From: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
To: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Cc: Sakari Ailus <sakari.ailus@linux.intel.com>,
linux-media@vger.kernel.org, laurent.pinchart@ideasonboard.com,
Dave Stevenson <dave.stevenson@raspberrypi.com>,
Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>,
Jai Luthra <jai.luthra@ideasonboard.com>,
Mehdi Djait <mehdi.djait@linux.intel.com>
Subject: Re: [PATCH 04/17] media: v4l2-subdev: Allocate frame descriptors based on the need
Date: Thu, 14 May 2026 12:25:22 +0200 [thread overview]
Message-ID: <agWi0wgNkfeZCV0t@zed> (raw)
In-Reply-To: <agWfFzXX9a5tcd1i@zed>
Hi Sakari
On Thu, May 14, 2026 at 12:16:51PM +0200, Jacopo Mondi wrote:
> Hi Sakari
>
> On Wed, May 13, 2026 at 01:43:45PM +0300, Sakari Ailus wrote:
> > Frame descriptors entries require a small amount of memory per entry (20
> > bytes), but if the number of entries in a frame descriptor is large, an
> > unreasonably large amount of memory would need to be allocated in the
> > stack. Therefore the number of entries has been limited to 8.
> >
> > Support larger frame descriptors by allocating as much memory as required.
> > The get_frame_desc() op can now set the num_entries to a number larger
> > than V4L2_FRAME_BUS_ENTRY_MAX and return -ENOSPC. The caller,
> > v4l2_subdev_get_frame_desc(), will then allocate memory for that amount of
> > memory and call the get_frame_desc() op again.
> >
> > The caller is also responsible for releasing the allocated memory by
> > calling v4l2_subdev_free_frame_desc().
> >
> > Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
> > ---
> > drivers/media/v4l2-core/v4l2-subdev.c | 49 +++++++++++++++++++++------
> > include/media/v4l2-subdev.h | 44 +++++++++++++++++++-----
> > 2 files changed, 74 insertions(+), 19 deletions(-)
> >
> > diff --git a/drivers/media/v4l2-core/v4l2-subdev.c b/drivers/media/v4l2-core/v4l2-subdev.c
> > index b8acce8f9c33..012ff6fb2988 100644
> > --- a/drivers/media/v4l2-core/v4l2-subdev.c
> > +++ b/drivers/media/v4l2-core/v4l2-subdev.c
> > @@ -63,10 +63,6 @@ static bool v4l2_subdev_enable_streams_api;
> > /*
> > * Maximum stream ID is 63 for now, as we use u64 bitmask to represent a set
> > * of streams.
> > - *
> > - * Note that V4L2_FRAME_DESC_ENTRY_MAX is related: V4L2_FRAME_DESC_ENTRY_MAX
> > - * restricts the total number of streams in a pad, although the stream ID is
> > - * not restricted.
> > */
> > #define V4L2_SUBDEV_MAX_STREAM_ID 63
> >
> > @@ -354,6 +350,7 @@ static int call_set_frame_interval(struct v4l2_subdev *sd,
> > static int call_get_frame_desc(struct v4l2_subdev *sd, unsigned int pad,
> > struct v4l2_mbus_frame_desc *fd)
> > {
> > + unsigned int type;
> > unsigned int i;
> > int ret;
> >
> > @@ -362,16 +359,38 @@ static int call_get_frame_desc(struct v4l2_subdev *sd, unsigned int pad,
> > return -EOPNOTSUPP;
> > #endif
> >
> > + type = fd->type;
> > memset(fd, 0, sizeof(*fd));
> > + if (!fd->entry) {
> > + fd->entry = fd->entry_mem;
> > + fd->len_entries = ARRAY_SIZE(fd->entry_mem);
> > + }
> >
> > ret = sd->ops->pad->get_frame_desc(sd, pad, fd);
> > + if (ret == -ENOSPC) {
