Contributors: 13
Author Tokens Token Proportion Commits Commit Proportion
Guoqing Li 2850 90.68% 3 16.67%
Jing Xiang 142 4.52% 3 16.67%
Bartlomiej Zolnierkiewicz 105 3.34% 1 5.56%
Zhou Zhu 13 0.41% 1 5.56%
Gustavo A. R. Silva 10 0.32% 1 5.56%
Gong Yuanjun 9 0.29% 1 5.56%
Christophe Jaillet 4 0.13% 2 11.11%
caihuoqing 3 0.10% 1 5.56%
SF Markus Elfring 2 0.06% 1 5.56%
Thomas Gleixner 2 0.06% 1 5.56%
Christoph Hellwig 1 0.03% 1 5.56%
Arnd Bergmann 1 0.03% 1 5.56%
Uwe Kleine-König 1 0.03% 1 5.56%
Total 3143 18


// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * linux/drivers/video/mmp/hw/mmp_ctrl.c
 * Marvell MMP series Display Controller support
 *
 * Copyright (C) 2012 Marvell Technology Group Ltd.
 * Authors:  Guoqing Li <ligq@marvell.com>
 *          Lisa Du <cldu@marvell.com>
 *          Zhou Zhu <zzhu3@marvell.com>
 */
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/string.h>
#include <linux/interrupt.h>
#include <linux/slab.h>
#include <linux/delay.h>
#include <linux/platform_device.h>
#include <linux/dma-mapping.h>
#include <linux/clk.h>
#include <linux/err.h>
#include <linux/vmalloc.h>
#include <linux/uaccess.h>
#include <linux/kthread.h>
#include <linux/io.h>

#include "mmp_ctrl.h"

static irqreturn_t ctrl_handle_irq(int irq, void *dev_id)
{
	struct mmphw_ctrl *ctrl = (struct mmphw_ctrl *)dev_id;
	u32 isr, imask, tmp;

	isr = readl_relaxed(ctrl->reg_base + SPU_IRQ_ISR);
	imask = readl_relaxed(ctrl->reg_base + SPU_IRQ_ENA);

	do {
		/* clear clock only */
		tmp = readl_relaxed(ctrl->reg_base + SPU_IRQ_ISR);
		if (tmp & isr)
			writel_relaxed(~isr, ctrl->reg_base + SPU_IRQ_ISR);
	} while ((isr = readl_relaxed(ctrl->reg_base + SPU_IRQ_ISR)) & imask);

	return IRQ_HANDLED;
}

static u32 fmt_to_reg(struct mmp_overlay *overlay, int pix_fmt)
{
	u32 rbswap = 0, uvswap = 0, yuvswap = 0,
		csc_en = 0, val = 0,
		vid = overlay_is_vid(overlay);

	switch (pix_fmt) {
	case PIXFMT_RGB565:
	case PIXFMT_RGB1555:
	case PIXFMT_RGB888PACK:
	case PIXFMT_RGB888UNPACK:
	case PIXFMT_RGBA888:
		rbswap = 1;
		break;
	case PIXFMT_VYUY:
	case PIXFMT_YVU422P:
	case PIXFMT_YVU420P:
		uvswap = 1;
		break;
	case PIXFMT_YUYV:
		yuvswap = 1;
		break;
	default:
		break;
	}

	switch (pix_fmt) {
	case PIXFMT_RGB565:
	case PIXFMT_BGR565:
		break;
	case PIXFMT_RGB1555:
	case PIXFMT_BGR1555:
		val = 0x1;
		break;
	case PIXFMT_RGB888PACK:
	case PIXFMT_BGR888PACK:
		val = 0x2;
		break;
	case PIXFMT_RGB888UNPACK:
	case PIXFMT_BGR888UNPACK:
		val = 0x3;
		break;
	case PIXFMT_RGBA888:
	case PIXFMT_BGRA888:
		val = 0x4;
		break;
	case PIXFMT_UYVY:
	case PIXFMT_VYUY:
	case PIXFMT_YUYV:
		val = 0x5;
		csc_en = 1;
		break;
	case PIXFMT_YUV422P:
	case PIXFMT_YVU422P:
		val = 0x6;
		csc_en = 1;
		break;
	case PIXFMT_YUV420P:
	case PIXFMT_YVU420P:
		val = 0x7;
		csc_en = 1;
		break;
	default:
		break;
	}

