[media] cx231xx: can't proceed if I2C bus register fails
The driver should not ignore errors while registering the I2C
bus, as this device can't even minimally work without the buses,
as it uses those buses internally to talk with the several IP
blocks inside the chip.
[media] cx231xx-i2c: handle errors with cx231xx_get_i2c_adap()
The cx231xx_get_i2c_adap() function should return the I2C
adapter that will be used to talk with a device. It should never
be NULL, as otherwise the driver will try to dereference a
null pointer.
We might instead fix the callers, but if this condition
ever happens, it is really a driver bug, because i2c_port
should always be a value from enum CX231XX_I2C_MASTER_PORT.
Found when checking the code due to this bug:
[ 39.769021] BUG: unable to handle kernel NULL pointer dereference at 0000000000000002
[ 39.769105] IP: [<ffffffff81638393>] i2c_master_send+0x13/0x70
The cx231xx_set_agc_analog_digital_mux_select() callers
expect it to return 0 or an error. Returning a positive value
makes the first attempt to switch between analog/digital to fail.
With the current settings, only one channel locks properly.
That's likely because, when this driver was written, Brazil
were still using experimental transmissions.
Change it to reproduce the settings used by the newer drivers.
That makes it lock on other channels.
Tested with both PixelView SBTVD Hybrid (cx231xx-based) and
C3Tech Digital Duo HDTV/SDTV (em28xx-based) devices.
On this frontend, it takes a while to start output normal
TS data. That only happens on state S9. On S8, the TS output
is enabled, but it is not reliable enough.
However, the zigzag loop is too fast to let it sync.
As, on practical tests, the zigzag software loop doesn't
seem to be helping, but just slowing down the tuning, let's
switch to hardware algorithm, as the tuners used on such
devices are capable of work with frequency drifts without
any help from software.
Merge branch 'docs-next' of git://git.lwn.net/linux into patchwork
* 'docs-next' of git://git.lwn.net/linux:
doc-rst: define PDF's of the media folder
doc-rst: generic way to build PDF of sub-folders
docs: sphinx-extensions: add metadata parallel-safe
docs-rst: kernel-doc: fix typedef output in RST format
docs-rst: improve typedef parser
docs: kernel-parameter: Improve the description of nr_cpus and maxcpus
docs-rst: kernel-doc: better output struct members
Markus Heiser [Wed, 24 Aug 2016 15:36:14 +0000 (17:36 +0200)]
doc-rst: generic way to build PDF of sub-folders
This extends the method to build only sub-folders to the targets
"latexdocs" and "pdfdocs". To do so, a conf.py in the sub-folder is
required, where the latex_documents of the sub-folder are
defined. E.g. to build only gpu's PDF add the following to the
Documentation/gpu/conf.py::
+latex_documents = [
+ ("index", "gpu.tex", "Linux GPU Driver Developer's Guide",
+ "The kernel development community", "manual"),
+]
and run:
make SPHINXDIRS=gpu pdfdocs
Signed-off-by: Markus Heiser <markus.heiser@darmarIT.de> Signed-off-by: Jonathan Corbet <corbet@lwn.net>
The setup() function of a Sphinx-extension can return a dictionary. This
is treated by Sphinx as metadata of the extension [1].
With metadata "parallel_read_safe = True" a extension is marked as
save for "parallel reading of source". This is needed if you want
build in parallel with N processes. E.g.:
make SPHINXOPTS=-j4 htmldocs
will no longer log warnings like:
WARNING: the foobar extension does not declare if it is safe for
parallel reading, assuming it isn't - please ask the extension author
to check and make it explicit.
docs-rst: kernel-doc: fix typedef output in RST format
When using a typedef function like this one:
typedef bool v4l2_check_dv_timings_fnc (const struct v4l2_dv_timings * t, void * handle);
The Sphinx C domain expects it to create a c:type: reference,
as that's the way it creates the type references when parsing
a c:function:: declaration.
Baoquan He [Wed, 24 Aug 2016 05:06:45 +0000 (13:06 +0800)]
docs: kernel-parameter: Improve the description of nr_cpus and maxcpus
From the old description people still can't get what's the exact
difference between nr_cpus and maxcpus. Especially in kdump kernel
nr_cpus is always suggested if it's implemented in the ARCH. The
reason is nr_cpus is used to limit the max number of possible cpu
in system, the sum of already plugged cpus and hot plug cpus can't
exceed its value. However maxcpus is used to limit how many cpus
are allowed to be brought up during bootup.
