Skip to contentSkip to conversation

media: v4l: Add support for the Cadence MIPI-CSI2 TX controller

PATCH Superseded Maxime Ripard 2 participants 8 messages Apr 4, 2018 – Apr 17, 2018

Subsystems (3)

v8 of 12 Superseded Latest is v12 →

lore
Share & tools
lore
#0 lore raw

v8 cover media: v4l: Add support for the Cadence MIPI-CSI2 TX controller

Hi,

Here is an attempt at supporting the MIPI-CSI2 TX block from Cadence.

This IP block is able to receive 4 video streams and stream them over a MIPI-CSI2 link using up to 4 lanes. Those streams are basically the interfaces to controllers generating some video signals, like a camera or a pattern generator.

It is able to map input streams to CSI2 virtual channels and datatypes dynamically. The streaming devices choose their virtual channels through an additional signal that is transparent to the CSI2-TX. The datatypes however are yet another additional input signal, and can be mapped to any CSI2 datatypes.

Since v4l2 doesn't really allow for that setup at the moment, this preliminary version is a rather dumb one in order to start the discussion on how to address this properly.

Let me know what you think!
Maxime

Changes from v7:
- Removed unused headers
- Fixed a function prefix inconsistency
- Added Niklas' Reviewed-by

Changes from v6:
- Added Niklas Reviewed-by on the bindings
- Added MODULE_LICENSE, MODULE_DESCRIPTION and MODULE_AUTHOR

Changes from v5:
- Changed the return type to void for csi2tx_stop
- Checked for the pad index in get/set_pad_format
- Made a comment that the DPHY registers are for the DPHY *interface* register

Changes from v4:
- After playing a bit with the pad multiplexing patches, found that it was making much more sense to have the subdev notifiers for the source subdev rather for the sink that might even be outside of Linux control. Removed the notifier for now.

Changes from v3:
- Added a comment about entity links walk concurrency
- Changed the default resolution to 1280x720
- Changed usleep_range calls to udelay
- Reworked the reference counting mechanism to remove a race condition by adding a mutex instead of an atomic count
- Changed the entity function to MEDIA_ENT_F_VID_IF_BRIDGE
- Changed the name of the reg variable in _get_resources to dev_cfg
- Removed the redundant error message in the devm_ioremap_resource error path
- Moved the subdev s_stream call before enabling the TX bridge
- Changed some int types to unsigned
- Init'd the pad formats properly
- Fixed typo in the CSI2TX_LANES_MAX define name
- Added Sakari Acked-by

Changes from v2:
- Use SPDX license header
- Use the lane mapping from DT

Changes from v1:
- Add a subdev notifier and start our downstream subdevice in s_stream
- Based the decision to enable the stream or not on the link state instead of whether a format was being set on the pad
- Put the controller back in reset when stopping the pipeline
- Clarified the enpoints number in the DT binding
- Added a default format for the pads
- Added some missing const
- Added more explicit comments
- Rebased on 4.15

Maxime Ripard (2):
dt-bindings: media: Add Cadence MIPI-CSI2 TX Device Tree bindings
v4l: cadence: Add Cadence MIPI-CSI2 TX driver

 .../devicetree/bindings/media/cdns,csi2tx.txt      |  98 ++++
 drivers/media/platform/cadence/Kconfig             |  11 +
 drivers/media/platform/cadence/Makefile            |   1 +
 drivers/media/platform/cadence/cdns-csi2tx.c       | 529 +++++++++++++++++++++
 4 files changed, 639 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/media/cdns,csi2tx.txt
 create mode 100644 drivers/media/platform/cadence/cdns-csi2tx.c
Signature · 1 line
2.14.3
#1 lore raw

PATCH v8 1/2 dt-bindings: media: Add Cadence MIPI-CSI2 TX Device Tree bindings

The Cadence MIPI-CSI2 TX controller is a CSI2 bridge that supports up to 4 video streams and can output on up to 4 CSI-2 lanes, depending on the hardware implementation.

It can operate with an external D-PHY, an internal one or no D-PHY at all in some configurations.