	return (dma_palette(0) | dma_fmt(vid, val) |
		dma_swaprb(vid, rbswap) | dma_swapuv(vid, uvswap) |
		dma_swapyuv(vid, yuvswap) | dma_csc(vid, csc_en));
}

static void dmafetch_set_fmt(struct mmp_overlay *overlay)
{
	u32 tmp;
	struct mmp_path *path = overlay->path;
	tmp = readl_relaxed(ctrl_regs(path) + dma_ctrl(0, path->id));
	tmp &= ~dma_mask(overlay_is_vid(overlay));
	tmp |= fmt_to_reg(overlay, overlay->win.pix_fmt);
	writel_relaxed(tmp, ctrl_regs(path) + dma_ctrl(0, path->id));
}

static void overlay_set_win(struct mmp_overlay *overlay, struct mmp_win *win)
{
	struct lcd_regs *regs = path_regs(overlay->path);

	/* assert win supported */
	memcpy(&overlay->win, win, sizeof(struct mmp_win));

	mutex_lock(&overlay->access_ok);

	if (overlay_is_vid(overlay)) {
		writel_relaxed(win->pitch[0],
				(void __iomem *)&regs->v_pitch_yc);
		writel_relaxed(win->pitch[2] << 16 | win->pitch[1],
				(void __iomem *)&regs->v_pitch_uv);

		writel_relaxed((win->ysrc << 16) | win->xsrc,
				(void __iomem *)&regs->v_size);
		writel_relaxed((win->ydst << 16) | win->xdst,
				(void __iomem *)&regs->v_size_z);
		writel_relaxed(win->ypos << 16 | win->xpos,
				(void __iomem *)&regs->v_start);
	} else {
		writel_relaxed(win->pitch[0], (void __iomem *)&regs->g_pitch);

		writel_relaxed((win->ysrc << 16) | win->xsrc,
				(void __iomem *)&regs->g_size);
		writel_relaxed((win->ydst << 16) | win->xdst,
				(void __iomem *)&regs->g_size_z);
		writel_relaxed(win->ypos << 16 | win->xpos,
				(void __iomem *)&regs->g_start);
	}

	dmafetch_set_fmt(overlay);
	mutex_unlock(&overlay->access_ok);
}

static void dmafetch_onoff(struct mmp_overlay *overlay, int on)
{
	u32 mask = overlay_is_vid(overlay) ? CFG_DMA_ENA_MASK :
		   CFG_GRA_ENA_MASK;
	u32 enable = overlay_is_vid(overlay) ? CFG_DMA_ENA(1) : CFG_GRA_ENA(1);
	u32 tmp;
	struct mmp_path *path = overlay->path;

	mutex_lock(&overlay->access_ok);
	tmp = readl_relaxed(ctrl_regs(path) + dma_ctrl(0, path->id));
	tmp &= ~mask;
	tmp |= (on ? enable : 0);
	writel(tmp, ctrl_regs(path) + dma_ctrl(0, path->id));
	mutex_unlock(&overlay->access_ok);
}

static void path_enabledisable(struct mmp_path *path, int on)
{
	u32 tmp;
	mutex_lock(&path->access_ok);
	tmp = readl_relaxed(ctrl_regs(path) + LCD_SCLK(path));
	if (on)
		tmp &= ~SCLK_DISABLE;
	else
		tmp |= SCLK_DISABLE;
	writel_relaxed(tmp, ctrl_regs(path) + LCD_SCLK(path));
	mutex_unlock(&path->access_ok);
}

static void path_onoff(struct mmp_path *path, int on)
{
	if (path->status == on) {
		dev_info(path->dev, "path %s is already %s\n",
				path->name, stat_name(path->status));
		return;
	}

	if (on) {
		path_enabledisable(path, 1);

		if (path->panel && path->panel->set_onoff)
			path->panel->set_onoff(path->panel, 1);
	} else {
		if (path->panel && path->panel->set_onoff)
			path->panel->set_onoff(path->panel, 0);

		path_enabledisable(path, 0);
	}
	path->status = on;
}

static void overlay_set_onoff(struct mmp_overlay *overlay, int on)
{
	if (overlay->status == on) {
		dev_info(overlay_to_ctrl(overlay)->dev, "overlay %s is already %s\n",
			overlay->path->name, stat_name(overlay->status));
		return;
	}
	overlay->status = on;
	dmafetch_onoff(overlay, on);
	if (overlay->path->ops.check_status(overlay->path)
			!= overlay->path->status)
		path_onoff(overlay->path, on);
}

static void overlay_set_fetch(struct mmp_overlay *overlay, int fetch_id)
{
	overlay->dmafetch_id = fetch_id;
}

static int overlay_set_addr(struct mmp_overlay *overlay, struct mmp_addr *addr)
{
	struct lcd_regs *regs = path_regs(overlay->path);

