Skip to contentSkip to conversation

Add T114 video support

PATCH In review Svyatoslav Ryhel 3 participants 47 messages Jan 23, 2024 – Apr 20, 2024

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v6 cover Add T114 video support

T114 is not that different from T30 and all T30 drivers will work on T114 as well with some adjustments.

Patches propose general improvements for existing Tegra DC and DSI drivers as well Tegra 114 video support (experimentl).

Commits pass buildman for tegra.

---
Changes from v5:
- backlight enable moved to the last step of setup for DSI
- parameterized V- and H-sync polarities
- added framebuffer clearing on probe to avoid glitches
- backlight enable moved after DC is fully configured
- fixed printing framebuffer pointer instead of address
- moved scdiv calculation to tegra DSI if it is used

Changes from v4:
- fixed typo in max rate to be divided (400KHz > 400MHz)

Changes from v3:
- fixed the clock divider calculation if PLLD/D2 is used
- removed unnecessary pre-configuration
- set correct video FIFO depth for DSI

Changes from v2:
- fixed image distortion on devices with 180deg rotation

Changes from v1:
- reworked patchset entirely
- diverged DC configuration per-SOC
- consolidated dc headers from different SOC gen
- initial support of DC detection (tegra has 2 DC)
- added PLLD2 support, resets and powergating
- added T114+ MIPI calibration
- added DSI detection (tegra has 2 DSI) and resets
---

Jonas Schwöbel (6):
video: tegra20: dc: fix printing of framebuffer address
video: tegra20: dc: enable backlight after DC is configured
video: tegra20: dc: clean framebuffer memory block
video: tegra20: dsi: remove pre-configuration
video: tegra20: dsi: set correct fifo depth
video: tegra20: dsi: use set_backlight for backlight only

Svyatoslav Ryhel (12):
video: tegra20: dc: diverge DC per-SOC
video: tegra20: dc: fix image shift on rotated panels
video: tegra20: consolidate DC header
video: tegra20: dc: pass DC id to internal devices
video: tegra20: dc: add PLLD2 parent support
video: tegra20: dc: add reset support
video: tegra20: dc: add powergate
video: tegra20: dc: configure behavior if PLLD/D2 is used
video: tegra20: dc: parameterize V- and H-sync polarities
video: tegra20: add MIPI calibration driver
video: tegra20: dsi: add T114 support
video: tegra20: dsi: add reset support

 arch/arm/dts/tegra114-u-boot.dtsi             |  13 +
 arch/arm/dts/tegra114.dtsi                    |   4 +-
 arch/arm/dts/tegra30-u-boot.dtsi              |   4 +
 arch/arm/dts/tegra30.dtsi                     |   2 +-
 arch/arm/include/asm/arch-tegra/dc.h          |  13 +-
 arch/arm/include/asm/arch-tegra114/pwm.h      |  13 +
 arch/arm/include/asm/arch-tegra20/display.h   |  28 --
 arch/arm/include/asm/arch-tegra30/display.h   |  28 --
 drivers/video/tegra20/Makefile                |   2 +-
 drivers/video/tegra20/tegra-dc.c              | 239 +++++++++++++-----
 drivers/video/tegra20/tegra-dc.h              |  45 ++++
 drivers/video/tegra20/tegra-dsi.c             | 122 ++++++++-
 .../video/tegra20/tegra-dsi.h                 |  24 +-
 drivers/video/tegra20/tegra-mipi.c            | 188 ++++++++++++++
 drivers/video/tegra20/tegra-pwm-backlight.c   |   3 +-
 15 files changed, 583 insertions(+), 145 deletions(-)
 create mode 100644 arch/arm/include/asm/arch-tegra114/pwm.h
 delete mode 100644 arch/arm/include/asm/arch-tegra20/display.h
 delete mode 100644 arch/arm/include/asm/arch-tegra30/display.h
 create mode 100644 drivers/video/tegra20/tegra-dc.h
 rename arch/arm/include/asm/arch-tegra30/dsi.h => drivers/video/tegra20/tegra-dsi.h (90%)
 create mode 100644 drivers/video/tegra20/tegra-mipi.c
Signature · 1 line
2.40.1
#1 lore raw

PATCH v6 1/18 video: tegra20: dc: diverge DC per-SOC

Diverge DC driver setup to better fit each of supported generations of Tegra SOC.

---
 arch/arm/dts/tegra114-u-boot.dtsi            |  13 +++
 arch/arm/dts/tegra114.dtsi                   |   4 +-
 arch/arm/dts/tegra30-u-boot.dtsi             |   4 +
 arch/arm/dts/tegra30.dtsi                    |   2 +-
 arch/arm/include/asm/arch-tegra114/display.h |  28 +++++
 arch/arm/include/asm/arch-tegra114/pwm.h     |  13 +++
 drivers/video/tegra20/tegra-dc.c             | 107 +++++++++++++------
 7 files changed, 133 insertions(+), 38 deletions(-)
 create mode 100644 arch/arm/include/asm/arch-tegra114/display.h
 create mode 100644 arch/arm/include/asm/arch-tegra114/pwm.h
arch/arm/dts/tegra114-u-boot.dtsi +13 -0
index 7c11972552..6a02714a25 100644
@@ -1,3 +1,16 @@
 #include <config.h>
 
 #include "tegra-u-boot.dtsi"
+
+/ {
+	host1x@50000000 {
+		bootph-all;
+		dc@54200000 {
+			bootph-all;
+		};
+
+		dc@54240000 {
+			bootph-all;
+		};
+	};
+};
arch/arm/dts/tegra114.dtsi +2 -2
index 68ee7f3165..250d692f6b 100644
@@ -42,7 +42,7 @@
 		};
 
 		dc@54200000 {
-			compatible = "nvidia,tegra114-dc", "nvidia,tegra20-dc";
+			compatible = "nvidia,tegra114-dc";
 			reg = <0x54200000 0x00040000>;
 			interrupts = <GIC_SPI 73 IRQ_TYPE_LEVEL_HIGH>;
 			clocks = <&tegra_car TEGRA114_CLK_DISP1>,
@@ -61,7 +61,7 @@
 		};
 
 		dc@54240000 {
-			compatible = "nvidia,tegra114-dc", "nvidia,tegra20-dc";
+			compatible = "nvidia,tegra114-dc";
 			reg = <0x54240000 0x00040000>;
 			interrupts = <GIC_SPI 74 IRQ_TYPE_LEVEL_HIGH>;
 			clocks = <&tegra_car TEGRA114_CLK_DISP2>,
arch/arm/dts/tegra30-u-boot.dtsi +4 -0
index 3038227dbe..6a02714a25 100644
@@ -8,5 +8,9 @@
 		dc@54200000 {
 			bootph-all;
 		};
+
+		dc@54240000 {
+			bootph-all;
+		};
 	};
 };
arch/arm/dts/tegra30.dtsi +1 -1
index f198bc0edb..1177e2ab1f 100644
@@ -158,7 +158,7 @@
 		};
 
 		dc@54200000 {
-			compatible = "nvidia,tegra30-dc", "nvidia,tegra20-dc";
+			compatible = "nvidia,tegra30-dc";
 			reg = <0x54200000 0x00040000>;
 			interrupts = <GIC_SPI 73 IRQ_TYPE_LEVEL_HIGH>;
 			clocks = <&tegra_car TEGRA30_CLK_DISP1>,
arch/arm/include/asm/arch-tegra114/display.h new file +28 -0
new file mode 100644
index 0000000000..9411525799
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ *  (C) Copyright 2010
+ *  NVIDIA Corporation <www.nvidia.com>
+ */
+
+#ifndef __ASM_ARCH_TEGRA_DISPLAY_H
+#define __ASM_ARCH_TEGRA_DISPLAY_H
+
+#include <asm/arch-tegra/dc.h>
+
+/* This holds information about a window which can be displayed */
+struct disp_ctl_win {
+	enum win_color_depth_id fmt;	/* Color depth/format */
+	unsigned int	bpp;		/* Bits per pixel */
+	phys_addr_t	phys_addr;	/* Physical address in memory */
+	unsigned int	x;		/* Horizontal address offset (bytes) */
+	unsigned int	y;		/* Veritical address offset (bytes) */
+	unsigned int	w;		/* Width of source window */
+	unsigned int	h;		/* Height of source window */
+	unsigned int	stride;		/* Number of bytes per line */
+	unsigned int	out_x;		/* Left edge of output window (col) */
+	unsigned int	out_y;		/* Top edge of output window (row) */
+	unsigned int	out_w;		/* Width of output window in pixels */
+	unsigned int	out_h;		/* Height of output window in pixels */
+};
+
+#endif /*__ASM_ARCH_TEGRA_DISPLAY_H*/
arch/arm/include/asm/arch-tegra114/pwm.h new file +13 -0
new file mode 100644
index 0000000000..af39151803
@@ -0,0 +1,13 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Tegra pulse width frequency modulator definitions
+ *
+ * Copyright (c) 2011 The Chromium OS Authors.
+ */
+
+#ifndef __ASM_ARCH_TEGRA114_PWM_H
+#define __ASM_ARCH_TEGRA114_PWM_H
+
+#include <asm/arch-tegra/pwm.h>
+
+#endif	/* __ASM_ARCH_TEGRA114_PWM_H */
drivers/video/tegra20/tegra-dc.c +72 -35

219 lines not on this page · raw message

Signature · 1 line
2.40.1
#2 lore raw

Re: [PATCH v6 01/18] video: tegra20: dc: diverge DC per-SOC

On Tue Jan 23, 2024 at 6:16 PM CET, Svyatoslav Ryhel wrote:
[...]

Quoted 24 lines
arch/arm/include/asm/arch-tegra114/display.h new file +18 -0
new file mode 100644
index 0000000000..9411525799
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ *  (C) Copyright 2010
+ *  NVIDIA Corporation <www.nvidia.com>
+ */
+
+#ifndef __ASM_ARCH_TEGRA_DISPLAY_H
+#define __ASM_ARCH_TEGRA_DISPLAY_H
+
+#include <asm/arch-tegra/dc.h>
+
+/* This holds information about a window which can be displayed */
+struct disp_ctl_win {
+	enum win_color_depth_id fmt;	/* Color depth/format */
+	unsigned int	bpp;		/* Bits per pixel */
+	phys_addr_t	phys_addr;	/* Physical address in memory */
+	unsigned int	x;		/* Horizontal address offset (bytes) */
+	unsigned int	y;		/* Veritical address offset (bytes) */

arch/arm/include/asm/arch-tegra114/display.h line 28

Quoted 10 lines
arch/arm/include/asm/arch-tegra114/display.h new file +10 -0
+	unsigned int	w;		/* Width of source window */
+	unsigned int	h;		/* Height of source window */
+	unsigned int	stride;		/* Number of bytes per line */
+	unsigned int	out_x;		/* Left edge of output window (col) */
+	unsigned int	out_y;		/* Top edge of output window (row) */
+	unsigned int	out_w;		/* Width of output window in pixels */
+	unsigned int	out_h;		/* Height of output window in pixels */
+};
+
+#endif /*__ASM_ARCH_TEGRA_DISPLAY_H*/

One of the earlier patches in the series gets rid of this per-SoC header file in favor of a common one. Did this end up here by mistake? It doesn't seem to be used.

Quoted 95 lines
drivers/video/tegra20/tegra-dc.c +22 -14
index f53ad46397..7605e77bc1 100644
@@ -3,7 +3,6 @@
  * Copyright (c) 2011 The Chromium OS Authors.
  */
 
-#include <common.h>
 #include <backlight.h>
 #include <dm.h>
 #include <fdtdec.h>
@@ -23,10 +22,15 @@
 #include <asm/arch/pinmux.h>
 #include <asm/arch/pwm.h>
 #include <asm/arch/display.h>
-#include <asm/arch-tegra/timer.h>
 
 DECLARE_GLOBAL_DATA_PTR;
 
+/* Holder of Tegra per-SOC DC differences */
+struct tegra_dc_soc_info {
+	bool has_timer;
+	bool has_rgb;
+};
+
 /* Information about the display controller */
 struct tegra_lcd_priv {
 	int width;			/* width in pixels */
@@ -35,6 +39,7 @@ struct tegra_lcd_priv {
 	struct display_timing timing;
 	struct udevice *panel;
 	struct dc_ctlr *dc;		/* Display controller regmap */
+	const struct tegra_dc_soc_info *soc;
 	fdt_addr_t frame_buffer;	/* Address of frame buffer */
 	unsigned pixel_clock;		/* Pixel clock in Hz */
 	int dc_clk[2];			/* Contains clk and its parent */
@@ -43,8 +48,8 @@ struct tegra_lcd_priv {
 
 enum {
 	/* Maximum LCD size we support */
-	LCD_MAX_WIDTH		= 1920,
-	LCD_MAX_HEIGHT		= 1200,
+	LCD_MAX_WIDTH		= 2560,
+	LCD_MAX_HEIGHT		= 1600,
 	LCD_MAX_LOG2_BPP	= VIDEO_BPP16,
 };
 
@@ -110,9 +115,9 @@ static void update_window(struct tegra_lcd_priv *priv,
 	writel(val, &dc->cmd.state_ctrl);
 }
 
-static int update_display_mode(struct dc_disp_reg *disp,
-			       struct tegra_lcd_priv *priv)
+static int update_display_mode(struct tegra_lcd_priv *priv)
 {
+	struct dc_disp_reg *disp = &priv->dc->disp;
 	struct display_timing *dt = &priv->timing;
 	unsigned long val;
 	unsigned long rate;
@@ -128,14 +133,16 @@ static int update_display_mode(struct dc_disp_reg *disp,
 	       &disp->front_porch);
 	writel(dt->hactive.typ | (dt->vactive.typ << 16), &disp->disp_active);
 
-	val = DE_SELECT_ACTIVE << DE_SELECT_SHIFT;
-	val |= DE_CONTROL_NORMAL << DE_CONTROL_SHIFT;
-	writel(val, &disp->data_enable_opt);
+	if (priv->soc->has_rgb) {
+		val = DE_SELECT_ACTIVE << DE_SELECT_SHIFT;
+		val |= DE_CONTROL_NORMAL << DE_CONTROL_SHIFT;
+		writel(val, &disp->data_enable_opt);
 
-	val = DATA_FORMAT_DF1P1C << DATA_FORMAT_SHIFT;
-	val |= DATA_ALIGNMENT_MSB << DATA_ALIGNMENT_SHIFT;
-	val |= DATA_ORDER_RED_BLUE << DATA_ORDER_SHIFT;
-	writel(val, &disp->disp_interface_ctrl);
+		val = DATA_FORMAT_DF1P1C << DATA_FORMAT_SHIFT;
+		val |= DATA_ALIGNMENT_MSB << DATA_ALIGNMENT_SHIFT;
+		val |= DATA_ORDER_RED_BLUE << DATA_ORDER_SHIFT;
+		writel(val, &disp->disp_interface_ctrl);
+	}
 
 	/*
 	 * The pixel clock divider is in 7.1 format (where the bottom bit
@@ -147,7 +154,8 @@ static int update_display_mode(struct dc_disp_reg *disp,
 	div = ((rate * 2 + priv->pixel_clock / 2) / priv->pixel_clock) - 2;
 	debug("Display clock %lu, divider %lu\n", rate, div);
 
-	writel(0x00010001, &disp->shift_clk_opt);
+	if (priv->soc->has_rgb)
+		writel(0x00010001, &disp->shift_clk_opt);
 
 	val = PIXEL_CLK_DIVIDER_PCD1 << PIXEL_CLK_DIVIDER_SHIFT;
 	val |= div << SHIFT_CLK_DIVIDER_SHIFT;

drivers/video/tegra20/tegra-dc.c static void basic_init(struct dc_cmd_reg *cmd) · line 186

Quoted 5 lines
drivers/video/tegra20/tegra-dc.c +1 -0
@@ -174,6 +182,7 @@ static void basic_init(struct dc_cmd_reg *cmd)
 	writel(val, &cmd->disp_pow_ctrl);
 
 	val = readl(&cmd->disp_cmd);
+	val &= ~CTRL_MODE_MASK;
 	val |= CTRL_MODE_C_DISPLAY << CTRL_MODE_SHIFT;

This seems unrelated to the rest here, but probably not worth splitting it into a separate patch. I vaguely recall that this wasn't really necessary because we do the reset prior to initialization and the register is all zeroes by default?