Signed-off-by: Baoquan He <bhe@redhat.com> Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Hans Verkuil [Tue, 23 Aug 2016 06:48:37 +0000 (03:48 -0300)]
[media] redrat3: fix sparse warning
Fix this sparse warning:
drivers/media/rc/redrat3.c:490:18: warning: incorrect type in assignment (different base types)
drivers/media/rc/redrat3.c:495:9: warning: cast to restricted __be32
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
docs-rst: kernel-doc: better output struct members
Right now, for a struct, kernel-doc produces the following output:
.. c:type:: struct v4l2_prio_state
stores the priority states
**Definition**
::
struct v4l2_prio_state {
atomic_t prios[4];
};
**Members**
``atomic_t prios[4]``
array with elements to store the array priorities
Putting a member name in verbatim and adding a continuation line
causes the LaTeX output to generate something like:
item[atomic_t prios\[4\]] array with elements to store the array priorities
Everything inside "item" is non-breakable, with may produce
lines bigger than the column width.
Markus Elfring [Fri, 19 Aug 2016 08:04:54 +0000 (05:04 -0300)]
[media] dvb_frontend: Use memdup_user() rather than duplicating its implementation
* Reuse existing functionality from memdup_user() instead of keeping
duplicate source code.
This issue was detected by using the Coccinelle software.
* Return directly if this copy operation failed.
* Replace the specification of data structures by pointer dereferences
to make the corresponding size determination a bit safer according to
the Linux coding style convention.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Colin Ian King [Sat, 13 Aug 2016 18:16:54 +0000 (15:16 -0300)]
[media] helene: fix memory leak when heleno_x_pon fails
The error return path of failed calls to heleno_x_pon leak
memory because priv is not kfree'd. Fix this by kfree'ing
priv before returning.
Signed-off-by: Colin Ian King <colin.king@canonical.com> Acked-by: Abylay Ospan <aospan@netup.ru> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The v4l2_subdev_core_ops .registered_async callback was added to notify
a subdev when its entity has been registered with the media device, to
allow for example to modify the media graph (i.e: adding entities/links).
But that's not needed since there is already a .registered callback in
struct v4l2_subdev_internal_ops that's called after the entity has been
registered with the media device in v4l2_device_register_subdev().
Suggested-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
[media] tvp5150: use sd internal ops .registered instead .registered_async
The driver is using the struct v4l2_subdev_core_ops .registered_async
callback to register the connector entities and create the pad links
after the subdev entity has been registered with the media device.
But the .registered_async callback isn't needed since the v4l2 core
already calls the struct v4l2_subdev_internal_ops .registered callback
in v4l2_device_register_subdev(), after media_device_register_entity().
So, use the .registered() callback instead of the .registered_async()
that is going to be removed in a following patch since isn't needed.
Suggested-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Sakari Ailus [Thu, 11 Aug 2016 10:18:37 +0000 (07:18 -0300)]
[media] v4l: Do not allow re-registering sub-devices
Albeit not prohibited explicitly, re-registering sub-devices generated a
big, loud warning which quite likely soon was followed by a crash. What
followed was re-initialising a media entity, driver's registered() callback
being called and re-adding a list entry to a list.
Prevent this by returning an error if a sub-device is already registered.
[mchehab@s-opensource.com: reorder logic to check if !sd before dereferencing it] Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Liu Ying [Thu, 11 Aug 2016 05:10:09 +0000 (02:10 -0300)]
[media] media-entity.h: Correct KernelDoc of media_entity_enum_empty()
The function media_entity_enum_empty() returns true when the bitmap
of the input parameter media entity enumeration is empty instead of marked.
This patch corrects the return value description of the function.
Signed-off-by: Liu Ying <gnuiyl@gmail.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Stephen Backway [Tue, 9 Aug 2016 22:43:40 +0000 (19:43 -0300)]
[media] cx23885: Add support for Hauppauge WinTV quadHD ATSC version
Add support for the Hauppauge WinTV quadHD ATSC version.
IR support has not been provided, all 4 tuners, demodulators etc are working.
Further documentation can be found on Linux TV wiki.