> > + if (fd->num_entries > V4L2_FRAME_DESC_ENTRY_PREALLOC &&
> > + fd->num_entries < V4L2_FRAME_DESC_ENTRY_MAX) {
>
> nit: you can invert the condition and save one indentation level
>
> if (fd->num_entries <= V4L2_FRAME_DESC_ENTRY_PREALLOC ||
> fd->num_entries >= V4L2_FRAME_DESC_ENTRY_MAX)
> return -E2BIG;
>
> > + fd->entry = kzalloc_objs(*fd->entry, fd->num_entries,
> > + GFP_KERNEL);
> > + if (!fd->entry)
> > + return -ENOMEM;
> > +
> > + fd->len_entries = fd->num_entries;
> > + fd->num_entries = 0;
> > +
> > + ret = sd->ops->pad->get_frame_desc(sd, pad, fd);
> > + } else {
> > + return -E2BIG;
> > + }
> > + }
> > if (ret)
> > return ret;
> >
> > dev_dbg(sd->dev, "Frame descriptor on pad %u, type %s\n", pad,
> > - fd->type == V4L2_MBUS_FRAME_DESC_TYPE_PARALLEL ? "parallel" :
> > - fd->type == V4L2_MBUS_FRAME_DESC_TYPE_CSI2 ? "CSI-2" :
> > - "unknown");
> > + type == V4L2_MBUS_FRAME_DESC_TYPE_PARALLEL ? "parallel" :
> > + type == V4L2_MBUS_FRAME_DESC_TYPE_CSI2 ? "CSI-2" : "unknown");
> > +
> > + fd->type = type;
>
> If we end up calling d->ops->pad->get_frame_desc() don't we overwrite
> the returned type ?
>
> >
> > for (i = 0; i < fd->num_entries; i++) {
> > struct v4l2_mbus_frame_desc_entry *entry = &fd->entry[i];
> > @@ -1086,9 +1105,9 @@ static long subdev_do_ioctl(struct file *file, unsigned int cmd, void *arg,
> > * descriptor accordingly, with up to one entry per route. Until
> > * the frame descriptors entries get allocated dynamically,
> > * limit the number of active routes to
> > - * V4L2_FRAME_DESC_ENTRY_MAX.
> > + * V4L2_FRAME_DESC_ENTRY_PREALLOC.
> > */
> > - if (num_active_routes > V4L2_FRAME_DESC_ENTRY_MAX)
> > + if (num_active_routes > V4L2_FRAME_DESC_ENTRY_PREALLOC)
> > return -E2BIG;
> >
> > /*
> > @@ -2638,7 +2657,7 @@ int __v4l2_subdev_get_frame_desc_passthrough(struct v4l2_subdev *sd,
> > return -EPIPE;
> > }
> >
> > - if (fd->num_entries >= V4L2_FRAME_DESC_ENTRY_MAX) {
> > + if (fd->num_entries >= V4L2_FRAME_DESC_ENTRY_PREALLOC) {
> > dev_dbg(dev, "Frame desc entry limit reached\n");
> > return -E2BIG;
> > }
> > @@ -2730,6 +2749,16 @@ int v4l2_subdev_get_frame_desc(struct v4l2_subdev *sd, unsigned int pad,
> > }
> > EXPORT_SYMBOL_GPL(v4l2_subdev_get_frame_desc);
> >
> > +void v4l2_subdev_free_frame_desc(struct v4l2_mbus_frame_desc *desc)
> > +{
>
> As this might be used as a __cleanup() should we check if desc is
> valid ?
>
> > + if (desc->entry != desc->entry_mem)
> > + kfree(desc->entry);
Sorry, in my below comment ....
> > +
> > + desc->entry = NULL;
> > + desc->len_entries = desc->num_entries = 0;
> > +}
> > +EXPORT_SYMBOL_GPL(v4l2_subdev_free_frame_desc);
> > +
> > #endif /* CONFIG_VIDEO_V4L2_SUBDEV_API */
> >
> > #endif /* CONFIG_MEDIA_CONTROLLER */
> > diff --git a/include/media/v4l2-subdev.h b/include/media/v4l2-subdev.h
> > index c9e74566c85a..bd97510c7024 100644
> > --- a/include/media/v4l2-subdev.h
> > +++ b/include/media/v4l2-subdev.h
> > @@ -365,11 +365,13 @@ struct v4l2_mbus_frame_desc_entry {
> > } bus;
> > };
> >
> > - /*
> > - * If this number is too small, it should be dropped altogether and the
> > - * API switched to a dynamic number of frame descriptor entries.