Quoted 67 lines
drivers/video/tegra20/tegra-dc.c +12 -10
 	writel(val, &cmd->disp_cmd);
 }
@@ -229,8 +238,8 @@ static void rgb_enable(struct dc_com_reg *com)
 		writel(rgb_sel_tab[i], &com->pin_output_sel[i]);
 }
 
-static int setup_window(struct disp_ctl_win *win,
-			struct tegra_lcd_priv *priv)
+static int setup_window(struct tegra_lcd_priv *priv,
+			struct disp_ctl_win *win)
 {
 	if (priv->rotation) {
 		win->x = priv->width * 2;
@@ -274,12 +283,11 @@ static int setup_window(struct disp_ctl_win *win,
  * You should pass in the U-Boot address here, and check the contents of
  * struct tegra_lcd_priv to see what was actually chosen.
  *
- * @param blob			Device tree blob
  * @param priv			Driver's private data
  * @param default_lcd_base	Default address of LCD frame buffer
  * Return: 0 if ok, -1 on error (unsupported bits per pixel)
  */
-static int tegra_display_probe(const void *blob, struct tegra_lcd_priv *priv,
+static int tegra_display_probe(struct tegra_lcd_priv *priv,
 			       void *default_lcd_base)
 {
 	struct disp_ctl_win window;
@@ -288,7 +296,7 @@ static int tegra_display_probe(const void *blob, struct tegra_lcd_priv *priv,
 	priv->frame_buffer = (u32)default_lcd_base;
 
 	/*
-	 * We halve the rate if DISP1 paret is PLLD, since actual parent
+	 * We halve the rate if DISP1 parent is PLLD, since actual parent
 	 * is plld_out0 which is PLLD divided by 2.
 	 */
 	if (priv->dc_clk[1] == CLOCK_ID_DISPLAY)
@@ -303,13 +311,17 @@ static int tegra_display_probe(const void *blob, struct tegra_lcd_priv *priv,
 			       rate);
 
 	basic_init(&priv->dc->cmd);
-	basic_init_timer(&priv->dc->disp);
-	rgb_enable(&priv->dc->com);
+
+	if (priv->soc->has_timer)
+		basic_init_timer(&priv->dc->disp);
+
+	if (priv->soc->has_rgb)
+		rgb_enable(&priv->dc->com);
 
 	if (priv->pixel_clock)
-		update_display_mode(&priv->dc->disp, priv);
+		update_display_mode(priv);
 
-	if (setup_window(&window, priv))
+	if (setup_window(priv, &window))
 		return -1;
 
 	update_window(priv, &window);
@@ -322,7 +334,6 @@ static int tegra_lcd_probe(struct udevice *dev)
 	struct video_uc_plat *plat = dev_get_uclass_plat(dev);
 	struct video_priv *uc_priv = dev_get_uclass_priv(dev);
 	struct tegra_lcd_priv *priv = dev_get_priv(dev);
-	const void *blob = gd->fdt_blob;
 	int ret;
 
 	/* Initialize the Tegra display controller */

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_probe(struct udevice *dev) · line 345

Quoted 7 lines
drivers/video/tegra20/tegra-dc.c +2 -2
@@ -330,8 +341,8 @@ static int tegra_lcd_probe(struct udevice *dev)
 	funcmux_select(PERIPH_ID_DISP1, FUNCMUX_DEFAULT);
 #endif
 
-	if (tegra_display_probe(blob, priv, (void *)plat->base)) {
-		printf("%s: Failed to probe display driver\n", __func__);
+	if (tegra_display_probe(priv, (void *)plat->base)) {
+		debug("%s: Failed to probe display driver\n", __func__);

Shouldn't this remain a printf() to make it more visible in the logs what's going on? I guess people will notice anyway when the display doesn't turn on, but this is good information to know when troubleshooting.

Quoted 20 lines
drivers/video/tegra20/tegra-dc.c +6 -0
 		return -1;
 	}
 
@@ -383,6 +394,8 @@ static int tegra_lcd_of_to_plat(struct udevice *dev)
 		return -EINVAL;
 	}
 
+	priv->soc = (struct tegra_dc_soc_info *)dev_get_driver_data(dev);
+
 	ret = clock_decode_pair(dev, priv->dc_clk);
 	if (ret < 0) {
 		debug("%s: Cannot decode clocks for '%s' (ret = %d)\n",
@@ -464,19 +477,43 @@ static int tegra_lcd_bind(struct udevice *dev)
 static const struct video_ops tegra_lcd_ops = {
 };
 
+static const struct tegra_dc_soc_info tegra20_dc_soc_info = {
+	.has_timer = true,
+	.has_rgb = true,
+};

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_bind(struct udevice *dev) · line 493

Quoted 10 lines
drivers/video/tegra20/tegra-dc.c +10 -0
+
+static const struct tegra_dc_soc_info tegra30_dc_soc_info = {
+	.has_timer = false,
+	.has_rgb = true,
+};
+
+static const struct tegra_dc_soc_info tegra114_dc_soc_info = {
+	.has_timer = false,
+	.has_rgb = false,
+};

My recollection is that technically Tegra114 still supports RGB, but it ends up never being used on any of the platforms that I know of. On Linux we base the decision to initialize the corresponding registers on the status property of the "rgb" node of each display controller. Perhaps that's something that U-Boot could do as well? That would avoid programming these registers on Tegra20 and Tegra30 devices that don't use RGB.

Thierry

#3 lore raw

Re: [PATCH v6 01/18] video: tegra20: dc: diverge DC per-SOC

Quoted 32 lines · Thierry Reding

On Tue Jan 23, 2024 at 6:16 PM CET, Svyatoslav Ryhel wrote:
[...]

Quoted 30 lines
arch/arm/include/asm/arch-tegra114/display.h new file +24 -0
new file mode 100644
index 0000000000..9411525799
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ *  (C) Copyright 2010
+ *  NVIDIA Corporation <www.nvidia.com>
+ */
+
+#ifndef __ASM_ARCH_TEGRA_DISPLAY_H
+#define __ASM_ARCH_TEGRA_DISPLAY_H
+
+#include <asm/arch-tegra/dc.h>
+
+/* This holds information about a window which can be displayed */
+struct disp_ctl_win {
+	enum win_color_depth_id fmt;	/* Color depth/format */
+	unsigned int	bpp;		/* Bits per pixel */
+	phys_addr_t	phys_addr;	/* Physical address in memory */
+	unsigned int	x;		/* Horizontal address offset (bytes) */
+	unsigned int	y;		/* Veritical address offset (bytes) */
+	unsigned int	w;		/* Width of source window */
+	unsigned int	h;		/* Height of source window */
+	unsigned int	stride;		/* Number of bytes per line */
+	unsigned int	out_x;		/* Left edge of output window (col) */
+	unsigned int	out_y;		/* Top edge of output window (row) */
+	unsigned int	out_w;		/* Width of output window in pixels */
Quoted 8 lines

arch/arm/include/asm/arch-tegra114/display.h line 28

Quoted 4 lines
arch/arm/include/asm/arch-tegra114/display.h new file +4 -0
+	unsigned int	out_h;		/* Height of output window in pixels */
+};
+
+#endif /*__ASM_ARCH_TEGRA_DISPLAY_H*/

One of the earlier patches in the series gets rid of this per-SoC header file in favor of a common one. Did this end up here by mistake? It doesn't seem to be used.

Nevermind, my MUA sorted these patches weirdly, so it appeared as if this was later in the series than it actually was.

Thierry

#4 lore raw

Re: [PATCH v6 01/18] video: tegra20: dc: diverge DC per-SOC

In reply to Thierry Reding

пт, 19 квіт. 2024 р. о 19:26 Thierry Reding <[email protected]> пише:

Quoted 218 lines · Svyatoslav Ryhel

[...]

Quoted 24 lines
arch/arm/include/asm/arch-tegra114/display.h new file +18 -0
new file mode 100644
index 0000000000..9411525799
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ *  (C) Copyright 2010
+ *  NVIDIA Corporation <www.nvidia.com>
+ */
+
+#ifndef __ASM_ARCH_TEGRA_DISPLAY_H
+#define __ASM_ARCH_TEGRA_DISPLAY_H
+
+#include <asm/arch-tegra/dc.h>
+
+/* This holds information about a window which can be displayed */
+struct disp_ctl_win {
+     enum win_color_depth_id fmt;    /* Color depth/format */
+     unsigned int    bpp;            /* Bits per pixel */
+     phys_addr_t     phys_addr;      /* Physical address in memory */
+     unsigned int    x;              /* Horizontal address offset (bytes) */
+     unsigned int    y;              /* Veritical address offset (bytes) */

arch/arm/include/asm/arch-tegra114/display.h line 28

Quoted 10 lines
arch/arm/include/asm/arch-tegra114/display.h new file +10 -0
+     unsigned int    w;              /* Width of source window */
+     unsigned int    h;              /* Height of source window */
+     unsigned int    stride;         /* Number of bytes per line */
+     unsigned int    out_x;          /* Left edge of output window (col) */
+     unsigned int    out_y;          /* Top edge of output window (row) */
+     unsigned int    out_w;          /* Width of output window in pixels */
+     unsigned int    out_h;          /* Height of output window in pixels */
+};
+
+#endif /*__ASM_ARCH_TEGRA_DISPLAY_H*/

One of the earlier patches in the series gets rid of this per-SoC header file in favor of a common one. Did this end up here by mistake? It doesn't seem to be used.

Quoted 95 lines
drivers/video/tegra20/tegra-dc.c +22 -14
index f53ad46397..7605e77bc1 100644
@@ -3,7 +3,6 @@
  * Copyright (c) 2011 The Chromium OS Authors.
  */
 
-#include <common.h>
 #include <backlight.h>
 #include <dm.h>
 #include <fdtdec.h>
@@ -23,10 +22,15 @@
 #include <asm/arch/pinmux.h>
 #include <asm/arch/pwm.h>
 #include <asm/arch/display.h>
-#include <asm/arch-tegra/timer.h>
 
 DECLARE_GLOBAL_DATA_PTR;
 
+/* Holder of Tegra per-SOC DC differences */
+struct tegra_dc_soc_info {
+     bool has_timer;
+     bool has_rgb;
+};
+
 /* Information about the display controller */
 struct tegra_lcd_priv {
      int width;                      /* width in pixels */
@@ -35,6 +39,7 @@ struct tegra_lcd_priv {
      struct display_timing timing;
      struct udevice *panel;
      struct dc_ctlr *dc;             /* Display controller regmap */
+     const struct tegra_dc_soc_info *soc;
      fdt_addr_t frame_buffer;        /* Address of frame buffer */
      unsigned pixel_clock;           /* Pixel clock in Hz */
      int dc_clk[2];                  /* Contains clk and its parent */
@@ -43,8 +48,8 @@ struct tegra_lcd_priv {
 
 enum {
      /* Maximum LCD size we support */
-     LCD_MAX_WIDTH           = 1920,
-     LCD_MAX_HEIGHT          = 1200,
+     LCD_MAX_WIDTH           = 2560,
+     LCD_MAX_HEIGHT          = 1600,
      LCD_MAX_LOG2_BPP        = VIDEO_BPP16,
 };
 
@@ -110,9 +115,9 @@ static void update_window(struct tegra_lcd_priv *priv,
      writel(val, &dc->cmd.state_ctrl);
 }
 
-static int update_display_mode(struct dc_disp_reg *disp,
-                            struct tegra_lcd_priv *priv)
+static int update_display_mode(struct tegra_lcd_priv *priv)
 {
+     struct dc_disp_reg *disp = &priv->dc->disp;
      struct display_timing *dt = &priv->timing;
      unsigned long val;
      unsigned long rate;
@@ -128,14 +133,16 @@ static int update_display_mode(struct dc_disp_reg *disp,
             &disp->front_porch);
      writel(dt->hactive.typ | (dt->vactive.typ << 16), &disp->disp_active);
 
-     val = DE_SELECT_ACTIVE << DE_SELECT_SHIFT;
-     val |= DE_CONTROL_NORMAL << DE_CONTROL_SHIFT;
-     writel(val, &disp->data_enable_opt);
+     if (priv->soc->has_rgb) {
+             val = DE_SELECT_ACTIVE << DE_SELECT_SHIFT;
+             val |= DE_CONTROL_NORMAL << DE_CONTROL_SHIFT;
+             writel(val, &disp->data_enable_opt);
 
-     val = DATA_FORMAT_DF1P1C << DATA_FORMAT_SHIFT;
-     val |= DATA_ALIGNMENT_MSB << DATA_ALIGNMENT_SHIFT;
-     val |= DATA_ORDER_RED_BLUE << DATA_ORDER_SHIFT;
-     writel(val, &disp->disp_interface_ctrl);
+             val = DATA_FORMAT_DF1P1C << DATA_FORMAT_SHIFT;
+             val |= DATA_ALIGNMENT_MSB << DATA_ALIGNMENT_SHIFT;
+             val |= DATA_ORDER_RED_BLUE << DATA_ORDER_SHIFT;
+             writel(val, &disp->disp_interface_ctrl);
+     }
 
      /*
       * The pixel clock divider is in 7.1 format (where the bottom bit
@@ -147,7 +154,8 @@ static int update_display_mode(struct dc_disp_reg *disp,
      div = ((rate * 2 + priv->pixel_clock / 2) / priv->pixel_clock) - 2;
      debug("Display clock %lu, divider %lu\n", rate, div);
 
-     writel(0x00010001, &disp->shift_clk_opt);
+     if (priv->soc->has_rgb)
+             writel(0x00010001, &disp->shift_clk_opt);
 
      val = PIXEL_CLK_DIVIDER_PCD1 << PIXEL_CLK_DIVIDER_SHIFT;
      val |= div << SHIFT_CLK_DIVIDER_SHIFT;

drivers/video/tegra20/tegra-dc.c static void basic_init(struct dc_cmd_reg *cmd) · line 186

Quoted 5 lines
drivers/video/tegra20/tegra-dc.c +1 -0
@@ -174,6 +182,7 @@ static void basic_init(struct dc_cmd_reg *cmd)
      writel(val, &cmd->disp_pow_ctrl);
 
      val = readl(&cmd->disp_cmd);
+     val &= ~CTRL_MODE_MASK;
      val |= CTRL_MODE_C_DISPLAY << CTRL_MODE_SHIFT;

This seems unrelated to the rest here, but probably not worth splitting it into a separate patch. I vaguely recall that this wasn't really necessary because we do the reset prior to initialization and the register is all zeroes by default?

