Hi Maxime,
Thank you for the patch.
On Thursday, 8 February 2018 17:08:30 EET Maxime Ripard wrote:
> The Cadence CSI-2 RX Controller is an hardware block meant to be used as a
> bridge between a CSI-2 bus and pixel grabbers.
>
> It supports operating with internal or external D-PHY, with up to 4 lanes,
> or without any D-PHY. The current code only supports the former case.
Don't you support the latter case, not the former ? The driver bails out with
an error if an internal or external PHY is present.
> It also support dynamic mapping of the CSI-2 virtual channels to the
> associated pixel grabbers, but that isn't allowed at the moment either.
>
> Acked-by: Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx>
> Signed-off-by: Maxime Ripard <maxime.ripard@xxxxxxxxxxx>
> ---
> drivers/media/platform/Kconfig | 1 +
> drivers/media/platform/Makefile | 2 +
> drivers/media/platform/cadence/Kconfig | 17 +
> drivers/media/platform/cadence/Makefile | 1 +
> drivers/media/platform/cadence/cdns-csi2rx.c | 465 ++++++++++++++++++++++++
> 5 files changed, 486 insertions(+)
> create mode 100644 drivers/media/platform/cadence/Kconfig
> create mode 100644 drivers/media/platform/cadence/Makefile
> create mode 100644 drivers/media/platform/cadence/cdns-csi2rx.c
>
> diff --git a/drivers/media/platform/Kconfig b/drivers/media/platform/Kconfig
> index fd0c99859d6f..6e790a317fbc 100644
> --- a/drivers/media/platform/Kconfig
> +++ b/drivers/media/platform/Kconfig
> @@ -26,6 +26,7 @@ config VIDEO_VIA_CAMERA
> #
> # Platform multimedia device configuration
> #
> +source "drivers/media/platform/cadence/Kconfig"
>
> source "drivers/media/platform/davinci/Kconfig"
>
> diff --git a/drivers/media/platform/Makefile
> b/drivers/media/platform/Makefile index 003b0bb2cddf..1cd2984c55d1 100644
> --- a/drivers/media/platform/Makefile
> +++ b/drivers/media/platform/Makefile
> @@ -3,6 +3,8 @@
> # Makefile for the video capture/playback device drivers.
> #
>
> +obj-$(CONFIG_VIDEO_CADENCE) += cadence/
> +
> obj-$(CONFIG_VIDEO_M32R_AR_M64278) += arv.o
>
> obj-$(CONFIG_VIDEO_VIA_CAMERA) += via-camera.o
> diff --git a/drivers/media/platform/cadence/Kconfig
> b/drivers/media/platform/cadence/Kconfig new file mode 100644
> index 000000000000..18f061e5cbd1
> --- /dev/null
> +++ b/drivers/media/platform/cadence/Kconfig
> @@ -0,0 +1,17 @@
> +config VIDEO_CADENCE
> + bool "Cadence Video Devices"
> +
> +if VIDEO_CADENCE
> +
> +config VIDEO_CADENCE_CSI2RX
> + tristate "Cadence MIPI-CSI2 RX Controller"
> + depends on MEDIA_CONTROLLER
> + depends on VIDEO_V4L2_SUBDEV_API
> + select V4L2_FWNODE
> + help
> + Support for the Cadence MIPI CSI2 Receiver controller.
> +
> + To compile this driver as a module, choose M here: the module will be
> + called cdns-csi2rx.
> +
> +endif
> diff --git a/drivers/media/platform/cadence/Makefile
> b/drivers/media/platform/cadence/Makefile new file mode 100644
> index 000000000000..99a4086b7448
> --- /dev/null
> +++ b/drivers/media/platform/cadence/Makefile
> @@ -0,0 +1 @@
> +obj-$(CONFIG_VIDEO_CADENCE_CSI2RX) += cdns-csi2rx.o
> diff --git a/drivers/media/platform/cadence/cdns-csi2rx.c
> b/drivers/media/platform/cadence/cdns-csi2rx.c new file mode 100644
> index 000000000000..353a9b10197b
> --- /dev/null
> +++ b/drivers/media/platform/cadence/cdns-csi2rx.c
> @@ -0,0 +1,465 @@
> +// SPDX-License-Identifier: GPL-2.0+
> +/*
> + * Driver for Cadence MIPI-CSI2 RX Controller v1.3
> + *
> + * Copyright (C) 2017 Cadence Design Systems Inc.