> > - */
> > -#define V4L2_FRAME_DESC_ENTRY_MAX 8
> > +/* Size of the statically allocated frame descriptor array. */
> > +#define V4L2_FRAME_DESC_ENTRY_PREALLOC 8
> > +/*
> > + * Maximum number of dynamically allocated frame descriptors. Note that
> > + * V4L2_SUBDEV_MAX_STREAM_ID is related to this limit as well.
> > + */
> > +#define V4L2_FRAME_DESC_ENTRY_MAX 64
> >
> > /**
> > * enum v4l2_mbus_frame_desc_type - media bus frame description type
> > @@ -392,13 +394,17 @@ enum v4l2_mbus_frame_desc_type {
> > /**
> > * struct v4l2_mbus_frame_desc - media bus data frame description
> > * @type: type of the bus (enum v4l2_mbus_frame_desc_type)
> > - * @entry: frame descriptors array
> > - * @num_entries: number of entries in @entry array
> > + * @entry_mem: memory for the frame descriptors (@entry)
> > + * @entry: pointer to the frame descriptors
> > + * @num_entries: number of entries in @entry
> > + * @len_entries: number of entries allocated for @entry
> > */
> > struct v4l2_mbus_frame_desc {
> > enum v4l2_mbus_frame_desc_type type;
> > - struct v4l2_mbus_frame_desc_entry entry[V4L2_FRAME_DESC_ENTRY_MAX];
> > + struct v4l2_mbus_frame_desc_entry entry_mem[V4L2_FRAME_DESC_ENTRY_PREALLOC];
> > + struct v4l2_mbus_frame_desc_entry *entry;
> > unsigned short num_entries;
> > + unsigned short len_entries;
> > };
> >
> > /**
> > @@ -780,7 +786,13 @@ struct v4l2_subdev_state {
> > * @link_validate: used by the media controller code to check if the links
> > * that belongs to a pipeline can be used for stream.
> > *
> > - * @get_frame_desc: get the current low level media bus frame parameters.
> > + * @get_frame_desc: get the current low level media bus frame parameters. The
> > + * callback is required to update the num_entries field to the
> > + * total number of entries in the frame descriptor. The
> > + * callback shall fill the first entries array up to
> > + * len_entries, which signifies the number of entries
> > + * allocated. If num_entries exceeds len_entries, the callback
> > + * shall return -ENOSPC.
> > *
> > * @set_frame_desc: set the low level media bus frame parameters, @fd array
> > * may be adjusted by the subdev driver to device capabilities.
> > @@ -1793,11 +1805,25 @@ int v4l2_subdev_get_frame_desc_passthrough(struct v4l2_subdev *sd,
> > *
> > * The caller is required to set @desc->type to the expected bus type.
> > *
> > + * The entries in the frame descriptor are allocated based on the need. The
> > + * caller is required to release the memory of the frame descriptor entries for
>
> Is it unconditionally true ? If fd->entry is not initialized aren't we
> deflecting to the pre-allocated entries in call_get_frame_desc() ?
>
> if (!fd->entry) {
> fd->entry = fd->entry_mem;
> fd->len_entries = ARRAY_SIZE(fd->entry_mem);
> }
>
> v4l2_subdev_free_frame_desc() shouldn't be called in this case, right ?
... here I missed this, so it's safe to call
v4l2_subdev_free_frame_desc() unconditionally
>
>
> > + * each frame descriptor obtained by calling this function using
> > + * v4l2_subdev_free_frame_desc().
> > + *
> > * Return: %0 on success or negative error code on failure.