Quoted 70 lines
drivers/video/tegra20/tegra-dc.c +12 -10
      writel(val, &cmd->disp_cmd);
 }
@@ -229,8 +238,8 @@ static void rgb_enable(struct dc_com_reg *com)
              writel(rgb_sel_tab[i], &com->pin_output_sel[i]);
 }
 
-static int setup_window(struct disp_ctl_win *win,
-                     struct tegra_lcd_priv *priv)
+static int setup_window(struct tegra_lcd_priv *priv,
+                     struct disp_ctl_win *win)
 {
      if (priv->rotation) {
              win->x = priv->width * 2;
@@ -274,12 +283,11 @@ static int setup_window(struct disp_ctl_win *win,
  * You should pass in the U-Boot address here, and check the contents of
  * struct tegra_lcd_priv to see what was actually chosen.
  *
- * @param blob                       Device tree blob
  * @param priv                       Driver's private data
  * @param default_lcd_base   Default address of LCD frame buffer
  * Return: 0 if ok, -1 on error (unsupported bits per pixel)
  */
-static int tegra_display_probe(const void *blob, struct tegra_lcd_priv *priv,
+static int tegra_display_probe(struct tegra_lcd_priv *priv,
                             void *default_lcd_base)
 {
      struct disp_ctl_win window;
@@ -288,7 +296,7 @@ static int tegra_display_probe(const void *blob, struct tegra_lcd_priv *priv,
      priv->frame_buffer = (u32)default_lcd_base;
 
      /*
-      * We halve the rate if DISP1 paret is PLLD, since actual parent
+      * We halve the rate if DISP1 parent is PLLD, since actual parent
       * is plld_out0 which is PLLD divided by 2.
       */
      if (priv->dc_clk[1] == CLOCK_ID_DISPLAY)
@@ -303,13 +311,17 @@ static int tegra_display_probe(const void *blob, struct tegra_lcd_priv *priv,
                             rate);
 
      basic_init(&priv->dc->cmd);
-     basic_init_timer(&priv->dc->disp);
-     rgb_enable(&priv->dc->com);
+
+     if (priv->soc->has_timer)
+             basic_init_timer(&priv->dc->disp);
+
+     if (priv->soc->has_rgb)
+             rgb_enable(&priv->dc->com);
 
      if (priv->pixel_clock)
-             update_display_mode(&priv->dc->disp, priv);
+             update_display_mode(priv);
 
-     if (setup_window(&window, priv))
+     if (setup_window(priv, &window))
              return -1;
 
      update_window(priv, &window);
@@ -322,7 +334,6 @@ static int tegra_lcd_probe(struct udevice *dev)
      struct video_uc_plat *plat = dev_get_uclass_plat(dev);
      struct video_priv *uc_priv = dev_get_uclass_priv(dev);
      struct tegra_lcd_priv *priv = dev_get_priv(dev);
-     const void *blob = gd->fdt_blob;
      int ret;
 
      /* Initialize the Tegra display controller */
@@ -330,8 +341,8 @@ static int tegra_lcd_probe(struct udevice *dev)
      funcmux_select(PERIPH_ID_DISP1, FUNCMUX_DEFAULT);
 #endif
 
Quoted 10 lines

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_probe(struct udevice *dev) · line 345

Quoted 4 lines
drivers/video/tegra20/tegra-dc.c +2 -2
-     if (tegra_display_probe(blob, priv, (void *)plat->base)) {
-             printf("%s: Failed to probe display driver\n", __func__);
+     if (tegra_display_probe(priv, (void *)plat->base)) {
+             debug("%s: Failed to probe display driver\n", __func__);

Shouldn't this remain a printf() to make it more visible in the logs what's going on? I guess people will notice anyway when the display doesn't turn on, but this is good information to know when troubleshooting.

debug() is preferred to reduce resulting image size.

Quoted 31 lines
Quoted 20 lines
drivers/video/tegra20/tegra-dc.c +6 -0
              return -1;
      }
 
@@ -383,6 +394,8 @@ static int tegra_lcd_of_to_plat(struct udevice *dev)
              return -EINVAL;
      }
 
+     priv->soc = (struct tegra_dc_soc_info *)dev_get_driver_data(dev);
+
      ret = clock_decode_pair(dev, priv->dc_clk);
      if (ret < 0) {
              debug("%s: Cannot decode clocks for '%s' (ret = %d)\n",
@@ -464,19 +477,43 @@ static int tegra_lcd_bind(struct udevice *dev)
 static const struct video_ops tegra_lcd_ops = {
 };
 
+static const struct tegra_dc_soc_info tegra20_dc_soc_info = {
+     .has_timer = true,
+     .has_rgb = true,
+};

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_bind(struct udevice *dev) · line 493

Quoted 10 lines
drivers/video/tegra20/tegra-dc.c +10 -0
+
+static const struct tegra_dc_soc_info tegra30_dc_soc_info = {
+     .has_timer = false,
+     .has_rgb = true,
+};
+
+static const struct tegra_dc_soc_info tegra114_dc_soc_info = {
+     .has_timer = false,
+     .has_rgb = false,
+};
Quoted 8 lines

My recollection is that technically Tegra114 still supports RGB, but it ends up never being used on any of the platforms that I know of. On Linux we base the decision to initialize the corresponding registers on the status property of the "rgb" node of each display controller. Perhaps that's something that U-Boot could do as well? That would avoid programming these registers on Tegra20 and Tegra30 devices that don't use RGB.

This may be implemented in follow up. Please keep in mind that video subsystem existing in U-Boot is much simpler then Linux framework and some stuff may not be easy to implement right away. Thanks.

Quoted 1 line

Thierry

#5 lore raw

PATCH v6 2/18 video: tegra20: dc: fix image shift on rotated panels

In reply to Svyatoslav Ryhel

Subtracting 1 from x and y fixes image shifting on rotated panels.

---
 drivers/video/tegra20/tegra-dc.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/video/tegra20/tegra-dc.c +2 -2
index 7605e77bc1..bcc2748623 100644
@@ -242,8 +242,8 @@ static int setup_window(struct tegra_lcd_priv *priv,
 			struct disp_ctl_win *win)
 {
 	if (priv->rotation) {
-		win->x = priv->width * 2;
-		win->y = priv->height;
+		win->x = priv->width * 2 - 1;
+		win->y = priv->height - 1;
 	} else {
 		win->x = 0;
 		win->y = 0;
Signature · 1 line
2.40.1
Reviewed-by on v6 2/18
#6 lore raw

Re: [PATCH v6 02/18] video: tegra20: dc: fix image shift on rotated panels

Quoted 8 lines · Svyatoslav Ryhel

Subtracting 1 from x and y fixes image shifting on rotated panels.

---
 drivers/video/tegra20/tegra-dc.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Yeah, looks like we have that same fixup in Linux:

#7 lore raw

PATCH v6 3/18 video: tegra20: consolidate DC header

In reply to Svyatoslav Ryhel

Consolidate HD headers and place the result into video/tegra20 since it is used only by devices from this directory.

---
 arch/arm/include/asm/arch-tegra/dc.h         |  8 ----
 arch/arm/include/asm/arch-tegra114/display.h | 28 -------------
 arch/arm/include/asm/arch-tegra20/display.h  | 28 -------------
 arch/arm/include/asm/arch-tegra30/display.h  | 28 -------------
 drivers/video/tegra20/tegra-dc.c             |  3 +-
 drivers/video/tegra20/tegra-dc.h             | 41 ++++++++++++++++++++
 drivers/video/tegra20/tegra-dsi.c            |  2 +-
 drivers/video/tegra20/tegra-pwm-backlight.c  |  3 +-
 8 files changed, 46 insertions(+), 95 deletions(-)
 delete mode 100644 arch/arm/include/asm/arch-tegra114/display.h
 delete mode 100644 arch/arm/include/asm/arch-tegra20/display.h
 delete mode 100644 arch/arm/include/asm/arch-tegra30/display.h
 create mode 100644 drivers/video/tegra20/tegra-dc.h
arch/arm/include/asm/arch-tegra/dc.h +0 -8
index 7613d84f22..6444af2993 100644
@@ -569,12 +569,4 @@ enum {
 #define DC_N_WINDOWS			5
 #define DC_REG_SAVE_SPACE		(DC_N_WINDOWS + 5)
 
-#define TEGRA_DSI_A		"dsi@54300000"
-#define TEGRA_DSI_B		"dsi@54400000"
-
-struct tegra_dc_plat {
-	struct udevice *dev;		/* Display controller device */
-	struct dc_ctlr *dc;		/* Display controller regmap */
-};
-
 #endif /* __ASM_ARCH_TEGRA_DC_H */
arch/arm/include/asm/arch-tegra114/display.h deleted +0 -28
deleted file mode 100644
index 9411525799..0000000000
@@ -1,28 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0+ */
-/*
- *  (C) Copyright 2010
- *  NVIDIA Corporation <www.nvidia.com>
- */
-
-#ifndef __ASM_ARCH_TEGRA_DISPLAY_H
-#define __ASM_ARCH_TEGRA_DISPLAY_H
-
-#include <asm/arch-tegra/dc.h>
-
-/* This holds information about a window which can be displayed */
-struct disp_ctl_win {
-	enum win_color_depth_id fmt;	/* Color depth/format */
-	unsigned int	bpp;		/* Bits per pixel */
-	phys_addr_t	phys_addr;	/* Physical address in memory */
-	unsigned int	x;		/* Horizontal address offset (bytes) */
-	unsigned int	y;		/* Veritical address offset (bytes) */
-	unsigned int	w;		/* Width of source window */
-	unsigned int	h;		/* Height of source window */
-	unsigned int	stride;		/* Number of bytes per line */
-	unsigned int	out_x;		/* Left edge of output window (col) */
-	unsigned int	out_y;		/* Top edge of output window (row) */
-	unsigned int	out_w;		/* Width of output window in pixels */
-	unsigned int	out_h;		/* Height of output window in pixels */
-};
-
-#endif /*__ASM_ARCH_TEGRA_DISPLAY_H*/
arch/arm/include/asm/arch-tegra20/display.h deleted +0 -28
deleted file mode 100644
index e7b3cffd46..0000000000
@@ -1,28 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0+ */
-/*
- *  (C) Copyright 2010
- *  NVIDIA Corporation <www.nvidia.com>
- */
-
-#ifndef __ASM_ARCH_TEGRA_DISPLAY_H
-#define __ASM_ARCH_TEGRA_DISPLAY_H
-
-#include <asm/arch-tegra/dc.h>
-
-/* This holds information about a window which can be displayed */
-struct disp_ctl_win {
-	enum win_color_depth_id fmt;	/* Color depth/format */
-	unsigned	bpp;		/* Bits per pixel */
-	phys_addr_t	phys_addr;	/* Physical address in memory */
-	unsigned	x;		/* Horizontal address offset (bytes) */
-	unsigned	y;		/* Veritical address offset (bytes) */
-	unsigned	w;		/* Width of source window */
-	unsigned	h;		/* Height of source window */
-	unsigned	stride;		/* Number of bytes per line */
-	unsigned	out_x;		/* Left edge of output window (col) */
-	unsigned	out_y;		/* Top edge of output window (row) */
-	unsigned	out_w;		/* Width of output window in pixels */
-	unsigned	out_h;		/* Height of output window in pixels */
-};
-
-#endif /*__ASM_ARCH_TEGRA_DISPLAY_H*/
arch/arm/include/asm/arch-tegra30/display.h deleted +0 -28
deleted file mode 100644
index 9411525799..0000000000
@@ -1,28 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0+ */
-/*
- *  (C) Copyright 2010
- *  NVIDIA Corporation <www.nvidia.com>
- */
-
-#ifndef __ASM_ARCH_TEGRA_DISPLAY_H
-#define __ASM_ARCH_TEGRA_DISPLAY_H
-
-#include <asm/arch-tegra/dc.h>
-
-/* This holds information about a window which can be displayed */
-struct disp_ctl_win {
-	enum win_color_depth_id fmt;	/* Color depth/format */
-	unsigned int	bpp;		/* Bits per pixel */
-	phys_addr_t	phys_addr;	/* Physical address in memory */
-	unsigned int	x;		/* Horizontal address offset (bytes) */
-	unsigned int	y;		/* Veritical address offset (bytes) */
-	unsigned int	w;		/* Width of source window */
-	unsigned int	h;		/* Height of source window */
-	unsigned int	stride;		/* Number of bytes per line */
-	unsigned int	out_x;		/* Left edge of output window (col) */
-	unsigned int	out_y;		/* Top edge of output window (row) */
-	unsigned int	out_w;		/* Width of output window in pixels */
-	unsigned int	out_h;		/* Height of output window in pixels */
-};
-
-#endif /*__ASM_ARCH_TEGRA_DISPLAY_H*/
drivers/video/tegra20/tegra-dc.c +2 -1
index bcc2748623..5d8874f323 100644
@@ -21,7 +21,8 @@
 #include <asm/arch/funcmux.h>
 #include <asm/arch/pinmux.h>
 #include <asm/arch/pwm.h>
-#include <asm/arch/display.h>
+
+#include "tegra-dc.h"
 
 DECLARE_GLOBAL_DATA_PTR;
 
drivers/video/tegra20/tegra-dc.h new file +41 -0

41 lines not on this page · raw message

drivers/video/tegra20/tegra-dsi.c +1 -1
index a48f9c85d0..72b91ed26b 100644
@@ -20,9 +20,9 @@
 #include <asm/gpio.h>
 #include <asm/io.h>
 #include <asm/arch/clock.h>
-#include <asm/arch/display.h>
 #include <asm/arch-tegra30/dsi.h>
 
+#include "tegra-dc.h"
 #include "mipi-phy.h"
 
 struct tegra_dsi_priv {
drivers/video/tegra20/tegra-pwm-backlight.c +2 -1
index bb677daa8a..5f93f57fe9 100644
@@ -15,7 +15,8 @@
 
 #include <asm/io.h>
 #include <asm/gpio.h>
-#include <asm/arch/display.h>
+
+#include "tegra-dc.h"
 
 #define TEGRA_DISPLAY_A_BASE		0x54200000
 #define TEGRA_DISPLAY_B_BASE		0x54240000
Signature · 1 line
2.40.1
Reviewed-by on v6 3/18
#8 lore raw

Re: [PATCH v6 03/18] video: tegra20: consolidate DC header

Quoted 1 line · Svyatoslav Ryhel

Consolidate HD headers and place the result into video/tegra20

Was this supposed to be "DC" headers like in the subject?