---
 .../devicetree/bindings/media/cdns,csi2tx.txt      | 98 ++++++++++++++++++++++
 1 file changed, 98 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/media/cdns,csi2tx.txt
Documentation/devicetree/bindings/media/cdns,csi2tx.txt new file +98 -0
new file mode 100644
index 000000000000..459c6e332f52
@@ -0,0 +1,98 @@
+Cadence MIPI-CSI2 TX controller
+===============================
+
+The Cadence MIPI-CSI2 TX controller is a CSI-2 bridge supporting up to
+4 CSI lanes in output, and up to 4 different pixel streams in input.
+
+Required properties:
+  - compatible: must be set to "cdns,csi2tx"
+  - reg: base address and size of the memory mapped region
+  - clocks: phandles to the clocks driving the controller
+  - clock-names: must contain:
+    * esc_clk: escape mode clock
+    * p_clk: register bank clock
+    * pixel_if[0-3]_clk: pixel stream output clock, one for each stream
+                         implemented in hardware, between 0 and 3
+
+Optional properties
+  - phys: phandle to the D-PHY. If it is set, phy-names need to be set
+  - phy-names: must contain "dphy"
+
+Required subnodes:
+  - ports: A ports node with one port child node per device input and output
+           port, in accordance with the video interface bindings defined in
+           Documentation/devicetree/bindings/media/video-interfaces.txt. The
+           port nodes are numbered as follows.
+
+           Port Description
+           -----------------------------
+           0    CSI-2 output
+           1    Stream 0 input
+           2    Stream 1 input
+           3    Stream 2 input
+           4    Stream 3 input
+
+           The stream input port nodes are optional if they are not
+           connected to anything at the hardware level or implemented
+           in the design. Since there is only one endpoint per port,
+           the endpoints are not numbered.
+
+Example:
+
+csi2tx: csi-bridge@0d0e1000 {
+	compatible = "cdns,csi2tx";
+	reg = <0x0d0e1000 0x1000>;
+	clocks = <&byteclock>, <&byteclock>,
+		 <&coreclock>, <&coreclock>,
+		 <&coreclock>, <&coreclock>;
+	clock-names = "p_clk", "esc_clk",
+		      "pixel_if0_clk", "pixel_if1_clk",
+		      "pixel_if2_clk", "pixel_if3_clk";
+
+	ports {
+		#address-cells = <1>;
+		#size-cells = <0>;
+
+		port@0 {
+			reg = <0>;
+
+			csi2tx_out: endpoint {
+				remote-endpoint = <&remote_in>;
+				clock-lanes = <0>;
+				data-lanes = <1 2>;
+			};
+		};
+
+		port@1 {
+			reg = <1>;
+
+			csi2tx_in_stream0: endpoint {
+				remote-endpoint = <&stream0_out>;
+			};
+		};
+
+		port@2 {
+			reg = <2>;
+
+			csi2tx_in_stream1: endpoint {
+				remote-endpoint = <&stream1_out>;
+			};
+		};
+
+		port@3 {
+			reg = <3>;
+
+			csi2tx_in_stream2: endpoint {
+				remote-endpoint = <&stream2_out>;
+			};
+		};
+
+		port@4 {
+			reg = <4>;
+
+			csi2tx_in_stream3: endpoint {
+				remote-endpoint = <&stream3_out>;
+			};
+		};
+	};
+};
Signature · 1 line
2.14.3
#2 lore raw

PATCH v8 2/2 v4l: cadence: Add Cadence MIPI-CSI2 TX driver

In reply to Maxime Ripard

The Cadence MIPI-CSI2 TX controller is an hardware block meant to be used as a bridge between pixel interfaces and a CSI-2 bus.

It supports operating with an internal or external D-PHY, with up to 4 lanes, or without any D-PHY. The current code only supports the latter case.

While the virtual channel input on the pixel interface can be directly mapped to CSI2, the datatype input is actually a selection signal (3-bits) mapping to a table of up to 8 preconfigured datatypes/formats (programmed at start-up)

The block supports up to 8 input datatypes.

---
 drivers/media/platform/cadence/Kconfig       |  11 +
 drivers/media/platform/cadence/Makefile      |   1 +
 drivers/media/platform/cadence/cdns-csi2tx.c | 529 +++++++++++++++++++++++++++
 3 files changed, 541 insertions(+)
 create mode 100644 drivers/media/platform/cadence/cdns-csi2tx.c
drivers/media/platform/cadence/Kconfig +11 -0
index 70c95d79c8f7..3bf0f2454384 100644
@@ -20,4 +20,15 @@ config VIDEO_CADENCE_CSI2RX
 	  To compile this driver as a module, choose M here: the module will be
 	  called cdns-csi2rx.
 
+config VIDEO_CADENCE_CSI2TX
+	tristate "Cadence MIPI-CSI2 TX Controller"
+	depends on MEDIA_CONTROLLER
+	depends on VIDEO_V4L2_SUBDEV_API
+	select V4L2_FWNODE
+	help
+	  Support for the Cadence MIPI CSI2 Transceiver controller.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called cdns-csi2tx.
+
 endif
drivers/media/platform/cadence/Makefile +1 -0
index 99a4086b7448..7fe992273162 100644
@@ -1 +1,2 @@
 obj-$(CONFIG_VIDEO_CADENCE_CSI2RX)	+= cdns-csi2rx.o
+obj-$(CONFIG_VIDEO_CADENCE_CSI2TX)	+= cdns-csi2tx.o
drivers/media/platform/cadence/cdns-csi2tx.c new file +529 -0

529 lines not on this page · raw message

Signature · 1 line
2.14.3
#3 lore raw

Re: [PATCH v8 2/2] v4l: cadence: Add Cadence MIPI-CSI2 TX driver

Hi Maxime,

Quoted 56 lines · Maxime Ripard

The Cadence MIPI-CSI2 TX controller is an hardware block meant to be used as a bridge between pixel interfaces and a CSI-2 bus.

It supports operating with an internal or external D-PHY, with up to 4 lanes, or without any D-PHY. The current code only supports the latter case.

While the virtual channel input on the pixel interface can be directly mapped to CSI2, the datatype input is actually a selection signal (3-bits) mapping to a table of up to 8 preconfigured datatypes/formats (programmed at start-up)

The block supports up to 8 input datatypes.

