[PATCH v4 00/12] media: cadence,ti: CSI2RX Multistream Support

[PATCH v4 00/12] media: cadence,ti: CSI2RX Multistream Support

[Rishikesh Donadkar]

This series adds multi-stream support for Cadence CSI2RX and TI CSI2RX
SHIM drivers.

PATCH 1-6:  Support multiple DMA contexts/video nodes in TI CSI2RX
PATCH 7-8:  Use get_frame_desc to propagate virtual channel
            information across Cadence and TI CSI-RX subdevs
PATCH 9-10: Use new multi-stream APIs across the drivers to support
            multiplexed cameras from sources like UB960 (FPDLink)
PATCH 11:   Optimize stream on by submitting all queued buffers to DMA
PATCH 12:   Change the drain architecture to support multi-stream

Testing for this series has been done on top of media tree with 4x IMX219
camera modules connected to TI's AM62A using V3 Link fusion mini board.

Overlay and defconfig changes for the same can be found below:
https://github.com/RISHI27-dot/linux/commits/u/multistream_v4

v4l2-compliance results:
https://gist.github.com/Rishikesh-D/4ddf957da2f62191eefe7a9e5ff212ab

---
Changes in v4:

[PATCH 01/13] dt-bindings: media: ti,j721e-csi2rx-shim: Support 32 dma chans
  - No change
[PATCH 02/13] media: ti: j721e-csi2rx: separate out device and context
  - Add ctx identifier in the dev_err() message
  - No change
[PATCH 04/13] media: ti: j721e-csi2rx: prepare SHIM code for multiple contexts
  - Reduced the name string lenght from 32 chars to 5 chars
[PATCH 05/13] media: ti: j721e-csi2rx: add a subdev for the core device
  - Add .enum_mbus_code callback
  - Replace statically allocated struct with a global static const struct
    v4l2_mbus_framefmt and used that in the _init_state() function
[PATCH 06/13] media: ti: j721e-csi2rx: get number of contexts from device tree
  - Fix the drain buffer being leaked
  - If the shows more number of ctx than the TI_CSI2RX_MAX_CTX, return an error
    instead of warning
[PATCH 07/13] media: cadence: csi2rx: add get_frame_desc wrapper
  - No change
[PATCH 08/13] media: ti: j721e-csi2rx: add support for processing virtual channels
  - Call ti_csi2rx_get_vc() only once on first stream start and cache the VC data in
    the driver, use the corresponding VC in all subsequent stream starts.
[PATCH 09/13] media: cadence: csi2rx: Use new enable stream APIs
[PATCH 10/13] media: cadence: csi2rx: Enable multi-stream support
  - Squash the above two patches into
    [PATCH v4 09/12] media: cadence: csi2rx: add multistream support
  - Use already obtained csi2rx->source_pad in enable_streams() and
    disable_streams() call
  - Update commit message with the reason for using a custom helper for s_stream
    instead of v4l2_subdev_s_stream_helper()
  - Use v4l2_get_link_freq() variant that takes pad of the source as its first
    argument instead of the one that takes v4l2_ctrl_handler
  - Call v4l2_get_link_freq() with bpp = 0 to prevent fallback to V4L2_CID_PIXEL_RATE
    in multi-stream case
  - Use lock guards to simplify error handling
  - Call csi2rx_update_vc_select() at first stream start before enabling the controller
[PATCH 11/13] media: ti: j721e-csi2rx: add multistream support
  - No change
[PATCH 12/13] media: ti: j721e-csi2rx: Submit all available buffers
  - No change
[PATCH 13/13] media: ti: j721e-csi2rx: Change the drain architecture for multistream
  - Fix checkpatch warning
  - Change commit message to give a better description of the patch

Link to (v3):
  https://lore.kernel.org/all/20250417065554.437541-1-r-donadkar@ti.com/

Changes in v3:

- Drop [PATCH v2 01/13] media: cadence: csi2rx: Support runtime PM from
  v2, support for runtime PM will be added in a separate series:
  https://lore.kernel.org/all/20250224-ti_csi_pm-v1-0-8f8c29ef646d@ideasonboard.com/
- Change the drain architecture to prevent FIFO overflow in multistream
  usecases.
- With the new drain architecture, we don't need the the driver to wait
  for userspace to start streaming on all "actively routed" video nodes
  before starting streaming on the source. So, revert back to the capture
  architecture where streams can be started and stopped independent
  to each other.

Link to (v2):
  https://lore.kernel.org/r/20240627-multistream-v2-0-6ae96c54c1c3@ti.com

Changes in v2:

- Change the multi-camera capture architecture to be similar to that of
  Tomi's RPi5 FE series, where the driver will wait for userspace to
  start streaming on all "actively routed" video nodes before starting
  streaming on the source. This simplifies things a lot from the HW
  perspective, which might run into deadlocks due to a shared FIFO
  between multiple DMA channels.

- Drop a few fixes that were posted separately and are already merged
- Fix dtschema warnings reported by Rob on [02/13]
- Fix warnings for uninitialized `used_vc` variable in cdns-csi2rx.c
- Return -EBUSY if someone updates routes for j721e-csi2rx subdev while
  streaming
- Only allow single-streams to be routed to the source pads (linked to
  video nodes) of the j721e-csi2rx device
- Squash the patches marked "SQUASH" in the v1 RFC series

Link to RFC (v1):
https://lore.kernel.org/r/20240222-multistream-v1-0-1837ed916eeb@ti.com


Jai Luthra (7):
  dt-bindings: media: ti,j721e-csi2rx-shim: Support 32 dma chans
  media: ti: j721e-csi2rx: separate out device and context
  media: ti: j721e-csi2rx: add a subdev for the core device
  media: ti: j721e-csi2rx: add support for processing virtual channels
  media: cadence: csi2rx: add multistream support
  media: ti: j721e-csi2rx: add multistream support
  media: ti: j721e-csi2rx: Submit all available buffers

Pratyush Yadav (4):
  media: ti: j721e-csi2rx: prepare SHIM code for multiple contexts
  media: ti: j721e-csi2rx: allocate DMA channel based on context index
  media: ti: j721e-csi2rx: get number of contexts from device tree
  media: cadence: csi2rx: add get_frame_desc wrapper

Rishikesh Donadkar (1):
  media: ti: j721e-csi2rx: Change the drain architecture for multistream

 .../bindings/media/ti,j721e-csi2rx-shim.yaml  |  39 +-
 drivers/media/platform/cadence/cdns-csi2rx.c  | 372 +++++--
 .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 958 +++++++++++++-----
 3 files changed, 1025 insertions(+), 344 deletions(-)

-- 
2.34.1

[PATCH v4 01/12] dt-bindings: media: ti,j721e-csi2rx-shim: Support 32 dma chans

[Rishikesh Donadkar]

From: Jai Luthra <j-luthra@ti.com>

The CSI2RX SHIM IP can support 32x DMA channels. These can be used to
split incoming "streams" of data on the CSI-RX port, distinguished by
MIPI Virtual Channel (or Data Type), into different locations in memory.

Actual number of DMA channels allocated to CSI-RX is dependent on the
usecase, and can be modified using the K3 Resource Partitioning tool [1].
So set the minimum channels as 1 and maximum as 32.

Link: https://software-dl.ti.com/processor-sdk-linux/esd/AM62X/10_00_07_04/exports/docs/linux/How_to_Guides/Host/K3_Resource_Partitioning_Tool.html [1]
Link: https://www.ti.com/lit/pdf/spruiv7
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Rob Herring (Arm) <robh@kernel.org>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../bindings/media/ti,j721e-csi2rx-shim.yaml  | 39 +++++++++++++++++--
 1 file changed, 36 insertions(+), 3 deletions(-)

diff --git a/Documentation/devicetree/bindings/media/ti,j721e-csi2rx-shim.yaml b/Documentation/devicetree/bindings/media/ti,j721e-csi2rx-shim.yaml
index b9f033f2f3ce4..bf62998b0445a 100644
--- a/Documentation/devicetree/bindings/media/ti,j721e-csi2rx-shim.yaml
+++ b/Documentation/devicetree/bindings/media/ti,j721e-csi2rx-shim.yaml
@@ -20,11 +20,44 @@ properties:
     const: ti,j721e-csi2rx-shim
 
   dmas:
-    maxItems: 1
+    minItems: 1
+    maxItems: 32
 
   dma-names:
+    minItems: 1
     items:
       - const: rx0
+      - const: rx1
+      - const: rx2
+      - const: rx3
+      - const: rx4
+      - const: rx5
+      - const: rx6
+      - const: rx7
+      - const: rx8
+      - const: rx9
+      - const: rx10
+      - const: rx11
+      - const: rx12
+      - const: rx13
+      - const: rx14
+      - const: rx15
+      - const: rx16
+      - const: rx17
+      - const: rx18
+      - const: rx19
+      - const: rx20
+      - const: rx21
+      - const: rx22
+      - const: rx23
+      - const: rx24
+      - const: rx25
+      - const: rx26
+      - const: rx27
+      - const: rx28
+      - const: rx29
+      - const: rx30
+      - const: rx31
 
   reg:
     maxItems: 1
@@ -62,8 +95,8 @@ examples:
 
     ti_csi2rx0: ticsi2rx@4500000 {
         compatible = "ti,j721e-csi2rx-shim";
-        dmas = <&main_udmap 0x4940>;
-        dma-names = "rx0";
+        dmas = <&main_udmap 0x4940>, <&main_udmap 0x4941>;
+        dma-names = "rx0", "rx1";
         reg = <0x4500000 0x1000>;
         power-domains = <&k3_pds 26 TI_SCI_PD_EXCLUSIVE>;
         #address-cells = <1>;
-- 
2.34.1

[PATCH v4 02/12] media: ti: j721e-csi2rx: separate out device and context

[Rishikesh Donadkar]

From: Jai Luthra <j-luthra@ti.com>

The TI CSI2RX wrapper has two parts: the main device and the DMA
contexts. The driver was originally written with single camera capture
in mind, so only one DMA context was needed. For the sake of simplicity,
the context specific stuff was not modeled different to the main device.

To enable multiplexed stream capture, the contexts need to be separated
out from the main device. Create a struct ti_csi2rx_ctx that holds the
DMA context specific things. Separate out functions handling the device
and context related functionality.

Co-developed-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 412 ++++++++++--------
 1 file changed, 228 insertions(+), 184 deletions(-)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index 6412a00be8eab..36cde2e87aabb 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -40,6 +40,8 @@
 #define SHIM_PSI_CFG0_DST_TAG		GENMASK(31, 16)
 
 #define PSIL_WORD_SIZE_BYTES		16
+#define TI_CSI2RX_NUM_CTX		1
+
 /*
  * There are no hard limits on the width or height. The DMA engine can handle
  * all sizes. The max width and height are arbitrary numbers for this driver.
@@ -64,7 +66,7 @@ struct ti_csi2rx_buffer {
 	/* Common v4l2 buffer. Must be first. */
 	struct vb2_v4l2_buffer		vb;
 	struct list_head		list;
-	struct ti_csi2rx_dev		*csi;
+	struct ti_csi2rx_ctx		*ctx;
 };
 
 enum ti_csi2rx_dma_state {
@@ -84,29 +86,37 @@ struct ti_csi2rx_dma {
 	 * Queue of buffers submitted to DMA engine.
 	 */
 	struct list_head		submitted;
-	/* Buffer to drain stale data from PSI-L endpoint */
-	struct {
-		void			*vaddr;
-		dma_addr_t		paddr;
-		size_t			len;
-	} drain;
+};
+
+struct ti_csi2rx_dev;
+
+struct ti_csi2rx_ctx {
+	struct ti_csi2rx_dev		*csi;
+	struct video_device		vdev;
+	struct vb2_queue		vidq;
+	struct mutex			mutex; /* To serialize ioctls. */
+	struct v4l2_format		v_fmt;
+	struct ti_csi2rx_dma		dma;
+	u32				sequence;
+	u32				idx;
 };
 
 struct ti_csi2rx_dev {
 	struct device			*dev;
 	void __iomem			*shim;
 	struct v4l2_device		v4l2_dev;
-	struct video_device		vdev;
 	struct media_device		mdev;
 	struct media_pipeline		pipe;
 	struct media_pad		pad;
 	struct v4l2_async_notifier	notifier;
 	struct v4l2_subdev		*source;
-	struct vb2_queue		vidq;
-	struct mutex			mutex; /* To serialize ioctls. */
-	struct v4l2_format		v_fmt;
-	struct ti_csi2rx_dma		dma;
-	u32				sequence;
+	struct ti_csi2rx_ctx		ctx[TI_CSI2RX_NUM_CTX];
+	/* Buffer to drain stale data from PSI-L endpoint */
+	struct {
+		void			*vaddr;
+		dma_addr_t		paddr;
+		size_t			len;
+	} drain;
 };
 
 static const struct ti_csi2rx_fmt ti_csi2rx_formats[] = {
@@ -212,7 +222,7 @@ static const struct ti_csi2rx_fmt ti_csi2rx_formats[] = {
 };
 
 /* Forward declaration needed by ti_csi2rx_dma_callback. */
-static int ti_csi2rx_start_dma(struct ti_csi2rx_dev *csi,
+static int ti_csi2rx_start_dma(struct ti_csi2rx_ctx *ctx,
 			       struct ti_csi2rx_buffer *buf);
 
 static const struct ti_csi2rx_fmt *find_format_by_fourcc(u32 pixelformat)
@@ -302,7 +312,7 @@ static int ti_csi2rx_enum_fmt_vid_cap(struct file *file, void *priv,
 static int ti_csi2rx_g_fmt_vid_cap(struct file *file, void *prov,
 				   struct v4l2_format *f)
 {
-	struct ti_csi2rx_dev *csi = video_drvdata(file);
+	struct ti_csi2rx_ctx *csi = video_drvdata(file);
 
 	*f = csi->v_fmt;
 
@@ -333,7 +343,7 @@ static int ti_csi2rx_try_fmt_vid_cap(struct file *file, void *priv,
 static int ti_csi2rx_s_fmt_vid_cap(struct file *file, void *priv,
 				   struct v4l2_format *f)
 {
-	struct ti_csi2rx_dev *csi = video_drvdata(file);
+	struct ti_csi2rx_ctx *csi = video_drvdata(file);
 	struct vb2_queue *q = &csi->vidq;
 	int ret;
 
@@ -419,25 +429,33 @@ static int csi_async_notifier_bound(struct v4l2_async_notifier *notifier,
 static int csi_async_notifier_complete(struct v4l2_async_notifier *notifier)
 {
 	struct ti_csi2rx_dev *csi = dev_get_drvdata(notifier->v4l2_dev->dev);
-	struct video_device *vdev = &csi->vdev;
-	int ret;
+	int ret, i;
 
-	ret = video_register_device(vdev, VFL_TYPE_VIDEO, -1);
-	if (ret)
-		return ret;
-
-	ret = v4l2_create_fwnode_links_to_pad(csi->source, &csi->pad,
-					      MEDIA_LNK_FL_IMMUTABLE | MEDIA_LNK_FL_ENABLED);
+	for (i = 0; i < TI_CSI2RX_NUM_CTX; i++) {
+		struct ti_csi2rx_ctx *ctx = &csi->ctx[i];
+		struct video_device *vdev = &ctx->vdev;
 
-	if (ret) {
-		video_unregister_device(vdev);
-		return ret;
+		ret = video_register_device(vdev, VFL_TYPE_VIDEO, -1);
+		if (ret)
+			goto unregister_dev;
 	}
 
+	ret = v4l2_create_fwnode_links_to_pad(csi->source, &csi->pad,
+					      MEDIA_LNK_FL_IMMUTABLE |
+					      MEDIA_LNK_FL_ENABLED);
+	if (ret)
+		goto unregister_dev;
+
 	ret = v4l2_device_register_subdev_nodes(&csi->v4l2_dev);
 	if (ret)
-		video_unregister_device(vdev);
+		goto unregister_dev;
 
+	return 0;
+
+unregister_dev:
+	i--;
+	for (; i >= 0; i--)
+		video_unregister_device(&csi->ctx[i].vdev);
 	return ret;
 }
 
@@ -483,12 +501,13 @@ static int ti_csi2rx_notifier_register(struct ti_csi2rx_dev *csi)
 	return 0;
 }
 
-static void ti_csi2rx_setup_shim(struct ti_csi2rx_dev *csi)
+static void ti_csi2rx_setup_shim(struct ti_csi2rx_ctx *ctx)
 {
+	struct ti_csi2rx_dev *csi = ctx->csi;
 	const struct ti_csi2rx_fmt *fmt;
 	unsigned int reg;
 
-	fmt = find_format_by_fourcc(csi->v_fmt.fmt.pix.pixelformat);
+	fmt = find_format_by_fourcc(ctx->v_fmt.fmt.pix.pixelformat);
 
 	/* De-assert the pixel interface reset. */
 	reg = SHIM_CNTL_PIX_RST;
@@ -555,8 +574,9 @@ static void ti_csi2rx_drain_callback(void *param)
  * To prevent that stale data corrupting the subsequent transactions, it is
  * required to issue DMA requests to drain it out.
  */
-static int ti_csi2rx_drain_dma(struct ti_csi2rx_dev *csi)
+static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
 {
+	struct ti_csi2rx_dev *csi = ctx->csi;
 	struct dma_async_tx_descriptor *desc;
 	struct completion drain_complete;
 	dma_cookie_t cookie;
@@ -564,8 +584,8 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_dev *csi)
 
 	init_completion(&drain_complete);
 
-	desc = dmaengine_prep_slave_single(csi->dma.chan, csi->dma.drain.paddr,
-					   csi->dma.drain.len, DMA_DEV_TO_MEM,
+	desc = dmaengine_prep_slave_single(ctx->dma.chan, csi->drain.paddr,
+					   csi->drain.len, DMA_DEV_TO_MEM,
 					   DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
 	if (!desc) {
 		ret = -EIO;
@@ -580,11 +600,11 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_dev *csi)
 	if (ret)
 		goto out;
 
-	dma_async_issue_pending(csi->dma.chan);
+	dma_async_issue_pending(ctx->dma.chan);
 
 	if (!wait_for_completion_timeout(&drain_complete,
 					 msecs_to_jiffies(DRAIN_TIMEOUT_MS))) {
-		dmaengine_terminate_sync(csi->dma.chan);
+		dmaengine_terminate_sync(ctx->dma.chan);
 		dev_dbg(csi->dev, "DMA transfer timed out for drain buffer\n");
 		ret = -ETIMEDOUT;
 		goto out;
@@ -596,8 +616,8 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_dev *csi)
 static void ti_csi2rx_dma_callback(void *param)
 {
 	struct ti_csi2rx_buffer *buf = param;
-	struct ti_csi2rx_dev *csi = buf->csi;
-	struct ti_csi2rx_dma *dma = &csi->dma;
+	struct ti_csi2rx_ctx *ctx = buf->ctx;
+	struct ti_csi2rx_dma *dma = &ctx->dma;
 	unsigned long flags;
 
 	/*
@@ -605,7 +625,7 @@ static void ti_csi2rx_dma_callback(void *param)
 	 * hardware monitor registers.
 	 */
 	buf->vb.vb2_buf.timestamp = ktime_get_ns();
-	buf->vb.sequence = csi->sequence++;
+	buf->vb.sequence = ctx->sequence++;
 
 	spin_lock_irqsave(&dma->lock, flags);
 
@@ -617,8 +637,9 @@ static void ti_csi2rx_dma_callback(void *param)
 	while (!list_empty(&dma->queue)) {
 		buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer, list);
 
-		if (ti_csi2rx_start_dma(csi, buf)) {
-			dev_err(csi->dev, "Failed to queue the next buffer for DMA\n");
+		if (ti_csi2rx_start_dma(ctx, buf)) {
+			dev_err(ctx->csi->dev,
+				"Failed to queue the next buffer for DMA\n");
 			vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_ERROR);
 		} else {
 			list_move_tail(&buf->list, &dma->submitted);
@@ -631,17 +652,17 @@ static void ti_csi2rx_dma_callback(void *param)
 	spin_unlock_irqrestore(&dma->lock, flags);
 }
 
-static int ti_csi2rx_start_dma(struct ti_csi2rx_dev *csi,
+static int ti_csi2rx_start_dma(struct ti_csi2rx_ctx *ctx,
 			       struct ti_csi2rx_buffer *buf)
 {
 	unsigned long addr;
 	struct dma_async_tx_descriptor *desc;
-	size_t len = csi->v_fmt.fmt.pix.sizeimage;
+	size_t len = ctx->v_fmt.fmt.pix.sizeimage;
 	dma_cookie_t cookie;
 	int ret = 0;
 
 	addr = vb2_dma_contig_plane_dma_addr(&buf->vb.vb2_buf, 0);
-	desc = dmaengine_prep_slave_single(csi->dma.chan, addr, len,
+	desc = dmaengine_prep_slave_single(ctx->dma.chan, addr, len,
 					   DMA_DEV_TO_MEM,
 					   DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
 	if (!desc)
@@ -655,20 +676,20 @@ static int ti_csi2rx_start_dma(struct ti_csi2rx_dev *csi,
 	if (ret)
 		return ret;
 
