public inbox for linux-media@vger.kernel.org
 help / color / mirror / Atom feed
From: Sakari Ailus <sakari.ailus@linux.intel.com>
To: linux-media@vger.kernel.org
Cc: hans@jjverkuil.nl, laurent.pinchart@ideasonboard.com,
	Prabhakar <prabhakar.csengg@gmail.com>,
	"Kate Hsuan" <hpa@redhat.com>,
	"Dave Stevenson" <dave.stevenson@raspberrypi.com>,
	"Tommaso Merciai" <tomm.merciai@gmail.com>,
	"Benjamin Mugnier" <benjamin.mugnier@foss.st.com>,
	"Sylvain Petinot" <sylvain.petinot@foss.st.com>,
	"Christophe JAILLET" <christophe.jaillet@wanadoo.fr>,
	"Julien Massot" <julien.massot@collabora.com>,
	"Naushir Patuck" <naush@raspberrypi.com>,
	"Yan, Dongcheng" <dongcheng.yan@intel.com>,
	"Cao, Bingbu" <bingbu.cao@intel.com>,
	"Qiu, Tian Shu" <tian.shu.qiu@intel.com>,
	"Stefan Klug" <stefan.klug@ideasonboard.com>,
	"Mirela Rabulea" <mirela.rabulea@nxp.com>,
	"André Apitzsch" <git@apitzsch.eu>,
	"Heimir Thor Sverrisson" <heimir.sverrisson@gmail.com>,
	"Kieran Bingham" <kieran.bingham@ideasonboard.com>,
	"Mehdi Djait" <mehdi.djait@linux.intel.com>,
	"Ricardo Ribalda Delgado" <ribalda@kernel.org>,
	"Hans de Goede" <hansg@kernel.org>,
	"Jacopo Mondi" <jacopo.mondi@ideasonboard.com>,
	"Tomi Valkeinen" <tomi.valkeinen@ideasonboard.com>,
	"David Plowman" <david.plowman@raspberrypi.com>,
	"Yu, Ong Hock" <ong.hock.yu@intel.com>,
	"Ng, Khai Wen" <khai.wen.ng@intel.com>,
	"Jai Luthra" <jai.luthra@ideasonboard.com>
Subject: [PATCH v3 16/22] media: mc: Add MEDIA_LNK_FL_VALIDATE_LATE
Date: Wed, 25 Mar 2026 12:58:13 +0200	[thread overview]
Message-ID: <20260325105818.1176816-18-sakari.ailus@linux.intel.com> (raw)
In-Reply-To: <20260325105818.1176816-1-sakari.ailus@linux.intel.com>

Add MEDIA_LNK_FL_VALIDATE_LATE flag to support late validation of links.
This is serving the use case where video devices are configured and
started streaming indepenently of each other but this sequence may be run
in series, in such a way that a video device in a pipeline starts
streaming before another one is configured.

Before this flag, drivers have resorted to implementing the link
validation separately for the video nodes as part of streaming start
sequence.

media_pipeline_start() shall be called on each leaf entity connected to
the graph with a link where MEDIA_LNK_FL_VALIDATE_LATE is set before
uphardware operation.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
---
 .../media/mediactl/media-ioc-setup-link.rst   |  4 +
 .../media/mediactl/media-types.rst            |  5 ++
 drivers/media/mc/mc-entity.c                  | 82 ++++++++++++++++++-
 include/uapi/linux/media.h                    |  1 +
 4 files changed, 88 insertions(+), 4 deletions(-)

diff --git a/Documentation/userspace-api/media/mediactl/media-ioc-setup-link.rst b/Documentation/userspace-api/media/mediactl/media-ioc-setup-link.rst
index 23208300cb61..7c2bced57e77 100644
--- a/Documentation/userspace-api/media/mediactl/media-ioc-setup-link.rst
+++ b/Documentation/userspace-api/media/mediactl/media-ioc-setup-link.rst
@@ -49,6 +49,10 @@ Only links marked with the ``DYNAMIC`` link flag can be enabled/disabled
 while streaming media data. Attempting to enable or disable a streaming
 non-dynamic link will return an ``EBUSY`` error code.
 