---
 drivers/media/platform/cadence/Kconfig       |  11 +
 drivers/media/platform/cadence/Makefile      |   1 +
 drivers/media/platform/cadence/cdns-csi2tx.c | 529 +++++++++++++++++++++++++++
 3 files changed, 541 insertions(+)
 create mode 100644 drivers/media/platform/cadence/cdns-csi2tx.c
drivers/media/platform/cadence/Kconfig +11 -0
index 70c95d79c8f7..3bf0f2454384 100644
@@ -20,4 +20,15 @@ config VIDEO_CADENCE_CSI2RX
 	  To compile this driver as a module, choose M here: the module will be
 	  called cdns-csi2rx.
 
+config VIDEO_CADENCE_CSI2TX
+	tristate "Cadence MIPI-CSI2 TX Controller"
+	depends on MEDIA_CONTROLLER
+	depends on VIDEO_V4L2_SUBDEV_API
+	select V4L2_FWNODE
+	help
+	  Support for the Cadence MIPI CSI2 Transceiver controller.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called cdns-csi2tx.
+
 endif
drivers/media/platform/cadence/Makefile +1 -0
index 99a4086b7448..7fe992273162 100644
@@ -1 +1,2 @@
 obj-$(CONFIG_VIDEO_CADENCE_CSI2RX)	+= cdns-csi2rx.o
+obj-$(CONFIG_VIDEO_CADENCE_CSI2TX)	+= cdns-csi2tx.o
drivers/media/platform/cadence/cdns-csi2tx.c new file +0 -0
new file mode 100644
index 000000000000..4495044e9f99

drivers/media/platform/cadence/cdns-csi2tx.c line 5

Quoted 5 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +5 -0
@@ -0,0 +1,529 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Driver for Cadence MIPI-CSI2 TX Controller
+ *
+ * Copyright (C) 2017 Cadence Design Systems Inc.

Maybe 2017--2018?

Quoted 108 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +108 -0
+ */
+
+#include <linux/clk.h>
+#include <linux/delay.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/of.h>
+#include <linux/of_graph.h>
+#include <linux/platform_device.h>
+
+#include <media/v4l2-ctrls.h>
+#include <media/v4l2-device.h>
+#include <media/v4l2-fwnode.h>
+#include <media/v4l2-subdev.h>
+
+#define CSI2TX_DEVICE_CONFIG_REG	0x00
+
+#define CSI2TX_CONFIG_REG		0x20
+#define CSI2TX_CONFIG_CFG_REQ			BIT(2)
+#define CSI2TX_CONFIG_SRST_REQ			BIT(1)
+
+#define CSI2TX_DPHY_CFG_REG		0x28
+#define CSI2TX_DPHY_CFG_CLK_RESET		BIT(16)
+#define CSI2TX_DPHY_CFG_LANE_RESET(n)		BIT((n) + 12)
+#define CSI2TX_DPHY_CFG_MODE_MASK		GENMASK(9, 8)
+#define CSI2TX_DPHY_CFG_MODE_LPDT		(2 << 8)
+#define CSI2TX_DPHY_CFG_MODE_HS			(1 << 8)
+#define CSI2TX_DPHY_CFG_MODE_ULPS		(0 << 8)
+#define CSI2TX_DPHY_CFG_CLK_ENABLE		BIT(4)
+#define CSI2TX_DPHY_CFG_LANE_ENABLE(n)		BIT(n)
+
+#define CSI2TX_DPHY_CLK_WAKEUP_REG	0x2c
+#define CSI2TX_DPHY_CLK_WAKEUP_ULPS_CYCLES(n)	((n) & 0xffff)
+
+#define CSI2TX_DT_CFG_REG(n)		(0x80 + (n) * 8)
+#define CSI2TX_DT_CFG_DT(n)			(((n) & 0x3f) << 2)
+
+#define CSI2TX_DT_FORMAT_REG(n)		(0x84 + (n) * 8)
+#define CSI2TX_DT_FORMAT_BYTES_PER_LINE(n)	(((n) & 0xffff) << 16)
+#define CSI2TX_DT_FORMAT_MAX_LINE_NUM(n)	((n) & 0xffff)
+
+#define CSI2TX_STREAM_IF_CFG_REG(n)	(0x100 + (n) * 4)
+#define CSI2TX_STREAM_IF_CFG_FILL_LEVEL(n)	((n) & 0x1f)
+
+#define CSI2TX_LANES_MAX	4
+#define CSI2TX_STREAMS_MAX	4
+
+enum csi2tx_pads {
+	CSI2TX_PAD_SOURCE,
+	CSI2TX_PAD_SINK_STREAM0,
+	CSI2TX_PAD_SINK_STREAM1,
+	CSI2TX_PAD_SINK_STREAM2,
+	CSI2TX_PAD_SINK_STREAM3,
+	CSI2TX_PAD_MAX,
+};
+
+struct csi2tx_fmt {
+	u32	mbus;
+	u32	dt;
+	u32	bpp;
+};
+
+struct csi2tx_priv {
+	struct device			*dev;
+	unsigned int			count;
+
+	/*
+	 * Used to prevent race conditions between multiple,
+	 * concurrent calls to start and stop.
+	 */
+	struct mutex			lock;
+
+	void __iomem			*base;
+
+	struct clk			*esc_clk;
+	struct clk			*p_clk;
+	struct clk			*pixel_clk[CSI2TX_STREAMS_MAX];
+
+	struct v4l2_subdev		subdev;
+	struct media_pad		pads[CSI2TX_PAD_MAX];
+	struct v4l2_mbus_framefmt	pad_fmts[CSI2TX_PAD_MAX];
+
+	bool				has_internal_dphy;
+	u8				lanes[CSI2TX_LANES_MAX];
+	unsigned int			num_lanes;
+	unsigned int			max_lanes;
+	unsigned int			max_streams;
+};
+
+static const struct csi2tx_fmt csi2tx_formats[] = {
+	{
+		.mbus	= MEDIA_BUS_FMT_UYVY8_1X16,
+		.bpp	= 2,
+		.dt	= 0x1e,
+	},
+	{
+		.mbus	= MEDIA_BUS_FMT_RGB888_1X24,
+		.bpp	= 3,
+		.dt	= 0x24,
+	},
+};
+
+static const struct v4l2_mbus_framefmt fmt_default = {
+	.width		= 1280,
+	.height		= 720,
+	.code		= MEDIA_BUS_FMT_RGB888_1X24,
+	.field		= V4L2_FIELD_NONE,

