[media] v4l2-dev: make it possible to skip locking for selected ioctls

Using the V4L2 core lock is a very robust method that is usually very good
at doing the right thing. But some drivers, particularly USB drivers, may
want to prevent the core from taking the lock for specific ioctls, particularly
buffer queuing ioctls.

The reason is that certain commands like S_CTRL can take a long time to process
over USB and all the time the core has the lock, preventing VIDIOC_DQBUF from
proceeding, even though a frame may be ready in the queue.

This introduces unwanted latency.

Since the buffer queuing commands often have their own internal lock it is
often not necessary to take the core lock. Drivers can now say that they don't
want the core to take the lock for specific ioctls.

As it is a specific opt-out it makes it clear to the reviewer that those
ioctls will need more care when reviewing.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>

1 files changed, 24 insertions, 3 deletions

diff --git a/Documentation/video4linux/v4l2-framework.txt b/Documentation/video4linux/v4l2-framework.txt
index fe53177f0d3c..e1e6a01d7ac6 100644
--- a/Documentation/video4linux/v4l2-framework.txt
+++ b/Documentation/video4linux/v4l2-framework.txt
@@ -580,19 +580,25 @@ allocated memory.
 You should also set these fields:
 
 - v4l2_dev: set to the v4l2_device parent device.
+
 - name: set to something descriptive and unique.
+
 - fops: set to the v4l2_file_operations struct.
+
 - ioctl_ops: if you use the v4l2_ioctl_ops to simplify ioctl maintenance
   (highly recommended to use this and it might become compulsory in the
   future!), then set this to your v4l2_ioctl_ops struct.
+
 - lock: leave to NULL if you want to do all the locking in the driver.
   Otherwise you give it a pointer to a struct mutex_lock and before any
   of the v4l2_file_operations is called this lock will be taken by the
-  core and released afterwards.
+  core and released afterwards. See the next section for more details.
+
 - prio: keeps track of the priorities. Used to implement VIDIOC_G/S_PRIORITY.
   If left to NULL, then it will use the struct v4l2_prio_state in v4l2_device.
   If you want to have a separate priority state per (group of) device node(s),
   then you can point it to your own struct v4l2_prio_state.
+
 - parent: you only set this if v4l2_device was registered with NULL as
   the parent device struct. This only happens in cases where one hardware
   device has multiple PCI devices that all share the same v4l2_device core.
@@ -602,6 +608,7 @@ You should also set these fields:
   (cx8802). Since the v4l2_device cannot be associated with a particular
   PCI device it is setup without a parent device. But when the struct
   video_device is setup you do know which parent PCI device to use.
+
 - flags: optional. Set to V4L2_FL_USE_FH_PRIO if you want to let the framework
   handle the VIDIOC_G/S_PRIORITY ioctls. This requires that you use struct
   v4l2_fh. Eventually this flag will disappear once all drivers use the core
@@ -634,8 +641,22 @@ v4l2_file_operations and locking
 --------------------------------
 
 You can set a pointer to a mutex_lock in struct video_device. Usually this
-will be either a top-level mutex or a mutex per device node. If you want
-finer-grained locking then you have to set it to NULL and do you own locking.
+will be either a top-level mutex or a mutex per device node. By default this
+lock will be used for each file operation and ioctl, but you can disable
+locking for selected ioctls by calling:
+
+        void v4l2_dont_use_lock(struct video_device *vdev, unsigned int cmd);
+
+E.g.: v4l2_dont_use_lock(vdev, VIDIOC_DQBUF);
+
+You have to call this before you register the video_device.
+
+Particularly with USB drivers where certain commands such as setting controls
+can take a long time you may want to do your own locking for the buffer queuing
+ioctls.
+
+If you want still finer-grained locking then you have to set mutex_lock to NULL
+and do you own locking completely.
 
 It is up to the driver developer to decide which method to use. However, if
 your driver has high-latency operations (for example, changing the exposure