Commit graph

18 commits

Author SHA1 Message Date
Hans Verkuil
6e6d76cdc5 [media] v4l2-event: fix regression with initial event handling
If the V4L2_EVENT_SUB_FL_SEND_INITIAL was set, then the application expects
to receive an initial event of the initial value of the control.

However, commit c53c254933 that added the new
v4l2_subscribed_event_ops introduced a regression: while the code still queued
that initial event the __v4l2_event_queue_fh() function was modified to ignore
such requests if sev->elems was 0 (meaning that the event subscription wasn't
finished yet).

And sev->elems was only set to a non-zero value after the add operation
returned.

This patch fixes this by passing the elems value to the add function. Then the
add function can set it before queuing the initial event.

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>
2012-05-15 08:43:06 -03:00
Hans de Goede
3e366149b8 [media] v4l2-ctrls: Use v4l2_subscribed_event_ops
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
[hans.verkuil@cisco.com: Fix a locking bug]
Acked-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-05-07 16:29:35 -03:00
Hans de Goede
c53c254933 [media] v4l2-event: Add v4l2_subscribed_event_ops
Just like with ctrl events, drivers may want to get called back on
listener add / remove for other event types too. Rather then special
casing all of this in subscribe / unsubscribe event it is better to
use ops for this.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-05-07 16:28:35 -03:00
Linus Torvalds
e6f1227e8b Merge branch 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media
* 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  [media] v4l2-ctrl: Send change events to all fh for auto cluster slave controls
  [media] v4l2-event: Don't set sev->fh to NULL on unsubscribe
  [media] v4l2-event: Remove pending events from fh event queue when unsubscribing
  [media] v4l2-event: Deny subscribing with a type of V4L2_EVENT_ALL
  [media] MAINTAINERS: add a maintainer for s5p-mfc driver
  [media] v4l: s5p-mfc: fix reported capabilities
  [media] media: vb2: reset queued list on REQBUFS(0) call
  [media] media: vb2: set buffer length correctly for all buffer types
  [media] media: vb2: add a check for uninitialized buffer
  [media] mxl111sf: fix build warning
  [media] mxl111sf: remove pointless if condition in mxl111sf_config_spi
  [media] mxl111sf: check for errors after mxl111sf_write_reg in mxl111sf_idac_config
  [media] mxl111sf: fix return value of mxl111sf_idac_config
  [media] uvcvideo: GET_RES should only be checked for BITMAP type menu controls
2011-11-12 00:03:50 -02:00
Hans de Goede
e3e72f39b6 [media] v4l2-event: Don't set sev->fh to NULL on unsubscribe
Setting sev->fh to NULL causes problems for the del op added in the next
patch of this series, since this op needs a way to get to its own data
structures, and typically this will be done by using container_of on an
embedded v4l2_fh struct.

The reason the original code is setting sev->fh to NULL is to signal
to users of the event framework that the unsubscription has happened,
but since their is no shared lock between the event framework and users
of it, this is inherently racy, and it also turns out to be unnecessary
as long as both the event framework and the user of the framework do their
own locking properly and the user guarantees that it holds no references
to the subcribed_event structure after its del operation has been called.

This is best explained by looking at the only code currently checking for
sev->fh being set to NULL on unsubscribe, which is the v4l2-ctrls.c send_event
function. Here is the relevant code from v4l2-ctrls: send_event():

	if (sev->fh && (sev->fh != fh ||
			(sev->flags & V4L2_EVENT_SUB_FL_ALLOW_FEEDBACK)))
		v4l2_event_queue_fh(sev->fh, &ev);

Now lets say that v4l2_event_unsubscribe and v4l2-ctrls: send_event() race
on the same sev, then the following could happens:

1) send_event checks sev->fh, finds it is not NULL
<thread switch>
2) v4l2_event_unsubscribe sets sev->fh NULL
3) v4l2_event_unsubscribe calls v4l2_ctrls del_event function, this blocks
   as the thread calling send_event holds the ctrl_lock
<thread switch>
4) send_event calls v4l2_event_queue_fh(sev->fh, &ev) which not is equivalent
   to calling: v4l2_event_queue_fh(NULL, &ev)
5) oops, NULL pointer deref.

Now again without setting sev->fh to NULL in v4l2_event_unsubscribe and
without the (now senseless since always true) sev->fh != NULL check in

1) send_event is about to call v4l2_event_queue_fh(sev->fh, &ev)
<thread switch>
2) v4l2_event_unsubscribe removes sev->list from the fh->subscribed list
<thread switch>
3) send_event calls v4l2_event_queue_fh(sev->fh, &ev)
4) v4l2_event_queue_fh blocks on the fh_lock spinlock
<thread switch>
5) v4l2_event_unsubscribe unlocks the fh_lock spinlock
6) v4l2_event_unsubscribe calls v4l2_ctrls del_event function, this blocks
   as the thread calling send_event holds the ctrl_lock
<thread switch>
8) v4l2_event_queue_fh takes the fh_lock
7) v4l2_event_queue_fh calls v4l2_event_subscribed, does not find it since
   sev->list has been removed from fh->subscribed already -> does nothing
9) v4l2_event_queue_fh releases the fh_lock
10) the caller of send_event releases the ctrl lock (mutex)
<thread switch>
11) v4l2_ctrls del_event takes the ctrl lock
12) v4l2_ctrls del_event removes sev->node from the ev_subs list
13) v4l2_ctrls del_event releases the ctrl lock
14) v4l2_event_unsubscribe frees the sev, to which no references are being
    held anymore

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-11-08 10:29:51 -02:00
Hans de Goede
78c87e863b [media] v4l2-event: Remove pending events from fh event queue when unsubscribing
The kev pointers inside the pending events queue (the available queue) of the
fh point to data inside the sev, unsubscribing frees the sev, thus making these
pointers point to freed memory!