-	dma_async_issue_pending(csi->dma.chan);
+	dma_async_issue_pending(ctx->dma.chan);
 
 	return 0;
 }
 
-static void ti_csi2rx_stop_dma(struct ti_csi2rx_dev *csi)
+static void ti_csi2rx_stop_dma(struct ti_csi2rx_ctx *ctx)
 {
-	struct ti_csi2rx_dma *dma = &csi->dma;
+	struct ti_csi2rx_dma *dma = &ctx->dma;
 	enum ti_csi2rx_dma_state state;
 	unsigned long flags;
 	int ret;
 
 	spin_lock_irqsave(&dma->lock, flags);
-	state = csi->dma.state;
+	state = ctx->dma.state;
 	dma->state = TI_CSI2RX_DMA_STOPPED;
 	spin_unlock_irqrestore(&dma->lock, flags);
 
@@ -679,30 +700,30 @@ static void ti_csi2rx_stop_dma(struct ti_csi2rx_dev *csi)
 		 * is stopped, as the module-level pixel reset cannot be
 		 * enforced before terminating DMA.
 		 */
-		ret = ti_csi2rx_drain_dma(csi);
+		ret = ti_csi2rx_drain_dma(ctx);
 		if (ret && ret != -ETIMEDOUT)
-			dev_warn(csi->dev,
+			dev_warn(ctx->csi->dev,
 				 "Failed to drain DMA. Next frame might be bogus\n");
 	}
 
-	ret = dmaengine_terminate_sync(csi->dma.chan);
+	ret = dmaengine_terminate_sync(ctx->dma.chan);
 	if (ret)
-		dev_err(csi->dev, "Failed to stop DMA: %d\n", ret);
+		dev_err(ctx->csi->dev, "Failed to stop DMA: %d\n", ret);
 }
 
-static void ti_csi2rx_cleanup_buffers(struct ti_csi2rx_dev *csi,
+static void ti_csi2rx_cleanup_buffers(struct ti_csi2rx_ctx *ctx,
 				      enum vb2_buffer_state state)
 {
-	struct ti_csi2rx_dma *dma = &csi->dma;
+	struct ti_csi2rx_dma *dma = &ctx->dma;
 	struct ti_csi2rx_buffer *buf, *tmp;
 	unsigned long flags;
 
 	spin_lock_irqsave(&dma->lock, flags);
-	list_for_each_entry_safe(buf, tmp, &csi->dma.queue, list) {
+	list_for_each_entry_safe(buf, tmp, &ctx->dma.queue, list) {
 		list_del(&buf->list);
 		vb2_buffer_done(&buf->vb.vb2_buf, state);
 	}
-	list_for_each_entry_safe(buf, tmp, &csi->dma.submitted, list) {
+	list_for_each_entry_safe(buf, tmp, &ctx->dma.submitted, list) {
 		list_del(&buf->list);
 		vb2_buffer_done(&buf->vb.vb2_buf, state);
 	}
@@ -713,8 +734,8 @@ static int ti_csi2rx_queue_setup(struct vb2_queue *q, unsigned int *nbuffers,
 				 unsigned int *nplanes, unsigned int sizes[],
 				 struct device *alloc_devs[])
 {
-	struct ti_csi2rx_dev *csi = vb2_get_drv_priv(q);
-	unsigned int size = csi->v_fmt.fmt.pix.sizeimage;
+	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(q);
+	unsigned int size = ctx->v_fmt.fmt.pix.sizeimage;
 
 	if (*nplanes) {
 		if (sizes[0] < size)
@@ -730,11 +751,11 @@ static int ti_csi2rx_queue_setup(struct vb2_queue *q, unsigned int *nbuffers,
 
 static int ti_csi2rx_buffer_prepare(struct vb2_buffer *vb)
 {
-	struct ti_csi2rx_dev *csi = vb2_get_drv_priv(vb->vb2_queue);
-	unsigned long size = csi->v_fmt.fmt.pix.sizeimage;
+	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
+	unsigned long size = ctx->v_fmt.fmt.pix.sizeimage;
 
 	if (vb2_plane_size(vb, 0) < size) {
-		dev_err(csi->dev, "Data will not fit into plane\n");
+		dev_err(ctx->csi->dev, "Data will not fit into plane\n");
 		return -EINVAL;
 	}
 
@@ -744,15 +765,15 @@ static int ti_csi2rx_buffer_prepare(struct vb2_buffer *vb)
 
 static void ti_csi2rx_buffer_queue(struct vb2_buffer *vb)
 {
-	struct ti_csi2rx_dev *csi = vb2_get_drv_priv(vb->vb2_queue);
+	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
 	struct ti_csi2rx_buffer *buf;
-	struct ti_csi2rx_dma *dma = &csi->dma;
+	struct ti_csi2rx_dma *dma = &ctx->dma;
 	bool restart_dma = false;
 	unsigned long flags = 0;
 	int ret;
 
 	buf = container_of(vb, struct ti_csi2rx_buffer, vb.vb2_buf);
-	buf->csi = csi;
+	buf->ctx = ctx;
 
 	spin_lock_irqsave(&dma->lock, flags);
 	/*
@@ -781,18 +802,18 @@ static void ti_csi2rx_buffer_queue(struct vb2_buffer *vb)
 		 * the application and will only confuse it. Issue a DMA
 		 * transaction to drain that up.
 		 */
-		ret = ti_csi2rx_drain_dma(csi);
+		ret = ti_csi2rx_drain_dma(ctx);
 		if (ret && ret != -ETIMEDOUT)
-			dev_warn(csi->dev,
+			dev_warn(ctx->csi->dev,
 				 "Failed to drain DMA. Next frame might be bogus\n");
 
 		spin_lock_irqsave(&dma->lock, flags);
-		ret = ti_csi2rx_start_dma(csi, buf);
+		ret = ti_csi2rx_start_dma(ctx, buf);
 		if (ret) {
 			vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_ERROR);
 			dma->state = TI_CSI2RX_DMA_IDLE;
 			spin_unlock_irqrestore(&dma->lock, flags);
-			dev_err(csi->dev, "Failed to start DMA: %d\n", ret);
+			dev_err(ctx->csi->dev, "Failed to start DMA: %d\n", ret);
 		} else {
 			list_add_tail(&buf->list, &dma->submitted);
 			spin_unlock_irqrestore(&dma->lock, flags);
@@ -802,8 +823,9 @@ static void ti_csi2rx_buffer_queue(struct vb2_buffer *vb)
 
 static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 {
-	struct ti_csi2rx_dev *csi = vb2_get_drv_priv(vq);
-	struct ti_csi2rx_dma *dma = &csi->dma;
+	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vq);
+	struct ti_csi2rx_dev *csi = ctx->csi;
+	struct ti_csi2rx_dma *dma = &ctx->dma;
 	struct ti_csi2rx_buffer *buf;
 	unsigned long flags;
 	int ret = 0;
@@ -815,18 +837,18 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 	if (ret)
 		return ret;
 
-	ret = video_device_pipeline_start(&csi->vdev, &csi->pipe);
+	ret = video_device_pipeline_start(&ctx->vdev, &csi->pipe);
 	if (ret)
 		goto err;
 
-	ti_csi2rx_setup_shim(csi);
+	ti_csi2rx_setup_shim(ctx);
 
-	csi->sequence = 0;
+	ctx->sequence = 0;
 
 	spin_lock_irqsave(&dma->lock, flags);
 	buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer, list);
 
-	ret = ti_csi2rx_start_dma(csi, buf);
+	ret = ti_csi2rx_start_dma(ctx, buf);
 	if (ret) {
 		dev_err(csi->dev, "Failed to start DMA: %d\n", ret);
 		spin_unlock_irqrestore(&dma->lock, flags);
@@ -844,22 +866,23 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 	return 0;
 
 err_dma:
-	ti_csi2rx_stop_dma(csi);
+	ti_csi2rx_stop_dma(ctx);
 err_pipeline:
-	video_device_pipeline_stop(&csi->vdev);
+	video_device_pipeline_stop(&ctx->vdev);
 	writel(0, csi->shim + SHIM_CNTL);
 	writel(0, csi->shim + SHIM_DMACNTX);
 err:
-	ti_csi2rx_cleanup_buffers(csi, VB2_BUF_STATE_QUEUED);
+	ti_csi2rx_cleanup_buffers(ctx, VB2_BUF_STATE_QUEUED);
 	return ret;
 }
 
 static void ti_csi2rx_stop_streaming(struct vb2_queue *vq)
 {
-	struct ti_csi2rx_dev *csi = vb2_get_drv_priv(vq);
+	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vq);
+	struct ti_csi2rx_dev *csi = ctx->csi;
 	int ret;
 
-	video_device_pipeline_stop(&csi->vdev);
+	video_device_pipeline_stop(&ctx->vdev);
 
 	writel(0, csi->shim + SHIM_CNTL);
 	writel(0, csi->shim + SHIM_DMACNTX);
@@ -868,8 +891,8 @@ static void ti_csi2rx_stop_streaming(struct vb2_queue *vq)
 	if (ret)
 		dev_err(csi->dev, "Failed to stop subdev stream\n");
 
-	ti_csi2rx_stop_dma(csi);
-	ti_csi2rx_cleanup_buffers(csi, VB2_BUF_STATE_ERROR);
+	ti_csi2rx_stop_dma(ctx);
+	ti_csi2rx_cleanup_buffers(ctx, VB2_BUF_STATE_ERROR);
 }
 
 static const struct vb2_ops csi_vb2_qops = {
@@ -880,27 +903,50 @@ static const struct vb2_ops csi_vb2_qops = {
 	.stop_streaming = ti_csi2rx_stop_streaming,
 };
 
-static int ti_csi2rx_init_vb2q(struct ti_csi2rx_dev *csi)
+static void ti_csi2rx_cleanup_v4l2(struct ti_csi2rx_dev *csi)
 {
-	struct vb2_queue *q = &csi->vidq;
+	media_device_unregister(&csi->mdev);
+	v4l2_device_unregister(&csi->v4l2_dev);
+	media_device_cleanup(&csi->mdev);
+}
+
+static void ti_csi2rx_cleanup_notifier(struct ti_csi2rx_dev *csi)
+{
+	v4l2_async_nf_unregister(&csi->notifier);
+	v4l2_async_nf_cleanup(&csi->notifier);
+}
+
+static void ti_csi2rx_cleanup_ctx(struct ti_csi2rx_ctx *ctx)
+{
+	dma_release_channel(ctx->dma.chan);
+	vb2_queue_release(&ctx->vidq);
+
+	video_unregister_device(&ctx->vdev);
+
+	mutex_destroy(&ctx->mutex);
+}
+
+static int ti_csi2rx_init_vb2q(struct ti_csi2rx_ctx *ctx)
+{
+	struct vb2_queue *q = &ctx->vidq;
 	int ret;
 
 	q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
 	q->io_modes = VB2_MMAP | VB2_DMABUF;
-	q->drv_priv = csi;
+	q->drv_priv = ctx;
 	q->buf_struct_size = sizeof(struct ti_csi2rx_buffer);
 	q->ops = &csi_vb2_qops;
 	q->mem_ops = &vb2_dma_contig_memops;
 	q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC;
-	q->dev = dmaengine_get_dma_device(csi->dma.chan);
-	q->lock = &csi->mutex;
+	q->dev = dmaengine_get_dma_device(ctx->dma.chan);
+	q->lock = &ctx->mutex;
 	q->min_queued_buffers = 1;
 
 	ret = vb2_queue_init(q);
 	if (ret)
 		return ret;
 
-	csi->vdev.queue = q;
+	ctx->vdev.queue = q;
 
 	return 0;
 }
@@ -909,8 +955,9 @@ static int ti_csi2rx_link_validate(struct media_link *link)
 {
 	struct media_entity *entity = link->sink->entity;
 	struct video_device *vdev = media_entity_to_video_device(entity);
-	struct ti_csi2rx_dev *csi = container_of(vdev, struct ti_csi2rx_dev, vdev);
-	struct v4l2_pix_format *csi_fmt = &csi->v_fmt.fmt.pix;
+	struct ti_csi2rx_ctx *ctx = container_of(vdev, struct ti_csi2rx_ctx, vdev);
+	struct ti_csi2rx_dev *csi = ctx->csi;
+	struct v4l2_pix_format *csi_fmt = &ctx->v_fmt.fmt.pix;
 	struct v4l2_subdev_format source_fmt = {
 		.which	= V4L2_SUBDEV_FORMAT_ACTIVE,
 		.pad	= link->source->index,
@@ -963,47 +1010,69 @@ static const struct media_entity_operations ti_csi2rx_video_entity_ops = {
 	.link_validate = ti_csi2rx_link_validate,
 };
 
-static int ti_csi2rx_init_dma(struct ti_csi2rx_dev *csi)
+static int ti_csi2rx_init_dma(struct ti_csi2rx_ctx *ctx)
 {
 	struct dma_slave_config cfg = {
 		.src_addr_width = DMA_SLAVE_BUSWIDTH_16_BYTES,
 	};
 	int ret;
 
-	INIT_LIST_HEAD(&csi->dma.queue);
-	INIT_LIST_HEAD(&csi->dma.submitted);
-	spin_lock_init(&csi->dma.lock);
+	INIT_LIST_HEAD(&ctx->dma.queue);
+	INIT_LIST_HEAD(&ctx->dma.submitted);
+	spin_lock_init(&ctx->dma.lock);
 
-	csi->dma.state = TI_CSI2RX_DMA_STOPPED;
+	ctx->dma.state = TI_CSI2RX_DMA_STOPPED;
 
-	csi->dma.chan = dma_request_chan(csi->dev, "rx0");
-	if (IS_ERR(csi->dma.chan))
-		return PTR_ERR(csi->dma.chan);
+	ctx->dma.chan = dma_request_chan(ctx->csi->dev, "rx0");
+	if (IS_ERR(ctx->dma.chan))
+		return PTR_ERR(ctx->dma.chan);
 
-	ret = dmaengine_slave_config(csi->dma.chan, &cfg);
+	ret = dmaengine_slave_config(ctx->dma.chan, &cfg);
 	if (ret) {
-		dma_release_channel(csi->dma.chan);
+		dma_release_channel(ctx->dma.chan);
 		return ret;
 	}
 
-	csi->dma.drain.len = DRAIN_BUFFER_SIZE;
-	csi->dma.drain.vaddr = dma_alloc_coherent(csi->dev, csi->dma.drain.len,
-						  &csi->dma.drain.paddr,
-						  GFP_KERNEL);
-	if (!csi->dma.drain.vaddr)
-		return -ENOMEM;
-
 	return 0;
 }
 
 static int ti_csi2rx_v4l2_init(struct ti_csi2rx_dev *csi)
 {
 	struct media_device *mdev = &csi->mdev;
-	struct video_device *vdev = &csi->vdev;
+	int ret;
+
+	mdev->dev = csi->dev;
+	mdev->hw_revision = 1;
+	strscpy(mdev->model, "TI-CSI2RX", sizeof(mdev->model));
+
+	media_device_init(mdev);
+
+	csi->v4l2_dev.mdev = mdev;
+
+	ret = v4l2_device_register(csi->dev, &csi->v4l2_dev);
+	if (ret)
+		return ret;
+
+	ret = media_device_register(mdev);
+	if (ret) {
+		v4l2_device_unregister(&csi->v4l2_dev);
+		media_device_cleanup(mdev);
+		return ret;
+	}
+
+	return 0;
+}
+
+static int ti_csi2rx_init_ctx(struct ti_csi2rx_ctx *ctx)
+{
+	struct ti_csi2rx_dev *csi = ctx->csi;
+	struct video_device *vdev = &ctx->vdev;
 	const struct ti_csi2rx_fmt *fmt;
-	struct v4l2_pix_format *pix_fmt = &csi->v_fmt.fmt.pix;
+	struct v4l2_pix_format *pix_fmt = &ctx->v_fmt.fmt.pix;
 	int ret;
 
+	mutex_init(&ctx->mutex);
+
 	fmt = find_format_by_fourcc(V4L2_PIX_FMT_UYVY);
 	if (!fmt)
 		return -EINVAL;
@@ -1012,19 +1081,20 @@ static int ti_csi2rx_v4l2_init(struct ti_csi2rx_dev *csi)
 	pix_fmt->height = 480;
 	pix_fmt->field = V4L2_FIELD_NONE;
 	pix_fmt->colorspace = V4L2_COLORSPACE_SRGB;
-	pix_fmt->ycbcr_enc = V4L2_YCBCR_ENC_601;
-	pix_fmt->quantization = V4L2_QUANTIZATION_LIM_RANGE;
-	pix_fmt->xfer_func = V4L2_XFER_FUNC_SRGB;
+	pix_fmt->ycbcr_enc = V4L2_YCBCR_ENC_601,
+	pix_fmt->quantization = V4L2_QUANTIZATION_LIM_RANGE,
+	pix_fmt->xfer_func = V4L2_XFER_FUNC_SRGB,
 
-	ti_csi2rx_fill_fmt(fmt, &csi->v_fmt);
+	ti_csi2rx_fill_fmt(fmt, &ctx->v_fmt);
 
-	mdev->dev = csi->dev;
-	mdev->hw_revision = 1;
-	strscpy(mdev->model, "TI-CSI2RX", sizeof(mdev->model));
-
-	media_device_init(mdev);
+	csi->pad.flags = MEDIA_PAD_FL_SINK;
+	vdev->entity.ops = &ti_csi2rx_video_entity_ops;
+	ret = media_entity_pads_init(&ctx->vdev.entity, 1, &csi->pad);
+	if (ret)
+		return ret;
 
-	strscpy(vdev->name, TI_CSI2RX_MODULE_NAME, sizeof(vdev->name));
+	snprintf(vdev->name, sizeof(vdev->name), "%s context %u",
+		 dev_name(csi->dev), ctx->idx);
 	vdev->v4l2_dev = &csi->v4l2_dev;
 	vdev->vfl_dir = VFL_DIR_RX;
 	vdev->fops = &csi_fops;
@@ -1032,61 +1102,28 @@ static int ti_csi2rx_v4l2_init(struct ti_csi2rx_dev *csi)
 	vdev->release = video_device_release_empty;
 	vdev->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_STREAMING |
 			    V4L2_CAP_IO_MC;
-	vdev->lock = &csi->mutex;
-	video_set_drvdata(vdev, csi);
+	vdev->lock = &ctx->mutex;
+	video_set_drvdata(vdev, ctx);
 
-	csi->pad.flags = MEDIA_PAD_FL_SINK;
-	vdev->entity.ops = &ti_csi2rx_video_entity_ops;
-	ret = media_entity_pads_init(&csi->vdev.entity, 1, &csi->pad);
+	ret = ti_csi2rx_init_dma(ctx);
 	if (ret)
 		return ret;
 
-	csi->v4l2_dev.mdev = mdev;
-
-	ret = v4l2_device_register(csi->dev, &csi->v4l2_dev);
+	ret = ti_csi2rx_init_vb2q(ctx);
 	if (ret)
-		return ret;
-
-	ret = media_device_register(mdev);
-	if (ret) {
-		v4l2_device_unregister(&csi->v4l2_dev);
-		media_device_cleanup(mdev);
-		return ret;
-	}
+		goto cleanup_dma;
 
 	return 0;
-}
-
-static void ti_csi2rx_cleanup_dma(struct ti_csi2rx_dev *csi)
-{
-	dma_free_coherent(csi->dev, csi->dma.drain.len,
-			  csi->dma.drain.vaddr, csi->dma.drain.paddr);
-	csi->dma.drain.vaddr = NULL;
-	dma_release_channel(csi->dma.chan);
-}
-
-static void ti_csi2rx_cleanup_v4l2(struct ti_csi2rx_dev *csi)
-{
-	media_device_unregister(&csi->mdev);
-	v4l2_device_unregister(&csi->v4l2_dev);
-	media_device_cleanup(&csi->mdev);
-}
 
-static void ti_csi2rx_cleanup_subdev(struct ti_csi2rx_dev *csi)
-{
-	v4l2_async_nf_unregister(&csi->notifier);
-	v4l2_async_nf_cleanup(&csi->notifier);
-}
-
-static void ti_csi2rx_cleanup_vb2q(struct ti_csi2rx_dev *csi)
-{
-	vb2_queue_release(&csi->vidq);
+cleanup_dma:
+	dma_release_channel(ctx->dma.chan);
+	return ret;
 }
 
 static int ti_csi2rx_probe(struct platform_device *pdev)
 {
 	struct ti_csi2rx_dev *csi;
-	int ret;
+	int ret, i;
 
 	csi = devm_kzalloc(&pdev->dev, sizeof(*csi), GFP_KERNEL);
 	if (!csi)
@@ -1095,62 +1132,69 @@ static int ti_csi2rx_probe(struct platform_device *pdev)
 	csi->dev = &pdev->dev;
 	platform_set_drvdata(pdev, csi);
 