	/* FIXME: assert addr supported */
	memcpy(&overlay->addr, addr, sizeof(struct mmp_addr));

	if (overlay_is_vid(overlay)) {
		writel_relaxed(addr->phys[0], (void __iomem *)&regs->v_y0);
		writel_relaxed(addr->phys[1], (void __iomem *)&regs->v_u0);
		writel_relaxed(addr->phys[2], (void __iomem *)&regs->v_v0);
	} else
		writel_relaxed(addr->phys[0], (void __iomem *)&regs->g_0);

	return overlay->addr.phys[0];
}

static void path_set_mode(struct mmp_path *path, struct mmp_mode *mode)
{
	struct lcd_regs *regs = path_regs(path);
	u32 total_x, total_y, vsync_ctrl, tmp, sclk_src, sclk_div,
		link_config = path_to_path_plat(path)->link_config,
		dsi_rbswap = path_to_path_plat(path)->link_config;

	/* FIXME: assert videomode supported */
	memcpy(&path->mode, mode, sizeof(struct mmp_mode));

	mutex_lock(&path->access_ok);

	/* polarity of timing signals */
	tmp = readl_relaxed(ctrl_regs(path) + intf_ctrl(path->id)) & 0x1;
	tmp |= mode->vsync_invert ? 0 : 0x8;
	tmp |= mode->hsync_invert ? 0 : 0x4;
	tmp |= link_config & CFG_DUMBMODE_MASK;
	tmp |= CFG_DUMB_ENA(1);
	writel_relaxed(tmp, ctrl_regs(path) + intf_ctrl(path->id));

	/* interface rb_swap setting */
	tmp = readl_relaxed(ctrl_regs(path) + intf_rbswap_ctrl(path->id)) &
		(~(CFG_INTFRBSWAP_MASK));
	tmp |= dsi_rbswap & CFG_INTFRBSWAP_MASK;
	writel_relaxed(tmp, ctrl_regs(path) + intf_rbswap_ctrl(path->id));

	writel_relaxed((mode->yres << 16) | mode->xres,
		(void __iomem *)&regs->screen_active);
	writel_relaxed((mode->left_margin << 16) | mode->right_margin,
		(void __iomem *)&regs->screen_h_porch);
	writel_relaxed((mode->upper_margin << 16) | mode->lower_margin,
		(void __iomem *)&regs->screen_v_porch);
	total_x = mode->xres + mode->left_margin + mode->right_margin +
		mode->hsync_len;
	total_y = mode->yres + mode->upper_margin + mode->lower_margin +
		mode->vsync_len;
	writel_relaxed((total_y << 16) | total_x,
		(void __iomem *)&regs->screen_size);

	/* vsync ctrl */
	if (path->output_type == PATH_OUT_DSI)
		vsync_ctrl = 0x01330133;
	else
		vsync_ctrl = ((mode->xres + mode->right_margin) << 16)
					| (mode->xres + mode->right_margin);
	writel_relaxed(vsync_ctrl, (void __iomem *)&regs->vsync_ctrl);

	/* set pixclock div */
	sclk_src = clk_get_rate(path_to_ctrl(path)->clk);
	sclk_div = sclk_src / mode->pixclock_freq;
	if (sclk_div * mode->pixclock_freq < sclk_src)
		sclk_div++;

	dev_info(path->dev, "%s sclk_src %d sclk_div 0x%x pclk %d\n",
			__func__, sclk_src, sclk_div, mode->pixclock_freq);

	tmp = readl_relaxed(ctrl_regs(path) + LCD_SCLK(path));
	tmp &= ~CLK_INT_DIV_MASK;
	tmp |= sclk_div;
	writel_relaxed(tmp, ctrl_regs(path) + LCD_SCLK(path));

	mutex_unlock(&path->access_ok);
}

static const struct mmp_overlay_ops mmphw_overlay_ops = {
	.set_fetch = overlay_set_fetch,
	.set_onoff = overlay_set_onoff,
	.set_win = overlay_set_win,
	.set_addr = overlay_set_addr,
};

static void ctrl_set_default(struct mmphw_ctrl *ctrl)
{
	u32 tmp, irq_mask;