> > */
> > int v4l2_subdev_get_frame_desc(struct v4l2_subdev *sd, unsigned int pad,
> > struct v4l2_mbus_frame_desc *desc);
> >
> > +/**
> > + * v4l2_subdev_free_frame_desc() - Release the memory of a frame descriptor
> > + * @desc: A pointer to a frame descriptor
> > + *
> > + * Release the frame descriptor entries in a frame descriptor. The number of
> > + * entries in the descriptor are set to 0 again.
> > + */
> > +void v4l2_subdev_free_frame_desc(struct v4l2_mbus_frame_desc *desc);
> > +
> > #endif /* CONFIG_VIDEO_V4L2_SUBDEV_API */
> >
> > #endif /* CONFIG_MEDIA_CONTROLLER */
> > --
> > 2.47.3
> >
> >
next prev parent reply other threads:[~2026-05-14 10:25 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-05-13 10:43 [PATCH 00/17] Rework frame descriptors Sakari Ailus
2026-05-13 10:43 ` [PATCH 01/17] media: v4l2-common: Add mipi_csi2_dt_for_mbus() Sakari Ailus
2026-05-13 20:46 ` Frank Li
2026-05-13 20:54 ` Sakari Ailus
2026-05-13 21:03 ` Frank Li
2026-05-13 10:43 ` [PATCH 02/17] media: v4l2-subdev: Align frame descriptor error codes with routing Sakari Ailus
2026-05-13 21:10 ` Frank Li
2026-05-13 10:43 ` [PATCH 03/17] media: v4l2-subdev: Prepare for changes in getting frame descriptors Sakari Ailus
2026-05-13 21:32 ` Frank Li
2026-05-13 22:03 ` Frank Li
2026-05-14 10:02 ` Jacopo Mondi
2026-05-13 10:43 ` [PATCH 04/17] media: v4l2-subdev: Allocate frame descriptors based on the need Sakari Ailus
2026-05-14 10:16 ` Jacopo Mondi
2026-05-14 10:25 ` Jacopo Mondi [this message]
2026-05-14 10:42 ` Jacopo Mondi
2026-05-13 10:43 ` [PATCH 05/17] media: v4l2-subdev: Provide a cleanup-friendly get_frame_desc Sakari Ailus
2026-05-13 10:43 ` [PATCH 06/17] media: v4l2-subdev: Change the maximum number of routes Sakari Ailus
2026-05-13 10:43 ` [PATCH 07/17] media: v4l2-subdev: Return dynamically allocated pass-through routes Sakari Ailus
2026-05-13 10:43 ` [PATCH 08/17] media: v4l2-subdev: Always return at least one frame descriptor Sakari Ailus
2026-05-13 10:43 ` [PATCH 09/17] media: bcm2835-unicam: Use v4l2_subdev_get_frame_desc() Sakari Ailus
2026-05-13 10:43 ` [PATCH 10/17] media: nxp: imx8-isi: " Sakari Ailus
2026-05-14 3:58 ` Frank Li
2026-05-13 10:43 ` [PATCH 11/17] media: raspberrypi: cfe: " Sakari Ailus
2026-05-13 10:43 ` [PATCH 12/17] media: rzg2l-cru: " Sakari Ailus
2026-05-13 10:43 ` [PATCH 13/17] media: rkisp1: " Sakari Ailus
2026-05-13 10:43 ` [PATCH 14/17] media: exynos4-is: " Sakari Ailus
2026-05-13 10:43 ` [PATCH 15/17] media: ti: cal: " Sakari Ailus
2026-05-13 10:43 ` [PATCH 16/17] media: ipu6: " Sakari Ailus
2026-05-13 10:43 ` [PATCH 17/17] staging: media: ipu7: " Sakari Ailus
2026-05-13 20:14 ` [PATCH 00/17] Rework frame descriptors 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=agWi0wgNkfeZCV0t@zed \
--to=jacopo.mondi@ideasonboard.com \
--cc=dave.stevenson@raspberrypi.com \
--cc=jai.luthra@ideasonboard.com \
--cc=laurent.pinchart@ideasonboard.com \
--cc=linux-media@vger.kernel.org \
--cc=mehdi.djait@linux.intel.com \
--cc=sakari.ailus@linux.intel.com \
--cc=tomi.valkeinen@ideasonboard.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