drivers/media/platform/cadence/cdns-csi2tx.c line 123

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+	.colorspace	= V4L2_COLORSPACE_DEFAULT,
+};
+
+static inline
+struct csi2tx_priv *v4l2_subdev_to_csi2tx(struct v4l2_subdev *subdev)
+{
+	return container_of(subdev, struct csi2tx_priv, subdev);
+}
+
+static const struct csi2tx_fmt *csi2tx_get_fmt_from_mbus(struct v4l2_mbus_framefmt *mfmt)

Over 80 characters, please wrap.

You could use just the mbus code as the argument if that's all you need.

Quoted 23 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +23 -0
+{
+	unsigned int i;
+
+	if (!mfmt)
+		return NULL;
+
+	for (i = 0; i < ARRAY_SIZE(csi2tx_formats); i++)
+		if (csi2tx_formats[i].mbus == mfmt->code)
+			return &csi2tx_formats[i];
+
+	return NULL;
+}
+
+static int csi2tx_enum_mbus_code(struct v4l2_subdev *subdev,
+				 struct v4l2_subdev_pad_config *cfg,
+				 struct v4l2_subdev_mbus_code_enum *code)
+{
+	if (code->pad || code->index >= ARRAY_SIZE(csi2tx_formats))
+		return -EINVAL;
+
+	code->code = csi2tx_formats[code->index].mbus;
+
+	return 0;

drivers/media/platform/cadence/cdns-csi2tx.c line 156

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+}
+
+static int csi2tx_get_pad_format(struct v4l2_subdev *subdev,
+				 struct v4l2_subdev_pad_config *cfg,
+				 struct v4l2_subdev_format *fmt)
+{
+	struct csi2tx_priv *csi2tx = v4l2_subdev_to_csi2tx(subdev);
+
+	if (fmt->pad >= CSI2TX_PAD_MAX)
+		return -EINVAL;

The V4L2 subdev IOCTL handlers do check the pad numbers, you can remove them here (and below).

Quoted 6 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +6 -0
+
+	fmt->format = csi2tx->pad_fmts[fmt->pad];
+
+	return 0;
+}
+

drivers/media/platform/cadence/cdns-csi2tx.c line 172

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+static int csi2tx_set_pad_format(struct v4l2_subdev *subdev,
+				 struct v4l2_subdev_pad_config *cfg,
+				 struct v4l2_subdev_format *fmt)
+{
+	struct csi2tx_priv *csi2tx = v4l2_subdev_to_csi2tx(subdev);
+
+	if (fmt->pad >= CSI2TX_PAD_MAX)
+		return -EINVAL;
+
+	csi2tx->pad_fmts[fmt->pad] = fmt->format;

Have I asked previously if there are any limitations with this?

The CSI-2 TX link has multiple formats so I wouldn't support formats on that pad in order to be compatible with the planned VC/data type support patchset. Or do you see issues with that?