	/*
	 * LCD Global control(LCD_TOP_CTRL) should be configed before
	 * any other LCD registers read/write, or there maybe issues.
	 */
	tmp = readl_relaxed(ctrl->reg_base + LCD_TOP_CTRL);
	tmp |= 0xfff0;
	writel_relaxed(tmp, ctrl->reg_base + LCD_TOP_CTRL);


	/* disable all interrupts */
	irq_mask = path_imasks(0) | err_imask(0) |
		   path_imasks(1) | err_imask(1);
	tmp = readl_relaxed(ctrl->reg_base + SPU_IRQ_ENA);
	tmp &= ~irq_mask;
	tmp |= irq_mask;
	writel_relaxed(tmp, ctrl->reg_base + SPU_IRQ_ENA);
}

static void path_set_default(struct mmp_path *path)
{
	struct lcd_regs *regs = path_regs(path);
	u32 dma_ctrl1, mask, tmp, path_config;

	path_config = path_to_path_plat(path)->path_config;

	/* Configure IOPAD: should be parallel only */
	if (PATH_OUT_PARALLEL == path->output_type) {
		mask = CFG_IOPADMODE_MASK | CFG_BURST_MASK | CFG_BOUNDARY_MASK;
		tmp = readl_relaxed(ctrl_regs(path) + SPU_IOPAD_CONTROL);
		tmp &= ~mask;
		tmp |= path_config;
		writel_relaxed(tmp, ctrl_regs(path) + SPU_IOPAD_CONTROL);
	}

	/* Select path clock source */
	tmp = readl_relaxed(ctrl_regs(path) + LCD_SCLK(path));
	tmp &= ~SCLK_SRC_SEL_MASK;
	tmp |= path_config;
	writel_relaxed(tmp, ctrl_regs(path) + LCD_SCLK(path));

	/*
	 * Configure default bits: vsync triggers DMA,
	 * power save enable, configure alpha registers to
	 * display 100% graphics, and set pixel command.
	 */
	dma_ctrl1 = 0x2032ff81;

	dma_ctrl1 |= CFG_VSYNC_INV_MASK;
	writel_relaxed(dma_ctrl1, ctrl_regs(path) + dma_ctrl(1, path->id));

	/* Configure default register values */
	writel_relaxed(0x00000000, (void __iomem *)&regs->blank_color);
	writel_relaxed(0x00000000, (void __iomem *)&regs->g_1);
	writel_relaxed(0x00000000, (void __iomem *)&regs->g_start);

	/*
	 * 1.enable multiple burst request in DMA AXI
	 * bus arbiter for faster read if not tv path;
	 * 2.enable horizontal smooth filter;
	 */
	mask = CFG_GRA_HSMOOTH_MASK | CFG_DMA_HSMOOTH_MASK | CFG_ARBFAST_ENA(1);
	tmp = readl_relaxed(ctrl_regs(path) + dma_ctrl(0, path->id));
	tmp |= mask;
	if (PATH_TV == path->id)
		tmp &= ~CFG_ARBFAST_ENA(1);
	writel_relaxed(tmp, ctrl_regs(path) + dma_ctrl(0, path->id));
}

static int path_init(struct mmphw_path_plat *path_plat,
		struct mmp_mach_path_config *config)
{
	struct mmphw_ctrl *ctrl = path_plat->ctrl;
	struct mmp_path_info *path_info;
	struct mmp_path *path = NULL;

	dev_info(ctrl->dev, "%s: %s\n", __func__, config->name);

	/* init driver data */
	path_info = kzalloc(sizeof(*path_info), GFP_KERNEL);
	if (!path_info)
		return 0;

	path_info->name = config->name;
	path_info->id = path_plat->id;
	path_info->dev = ctrl->dev;
	path_info->overlay_num = config->overlay_num;
	path_info->overlay_ops = &mmphw_overlay_ops;
	path_info->set_mode = path_set_mode;
	path_info->plat_data = path_plat;