+Pipeline validation may be delayed for links marked with the ``VALIDATE_LATE``
+flag until actual hardware operation even if the rest of the pipeline would be
+validated at an earlier point of time.
+
 If the specified link can't be found the driver returns with an ``EINVAL``
 error code.
 
diff --git a/Documentation/userspace-api/media/mediactl/media-types.rst b/Documentation/userspace-api/media/mediactl/media-types.rst
index 6332e8395263..d6a690655a01 100644
--- a/Documentation/userspace-api/media/mediactl/media-types.rst
+++ b/Documentation/userspace-api/media/mediactl/media-types.rst
@@ -391,6 +391,7 @@ must be set for every pad.
 .. _MEDIA-LNK-FL-ENABLED:
 .. _MEDIA-LNK-FL-IMMUTABLE:
 .. _MEDIA-LNK-FL-DYNAMIC:
+.. _MEDIA-LNK-FL-VALIDATE-LATE:
 .. _MEDIA-LNK-FL-LINK-TYPE:
 
 .. flat-table:: Media link flags
@@ -410,6 +411,10 @@ must be set for every pad.
        -  The link enabled state can be modified during streaming. This flag
 	  is set by drivers and is read-only for applications.
 
+    *  -  ``MEDIA_LNK_FL_VALIDATE_LATE``
+       -  The validation of the link may be delayed up to until the start of
+	  hardware operation.
+
     *  -  ``MEDIA_LNK_FL_LINK_TYPE``
        -  This is a bitmask that defines the type of the link. The following
 	  link types are currently supported:
diff --git a/drivers/media/mc/mc-entity.c b/drivers/media/mc/mc-entity.c
index 287eded356bb..bb639ce4d86e 100644
--- a/drivers/media/mc/mc-entity.c
+++ b/drivers/media/mc/mc-entity.c
@@ -771,7 +771,7 @@ static int media_pipeline_populate(struct media_pipeline *pipe,
 static int
 __media_pipeline_validate_one(struct media_pad *origin,
 			      struct media_pad *pad, struct media_link *link,
-			      bool *has_enabled_link)
+			      bool *has_enabled_link, bool skip_validation)
 {
 	struct media_device *mdev = origin->graph_obj.mdev;
 	struct media_entity *entity = pad->entity;
@@ -784,6 +784,9 @@ __media_pipeline_validate_one(struct media_pad *origin,
 	if (has_enabled_link)
 		*has_enabled_link = true;
 
+	if (skip_validation)
+		return 0;
+
 	/* Skip validation if the current pad isn't the sink pad of the link. */
 	if (link->sink != pad)
 		return 0;
@@ -825,11 +828,48 @@ __must_check int __media_pipeline_start(struct media_pad *origin,
 		return -EINVAL;
 
 	/*
-	 * If the pipeline has already been started, it is guaranteed to be
-	 * valid, so just increase the start count.
+	 * Increase start count on pipelines that have been validated
+	 * earlier. Also check links with the VALIDATE_LATE flag here.
 	 */
 	if (pipe->start_count) {
+		struct media_link *link;
+
+		link = __media_entity_next_link(origin->entity, NULL,
+						MEDIA_LNK_FL_DATA_LINK);
+		if (link && link->flags & MEDIA_LNK_FL_VALIDATE_LATE) {
+			struct media_link *link2 =
+				__media_entity_next_link(origin->entity, link,
+							 MEDIA_LNK_FL_DATA_LINK);
+			bool has_enabled_link = false;
+
+			/*
+			 * Only a single pad is allowed for VALIDATE_LATE
+			 * links. That pad needs to have exactly one link.
+			 */
+			if (origin->entity->num_pads != 1)
+				return -EINVAL;
+
+			if (!link || link2)
+				return -EINVAL;
+
+			dev_dbg(mdev->dev,
+				"Validating pad '%s':%u late\n",
+				origin->entity->name, origin->index);
+
+			ret = __media_pipeline_validate_one(link->sink,
+							    link->sink, link,
+							    &has_enabled_link,
+							    false);
+			if (ret)
+				return ret;
+
+			if (origin->flags & MEDIA_PAD_FL_MUST_CONNECT &&
+			    !has_enabled_link)
+				return -ENOLINK;
+		}
+
 		pipe->start_count++;
+
 		return 0;
 	}
 