-	mutex_init(&csi->mutex);
 	csi->shim = devm_platform_ioremap_resource(pdev, 0);
 	if (IS_ERR(csi->shim)) {
 		ret = PTR_ERR(csi->shim);
-		goto err_mutex;
+		return ret;
 	}
 
-	ret = ti_csi2rx_init_dma(csi);
-	if (ret)
-		goto err_mutex;
+	csi->drain.len = DRAIN_BUFFER_SIZE;
+	csi->drain.vaddr = dma_alloc_coherent(csi->dev, csi->drain.len,
+					      &csi->drain.paddr,
+					      GFP_KERNEL);
+	if (!csi->drain.vaddr)
+		return -ENOMEM;
 
 	ret = ti_csi2rx_v4l2_init(csi);
-	if (ret)
-		goto err_dma;
-
-	ret = ti_csi2rx_init_vb2q(csi);
 	if (ret)
 		goto err_v4l2;
 
+	for (i = 0; i < TI_CSI2RX_NUM_CTX; i++) {
+		csi->ctx[i].idx = i;
+		csi->ctx[i].csi = csi;
+		ret = ti_csi2rx_init_ctx(&csi->ctx[i]);
+		if (ret)
+			goto err_ctx;
+	}
+
 	ret = ti_csi2rx_notifier_register(csi);
 	if (ret)
-		goto err_vb2q;
+		goto err_ctx;
 
 	ret = of_platform_populate(csi->dev->of_node, NULL, NULL, csi->dev);
 	if (ret) {
 		dev_err(csi->dev, "Failed to create children: %d\n", ret);
-		goto err_subdev;
+		goto err_notifier;
 	}
 
 	return 0;
 
-err_subdev:
-	ti_csi2rx_cleanup_subdev(csi);
-err_vb2q:
-	ti_csi2rx_cleanup_vb2q(csi);
-err_v4l2:
+err_notifier:
+	ti_csi2rx_cleanup_notifier(csi);
+err_ctx:
+	i--;
+	for (; i >= 0; i--)
+		ti_csi2rx_cleanup_ctx(&csi->ctx[i]);
 	ti_csi2rx_cleanup_v4l2(csi);
-err_dma:
-	ti_csi2rx_cleanup_dma(csi);
-err_mutex:
-	mutex_destroy(&csi->mutex);
+err_v4l2:
+	dma_free_coherent(csi->dev, csi->drain.len, csi->drain.vaddr,
+			  csi->drain.paddr);
 	return ret;
 }
 
 static void ti_csi2rx_remove(struct platform_device *pdev)
 {
 	struct ti_csi2rx_dev *csi = platform_get_drvdata(pdev);
+	unsigned int i;
 
-	video_unregister_device(&csi->vdev);
+	for (i = 0; i < TI_CSI2RX_NUM_CTX; i++)
+		ti_csi2rx_cleanup_ctx(&csi->ctx[i]);
 
-	ti_csi2rx_cleanup_vb2q(csi);
-	ti_csi2rx_cleanup_subdev(csi);
+	ti_csi2rx_cleanup_notifier(csi);
 	ti_csi2rx_cleanup_v4l2(csi);
-	ti_csi2rx_cleanup_dma(csi);
 
-	mutex_destroy(&csi->mutex);
+	dma_free_coherent(csi->dev, csi->drain.len, csi->drain.vaddr,
+			  csi->drain.paddr);
 }
 
 static const struct of_device_id ti_csi2rx_of_match[] = {
-- 
2.34.1

[PATCH v4 03/12] media: ti: j721e-csi2rx: prepare SHIM code for multiple contexts

[Rishikesh Donadkar]

From: Pratyush Yadav <p.yadav@ti.com>

Currently the SHIM code to configure the context only touches the first
context. Add support for writing to the context's registers based on the
context index.

Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../media/platform/ti/j721e-csi2rx/j721e-csi2rx.c    | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index 36cde2e87aabb..d03dc4e56d306 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -25,7 +25,7 @@
 #define SHIM_CNTL			0x10
 #define SHIM_CNTL_PIX_RST		BIT(0)
 
-#define SHIM_DMACNTX			0x20
+#define SHIM_DMACNTX(i)			(0x20 + ((i) * 0x20))
 #define SHIM_DMACNTX_EN			BIT(31)
 #define SHIM_DMACNTX_YUV422		GENMASK(27, 26)
 #define SHIM_DMACNTX_SIZE		GENMASK(21, 20)
@@ -35,7 +35,7 @@
 #define SHIM_DMACNTX_SIZE_16		1
 #define SHIM_DMACNTX_SIZE_32		2
 
-#define SHIM_PSI_CFG0			0x24
+#define SHIM_PSI_CFG0(i)		(0x24 + ((i) * 0x20))
 #define SHIM_PSI_CFG0_SRC_TAG		GENMASK(15, 0)
 #define SHIM_PSI_CFG0_DST_TAG		GENMASK(31, 16)
 
@@ -549,11 +549,11 @@ static void ti_csi2rx_setup_shim(struct ti_csi2rx_ctx *ctx)
 
 	reg |= FIELD_PREP(SHIM_DMACNTX_SIZE, fmt->size);
 
-	writel(reg, csi->shim + SHIM_DMACNTX);
+	writel(reg, csi->shim + SHIM_DMACNTX(ctx->idx));
 
 	reg = FIELD_PREP(SHIM_PSI_CFG0_SRC_TAG, 0) |
 	      FIELD_PREP(SHIM_PSI_CFG0_DST_TAG, 0);
-	writel(reg, csi->shim + SHIM_PSI_CFG0);
+	writel(reg, csi->shim + SHIM_PSI_CFG0(ctx->idx));
 }
 
 static void ti_csi2rx_drain_callback(void *param)
@@ -870,7 +870,7 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 err_pipeline:
 	video_device_pipeline_stop(&ctx->vdev);
 	writel(0, csi->shim + SHIM_CNTL);
-	writel(0, csi->shim + SHIM_DMACNTX);
+	writel(0, csi->shim + SHIM_DMACNTX(ctx->idx));
 err:
 	ti_csi2rx_cleanup_buffers(ctx, VB2_BUF_STATE_QUEUED);
 	return ret;
@@ -885,7 +885,7 @@ static void ti_csi2rx_stop_streaming(struct vb2_queue *vq)
 	video_device_pipeline_stop(&ctx->vdev);
 
 	writel(0, csi->shim + SHIM_CNTL);
-	writel(0, csi->shim + SHIM_DMACNTX);
+	writel(0, csi->shim + SHIM_DMACNTX(ctx->idx));
 
 	ret = v4l2_subdev_call(csi->source, video, s_stream, 0);
 	if (ret)
-- 
2.34.1

[PATCH v4 04/12] media: ti: j721e-csi2rx: allocate DMA channel based on context index

[Rishikesh Donadkar]

From: Pratyush Yadav <p.yadav@ti.com>

With multiple contexts, there needs to be a different DMA channel for
each context. Earlier, the DMA channel name was hard coded to "rx0" for
the sake of simplicity. Generate the DMA channel name based on its index
and get the channel corresponding to the context.

Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index d03dc4e56d306..cc6eb3d4c2b26 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -1015,6 +1015,7 @@ static int ti_csi2rx_init_dma(struct ti_csi2rx_ctx *ctx)
 	struct dma_slave_config cfg = {
 		.src_addr_width = DMA_SLAVE_BUSWIDTH_16_BYTES,
 	};
+	char name[5];
 	int ret;
 
 	INIT_LIST_HEAD(&ctx->dma.queue);
@@ -1023,7 +1024,8 @@ static int ti_csi2rx_init_dma(struct ti_csi2rx_ctx *ctx)
 
 	ctx->dma.state = TI_CSI2RX_DMA_STOPPED;
 
-	ctx->dma.chan = dma_request_chan(ctx->csi->dev, "rx0");
+	snprintf(name, sizeof(name), "rx%u", ctx->idx);
+	ctx->dma.chan = dma_request_chan(ctx->csi->dev, name);
 	if (IS_ERR(ctx->dma.chan))
 		return PTR_ERR(ctx->dma.chan);
 
-- 
2.34.1

[PATCH v4 05/12] media: ti: j721e-csi2rx: add a subdev for the core device

[Rishikesh Donadkar]

From: Jai Luthra <j-luthra@ti.com>

With single stream capture, it was simpler to use the video device as
the media entity representing the main TI CSI2RX device. Now with multi
stream capture coming into the picture, the model has shifted to each
video device having a link to the main device's subdev. The routing
would then be set on this subdev.

Add this subdev, link each context to this subdev's entity and link the
subdev's entity to the source. Also add an array of media pads. It will
have one sink pad and source pads equal to the number of contexts.

Co-developed-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 248 ++++++++++++++++--
 1 file changed, 223 insertions(+), 25 deletions(-)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index cc6eb3d4c2b26..d746a3ff25934 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -51,6 +51,11 @@
 #define MAX_WIDTH_BYTES			SZ_16K
 #define MAX_HEIGHT_LINES		SZ_16K
 
+#define TI_CSI2RX_PAD_SINK		0
+#define TI_CSI2RX_PAD_FIRST_SOURCE	1
+#define TI_CSI2RX_NUM_SOURCE_PADS	1
+#define TI_CSI2RX_NUM_PADS		(1 + TI_CSI2RX_NUM_SOURCE_PADS)
+
 #define DRAIN_TIMEOUT_MS		50
 #define DRAIN_BUFFER_SIZE		SZ_32K
 
@@ -97,6 +102,7 @@ struct ti_csi2rx_ctx {
 	struct mutex			mutex; /* To serialize ioctls. */
 	struct v4l2_format		v_fmt;
 	struct ti_csi2rx_dma		dma;
+	struct media_pad		pad;
 	u32				sequence;
 	u32				idx;
 };
@@ -104,12 +110,15 @@ struct ti_csi2rx_ctx {
 struct ti_csi2rx_dev {
 	struct device			*dev;
 	void __iomem			*shim;
+	struct mutex			mutex; /* To serialize ioctls. */
+	unsigned int			enable_count;
 	struct v4l2_device		v4l2_dev;
 	struct media_device		mdev;
 	struct media_pipeline		pipe;
-	struct media_pad		pad;
+	struct media_pad		pads[TI_CSI2RX_NUM_PADS];
 	struct v4l2_async_notifier	notifier;
 	struct v4l2_subdev		*source;
+	struct v4l2_subdev		subdev;
 	struct ti_csi2rx_ctx		ctx[TI_CSI2RX_NUM_CTX];
 	/* Buffer to drain stale data from PSI-L endpoint */
 	struct {
@@ -119,6 +128,22 @@ struct ti_csi2rx_dev {
 	} drain;
 };
 
+static inline struct ti_csi2rx_dev *to_csi2rx_dev(struct v4l2_subdev *sd)
+{
+	return container_of(sd, struct ti_csi2rx_dev, subdev);
+}
+
+static const struct v4l2_mbus_framefmt ti_csi2rx_default_fmt = {
+	.width = 640,
+	.height = 480,
+	.code = MEDIA_BUS_FMT_UYVY8_1X16,
+	.field = V4L2_FIELD_NONE,
+	.colorspace = V4L2_COLORSPACE_SRGB,
+	.ycbcr_enc = V4L2_YCBCR_ENC_601,
+	.quantization = V4L2_QUANTIZATION_LIM_RANGE,
+	.xfer_func = V4L2_XFER_FUNC_SRGB,
+};
+
 static const struct ti_csi2rx_fmt ti_csi2rx_formats[] = {
 	{
 		.fourcc			= V4L2_PIX_FMT_YUYV,
@@ -431,6 +456,15 @@ static int csi_async_notifier_complete(struct v4l2_async_notifier *notifier)
 	struct ti_csi2rx_dev *csi = dev_get_drvdata(notifier->v4l2_dev->dev);
 	int ret, i;
 
+	/* Create link from source to subdev */
+	ret = v4l2_create_fwnode_links_to_pad(csi->source,
+					      &csi->pads[TI_CSI2RX_PAD_SINK],
+					      MEDIA_LNK_FL_IMMUTABLE |
+					      MEDIA_LNK_FL_ENABLED);
+	if (ret)
+		return ret;
+
+	/* Create and link video nodes for all DMA contexts */
 	for (i = 0; i < TI_CSI2RX_NUM_CTX; i++) {
 		struct ti_csi2rx_ctx *ctx = &csi->ctx[i];
 		struct video_device *vdev = &ctx->vdev;
@@ -438,13 +472,17 @@ static int csi_async_notifier_complete(struct v4l2_async_notifier *notifier)
 		ret = video_register_device(vdev, VFL_TYPE_VIDEO, -1);
 		if (ret)
 			goto unregister_dev;
-	}
 
-	ret = v4l2_create_fwnode_links_to_pad(csi->source, &csi->pad,
-					      MEDIA_LNK_FL_IMMUTABLE |
-					      MEDIA_LNK_FL_ENABLED);
-	if (ret)
-		goto unregister_dev;
+		ret = media_create_pad_link(&csi->subdev.entity,
+					    TI_CSI2RX_PAD_FIRST_SOURCE + ctx->idx,
+					    &vdev->entity, 0,
+					    MEDIA_LNK_FL_IMMUTABLE |
+					    MEDIA_LNK_FL_ENABLED);
+		if (ret) {
+			video_unregister_device(vdev);
+			goto unregister_dev;
+		}
+	}
 
 	ret = v4l2_device_register_subdev_nodes(&csi->v4l2_dev);
 	if (ret)
@@ -454,8 +492,10 @@ static int csi_async_notifier_complete(struct v4l2_async_notifier *notifier)
 
 unregister_dev:
 	i--;
-	for (; i >= 0; i--)
+	for (; i >= 0; i--) {
+		media_entity_remove_links(&csi->ctx[i].vdev.entity);
 		video_unregister_device(&csi->ctx[i].vdev);
+	}
 	return ret;
 }
 
@@ -859,7 +899,7 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 	dma->state = TI_CSI2RX_DMA_ACTIVE;
 	spin_unlock_irqrestore(&dma->lock, flags);
 
-	ret = v4l2_subdev_call(csi->source, video, s_stream, 1);
+	ret = v4l2_subdev_call(&csi->subdev, video, s_stream, 1);
 	if (ret)
 		goto err_dma;
 
@@ -887,7 +927,7 @@ static void ti_csi2rx_stop_streaming(struct vb2_queue *vq)
 	writel(0, csi->shim + SHIM_CNTL);
 	writel(0, csi->shim + SHIM_DMACNTX(ctx->idx));
 
-	ret = v4l2_subdev_call(csi->source, video, s_stream, 0);
+	ret = v4l2_subdev_call(&csi->subdev, video, s_stream, 0);
 	if (ret)
 		dev_err(csi->dev, "Failed to stop subdev stream\n");
 
@@ -903,8 +943,114 @@ static const struct vb2_ops csi_vb2_qops = {
 	.stop_streaming = ti_csi2rx_stop_streaming,
 };
 
+static int ti_csi2rx_enum_mbus_code(struct v4l2_subdev *subdev,
+				    struct v4l2_subdev_state *state,
+				    struct v4l2_subdev_mbus_code_enum *code_enum)
+{
+	if (code_enum->index >= ARRAY_SIZE(ti_csi2rx_formats))
+		return -EINVAL;
+
+	code_enum->code = ti_csi2rx_formats[code_enum->index].code;
+
+	return 0;
+}
+
+static int ti_csi2rx_sd_set_fmt(struct v4l2_subdev *sd,
+				struct v4l2_subdev_state *state,
+				struct v4l2_subdev_format *format)
+{
+	struct v4l2_mbus_framefmt *fmt;
+
+	/* No transcoding, don't allow setting source fmt */
+	if (format->pad > TI_CSI2RX_PAD_SINK)
+		return v4l2_subdev_get_fmt(sd, state, format);
+
+	if (!find_format_by_code(format->format.code))
+		format->format.code = ti_csi2rx_formats[0].code;
+
+	format->format.field = V4L2_FIELD_NONE;
+
+	fmt = v4l2_subdev_state_get_format(state, format->pad, format->stream);
+	*fmt = format->format;
+
+	fmt = v4l2_subdev_state_get_format(state, TI_CSI2RX_PAD_FIRST_SOURCE,
+					   format->stream);
+	*fmt = format->format;
+
+	return 0;
+}
+
+static int ti_csi2rx_sd_init_state(struct v4l2_subdev *sd,
+				   struct v4l2_subdev_state *state)
+{
+	struct v4l2_mbus_framefmt *fmt;
+
+	fmt = v4l2_subdev_state_get_format(state, TI_CSI2RX_PAD_SINK);
+	*fmt = ti_csi2rx_default_fmt;
+
+	fmt = v4l2_subdev_state_get_format(state, TI_CSI2RX_PAD_FIRST_SOURCE);
+	*fmt = ti_csi2rx_default_fmt;
+
+	return 0;
+}
+
+static int ti_csi2rx_sd_s_stream(struct v4l2_subdev *sd, int enable)
+{
+	struct ti_csi2rx_dev *csi = to_csi2rx_dev(sd);
+	int ret = 0;
+
+	mutex_lock(&csi->mutex);
+
+	if (enable) {
+		if (csi->enable_count > 0) {
+			csi->enable_count++;
+			goto out;
+		}
+
+		ret = v4l2_subdev_call(csi->source, video, s_stream, 1);
+		if (ret)
+			goto out;
+
+		csi->enable_count++;
+	} else {
+		if (csi->enable_count == 0) {
+			ret = -EINVAL;
+			goto out;
+		}
+
+		if (--csi->enable_count > 0)
+			goto out;
+
+		ret = v4l2_subdev_call(csi->source, video, s_stream, 0);
+	}
+
+out:
+	mutex_unlock(&csi->mutex);
+	return ret;
+}
+
+static const struct v4l2_subdev_pad_ops ti_csi2rx_subdev_pad_ops = {
+	.enum_mbus_code	= ti_csi2rx_enum_mbus_code,
+	.get_fmt = v4l2_subdev_get_fmt,
+	.set_fmt = ti_csi2rx_sd_set_fmt,
+};
+
+static const struct v4l2_subdev_video_ops ti_csi2rx_subdev_video_ops = {
+	.s_stream = ti_csi2rx_sd_s_stream,
+};
+
+static const struct v4l2_subdev_ops ti_csi2rx_subdev_ops = {
+	.video = &ti_csi2rx_subdev_video_ops,
+	.pad = &ti_csi2rx_subdev_pad_ops,
+};
+
+static const struct v4l2_subdev_internal_ops ti_csi2rx_internal_ops = {
+	.init_state = ti_csi2rx_sd_init_state,
+};
+
 static void ti_csi2rx_cleanup_v4l2(struct ti_csi2rx_dev *csi)
 {
+	v4l2_subdev_cleanup(&csi->subdev);
 	media_device_unregister(&csi->mdev);
 	v4l2_device_unregister(&csi->v4l2_dev);
 	media_device_cleanup(&csi->mdev);
@@ -961,14 +1107,22 @@ static int ti_csi2rx_link_validate(struct media_link *link)
 	struct v4l2_subdev_format source_fmt = {
 		.which	= V4L2_SUBDEV_FORMAT_ACTIVE,
 		.pad	= link->source->index,
+		.stream = 0,
 	};
+	struct v4l2_subdev_state *state;
 	const struct ti_csi2rx_fmt *ti_fmt;
 	int ret;
 
-	ret = v4l2_subdev_call_state_active(csi->source, pad,
-					    get_fmt, &source_fmt);
-	if (ret)
-		return ret;
+	state = v4l2_subdev_lock_and_get_active_state(&csi->subdev);
+	ret = v4l2_subdev_call(&csi->subdev, pad, get_fmt, state, &source_fmt);
+	v4l2_subdev_unlock_state(state);
+
+	if (ret) {
+		dev_dbg(csi->dev,
+			"Skipping validation as no format present on \"%s\":%u:0\n",
+			link->source->entity->name, link->source->index);
+		return 0;
+	}
 
 	if (source_fmt.format.width != csi_fmt->width) {
 		dev_dbg(csi->dev, "Width does not match (source %u, sink %u)\n",
@@ -998,8 +1152,9 @@ static int ti_csi2rx_link_validate(struct media_link *link)
 
 	if (ti_fmt->fourcc != csi_fmt->pixelformat) {
 		dev_dbg(csi->dev,
-			"Cannot transform source fmt 0x%x to sink fmt 0x%x\n",
-			ti_fmt->fourcc, csi_fmt->pixelformat);
+			"Cannot transform \"%s\":%u format %p4cc to %p4cc\n",
+			link->source->entity->name, link->source->index,
+			&ti_fmt->fourcc, &csi_fmt->pixelformat);
 		return -EPIPE;
 	}
 
@@ -1010,6 +1165,10 @@ static const struct media_entity_operations ti_csi2rx_video_entity_ops = {
 	.link_validate = ti_csi2rx_link_validate,
 };
 
+static const struct media_entity_operations ti_csi2rx_subdev_entity_ops = {
+	.link_validate = v4l2_subdev_link_validate,
+};
+
 static int ti_csi2rx_init_dma(struct ti_csi2rx_ctx *ctx)
 {
 	struct dma_slave_config cfg = {
@@ -1041,6 +1200,7 @@ static int ti_csi2rx_init_dma(struct ti_csi2rx_ctx *ctx)
 static int ti_csi2rx_v4l2_init(struct ti_csi2rx_dev *csi)
 {
 	struct media_device *mdev = &csi->mdev;
+	struct v4l2_subdev *sd = &csi->subdev;
 	int ret;
 
 	mdev->dev = csi->dev;
@@ -1053,16 +1213,51 @@ static int ti_csi2rx_v4l2_init(struct ti_csi2rx_dev *csi)
 
 	ret = v4l2_device_register(csi->dev, &csi->v4l2_dev);
 	if (ret)
-		return ret;
+		goto cleanup_media;
 
 	ret = media_device_register(mdev);
-	if (ret) {
-		v4l2_device_unregister(&csi->v4l2_dev);
-		media_device_cleanup(mdev);
-		return ret;
-	}
+	if (ret)
+		goto unregister_v4l2;
+
+	v4l2_subdev_init(sd, &ti_csi2rx_subdev_ops);
+	sd->internal_ops = &ti_csi2rx_internal_ops;
+	sd->entity.function = MEDIA_ENT_F_VID_IF_BRIDGE;
+	sd->flags = V4L2_SUBDEV_FL_HAS_DEVNODE;
+	strscpy(sd->name, dev_name(csi->dev), sizeof(sd->name));
+	sd->dev = csi->dev;
+	sd->entity.ops = &ti_csi2rx_subdev_entity_ops;
+
+	csi->pads[TI_CSI2RX_PAD_SINK].flags = MEDIA_PAD_FL_SINK;
+
+	for (unsigned int i = TI_CSI2RX_PAD_FIRST_SOURCE;
+	     i < TI_CSI2RX_NUM_PADS; i++)
+		csi->pads[i].flags = MEDIA_PAD_FL_SOURCE;
+
+	ret = media_entity_pads_init(&sd->entity, ARRAY_SIZE(csi->pads),
+				     csi->pads);
+	if (ret)
+		goto unregister_media;
+
+	ret = v4l2_subdev_init_finalize(sd);
+	if (ret)
+		goto unregister_media;
+
+	ret = v4l2_device_register_subdev(&csi->v4l2_dev, sd);
+	if (ret)
+		goto cleanup_subdev;
 
 	return 0;
+
+cleanup_subdev:
+	v4l2_subdev_cleanup(sd);
+unregister_media:
+	media_device_unregister(mdev);
+unregister_v4l2:
+	v4l2_device_unregister(&csi->v4l2_dev);
+cleanup_media:
+	media_device_cleanup(mdev);
+
+	return ret;
 }
 
 static int ti_csi2rx_init_ctx(struct ti_csi2rx_ctx *ctx)
@@ -1089,9 +1284,9 @@ static int ti_csi2rx_init_ctx(struct ti_csi2rx_ctx *ctx)
 
 	ti_csi2rx_fill_fmt(fmt, &ctx->v_fmt);
 
-	csi->pad.flags = MEDIA_PAD_FL_SINK;
+	ctx->pad.flags = MEDIA_PAD_FL_SINK;
 	vdev->entity.ops = &ti_csi2rx_video_entity_ops;
-	ret = media_entity_pads_init(&ctx->vdev.entity, 1, &csi->pad);
+	ret = media_entity_pads_init(&ctx->vdev.entity, 1, &ctx->pad);
 	if (ret)
 		return ret;
 
@@ -1147,6 +1342,8 @@ static int ti_csi2rx_probe(struct platform_device *pdev)
 	if (!csi->drain.vaddr)
 		return -ENOMEM;
 
+	mutex_init(&csi->mutex);
+
 	ret = ti_csi2rx_v4l2_init(csi);
 	if (ret)
 		goto err_v4l2;
@@ -1179,6 +1376,7 @@ static int ti_csi2rx_probe(struct platform_device *pdev)
 		ti_csi2rx_cleanup_ctx(&csi->ctx[i]);
 	ti_csi2rx_cleanup_v4l2(csi);
 err_v4l2:
+	mutex_destroy(&csi->mutex);
 	dma_free_coherent(csi->dev, csi->drain.len, csi->drain.vaddr,
 			  csi->drain.paddr);
 	return ret;
@@ -1194,7 +1392,7 @@ static void ti_csi2rx_remove(struct platform_device *pdev)
 
 	ti_csi2rx_cleanup_notifier(csi);
 	ti_csi2rx_cleanup_v4l2(csi);
-
+	mutex_destroy(&csi->mutex);
 	dma_free_coherent(csi->dev, csi->drain.len, csi->drain.vaddr,
 			  csi->drain.paddr);
 }
-- 
2.34.1

[PATCH v4 06/12] media: ti: j721e-csi2rx: get number of contexts from device tree

[Rishikesh Donadkar]

From: Pratyush Yadav <p.yadav@ti.com>

Different platforms that use this driver might have different number of
DMA channels allocated for CSI. So only as many DMA contexts can be used
as the number of DMA channels available. Get the number of channels
provided via device tree and only configure that many contexts, and
hence only that many pads.

Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
Co-developed-by: Jai Luthra <j-luthra@ti.com>
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 42 ++++++++++++++-----
 1 file changed, 31 insertions(+), 11 deletions(-)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index d746a3ff25934..9c20a5421fa20 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -40,7 +40,7 @@
 #define SHIM_PSI_CFG0_DST_TAG		GENMASK(31, 16)
 
 #define PSIL_WORD_SIZE_BYTES		16
-#define TI_CSI2RX_NUM_CTX		1
+#define TI_CSI2RX_MAX_CTX		32
 
 /*
  * There are no hard limits on the width or height. The DMA engine can handle
@@ -53,8 +53,8 @@
 
 #define TI_CSI2RX_PAD_SINK		0
 #define TI_CSI2RX_PAD_FIRST_SOURCE	1
-#define TI_CSI2RX_NUM_SOURCE_PADS	1
-#define TI_CSI2RX_NUM_PADS		(1 + TI_CSI2RX_NUM_SOURCE_PADS)
+#define TI_CSI2RX_MAX_SOURCE_PADS	TI_CSI2RX_MAX_CTX
+#define TI_CSI2RX_MAX_PADS		(1 + TI_CSI2RX_MAX_SOURCE_PADS)
 
 #define DRAIN_TIMEOUT_MS		50
 #define DRAIN_BUFFER_SIZE		SZ_32K
@@ -112,14 +112,15 @@ struct ti_csi2rx_dev {
 	void __iomem			*shim;
 	struct mutex			mutex; /* To serialize ioctls. */
 	unsigned int			enable_count;
+	unsigned int			num_ctx;
 	struct v4l2_device		v4l2_dev;
 	struct media_device		mdev;
 	struct media_pipeline		pipe;
-	struct media_pad		pads[TI_CSI2RX_NUM_PADS];
+	struct media_pad		pads[TI_CSI2RX_MAX_PADS];
 	struct v4l2_async_notifier	notifier;
 	struct v4l2_subdev		*source;
 	struct v4l2_subdev		subdev;
-	struct ti_csi2rx_ctx		ctx[TI_CSI2RX_NUM_CTX];
+	struct ti_csi2rx_ctx		ctx[TI_CSI2RX_MAX_CTX];
 	/* Buffer to drain stale data from PSI-L endpoint */
 	struct {
 		void			*vaddr;
@@ -465,7 +466,7 @@ static int csi_async_notifier_complete(struct v4l2_async_notifier *notifier)
 		return ret;
 
 	/* Create and link video nodes for all DMA contexts */
-	for (i = 0; i < TI_CSI2RX_NUM_CTX; i++) {
+	for (i = 0; i < csi->num_ctx; i++) {
 		struct ti_csi2rx_ctx *ctx = &csi->ctx[i];
 		struct video_device *vdev = &ctx->vdev;
 
@@ -1230,10 +1231,11 @@ static int ti_csi2rx_v4l2_init(struct ti_csi2rx_dev *csi)
 	csi->pads[TI_CSI2RX_PAD_SINK].flags = MEDIA_PAD_FL_SINK;
 
 	for (unsigned int i = TI_CSI2RX_PAD_FIRST_SOURCE;
-	     i < TI_CSI2RX_NUM_PADS; i++)
+	     i < TI_CSI2RX_PAD_FIRST_SOURCE + csi->num_ctx; i++)
 		csi->pads[i].flags = MEDIA_PAD_FL_SOURCE;
 
-	ret = media_entity_pads_init(&sd->entity, ARRAY_SIZE(csi->pads),
+	ret = media_entity_pads_init(&sd->entity,
+				     TI_CSI2RX_PAD_FIRST_SOURCE + csi->num_ctx,
 				     csi->pads);
 	if (ret)
 		goto unregister_media;
@@ -1319,8 +1321,9 @@ static int ti_csi2rx_init_ctx(struct ti_csi2rx_ctx *ctx)
 
 static int ti_csi2rx_probe(struct platform_device *pdev)
 {
+	struct device_node *np = pdev->dev.of_node;
 	struct ti_csi2rx_dev *csi;
-	int ret, i;
+	int ret, i, count;
 
 	csi = devm_kzalloc(&pdev->dev, sizeof(*csi), GFP_KERNEL);
 	if (!csi)
@@ -1342,13 +1345,29 @@ static int ti_csi2rx_probe(struct platform_device *pdev)
 	if (!csi->drain.vaddr)
 		return -ENOMEM;
 
+	/* Only use as many contexts as the number of DMA channels allocated. */
+	count = of_property_count_strings(np, "dma-names");
+	if (count < 0) {
+		dev_err(csi->dev, "Failed to get DMA channel count: %d\n", count);
+		ret = count;
+		goto err_dma_chan;
+	}
+
+	csi->num_ctx = count;
+	if (csi->num_ctx > TI_CSI2RX_MAX_CTX) {
+		dev_err(csi->dev,
+			"%u DMA channels passed. Maximum is %u.\n",
+			csi->num_ctx, TI_CSI2RX_MAX_CTX);
+		goto err_dma_chan;
+	}
+
 	mutex_init(&csi->mutex);
 
 	ret = ti_csi2rx_v4l2_init(csi);
 	if (ret)
 		goto err_v4l2;
 
-	for (i = 0; i < TI_CSI2RX_NUM_CTX; i++) {
+	for (i = 0; i < csi->num_ctx; i++) {
 		csi->ctx[i].idx = i;
 		csi->ctx[i].csi = csi;
 		ret = ti_csi2rx_init_ctx(&csi->ctx[i]);
@@ -1377,6 +1396,7 @@ static int ti_csi2rx_probe(struct platform_device *pdev)
 	ti_csi2rx_cleanup_v4l2(csi);
 err_v4l2:
 	mutex_destroy(&csi->mutex);
+err_dma_chan:
 	dma_free_coherent(csi->dev, csi->drain.len, csi->drain.vaddr,
 			  csi->drain.paddr);
 	return ret;
@@ -1387,7 +1407,7 @@ static void ti_csi2rx_remove(struct platform_device *pdev)
 	struct ti_csi2rx_dev *csi = platform_get_drvdata(pdev);
 	unsigned int i;
 
-	for (i = 0; i < TI_CSI2RX_NUM_CTX; i++)
+	for (i = 0; i < csi->num_ctx; i++)
 		ti_csi2rx_cleanup_ctx(&csi->ctx[i]);
 
 	ti_csi2rx_cleanup_notifier(csi);
-- 
2.34.1

[PATCH v4 07/12] media: cadence: csi2rx: add get_frame_desc wrapper

[Rishikesh Donadkar]

From: Pratyush Yadav <p.yadav@ti.com>

J721E wrapper CSI2RX driver needs to get the frame descriptor from the
source to find out info about virtual channel. This driver itself does
not touch the routing or virtual channels in any way. So simply pass the
descriptor through from the source.

Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Reviewed-by: Changhuang Liang <changhuang.liang@starfivetech.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 drivers/media/platform/cadence/cdns-csi2rx.c | 24 ++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/drivers/media/platform/cadence/cdns-csi2rx.c b/drivers/media/platform/cadence/cdns-csi2rx.c
index cebcae196eecc..b7e9225f66a41 100644
--- a/drivers/media/platform/cadence/cdns-csi2rx.c
+++ b/drivers/media/platform/cadence/cdns-csi2rx.c
@@ -135,6 +135,21 @@ static const struct csi2rx_fmt *csi2rx_get_fmt_by_code(u32 code)
 	return NULL;
 }
 
+static int csi2rx_get_frame_desc_from_source(struct csi2rx_priv *csi2rx,
+					     struct v4l2_mbus_frame_desc *fd)
+{
+	struct media_pad *remote_pad;
+
+	remote_pad = media_entity_remote_source_pad_unique(&csi2rx->subdev.entity);
+	if (!remote_pad) {
+		dev_err(csi2rx->dev, "No remote pad found for sink\n");
+		return -ENODEV;
+	}
+
+	return v4l2_subdev_call(csi2rx->source_subdev, pad, get_frame_desc,
+				remote_pad->index, fd);
+}
+
 static inline
 struct csi2rx_priv *v4l2_subdev_to_csi2rx(struct v4l2_subdev *subdev)
 {
@@ -458,10 +473,19 @@ static int csi2rx_init_state(struct v4l2_subdev *subdev,
 	return csi2rx_set_fmt(subdev, state, &format);
 }
 
+static int csi2rx_get_frame_desc(struct v4l2_subdev *subdev, unsigned int pad,
+				 struct v4l2_mbus_frame_desc *fd)
+{
+	struct csi2rx_priv *csi2rx = v4l2_subdev_to_csi2rx(subdev);
+
+	return csi2rx_get_frame_desc_from_source(csi2rx, fd);
+}
+
 static const struct v4l2_subdev_pad_ops csi2rx_pad_ops = {
 	.enum_mbus_code	= csi2rx_enum_mbus_code,
 	.get_fmt	= v4l2_subdev_get_fmt,
 	.set_fmt	= csi2rx_set_fmt,
+	.get_frame_desc	= csi2rx_get_frame_desc,
 };
 
 static const struct v4l2_subdev_video_ops csi2rx_video_ops = {
-- 
2.34.1

[PATCH v4 08/12] media: ti: j721e-csi2rx: add support for processing virtual channels

[Rishikesh Donadkar]

From: Jai Luthra <j-luthra@ti.com>

Use get_frame_desc() to get the frame desc from the connected source,
and use the provided virtual channel instead of hardcoded one.

get_frame_desc() returns the same information when called on each stream
start, so instead get the VCs for all the routed stream at first
stream start and cache this information in the driver.

get_frame_desc() works per stream, but as we don't support multiple
streams yet, we will just always use stream 0. If the source doesn't
support get_frame_desc(), fall back to the previous method of always
capturing virtual channel 0.

Co-developed-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 108 ++++++++++++++++++
 1 file changed, 108 insertions(+)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index 9c20a5421fa20..9f05d16ef8c1d 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -29,6 +29,7 @@
 #define SHIM_DMACNTX_EN			BIT(31)
 #define SHIM_DMACNTX_YUV422		GENMASK(27, 26)
 #define SHIM_DMACNTX_SIZE		GENMASK(21, 20)
+#define SHIM_DMACNTX_VC			GENMASK(9, 6)
 #define SHIM_DMACNTX_FMT		GENMASK(5, 0)
 #define SHIM_DMACNTX_YUV422_MODE_11	3
 #define SHIM_DMACNTX_SIZE_8		0
@@ -97,6 +98,7 @@ struct ti_csi2rx_dev;
 
 struct ti_csi2rx_ctx {
 	struct ti_csi2rx_dev		*csi;
+	struct v4l2_subdev_route	*route;
 	struct video_device		vdev;
 	struct vb2_queue		vidq;
 	struct mutex			mutex; /* To serialize ioctls. */
@@ -105,6 +107,8 @@ struct ti_csi2rx_ctx {
 	struct media_pad		pad;
 	u32				sequence;
 	u32				idx;
+	u32				vc;
+	u32				stream;
 };
 
 struct ti_csi2rx_dev {
@@ -127,6 +131,7 @@ struct ti_csi2rx_dev {
 		dma_addr_t		paddr;
 		size_t			len;
 	} drain;
+	bool				vc_cached;
 };
 
 static inline struct ti_csi2rx_dev *to_csi2rx_dev(struct v4l2_subdev *sd)
@@ -589,6 +594,7 @@ static void ti_csi2rx_setup_shim(struct ti_csi2rx_ctx *ctx)
 	}
 
 	reg |= FIELD_PREP(SHIM_DMACNTX_SIZE, fmt->size);
+	reg |= FIELD_PREP(SHIM_DMACNTX_VC, ctx->vc);
 
 	writel(reg, csi->shim + SHIM_DMACNTX(ctx->idx));
 
@@ -862,6 +868,83 @@ static void ti_csi2rx_buffer_queue(struct vb2_buffer *vb)
 	}
 }
 
+static int ti_csi2rx_get_route(struct ti_csi2rx_ctx *ctx)
+{
+	struct ti_csi2rx_dev *csi = ctx->csi;
+	struct media_pad *pad;
+	struct v4l2_subdev_state *state;
+	struct v4l2_subdev_route *r;
+
+	/* Get the source pad connected to this ctx */
+	pad = media_entity_remote_source_pad_unique(ctx->pad.entity);
+	if (!pad) {
+		dev_err(csi->dev, "No pad connected to ctx %d\n", ctx->idx);
+		v4l2_subdev_unlock_state(state);
+		return -ENODEV;
+	}
+
+	state = v4l2_subdev_lock_and_get_active_state(&csi->subdev);
+
+	for_each_active_route(&state->routing, r) {
+		if (!(r->flags & V4L2_SUBDEV_ROUTE_FL_ACTIVE))
+			continue;
+		if (r->source_pad != pad->index)
+			continue;
+
+		ctx->route = r;
+	}
+
+	v4l2_subdev_unlock_state(state);
+
+	if (!ctx->route)
+		return -ENODEV;
+
+	return 0;
+}
+
+static int ti_csi2rx_get_vc(struct ti_csi2rx_ctx *ctx)
+{
+	struct ti_csi2rx_dev *csi = ctx->csi;
+	struct ti_csi2rx_ctx *curr_ctx;
+	struct v4l2_mbus_frame_desc fd;
+	struct media_pad *source_pad;
+	struct v4l2_subdev_route *curr_route;
+	int ret;
+	unsigned int i, j;
+
+	/* Get the frame desc form source */
+	source_pad = media_entity_remote_pad_unique(&csi->subdev.entity, MEDIA_PAD_FL_SOURCE);
+	if (!source_pad)
+		return -ENODEV;
+
+	ret = v4l2_subdev_call(csi->source, pad, get_frame_desc, source_pad->index, &fd);
+	if (ret)
+		return ret;
+
+	if (fd.type != V4L2_MBUS_FRAME_DESC_TYPE_CSI2)
+		return -EINVAL;
+
+	for (i = 0; i < csi->num_ctx; i++) {
+		curr_ctx = &csi->ctx[i];
+
+		/* Capture VC 0 by default */
+		curr_ctx->vc = 0;
+
+		ret = ti_csi2rx_get_route(curr_ctx);
+		if (ret)
+			continue;
+
+		curr_route = curr_ctx->route;
+		curr_ctx->stream = curr_route->sink_stream;
+
+		for (j = 0; j < fd.num_entries; j++)
+			if (curr_ctx->stream == fd.entry[j].stream)
+				curr_ctx->vc = fd.entry[j].bus.csi2.vc;
+	}
+
+	return 0;
+}
+
 static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 {
 	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vq);
@@ -882,6 +965,25 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 	if (ret)
 		goto err;
 
+	/* If no stream is routed to this ctx, exit early */
+	ret = ti_csi2rx_get_route(ctx);
+	if (ret)
+		goto err;
+
+	/* Get the VC for all enabled ctx on first stream start */
+	mutex_lock(&csi->mutex);
+	if (!csi->vc_cached) {
+		ret = ti_csi2rx_get_vc(ctx);
+		if (ret == -ENOIOCTLCMD) {
+			ctx->vc = 0;
+		} else if (ret < 0) {
+			mutex_unlock(&csi->mutex);
+			goto err;
+		}
+		csi->vc_cached = true;
+	}
+	mutex_unlock(&csi->mutex);
+
 	ti_csi2rx_setup_shim(ctx);
 
 	ctx->sequence = 0;
@@ -928,6 +1030,10 @@ static void ti_csi2rx_stop_streaming(struct vb2_queue *vq)
 	writel(0, csi->shim + SHIM_CNTL);
 	writel(0, csi->shim + SHIM_DMACNTX(ctx->idx));
 
+	mutex_lock(&csi->mutex);
+	csi->vc_cached = false;
+	mutex_unlock(&csi->mutex);
+
 	ret = v4l2_subdev_call(&csi->subdev, video, s_stream, 0);
 	if (ret)
 		dev_err(csi->dev, "Failed to stop subdev stream\n");
@@ -1286,6 +1392,8 @@ static int ti_csi2rx_init_ctx(struct ti_csi2rx_ctx *ctx)
 
 	ti_csi2rx_fill_fmt(fmt, &ctx->v_fmt);
 
+	ctx->route = NULL;
+
 	ctx->pad.flags = MEDIA_PAD_FL_SINK;
 	vdev->entity.ops = &ti_csi2rx_video_entity_ops;
 	ret = media_entity_pads_init(&ctx->vdev.entity, 1, &ctx->pad);
-- 
2.34.1

[PATCH v4 09/12] media: cadence: csi2rx: add multistream support

[Rishikesh Donadkar]

From: Jai Luthra <j-luthra@ti.com>

Cadence CSI-2 bridge IP supports capturing multiple virtual "streams"
of data over the same physical interface using MIPI Virtual Channels.

While the hardware IP supports usecases where streams coming in the sink
pad can be broadcasted to multiple source pads, the driver will need
significant re-architecture to make that possible. The two users of this
IP in mainline linux are TI Shim and StarFive JH7110 CAMSS, and both
have only integrated the first source pad i.e stream0 of this IP. So for
now keep it simple and only allow 1-to-1 mapping of streams from sink to
source, without any broadcasting.

The enable_streams() API in v4l2 supports passing a bitmask to enable
each pad/stream combination individually on any media subdev. Use this
API instead of  s_stream() API.

Implement the enable_stream and disable_stream hooks in place of the
stream-unaware s_stream hook.

Implement a fallback s_stream hook that internally calls enable_stream
on each source pad, for consumer drivers that don't use multi-stream
APIs to still work. The helper function v4l2_subdev_s_stream_helper()
form the v4l2 framework is not used here as it is meant only for the
subedvs that have a single source pad and this hardware IP supports
having multiple source pads.

Signed-off-by: Jai Luthra <j-luthra@ti.com>
Reviewed-by: Changhuang Liang <changhuang.liang@starfivetech.com>
Co-developed-by: Rishikesh Donadkar <r-donadkar@ti.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 drivers/media/platform/cadence/cdns-csi2rx.c | 352 +++++++++++++++----
 1 file changed, 275 insertions(+), 77 deletions(-)

diff --git a/drivers/media/platform/cadence/cdns-csi2rx.c b/drivers/media/platform/cadence/cdns-csi2rx.c
index b7e9225f66a41..103c3bda653b9 100644
--- a/drivers/media/platform/cadence/cdns-csi2rx.c
+++ b/drivers/media/platform/cadence/cdns-csi2rx.c
@@ -90,6 +90,7 @@ struct csi2rx_priv {
 	struct reset_control		*pixel_rst[CSI2RX_STREAMS_MAX];
 	struct phy			*dphy;
 
+	u32				vc_select[CSI2RX_STREAMS_MAX];
 	u8				lanes[CSI2RX_LANES_MAX];
 	u8				num_lanes;
 	u8				max_lanes;
@@ -179,29 +180,43 @@ static void csi2rx_reset(struct csi2rx_priv *csi2rx)
 
 static int csi2rx_configure_ext_dphy(struct csi2rx_priv *csi2rx)
 {
-	struct media_pad *src_pad =
-		&csi2rx->source_subdev->entity.pads[csi2rx->source_pad];
 	union phy_configure_opts opts = { };
 	struct phy_configure_opts_mipi_dphy *cfg = &opts.mipi_dphy;
-	struct v4l2_subdev_format sd_fmt = {
-		.which	= V4L2_SUBDEV_FORMAT_ACTIVE,
-		.pad	= CSI2RX_PAD_SINK,
-	};
+	struct v4l2_mbus_framefmt *framefmt;
+	struct v4l2_subdev_state *state;
 	const struct csi2rx_fmt *fmt;
+	int source_pad = csi2rx->source_pad;
+	struct media_pad *pad = &csi2rx->source_subdev->entity.pads[source_pad];
 	s64 link_freq;
 	int ret;
+	u32 bpp;
 
-	ret = v4l2_subdev_call_state_active(&csi2rx->subdev, pad, get_fmt,
-					    &sd_fmt);
-	if (ret < 0)
-		return ret;
+	state = v4l2_subdev_get_locked_active_state(&csi2rx->subdev);
+
+	/*
+	 * For multi-stream transmitters there is no single pixel rate.
+	 *
+	 * In multistream usecase pass bpp as 0 so that v4l2_get_link_freq()
+	 * returns an error if it falls back to V4L2_CID_PIXEL_RATE.
+	 */
+	if (state->routing.num_routes > 1) {
+		bpp = 0;
+	} else {
+		framefmt = v4l2_subdev_state_get_format(state, CSI2RX_PAD_SINK, 0);
+		if (!framefmt) {
+			dev_err(csi2rx->dev, "Did not find active sink format\n");
+			return -EINVAL;
+		}
 
-	fmt = csi2rx_get_fmt_by_code(sd_fmt.format.code);
+		fmt = csi2rx_get_fmt_by_code(framefmt->code);
+		bpp = fmt->bpp;
+	}
 
-	link_freq = v4l2_get_link_freq(src_pad,
-				       fmt->bpp, 2 * csi2rx->num_lanes);
-	if (link_freq < 0)
+	link_freq = v4l2_get_link_freq(pad, bpp, 2 * csi2rx->num_lanes);
+	if (link_freq < 0) {
+		dev_err(csi2rx->dev, "Unable to calculate link frequency\n");
 		return link_freq;
+	}
 
 	ret = phy_mipi_dphy_get_default_config_for_hsclk(link_freq,
 							 csi2rx->num_lanes, cfg);
@@ -294,11 +309,7 @@ static int csi2rx_start(struct csi2rx_priv *csi2rx)
 		writel(CSI2RX_STREAM_CFG_FIFO_MODE_LARGE_BUF,
 		       csi2rx->base + CSI2RX_STREAM_CFG_REG(i));
 
-		/*
-		 * Enable one virtual channel. When multiple virtual channels
-		 * are supported this will have to be changed.
-		 */
-		writel(CSI2RX_STREAM_DATA_CFG_VC_SELECT(0),
+		writel(csi2rx->vc_select[i],
 		       csi2rx->base + CSI2RX_STREAM_DATA_CFG_REG(i));
 
 		writel(CSI2RX_STREAM_CTRL_START,
@@ -311,16 +322,10 @@ static int csi2rx_start(struct csi2rx_priv *csi2rx)
 
 	reset_control_deassert(csi2rx->sys_rst);
 
-	ret = v4l2_subdev_call(csi2rx->source_subdev, video, s_stream, true);
-	if (ret)
-		goto err_disable_sysclk;
-
 	clk_disable_unprepare(csi2rx->p_clk);
 
 	return 0;
 
-err_disable_sysclk:
-	clk_disable_unprepare(csi2rx->sys_clk);
 err_disable_pixclk:
 	for (; i > 0; i--) {
 		reset_control_assert(csi2rx->pixel_rst[i - 1]);
@@ -367,9 +372,6 @@ static void csi2rx_stop(struct csi2rx_priv *csi2rx)
 	reset_control_assert(csi2rx->p_rst);
 	clk_disable_unprepare(csi2rx->p_clk);
 
-	if (v4l2_subdev_call(csi2rx->source_subdev, video, s_stream, false))
-		dev_warn(csi2rx->dev, "Couldn't disable our subdev\n");
-
 	if (csi2rx->dphy) {
 		writel(0, csi2rx->base + CSI2RX_DPHY_LANE_CTRL_REG);
 
@@ -378,37 +380,134 @@ static void csi2rx_stop(struct csi2rx_priv *csi2rx)
 	}
 }
 
-static int csi2rx_s_stream(struct v4l2_subdev *subdev, int enable)
+static void csi2rx_update_vc_select(struct csi2rx_priv *csi2rx,
+				    struct v4l2_subdev_state *state)
 {
-	struct csi2rx_priv *csi2rx = v4l2_subdev_to_csi2rx(subdev);
-	int ret = 0;
-
-	mutex_lock(&csi2rx->lock);
-
-	if (enable) {
-		/*
-		 * If we're not the first users, there's no need to
-		 * enable the whole controller.
-		 */
-		if (!csi2rx->count) {
-			ret = csi2rx_start(csi2rx);
-			if (ret)
-				goto out;
+	struct v4l2_mbus_frame_desc fd = {0};
+	struct v4l2_subdev_route *route;
+	unsigned int i;
+	int ret;
+
+	/* Capture VC=0 by default */
+	for (i = 0; i < CSI2RX_STREAMS_MAX; i++)
+		csi2rx->vc_select[i] = CSI2RX_STREAM_DATA_CFG_VC_SELECT(0);
+
+	ret = csi2rx_get_frame_desc_from_source(csi2rx, &fd);
+	if (ret || fd.type != V4L2_MBUS_FRAME_DESC_TYPE_CSI2) {
+		dev_dbg(csi2rx->dev,
+			"Failed to get source frame desc, allowing only VC=0\n");
+		return;
+	}
+
+	/* If source provides per-stream VC info, use it to filter by VC */
+	memset(csi2rx->vc_select, 0, sizeof(csi2rx->vc_select));
+
+	for_each_active_route(&state->routing, route) {
+		u32 cdns_stream = route->source_pad - CSI2RX_PAD_SOURCE_STREAM0;
+
+		for (i = 0; i < fd.num_entries; i++) {
+			if (fd.entry[i].stream != route->sink_stream)
+				continue;
+
+			csi2rx->vc_select[cdns_stream] |=
+				CSI2RX_STREAM_DATA_CFG_VC_SELECT(fd.entry[i].bus.csi2.vc);
 		}
+	}
+}
 
-		csi2rx->count++;
-	} else {
-		csi2rx->count--;
+static int csi2rx_enable_streams(struct v4l2_subdev *subdev,
+				 struct v4l2_subdev_state *state, u32 pad,
+				 u64 streams_mask)
+{
+	struct csi2rx_priv *csi2rx = v4l2_subdev_to_csi2rx(subdev);
+	u64 sink_streams;
+	int ret;
+
+	sink_streams = v4l2_subdev_state_xlate_streams(state, pad,
+						       CSI2RX_PAD_SINK,
+						       &streams_mask);
+
+	guard(mutex)(&csi2rx->lock);
+	/*
+	 * If we're not the first users, there's no need to
+	 * enable the whole controller.
+	 */
+	if (!csi2rx->count) {
+		csi2rx_update_vc_select(csi2rx, state);
+		ret = csi2rx_start(csi2rx);
+		if (ret)
+			return ret;
+	}
 
-		/*
-		 * Let the last user turn off the lights.
-		 */
+	/* Start streaming on the source */
+	ret = v4l2_subdev_enable_streams(csi2rx->source_subdev, csi2rx->source_pad,
+					 sink_streams);
+	if (ret) {
+		dev_err(csi2rx->dev,
+			"Failed to start streams %#llx on subdev\n",
+			sink_streams);
 		if (!csi2rx->count)
 			csi2rx_stop(csi2rx);
+		return ret;
+	}
+
+	csi2rx->count++;
+	return 0;
+}
+
+static int csi2rx_disable_streams(struct v4l2_subdev *subdev,
+				  struct v4l2_subdev_state *state, u32 pad,
+				  u64 streams_mask)
+{
+	struct csi2rx_priv *csi2rx = v4l2_subdev_to_csi2rx(subdev);
+	u64 sink_streams;
+
+	sink_streams = v4l2_subdev_state_xlate_streams(state, pad,
+						       CSI2RX_PAD_SINK,
+						       &streams_mask);
+
+	if (v4l2_subdev_disable_streams(csi2rx->source_subdev,
+					csi2rx->source_pad, sink_streams)) {
+		dev_err(csi2rx->dev, "Couldn't disable our subdev\n");
+	}
+
+	guard(mutex)(&csi2rx->lock);
+	csi2rx->count--;
+
+	/* Let the last user turn off the lights. */
+	if (!csi2rx->count)
+		csi2rx_stop(csi2rx);
+
+	return 0;
+}
+
+static int csi2rx_s_stream_fallback(struct v4l2_subdev *sd, int enable)
+{
+	struct v4l2_subdev_state *state;
+	struct v4l2_subdev_route *route;
+	u64 mask[CSI2RX_PAD_MAX] = {0};
+	int i, ret;
+
+	/* Find the stream mask on all source pads */
+	state = v4l2_subdev_lock_and_get_active_state(sd);
+	for (i = CSI2RX_PAD_SOURCE_STREAM0; i < CSI2RX_PAD_MAX; i++) {
+		for_each_active_route(&state->routing, route) {
+			if (route->source_pad == i)
+				mask[i] |= BIT_ULL(route->source_stream);
+		}
+	}
+	v4l2_subdev_unlock_state(state);
+
+	/* Start streaming on each pad */
+	for (i = CSI2RX_PAD_SOURCE_STREAM0; i < CSI2RX_PAD_MAX; i++) {
+		if (enable)
+			ret = v4l2_subdev_enable_streams(sd, i, mask[i]);
+		else
+			ret = v4l2_subdev_disable_streams(sd, i, mask[i]);
+		if (ret)
+			return ret;
 	}
 
-out:
-	mutex_unlock(&csi2rx->lock);
 	return ret;
 }
 
@@ -424,12 +523,56 @@ static int csi2rx_enum_mbus_code(struct v4l2_subdev *subdev,
 	return 0;
 }
 
+static int _csi2rx_set_routing(struct v4l2_subdev *subdev,
+			       struct v4l2_subdev_state *state,
+			       struct v4l2_subdev_krouting *routing)
+{
+	static const struct v4l2_mbus_framefmt format = {
+		.width = 640,
+		.height = 480,
+		.code = MEDIA_BUS_FMT_UYVY8_1X16,
+		.field = V4L2_FIELD_NONE,
+		.colorspace = V4L2_COLORSPACE_SRGB,
+		.ycbcr_enc = V4L2_YCBCR_ENC_601,
+		.quantization = V4L2_QUANTIZATION_LIM_RANGE,
+		.xfer_func = V4L2_XFER_FUNC_SRGB,
+	};
+	int ret;
+
+	if (routing->num_routes > V4L2_FRAME_DESC_ENTRY_MAX)
+		return -EINVAL;
+
+	ret = v4l2_subdev_routing_validate(subdev, routing,
+					   V4L2_SUBDEV_ROUTING_ONLY_1_TO_1);
+	if (ret)
+		return ret;
+
+	return v4l2_subdev_set_routing_with_fmt(subdev, state, routing, &format);
+}
+
+static int csi2rx_set_routing(struct v4l2_subdev *subdev,
+			      struct v4l2_subdev_state *state,
+			      enum v4l2_subdev_format_whence which,
+			      struct v4l2_subdev_krouting *routing)
+{
+	struct csi2rx_priv *csi2rx = v4l2_subdev_to_csi2rx(subdev);
+	int ret;
+
+	if (which == V4L2_SUBDEV_FORMAT_ACTIVE && csi2rx->count)
+		return -EBUSY;
+
+	ret = _csi2rx_set_routing(subdev, state, routing);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
 static int csi2rx_set_fmt(struct v4l2_subdev *subdev,
 			  struct v4l2_subdev_state *state,
 			  struct v4l2_subdev_format *format)
 {
 	struct v4l2_mbus_framefmt *fmt;
-	unsigned int i;
 
 	/* No transcoding, source and sink formats must match. */
 	if (format->pad != CSI2RX_PAD_SINK)
@@ -441,14 +584,16 @@ static int csi2rx_set_fmt(struct v4l2_subdev *subdev,
 	format->format.field = V4L2_FIELD_NONE;
 
 	/* Set sink format */
-	fmt = v4l2_subdev_state_get_format(state, format->pad);
+	fmt = v4l2_subdev_state_get_format(state, format->pad, format->stream);
 	*fmt = format->format;
 
-	/* Propagate to source formats */
-	for (i = CSI2RX_PAD_SOURCE_STREAM0; i < CSI2RX_PAD_MAX; i++) {
-		fmt = v4l2_subdev_state_get_format(state, i);
-		*fmt = format->format;
-	}
+	/* Propagate to source format */
+	fmt = v4l2_subdev_state_get_opposite_stream_format(state, format->pad,
+							   format->stream);
+	if (!fmt)
+		return -EINVAL;
+
+	*fmt = format->format;
 
 	return 0;
 }
@@ -456,40 +601,92 @@ static int csi2rx_set_fmt(struct v4l2_subdev *subdev,
 static int csi2rx_init_state(struct v4l2_subdev *subdev,
 			     struct v4l2_subdev_state *state)
 {
-	struct v4l2_subdev_format format = {
-		.pad = CSI2RX_PAD_SINK,
-		.format = {
-			.width = 640,
-			.height = 480,
-			.code = MEDIA_BUS_FMT_UYVY8_1X16,
-			.field = V4L2_FIELD_NONE,
-			.colorspace = V4L2_COLORSPACE_SRGB,
-			.ycbcr_enc = V4L2_YCBCR_ENC_601,
-			.quantization = V4L2_QUANTIZATION_LIM_RANGE,
-			.xfer_func = V4L2_XFER_FUNC_SRGB,
+	struct v4l2_subdev_route routes[] = {
+		{
+			.sink_pad = CSI2RX_PAD_SINK,
+			.sink_stream = 0,
+			.source_pad = CSI2RX_PAD_SOURCE_STREAM0,
+			.source_stream = 0,
+			.flags = V4L2_SUBDEV_ROUTE_FL_ACTIVE,
 		},
 	};
 
-	return csi2rx_set_fmt(subdev, state, &format);
+	struct v4l2_subdev_krouting routing = {
+		.num_routes = ARRAY_SIZE(routes),
+		.routes = routes,
+	};
+
+	return _csi2rx_set_routing(subdev, state, &routing);
 }
 
 static int csi2rx_get_frame_desc(struct v4l2_subdev *subdev, unsigned int pad,
 				 struct v4l2_mbus_frame_desc *fd)
 {
 	struct csi2rx_priv *csi2rx = v4l2_subdev_to_csi2rx(subdev);
+	struct v4l2_mbus_frame_desc source_fd = {0};
+	struct v4l2_subdev_route *route;
+	struct v4l2_subdev_state *state;
+	int ret;
+
+	ret = csi2rx_get_frame_desc_from_source(csi2rx, &source_fd);
+	if (ret)
+		return ret;
+
+	fd->type = V4L2_MBUS_FRAME_DESC_TYPE_CSI2;
 
-	return csi2rx_get_frame_desc_from_source(csi2rx, fd);
+	state = v4l2_subdev_lock_and_get_active_state(subdev);
+
+	for_each_active_route(&state->routing, route) {
+		struct v4l2_mbus_frame_desc_entry *source_entry = NULL;
+		unsigned int i;
+
+		if (route->source_pad != pad)
+			continue;
+
+		for (i = 0; i < source_fd.num_entries; i++) {
+			if (source_fd.entry[i].stream == route->sink_stream) {
+				source_entry = &source_fd.entry[i];
+				break;
+			}
+		}
+
+		if (!source_entry) {
+			dev_err(csi2rx->dev,
+				"Failed to find stream from source frame desc\n");
+			ret = -EPIPE;
+			goto err_missing_stream;
+		}
+
+		fd->entry[fd->num_entries].stream = route->source_stream;
+		fd->entry[fd->num_entries].flags = source_entry->flags;
+		fd->entry[fd->num_entries].length = source_entry->length;
+		fd->entry[fd->num_entries].pixelcode = source_entry->pixelcode;
+		fd->entry[fd->num_entries].bus.csi2.vc =
+			source_entry->bus.csi2.vc;
+		fd->entry[fd->num_entries].bus.csi2.dt =
+			source_entry->bus.csi2.dt;
+
+		fd->num_entries++;
+	}
+
+err_missing_stream:
+	v4l2_subdev_unlock_state(state);
+
+	return ret;
 }
 
 static const struct v4l2_subdev_pad_ops csi2rx_pad_ops = {
-	.enum_mbus_code	= csi2rx_enum_mbus_code,
-	.get_fmt	= v4l2_subdev_get_fmt,
-	.set_fmt	= csi2rx_set_fmt,
-	.get_frame_desc	= csi2rx_get_frame_desc,
+	.enum_mbus_code		= csi2rx_enum_mbus_code,
+	.get_fmt		= v4l2_subdev_get_fmt,
+	.set_fmt		= csi2rx_set_fmt,
+	.get_frame_desc		= csi2rx_get_frame_desc,
+	.set_routing		= csi2rx_set_routing,
+	.enable_streams		= csi2rx_enable_streams,
+	.disable_streams	= csi2rx_disable_streams,
 };
 
 static const struct v4l2_subdev_video_ops csi2rx_video_ops = {
-	.s_stream	= csi2rx_s_stream,
+	.s_stream	= csi2rx_s_stream_fallback,
 };
 
 static const struct v4l2_subdev_ops csi2rx_subdev_ops = {
@@ -721,7 +918,8 @@ static int csi2rx_probe(struct platform_device *pdev)
 	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;
-	csi2rx->subdev.flags |= V4L2_SUBDEV_FL_HAS_DEVNODE;
+	csi2rx->subdev.flags = V4L2_SUBDEV_FL_HAS_DEVNODE |
+		V4L2_SUBDEV_FL_STREAMS;
 	csi2rx->subdev.entity.ops = &csi2rx_media_ops;
 
 	ret = media_entity_pads_init(&csi2rx->subdev.entity, CSI2RX_PAD_MAX,
-- 
2.34.1

[PATCH v4 10/12] media: ti: j721e-csi2rx: add multistream support

[Rishikesh Donadkar]

From: Jai Luthra <j-luthra@ti.com>

Each CSI2 stream can be multiplexed into 4 independent streams, each
identified by its virtual channel number. To capture this multiplexed
stream, the application needs to tell the driver how it wants to route
the data. It needs to specify which context should process which stream.
This is done via the new routing APIs.

