h % 2))
			rga_job_log(job, "yuv10bit err win1 wstride is not align\n");
	if (rga_is_yuv8bit_format(req->win0.format))
		if ((req->win0.vir_w % 16) || (req->win0.x_offset % 2) ||
			(req->win0.src_act_w % 2) || (req->win0.y_offset % 2) ||
			(req->win0.src_act_h % 2) || (req->win0.vir_h % 2))
			rga_job_log(job, "yuv8bit err win0 wstride is not align\n");
	if (rga_is_yuv8bit_format(req->win1.format))
		if ((req->win1.vir_w % 16) || (req->win1.x_offset % 2) ||
			(req->win1.src_act_w % 2) || (req->win1.y_offset % 2) ||
			(req->win1.src_act_h % 2) || (req->win1.vir_h % 2))
			rga_job_log(job, "yuv8bit err win1 wstride is not align\n");
	return 0;
}

static int rga3_init_reg(struct rga_job *job)
{
	struct rga3_req req;
	int ret = 0;
	struct rga_scheduler_t *scheduler = NULL;
	ktime_t timestamp = ktime_get();

	scheduler = job->scheduler;
	if (unlikely(scheduler == NULL)) {
		rga_job_err(job, "failed to get scheduler, %s(%d)\n", __func__, __LINE__);
		return -EINVAL;
	}

	memset(&req, 0x0, sizeof(req));

	rga_cmd_to_rga3_cmd(&job->rga_command_base, &req);

	/* check value if legal */
	ret = rga3_check_param(job, scheduler->data, &req);
	if (ret == -EINVAL) {
		rga_job_err(job, "req argument is inval\n");
		return ret;
	}

	rga3_align_check(job, &req);

	/* for debug */
	if (DEBUGGER_EN(MSG))
		print_debug_info(job, &req);

	if (rga3_gen_reg_info((uint8_t *) job->cmd_buf->vaddr, &req) == -1) {
		rga_job_err(job, "RKA: gen reg info error\n");
		return -EINVAL;
	}

	if (DEBUGGER_EN(TIME))
		rga_job_log(job, "generate register cost time %lld us\n",
			ktime_us_delta(ktime_get(), timestamp));

	return ret;
}

static void rga3_dump_read_back_reg(struct rga_job *job, struct rga_scheduler_t *scheduler)
{
	int i;
	unsigned long flags;
	uint32_t cmd_reg[48] = {0};

	spin_lock_irqsave(&scheduler->irq_lock, flags);

	for (i = 0; i < 48; i++)
		cmd_reg[i] = rga_read(0x100 + i * 4, scheduler);

	spin_unlock_irqrestore(&scheduler->irq_lock, flags);

	rga_job_log(job, "CMD_READ_BACK_REG\n");
	for (i = 0; i < 12; i++)
		rga_job_log(job, "i = %x : %.8x %.8x %.8x %.8x\n", i,
			cmd_reg[0 + i * 4], cmd_reg[1 + i * 4],
			cmd_reg[2 + i * 4], cmd_reg[3 + i * 4]);
}

static int rga3_set_reg(struct rga_job *job, struct rga_scheduler_t *scheduler)
{
	int i;
	bool master_mode_en;
	uint32_t sys_ctrl;
	uint32_t *cmd;
	ktime_t now = ktime_get();

	cmd = job->cmd_buf->vaddr;

	/*
	 * Currently there is no iova allocated for storing cmd for the IOMMU device,
	 * so the iommu device needs to use the slave mode.
	 */
	if (scheduler->data->mmu != RGA_IOMMU)
		master_mode_en = true;
	else
		master_mode_en = false;

	if (DEBUGGER_EN(REG)) {
		rga_job_log(job, "CMD_REG\n");
		for (i = 0; i < 12; i++)
			rga_job_log(job, "i = %x : %.8x %.8x %.8x %.8x\n", i,
				cmd[0 + i * 4], cmd[1 + i * 4],
				cmd[2 + i * 4], cmd[3 + i * 4]);
	}

	/* All CMD finish int */
	rga_write(m_RGA3_INT_FRM_DONE | m_RGA3_INT_CMD_LINE_FINISH | m_RGA3_INT_ERROR_MASK,
		  RGA3_INT_EN, scheduler);

	if (master_mode_en) {
		/* master mode */
		sys_ctrl = s_RGA3_SYS_CTRL_CMD_MODE(1);

		rga_write(job->cmd_buf->dma_addr, RGA3_CMD_ADDR, scheduler);
		rga_write(sys_ctrl, RGA3_SYS_CTRL, scheduler);
		rga_write(m_RGA3_CMD_CTRL_CMD_LINE_ST_P, RGA3_CMD_CTRL, scheduler);
	} else {
		/* slave mode */
		sys_ctrl = s_RGA3_SYS_CTRL_CMD_MODE(0) | m_RGA3_SYS_CTRL_RGA_SART;

		for (i = 0; i <= 50; i++)
			rga_write(cmd[i], 0x100 + i * 4, scheduler);

		rga_write(sys_ctrl, RGA3_SYS_CTRL, scheduler);
	}

	if (DEBUGGER_EN(REG)) {
		rga_job_log(job, "sys_ctrl = 0x%x, int_en = 0x%x, int_raw = 0x%x\n",
			rga_read(RGA3_SYS_CTRL, scheduler),
			rga_read(RGA3_INT_EN, scheduler),
			rga_read(RGA3_INT_RAW, scheduler));

		rga_job_log(job, "hw_status = 0x%x, cmd_status = 0x%x\n",
			rga_read(RGA3_STATUS0, scheduler),
			rga_read(RGA3_CMD_STATE, scheduler));
	}

	if (DEBUGGER_EN(TIME))
		rga_job_log(job, "set register cost time %lld us\n",
			ktime_us_delta(ktime_get(), now));