@@ -873,12 +913,19 @@ __must_check int __media_pipeline_start(struct media_pad *origin,
 		 * the connected sink pad to avoid duplicating checks.
 		 */
 		for_each_media_entity_data_link(entity, link) {
+			/* Skip late-validated links not connected to origin. */
+			bool skip_validation =
+				link->flags & MEDIA_LNK_FL_VALIDATE_LATE &&
+				(link->sink == origin ||
+				 link->source == origin);
+
 			/* Skip links unrelated to the current pad. */
 			if (link->sink != pad && link->source != pad)
 				continue;
 
 			ret = __media_pipeline_validate_one(origin, pad, link,
-							    &has_enabled_link);
+							    &has_enabled_link,
+							    skip_validation);
 			if (ret)
 				goto error;
 		}
@@ -1158,6 +1205,33 @@ media_create_pad_link(struct media_entity *source, u16 source_pad,
 	if (WARN_ON(!(sink->pads[sink_pad].flags & MEDIA_PAD_FL_SINK)))
 		return -EINVAL;
 
+	/*
+	 * With the late validate flag, either source or sink shall have exactly
+	 * one pad and no links before this one. Similarly, no links may be
+	 * added to entities with a single pad and an existing late-validated
+	 * link.
+	 */
+	if (flags & MEDIA_LNK_FL_VALIDATE_LATE) {
+		if (!(source->num_pads == 1 && !source->num_links) &&
+		    !(sink->num_pads == 1 && !sink->num_links))
+			return -EINVAL;
+	} else {
+		struct media_entity *entities[] = { source, sink };
+
+		for (unsigned int i = 0; i < ARRAY_SIZE(entities); i++) {
+			if (entities[i]->num_pads != 1)
+				continue;
+
+			struct media_link *__link =
+				__media_entity_next_link(entities[i], NULL,
+							 MEDIA_LNK_FL_DATA_LINK);
+
+			if (__link &&
+			    __link->flags & MEDIA_LNK_FL_VALIDATE_LATE)
+				return -EINVAL;
+		}
+	}
+
 	link = media_add_link(&source->links);
 	if (link == NULL)
 		return -ENOMEM;
diff --git a/include/uapi/linux/media.h b/include/uapi/linux/media.h
index 1c80b1d6bbaf..c96e2118ea99 100644
--- a/include/uapi/linux/media.h
+++ b/include/uapi/linux/media.h
@@ -219,6 +219,7 @@ struct media_pad_desc {
 #define MEDIA_LNK_FL_ENABLED			(1U << 0)
 #define MEDIA_LNK_FL_IMMUTABLE			(1U << 1)
 #define MEDIA_LNK_FL_DYNAMIC			(1U << 2)
+#define MEDIA_LNK_FL_VALIDATE_LATE		(1U << 3)
 