Other than that this makes sense, so:

#9 lore raw

PATCH v6 4/18 video: tegra20: dc: pass DC id to internal devices

In reply to Svyatoslav Ryhel

Tegra SoC has 2 independent display controllers called DC_A and DC_B, they are handled differently by internal video devices like DSI and HDMI controllers so it is important for last to know which display controller is used to properly set up registers. To achieve this, a pipe field was added to pdata to pass display controller id to internal Tegra SoC devices.

---
 drivers/video/tegra20/tegra-dc.c | 6 ++++++
 drivers/video/tegra20/tegra-dc.h | 3 +++
 2 files changed, 9 insertions(+)
drivers/video/tegra20/tegra-dc.c +6 -0
index 5d8874f323..0e94e665ef 100644
@@ -45,6 +45,7 @@ struct tegra_lcd_priv {
 	unsigned pixel_clock;		/* Pixel clock in Hz */
 	int dc_clk[2];			/* Contains clk and its parent */
 	bool rotation;			/* 180 degree panel turn */
+	bool pipe;			/* DC controller: 0 for A, 1 for B */
 };
 
 enum {
@@ -406,6 +407,9 @@ static int tegra_lcd_of_to_plat(struct udevice *dev)
 
 	priv->rotation = dev_read_bool(dev, "nvidia,180-rotation");
 
+	if (!strcmp(dev->name, TEGRA_DC_B))
+		priv->pipe = 1;
+
 	rgb = fdt_subnode_offset(blob, node, "rgb");
 	if (rgb < 0) {
 		debug("%s: Cannot find rgb subnode for '%s' (ret=%d)\n",
@@ -431,12 +435,14 @@ static int tegra_lcd_of_to_plat(struct udevice *dev)
 		return ret;
 	}
 
+	/* Fill the platform data for internal devices */
 	if (!strcmp(priv->panel->name, TEGRA_DSI_A) ||
 	    !strcmp(priv->panel->name, TEGRA_DSI_B)) {
 		struct tegra_dc_plat *dc_plat = dev_get_plat(priv->panel);
 
 		dc_plat->dev = dev;
 		dc_plat->dc = priv->dc;
+		dc_plat->pipe = priv->pipe;
 	}
 
 	ret = panel_get_display_timing(priv->panel, &priv->timing);
drivers/video/tegra20/tegra-dc.h +3 -0
index 5c05221038..75fc0fa4de 100644
@@ -14,12 +14,15 @@
 /* arch-tegra/dc exists only because T124 uses it */
 #include <asm/arch-tegra/dc.h>
 
+#define TEGRA_DC_A		"dc@54200000"
+#define TEGRA_DC_B		"dc@54240000"
 #define TEGRA_DSI_A		"dsi@54300000"
 #define TEGRA_DSI_B		"dsi@54400000"
 
 struct tegra_dc_plat {
 	struct udevice *dev;		/* Display controller device */
 	struct dc_ctlr *dc;		/* Display controller regmap */
+	bool pipe;			/* DC number: 0 for A, 1 for B */
 };
 
 /* This holds information about a window which can be displayed */
Signature · 1 line
2.40.1
#10 lore raw

Re: [PATCH v6 04/18] video: tegra20: dc: pass DC id to internal devices

Quoted 23 lines · Svyatoslav Ryhel

Tegra SoC has 2 independent display controllers called DC_A and DC_B, they are handled differently by internal video devices like DSI and HDMI controllers so it is important for last to know which display controller is used to properly set up registers. To achieve this, a pipe field was added to pdata to pass display controller id to internal Tegra SoC devices.

---
 drivers/video/tegra20/tegra-dc.c | 6 ++++++
 drivers/video/tegra20/tegra-dc.h | 3 +++
 2 files changed, 9 insertions(+)
drivers/video/tegra20/tegra-dc.c +0 -0
index 5d8874f323..0e94e665ef 100644

drivers/video/tegra20/tegra-dc.c struct tegra_lcd_priv { · line 48

Quoted 4 lines
drivers/video/tegra20/tegra-dc.c +1 -0
@@ -45,6 +45,7 @@ struct tegra_lcd_priv {
 	unsigned pixel_clock;		/* Pixel clock in Hz */
 	int dc_clk[2];			/* Contains clk and its parent */
 	bool rotation;			/* 180 degree panel turn */
+	bool pipe;			/* DC controller: 0 for A, 1 for B */

Bool is a poor choice, even if there's only two of them. This is a proper index, so it should be some sort of integer.

Also, the device tree bindings for the display controller specify a "nvidia,head" property that can be used to identify these. If you add that to the U-Boot DT you can avoid looking up by name to map this value.

Thierry

#11 lore raw

Re: [PATCH v6 04/18] video: tegra20: dc: pass DC id to internal devices

пт, 19 квіт. 2024 р. о 19:38 Thierry Reding <[email protected]> пише:

Quoted 33 lines · Svyatoslav Ryhel
Quoted 23 lines

Tegra SoC has 2 independent display controllers called DC_A and DC_B, they are handled differently by internal video devices like DSI and HDMI controllers so it is important for last to know which display controller is used to properly set up registers. To achieve this, a pipe field was added to pdata to pass display controller id to internal Tegra SoC devices.

---
 drivers/video/tegra20/tegra-dc.c | 6 ++++++
 drivers/video/tegra20/tegra-dc.h | 3 +++
 2 files changed, 9 insertions(+)
drivers/video/tegra20/tegra-dc.c +0 -0
index 5d8874f323..0e94e665ef 100644

drivers/video/tegra20/tegra-dc.c struct tegra_lcd_priv { · line 48

Quoted 4 lines
drivers/video/tegra20/tegra-dc.c +1 -0
@@ -45,6 +45,7 @@ struct tegra_lcd_priv {
      unsigned pixel_clock;           /* Pixel clock in Hz */
      int dc_clk[2];                  /* Contains clk and its parent */
      bool rotation;                  /* 180 degree panel turn */
+     bool pipe;                      /* DC controller: 0 for A, 1 for B */

Bool is a poor choice, even if there's only two of them. This is a proper index, so it should be some sort of integer.

Quoted 5 lines

Also, the device tree bindings for the display controller specify a "nvidia,head" property that can be used to identify these. If you add that to the U-Boot DT you can avoid looking up by name to map this value.

Thanks for pointing to this property. May we apply this patch set as is since it is well tested and confirmed to work and I will prepare a follow up patches to adjust device tree relations? Would what be ok?

Quoted 1 line

Thierry

#12 lore raw

Re: [PATCH v6 04/18] video: tegra20: dc: pass DC id to internal devices

Quoted 40 lines · Svyatoslav Ryhel

пт, 19 квіт. 2024 р. о 19:38 Thierry Reding <[email protected]> пише:

Quoted 33 lines · Svyatoslav Ryhel
Quoted 23 lines

Tegra SoC has 2 independent display controllers called DC_A and DC_B, they are handled differently by internal video devices like DSI and HDMI controllers so it is important for last to know which display controller is used to properly set up registers. To achieve this, a pipe field was added to pdata to pass display controller id to internal Tegra SoC devices.

---
 drivers/video/tegra20/tegra-dc.c | 6 ++++++
 drivers/video/tegra20/tegra-dc.h | 3 +++
 2 files changed, 9 insertions(+)
drivers/video/tegra20/tegra-dc.c +0 -0
index 5d8874f323..0e94e665ef 100644

drivers/video/tegra20/tegra-dc.c struct tegra_lcd_priv { · line 48

Quoted 4 lines
drivers/video/tegra20/tegra-dc.c +1 -0
@@ -45,6 +45,7 @@ struct tegra_lcd_priv {
      unsigned pixel_clock;           /* Pixel clock in Hz */
      int dc_clk[2];                  /* Contains clk and its parent */
      bool rotation;                  /* 180 degree panel turn */
+     bool pipe;                      /* DC controller: 0 for A, 1 for B */

Bool is a poor choice, even if there's only two of them. This is a proper index, so it should be some sort of integer.

Quoted 5 lines

Also, the device tree bindings for the display controller specify a "nvidia,head" property that can be used to identify these. If you add that to the U-Boot DT you can avoid looking up by name to map this value.

Quoted 3 lines

Thanks for pointing to this property. May we apply this patch set as is since it is well tested and confirmed to work and I will prepare a follow up patches to adjust device tree relations? Would what be ok?

Well, there's a few other things that I think should be addressed, but if you'd like to keep this one patch as-is and clean this up later, I guess that's fine.

Thierry

#13 lore raw

Re: [PATCH v6 04/18] video: tegra20: dc: pass DC id to internal devices

пт, 19 квіт. 2024 р. о 19:58 Thierry Reding <[email protected]> пише:

Quoted 42 lines · Svyatoslav Ryhel
Quoted 40 lines

пт, 19 квіт. 2024 р. о 19:38 Thierry Reding <[email protected]> пише:

Quoted 38 lines · Svyatoslav Ryhel
Quoted 23 lines

Tegra SoC has 2 independent display controllers called DC_A and DC_B, they are handled differently by internal video devices like DSI and HDMI controllers so it is important for last to know which display controller is used to properly set up registers. To achieve this, a pipe field was added to pdata to pass display controller id to internal Tegra SoC devices.

---
 drivers/video/tegra20/tegra-dc.c | 6 ++++++
 drivers/video/tegra20/tegra-dc.h | 3 +++
 2 files changed, 9 insertions(+)
drivers/video/tegra20/tegra-dc.c +0 -0
index 5d8874f323..0e94e665ef 100644

drivers/video/tegra20/tegra-dc.c struct tegra_lcd_priv { · line 48

Quoted 4 lines
drivers/video/tegra20/tegra-dc.c +1 -0
@@ -45,6 +45,7 @@ struct tegra_lcd_priv {
      unsigned pixel_clock;           /* Pixel clock in Hz */
      int dc_clk[2];                  /* Contains clk and its parent */
      bool rotation;                  /* 180 degree panel turn */
+     bool pipe;                      /* DC controller: 0 for A, 1 for B */

Bool is a poor choice, even if there's only two of them. This is a proper index, so it should be some sort of integer.

Also, the device tree bindings for the display controller specify a "nvidia,head" property that can be used to identify these. If you add that to the U-Boot DT you can avoid looking up by name to map this value.

Quoted 7 lines
Quoted 3 lines

Thanks for pointing to this property. May we apply this patch set as is since it is well tested and confirmed to work and I will prepare a follow up patches to adjust device tree relations? Would what be ok?

Well, there's a few other things that I think should be addressed, but if you'd like to keep this one patch as-is and clean this up later, I guess that's fine.

May you send me a list of stuff which you think may be improved? I would gladly adjust whatever I can in follow up.

Quoted 1 line

Thierry

#14 lore raw

PATCH v6 5/18 video: tegra20: dc: add PLLD2 parent support

In reply to Svyatoslav Ryhel

T30+ SOC have second PLLD - PLLD2 which can be actively used by DC and act as main DISP1/2 clock parent.

---
 drivers/video/tegra20/tegra-dc.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/video/tegra20/tegra-dc.c +6 -0
index 0e94e665ef..56a23b3c97 100644
@@ -304,6 +304,12 @@ static int tegra_display_probe(struct tegra_lcd_priv *priv,
 	if (priv->dc_clk[1] == CLOCK_ID_DISPLAY)
 		rate /= 2;
 
+#ifndef CONFIG_TEGRA20
+	/* PLLD2 obeys same rules as PLLD but it is present only on T30+ */
+	if (priv->dc_clk[1] == CLOCK_ID_DISPLAY2)
+		rate /= 2;
+#endif
+
 	/*
 	 * HOST1X is init by default at 150MHz with PLLC as parent
 	 */
Signature · 1 line
2.40.1
Reviewed-by on v6 5/18
#15 lore raw

Re: [PATCH v6 05/18] video: tegra20: dc: add PLLD2 parent support

Quoted 9 lines · Svyatoslav Ryhel

T30+ SOC have second PLLD - PLLD2 which can be actively used by DC and act as main DISP1/2 clock parent.

Quoted 3 lines
---
 drivers/video/tegra20/tegra-dc.c | 6 ++++++
 1 file changed, 6 insertions(+)
#16 lore raw

PATCH v6 6/18 video: tegra20: dc: add reset support

In reply to Svyatoslav Ryhel

Implement reset use to discard any changes which could have been applied to DC before and can interfere with current configuration.

---
 drivers/video/tegra20/tegra-dc.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
drivers/video/tegra20/tegra-dc.c +17 -0
index 56a23b3c97..35abb6fe46 100644
@@ -10,7 +10,9 @@
 #include <panel.h>
 #include <part.h>
 #include <pwm.h>
+#include <reset.h>
 #include <video.h>
+#include <linux/delay.h>
 #include <asm/cache.h>
 #include <asm/global_data.h>
 #include <asm/system.h>
@@ -342,6 +344,7 @@ static int tegra_lcd_probe(struct udevice *dev)
 	struct video_uc_plat *plat = dev_get_uclass_plat(dev);
 	struct video_priv *uc_priv = dev_get_uclass_priv(dev);
 	struct tegra_lcd_priv *priv = dev_get_priv(dev);
+	struct reset_ctl reset_ctl;
 	int ret;
 
 	/* Initialize the Tegra display controller */
@@ -349,6 +352,20 @@ static int tegra_lcd_probe(struct udevice *dev)
 	funcmux_select(PERIPH_ID_DISP1, FUNCMUX_DEFAULT);
 #endif
 
+	ret = reset_get_by_name(dev, "dc", &reset_ctl);
+	if (ret) {
+		log_err("reset_get_by_name() failed: %d\n", ret);
+		return ret;
+	}
+
+	clock_disable(priv->dc_clk[0]);
+
+	/* Reset everything set before */
+	reset_assert(&reset_ctl);
+	mdelay(4);
+	reset_deassert(&reset_ctl);
+	mdelay(4);
+
 	if (tegra_display_probe(priv, (void *)plat->base)) {
 		debug("%s: Failed to probe display driver\n", __func__);
 		return -1;
Signature · 1 line
2.40.1
#17 lore raw

Re: [PATCH v6 06/18] video: tegra20: dc: add reset support

Quoted 42 lines · Svyatoslav Ryhel

Implement reset use to discard any changes which could have been applied to DC before and can interfere with current configuration.

---
 drivers/video/tegra20/tegra-dc.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
drivers/video/tegra20/tegra-dc.c +6 -0
index 56a23b3c97..35abb6fe46 100644
@@ -10,7 +10,9 @@
 #include <panel.h>
 #include <part.h>
 #include <pwm.h>
+#include <reset.h>
 #include <video.h>
+#include <linux/delay.h>
 #include <asm/cache.h>
 #include <asm/global_data.h>
 #include <asm/system.h>
@@ -342,6 +344,7 @@ static int tegra_lcd_probe(struct udevice *dev)
 	struct video_uc_plat *plat = dev_get_uclass_plat(dev);
 	struct video_priv *uc_priv = dev_get_uclass_priv(dev);
 	struct tegra_lcd_priv *priv = dev_get_priv(dev);
+	struct reset_ctl reset_ctl;
 	int ret;
 
 	/* Initialize the Tegra display controller */
@@ -349,6 +352,20 @@ static int tegra_lcd_probe(struct udevice *dev)
 	funcmux_select(PERIPH_ID_DISP1, FUNCMUX_DEFAULT);
 #endif
 
+	ret = reset_get_by_name(dev, "dc", &reset_ctl);
+	if (ret) {
+		log_err("reset_get_by_name() failed: %d\n", ret);

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_probe(struct udevice *dev) · line 367

Quoted 10 lines
drivers/video/tegra20/tegra-dc.c +10 -0
+		return ret;
+	}
+
+	clock_disable(priv->dc_clk[0]);
+
+	/* Reset everything set before */
+	reset_assert(&reset_ctl);
+	mdelay(4);
+	reset_deassert(&reset_ctl);
+	mdelay(4);

Are you sure this works as intended? It's been a long time since I worked on this, but I seem to recall that most of these resets are actually synchronous, so in order for them to do what they're supposed to the clock needs to be kept running.

The Linux driver certainly does this differently.

Thierry

#18 lore raw

Re: [PATCH v6 06/18] video: tegra20: dc: add reset support

пт, 19 квіт. 2024 р. о 19:05 Thierry Reding <[email protected]> пише:

Quoted 60 lines · Svyatoslav Ryhel
Quoted 42 lines

Implement reset use to discard any changes which could have been applied to DC before and can interfere with current configuration.

---
 drivers/video/tegra20/tegra-dc.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
drivers/video/tegra20/tegra-dc.c +6 -0
index 56a23b3c97..35abb6fe46 100644
@@ -10,7 +10,9 @@
 #include <panel.h>
 #include <part.h>
 #include <pwm.h>
+#include <reset.h>
 #include <video.h>
+#include <linux/delay.h>
 #include <asm/cache.h>
 #include <asm/global_data.h>
 #include <asm/system.h>
@@ -342,6 +344,7 @@ static int tegra_lcd_probe(struct udevice *dev)
      struct video_uc_plat *plat = dev_get_uclass_plat(dev);
      struct video_priv *uc_priv = dev_get_uclass_priv(dev);
      struct tegra_lcd_priv *priv = dev_get_priv(dev);
+     struct reset_ctl reset_ctl;
      int ret;
 
      /* Initialize the Tegra display controller */
@@ -349,6 +352,20 @@ static int tegra_lcd_probe(struct udevice *dev)
      funcmux_select(PERIPH_ID_DISP1, FUNCMUX_DEFAULT);
 #endif
 
+     ret = reset_get_by_name(dev, "dc", &reset_ctl);
+     if (ret) {
+             log_err("reset_get_by_name() failed: %d\n", ret);

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_probe(struct udevice *dev) · line 367

Quoted 10 lines
drivers/video/tegra20/tegra-dc.c +10 -0
+             return ret;
+     }
+
+     clock_disable(priv->dc_clk[0]);
+
+     /* Reset everything set before */
+     reset_assert(&reset_ctl);
+     mdelay(4);
+     reset_deassert(&reset_ctl);
+     mdelay(4);

Are you sure this works as intended? It's been a long time since I worked on this, but I seem to recall that most of these resets are actually synchronous, so in order for them to do what they're supposed to the clock needs to be kept running.

Quoted 1 line

The Linux driver certainly does this differently.

You have point, but I have tried to mostly adapt Linux tegra dc driver, which has same logic in probe. Maybe I have not understood it properly. Testing on T20, T30 and T114 passed without issues so far.

Quoted 1 line

Thierry

#19 lore raw

Re: [PATCH v6 06/18] video: tegra20: dc: add reset support

Quoted 61 lines · Svyatoslav Ryhel

пт, 19 квіт. 2024 р. о 19:05 Thierry Reding <[email protected]> пише:

Quoted 60 lines · Svyatoslav Ryhel
Quoted 42 lines

Implement reset use to discard any changes which could have been applied to DC before and can interfere with current configuration.

---
 drivers/video/tegra20/tegra-dc.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
drivers/video/tegra20/tegra-dc.c +6 -0
index 56a23b3c97..35abb6fe46 100644
@@ -10,7 +10,9 @@
 #include <panel.h>
 #include <part.h>
 #include <pwm.h>
+#include <reset.h>
 #include <video.h>
+#include <linux/delay.h>
 #include <asm/cache.h>
 #include <asm/global_data.h>
 #include <asm/system.h>
@@ -342,6 +344,7 @@ static int tegra_lcd_probe(struct udevice *dev)
      struct video_uc_plat *plat = dev_get_uclass_plat(dev);
      struct video_priv *uc_priv = dev_get_uclass_priv(dev);
      struct tegra_lcd_priv *priv = dev_get_priv(dev);
+     struct reset_ctl reset_ctl;
      int ret;
 
      /* Initialize the Tegra display controller */
@@ -349,6 +352,20 @@ static int tegra_lcd_probe(struct udevice *dev)
      funcmux_select(PERIPH_ID_DISP1, FUNCMUX_DEFAULT);
 #endif
 
+     ret = reset_get_by_name(dev, "dc", &reset_ctl);
+     if (ret) {
+             log_err("reset_get_by_name() failed: %d\n", ret);

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_probe(struct udevice *dev) · line 367

Quoted 10 lines
drivers/video/tegra20/tegra-dc.c +10 -0
+             return ret;
+     }
+
+     clock_disable(priv->dc_clk[0]);
+
+     /* Reset everything set before */
+     reset_assert(&reset_ctl);
+     mdelay(4);
+     reset_deassert(&reset_ctl);
+     mdelay(4);

Are you sure this works as intended? It's been a long time since I worked on this, but I seem to recall that most of these resets are actually synchronous, so in order for them to do what they're supposed to the clock needs to be kept running.

Quoted 5 lines
Quoted 1 line

The Linux driver certainly does this differently.

You have point, but I have tried to mostly adapt Linux tegra dc driver, which has same logic in probe. Maybe I have not understood it properly. Testing on T20, T30 and T114 passed without issues so far.

Maybe look again. What it does is (basically):

	clock_enable();
	mdelay(4);
	reset_assert();
	mdelay(4);
	clock_disable();

That should ensure that it's completely reset at that point. Now before any subsequent register accesses happen it will do this:

	clock_enable();
	reset_deassert();

to take it out of reset again. Perhaps that's something you want to keep doing in probe() in U-Boot. In that case maybe you want something like this instead:

	clock_enable();
	mdelay(4);
	reset_assert();
	mdelay(4);
	reset_deassert();

Thierry

#20 lore raw

Re: [PATCH v6 06/18] video: tegra20: dc: add reset support

пт, 19 квіт. 2024 р. о 19:46 Thierry Reding <[email protected]> пише:

Quoted 87 lines · Svyatoslav Ryhel
Quoted 61 lines

пт, 19 квіт. 2024 р. о 19:05 Thierry Reding <[email protected]> пише:

Quoted 60 lines · Svyatoslav Ryhel
Quoted 42 lines

Implement reset use to discard any changes which could have been applied to DC before and can interfere with current configuration.

---
 drivers/video/tegra20/tegra-dc.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)
drivers/video/tegra20/tegra-dc.c +6 -0
index 56a23b3c97..35abb6fe46 100644
@@ -10,7 +10,9 @@
 #include <panel.h>
 #include <part.h>
 #include <pwm.h>
+#include <reset.h>
 #include <video.h>
+#include <linux/delay.h>
 #include <asm/cache.h>
 #include <asm/global_data.h>
 #include <asm/system.h>
@@ -342,6 +344,7 @@ static int tegra_lcd_probe(struct udevice *dev)
      struct video_uc_plat *plat = dev_get_uclass_plat(dev);
      struct video_priv *uc_priv = dev_get_uclass_priv(dev);
      struct tegra_lcd_priv *priv = dev_get_priv(dev);
+     struct reset_ctl reset_ctl;
      int ret;
 
      /* Initialize the Tegra display controller */
@@ -349,6 +352,20 @@ static int tegra_lcd_probe(struct udevice *dev)
      funcmux_select(PERIPH_ID_DISP1, FUNCMUX_DEFAULT);
 #endif
 
+     ret = reset_get_by_name(dev, "dc", &reset_ctl);
+     if (ret) {
+             log_err("reset_get_by_name() failed: %d\n", ret);

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_probe(struct udevice *dev) · line 367

Quoted 10 lines
drivers/video/tegra20/tegra-dc.c +10 -0
+             return ret;
+     }
+
+     clock_disable(priv->dc_clk[0]);
+
+     /* Reset everything set before */
+     reset_assert(&reset_ctl);
+     mdelay(4);
+     reset_deassert(&reset_ctl);
+     mdelay(4);

Are you sure this works as intended? It's been a long time since I worked on this, but I seem to recall that most of these resets are actually synchronous, so in order for them to do what they're supposed to the clock needs to be kept running.

Quoted 5 lines
Quoted 1 line

The Linux driver certainly does this differently.

You have point, but I have tried to mostly adapt Linux tegra dc driver, which has same logic in probe. Maybe I have not understood it properly. Testing on T20, T30 and T114 passed without issues so far.

Maybe look again. What it does is (basically):

        clock_enable();
        mdelay(4);
        reset_assert();
        mdelay(4);
        clock_disable();

That should ensure that it's completely reset at that point. Now before any subsequent register accesses happen it will do this:

        clock_enable();
        reset_deassert();

to take it out of reset again. Perhaps that's something you want to keep doing in probe() in U-Boot. In that case maybe you want something like this instead:

Quoted 6 lines
        clock_enable();
        mdelay(4);
        reset_assert();
        mdelay(4);
        reset_deassert();

You are correct. I assume This patch may be dropped entirely. Thanks.

Quoted 1 line

Thierry

#21 lore raw

PATCH v6 7/18 video: tegra20: dc: add powergate

In reply to Svyatoslav Ryhel

Add powergate use on T114 to complete resetting of DC.

---
 drivers/video/tegra20/tegra-dc.c | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)
drivers/video/tegra20/tegra-dc.c +27 -0
index 35abb6fe46..9a18e38cd8 100644
@@ -22,6 +22,7 @@
 #include <asm/arch/clock.h>
 #include <asm/arch/funcmux.h>
 #include <asm/arch/pinmux.h>
+#include <asm/arch/powergate.h>
 #include <asm/arch/pwm.h>
 
 #include "tegra-dc.h"
@@ -32,6 +33,7 @@ DECLARE_GLOBAL_DATA_PTR;
 struct tegra_dc_soc_info {
 	bool has_timer;
 	bool has_rgb;
+	bool has_pgate;
 };
 
 /* Information about the display controller */
@@ -360,6 +362,28 @@ static int tegra_lcd_probe(struct udevice *dev)
 
 	clock_disable(priv->dc_clk[0]);
 
+	if (priv->soc->has_pgate) {
+		uint powergate;
+
+		if (priv->pipe)
+			powergate = TEGRA_POWERGATE_DISB;
+		else
+			powergate = TEGRA_POWERGATE_DIS;
+
+		ret = tegra_powergate_power_off(powergate);
+		if (ret < 0) {
+			log_err("failed to power off DISP gate: %d", ret);
+			return ret;
+		}
+
+		ret = tegra_powergate_sequence_power_up(powergate,
+							priv->dc_clk[0]);
+		if (ret < 0) {
+			log_err("failed to power up DISP gate: %d", ret);
+			return ret;
+		}
+	}
+
 	/* Reset everything set before */
 	reset_assert(&reset_ctl);
 	mdelay(4);
@@ -510,16 +534,19 @@ static const struct video_ops tegra_lcd_ops = {
 static const struct tegra_dc_soc_info tegra20_dc_soc_info = {
 	.has_timer = true,
 	.has_rgb = true,
+	.has_pgate = false,
 };
 
 static const struct tegra_dc_soc_info tegra30_dc_soc_info = {
 	.has_timer = false,
 	.has_rgb = true,
+	.has_pgate = false,
 };
 
 static const struct tegra_dc_soc_info tegra114_dc_soc_info = {
 	.has_timer = false,
 	.has_rgb = false,
+	.has_pgate = true,
 };
 
 static const struct udevice_id tegra_lcd_ids[] = {
Signature · 1 line
2.40.1
#23 lore raw

PATCH v6 8/18 video: tegra20: dc: configure behavior if PLLD/D2 is used

In reply to Svyatoslav Ryhel

If DISP1 is a PLLD/D2 child, it cannot go over 370MHz. The cause of this is not quite clear. This can be overcomed by further halving the PLLD/D2 if the target parent rate is over 800MHz. This way DISP1 and DSI clocks will have the same frequency. The shift divider in this case has to be calculated from the original PLLD/D2 frequency and is passed from the DSI driver.

---
 drivers/video/tegra20/tegra-dc.c  | 34 +++++++++++++++++++------------
 drivers/video/tegra20/tegra-dc.h  |  1 +
 drivers/video/tegra20/tegra-dsi.c | 14 +++++++++++++
 3 files changed, 36 insertions(+), 13 deletions(-)
drivers/video/tegra20/tegra-dc.c +21 -13
index 9a18e38cd8..a8e32e6893 100644
@@ -48,6 +48,7 @@ struct tegra_lcd_priv {
 	fdt_addr_t frame_buffer;	/* Address of frame buffer */
 	unsigned pixel_clock;		/* Pixel clock in Hz */
 	int dc_clk[2];			/* Contains clk and its parent */
+	ulong scdiv;			/* Clock divider used by disp_clk_ctrl */
 	bool rotation;			/* 180 degree panel turn */
 	bool pipe;			/* DC controller: 0 for A, 1 for B */
 };
@@ -126,8 +127,6 @@ static int update_display_mode(struct tegra_lcd_priv *priv)
 	struct dc_disp_reg *disp = &priv->dc->disp;
 	struct display_timing *dt = &priv->timing;
 	unsigned long val;
-	unsigned long rate;
-	unsigned long div;
 
 	writel(0x0, &disp->disp_timing_opt);
 
@@ -150,21 +149,11 @@ static int update_display_mode(struct tegra_lcd_priv *priv)
 		writel(val, &disp->disp_interface_ctrl);
 	}
 
-	/*
-	 * The pixel clock divider is in 7.1 format (where the bottom bit
-	 * represents 0.5). Here we calculate the divider needed to get from
-	 * the display clock (typically 600MHz) to the pixel clock. We round
-	 * up or down as requried.
-	 */
-	rate = clock_get_periph_rate(priv->dc_clk[0], priv->dc_clk[1]);
-	div = ((rate * 2 + priv->pixel_clock / 2) / priv->pixel_clock) - 2;
-	debug("Display clock %lu, divider %lu\n", rate, div);
-
 	if (priv->soc->has_rgb)
 		writel(0x00010001, &disp->shift_clk_opt);
 
 	val = PIXEL_CLK_DIVIDER_PCD1 << PIXEL_CLK_DIVIDER_SHIFT;
-	val |= div << SHIFT_CLK_DIVIDER_SHIFT;
+	val |= priv->scdiv << SHIFT_CLK_DIVIDER_SHIFT;
 	writel(val, &disp->disp_clk_ctrl);
 
 	return 0;
@@ -314,6 +303,17 @@ static int tegra_display_probe(struct tegra_lcd_priv *priv,
 		rate /= 2;
 #endif
 
+	/*
+	 * The pixel clock divider is in 7.1 format (where the bottom bit
+	 * represents 0.5). Here we calculate the divider needed to get from
+	 * the display clock (typically 600MHz) to the pixel clock. We round
+	 * up or down as required.
+	 */
+	if (!priv->scdiv)
+		priv->scdiv = ((rate * 2 + priv->pixel_clock / 2)
+						/ priv->pixel_clock) - 2;
+	debug("Display clock %lu, divider %lu\n", rate, priv->scdiv);
+
 	/*
 	 * HOST1X is init by default at 150MHz with PLLC as parent
 	 */
@@ -390,6 +390,14 @@ static int tegra_lcd_probe(struct udevice *dev)
 	reset_deassert(&reset_ctl);
 	mdelay(4);
 
+	/* Get shift clock divider from Tegra DSI if used */
+	if (!strcmp(priv->panel->name, TEGRA_DSI_A) ||
+	    !strcmp(priv->panel->name, TEGRA_DSI_B)) {
+		struct tegra_dc_plat *dc_plat = dev_get_plat(priv->panel);
+
+		priv->scdiv = dc_plat->scdiv;
+	}
+
 	if (tegra_display_probe(priv, (void *)plat->base)) {
 		debug("%s: Failed to probe display driver\n", __func__);
 		return -1;
drivers/video/tegra20/tegra-dc.h +1 -0
index 75fc0fa4de..05042dab1c 100644
@@ -23,6 +23,7 @@ struct tegra_dc_plat {
 	struct udevice *dev;		/* Display controller device */
 	struct dc_ctlr *dc;		/* Display controller regmap */
 	bool pipe;			/* DC number: 0 for A, 1 for B */
+	ulong scdiv;			/* Shift clock divider */
 };
 
 /* This holds information about a window which can be displayed */
drivers/video/tegra20/tegra-dsi.c +14 -0
index 72b91ed26b..de225ed376 100644
@@ -743,6 +743,7 @@ static int tegra_dsi_panel_timings(struct udevice *dev,
 static void tegra_dsi_init_clocks(struct udevice *dev)
 {
 	struct tegra_dsi_priv *priv = dev_get_priv(dev);
+	struct tegra_dc_plat *dc_plat = dev_get_plat(dev);
 	struct mipi_dsi_device *device = &priv->device;
 	unsigned int mul, div;
 	unsigned long bclk, plld;
@@ -754,6 +755,19 @@ static void tegra_dsi_init_clocks(struct udevice *dev)
 
 	plld = DIV_ROUND_UP(bclk * 8, USEC_PER_SEC);
 
+	dc_plat->scdiv = ((plld * USEC_PER_SEC +
+		priv->timing.pixelclock.typ / 2) /
+		priv->timing.pixelclock.typ) - 2;
+
+	/*
+	 * BUG: If DISP1 is a PLLD/D2 child, it cannot go over 370MHz. The
+	 * cause of this is not quite clear. This can be overcomed by
+	 * halving the PLLD/D2 if the target rate is > 800MHz. This way
+	 * DISP1 and DSI clocks will be equal.
+	 */
+	if (plld > 800)
+		plld /= 2;
+
 	switch (clock_get_osc_freq()) {
 	case CLOCK_OSC_FREQ_12_0: /* OSC is 12Mhz */
 	case CLOCK_OSC_FREQ_48_0: /* OSC is 48Mhz */
Signature · 1 line
2.40.1
Acked-by on v6 8/18
#24 lore raw

Re: [PATCH v6 08/18] video: tegra20: dc: configure behavior if PLLD/D2 is used

Quoted 13 lines · Svyatoslav Ryhel

If DISP1 is a PLLD/D2 child, it cannot go over 370MHz. The cause of this is not quite clear. This can be overcomed by further halving the PLLD/D2 if the target parent rate is over 800MHz. This way DISP1 and DSI clocks will have the same frequency. The shift divider in this case has to be calculated from the original PLLD/D2 frequency and is passed from the DSI driver.

Quoted 5 lines
---
 drivers/video/tegra20/tegra-dc.c  | 34 +++++++++++++++++++------------
 drivers/video/tegra20/tegra-dc.h  |  1 +
 drivers/video/tegra20/tegra-dsi.c | 14 +++++++++++++
 3 files changed, 36 insertions(+), 13 deletions(-)

I'm not very familiar with these clocks, but seeing that this was extensively tested, I guess this is okay, so:

#25 lore raw

PATCH v6 9/18 video: tegra20: dc: fix printing of framebuffer address

In reply to Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

Framebuffer address should not be a pointer.

---
 drivers/video/tegra20/tegra-dc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/video/tegra20/tegra-dc.c +1 -1
index a8e32e6893..000a0e02f8 100644
@@ -429,7 +429,7 @@ static int tegra_lcd_probe(struct udevice *dev)
 	uc_priv->xsize = priv->width;
 	uc_priv->ysize = priv->height;
 	uc_priv->bpix = priv->log2_bpp;
-	debug("LCD frame buffer at %pa, size %x\n", &priv->frame_buffer,
+	debug("LCD frame buffer at %08x, size %x\n", priv->frame_buffer,
 	      plat->size);
 
 	return 0;
Signature · 1 line
2.40.1
#26 lore raw

Re: [PATCH v6 09/18] video: tegra20: dc: fix printing of framebuffer address

Quoted 15 lines · Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

Framebuffer address should not be a pointer.

---
 drivers/video/tegra20/tegra-dc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/video/tegra20/tegra-dc.c +0 -0
index a8e32e6893..000a0e02f8 100644

drivers/video/tegra20/tegra-dc.c static int tegra_lcd_probe(struct udevice *dev) · line 432

Quoted 5 lines
drivers/video/tegra20/tegra-dc.c +1 -1
@@ -429,7 +429,7 @@ static int tegra_lcd_probe(struct udevice *dev)
 	uc_priv->xsize = priv->width;
 	uc_priv->ysize = priv->height;
 	uc_priv->bpix = priv->log2_bpp;
-	debug("LCD frame buffer at %pa, size %x\n", &priv->frame_buffer,
+	debug("LCD frame buffer at %08x, size %x\n", priv->frame_buffer,

%pa is usually treated in a special way. At least it is on Linux, not sure if the U-Boot printf implementation uses this, too. Looking at doc/develop/printf.rst, it seems like this is supported, and given that priv->frame_buffer is fdt_addr_t, %pa seems totally appropriate here.

Thierry

#27 lore raw

PATCH v6 10/18 video: tegra20: dc: enable backlight after DC is configured

In reply to Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

The goal of panel_set_backlight() is to enable backlight. Hence, it should be called at the probe end.

---
 drivers/video/tegra20/tegra-dc.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)
drivers/video/tegra20/tegra-dc.c +1 -7
index 000a0e02f8..ac6833229a 100644
@@ -414,12 +414,6 @@ static int tegra_lcd_probe(struct udevice *dev)
 		return ret;
 	}
 
-	ret = panel_set_backlight(priv->panel, BACKLIGHT_DEFAULT);
-	if (ret) {
-		debug("%s: Cannot set backlight to default, ret=%d\n", __func__, ret);
-		return ret;
-	}
-
 	mmu_set_region_dcache_behaviour(priv->frame_buffer, plat->size,
 					DCACHE_WRITETHROUGH);
 
@@ -432,7 +426,7 @@ static int tegra_lcd_probe(struct udevice *dev)
 	debug("LCD frame buffer at %08x, size %x\n", priv->frame_buffer,
 	      plat->size);
 
-	return 0;
+	return panel_set_backlight(priv->panel, BACKLIGHT_DEFAULT);
 }
 
 static int tegra_lcd_of_to_plat(struct udevice *dev)
Signature · 1 line
2.40.1
Reviewed-by on v6 10/18
#28 lore raw

Re: [PATCH v6 10/18] video: tegra20: dc: enable backlight after DC is configured

Quoted 10 lines · Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

The goal of panel_set_backlight() is to enable backlight. Hence, it should be called at the probe end.

---
 drivers/video/tegra20/tegra-dc.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)
#29 lore raw

PATCH v6 11/18 video: tegra20: dc: clean framebuffer memory block

In reply to Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

Fill the framebuffer memory with zeros to avoid visual glitches.

---
 drivers/video/tegra20/tegra-dc.c | 5 +++++
 1 file changed, 5 insertions(+)
drivers/video/tegra20/tegra-dc.c +5 -0
index ac6833229a..caeec94468 100644
@@ -4,6 +4,7 @@
  */
 
 #include <backlight.h>
+#include <cpu_func.h>
 #include <dm.h>
 #include <fdtdec.h>
 #include <log.h>
@@ -398,6 +399,10 @@ static int tegra_lcd_probe(struct udevice *dev)
 		priv->scdiv = dc_plat->scdiv;
 	}
 
+	/* Clean the framebuffer area */
+	memset((u8 *)plat->base, 0, plat->size);
+	flush_dcache_all();
+
 	if (tegra_display_probe(priv, (void *)plat->base)) {
 		debug("%s: Failed to probe display driver\n", __func__);
 		return -1;
Signature · 1 line
2.40.1
Reviewed-by on v6 11/18
#30 lore raw

Re: [PATCH v6 11/18] video: tegra20: dc: clean framebuffer memory block

Quoted 9 lines · Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

Fill the framebuffer memory with zeros to avoid visual glitches.