This patch fixes these dangling pointers in the available queue by removing
all matching pending events on unsubscription.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-11-08 10:29:29 -02:00
Hans de Goede
b36b505965 [media] v4l2-event: Deny subscribing with a type of V4L2_EVENT_ALL
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-11-08 10:29:05 -02:00
Paul Gortmaker
35a246363e drivers/media: Add export.h for EXPORT_SYMBOL/THIS_MODULE as required
These two macros were in module.h but now module.h is no longer
sprayed across every source file imaginable, so the users need
to expicitly call out their use of them.

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
2011-10-31 19:31:51 -04:00
Hans Verkuil
3f66f0ed31 [media] v4l2-ctrls/v4l2-events: small coding style cleanups
Thanks to Laurent Pinchart <laurent.pinchart@ideasonboard.com>.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-07-27 17:53:34 -03:00
Hans Verkuil
2151bdc887 [media] v4l2-event: add optional merge and replace callbacks
When the event queue for a subscribed event is full, then the oldest
event is dropped. It would be nice if the contents of that oldest
event could be merged with the next-oldest. That way no information is
lost, only intermediate steps are lost.

This patch adds optional replace() (called when only one kevent was allocated)
and merge() (called when more than one kevent was allocated) callbacks that
will be called to do this job.

These two callbacks are implemented for the V4L2_EVENT_CTRL event.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-07-27 17:53:31 -03:00
Hans Verkuil
f1e393de38 [media] v4l2-event/ctrls/fh: allocate events per fh and per type instead of just per-fh
The driver had to decide how many events to allocate when the v4l2_fh struct
was created. It was possible to add more events afterwards, but there was no
way to ensure that you wouldn't miss important events if the event queue
would fill up for that filehandle.

In addition, once there were no more free events, any new events were simply
dropped on the floor.

For the control event in particular this made life very difficult since
control status/value changes could just be missed if the number of allocated
events and the speed at which the application read events was too low to keep
up with the number of generated events. The application would have no idea
what the latest state was for a control since it could have missed the latest
control change.

So this patch makes some major changes in how events are allocated. Instead
of allocating events per-filehandle they are now allocated when subscribing an
event. So for that particular event type N events (determined by the driver)
are allocated. Those events are reserved for that particular event type.
This ensures that you will not miss events for a particular type altogether.

In addition, if there are N events in use and a new event is raised, then
the oldest event is dropped and the new one is added. So the latest event
is always available.

This can be further improved by adding the ability to merge the state of
two events together, ensuring that no data is lost at all. This will be
added in the next patch.

This also makes it possible to allow the user to determine the number of
events that will be allocated. This is not implemented at the moment, but
would be trivial.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-07-27 17:53:31 -03:00
Hans Verkuil
77068d36d8 [media] v4l2-ctrls/event: remove struct v4l2_ctrl_fh, instead use v4l2_subscribed_event
The v4l2_ctrl_fh struct connected v4l2_ctrl with v4l2_fh so the control
would know which filehandles subscribed to it. However, it is much easier
to use struct v4l2_subscribed_event directly for that and get rid of that
intermediate struct.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-07-27 17:53:31 -03:00
Hans Verkuil
523f46d6ab [media] v4l2-events/fh: merge v4l2_events into v4l2_fh
Drivers that supported events used to be rare, but now that controls can also
raise events this will become much more common since almost all drivers have
controls.

This means that keeping struct v4l2_events as a separate struct make no more
sense. Merging it into struct v4l2_fh simplifies things substantially as it
is now an integral part of the filehandle struct.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-07-27 17:53:31 -03:00
Hans Verkuil
6e239399e5 [media] v4l2-ctrls: add control events
Whenever a control changes value or state an event is sent to anyone
that subscribed to it.

This functionality is useful for control panels but also for applications
that need to wait for (usually status) controls to change value.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2011-07-27 17:53:20 -03:00
Hans Verkuil
ee6869afc9 V4L/DVB: v4l2: add core serialization lock
Drivers can optionally set a pointer to a mutex in struct video_device.
The core will use that to lock before calling open, read, write, unlocked_ioctl,
poll, mmap or release.

Updated the documentation as well and ensure that v4l2-event knows about the
lock: it will unlock it before doing a blocking wait on an event and relock it
afterwards.

Ensure that the 'video_is_registered' check is done when the lock is held:
a typical disconnect will take the lock as well before unregistering the
device nodes, so to prevent race conditions the video_is_registered check
should also be done with the lock held.

Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2010-10-21 01:06:14 -02:00
Laurent Pinchart
0a4f8d0798 V4L/DVB: event: Export the v4l2_event_init and v4l2_event_dequeue functions
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Sakari Ailus <sakari.ailus@maxwell.research.nokia.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2010-05-19 12:58:24 -03:00
Sakari Ailus
f3cd385a9c V4L/DVB: V4L: Events: Replace bad WARN_ON() with assert_spin_locked()
spin_is_locked() always returns zero when spinlock debugging is
disabled on a single CPU machine. Replace WARN_ON() with
assert_spin_locked().

Thanks to Laurent Pinchart for spotting this!

Signed-off-by: Sakari Ailus <sakari.ailus@maxwell.research.nokia.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2010-05-19 12:58:23 -03:00
Sakari Ailus
c3b5b0241f V4L/DVB: V4L: Events: Add backend
Add event handling backend to V4L2. The backend handles event subscription
and delivery to file handles. Event subscriptions are based on file handle.
Events may be delivered to all subscribed file handles on a device
independent of where they originate from.

Signed-off-by: Sakari Ailus <sakari.ailus@maxwell.research.nokia.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2010-05-19 12:58:06 -03:00