> + */
> +
> +#include <linux/atomic.h>
> +#include <linux/clk.h>
> +#include <linux/delay.h>
> +#include <linux/io.h>
> +#include <linux/module.h>
> +#include <linux/of.h>
> +#include <linux/of_graph.h>
> +#include <linux/phy/phy.h>
> +#include <linux/platform_device.h>
> +#include <linux/slab.h>
> +
> +#include <media/v4l2-ctrls.h>
> +#include <media/v4l2-device.h>
> +#include <media/v4l2-fwnode.h>
> +#include <media/v4l2-subdev.h>
> +
> +#define CSI2RX_DEVICE_CFG_REG 0x000
> +
> +#define CSI2RX_SOFT_RESET_REG 0x004
> +#define CSI2RX_SOFT_RESET_PROTOCOL BIT(1)
> +#define CSI2RX_SOFT_RESET_FRONT BIT(0)
> +
> +#define CSI2RX_STATIC_CFG_REG 0x008
> +#define CSI2RX_STATIC_CFG_DLANE_MAP(llane, plane) ((plane) << (16 +
(llane)
> * 4)) +#define CSI2RX_STATIC_CFG_LANES_MASK GENMASK(11, 8)
> +
> +#define CSI2RX_STREAM_BASE(n) (((n) + 1) * 0x100)
> +
> +#define CSI2RX_STREAM_CTRL_REG(n) (CSI2RX_STREAM_BASE(n) + 0x000)
> +#define CSI2RX_STREAM_CTRL_START BIT(0)
> +
> +#define CSI2RX_STREAM_DATA_CFG_REG(n) (CSI2RX_STREAM_BASE(n) + 0x008)
> +#define CSI2RX_STREAM_DATA_CFG_EN_VC_SELECT BIT(31)
> +#define CSI2RX_STREAM_DATA_CFG_VC_SELECT(n) BIT((n) + 16)
> +
> +#define CSI2RX_STREAM_CFG_REG(n) (CSI2RX_STREAM_BASE(n) + 0x00c)
> +#define CSI2RX_STREAM_CFG_FIFO_MODE_LARGE_BUF (1 << 8)
> +
> +#define CSI2RX_LANES_MAX 4
> +#define CSI2RX_STREAMS_MAX 4
> +
> +enum csi2rx_pads {
> + CSI2RX_PAD_SINK,
> + CSI2RX_PAD_SOURCE_STREAM0,
> + CSI2RX_PAD_SOURCE_STREAM1,
> + CSI2RX_PAD_SOURCE_STREAM2,
> + CSI2RX_PAD_SOURCE_STREAM3,
> + CSI2RX_PAD_MAX,
> +};
> +
> +struct csi2rx_priv {
> + struct device *dev;
> + atomic_t count;
> +
> + void __iomem *base;
> + struct clk *sys_clk;
> + struct clk *p_clk;
> + struct clk *pixel_clk[CSI2RX_STREAMS_MAX];
> + struct phy *dphy;
> +
> + u8 lanes[CSI2RX_LANES_MAX];
> + u8 num_lanes;
> + u8 max_lanes;
> + u8 max_streams;
> + bool has_internal_dphy;
> +
> + struct v4l2_subdev subdev;
> + struct v4l2_async_notifier notifier;
> + struct media_pad pads[CSI2RX_PAD_MAX];
> +
> + /* Remote source */
> + struct v4l2_async_subdev asd;
> + struct v4l2_subdev *source_subdev;
> + int source_pad;
> +};
> +
> +static inline
> +struct csi2rx_priv *v4l2_subdev_to_csi2rx(struct v4l2_subdev *subdev)
> +{
> + return container_of(subdev, struct csi2rx_priv, subdev);
> +}
> +
> +static void csi2rx_reset(struct csi2rx_priv *csi2rx)
> +{
> + writel(CSI2RX_SOFT_RESET_PROTOCOL | CSI2RX_SOFT_RESET_FRONT,
> + csi2rx->base + CSI2RX_SOFT_RESET_REG);
> +
> + usleep_range(10, 20);
That's a very short range. Is it worth getting the scheduler involved,
wouldn't udelay() be better ?
> + writel(0, csi2rx->base + CSI2RX_SOFT_RESET_REG);
> +}
> +
> +static int csi2rx_start(struct csi2rx_priv *csi2rx)
> +{
> + unsigned int i;
> + unsigned long lanes_used = 0;
> + u32 reg;
> + int ret;
> +
> + /*
> + * We're not the first users, there's no need to enable the
> + * whole controller.
> + */
> + if (atomic_inc_return(&csi2rx->count) > 1)
> + return 0;
You have a race condition here. If s_stream(1) is called twice in parallel,
the second call will return immediately while the first call might still be
starting the stream. The caller could thus proceed with the stream not enabled
yet, which in some cases could be an issue. If you want to fix this you'll
need a mutex.
> + clk_prepare_enable(csi2rx->p_clk);
No need to check for errors ?
> + csi2rx_reset(csi2rx);
> +
> + reg = csi2rx->num_lanes << 8;
> + for (i = 0; i < csi2rx->num_lanes; i++) {
> + reg |= CSI2RX_STATIC_CFG_DLANE_MAP(i, csi2rx->lanes[i]);
> + set_bit(csi2rx->lanes[i], &lanes_used);
> + }
> +
> + /*
> + * Even the unused lanes need to be mapped. In order to avoid
> + * to map twice to the same physical lane, keep the lanes used
> + * in the previous loop, and only map unused physical lanes to
> + * the rest of our logical lanes.
> + */
> + for (i = csi2rx->num_lanes; i < csi2rx->max_lanes; i++) {
> + unsigned int idx = find_first_zero_bit(&lanes_used,
> + sizeof(lanes_used));
> + set_bit(idx, &lanes_used);
> + reg |= CSI2RX_STATIC_CFG_DLANE_MAP(i, i + 1);
> + }
> +
> + writel(reg, csi2rx->base + CSI2RX_STATIC_CFG_REG);
> +
> + ret = v4l2_subdev_call(csi2rx->source_subdev, video, s_stream, true);
> + if (ret)
> + return ret;
Doesn't the CSI-2 receiver need to be fully started before the sensor ?
Otherwise I believe it could have trouble synchronizing to the LP-11 state.
That's a PHY issue though, so maybe you plan to fix this later when adding
support for PHYs ?
> + /*
> + * Create a static mapping between the CSI virtual channels
> + * and the output stream.
> + *
> + * This should be enhanced, but v4l2 lacks the support for
> + * changing that mapping dynamically.
> + *
> + * We also cannot enable and disable independant streams here,
> + * hence the reference counting.
> + */
If you start all streams in one go, will s_stream(1) be called multiple times
? If not, you could possibly skip the whole reference counting and avoid
locking.