	job->timestamp.hw_execute = now;
	job->timestamp.hw_recode = now;
	job->session->last_active = now;

	if (DEBUGGER_EN(REG))
		rga3_dump_read_back_reg(job, scheduler);

	return 0;
}

static int rga3_get_version(struct rga_scheduler_t *scheduler)
{
	u32 major_version, minor_version, svn_version;
	u32 reg_version;

	if (!scheduler) {
		rga_err("scheduler is null\n");
		return -EINVAL;
	}

	reg_version = rga_read(RGA3_VERSION_NUM, scheduler);

	major_version = (reg_version & RGA3_MAJOR_VERSION_MASK) >> 28;
	minor_version = (reg_version & RGA3_MINOR_VERSION_MASK) >> 20;
	svn_version = (reg_version & RGA3_SVN_VERSION_MASK);

	snprintf(scheduler->version.str, 10, "%x.%01x.%05x", major_version,
		 minor_version, svn_version);

	scheduler->version.major = major_version;
	scheduler->version.minor = minor_version;
	scheduler->version.revision = svn_version;

	return 0;
}

static int rga3_read_status(struct rga_job *job, struct rga_scheduler_t *scheduler)
{
	job->intr_status = rga_read(RGA3_INT_RAW, scheduler);
	job->hw_status = rga_read(RGA3_STATUS0, scheduler);
	job->cmd_status = rga_read(RGA3_CMD_STATE, scheduler);
	job->work_cycle = 0;

	return 0;
}

static void rga3_clear_intr(struct rga_scheduler_t *scheduler)
{
	rga_write(m_RGA3_INT_FRM_DONE | m_RGA3_INT_CMD_LINE_FINISH | m_RGA3_INT_ERROR_MASK,
		  RGA3_INT_CLR, scheduler);
}

static int rga3_irq(struct rga_scheduler_t *scheduler)
{
	struct rga_job *job = scheduler->running_job;

	if (job == NULL) {
		rga3_clear_intr(scheduler);
		rga_err("core[%d], invalid job, INTR[0x%x], HW_STATUS[0x%x], CMD_STATUS[0x%x]\n",
			scheduler->core, rga_read(RGA3_INT_RAW, scheduler),
			rga_read(RGA3_STATUS0, scheduler), rga_read(RGA3_CMD_STATE, scheduler));

		return IRQ_HANDLED;
	}

	if (test_bit(RGA_JOB_STATE_INTR_ERR, &job->state)) {
		rga3_clear_intr(scheduler);
		return IRQ_WAKE_THREAD;
	}

	scheduler->ops->read_status(job, scheduler);

	if (DEBUGGER_EN(INT_FLAG))
		rga_job_log(job, "irq handler, INTR[0x%x], HW_STATUS[0x%x], CMD_STATUS[0x%x]\n",
			job->intr_status, job->hw_status, job->cmd_status);

	if (job->intr_status & (m_RGA3_INT_FRM_DONE | m_RGA3_INT_CMD_LINE_FINISH)) {
		set_bit(RGA_JOB_STATE_FINISH, &job->state);
	} else if (job->intr_status & m_RGA3_INT_ERROR_MASK) {
		set_bit(RGA_JOB_STATE_INTR_ERR, &job->state);

		rga_job_err(job, "irq handler err! INTR[0x%x], HW_STATUS[0x%x], CMD_STATUS[0x%x]\n",
		       job->intr_status, job->hw_status, job->cmd_status);
		scheduler->ops->soft_reset(scheduler);
	}

	rga3_clear_intr(scheduler);

	return IRQ_WAKE_THREAD;
}

static int rga3_isr_thread(struct rga_job *job, struct rga_scheduler_t *scheduler)
{
	if (DEBUGGER_EN(INT_FLAG))
		rga_job_log(job, "isr thread, INTR[0x%x], HW_STATUS[0x%x], CMD_STATUS[0x%x]\n",
			rga_read(RGA3_INT_RAW, scheduler),
			rga_read(RGA3_STATUS0, scheduler),
			rga_read(RGA3_CMD_STATE, scheduler));

	if (test_bit(RGA_JOB_STATE_INTR_ERR, &job->state)) {
		if (job->intr_status & m_RGA3_INT_RAG_MI_RD_BUS_ERR) {
			rga_job_err(job, "DMA read bus error, please check size of the input_buffer or whether the buffer has been freed.\n");
			job->ret = -EFAULT;
		} else if (job->intr_status & m_RGA3_INT_WIN0_FBCD_DEC_ERR) {
			rga_job_err(job, "win0 FBC decoder error, please check the fbc image of the source.\n");
			job->ret = -EFAULT;
		} else if (job->intr_status & m_RGA3_INT_WIN1_FBCD_DEC_ERR) {
			rga_job_err(job, "win1 FBC decoder error, please check the fbc image of the source.\n");
			job->ret = -EFAULT;
		} else if (job->intr_status & m_RGA3_INT_RGA_MI_WR_BUS_ERR) {
			rga_job_err(job, "wr buss error, please check size of the output_buffer or whether the buffer has been freed.\n");
			job->ret = -EFAULT;
		}

		if (job->ret == 0) {
			rga_job_err(job, "rga intr error[0x%x]!\n", job->intr_status);
			job->ret = -EFAULT;
		}
	}

	return IRQ_HANDLED;
}

const struct rga_backend_ops rga3_ops = {
	.get_version = rga3_get_version,
	.set_reg = rga3_set_reg,
	.init_reg = rga3_init_reg,
	.soft_reset = rga3_soft_reset,
	.read_back_reg = NULL,
	.read_status = rga3_read_status,
	.irq = rga3_irq,
	.isr_thread = rga3_isr_thread,
};