Quoted 191 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +191 -0
+
+	return 0;
+}
+
+static const struct v4l2_subdev_pad_ops csi2tx_pad_ops = {
+	.enum_mbus_code	= csi2tx_enum_mbus_code,
+	.get_fmt	= csi2tx_get_pad_format,
+	.set_fmt	= csi2tx_set_pad_format,
+};
+
+static void csi2tx_reset(struct csi2tx_priv *csi2tx)
+{
+	writel(CSI2TX_CONFIG_SRST_REQ, csi2tx->base + CSI2TX_CONFIG_REG);
+
+	udelay(10);
+}
+
+static int csi2tx_start(struct csi2tx_priv *csi2tx)
+{
+	struct media_entity *entity = &csi2tx->subdev.entity;
+	struct media_link *link;
+	unsigned int i;
+	u32 reg;
+
+	csi2tx_reset(csi2tx);
+
+	writel(CSI2TX_CONFIG_CFG_REQ, csi2tx->base + CSI2TX_CONFIG_REG);
+
+	udelay(10);
+
+	/* Configure our PPI interface with the D-PHY */
+	writel(CSI2TX_DPHY_CLK_WAKEUP_ULPS_CYCLES(32),
+	       csi2tx->base + CSI2TX_DPHY_CLK_WAKEUP_REG);
+
+	/* Put our lanes (clock and data) out of reset */
+	reg = CSI2TX_DPHY_CFG_CLK_RESET | CSI2TX_DPHY_CFG_MODE_LPDT;
+	for (i = 0; i < csi2tx->num_lanes; i++)
+		reg |= CSI2TX_DPHY_CFG_LANE_RESET(csi2tx->lanes[i]);
+	writel(reg, csi2tx->base + CSI2TX_DPHY_CFG_REG);
+
+	udelay(10);
+
+	/* Enable our (clock and data) lanes */
+	reg |= CSI2TX_DPHY_CFG_CLK_ENABLE;
+	for (i = 0; i < csi2tx->num_lanes; i++)
+		reg |= CSI2TX_DPHY_CFG_LANE_ENABLE(csi2tx->lanes[i]);
+	writel(reg, csi2tx->base + CSI2TX_DPHY_CFG_REG);
+
+	udelay(10);
+
+	/* Switch to HS mode */
+	reg &= ~CSI2TX_DPHY_CFG_MODE_MASK;
+	writel(reg | CSI2TX_DPHY_CFG_MODE_HS,
+	       csi2tx->base + CSI2TX_DPHY_CFG_REG);
+
+	udelay(10);
+
+	/*
+	 * Create a static mapping between the CSI virtual channels
+	 * and the input streams.
+	 *
+	 * This should be enhanced, but v4l2 lacks the support for
+	 * changing that mapping dynamically at the moment.
+	 *
+	 * We're protected from the userspace setting up links at the
+	 * same time by the upper layer having called
+	 * media_pipeline_start().
+	 */
+	list_for_each_entry(link, &entity->links, list) {
+		struct v4l2_mbus_framefmt *mfmt;
+		const struct csi2tx_fmt *fmt;
+		unsigned int stream;
+		int pad_idx = -1;
+
+		/* Only consider our enabled input pads */
+		for (i = CSI2TX_PAD_SINK_STREAM0; i < CSI2TX_PAD_MAX; i++) {
+			struct media_pad *pad = &csi2tx->pads[i];
+
+			if ((pad == link->sink) &&
+			    (link->flags & MEDIA_LNK_FL_ENABLED)) {
+				pad_idx = i;
+				break;
+			}
+		}
+
+		if (pad_idx < 0)
+			continue;
+
+		mfmt = &csi2tx->pad_fmts[pad_idx];
+		fmt = csitx_get_fmt_from_mbus(mfmt);
+		if (!fmt)
+			continue;
+
+		stream = pad_idx - CSI2TX_PAD_SINK_STREAM0;
+
+		/*
+		 * We use the stream ID there, but it's wrong.
+		 *
+		 * A stream could very well send a data type that is
+		 * not equal to its stream ID. We need to find a
+		 * proper way to address it.
+		 */
+		writel(CSI2TX_DT_CFG_DT(fmt->dt),
+		       csi2tx->base + CSI2TX_DT_CFG_REG(stream));
+
+		writel(CSI2TX_DT_FORMAT_BYTES_PER_LINE(mfmt->width * fmt->bpp) |
+		       CSI2TX_DT_FORMAT_MAX_LINE_NUM(mfmt->height + 1),
+		       csi2tx->base + CSI2TX_DT_FORMAT_REG(stream));
+
+		/*
+		 * TODO: This needs to be calculated based on the
+		 * output CSI2 clock rate.
+		 */
+		writel(CSI2TX_STREAM_IF_CFG_FILL_LEVEL(4),
+		       csi2tx->base + CSI2TX_STREAM_IF_CFG_REG(stream));
+	}
+
+	/* Disable the configuration mode */
+	writel(0, csi2tx->base + CSI2TX_CONFIG_REG);
+
+	return 0;
+}
+
+static void csi2tx_stop(struct csi2tx_priv *csi2tx)
+{
+	writel(CSI2TX_CONFIG_CFG_REQ | CSI2TX_CONFIG_SRST_REQ,
+	       csi2tx->base + CSI2TX_CONFIG_REG);
+}
+
+static int csi2tx_s_stream(struct v4l2_subdev *subdev, int enable)
+{
+	struct csi2tx_priv *csi2tx = v4l2_subdev_to_csi2tx(subdev);
+	int ret = 0;
+
+	mutex_lock(&csi2tx->lock);
+
+	if (enable) {
+		/*
+		 * If we're not the first users, there's no need to
+		 * enable the whole controller.
+		 */
+		if (!csi2tx->count) {
+			ret = csi2tx_start(csi2tx);
+			if (ret)
+				goto out;
+		}
+
+		csi2tx->count++;
+	} else {
+		csi2tx->count--;
+
+		/*
+		 * Let the last user turn off the lights.
+		 */
+		if (!csi2tx->count)
+			csi2tx_stop(csi2tx);
+	}
+
+out:
+	mutex_unlock(&csi2tx->lock);
+	return ret;
+}
+
+static const struct v4l2_subdev_video_ops csi2tx_video_ops = {
+	.s_stream	= csi2tx_s_stream,
+};
+
+static const struct v4l2_subdev_ops csi2tx_subdev_ops = {
+	.pad		= &csi2tx_pad_ops,
+	.video		= &csi2tx_video_ops,
+};
+
+static int csi2tx_get_resources(struct csi2tx_priv *csi2tx,
+				struct platform_device *pdev)
+{
+	struct resource *res;
+	unsigned int i;
+	u32 dev_cfg;
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	csi2tx->base = devm_ioremap_resource(&pdev->dev, res);
+	if (IS_ERR(csi2tx->base))
+		return PTR_ERR(csi2tx->base);
+
+	csi2tx->p_clk = devm_clk_get(&pdev->dev, "p_clk");
+	if (IS_ERR(csi2tx->p_clk)) {
+		dev_err(&pdev->dev, "Couldn't get p_clk\n");
+		return PTR_ERR(csi2tx->p_clk);
+	}
+
+	csi2tx->esc_clk = devm_clk_get(&pdev->dev, "esc_clk");