---
 drivers/video/tegra20/tegra-dc.c | 5 +++++
 1 file changed, 5 insertions(+)

Makes sense:

#31 lore raw

PATCH v6 12/18 video: tegra20: dc: parameterize V- and H-sync polarities

In reply to Svyatoslav Ryhel

Based on Thierry Reding's Linux commit:

'commit 1716b1891e1de05e2c20ccafa9f58550f3539717 ("drm/tegra: rgb: Parameterize V- and H-sync polarities")'

---
 arch/arm/include/asm/arch-tegra/dc.h |  5 +++++
 drivers/video/tegra20/tegra-dc.c     | 22 ++++++++++++++++++++--
 2 files changed, 25 insertions(+), 2 deletions(-)
arch/arm/include/asm/arch-tegra/dc.h +5 -0
index 6444af2993..ca3718411a 100644
@@ -443,6 +443,11 @@ enum win_color_depth_id {
 #define	WINDOW_D_SELECT		BIT(7)
 #define	WINDOW_H_SELECT		BIT(8)
 
+/* DC_COM_PIN_OUTPUT_POLARITY1 0x307 */
+#define LHS_OUTPUT_POLARITY_LOW		BIT(30)
+#define LVS_OUTPUT_POLARITY_LOW		BIT(28)
+#define LSC0_OUTPUT_POLARITY_LOW	BIT(24)
+
 /* DC_DISP_DISP_WIN_OPTIONS 0x402 */
 #define	CURSOR_ENABLE		BIT(16)
 #define	SOR_ENABLE		BIT(25)
drivers/video/tegra20/tegra-dc.c +20 -2
index caeec94468..ffa81eeac3 100644
@@ -220,8 +220,11 @@ static const u32 rgb_sel_tab[PIN_OUTPUT_SEL_COUNT] = {
 	0x00020000,
 };
 
-static void rgb_enable(struct dc_com_reg *com)
+static void rgb_enable(struct tegra_lcd_priv *priv)
 {
+	struct dc_com_reg *com = &priv->dc->com;
+	struct display_timing *dt = &priv->timing;
+	u32 value;
 	int i;
 
 	for (i = 0; i < PIN_REG_COUNT; i++) {
@@ -230,6 +233,21 @@ static void rgb_enable(struct dc_com_reg *com)
 		writel(rgb_data_tab[i], &com->pin_output_data[i]);
 	}
 
+	/* configure H- and V-sync signal polarities */
+	value = readl(&com->pin_output_polarity[1]);
+
+	if (dt->flags & DISPLAY_FLAGS_HSYNC_LOW)
+		value |= LHS_OUTPUT_POLARITY_LOW;
+	else
+		value &= ~LHS_OUTPUT_POLARITY_LOW;
+
+	if (dt->flags & DISPLAY_FLAGS_VSYNC_LOW)
+		value |= LVS_OUTPUT_POLARITY_LOW;
+	else
+		value &= ~LVS_OUTPUT_POLARITY_LOW;
+
+	writel(value, &com->pin_output_polarity[1]);
+
 	for (i = 0; i < PIN_OUTPUT_SEL_COUNT; i++)
 		writel(rgb_sel_tab[i], &com->pin_output_sel[i]);
 }
@@ -329,7 +347,7 @@ static int tegra_display_probe(struct tegra_lcd_priv *priv,
 		basic_init_timer(&priv->dc->disp);
 
 	if (priv->soc->has_rgb)
-		rgb_enable(&priv->dc->com);
+		rgb_enable(priv);
 
 	if (priv->pixel_clock)
 		update_display_mode(priv);
Signature · 1 line
2.40.1
Reviewed-by on v6 12/18
#32 lore raw

Re: [PATCH v6 12/18] video: tegra20: dc: parameterize V- and H-sync polarities

Quoted 16 lines · Svyatoslav Ryhel

Based on Thierry Reding's Linux commit:

'commit 1716b1891e1de05e2c20ccafa9f58550f3539717 ("drm/tegra: rgb: Parameterize V- and H-sync polarities")'

---
 arch/arm/include/asm/arch-tegra/dc.h |  5 +++++
 drivers/video/tegra20/tegra-dc.c     | 22 ++++++++++++++++++++--
 2 files changed, 25 insertions(+), 2 deletions(-)
arch/arm/include/asm/arch-tegra/dc.h +0 -0
index 6444af2993..ca3718411a 100644

arch/arm/include/asm/arch-tegra/dc.h enum win_color_depth_id { · line 449

Quoted 7 lines
arch/arm/include/asm/arch-tegra/dc.h +4 -0
@@ -443,6 +443,11 @@ enum win_color_depth_id {
 #define	WINDOW_D_SELECT		BIT(7)
 #define	WINDOW_H_SELECT		BIT(8)
 
+/* DC_COM_PIN_OUTPUT_POLARITY1 0x307 */
+#define LHS_OUTPUT_POLARITY_LOW		BIT(30)
+#define LVS_OUTPUT_POLARITY_LOW		BIT(28)
+#define LSC0_OUTPUT_POLARITY_LOW	BIT(24)

This definition seems to be unused, other than that:

#33 lore raw

PATCH v6 13/18 video: tegra20: add MIPI calibration driver

In reply to Svyatoslav Ryhel

Dedicated MIPI calibration driver is used on T114 and newer. Before T114 MIPI calibration registers were part of VI and CSI.

---
 drivers/video/tegra20/Makefile     |   2 +-
 drivers/video/tegra20/tegra-mipi.c | 188 +++++++++++++++++++++++++++++
 2 files changed, 189 insertions(+), 1 deletion(-)
 create mode 100644 drivers/video/tegra20/tegra-mipi.c
drivers/video/tegra20/Makefile +1 -1
index f0b534c579..a75aea2a87 100644
@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0+
 
 obj-$(CONFIG_VIDEO_TEGRA20) += tegra-dc.o
-obj-$(CONFIG_VIDEO_DSI_TEGRA30) += tegra-dsi.o mipi-phy.o
+obj-$(CONFIG_VIDEO_DSI_TEGRA30) += tegra-dsi.o tegra-mipi.o mipi-phy.o
 obj-$(CONFIG_TEGRA_BACKLIGHT_PWM) += tegra-pwm-backlight.o
drivers/video/tegra20/tegra-mipi.c new file +188 -0

188 lines not on this page · raw message

Signature · 1 line
2.40.1
Reviewed-by on v6 13/18
#34 lore raw

Re: [PATCH v6 13/18] video: tegra20: add MIPI calibration driver

Quoted 10 lines · Svyatoslav Ryhel

Dedicated MIPI calibration driver is used on T114 and newer. Before T114 MIPI calibration registers were part of VI and CSI.

---
 drivers/video/tegra20/Makefile     |   2 +-
 drivers/video/tegra20/tegra-mipi.c | 188 +++++++++++++++++++++++++++++
 2 files changed, 189 insertions(+), 1 deletion(-)
 create mode 100644 drivers/video/tegra20/tegra-mipi.c

This looks like an adequate copy of what we have in the Linux kernel. It's slightly suboptimal because most of the register values are hard- coded for Tegra114, but I guess that can always be improved later on.

#35 lore raw

PATCH v6 14/18 video: tegra20: dsi: add T114 support

In reply to Svyatoslav Ryhel

Existing Tegra DSI driver mostly fits T114 apart MIPI calibration which on T114 has dedicated driver. To resolve this MIPI calibration logic was split for pre-T114 and T114+ devices.

---
 drivers/video/tegra20/tegra-dsi.c             | 78 ++++++++++++++++++-
 .../video/tegra20/tegra-dsi.h                 | 24 +++++-
 2 files changed, 96 insertions(+), 6 deletions(-)
 rename arch/arm/include/asm/arch-tegra30/dsi.h => drivers/video/tegra20/tegra-dsi.h (90%)
drivers/video/tegra20/tegra-dsi.c +75 -3
index de225ed376..25a629535e 100644
@@ -20,17 +20,24 @@
 #include <asm/gpio.h>
 #include <asm/io.h>
 #include <asm/arch/clock.h>
-#include <asm/arch-tegra30/dsi.h>
 
 #include "tegra-dc.h"
+#include "tegra-dsi.h"
 #include "mipi-phy.h"
 
+/* List of supported DSI bridges */
+enum {
+	DSI_V0,
+	DSI_V1,
+};
+
 struct tegra_dsi_priv {
 	struct mipi_dsi_host host;
 	struct mipi_dsi_device device;
 	struct mipi_dphy_timing dphy_timing;
 
 	struct udevice *panel;
+	struct udevice *mipi;
 	struct display_timing timing;
 
 	struct dsi_ctlr *dsi;
@@ -41,6 +48,8 @@ struct tegra_dsi_priv {
 	int dsi_clk;
 	int video_fifo_depth;
 	int host_fifo_depth;
+
+	u32 version;
 };
 
 static void tegra_dc_enable_controller(struct udevice *dev)
@@ -501,6 +510,41 @@ static void tegra_dsi_pad_calibrate(struct dsi_pad_ctrl_reg *pad)
 	writel(value, TEGRA_VI_BASE + (CSI_CIL_PAD_CONFIG << 2));
 }
 
+static void tegra_dsi_mipi_calibrate(struct tegra_dsi_priv *priv)
+{
+	struct dsi_pad_ctrl_reg *pad = &priv->dsi->pad;
+	u32 value;
+	int ret;
+
+	ret = misc_set_enabled(priv->mipi, true);
+	if (ret)
+		log_debug("%s: failed to enable MIPI calibration: %d\n",
+			  __func__, ret);
+
+	writel(0, &pad->pad_ctrl);
+	writel(0, &pad->pad_ctrl_1);
+	writel(0, &pad->pad_ctrl_2);
+	writel(0, &pad->pad_ctrl_3);
+	writel(0, &pad->pad_ctrl_4);
+
+	/* DSI pad enable */
+	value = DSI_PAD_CONTROL_VS1_PULLDN(0) | DSI_PAD_CONTROL_VS1_PDIO(0);
+	writel(value, &pad->pad_ctrl);
+
+	value = DSI_PAD_SLEW_UP(0x7) | DSI_PAD_SLEW_DN(0x7) |
+		DSI_PAD_LP_UP(0x1) | DSI_PAD_LP_DN(0x1) |
+		DSI_PAD_OUT_CLK(0x0);
+	writel(value, &pad->pad_ctrl_2);
+
+	value = DSI_PAD_PREEMP_PD_CLK(0x3) | DSI_PAD_PREEMP_PU_CLK(0x3) |
+		DSI_PAD_PREEMP_PD(0x03) | DSI_PAD_PREEMP_PU(0x3);
+	writel(value, &pad->pad_ctrl_3);
+
+	ret = misc_write(priv->mipi, 0, NULL, 0);
+	if (ret)
+		log_debug("%s: MIPI calibration failed %d\n", __func__, ret);
+}
+
 static void tegra_dsi_set_timeout(struct dsi_timeout_reg *rtimeout,
 				  unsigned long bclk,
 				  unsigned int vrefresh)
@@ -664,10 +708,25 @@ static int tegra_dsi_encoder_enable(struct udevice *dev)
 	u32 value;
 	int ret;
 
+	/* If for some reasone DSI is enabled then it needs to
+	 * be disabled in order for the panel initialization
+	 * commands to be properly sent.
+	 */
+	value = readl(&misc->dsi_pwr_ctrl);
+
+	if (value & DSI_POWER_CONTROL_ENABLE) {
+		value = readl(&misc->dsi_pwr_ctrl);
+		value &= ~DSI_POWER_CONTROL_ENABLE;
+		writel(value, &misc->dsi_pwr_ctrl);
+	}
+
 	/* Disable interrupt */
 	writel(0, &misc->int_enable);
 
-	tegra_dsi_pad_calibrate(&priv->dsi->pad);
+	if (priv->version)
+		tegra_dsi_mipi_calibrate(priv);
+	else
+		tegra_dsi_pad_calibrate(&priv->dsi->pad);
 
 	tegra_dsi_get_muldiv(device->format, &mul, &div);
 
@@ -806,6 +865,8 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 	struct mipi_dsi_panel_plat *mipi_plat;
 	int ret;
 
+	priv->version = dev_get_driver_data(dev);
+
 	priv->dsi = (struct dsi_ctlr *)dev_read_addr_ptr(dev);
 	if (!priv->dsi) {
 		printf("%s: No display controller address\n", __func__);
@@ -828,6 +889,16 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 		return log_ret(ret);
 	}
 
+	if (priv->version) {
+		ret = uclass_get_device_by_phandle(UCLASS_MISC, dev,
+						   "nvidia,mipi-calibrate",
+						   &priv->mipi);
+		if (ret) {
+			log_debug("%s: cannot get MIPI: error %d\n", __func__, ret);
+			return ret;
+		}
+	}
+
 	panel_get_display_timing(priv->panel, &priv->timing);
 
 	mipi_plat = dev_get_plat(priv->panel);
@@ -859,7 +930,8 @@ static const struct panel_ops tegra_dsi_bridge_ops = {
 };
 
 static const struct udevice_id tegra_dsi_bridge_ids[] = {
-	{ .compatible = "nvidia,tegra30-dsi" },
+	{ .compatible = "nvidia,tegra30-dsi", .data = DSI_V0 },
+	{ .compatible = "nvidia,tegra114-dsi", .data = DSI_V1 },
 	{ }
 };
 
drivers/video/tegra20/tegra-dsi.h renamed from arch/arm/include/asm/arch-tegra30/dsi.h +21 -3
similarity index 90%
rename from arch/arm/include/asm/arch-tegra30/dsi.h
rename to drivers/video/tegra20/tegra-dsi.h
index 7ade132613..69dac4bd1b 100644
@@ -4,8 +4,8 @@
  *  NVIDIA Corporation <www.nvidia.com>
  */
 
-#ifndef __ASM_ARCH_TEGRA_DSI_H
-#define __ASM_ARCH_TEGRA_DSI_H
+#ifndef _TEGRA_DSI_H
+#define _TEGRA_DSI_H
 
 #ifndef __ASSEMBLY__
 #include <linux/bitops.h>
@@ -105,6 +105,10 @@ struct dsi_pad_ctrl_reg {
 	uint pad_ctrl_cd;		/* _PAD_CONTROL_CD_0 */
 	uint pad_cd_status;		/* _PAD_CD_STATUS_0 */
 	uint dsi_vid_mode_control;	/* _DSI_VID_MODE_CONTROL_0 */
+	uint pad_ctrl_1;		/* _PAD_CONTROL_1 */
+	uint pad_ctrl_2;		/* _PAD_CONTROL_2 */
+	uint pad_ctrl_3;		/* _PAD_CONTROL_3 */
+	uint pad_ctrl_4;		/* _PAD_CONTROL_4 */
 };
 
 /* Display Serial Interface (DSI_) regs */
@@ -184,6 +188,20 @@ struct dsi_ctlr {
 #define DSI_PAD_CONTROL_PAD_LPUPADJ(x)		(((x) & 0x3) << 14)
 #define DSI_PAD_CONTROL_PAD_LPDNADJ(x)		(((x) & 0x3) << 12)
 
+#define DSI_PAD_CONTROL_VS1_PDIO(x)	(((x) & 0xf) <<  0)
+#define DSI_PAD_CONTROL_VS1_PULLDN(x)	(((x) & 0xf) << 16)
+
+#define DSI_PAD_OUT_CLK(x)		(((x) & 0x7) <<  0)
+#define DSI_PAD_LP_DN(x)		(((x) & 0x7) <<  4)
+#define DSI_PAD_LP_UP(x)		(((x) & 0x7) <<  8)
+#define DSI_PAD_SLEW_DN(x)		(((x) & 0x7) << 12)
+#define DSI_PAD_SLEW_UP(x)		(((x) & 0x7) << 16)
+
+#define DSI_PAD_PREEMP_PD_CLK(x)	(((x) & 0x3) << 12)
+#define DSI_PAD_PREEMP_PU_CLK(x)	(((x) & 0x3) << 8)
+#define DSI_PAD_PREEMP_PD(x)		(((x) & 0x3) << 4)
+#define DSI_PAD_PREEMP_PU(x)		(((x) & 0x3) << 0)
+
 /*
  * pixel format as used in the DSI_CONTROL_FORMAT field
  */
@@ -214,4 +232,4 @@ enum tegra_dsi_format {
 #define  PAD_DRIV_DN_REF(x)		(((x) & 0x7) << 16)
 #define  PAD_DRIV_UP_REF(x)		(((x) & 0x7) << 8)
 
-#endif /* __ASM_ARCH_TEGRA_DSI_H */
+#endif /* _TEGRA_DSI_H */
Signature · 1 line
2.40.1
Reviewed-by on v6 14/18
#36 lore raw

Re: [PATCH v6 14/18] video: tegra20: dsi: add T114 support

Quoted 4 lines · Svyatoslav Ryhel

Existing Tegra DSI driver mostly fits T114 apart MIPI calibration which on T114 has dedicated driver. To resolve this MIPI calibration logic was split for pre-T114 and T114+ devices.