 #define MEDIA_LNK_FL_LINK_TYPE			(0xf << 28)
 #  define MEDIA_LNK_FL_DATA_LINK		(0U << 28)
-- 
2.47.3


  parent reply	other threads:[~2026-03-25 10:58 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-25 10:57 [PATCH v3 00/22] Metadata series preparation Sakari Ailus
2026-03-25 10:57 ` [PATCH v3 01/22] media: imx219: Rename "PIXEL_ARRAY" as "VISIBLE" Sakari Ailus
2026-03-25 10:57 ` [PATCH v3 02/22] media: imx219: Fix maximum frame length in lines Sakari Ailus
2026-03-25 11:36   ` Dave Stevenson
2026-03-25 10:57 ` [PATCH v3 03/22] media: imx219: Set horizontal blanking on mode change Sakari Ailus
2026-03-25 11:48   ` Dave Stevenson
2026-03-25 12:20     ` Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 04/22] media: imx219: Scale the vblank limits according to rate_factor Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 05/22] media: imx219: Fix vertical blanking and exposure for analogue binning Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 06/22] media: imx219: Don't update exposure limits while setting format Sakari Ailus
2026-03-25 12:37   ` Dave Stevenson
2026-03-25 12:47     ` Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 07/22] media: imx219: Rename "binning" as "bin_hv" in imx219_set_pad_format Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 08/22] media: Documentation: Improve LINK_FREQ documentation Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 09/22] media: Documentation: Improve pixel rate calculation documentation Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 10/22] media: v4l2-subdev: Refactor returning routes Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 11/22] media: v4l2-subdev: Allow accessing routes with STREAMS client capability Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 12/22] media: mc: Simplify link processing in __media_pipeline_start() Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 13/22] media: mc: Separate single link validation into a new function Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 14/14] media: Improve enable_streams and disable_streams documentation Sakari Ailus
2026-03-25 12:34   ` Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 14/22] media: uapi: Bump the STREAMS bit a little Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 15/22] media: mc: Don't care about unsettable flags in MEDIA_IOC_LINK_SETUP Sakari Ailus
2026-03-25 10:58 ` Sakari Ailus [this message]
2026-03-25 10:58 ` [PATCH v3 17/22] media: Improve enable_streams and disable_streams documentation Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 18/22] media: v4l2-subdev: Move subdev client capabilities into a new struct Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 19/22] media: v4l2-subdev: Add struct v4l2_subdev_client_info pointer to pad ops Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 20/22] media: v4l2-subdev: Add v4l2_subdev_call_ci_active_state Sakari Ailus
2026-03-25 10:58 ` [PATCH v3 21/22] media: v4l2-subdev: Perform client info changes to i2c drivers Sakari Ailus

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260325105818.1176816-18-sakari.ailus@linux.intel.com \
    --to=sakari.ailus@linux.intel.com \
    --cc=benjamin.mugnier@foss.st.com \
    --cc=bingbu.cao@intel.com \
    --cc=christophe.jaillet@wanadoo.fr \
    --cc=dave.stevenson@raspberrypi.com \
    --cc=david.plowman@raspberrypi.com \
    --cc=dongcheng.yan@intel.com \
    --cc=git@apitzsch.eu \
    --cc=hans@jjverkuil.nl \
    --cc=hansg@kernel.org \
    --cc=heimir.sverrisson@gmail.com \
    --cc=hpa@redhat.com \
    --cc=jacopo.mondi@ideasonboard.com \
    --cc=jai.luthra@ideasonboard.com \
    --cc=julien.massot@collabora.com \
    --cc=khai.wen.ng@intel.com \
    --cc=kieran.bingham@ideasonboard.com \
    --cc=laurent.pinchart@ideasonboard.com \
    --cc=linux-media@vger.kernel.org \
    --cc=mehdi.djait@linux.intel.com \
    --cc=mirela.rabulea@nxp.com \
    --cc=naush@raspberrypi.com \
    --cc=ong.hock.yu@intel.com \
    --cc=prabhakar.csengg@gmail.com \
    --cc=ribalda@kernel.org \
    --cc=stefan.klug@ideasonboard.com \
    --cc=sylvain.petinot@foss.st.com \
    --cc=tian.shu.qiu@intel.com \
    --cc=tomi.valkeinen@ideasonboard.com \
    --cc=tomm.merciai@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox