Re: [PATCH 1/3] media: subdev: Drop implicit zeroing of stream field

From: Tomi Valkeinen
Date: Mon Jul 17 2023 - 10:39:30 EST


On 17/07/2023 17:11, Jacopo Mondi wrote:
Hi Tomi

On Mon, Jun 19, 2023 at 02:27:05PM +0300, Tomi Valkeinen wrote:
Now that the kernel drivers have been fixed to initialize the stream
field, and we have the client capability which the userspace uses to say

Not sure I got this. Isn't the capabilities flag intended for drivers
to tell userspace it support streams ? This seems to suggest it is
userspace setting it ?

Client capabilities tell the capabilities of the client. It's the new VIDIOC_SUBDEV_S_CLIENT_CAP/VIDIOC_SUBDEV_G_CLIENT_CAP ioctl.

it has initialized the stream field, we can drop the implicit zeroing of
the stream field in the various check functions.


I guess this is safe, but I'm not sure why it wasn't before. If a
driver doesn't support streams (!(sd->flags & V4L2_SUBDEV_FL_STREAMS))
then it should have ignored the 'stream' field even if it wasn't
zeroed. So I suspect I am missing the reason for zeroing in first
place...

The code being removed here was a quick fix. The issue was that before we had the client capability flag for "userspace supports streams", the 'stream' field could contain garbage. Also some kernel drivers were not properly initializing struct v4l2_subdev_format to zero, so again the 'stream' field could contain garbage.

The code removed here made sure that if a non-streams-supporting device was used, the 'stream' field would be zero as expected, and the v4l2 framework would not get confused by seeing a non-zero stream. The non-streams-enabled drivers themselves would not use the field anyway, of course, but the framework has code that expects the 'stream' to be zero (e.g. check_state() checks that stream == 0 if the device hasn't set V4L2_SUBDEV_FL_STREAMS).

Now the kernel drivers have been fixed to initialize the struct properly, and we have the VIDIOC_SUBDEV_S_CLIENT_CAP to handle the userspace part. Thus this code is no longer needed, and, I think, just might confused the reader.

And, in fact, I think it might hide an error. If a subdev is used that does not support streams, but the userspace supports streams. If the userspace uses an ioctl with stream != 0 for that subdev, it's clearly an error. However, with the code removed here, the error would go unnoticed as the kernel clears the stream field.

Tomi

Signed-off-by: Tomi Valkeinen <tomi.valkeinen@xxxxxxxxxxxxxxxx>
---
drivers/media/v4l2-core/v4l2-subdev.c | 15 ---------------
1 file changed, 15 deletions(-)

diff --git a/drivers/media/v4l2-core/v4l2-subdev.c b/drivers/media/v4l2-core/v4l2-subdev.c
index 2ec179cd1264..c1ac6d7a63d2 100644
--- a/drivers/media/v4l2-core/v4l2-subdev.c
+++ b/drivers/media/v4l2-core/v4l2-subdev.c
@@ -200,9 +200,6 @@ static inline int check_format(struct v4l2_subdev *sd,
if (!format)
return -EINVAL;

- if (!(sd->flags & V4L2_SUBDEV_FL_STREAMS))
- format->stream = 0;
-
return check_which(format->which) ? : check_pad(sd, format->pad) ? :
check_state(sd, state, format->which, format->pad, format->stream);
}
@@ -230,9 +227,6 @@ static int call_enum_mbus_code(struct v4l2_subdev *sd,
if (!code)
return -EINVAL;

- if (!(sd->flags & V4L2_SUBDEV_FL_STREAMS))
- code->stream = 0;
-
return check_which(code->which) ? : check_pad(sd, code->pad) ? :
check_state(sd, state, code->which, code->pad, code->stream) ? :
sd->ops->pad->enum_mbus_code(sd, state, code);
@@ -245,9 +239,6 @@ static int call_enum_frame_size(struct v4l2_subdev *sd,
if (!fse)
return -EINVAL;

- if (!(sd->flags & V4L2_SUBDEV_FL_STREAMS))
- fse->stream = 0;
-
return check_which(fse->which) ? : check_pad(sd, fse->pad) ? :
check_state(sd, state, fse->which, fse->pad, fse->stream) ? :
sd->ops->pad->enum_frame_size(sd, state, fse);
@@ -283,9 +274,6 @@ static int call_enum_frame_interval(struct v4l2_subdev *sd,
if (!fie)
return -EINVAL;

- if (!(sd->flags & V4L2_SUBDEV_FL_STREAMS))
- fie->stream = 0;
-
return check_which(fie->which) ? : check_pad(sd, fie->pad) ? :
check_state(sd, state, fie->which, fie->pad, fie->stream) ? :
sd->ops->pad->enum_frame_interval(sd, state, fie);
@@ -298,9 +286,6 @@ static inline int check_selection(struct v4l2_subdev *sd,
if (!sel)
return -EINVAL;

- if (!(sd->flags & V4L2_SUBDEV_FL_STREAMS))
- sel->stream = 0;
-
return check_which(sel->which) ? : check_pad(sd, sel->pad) ? :
check_state(sd, state, sel->which, sel->pad, sel->stream);
}
--
2.34.1