> + for (i = 0; i < csi2rx->max_streams; i++) {
> + clk_prepare_enable(csi2rx->pixel_clk[i]);
> +
> + writel(CSI2RX_STREAM_CFG_FIFO_MODE_LARGE_BUF,
> + csi2rx->base + CSI2RX_STREAM_CFG_REG(i));
> +
> + writel(CSI2RX_STREAM_DATA_CFG_EN_VC_SELECT |
> + CSI2RX_STREAM_DATA_CFG_VC_SELECT(i),
> + csi2rx->base + CSI2RX_STREAM_DATA_CFG_REG(i));
> +
> + writel(CSI2RX_STREAM_CTRL_START,
> + csi2rx->base + CSI2RX_STREAM_CTRL_REG(i));
> + }
> +
> + clk_prepare_enable(csi2rx->sys_clk);
> +
> + clk_disable_unprepare(csi2rx->p_clk);
> +
> + return 0;
> +}
> +
> +static int csi2rx_stop(struct csi2rx_priv *csi2rx)
> +{
> + unsigned int i;
> +
> + /*
> + * Let the last user turn off the lights
> + */
> + if (!atomic_dec_and_test(&csi2rx->count))
> + return 0;
> +
> + clk_prepare_enable(csi2rx->p_clk);
> +
> + for (i = 0; i < csi2rx->max_streams; i++) {
> + writel(0, csi2rx->base + CSI2RX_STREAM_CTRL_REG(i));
> +
> + clk_disable_unprepare(csi2rx->pixel_clk[i]);
> + }
> +
> + clk_disable_unprepare(csi2rx->p_clk);
> +
> + return v4l2_subdev_call(csi2rx->source_subdev, video, s_stream, false);
> +}
> +
> +static int csi2rx_s_stream(struct v4l2_subdev *sd, int enable)
> +{
> + struct csi2rx_priv *csi2rx = v4l2_subdev_to_csi2rx(sd);
> + int ret;
> +
> + if (enable)
> + ret = csi2rx_start(csi2rx);
> + else
> + ret = csi2rx_stop(csi2rx);
Nitpicking, you could
if (enable)
return csi2rx_start(csi2rx);
else
return csi2tx_start(csi2tx);
> +
> + return ret;
> +}
> +
> +static const struct v4l2_subdev_video_ops csi2rx_video_ops = {
> + .s_stream = csi2rx_s_stream,
> +};
> +
> +static const struct v4l2_subdev_ops csi2rx_subdev_ops = {
> + .video = &csi2rx_video_ops,
> +};
> +
> +static int csi2rx_async_bound(struct v4l2_async_notifier *notifier,
> + struct v4l2_subdev *s_subdev,
> + struct v4l2_async_subdev *asd)
> +{
> + struct v4l2_subdev *subdev = notifier->sd;
> + struct csi2rx_priv *csi2rx = v4l2_subdev_to_csi2rx(subdev);
> +
> + csi2rx->source_pad = media_entity_get_fwnode_pad(&s_subdev->entity,
> + s_subdev->fwnode,
> + MEDIA_PAD_FL_SOURCE);
> + if (csi2rx->source_pad < 0) {
> + dev_err(csi2rx->dev, "Couldn't find output pad for subdev %s\n",
> + s_subdev->name);
> + return csi2rx->source_pad;
> + }
> +
> + csi2rx->source_subdev = s_subdev;
> +
> + dev_dbg(csi2rx->dev, "Bound %s pad: %d\n", s_subdev->name,
> + csi2rx->source_pad);
> +
> + return media_create_pad_link(&csi2rx->source_subdev->entity,
> + csi2rx->source_pad,
> + &csi2rx->subdev.entity, 0,
> + MEDIA_LNK_FL_ENABLED |
> + MEDIA_LNK_FL_IMMUTABLE);
> +}
> +
> +static const struct v4l2_async_notifier_operations csi2rx_notifier_ops = {
> + .bound = csi2rx_async_bound,
> +};
> +
> +static int csi2rx_get_resources(struct csi2rx_priv *csi2rx,
> + struct platform_device *pdev)
> +{
> + struct resource *res;
> + unsigned char i;
> + u32 reg;
How about naming the variable dev_cfg to be a bit more explicit ?
> +
> + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + csi2rx->base = devm_ioremap_resource(&pdev->dev, res);
> + if (IS_ERR(csi2rx->base))
> + return PTR_ERR(csi2rx->base);
> +
> + csi2rx->sys_clk = devm_clk_get(&pdev->dev, "sys_clk");
> + if (IS_ERR(csi2rx->sys_clk)) {
> + dev_err(&pdev->dev, "Couldn't get sys clock\n");
> + return PTR_ERR(csi2rx->sys_clk);
> + }
> +
> + csi2rx->p_clk = devm_clk_get(&pdev->dev, "p_clk");
> + if (IS_ERR(csi2rx->p_clk)) {
> + dev_err(&pdev->dev, "Couldn't get P clock\n");
> + return PTR_ERR(csi2rx->p_clk);
> + }
> +
> + csi2rx->dphy = devm_phy_optional_get(&pdev->dev, "dphy");
> + if (IS_ERR(csi2rx->dphy)) {
> + dev_err(&pdev->dev, "Couldn't get external D-PHY\n");
> + return PTR_ERR(csi2rx->dphy);
> + }
> +
> + /*
> + * FIXME: Once we'll have external D-PHY support, the check
> + * will need to be removed.