drivers/media/platform/cadence/cdns-csi2tx.c line 373

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+	if (IS_ERR(csi2tx->esc_clk)) {
+		dev_err(&pdev->dev, "Couldn't get the esc_clk\n");
+		return PTR_ERR(csi2tx->esc_clk);
+	}
+
+	clk_prepare_enable(csi2tx->p_clk);
+	dev_cfg = readl(csi2tx->base + CSI2TX_DEVICE_CONFIG_REG);
+	clk_disable_unprepare(csi2tx->p_clk);
+
+	csi2tx->max_lanes = dev_cfg & 7;

Some defines would be nice for this (instead of magic numbers).

Quoted 56 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +56 -0
+	if (csi2tx->max_lanes > CSI2TX_LANES_MAX) {
+		dev_err(&pdev->dev, "Invalid number of lanes: %u\n",
+			csi2tx->max_lanes);
+		return -EINVAL;
+	}
+
+	csi2tx->max_streams = (dev_cfg >> 4) & 7;
+	if (csi2tx->max_streams > CSI2TX_STREAMS_MAX) {
+		dev_err(&pdev->dev, "Invalid number of streams: %u\n",
+			csi2tx->max_streams);
+		return -EINVAL;
+	}
+
+	csi2tx->has_internal_dphy = (dev_cfg & BIT(3)) ? true : false;
+
+	for (i = 0; i < csi2tx->max_streams; i++) {
+		char clk_name[16];
+
+		snprintf(clk_name, sizeof(clk_name), "pixel_if%u_clk", i);
+		csi2tx->pixel_clk[i] = devm_clk_get(&pdev->dev, clk_name);
+		if (IS_ERR(csi2tx->pixel_clk[i])) {
+			dev_err(&pdev->dev, "Couldn't get clock %s\n",
+				clk_name);
+			return PTR_ERR(csi2tx->pixel_clk[i]);
+		}
+	}
+
+	return 0;
+}
+
+static int csi2tx_check_lanes(struct csi2tx_priv *csi2tx)
+{
+	struct v4l2_fwnode_endpoint v4l2_ep;
+	struct device_node *ep;
+	int ret;
+
+	ep = of_graph_get_endpoint_by_regs(csi2tx->dev->of_node, 0, 0);
+	if (!ep)
+		return -EINVAL;
+
+	ret = v4l2_fwnode_endpoint_parse(of_fwnode_handle(ep), &v4l2_ep);
+	if (ret) {
+		dev_err(csi2tx->dev, "Could not parse v4l2 endpoint\n");
+		goto out;
+	}
+
+	if (v4l2_ep.bus_type != V4L2_MBUS_CSI2) {
+		dev_err(csi2tx->dev, "Unsupported media bus type: 0x%x\n",
+			v4l2_ep.bus_type);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	csi2tx->num_lanes = v4l2_ep.bus.mipi_csi2.num_data_lanes;
+	if (csi2tx->num_lanes > csi2tx->max_lanes) {
+		dev_err(csi2tx->dev,

drivers/media/platform/cadence/cdns-csi2tx.c line 439

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+			"Current configuration uses more lanes than supported\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	memcpy(csi2tx->lanes, v4l2_ep.bus.mipi_csi2.data_lanes,
+	       sizeof(csi2tx->lanes));
+
+out:
+	of_node_put(ep);

Newline here?

drivers/media/platform/cadence/cdns-csi2tx.c line 451

Quoted 12 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +12 -0
+	return ret;
+}
+
+static int csi2tx_probe(struct platform_device *pdev)
+{
+	struct csi2tx_priv *csi2tx;
+	unsigned int i;
+	int ret;
+
+	csi2tx = kzalloc(sizeof(*csi2tx), GFP_KERNEL);
+	if (!csi2tx)
+		return -ENOMEM;

And here?