Add ioctls to accept routing information from the application and save
that in the driver. This can be used when starting streaming on a
context to determine which route and consequently which virtual channel
it should process.

Support the new enable_stream()/disable_stream() APIs in the subdev
instead of s_stream() hook.

De-assert the pixel interface reset on first start_streaming() and assert
it on the last stop_streaming().

Co-developed-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
Signed-off-by: Jai Luthra <j-luthra@ti.com>
Co-developed-by: Rishikesh Donadkar <r-donadkar@ti.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 191 +++++++++++++-----
 1 file changed, 135 insertions(+), 56 deletions(-)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index 9f05d16ef8c1d..7986f96c5e11b 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -139,17 +139,6 @@ static inline struct ti_csi2rx_dev *to_csi2rx_dev(struct v4l2_subdev *sd)
 	return container_of(sd, struct ti_csi2rx_dev, subdev);
 }
 
-static const struct v4l2_mbus_framefmt ti_csi2rx_default_fmt = {
-	.width = 640,
-	.height = 480,
-	.code = MEDIA_BUS_FMT_UYVY8_1X16,
-	.field = V4L2_FIELD_NONE,
-	.colorspace = V4L2_COLORSPACE_SRGB,
-	.ycbcr_enc = V4L2_YCBCR_ENC_601,
-	.quantization = V4L2_QUANTIZATION_LIM_RANGE,
-	.xfer_func = V4L2_XFER_FUNC_SRGB,
-};
-
 static const struct ti_csi2rx_fmt ti_csi2rx_formats[] = {
 	{
 		.fourcc			= V4L2_PIX_FMT_YUYV,
@@ -556,8 +545,10 @@ static void ti_csi2rx_setup_shim(struct ti_csi2rx_ctx *ctx)
 	fmt = find_format_by_fourcc(ctx->v_fmt.fmt.pix.pixelformat);
 
 	/* De-assert the pixel interface reset. */
-	reg = SHIM_CNTL_PIX_RST;
-	writel(reg, csi->shim + SHIM_CNTL);
+	if (!csi->enable_count) {
+		reg = SHIM_CNTL_PIX_RST;
+		writel(reg, csi->shim + SHIM_CNTL);
+	}
 
 	reg = SHIM_DMACNTX_EN;
 	reg |= FIELD_PREP(SHIM_DMACNTX_FMT, fmt->csi_dt);
@@ -1002,7 +993,10 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 	dma->state = TI_CSI2RX_DMA_ACTIVE;
 	spin_unlock_irqrestore(&dma->lock, flags);
 
-	ret = v4l2_subdev_call(&csi->subdev, video, s_stream, 1);
+	/* Start stream 0, we don't allow multiple streams on the source pad */
+	ret = v4l2_subdev_enable_streams(&csi->subdev,
+					 TI_CSI2RX_PAD_FIRST_SOURCE + ctx->idx,
+					 BIT(0));
 	if (ret)
 		goto err_dma;
 
@@ -1025,16 +1019,20 @@ static void ti_csi2rx_stop_streaming(struct vb2_queue *vq)
 	struct ti_csi2rx_dev *csi = ctx->csi;
 	int ret;
 
-	video_device_pipeline_stop(&ctx->vdev);
+	/* assert pixel reset to prevent stale data */
+	if (csi->enable_count == 1) {
+		writel(0, csi->shim + SHIM_CNTL);
 
-	writel(0, csi->shim + SHIM_CNTL);
-	writel(0, csi->shim + SHIM_DMACNTX(ctx->idx));
+		mutex_lock(&csi->mutex);
+		csi->vc_cached = false;
+		mutex_unlock(&csi->mutex);
+	}
 
-	mutex_lock(&csi->mutex);
-	csi->vc_cached = false;
-	mutex_unlock(&csi->mutex);
+	video_device_pipeline_stop(&ctx->vdev);
 
-	ret = v4l2_subdev_call(&csi->subdev, video, s_stream, 0);
+	ret = v4l2_subdev_disable_streams(&csi->subdev,
+					  TI_CSI2RX_PAD_FIRST_SOURCE + ctx->idx,
+					  BIT(0));
 	if (ret)
 		dev_err(csi->dev, "Failed to stop subdev stream\n");
 
@@ -1080,57 +1078,140 @@ static int ti_csi2rx_sd_set_fmt(struct v4l2_subdev *sd,
 	fmt = v4l2_subdev_state_get_format(state, format->pad, format->stream);
 	*fmt = format->format;
 
-	fmt = v4l2_subdev_state_get_format(state, TI_CSI2RX_PAD_FIRST_SOURCE,
-					   format->stream);
+	fmt = v4l2_subdev_state_get_opposite_stream_format(state, format->pad,
+							   format->stream);
+	if (!fmt)
+		return -EINVAL;
+
 	*fmt = format->format;
 
 	return 0;
 }
 
-static int ti_csi2rx_sd_init_state(struct v4l2_subdev *sd,
-				   struct v4l2_subdev_state *state)
+static int _ti_csi2rx_sd_set_routing(struct v4l2_subdev *sd,
+				     struct v4l2_subdev_state *state,
+				     struct v4l2_subdev_krouting *routing)
 {
-	struct v4l2_mbus_framefmt *fmt;
+	int ret;
 
-	fmt = v4l2_subdev_state_get_format(state, TI_CSI2RX_PAD_SINK);
-	*fmt = ti_csi2rx_default_fmt;
+	const struct v4l2_mbus_framefmt format = {
+		.width = 640,
+		.height = 480,
+		.code = MEDIA_BUS_FMT_UYVY8_1X16,
+		.field = V4L2_FIELD_NONE,
+		.colorspace = V4L2_COLORSPACE_SRGB,
+		.ycbcr_enc = V4L2_YCBCR_ENC_601,
+		.quantization = V4L2_QUANTIZATION_LIM_RANGE,
+		.xfer_func = V4L2_XFER_FUNC_SRGB,
+	};
 
-	fmt = v4l2_subdev_state_get_format(state, TI_CSI2RX_PAD_FIRST_SOURCE);
-	*fmt = ti_csi2rx_default_fmt;
+	ret = v4l2_subdev_routing_validate(sd, routing,
+					   V4L2_SUBDEV_ROUTING_ONLY_1_TO_1 |
+					   V4L2_SUBDEV_ROUTING_NO_SOURCE_MULTIPLEXING);
 
-	return 0;
+	if (ret)
+		return ret;
+
+	/* Only stream ID 0 allowed on source pads */
+	for (unsigned int i = 0; i < routing->num_routes; ++i) {
+		const struct v4l2_subdev_route *route = &routing->routes[i];
+
+		if (route->source_stream != 0)
+			return -EINVAL;
+	}
+
+	ret = v4l2_subdev_set_routing_with_fmt(sd, state, routing, &format);
+
+	return ret;
+}
+
+static int ti_csi2rx_sd_set_routing(struct v4l2_subdev *sd,
+				    struct v4l2_subdev_state *state,
+				    enum v4l2_subdev_format_whence which,
+				    struct v4l2_subdev_krouting *routing)
+{
+	struct ti_csi2rx_dev *csi = to_csi2rx_dev(sd);
+
+	if (csi->enable_count > 0)
+		return -EBUSY;
+
+	return _ti_csi2rx_sd_set_routing(sd, state, routing);
+}
+
+static int ti_csi2rx_sd_init_state(struct v4l2_subdev *sd,
+				   struct v4l2_subdev_state *state)
+{
+	struct v4l2_subdev_route routes[] = { {
+		.sink_pad = 0,
+		.sink_stream = 0,
+		.source_pad = TI_CSI2RX_PAD_FIRST_SOURCE,
+		.source_stream = 0,
+		.flags = V4L2_SUBDEV_ROUTE_FL_ACTIVE,
+	} };
+
+	struct v4l2_subdev_krouting routing = {
+		.num_routes = 1,
+		.routes = routes,
+	};
+
+	/* Initialize routing to single route to the fist source pad */
+	return _ti_csi2rx_sd_set_routing(sd, state, &routing);
 }
 
-static int ti_csi2rx_sd_s_stream(struct v4l2_subdev *sd, int enable)
+static int ti_csi2rx_sd_enable_streams(struct v4l2_subdev *sd,
+				       struct v4l2_subdev_state *state,
+				       u32 pad, u64 streams_mask)
 {
 	struct ti_csi2rx_dev *csi = to_csi2rx_dev(sd);
+	struct media_pad *remote_pad;
+	u64 sink_streams;
 	int ret = 0;
 
-	mutex_lock(&csi->mutex);
+	remote_pad = media_entity_remote_source_pad_unique(&csi->subdev.entity);
+	if (!remote_pad)
+		return -ENODEV;
+	sink_streams = v4l2_subdev_state_xlate_streams(state, pad,
+						       TI_CSI2RX_PAD_SINK,
+						       &streams_mask);
 
-	if (enable) {
-		if (csi->enable_count > 0) {
-			csi->enable_count++;
-			goto out;
-		}
+	ret = v4l2_subdev_enable_streams(csi->source, remote_pad->index,
+					 sink_streams);
+	if (ret)
+		return ret;
 
-		ret = v4l2_subdev_call(csi->source, video, s_stream, 1);
-		if (ret)
-			goto out;
+	mutex_lock(&csi->mutex);
+	csi->enable_count++;
+	mutex_unlock(&csi->mutex);
 
-		csi->enable_count++;
-	} else {
-		if (csi->enable_count == 0) {
-			ret = -EINVAL;
-			goto out;
-		}
+	return 0;
+}
+
+static int ti_csi2rx_sd_disable_streams(struct v4l2_subdev *sd,
+					struct v4l2_subdev_state *state,
+					u32 pad, u64 streams_mask)
+{
+	struct ti_csi2rx_dev *csi = to_csi2rx_dev(sd);
+	struct media_pad *remote_pad;
+	u64 sink_streams;
+	int ret = 0;
 
-		if (--csi->enable_count > 0)
-			goto out;
+	remote_pad = media_entity_remote_source_pad_unique(&csi->subdev.entity);
+	if (!remote_pad)
+		return -ENODEV;
+	sink_streams = v4l2_subdev_state_xlate_streams(state, pad,
+						       TI_CSI2RX_PAD_SINK,
+						       &streams_mask);
 
-		ret = v4l2_subdev_call(csi->source, video, s_stream, 0);
+	mutex_lock(&csi->mutex);
+	if (csi->enable_count == 0) {
+		ret = -EINVAL;
+		goto out;
 	}
 
+	ret = v4l2_subdev_disable_streams(csi->source, remote_pad->index,
+					  sink_streams);
+	if (!ret)
+		--csi->enable_count;
 out:
 	mutex_unlock(&csi->mutex);
 	return ret;
@@ -1138,16 +1219,14 @@ static int ti_csi2rx_sd_s_stream(struct v4l2_subdev *sd, int enable)
 
 static const struct v4l2_subdev_pad_ops ti_csi2rx_subdev_pad_ops = {
 	.enum_mbus_code	= ti_csi2rx_enum_mbus_code,
+	.set_routing = ti_csi2rx_sd_set_routing,
 	.get_fmt = v4l2_subdev_get_fmt,
 	.set_fmt = ti_csi2rx_sd_set_fmt,
-};
-
-static const struct v4l2_subdev_video_ops ti_csi2rx_subdev_video_ops = {
-	.s_stream = ti_csi2rx_sd_s_stream,
+	.enable_streams = ti_csi2rx_sd_enable_streams,
+	.disable_streams = ti_csi2rx_sd_disable_streams,
 };
 
 static const struct v4l2_subdev_ops ti_csi2rx_subdev_ops = {
-	.video = &ti_csi2rx_subdev_video_ops,
 	.pad = &ti_csi2rx_subdev_pad_ops,
 };
 
@@ -1329,7 +1408,7 @@ static int ti_csi2rx_v4l2_init(struct ti_csi2rx_dev *csi)
 	v4l2_subdev_init(sd, &ti_csi2rx_subdev_ops);
 	sd->internal_ops = &ti_csi2rx_internal_ops;
 	sd->entity.function = MEDIA_ENT_F_VID_IF_BRIDGE;
-	sd->flags = V4L2_SUBDEV_FL_HAS_DEVNODE;
+	sd->flags = V4L2_SUBDEV_FL_HAS_DEVNODE | V4L2_SUBDEV_FL_STREAMS;
 	strscpy(sd->name, dev_name(csi->dev), sizeof(sd->name));
 	sd->dev = csi->dev;
 	sd->entity.ops = &ti_csi2rx_subdev_entity_ops;
-- 
2.34.1

[PATCH v4 11/12] media: ti: j721e-csi2rx: Submit all available buffers

[Rishikesh Donadkar]

From: Jai Luthra <j-luthra@ti.com>

We already make sure to submit all available buffers to DMA in each DMA
completion callback.

Move that logic in a separate function, and use it during stream start
as well, as most application queue all their buffers before stream on.

Signed-off-by: Jai Luthra <j-luthra@ti.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 43 +++++++++++--------
 1 file changed, 24 insertions(+), 19 deletions(-)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index 7986f96c5e11b..ba2a30bfed37d 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -651,6 +651,27 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
 	return ret;
 }
 
+static int ti_csi2rx_dma_submit_pending(struct ti_csi2rx_ctx *ctx)
+{
+	struct ti_csi2rx_dma *dma = &ctx->dma;
+	struct ti_csi2rx_buffer *buf;
+	int ret = 0;
+
+	/* If there are more buffers to process then start their transfer. */
+	while (!list_empty(&dma->queue)) {
+		buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer, list);
+		ret = ti_csi2rx_start_dma(ctx, buf);
+		if (ret) {
+			dev_err(ctx->csi->dev,
+				"Failed to queue the next buffer for DMA\n");
+			vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_ERROR);
+			break;
+		}
+		list_move_tail(&buf->list, &dma->submitted);
+	}
+	return ret;
+}
+
 static void ti_csi2rx_dma_callback(void *param)
 {
 	struct ti_csi2rx_buffer *buf = param;
@@ -671,18 +692,7 @@ static void ti_csi2rx_dma_callback(void *param)
 	vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_DONE);
 	list_del(&buf->list);
 
-	/* If there are more buffers to process then start their transfer. */
-	while (!list_empty(&dma->queue)) {
-		buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer, list);
-
-		if (ti_csi2rx_start_dma(ctx, buf)) {
-			dev_err(ctx->csi->dev,
-				"Failed to queue the next buffer for DMA\n");
-			vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_ERROR);
-		} else {
-			list_move_tail(&buf->list, &dma->submitted);
-		}
-	}
+	ti_csi2rx_dma_submit_pending(ctx);
 
 	if (list_empty(&dma->submitted))
 		dma->state = TI_CSI2RX_DMA_IDLE;
@@ -941,7 +951,6 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vq);
 	struct ti_csi2rx_dev *csi = ctx->csi;
 	struct ti_csi2rx_dma *dma = &ctx->dma;
-	struct ti_csi2rx_buffer *buf;
 	unsigned long flags;
 	int ret = 0;
 
@@ -980,16 +989,13 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 	ctx->sequence = 0;
 
 	spin_lock_irqsave(&dma->lock, flags);
-	buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer, list);
 
-	ret = ti_csi2rx_start_dma(ctx, buf);
+	ret = ti_csi2rx_dma_submit_pending(ctx);
 	if (ret) {
-		dev_err(csi->dev, "Failed to start DMA: %d\n", ret);
 		spin_unlock_irqrestore(&dma->lock, flags);
-		goto err_pipeline;
+		goto err_dma;
 	}
 
-	list_move_tail(&buf->list, &dma->submitted);
 	dma->state = TI_CSI2RX_DMA_ACTIVE;
 	spin_unlock_irqrestore(&dma->lock, flags);
 
@@ -1004,7 +1010,6 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq, unsigned int count)
 
 err_dma:
 	ti_csi2rx_stop_dma(ctx);
-err_pipeline:
 	video_device_pipeline_stop(&ctx->vdev);
 	writel(0, csi->shim + SHIM_CNTL);
 	writel(0, csi->shim + SHIM_DMACNTX(ctx->idx));
-- 
2.34.1

[PATCH v4 12/12] media: ti: j721e-csi2rx: Change the drain architecture for multistream

[Rishikesh Donadkar]

On buffer starvation the DMA is marked IDLE, and the stale data in the
internal FIFOs gets drained only on the next VIDIOC_QBUF call from the
userspace. This approach works fine for a single stream case.

But in multistream scenarios, buffer starvation for one stream i.e. one
virtual channel, can block the shared HW FIFO of the CSI2RX IP. This can
stall the pipeline for all other virtual channels, even if buffers are
available for them.

This patch introduces a new architecture, that continuously drains data
from the shared HW FIFO into a small (32KiB) buffer if no buffers are made
available to the driver from the userspace. This ensures independence
between different streams, where a slower downstream element for one
camera does not block streaming for other cameras.

Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
---
 .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 96 +++++++------------
 1 file changed, 33 insertions(+), 63 deletions(-)

diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
index ba2a30bfed37d..3b046d3cf7e5a 100644
--- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
+++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
@@ -57,7 +57,6 @@
 #define TI_CSI2RX_MAX_SOURCE_PADS	TI_CSI2RX_MAX_CTX
 #define TI_CSI2RX_MAX_PADS		(1 + TI_CSI2RX_MAX_SOURCE_PADS)
 
-#define DRAIN_TIMEOUT_MS		50
 #define DRAIN_BUFFER_SIZE		SZ_32K
 
 struct ti_csi2rx_fmt {
@@ -77,7 +76,6 @@ struct ti_csi2rx_buffer {
 
 enum ti_csi2rx_dma_state {
 	TI_CSI2RX_DMA_STOPPED,	/* Streaming not started yet. */
-	TI_CSI2RX_DMA_IDLE,	/* Streaming but no pending DMA operation. */
 	TI_CSI2RX_DMA_ACTIVE,	/* Streaming and pending DMA operation. */
 };
 
@@ -245,6 +243,10 @@ static const struct ti_csi2rx_fmt ti_csi2rx_formats[] = {
 static int ti_csi2rx_start_dma(struct ti_csi2rx_ctx *ctx,
 			       struct ti_csi2rx_buffer *buf);
 
+/* Forward declarations needed by ti_csi2rx_drain_callback. */
+static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx);
+static int ti_csi2rx_dma_submit_pending(struct ti_csi2rx_ctx *ctx);
+
 static const struct ti_csi2rx_fmt *find_format_by_fourcc(u32 pixelformat)
 {
 	unsigned int i;
@@ -596,9 +598,28 @@ static void ti_csi2rx_setup_shim(struct ti_csi2rx_ctx *ctx)
 
 static void ti_csi2rx_drain_callback(void *param)
 {
-	struct completion *drain_complete = param;
+	struct ti_csi2rx_ctx *ctx = param;
+	struct ti_csi2rx_dma *dma = &ctx->dma;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dma->lock, flags);
 
-	complete(drain_complete);
+	if (dma->state == TI_CSI2RX_DMA_STOPPED) {
+		spin_unlock_irqrestore(&dma->lock, flags);
+		return;
+	}
+
+	/*
+	 * If dma->queue is empty, it signals no buffer has arrived from
+	 * user space, so, queue more transaction to drain dma
+	 */
+	if (list_empty(&dma->queue)) {
+		if (ti_csi2rx_drain_dma(ctx))
+			dev_warn(ctx->csi->dev, "DMA drain failed\n");
+	} else {
+		ti_csi2rx_dma_submit_pending(ctx);
+	}
+	spin_unlock_irqrestore(&dma->lock, flags);
 }
 
 /*
@@ -616,12 +637,9 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
 {
 	struct ti_csi2rx_dev *csi = ctx->csi;
 	struct dma_async_tx_descriptor *desc;
-	struct completion drain_complete;
 	dma_cookie_t cookie;
 	int ret;
 
-	init_completion(&drain_complete);
-
 	desc = dmaengine_prep_slave_single(ctx->dma.chan, csi->drain.paddr,
 					   csi->drain.len, DMA_DEV_TO_MEM,
 					   DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
@@ -631,7 +649,7 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
 	}
 
 	desc->callback = ti_csi2rx_drain_callback;
-	desc->callback_param = &drain_complete;
+	desc->callback_param = ctx;
 
 	cookie = dmaengine_submit(desc);
 	ret = dma_submit_error(cookie);
@@ -640,13 +658,6 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
 
 	dma_async_issue_pending(ctx->dma.chan);
 
-	if (!wait_for_completion_timeout(&drain_complete,
-					 msecs_to_jiffies(DRAIN_TIMEOUT_MS))) {
-		dmaengine_terminate_sync(ctx->dma.chan);
-		dev_dbg(csi->dev, "DMA transfer timed out for drain buffer\n");
-		ret = -ETIMEDOUT;
-		goto out;
-	}
 out:
 	return ret;
 }
@@ -694,9 +705,11 @@ static void ti_csi2rx_dma_callback(void *param)
 
 	ti_csi2rx_dma_submit_pending(ctx);
 
-	if (list_empty(&dma->submitted))
-		dma->state = TI_CSI2RX_DMA_IDLE;
-
+	if (list_empty(&dma->submitted)) {
+		if (ti_csi2rx_drain_dma(ctx))
+			dev_warn(ctx->csi->dev,
+				 "DMA drain failed on one of the transactions\n");
+	}
 	spin_unlock_irqrestore(&dma->lock, flags);
 }
 
@@ -749,7 +762,7 @@ static void ti_csi2rx_stop_dma(struct ti_csi2rx_ctx *ctx)
 		 * enforced before terminating DMA.
 		 */
 		ret = ti_csi2rx_drain_dma(ctx);
-		if (ret && ret != -ETIMEDOUT)
+		if (ret)
 			dev_warn(ctx->csi->dev,
 				 "Failed to drain DMA. Next frame might be bogus\n");
 	}
@@ -816,57 +829,14 @@ static void ti_csi2rx_buffer_queue(struct vb2_buffer *vb)
 	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
 	struct ti_csi2rx_buffer *buf;
 	struct ti_csi2rx_dma *dma = &ctx->dma;
-	bool restart_dma = false;
 	unsigned long flags = 0;
-	int ret;
 
 	buf = container_of(vb, struct ti_csi2rx_buffer, vb.vb2_buf);
 	buf->ctx = ctx;
 
 	spin_lock_irqsave(&dma->lock, flags);
-	/*
-	 * Usually the DMA callback takes care of queueing the pending buffers.
-	 * But if DMA has stalled due to lack of buffers, restart it now.
-	 */
-	if (dma->state == TI_CSI2RX_DMA_IDLE) {
-		/*
-		 * Do not restart DMA with the lock held because
-		 * ti_csi2rx_drain_dma() might block for completion.
-		 * There won't be a race on queueing DMA anyway since the
-		 * callback is not being fired.
-		 */
-		restart_dma = true;
-		dma->state = TI_CSI2RX_DMA_ACTIVE;
-	} else {
-		list_add_tail(&buf->list, &dma->queue);
-	}
+	list_add_tail(&buf->list, &dma->queue);
 	spin_unlock_irqrestore(&dma->lock, flags);
-
-	if (restart_dma) {
-		/*
-		 * Once frames start dropping, some data gets stuck in the DMA
-		 * pipeline somewhere. So the first DMA transfer after frame
-		 * drops gives a partial frame. This is obviously not useful to
-		 * the application and will only confuse it. Issue a DMA
-		 * transaction to drain that up.
-		 */
-		ret = ti_csi2rx_drain_dma(ctx);
-		if (ret && ret != -ETIMEDOUT)
-			dev_warn(ctx->csi->dev,
-				 "Failed to drain DMA. Next frame might be bogus\n");
-
-		spin_lock_irqsave(&dma->lock, flags);
-		ret = ti_csi2rx_start_dma(ctx, buf);
-		if (ret) {
-			vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_ERROR);
-			dma->state = TI_CSI2RX_DMA_IDLE;
-			spin_unlock_irqrestore(&dma->lock, flags);
-			dev_err(ctx->csi->dev, "Failed to start DMA: %d\n", ret);
-		} else {
-			list_add_tail(&buf->list, &dma->submitted);
-			spin_unlock_irqrestore(&dma->lock, flags);
-		}
-	}
 }
 
 static int ti_csi2rx_get_route(struct ti_csi2rx_ctx *ctx)
-- 
2.34.1

Re: [PATCH v4 00/12] media: cadence,ti: CSI2RX Multistream Support

[Yemike Abhilash Chandra]

Hi Rishikesh,

Thank you for the patch series.

On 14/05/25 16:55, Rishikesh Donadkar wrote:
> This series adds multi-stream support for Cadence CSI2RX and TI CSI2RX
> SHIM drivers.
> 
> PATCH 1-6:  Support multiple DMA contexts/video nodes in TI CSI2RX
> PATCH 7-8:  Use get_frame_desc to propagate virtual channel
>              information across Cadence and TI CSI-RX subdevs
> PATCH 9-10: Use new multi-stream APIs across the drivers to support
>              multiplexed cameras from sources like UB960 (FPDLink)
> PATCH 11:   Optimize stream on by submitting all queued buffers to DMA
> PATCH 12:   Change the drain architecture to support multi-stream
> 
> Testing for this series has been done on top of media tree with 4x IMX219
> camera modules connected to TI's AM62A using V3 Link fusion mini board.
> 
> Overlay and defconfig changes for the same can be found below:
> https://github.com/RISHI27-dot/linux/commits/u/multistream_v4
> 
> v4l2-compliance results:
> https://gist.github.com/Rishikesh-D/4ddf957da2f62191eefe7a9e5ff212ab
> 

For the entire series,

Reviewed-by: Yemike Abhilash Chandra <y-abhilashchandra@ti.com>

> ---
> Changes in v4:
> 
> [PATCH 01/13] dt-bindings: media: ti,j721e-csi2rx-shim: Support 32 dma chans
>    - No change
> [PATCH 02/13] media: ti: j721e-csi2rx: separate out device and context
>    - Add ctx identifier in the dev_err() message
>    - No change
> [PATCH 04/13] media: ti: j721e-csi2rx: prepare SHIM code for multiple contexts
>    - Reduced the name string lenght from 32 chars to 5 chars
> [PATCH 05/13] media: ti: j721e-csi2rx: add a subdev for the core device
>    - Add .enum_mbus_code callback
>    - Replace statically allocated struct with a global static const struct
>      v4l2_mbus_framefmt and used that in the _init_state() function
> [PATCH 06/13] media: ti: j721e-csi2rx: get number of contexts from device tree
>    - Fix the drain buffer being leaked
>    - If the shows more number of ctx than the TI_CSI2RX_MAX_CTX, return an error
>      instead of warning
> [PATCH 07/13] media: cadence: csi2rx: add get_frame_desc wrapper
>    - No change
> [PATCH 08/13] media: ti: j721e-csi2rx: add support for processing virtual channels
>    - Call ti_csi2rx_get_vc() only once on first stream start and cache the VC data in
>      the driver, use the corresponding VC in all subsequent stream starts.
> [PATCH 09/13] media: cadence: csi2rx: Use new enable stream APIs
> [PATCH 10/13] media: cadence: csi2rx: Enable multi-stream support
>    - Squash the above two patches into
>      [PATCH v4 09/12] media: cadence: csi2rx: add multistream support
>    - Use already obtained csi2rx->source_pad in enable_streams() and
>      disable_streams() call
>    - Update commit message with the reason for using a custom helper for s_stream
>      instead of v4l2_subdev_s_stream_helper()
>    - Use v4l2_get_link_freq() variant that takes pad of the source as its first
>      argument instead of the one that takes v4l2_ctrl_handler
>    - Call v4l2_get_link_freq() with bpp = 0 to prevent fallback to V4L2_CID_PIXEL_RATE
>      in multi-stream case
>    - Use lock guards to simplify error handling
>    - Call csi2rx_update_vc_select() at first stream start before enabling the controller
> [PATCH 11/13] media: ti: j721e-csi2rx: add multistream support
>    - No change
> [PATCH 12/13] media: ti: j721e-csi2rx: Submit all available buffers
>    - No change
> [PATCH 13/13] media: ti: j721e-csi2rx: Change the drain architecture for multistream
>    - Fix checkpatch warning
>    - Change commit message to give a better description of the patch
> 
> Link to (v3):
>    https://lore.kernel.org/all/20250417065554.437541-1-r-donadkar@ti.com/
> 
> Changes in v3:
> 
> - Drop [PATCH v2 01/13] media: cadence: csi2rx: Support runtime PM from
>    v2, support for runtime PM will be added in a separate series:
>    https://lore.kernel.org/all/20250224-ti_csi_pm-v1-0-8f8c29ef646d@ideasonboard.com/
> - Change the drain architecture to prevent FIFO overflow in multistream
>    usecases.
> - With the new drain architecture, we don't need the the driver to wait
>    for userspace to start streaming on all "actively routed" video nodes
>    before starting streaming on the source. So, revert back to the capture
>    architecture where streams can be started and stopped independent
>    to each other.
> 
> Link to (v2):
>    https://lore.kernel.org/r/20240627-multistream-v2-0-6ae96c54c1c3@ti.com
> 
> Changes in v2:
> 
> - Change the multi-camera capture architecture to be similar to that of
>    Tomi's RPi5 FE series, where the driver will wait for userspace to
>    start streaming on all "actively routed" video nodes before starting
>    streaming on the source. This simplifies things a lot from the HW
>    perspective, which might run into deadlocks due to a shared FIFO
>    between multiple DMA channels.
> 
> - Drop a few fixes that were posted separately and are already merged
> - Fix dtschema warnings reported by Rob on [02/13]
> - Fix warnings for uninitialized `used_vc` variable in cdns-csi2rx.c
> - Return -EBUSY if someone updates routes for j721e-csi2rx subdev while
>    streaming
> - Only allow single-streams to be routed to the source pads (linked to
>    video nodes) of the j721e-csi2rx device
> - Squash the patches marked "SQUASH" in the v1 RFC series
> 
> Link to RFC (v1):
> https://lore.kernel.org/r/20240222-multistream-v1-0-1837ed916eeb@ti.com
> 
> 
> Jai Luthra (7):
>    dt-bindings: media: ti,j721e-csi2rx-shim: Support 32 dma chans
>    media: ti: j721e-csi2rx: separate out device and context
>    media: ti: j721e-csi2rx: add a subdev for the core device
>    media: ti: j721e-csi2rx: add support for processing virtual channels
>    media: cadence: csi2rx: add multistream support
>    media: ti: j721e-csi2rx: add multistream support
>    media: ti: j721e-csi2rx: Submit all available buffers
> 
> Pratyush Yadav (4):
>    media: ti: j721e-csi2rx: prepare SHIM code for multiple contexts
>    media: ti: j721e-csi2rx: allocate DMA channel based on context index
>    media: ti: j721e-csi2rx: get number of contexts from device tree
>    media: cadence: csi2rx: add get_frame_desc wrapper
> 
> Rishikesh Donadkar (1):
>    media: ti: j721e-csi2rx: Change the drain architecture for multistream
> 
>   .../bindings/media/ti,j721e-csi2rx-shim.yaml  |  39 +-
>   drivers/media/platform/cadence/cdns-csi2rx.c  | 372 +++++--
>   .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 958 +++++++++++++-----
>   3 files changed, 1025 insertions(+), 344 deletions(-)
> 

Re: [PATCH v4 12/12] media: ti: j721e-csi2rx: Change the drain architecture for multistream

[Jai Luthra]

[-- Attachment #1: Type: text/plain, Size: 9111 bytes --]

Hi Rishikesh,

Thanks for the patch.

Quoting Rishikesh Donadkar (2025-05-14 04:25:27)
> On buffer starvation the DMA is marked IDLE, and the stale data in the
> internal FIFOs gets drained only on the next VIDIOC_QBUF call from the
> userspace. This approach works fine for a single stream case.
> 
> But in multistream scenarios, buffer starvation for one stream i.e. one
> virtual channel, can block the shared HW FIFO of the CSI2RX IP. This can
> stall the pipeline for all other virtual channels, even if buffers are
> available for them.
> 
> This patch introduces a new architecture, that continuously drains data
> from the shared HW FIFO into a small (32KiB) buffer if no buffers are made
> available to the driver from the userspace. This ensures independence
> between different streams, where a slower downstream element for one
> camera does not block streaming for other cameras.
> 
> Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
> ---
>  .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 96 +++++++------------
>  1 file changed, 33 insertions(+), 63 deletions(-)
> 
> diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> index ba2a30bfed37d..3b046d3cf7e5a 100644
> --- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> +++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> @@ -57,7 +57,6 @@
>  #define TI_CSI2RX_MAX_SOURCE_PADS      TI_CSI2RX_MAX_CTX
>  #define TI_CSI2RX_MAX_PADS             (1 + TI_CSI2RX_MAX_SOURCE_PADS)
>  
> -#define DRAIN_TIMEOUT_MS               50
>  #define DRAIN_BUFFER_SIZE              SZ_32K
>  
>  struct ti_csi2rx_fmt {
> @@ -77,7 +76,6 @@ struct ti_csi2rx_buffer {
>  
>  enum ti_csi2rx_dma_state {
>         TI_CSI2RX_DMA_STOPPED,  /* Streaming not started yet. */
> -       TI_CSI2RX_DMA_IDLE,     /* Streaming but no pending DMA operation. */
>         TI_CSI2RX_DMA_ACTIVE,   /* Streaming and pending DMA operation. */
>  };
>  
> @@ -245,6 +243,10 @@ static const struct ti_csi2rx_fmt ti_csi2rx_formats[] = {
>  static int ti_csi2rx_start_dma(struct ti_csi2rx_ctx *ctx,
>                                struct ti_csi2rx_buffer *buf);
>  
> +/* Forward declarations needed by ti_csi2rx_drain_callback. */
> +static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx);
> +static int ti_csi2rx_dma_submit_pending(struct ti_csi2rx_ctx *ctx);
> +
>  static const struct ti_csi2rx_fmt *find_format_by_fourcc(u32 pixelformat)
>  {
>         unsigned int i;
> @@ -596,9 +598,28 @@ static void ti_csi2rx_setup_shim(struct ti_csi2rx_ctx *ctx)
>  
>  static void ti_csi2rx_drain_callback(void *param)
>  {
> -       struct completion *drain_complete = param;
> +       struct ti_csi2rx_ctx *ctx = param;
> +       struct ti_csi2rx_dma *dma = &ctx->dma;
> +       unsigned long flags;
> +
> +       spin_lock_irqsave(&dma->lock, flags);
>  
> -       complete(drain_complete);
> +       if (dma->state == TI_CSI2RX_DMA_STOPPED) {
> +               spin_unlock_irqrestore(&dma->lock, flags);
> +               return;
> +       }
> +
> +       /*
> +        * If dma->queue is empty, it signals no buffer has arrived from
> +        * user space, so, queue more transaction to drain dma
> +        */
> +       if (list_empty(&dma->queue)) {
> +               if (ti_csi2rx_drain_dma(ctx))
> +                       dev_warn(ctx->csi->dev, "DMA drain failed\n");
> +       } else {
> +               ti_csi2rx_dma_submit_pending(ctx);
> +       }
> +       spin_unlock_irqrestore(&dma->lock, flags);
>  }
>  
>  /*
> @@ -616,12 +637,9 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
>  {
>         struct ti_csi2rx_dev *csi = ctx->csi;
>         struct dma_async_tx_descriptor *desc;
> -       struct completion drain_complete;
>         dma_cookie_t cookie;
>         int ret;
>  
> -       init_completion(&drain_complete);
> -
>         desc = dmaengine_prep_slave_single(ctx->dma.chan, csi->drain.paddr,
>                                            csi->drain.len, DMA_DEV_TO_MEM,
>                                            DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
> @@ -631,7 +649,7 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
>         }
>  
>         desc->callback = ti_csi2rx_drain_callback;
> -       desc->callback_param = &drain_complete;
> +       desc->callback_param = ctx;
>  
>         cookie = dmaengine_submit(desc);
>         ret = dma_submit_error(cookie);
> @@ -640,13 +658,6 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
>  
>         dma_async_issue_pending(ctx->dma.chan);
>  
> -       if (!wait_for_completion_timeout(&drain_complete,
> -                                        msecs_to_jiffies(DRAIN_TIMEOUT_MS))) {
> -               dmaengine_terminate_sync(ctx->dma.chan);
> -               dev_dbg(csi->dev, "DMA transfer timed out for drain buffer\n");
> -               ret = -ETIMEDOUT;
> -               goto out;
> -       }
>  out:
>         return ret;
>  }
> @@ -694,9 +705,11 @@ static void ti_csi2rx_dma_callback(void *param)
>  
>         ti_csi2rx_dma_submit_pending(ctx);
>  
> -       if (list_empty(&dma->submitted))
> -               dma->state = TI_CSI2RX_DMA_IDLE;
> -
> +       if (list_empty(&dma->submitted)) {
> +               if (ti_csi2rx_drain_dma(ctx))
> +                       dev_warn(ctx->csi->dev,
> +                                "DMA drain failed on one of the transactions\n");
> +       }
>         spin_unlock_irqrestore(&dma->lock, flags);
>  }
>  
> @@ -749,7 +762,7 @@ static void ti_csi2rx_stop_dma(struct ti_csi2rx_ctx *ctx)
>                  * enforced before terminating DMA.
>                  */
>                 ret = ti_csi2rx_drain_dma(ctx);
> -               if (ret && ret != -ETIMEDOUT)
> +               if (ret)
>                         dev_warn(ctx->csi->dev,
>                                  "Failed to drain DMA. Next frame might be bogus\n");
>         }
> @@ -816,57 +829,14 @@ static void ti_csi2rx_buffer_queue(struct vb2_buffer *vb)
>         struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
>         struct ti_csi2rx_buffer *buf;
>         struct ti_csi2rx_dma *dma = &ctx->dma;
> -       bool restart_dma = false;
>         unsigned long flags = 0;
> -       int ret;
>  
>         buf = container_of(vb, struct ti_csi2rx_buffer, vb.vb2_buf);
>         buf->ctx = ctx;
>  
>         spin_lock_irqsave(&dma->lock, flags);
> -       /*
> -        * Usually the DMA callback takes care of queueing the pending buffers.
> -        * But if DMA has stalled due to lack of buffers, restart it now.
> -        */
> -       if (dma->state == TI_CSI2RX_DMA_IDLE) {
> -               /*
> -                * Do not restart DMA with the lock held because
> -                * ti_csi2rx_drain_dma() might block for completion.
> -                * There won't be a race on queueing DMA anyway since the
> -                * callback is not being fired.
> -                */
> -               restart_dma = true;
> -               dma->state = TI_CSI2RX_DMA_ACTIVE;
> -       } else {
> -               list_add_tail(&buf->list, &dma->queue);
> -       }
> +       list_add_tail(&buf->list, &dma->queue);
>         spin_unlock_irqrestore(&dma->lock, flags);
> -
> -       if (restart_dma) {
> -               /*
> -                * Once frames start dropping, some data gets stuck in the DMA
> -                * pipeline somewhere. So the first DMA transfer after frame
> -                * drops gives a partial frame. This is obviously not useful to

I think it would be good to retain a similar comment in the code, and also
in the commit message, as there is a possibility of returning a partial frame
to the user post drain.