> + */
> + if (csi2rx->dphy) {
> + dev_err(&pdev->dev, "External D-PHY not supported yet\n");
> + return -EINVAL;
> + }
> +
> + clk_prepare_enable(csi2rx->p_clk);
> + reg = readl(csi2rx->base + CSI2RX_DEVICE_CFG_REG);
> + clk_disable_unprepare(csi2rx->p_clk);
> +
> + csi2rx->max_lanes = (reg & 7);
No need for outer parentheses.
> + if (csi2rx->max_lanes > CSI2RX_LANES_MAX) {
> + dev_err(&pdev->dev, "Invalid number of lanes: %u\n",
> + csi2rx->max_lanes);
> + return -EINVAL;
> + }
> +
> + csi2rx->max_streams = ((reg >> 4) & 7);
Same here.
> + if (csi2rx->max_streams > CSI2RX_STREAMS_MAX) {
> + dev_err(&pdev->dev, "Invalid number of streams: %u\n",
> + csi2rx->max_streams);
> + return -EINVAL;
> + }
> +
> + csi2rx->has_internal_dphy = (reg & BIT(3)) ? true : false;
Same here.
> +
> + /*
> + * FIXME: Once we'll have internal D-PHY support, the check
> + * will need to be removed.
> + */
> + if (csi2rx->has_internal_dphy) {
> + dev_err(&pdev->dev, "Internal D-PHY not supported yet\n");
> + return -EINVAL;
> + }
> +
> + for (i = 0; i < csi2rx->max_streams; i++) {
> + char clk_name[16];
> +
> + snprintf(clk_name, sizeof(clk_name), "pixel_if%u_clk", i);
> + csi2rx->pixel_clk[i] = devm_clk_get(&pdev->dev, clk_name);
> + if (IS_ERR(csi2rx->pixel_clk[i])) {
> + dev_err(&pdev->dev, "Couldn't get clock %s\n", clk_name);
> + return PTR_ERR(csi2rx->pixel_clk[i]);
> + }
> + }
> +
> + return 0;
> +}
> +
> +static int csi2rx_parse_dt(struct csi2rx_priv *csi2rx)
> +{
> + struct v4l2_fwnode_endpoint v4l2_ep;
> + struct fwnode_handle *fwh;
> + struct device_node *ep;
> + int ret;
> +
> + ep = of_graph_get_endpoint_by_regs(csi2rx->dev->of_node, 0, 0);
> + if (!ep)
> + return -EINVAL;
> +
> + fwh = of_fwnode_handle(ep);
> + ret = v4l2_fwnode_endpoint_parse(fwh, &v4l2_ep);
> + if (ret) {
> + dev_err(csi2rx->dev, "Could not parse v4l2 endpoint\n");
> + of_node_put(ep);
> + return ret;
> + }
> +
> + if (v4l2_ep.bus_type != V4L2_MBUS_CSI2) {
> + dev_err(csi2rx->dev, "Unsupported media bus type: 0x%x\n",
> + v4l2_ep.bus_type);
> + of_node_put(ep);
> + return -EINVAL;
> + }
> +
> + memcpy(csi2rx->lanes, v4l2_ep.bus.mipi_csi2.data_lanes,
> + sizeof(csi2rx->lanes));
> + csi2rx->num_lanes = v4l2_ep.bus.mipi_csi2.num_data_lanes;
> + if (csi2rx->num_lanes > csi2rx->max_lanes) {
> + dev_err(csi2rx->dev, "Unsupported number of data-lanes: %d\n",
> + csi2rx->num_lanes);
> + of_node_put(ep);
> + return -EINVAL;
> + }
> +
> + csi2rx->asd.match.fwnode.fwnode =
> fwnode_graph_get_remote_port_parent(fwh);
> + csi2rx->asd.match_type = V4L2_ASYNC_MATCH_FWNODE;
> + of_node_put(ep);
> +
> + csi2rx->notifier.subdevs = devm_kzalloc(csi2rx->dev,
> + sizeof(*csi2rx->notifier.subdevs),
> + GFP_KERNEL);
> + if (!csi2rx->notifier.subdevs)
> + return -ENOMEM;
> +
> + csi2rx->notifier.subdevs[0] = &csi2rx->asd;