	/* create/register platform device */
	path = mmp_register_path(path_info);
	if (!path) {
		kfree(path_info);
		return 0;
	}
	path_plat->path = path;
	path_plat->path_config = config->path_config;
	path_plat->link_config = config->link_config;
	path_plat->dsi_rbswap = config->dsi_rbswap;
	path_set_default(path);

	kfree(path_info);
	return 1;
}

static void path_deinit(struct mmphw_path_plat *path_plat)
{
	if (!path_plat)
		return;

	mmp_unregister_path(path_plat->path);
}

static int mmphw_probe(struct platform_device *pdev)
{
	struct mmp_mach_plat_info *mi;
	struct resource *res;
	int ret, i, irq;
	struct mmphw_path_plat *path_plat;
	struct mmphw_ctrl *ctrl = NULL;

	/* get resources from platform data */
	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
	if (res == NULL) {
		dev_err(&pdev->dev, "%s: no IO memory defined\n", __func__);
		ret = -ENOENT;
		goto failed;
	}

	irq = platform_get_irq(pdev, 0);
	if (irq < 0) {
		ret = -ENOENT;
		goto failed;
	}

	/* get configs from platform data */
	mi = pdev->dev.platform_data;
	if (mi == NULL || !mi->path_num || !mi->paths) {
		dev_err(&pdev->dev, "%s: no platform data defined\n", __func__);
		ret = -EINVAL;
		goto failed;
	}

	/* allocate */
	ctrl = devm_kzalloc(&pdev->dev,
			    struct_size(ctrl, path_plats, mi->path_num),
			    GFP_KERNEL);
	if (!ctrl) {
		ret = -ENOMEM;
		goto failed;
	}

	ctrl->name = mi->name;
	ctrl->path_num = mi->path_num;
	ctrl->dev = &pdev->dev;
	ctrl->irq = irq;
	platform_set_drvdata(pdev, ctrl);
	mutex_init(&ctrl->access_ok);

	/* map registers.*/
	if (!devm_request_mem_region(ctrl->dev, res->start,
			resource_size(res), ctrl->name)) {
		dev_err(ctrl->dev,
			"can't request region for resource %pR\n", res);
		ret = -EINVAL;
		goto failed;
	}

	ctrl->reg_base = devm_ioremap(ctrl->dev,
			res->start, resource_size(res));
	if (ctrl->reg_base == NULL) {
		dev_err(ctrl->dev, "%s: res %pR map failed\n", __func__, res);
		ret = -ENOMEM;
		goto failed;
	}

	/* request irq */
	ret = devm_request_irq(ctrl->dev, ctrl->irq, ctrl_handle_irq,
		IRQF_SHARED, "lcd_controller", ctrl);
	if (ret < 0) {
		dev_err(ctrl->dev, "%s unable to request IRQ %d\n",
				__func__, ctrl->irq);
		ret = -ENXIO;
		goto failed;
	}

	/* get clock */
	ctrl->clk = devm_clk_get(ctrl->dev, mi->clk_name);
	if (IS_ERR(ctrl->clk)) {
		ret = PTR_ERR(ctrl->clk);
		dev_err_probe(ctrl->dev, ret,
			      "unable to get clk %s\n", mi->clk_name);
		goto failed;
	}
	ret = clk_prepare_enable(ctrl->clk);
	if (ret)
		goto failed;

	/* init global regs */
	ctrl_set_default(ctrl);

	/* init pathes from machine info and register them */
	for (i = 0; i < ctrl->path_num; i++) {
		/* get from config and machine info */
		path_plat = &ctrl->path_plats[i];
		path_plat->id = i;
		path_plat->ctrl = ctrl;

		/* path init */
		if (!path_init(path_plat, &mi->paths[i])) {
			ret = -EINVAL;
			goto failed_path_init;
		}
	}

#ifdef CONFIG_MMP_DISP_SPI
	ret = lcd_spi_register(ctrl);
	if (ret < 0)
		goto failed_path_init;
#endif

	dev_info(ctrl->dev, "device init done\n");

	return 0;

failed_path_init:
	for (i = 0; i < ctrl->path_num; i++) {
		path_plat = &ctrl->path_plats[i];
		path_deinit(path_plat);
	}

	clk_disable_unprepare(ctrl->clk);
failed:
	dev_err(&pdev->dev, "device init failed\n");

	return ret;
}

static struct platform_driver mmphw_driver = {
	.driver		= {
		.name	= "mmp-disp",
	},
	.probe		= mmphw_probe,
};

static int mmphw_init(void)
{
	return platform_driver_register(&mmphw_driver);
}
module_init(mmphw_init);

MODULE_AUTHOR("Li Guoqing<ligq@marvell.com>");
MODULE_DESCRIPTION("Framebuffer driver for mmp");
MODULE_LICENSE("GPL");