Quoted 8 lines
---
 drivers/video/tegra20/tegra-dsi.c             | 78 ++++++++++++++++++-
 .../video/tegra20/tegra-dsi.h                 | 24 +++++-
 2 files changed, 96 insertions(+), 6 deletions(-)
 rename arch/arm/include/asm/arch-tegra30/dsi.h => drivers/video/tegra20/tegra-dsi.h (90%)
#37 lore raw

PATCH v6 15/18 video: tegra20: dsi: add reset support

In reply to Svyatoslav Ryhel

Implement reset use to discard any changes which could have been applied to DSI before and can interfere with current configuration.

---
 drivers/video/tegra20/tegra-dsi.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
drivers/video/tegra20/tegra-dsi.c +14 -0
index 25a629535e..fc9ca1310a 100644
@@ -12,6 +12,7 @@
 #include <mipi_dsi.h>
 #include <backlight.h>
 #include <panel.h>
+#include <reset.h>
 #include <linux/delay.h>
 #include <linux/err.h>
 #include <linux/time.h>
@@ -863,6 +864,7 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 	struct tegra_dsi_priv *priv = dev_get_priv(dev);
 	struct mipi_dsi_device *device = &priv->device;
 	struct mipi_dsi_panel_plat *mipi_plat;
+	struct reset_ctl reset_ctl;
 	int ret;
 
 	priv->version = dev_get_driver_data(dev);
@@ -876,6 +878,13 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 	priv->video_fifo_depth = 480;
 	priv->host_fifo_depth = 64;
 
+	ret = reset_get_by_name(dev, "dsi", &reset_ctl);
+	if (ret) {
+		log_debug("%s: reset_get_by_name() failed: %d\n",
+			  __func__, ret);
+		return ret;
+	}
+
 	ret = uclass_get_device_by_phandle(UCLASS_REGULATOR, dev,
 					   "avdd-dsi-csi-supply", &priv->avdd);
 	if (ret)
@@ -914,12 +923,17 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 
 	tegra_dsi_get_format(device->format, &priv->format);
 
+	reset_assert(&reset_ctl);
+
 	ret = regulator_set_enable_if_allowed(priv->avdd, true);
 	if (ret && ret != -ENOSYS)
 		return ret;
 
 	tegra_dsi_init_clocks(dev);
 
+	mdelay(2);
+	reset_deassert(&reset_ctl);
+
 	return 0;
 }
 
Signature · 1 line
2.40.1
#38 lore raw

Re: [PATCH v6 15/18] video: tegra20: dsi: add reset support

Quoted 51 lines · Svyatoslav Ryhel

Implement reset use to discard any changes which could have been applied to DSI before and can interfere with current configuration.

---
 drivers/video/tegra20/tegra-dsi.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
drivers/video/tegra20/tegra-dsi.c +11 -0
index 25a629535e..fc9ca1310a 100644
@@ -12,6 +12,7 @@
 #include <mipi_dsi.h>
 #include <backlight.h>
 #include <panel.h>
+#include <reset.h>
 #include <linux/delay.h>
 #include <linux/err.h>
 #include <linux/time.h>
@@ -863,6 +864,7 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 	struct tegra_dsi_priv *priv = dev_get_priv(dev);
 	struct mipi_dsi_device *device = &priv->device;
 	struct mipi_dsi_panel_plat *mipi_plat;
+	struct reset_ctl reset_ctl;
 	int ret;
 
 	priv->version = dev_get_driver_data(dev);
@@ -876,6 +878,13 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 	priv->video_fifo_depth = 480;
 	priv->host_fifo_depth = 64;
 
+	ret = reset_get_by_name(dev, "dsi", &reset_ctl);
+	if (ret) {
+		log_debug("%s: reset_get_by_name() failed: %d\n",
+			  __func__, ret);
+		return ret;
+	}
+
 	ret = uclass_get_device_by_phandle(UCLASS_REGULATOR, dev,
 					   "avdd-dsi-csi-supply", &priv->avdd);
 	if (ret)
@@ -914,12 +923,17 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 
 	tegra_dsi_get_format(device->format, &priv->format);
 
+	reset_assert(&reset_ctl);
+
 	ret = regulator_set_enable_if_allowed(priv->avdd, true);

drivers/video/tegra20/tegra-dsi.c static int tegra_dsi_bridge_probe(struct udevice *dev) · line 938

Quoted 10 lines
drivers/video/tegra20/tegra-dsi.c +3 -0
 	if (ret && ret != -ENOSYS)
 		return ret;
 
 	tegra_dsi_init_clocks(dev);
 
+	mdelay(2);
+	reset_deassert(&reset_ctl);
+
 	return 0;
 }

Looks like tegra_dsi_init_clocks() already does the mdelay() and reset_deassert()? Or perhaps I'm again looking at things in the wrong order?

Thierry

#39 lore raw

PATCH v6 16/18 video: tegra20: dsi: remove pre-configuration

In reply to Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

Configuration for DC driver command mode is not required for every panel. Removed.

---
 drivers/video/tegra20/tegra-dsi.c | 2 --
 1 file changed, 2 deletions(-)
drivers/video/tegra20/tegra-dsi.c +0 -2
index fc9ca1310a..ddd74540f2 100644
@@ -766,8 +766,6 @@ static int tegra_dsi_encoder_enable(struct udevice *dev)
 	if (ret)
 		return ret;
 
-	tegra_dsi_configure(dev, 0);
-
 	ret = panel_set_backlight(priv->panel, BACKLIGHT_DEFAULT);
 	if (ret)
 		return ret;
Signature · 1 line
2.40.1
Reviewed-by on v6 16/18
#40 lore raw

Re: [PATCH v6 16/18] video: tegra20: dsi: remove pre-configuration

Quoted 5 lines · Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

Configuration for DC driver command mode is not required for every panel. Removed.

Quoted 7 lines
---
 drivers/video/tegra20/tegra-dsi.c | 2 --
 1 file changed, 2 deletions(-)

Yeah, this seems completely unnecessary.

#41 lore raw

PATCH v6 17/18 video: tegra20: dsi: set correct fifo depth

In reply to Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

According to Thierry Reding's commit in the linux kernel

976cebc35bed0456a42bf96073a26f251d23b264 "drm/tegra: dsi: Make FIFO depths host parameters"

correct depth of the video FIFO is 1920 *words* no *bytes*

---
 drivers/video/tegra20/tegra-dsi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/video/tegra20/tegra-dsi.c +1 -1
index ddd74540f2..7d63557d1b 100644
@@ -873,7 +873,7 @@ static int tegra_dsi_bridge_probe(struct udevice *dev)
 		return -EINVAL;
 	}
 
-	priv->video_fifo_depth = 480;
+	priv->video_fifo_depth = 1920;
 	priv->host_fifo_depth = 64;
 
 	ret = reset_get_by_name(dev, "dsi", &reset_ctl);
Signature · 1 line
2.40.1
Reviewed-by on v6 17/18
#42 lore raw

Re: [PATCH v6 17/18] video: tegra20: dsi: set correct fifo depth

Quoted 9 lines · Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

According to Thierry Reding's commit in the linux kernel

976cebc35bed0456a42bf96073a26f251d23b264 "drm/tegra: dsi: Make FIFO depths host parameters"

correct depth of the video FIFO is 1920 *words* no *bytes*

Quoted 7 lines
---
 drivers/video/tegra20/tegra-dsi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
#43 lore raw

PATCH v6 18/18 video: tegra20: dsi: use set_backlight for backlight only

In reply to Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

Shift the backlight set further to prevent visual glitches on panel init.

---
 drivers/video/tegra20/tegra-dsi.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/video/tegra20/tegra-dsi.c +4 -6
index 7d63557d1b..13dae37806 100644
@@ -766,10 +766,6 @@ static int tegra_dsi_encoder_enable(struct udevice *dev)
 	if (ret)
 		return ret;
 
-	ret = panel_set_backlight(priv->panel, BACKLIGHT_DEFAULT);
-	if (ret)
-		return ret;
-
 	tegra_dsi_configure(dev, device->mode_flags);
 
 	tegra_dc_enable_controller(dev);
@@ -784,8 +780,10 @@ static int tegra_dsi_encoder_enable(struct udevice *dev)
 
 static int tegra_dsi_bridge_set_panel(struct udevice *dev, int percent)
 {
-	/* Is not used in tegra dc */
-	return 0;
+	struct tegra_dsi_priv *priv = dev_get_priv(dev);
+
+	/* Turn on/off backlight */
+	return panel_set_backlight(priv->panel, percent);
 }
 
 static int tegra_dsi_panel_timings(struct udevice *dev,
Signature · 1 line
2.40.1
Reviewed-by on v6 18/18
#44 lore raw

Re: [PATCH v6 18/18] video: tegra20: dsi: use set_backlight for backlight only

Quoted 10 lines · Svyatoslav Ryhel

From: Jonas Schwöbel <[email protected]>

Shift the backlight set further to prevent visual glitches on panel init.

---
 drivers/video/tegra20/tegra-dsi.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

I vaguely recall that some devices may have had the panel and the backlight hooked up to the same regulator or enable GPIO, so calling panel_set_backlight() too late may cause the DSI panel to malfunction.

That said, I'm not familiar with panel_set_backlight(), so perhaps it always only sets the brightness and the power may already be on earlier?

Ah... nevermind... I see that panel_enable_backlight() is called prior to panel_set_backlight(), so this looks like it should be fine.

#45 lore raw
In reply to Svyatoslav Ryhel

Hello Tom! This patch set is hanging in patchwork for 3 month without any comments. If no one has anything to say, may you pick it into master?

Best regards,
Svyatoslav R.

вт, 23 січ. 2024 р. о 19:17 Svyatoslav Ryhel <[email protected]> пише:

Quoted 87 lines

T114 is not that different from T30 and all T30 drivers will work on T114 as well with some adjustments.

Patches propose general improvements for existing Tegra DC and DSI drivers as well Tegra 114 video support (experimentl).

Commits pass buildman for tegra.

---
Changes from v5:
- backlight enable moved to the last step of setup for DSI
- parameterized V- and H-sync polarities
- added framebuffer clearing on probe to avoid glitches
- backlight enable moved after DC is fully configured
- fixed printing framebuffer pointer instead of address
- moved scdiv calculation to tegra DSI if it is used

Changes from v4:
- fixed typo in max rate to be divided (400KHz > 400MHz)

Changes from v3:
- fixed the clock divider calculation if PLLD/D2 is used
- removed unnecessary pre-configuration
- set correct video FIFO depth for DSI

Changes from v2:
- fixed image distortion on devices with 180deg rotation

Changes from v1:
- reworked patchset entirely
- diverged DC configuration per-SOC
- consolidated dc headers from different SOC gen
- initial support of DC detection (tegra has 2 DC)
- added PLLD2 support, resets and powergating
- added T114+ MIPI calibration
- added DSI detection (tegra has 2 DSI) and resets
---

Jonas Schwöbel (6):
video: tegra20: dc: fix printing of framebuffer address
video: tegra20: dc: enable backlight after DC is configured
video: tegra20: dc: clean framebuffer memory block
video: tegra20: dsi: remove pre-configuration
video: tegra20: dsi: set correct fifo depth
video: tegra20: dsi: use set_backlight for backlight only

Svyatoslav Ryhel (12):
video: tegra20: dc: diverge DC per-SOC
video: tegra20: dc: fix image shift on rotated panels
video: tegra20: consolidate DC header
video: tegra20: dc: pass DC id to internal devices
video: tegra20: dc: add PLLD2 parent support
video: tegra20: dc: add reset support
video: tegra20: dc: add powergate
video: tegra20: dc: configure behavior if PLLD/D2 is used
video: tegra20: dc: parameterize V- and H-sync polarities
video: tegra20: add MIPI calibration driver
video: tegra20: dsi: add T114 support
video: tegra20: dsi: add reset support

 arch/arm/dts/tegra114-u-boot.dtsi             |  13 +
 arch/arm/dts/tegra114.dtsi                    |   4 +-
 arch/arm/dts/tegra30-u-boot.dtsi              |   4 +
 arch/arm/dts/tegra30.dtsi                     |   2 +-
 arch/arm/include/asm/arch-tegra/dc.h          |  13 +-
 arch/arm/include/asm/arch-tegra114/pwm.h      |  13 +
 arch/arm/include/asm/arch-tegra20/display.h   |  28 --
 arch/arm/include/asm/arch-tegra30/display.h   |  28 --
 drivers/video/tegra20/Makefile                |   2 +-
 drivers/video/tegra20/tegra-dc.c              | 239 +++++++++++++-----
 drivers/video/tegra20/tegra-dc.h              |  45 ++++
 drivers/video/tegra20/tegra-dsi.c             | 122 ++++++++-
 .../video/tegra20/tegra-dsi.h                 |  24 +-
 drivers/video/tegra20/tegra-mipi.c            | 188 ++++++++++++++
 drivers/video/tegra20/tegra-pwm-backlight.c   |   3 +-
 15 files changed, 583 insertions(+), 145 deletions(-)
 create mode 100644 arch/arm/include/asm/arch-tegra114/pwm.h
 delete mode 100644 arch/arm/include/asm/arch-tegra20/display.h
 delete mode 100644 arch/arm/include/asm/arch-tegra30/display.h
 create mode 100644 drivers/video/tegra20/tegra-dc.h
 rename arch/arm/include/asm/arch-tegra30/dsi.h => drivers/video/tegra20/tegra-dsi.h (90%)
 create mode 100644 drivers/video/tegra20/tegra-mipi.c

--
2.40.1

#46 lore raw
In reply to Svyatoslav Ryhel
Quoted 7 lines · On Tue, 23 Jan 2024 19:16:15 +0200 Svyatoslav Ryhel [email protected] wrote:

T114 is not that different from T30 and all T30 drivers will work on T114 as well with some adjustments.

Patches propose general improvements for existing Tegra DC and DSI drivers as well Tegra 114 video support (experimentl).

Commits pass buildman for tegra.

Series applied to u-boot-video, except patch 06/18.

--
Anatolij