> + csi2rx->notifier.num_subdevs = 1;
> + csi2rx->notifier.ops = &csi2rx_notifier_ops;
> +
> + return v4l2_async_subdev_notifier_register(&csi2rx->subdev,
> + &csi2rx->notifier);
> +}
> +
> +static int csi2rx_probe(struct platform_device *pdev)
> +{
> + struct csi2rx_priv *csi2rx;
> + unsigned int i;
> + int ret;
> +
> + csi2rx = kzalloc(sizeof(*csi2rx), GFP_KERNEL);
> + if (!csi2rx)
> + return -ENOMEM;
> + platform_set_drvdata(pdev, csi2rx);
> + csi2rx->dev = &pdev->dev;
> +
> + ret = csi2rx_get_resources(csi2rx, pdev);
> + if (ret)
> + goto err_free_priv;
> +
> + ret = csi2rx_parse_dt(csi2rx);
> + if (ret)
> + goto err_free_priv;
> +
> + csi2rx->subdev.owner = THIS_MODULE;
> + csi2rx->subdev.dev = &pdev->dev;
> + v4l2_subdev_init(&csi2rx->subdev, &csi2rx_subdev_ops);
> + v4l2_set_subdevdata(&csi2rx->subdev, &pdev->dev);
> + snprintf(csi2rx->subdev.name, V4L2_SUBDEV_NAME_SIZE, "%s.%s",
> + KBUILD_MODNAME, dev_name(&pdev->dev));
> +
> + /* Create our media pads */
> + csi2rx->subdev.entity.function = MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER;
MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER is documented as
- Video pixel formatter. An entity capable of pixel formatting
must have at least one sink pad and one source pad. Read
pixel formatters read pixels from memory and perform a subset
of unpacking, cropping, color keying, alpha multiplication
and pixel encoding conversion. Write pixel formatters perform
a subset of dithering, pixel encoding conversion and packing
and write pixels to memory.
I don't think it's a good match in this case. Wouldn't
MEDIA_ENT_F_VID_IF_BRIDGE be better ?
- Video interface bridge. A video interface bridge entity must have at
least one sink pad and at least one source pad. It receives video
frames on its sink pad from an input video bus of one type (HDMI,
eDP, MIPI CSI-2, ...), and outputs them on its source pad to an
output video bus of another type (eDP, MIPI CSI-2, parallel, ...).
(I'm not sure I like the name "bridge" too much in this context, but that's a
separate issue)
> + csi2rx->pads[CSI2RX_PAD_SINK].flags = MEDIA_PAD_FL_SINK;
> + for (i = CSI2RX_PAD_SOURCE_STREAM0; i < CSI2RX_PAD_MAX; i++)
> + csi2rx->pads[i].flags = MEDIA_PAD_FL_SOURCE;
> +
> + ret = media_entity_pads_init(&csi2rx->subdev.entity, CSI2RX_PAD_MAX,
> + csi2rx->pads);
> + if (ret)
> + goto err_free_priv;
> +
> + ret = v4l2_async_register_subdev(&csi2rx->subdev);
> + if (ret < 0)
> + goto err_free_priv;
> +
> + dev_info(&pdev->dev,
> + "Probed CSI2RX with %u/%u lanes, %u streams, %s D-PHY\n",
> + csi2rx->num_lanes, csi2rx->max_lanes, csi2rx->max_streams,
> + csi2rx->has_internal_dphy ? "internal" : "no");
> +
> + return 0;
> +
> +err_free_priv:
> + kfree(csi2rx);
> + return ret;
> +}
[snip]
--
Regards,
Laurent Pinchart