With that change,

Reviewed-by: Jai Luthra <jai.luthra@ideasonboard.com>

> -                * the application and will only confuse it. Issue a DMA
> -                * transaction to drain that up.
> -                */
> -               ret = ti_csi2rx_drain_dma(ctx);
> -               if (ret && ret != -ETIMEDOUT)
> -                       dev_warn(ctx->csi->dev,
> -                                "Failed to drain DMA. Next frame might be bogus\n");
> -
> -               spin_lock_irqsave(&dma->lock, flags);
> -               ret = ti_csi2rx_start_dma(ctx, buf);
> -               if (ret) {
> -                       vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_ERROR);
> -                       dma->state = TI_CSI2RX_DMA_IDLE;
> -                       spin_unlock_irqrestore(&dma->lock, flags);
> -                       dev_err(ctx->csi->dev, "Failed to start DMA: %d\n", ret);
> -               } else {
> -                       list_add_tail(&buf->list, &dma->submitted);
> -                       spin_unlock_irqrestore(&dma->lock, flags);
> -               }
> -       }
>  }
>  
>  static int ti_csi2rx_get_route(struct ti_csi2rx_ctx *ctx)
> -- 
> 2.34.1
>

[-- Attachment #2: signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

-----BEGIN PGP SIGNATURE-----
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=jt1I
-----END PGP SIGNATURE-----

Re: [PATCH v4 11/12] media: ti: j721e-csi2rx: Submit all available buffers

[Sjoerd Simons]

Hey,

On Wed, 2025-05-14 at 16:55 +0530, Rishikesh Donadkar wrote:
> From: Jai Luthra <j-luthra@ti.com>
> 
> We already make sure to submit all available buffers to DMA in each DMA
> completion callback.
> 
> Move that logic in a separate function, and use it during stream start
> as well, as most application queue all their buffers before stream on.
> 
> Signed-off-by: Jai Luthra <j-luthra@ti.com>
> Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
> Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
> ---
>  .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 43 +++++++++++--------
>  1 file changed, 24 insertions(+), 19 deletions(-)
> 
> diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> index 7986f96c5e11b..ba2a30bfed37d 100644
> --- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> +++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> @@ -651,6 +651,27 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
>  	return ret;
>  }
>  
> +static int ti_csi2rx_dma_submit_pending(struct ti_csi2rx_ctx *ctx)
> +{
> +	struct ti_csi2rx_dma *dma = &ctx->dma;
> +	struct ti_csi2rx_buffer *buf;
> +	int ret = 0;
> +
> +	/* If there are more buffers to process then start their transfer. */
> +	while (!list_empty(&dma->queue)) {
> +		buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer,
> list);
> +		ret = ti_csi2rx_start_dma(ctx, buf);
> +		if (ret) {
> +			dev_err(ctx->csi->dev,
> +				"Failed to queue the next buffer for DMA\n");
> +			vb2_buffer_done(&buf->vb.vb2_buf,
> VB2_BUF_STATE_ERROR);
> +			break;

The break here seems wrong and does change the previous logic; It means once *a*
buffer fails to start DMA, you'll no longer try to submit the other (queued)
buffers. If this was called from the DMA callback of the last submitted buffer
and userspace doesn't re-queue the error buffer, then capturing will stop, even
if there were still queued up buffers from a userspace pov.


For a potential next iteration you probably also want to wrap in the changes
from to fix list_del corruption:
https://lore.kernel.org/all/20250630-j721e-dma-fixup-v1-1-591e378ab3a8@collabora.com/



> +		}
> +		list_move_tail(&buf->list, &dma->submitted);
> +	}
> +	return ret;
> +}
> +
>  static void ti_csi2rx_dma_callback(void *param)
>  {
>  	struct ti_csi2rx_buffer *buf = param;
> @@ -671,18 +692,7 @@ static void ti_csi2rx_dma_callback(void *param)
>  	vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_DONE);
>  	list_del(&buf->list);
>  
> -	/* If there are more buffers to process then start their transfer. */
> -	while (!list_empty(&dma->queue)) {
> -		buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer,
> list);
> -
> -		if (ti_csi2rx_start_dma(ctx, buf)) {
> -			dev_err(ctx->csi->dev,
> -				"Failed to queue the next buffer for DMA\n");
> -			vb2_buffer_done(&buf->vb.vb2_buf,
> VB2_BUF_STATE_ERROR);
> -		} else {
> -			list_move_tail(&buf->list, &dma->submitted);
> -		}
> -	}
> +	ti_csi2rx_dma_submit_pending(ctx);
>  
>  	if (list_empty(&dma->submitted))
>  		dma->state = TI_CSI2RX_DMA_IDLE;
> @@ -941,7 +951,6 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq,
> unsigned int count)
>  	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vq);
>  	struct ti_csi2rx_dev *csi = ctx->csi;
>  	struct ti_csi2rx_dma *dma = &ctx->dma;
> -	struct ti_csi2rx_buffer *buf;
>  	unsigned long flags;
>  	int ret = 0;
>  
> @@ -980,16 +989,13 @@ static int ti_csi2rx_start_streaming(struct vb2_queue
> *vq, unsigned int count)
>  	ctx->sequence = 0;
>  
>  	spin_lock_irqsave(&dma->lock, flags);
> -	buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer, list);
>  
> -	ret = ti_csi2rx_start_dma(ctx, buf);
> +	ret = ti_csi2rx_dma_submit_pending(ctx);
>  	if (ret) {
> -		dev_err(csi->dev, "Failed to start DMA: %d\n", ret);
>  		spin_unlock_irqrestore(&dma->lock, flags);
> -		goto err_pipeline;
> +		goto err_dma;
>  	}
>  
> -	list_move_tail(&buf->list, &dma->submitted);
>  	dma->state = TI_CSI2RX_DMA_ACTIVE;
>  	spin_unlock_irqrestore(&dma->lock, flags);
>  
> @@ -1004,7 +1010,6 @@ static int ti_csi2rx_start_streaming(struct vb2_queue
> *vq, unsigned int count)
>  
>  err_dma:
>  	ti_csi2rx_stop_dma(ctx);
> -err_pipeline:
>  	video_device_pipeline_stop(&ctx->vdev);
>  	writel(0, csi->shim + SHIM_CNTL);
>  	writel(0, csi->shim + SHIM_DMACNTX(ctx->idx));

Re: [PATCH v4 12/12] media: ti: j721e-csi2rx: Change the drain architecture for multistream

[Rishikesh Donadkar]

On 25/06/25 07:03, Jai Luthra wrote:
> Hi Rishikesh,


Hi Jai,

Thank you for the review !


>
> Thanks for the patch.
>
> Quoting Rishikesh Donadkar (2025-05-14 04:25:27)
>> On buffer starvation the DMA is marked IDLE, and the stale data in the
>> internal FIFOs gets drained only on the next VIDIOC_QBUF call from the
>> userspace. This approach works fine for a single stream case.
>>
>> But in multistream scenarios, buffer starvation for one stream i.e. one
>> virtual channel, can block the shared HW FIFO of the CSI2RX IP. This can
>> stall the pipeline for all other virtual channels, even if buffers are
>> available for them.
>>
>> This patch introduces a new architecture, that continuously drains data
>> from the shared HW FIFO into a small (32KiB) buffer if no buffers are made
>> available to the driver from the userspace. This ensures independence
>> between different streams, where a slower downstream element for one
>> camera does not block streaming for other cameras.
>>
>> Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
>> ---
>>   .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 96 +++++++------------
>>   1 file changed, 33 insertions(+), 63 deletions(-)
>>
>> diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
>> index ba2a30bfed37d..3b046d3cf7e5a 100644
>> --- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
>> +++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
>> @@ -57,7 +57,6 @@
>>   #define TI_CSI2RX_MAX_SOURCE_PADS      TI_CSI2RX_MAX_CTX
>>   #define TI_CSI2RX_MAX_PADS             (1 + TI_CSI2RX_MAX_SOURCE_PADS)
>>   
>> -#define DRAIN_TIMEOUT_MS               50
>>   #define DRAIN_BUFFER_SIZE              SZ_32K
>>   
>>   struct ti_csi2rx_fmt {
>> @@ -77,7 +76,6 @@ struct ti_csi2rx_buffer {
>>   
>>   enum ti_csi2rx_dma_state {
>>          TI_CSI2RX_DMA_STOPPED,  /* Streaming not started yet. */
>> -       TI_CSI2RX_DMA_IDLE,     /* Streaming but no pending DMA operation. */
>>          TI_CSI2RX_DMA_ACTIVE,   /* Streaming and pending DMA operation. */
>>   };
>>   
>> @@ -245,6 +243,10 @@ static const struct ti_csi2rx_fmt ti_csi2rx_formats[] = {
>>   static int ti_csi2rx_start_dma(struct ti_csi2rx_ctx *ctx,
>>                                 struct ti_csi2rx_buffer *buf);
>>   
>> +/* Forward declarations needed by ti_csi2rx_drain_callback. */
>> +static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx);
>> +static int ti_csi2rx_dma_submit_pending(struct ti_csi2rx_ctx *ctx);
>> +
>>   static const struct ti_csi2rx_fmt *find_format_by_fourcc(u32 pixelformat)
>>   {
>>          unsigned int i;
>> @@ -596,9 +598,28 @@ static void ti_csi2rx_setup_shim(struct ti_csi2rx_ctx *ctx)
>>   
>>   static void ti_csi2rx_drain_callback(void *param)
>>   {
>> -       struct completion *drain_complete = param;
>> +       struct ti_csi2rx_ctx *ctx = param;
>> +       struct ti_csi2rx_dma *dma = &ctx->dma;
>> +       unsigned long flags;
>> +
>> +       spin_lock_irqsave(&dma->lock, flags);
>>   
>> -       complete(drain_complete);
>> +       if (dma->state == TI_CSI2RX_DMA_STOPPED) {
>> +               spin_unlock_irqrestore(&dma->lock, flags);
>> +               return;
>> +       }
>> +
>> +       /*
>> +        * If dma->queue is empty, it signals no buffer has arrived from
>> +        * user space, so, queue more transaction to drain dma
>> +        */
>> +       if (list_empty(&dma->queue)) {
>> +               if (ti_csi2rx_drain_dma(ctx))
>> +                       dev_warn(ctx->csi->dev, "DMA drain failed\n");
>> +       } else {
>> +               ti_csi2rx_dma_submit_pending(ctx);
>> +       }
>> +       spin_unlock_irqrestore(&dma->lock, flags);
>>   }
>>   
>>   /*
>> @@ -616,12 +637,9 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
>>   {
>>          struct ti_csi2rx_dev *csi = ctx->csi;
>>          struct dma_async_tx_descriptor *desc;
>> -       struct completion drain_complete;
>>          dma_cookie_t cookie;
>>          int ret;
>>   
>> -       init_completion(&drain_complete);
>> -
>>          desc = dmaengine_prep_slave_single(ctx->dma.chan, csi->drain.paddr,
>>                                             csi->drain.len, DMA_DEV_TO_MEM,
>>                                             DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
>> @@ -631,7 +649,7 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
>>          }
>>   
>>          desc->callback = ti_csi2rx_drain_callback;
>> -       desc->callback_param = &drain_complete;
>> +       desc->callback_param = ctx;
>>   
>>          cookie = dmaengine_submit(desc);
>>          ret = dma_submit_error(cookie);
>> @@ -640,13 +658,6 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
>>   
>>          dma_async_issue_pending(ctx->dma.chan);
>>   
>> -       if (!wait_for_completion_timeout(&drain_complete,
>> -                                        msecs_to_jiffies(DRAIN_TIMEOUT_MS))) {
>> -               dmaengine_terminate_sync(ctx->dma.chan);
>> -               dev_dbg(csi->dev, "DMA transfer timed out for drain buffer\n");
>> -               ret = -ETIMEDOUT;
>> -               goto out;
>> -       }
>>   out:
>>          return ret;
>>   }
>> @@ -694,9 +705,11 @@ static void ti_csi2rx_dma_callback(void *param)
>>   
>>          ti_csi2rx_dma_submit_pending(ctx);
>>   
>> -       if (list_empty(&dma->submitted))
>> -               dma->state = TI_CSI2RX_DMA_IDLE;
>> -
>> +       if (list_empty(&dma->submitted)) {
>> +               if (ti_csi2rx_drain_dma(ctx))
>> +                       dev_warn(ctx->csi->dev,
>> +                                "DMA drain failed on one of the transactions\n");
>> +       }
>>          spin_unlock_irqrestore(&dma->lock, flags);
>>   }
>>   
>> @@ -749,7 +762,7 @@ static void ti_csi2rx_stop_dma(struct ti_csi2rx_ctx *ctx)
>>                   * enforced before terminating DMA.
>>                   */
>>                  ret = ti_csi2rx_drain_dma(ctx);
>> -               if (ret && ret != -ETIMEDOUT)
>> +               if (ret)
>>                          dev_warn(ctx->csi->dev,
>>                                   "Failed to drain DMA. Next frame might be bogus\n");
>>          }
>> @@ -816,57 +829,14 @@ static void ti_csi2rx_buffer_queue(struct vb2_buffer *vb)
>>          struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
>>          struct ti_csi2rx_buffer *buf;
>>          struct ti_csi2rx_dma *dma = &ctx->dma;
>> -       bool restart_dma = false;
>>          unsigned long flags = 0;
>> -       int ret;
>>   
>>          buf = container_of(vb, struct ti_csi2rx_buffer, vb.vb2_buf);
>>          buf->ctx = ctx;
>>   
>>          spin_lock_irqsave(&dma->lock, flags);
>> -       /*
>> -        * Usually the DMA callback takes care of queueing the pending buffers.
>> -        * But if DMA has stalled due to lack of buffers, restart it now.
>> -        */
>> -       if (dma->state == TI_CSI2RX_DMA_IDLE) {
>> -               /*
>> -                * Do not restart DMA with the lock held because
>> -                * ti_csi2rx_drain_dma() might block for completion.
>> -                * There won't be a race on queueing DMA anyway since the
>> -                * callback is not being fired.
>> -                */
>> -               restart_dma = true;
>> -               dma->state = TI_CSI2RX_DMA_ACTIVE;
>> -       } else {
>> -               list_add_tail(&buf->list, &dma->queue);
>> -       }
>> +       list_add_tail(&buf->list, &dma->queue);
>>          spin_unlock_irqrestore(&dma->lock, flags);
>> -
>> -       if (restart_dma) {
>> -               /*
>> -                * Once frames start dropping, some data gets stuck in the DMA
>> -                * pipeline somewhere. So the first DMA transfer after frame
>> -                * drops gives a partial frame. This is obviously not useful to
> I think it would be good to retain a similar comment in the code, and also
> in the commit message, as there is a possibility of returning a partial frame
> to the user post drain.


Noted, Will add this info in the next revision of this series


Regards,

Rishikesh

>
> With that change,
>
> Reviewed-by: Jai Luthra <jai.luthra@ideasonboard.com>
>
>> -                * the application and will only confuse it. Issue a DMA
>> -                * transaction to drain that up.
>> -                */
>> -               ret = ti_csi2rx_drain_dma(ctx);
>> -               if (ret && ret != -ETIMEDOUT)
>> -                       dev_warn(ctx->csi->dev,
>> -                                "Failed to drain DMA. Next frame might be bogus\n");
>> -
>> -               spin_lock_irqsave(&dma->lock, flags);
>> -               ret = ti_csi2rx_start_dma(ctx, buf);
>> -               if (ret) {
>> -                       vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_ERROR);
>> -                       dma->state = TI_CSI2RX_DMA_IDLE;
>> -                       spin_unlock_irqrestore(&dma->lock, flags);
>> -                       dev_err(ctx->csi->dev, "Failed to start DMA: %d\n", ret);
>> -               } else {
>> -                       list_add_tail(&buf->list, &dma->submitted);
>> -                       spin_unlock_irqrestore(&dma->lock, flags);
>> -               }
>> -       }
>>   }
>>   
>>   static int ti_csi2rx_get_route(struct ti_csi2rx_ctx *ctx)
>> -- 
>> 2.34.1
> >

Re: [PATCH v4 11/12] media: ti: j721e-csi2rx: Submit all available buffers

[Rishikesh Donadkar]

On 01/07/25 13:39, Sjoerd Simons wrote:
> Hey,


Hi Sjoerd,

>
> On Wed, 2025-05-14 at 16:55 +0530, Rishikesh Donadkar wrote:
>> From: Jai Luthra <j-luthra@ti.com>
>>
>> We already make sure to submit all available buffers to DMA in each DMA
>> completion callback.
>>
>> Move that logic in a separate function, and use it during stream start
>> as well, as most application queue all their buffers before stream on.
>>
>> Signed-off-by: Jai Luthra <j-luthra@ti.com>
>> Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
>> Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
>> ---
>>   .../platform/ti/j721e-csi2rx/j721e-csi2rx.c   | 43 +++++++++++--------
>>   1 file changed, 24 insertions(+), 19 deletions(-)
>>
>> diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
>> b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
>> index 7986f96c5e11b..ba2a30bfed37d 100644
>> --- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
>> +++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
>> @@ -651,6 +651,27 @@ static int ti_csi2rx_drain_dma(struct ti_csi2rx_ctx *ctx)
>>   	return ret;
>>   }
>>   
>> +static int ti_csi2rx_dma_submit_pending(struct ti_csi2rx_ctx *ctx)
>> +{
>> +	struct ti_csi2rx_dma *dma = &ctx->dma;
>> +	struct ti_csi2rx_buffer *buf;
>> +	int ret = 0;
>> +
>> +	/* If there are more buffers to process then start their transfer. */
>> +	while (!list_empty(&dma->queue)) {
>> +		buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer,
>> list);
>> +		ret = ti_csi2rx_start_dma(ctx, buf);
>> +		if (ret) {
>> +			dev_err(ctx->csi->dev,
>> +				"Failed to queue the next buffer for DMA\n");
>> +			vb2_buffer_done(&buf->vb.vb2_buf,
>> VB2_BUF_STATE_ERROR);
>> +			break;
> The break here seems wrong and does change the previous logic; It means once *a*
> buffer fails to start DMA, you'll no longer try to submit the other (queued)
> buffers. If this was called from the DMA callback of the last submitted buffer
> and userspace doesn't re-queue the error buffer, then capturing will stop, even
> if there were still queued up buffers from a userspace pov.
>
>
> For a potential next iteration you probably also want to wrap in the changes
> from to fix list_del corruption:
> https://lore.kernel.org/all/20250630-j721e-dma-fixup-v1-1-591e378ab3a8@collabora.com/


Thank you for the pointer !

I will make these changes in the next revision


Regards,

Rishikesh

>
>
>
>> +		}
>> +		list_move_tail(&buf->list, &dma->submitted);
>> +	}
>> +	return ret;
>> +}
>> +
>>   static void ti_csi2rx_dma_callback(void *param)
>>   {
>>   	struct ti_csi2rx_buffer *buf = param;
>> @@ -671,18 +692,7 @@ static void ti_csi2rx_dma_callback(void *param)
>>   	vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_DONE);
>>   	list_del(&buf->list);
>>   
>> -	/* If there are more buffers to process then start their transfer. */
>> -	while (!list_empty(&dma->queue)) {
>> -		buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer,
>> list);
>> -
>> -		if (ti_csi2rx_start_dma(ctx, buf)) {
>> -			dev_err(ctx->csi->dev,
>> -				"Failed to queue the next buffer for DMA\n");
>> -			vb2_buffer_done(&buf->vb.vb2_buf,
>> VB2_BUF_STATE_ERROR);
>> -		} else {
>> -			list_move_tail(&buf->list, &dma->submitted);
>> -		}
>> -	}
>> +	ti_csi2rx_dma_submit_pending(ctx);
>>   
>>   	if (list_empty(&dma->submitted))
>>   		dma->state = TI_CSI2RX_DMA_IDLE;
>> @@ -941,7 +951,6 @@ static int ti_csi2rx_start_streaming(struct vb2_queue *vq,
>> unsigned int count)
>>   	struct ti_csi2rx_ctx *ctx = vb2_get_drv_priv(vq);
>>   	struct ti_csi2rx_dev *csi = ctx->csi;
>>   	struct ti_csi2rx_dma *dma = &ctx->dma;
>> -	struct ti_csi2rx_buffer *buf;
>>   	unsigned long flags;
>>   	int ret = 0;
>>   
>> @@ -980,16 +989,13 @@ static int ti_csi2rx_start_streaming(struct vb2_queue
>> *vq, unsigned int count)
>>   	ctx->sequence = 0;
>>   
>>   	spin_lock_irqsave(&dma->lock, flags);
>> -	buf = list_entry(dma->queue.next, struct ti_csi2rx_buffer, list);
>>   
>> -	ret = ti_csi2rx_start_dma(ctx, buf);
>> +	ret = ti_csi2rx_dma_submit_pending(ctx);
>>   	if (ret) {
>> -		dev_err(csi->dev, "Failed to start DMA: %d\n", ret);
>>   		spin_unlock_irqrestore(&dma->lock, flags);
>> -		goto err_pipeline;
>> +		goto err_dma;
>>   	}
>>   
>> -	list_move_tail(&buf->list, &dma->submitted);
>>   	dma->state = TI_CSI2RX_DMA_ACTIVE;
>>   	spin_unlock_irqrestore(&dma->lock, flags);
>>   
>> @@ -1004,7 +1010,6 @@ static int ti_csi2rx_start_streaming(struct vb2_queue
>> *vq, unsigned int count)
>>   
>>   err_dma:
>>   	ti_csi2rx_stop_dma(ctx);
>> -err_pipeline:
>>   	video_device_pipeline_stop(&ctx->vdev);
>>   	writel(0, csi->shim + SHIM_CNTL);
>>   	writel(0, csi->shim + SHIM_DMACNTX(ctx->idx));