Quoted 36 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +36 -0
+	platform_set_drvdata(pdev, csi2tx);
+	mutex_init(&csi2tx->lock);
+	csi2tx->dev = &pdev->dev;
+
+	ret = csi2tx_get_resources(csi2tx, pdev);
+	if (ret)
+		goto err_free_priv;
+
+	v4l2_subdev_init(&csi2tx->subdev, &csi2tx_subdev_ops);
+	csi2tx->subdev.owner = THIS_MODULE;
+	csi2tx->subdev.dev = &pdev->dev;
+	csi2tx->subdev.flags |= V4L2_SUBDEV_FL_HAS_DEVNODE;
+	snprintf(csi2tx->subdev.name, V4L2_SUBDEV_NAME_SIZE, "%s.%s",
+		 KBUILD_MODNAME, dev_name(&pdev->dev));
+
+	ret = csi2tx_check_lanes(csi2tx);
+	if (ret)
+		goto err_free_priv;
+
+	/* Create our media pads */
+	csi2tx->subdev.entity.function = MEDIA_ENT_F_VID_IF_BRIDGE;
+	csi2tx->pads[CSI2TX_PAD_SOURCE].flags = MEDIA_PAD_FL_SOURCE;
+	for (i = CSI2TX_PAD_SINK_STREAM0; i < CSI2TX_PAD_MAX; i++)
+		csi2tx->pads[i].flags = MEDIA_PAD_FL_SINK;
+
+	for (i = 0; i < CSI2TX_PAD_MAX; i++)
+		csi2tx->pad_fmts[i] = fmt_default;
+
+	ret = media_entity_pads_init(&csi2tx->subdev.entity, CSI2TX_PAD_MAX,
+				     csi2tx->pads);
+	if (ret)
+		goto err_free_priv;
+
+	ret = v4l2_async_register_subdev(&csi2tx->subdev);
+	if (ret < 0)
+		goto err_free_priv;

drivers/media/platform/cadence/cdns-csi2tx.c line 497

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+
+	dev_info(&pdev->dev,
+		 "Probed CSI2TX with %u/%u lanes, %u streams, %s D-PHY\n",
+		 csi2tx->num_lanes, csi2tx->max_lanes, csi2tx->max_streams,
+		 csi2tx->has_internal_dphy ? "internal" : "no");
+
+	return 0;
+
+err_free_priv:
+	kfree(csi2tx);

And here.

Quoted 32 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +32 -0
+	return ret;
+}
+
+static int csi2tx_remove(struct platform_device *pdev)
+{
+	struct csi2tx_priv *csi2tx = platform_get_drvdata(pdev);
+
+	v4l2_async_unregister_subdev(&csi2tx->subdev);
+	kfree(csi2tx);
+
+	return 0;
+}
+
+static const struct of_device_id csi2tx_of_table[] = {
+	{ .compatible = "cdns,csi2tx" },
+	{ },
+};
+MODULE_DEVICE_TABLE(of, csi2tx_of_table);
+
+static struct platform_driver csi2tx_driver = {
+	.probe	= csi2tx_probe,
+	.remove	= csi2tx_remove,
+
+	.driver	= {
+		.name		= "cdns-csi2tx",
+		.of_match_table	= csi2tx_of_table,
+	},
+};
+module_platform_driver(csi2tx_driver);
+MODULE_AUTHOR("Maxime Ripard <[email protected]>");
+MODULE_DESCRIPTION("Cadence CSI2-TX controller");
+MODULE_LICENSE("GPL");
Signature · 4 lines
Kind regards,

Sakari Ailus
e-mail: [email protected]
#4 lore raw

Re: [PATCH v8 2/2] v4l: cadence: Add Cadence MIPI-CSI2 TX driver

Hi Sakari,

Quoted 13 lines · Sakari Ailus

drivers/media/platform/cadence/cdns-csi2tx.c line 172

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+static int csi2tx_set_pad_format(struct v4l2_subdev *subdev,
+				 struct v4l2_subdev_pad_config *cfg,
+				 struct v4l2_subdev_format *fmt)
+{
+	struct csi2tx_priv *csi2tx = v4l2_subdev_to_csi2tx(subdev);
+
+	if (fmt->pad >= CSI2TX_PAD_MAX)
+		return -EINVAL;
+
+	csi2tx->pad_fmts[fmt->pad] = fmt->format;

Have I asked previously if there are any limitations with this?

Quoted 3 lines

The CSI-2 TX link has multiple formats so I wouldn't support formats on that pad in order to be compatible with the planned VC/data type support patchset. Or do you see issues with that?

It's not just about the CSI-2 link, but more about the input pads as well, that can be configured (and we need to know the format in order to configure the IP properly).

Maybe we can simply prevent the format change on the CSI-2 pad, but not the others?

Maxime

Signature · 3 lines
Maxime Ripard, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com
#5 lore raw

Re: [PATCH v8 2/2] v4l: cadence: Add Cadence MIPI-CSI2 TX driver

Quoted 20 lines · Maxime Ripard

Hi Sakari,

Quoted 13 lines · Sakari Ailus

drivers/media/platform/cadence/cdns-csi2tx.c line 172

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+static int csi2tx_set_pad_format(struct v4l2_subdev *subdev,
+				 struct v4l2_subdev_pad_config *cfg,
+				 struct v4l2_subdev_format *fmt)
+{
+	struct csi2tx_priv *csi2tx = v4l2_subdev_to_csi2tx(subdev);
+
+	if (fmt->pad >= CSI2TX_PAD_MAX)
+		return -EINVAL;
+
+	csi2tx->pad_fmts[fmt->pad] = fmt->format;

Have I asked previously if there are any limitations with this?

Quoted 3 lines

The CSI-2 TX link has multiple formats so I wouldn't support formats on that pad in order to be compatible with the planned VC/data type support patchset. Or do you see issues with that?

Quoted 6 lines

It's not just about the CSI-2 link, but more about the input pads as well, that can be configured (and we need to know the format in order to configure the IP properly).

Maybe we can simply prevent the format change on the CSI-2 pad, but not the others?

Yes, that was what I wanted to suggest. It's in line with the intended way to support multiplexed pads.

The latest set is here:

<URL:https://git.linuxtv.org/sailus/media_tree.git/log/?h=vc>

Signature · 2 lines
Sakari Ailus
e-mail: [email protected]
#6 lore raw

Re: [PATCH v8 2/2] v4l: cadence: Add Cadence MIPI-CSI2 TX driver

Quoted 28 lines · Sakari Ailus
Quoted 20 lines · Maxime Ripard

Hi Sakari,

Quoted 13 lines · Sakari Ailus

drivers/media/platform/cadence/cdns-csi2tx.c line 172

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+static int csi2tx_set_pad_format(struct v4l2_subdev *subdev,
+				 struct v4l2_subdev_pad_config *cfg,
+				 struct v4l2_subdev_format *fmt)
+{
+	struct csi2tx_priv *csi2tx = v4l2_subdev_to_csi2tx(subdev);
+
+	if (fmt->pad >= CSI2TX_PAD_MAX)
+		return -EINVAL;
+
+	csi2tx->pad_fmts[fmt->pad] = fmt->format;

Have I asked previously if there are any limitations with this?

Quoted 3 lines

The CSI-2 TX link has multiple formats so I wouldn't support formats on that pad in order to be compatible with the planned VC/data type support patchset. Or do you see issues with that?

Quoted 6 lines

It's not just about the CSI-2 link, but more about the input pads as well, that can be configured (and we need to know the format in order to configure the IP properly).

Maybe we can simply prevent the format change on the CSI-2 pad, but not the others?

Quoted 6 lines

Yes, that was what I wanted to suggest. It's in line with the intended way to support multiplexed pads.

The latest set is here:

<URL:https://git.linuxtv.org/sailus/media_tree.git/log/?h=vc>

Thanks for the pointer.

I've looked at the smiapp set_format hook, and especially:
https://git.linuxtv.org/sailus/media_tree.git/tree/drivers/media/i2c/smiapp/smiapp-core.c?h=vc&id=cb864a1d8e2d19b793d8f550b026dcf8d2f78f11#n1817

After reading this, I'm not quite sure to get what I should do for the CSI-2 pad. Should I ignore all formats change (and thus return 0), or should I return EINVAL (which would probably be a bit confusing to the userspace)?

Maxime

Signature · 3 lines
Maxime Ripard, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com
#7 lore raw

Re: [PATCH v8 2/2] v4l: cadence: Add Cadence MIPI-CSI2 TX driver

Quoted 38 lines · Maxime Ripard
Quoted 28 lines · Sakari Ailus
Quoted 20 lines · Maxime Ripard

Hi Sakari,

Quoted 13 lines · Sakari Ailus

drivers/media/platform/cadence/cdns-csi2tx.c line 172

Quoted 10 lines
drivers/media/platform/cadence/cdns-csi2tx.c new file +10 -0
+static int csi2tx_set_pad_format(struct v4l2_subdev *subdev,
+				 struct v4l2_subdev_pad_config *cfg,
+				 struct v4l2_subdev_format *fmt)
+{
+	struct csi2tx_priv *csi2tx = v4l2_subdev_to_csi2tx(subdev);
+
+	if (fmt->pad >= CSI2TX_PAD_MAX)
+		return -EINVAL;
+
+	csi2tx->pad_fmts[fmt->pad] = fmt->format;

Have I asked previously if there are any limitations with this?

Quoted 3 lines

The CSI-2 TX link has multiple formats so I wouldn't support formats on that pad in order to be compatible with the planned VC/data type support patchset. Or do you see issues with that?

Quoted 6 lines

It's not just about the CSI-2 link, but more about the input pads as well, that can be configured (and we need to know the format in order to configure the IP properly).

Maybe we can simply prevent the format change on the CSI-2 pad, but not the others?

Quoted 6 lines

Yes, that was what I wanted to suggest. It's in line with the intended way to support multiplexed pads.

The latest set is here:

<URL:https://git.linuxtv.org/sailus/media_tree.git/log/?h=vc>

Thanks for the pointer.

Quoted 7 lines

I've looked at the smiapp set_format hook, and especially:
https://git.linuxtv.org/sailus/media_tree.git/tree/drivers/media/i2c/smiapp/smiapp-core.c?h=vc&id=cb864a1d8e2d19b793d8f550b026dcf8d2f78f11#n1817

After reading this, I'm not quite sure to get what I should do for the CSI-2 pad. Should I ignore all formats change (and thus return 0), or should I return EINVAL (which would probably be a bit confusing to the userspace)?

-EINVAL, please. The subdev uAPI does currently not (nor it is intended to) support accessing multiple formats on a single pad, the logical choice is to return an error when accessing a format here. That is aligned with the stream support patchset (link above).

If you want to access the format, you need to use the pads on the other side of the subdev.

Signature · 2 lines
Sakari Ailus
e-mail: [email protected]