Signed-off-by: Stephen Backway <stev391@gmail.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
According to its documentation, rcar_fcp_enable() returns 0 on success
or a negative error code if an error occurs. Hence
fdp1_pm_runtime_resume() and vsp1_pm_runtime_resume() forward its return
value to their callers.
However, rcar_fcp_enable() forwards the return value of
pm_runtime_get_sync(), which can actually be 1 on success, leading to
the resume failure above.
To fix this, consider only negative values returned by
pm_runtime_get_sync() to be failures.
The MFC device driver is a v4l2 driver which can encode/decode video
raw/elementary streams and has support for all popular video codecs.
The driver's watchdog_workqueue has been replaced with system_wq since
it queues a single work item, &dev->watchdog_work, which calls for no
ordering requirement. The work item is involved in running the watchdog
timer and is not being used on a memory reclaim path.
Work item has been flushed in s5p_mfc_remove() to ensure
that there are no pending tasks while disconnecting the driver.
[media] media: s5p-mfc Fix misspelled error message and checkpatch errors
Fix misspelled error message and existing checkpatch errors in the
error message conditional.
WARNING: suspect code indent for conditional statements (8, 24)
if (ctx->state != MFCINST_HEAD_PARSED &&
[...]
mfc_err("Can not get crop information\n");
[media] media: s5p-mfc fix invalid memory access from s5p_mfc_release()
If s5p_mfc_release() runs after s5p_mfc_remove(), the former will access
invalid s5p_mfc_dev pointer saved in the s5p_mfc_ctx and runs into kernel
paging request errors.
Clear ctx dev pointer in s5p_mfc_remove() and change s5p_mfc_release() to
avoid work that requires ctx->dev.
odroid kernel: Unable to handle kernel paging request at virtual address f17c1104
odroid kernel: pgd = ebca4000
odroid kernel: [f17c1104] *pgd=6e23d811, *pte=00000000, *ppte=00000000
odroid kernel: Internal error: Oops: 807 [#1] PREEMPT SMP ARM
odroid kernel: Modules linked in: cpufreq_userspace cpufreq_powersave
cpufreq_conservative s5p_mfc s5p_jpeg v4l2_mem2mem
videobuf2_dma_contig videobuf2_memops videobuf2_v4l2 videobuf2_core
v4l2_common videodev media
odroid kernel: Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
odroid kernel: task: c2241400 ti: e7018000 task.ti: e7018000
odroid kernel: PC is at s5p_mfc_reset+0x40/0x28c [s5p_mfc]
odroid kernel: LR is at s5p_mfc_reset+0x34/0x28c [s5p_mfc]
odroid kernel: pc : [<bf15bfbc>] lr : [<bf15bfb0>] psr: 60010013
odroid kernel: [<bf15bfbc>] (s5p_mfc_reset [s5p_mfc]) from [<bf15c62c>]
(s5p_mfc_deinit_hw+0x14/0x3c [s5p_mfc])
odroid kernel: [<bf15c62c>] (s5p_mfc_deinit_hw [s5p_mfc]) from [<bf155958>]
(s5p_mfc_release+0xac/0x1c4 [s5p_mfc])
odroid kernel: [<bf155958>] (s5p_mfc_release [s5p_mfc]) from [<bf021344>]
(v4l2_release+0x38/0x74 [videodev])
odroid kernel: [<bf021344>] (v4l2_release [videodev]) from [<c01e4274>]
(__fput+0x80/0x1c8)
odroid kernel: [<c01e4274>] (__fput) from [<c0135c58>]
(task_work_run+0x94/0xc8)
odroid kernel: [<c0135c58>] (task_work_run) from [<c010a9d4>]
(do_work_pending+0x7c/0xa4)
odroid kernel: [<c010a9d4>] (do_work_pending) from [<c0107794>]
(slow_work_pending+0xc/0x20)
odroid kernel: Code: eb3edacce5953078e3a06000e2833c11 (e5836004)
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com> Tested-by: Luis de Bethencourt <luisbg@osg.samsung.com> Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
[media] gsc-m2m: improve v4l2_capability driver and card fields
According to the V4L2 documentation the driver and card fields should be
used to identify the driver and the device but the gsc-m2m driver fills
those field using the platform device name, which in turn is the name of
the device DT node.
So not only the filled information isn't correct but also the same values
are used in the driver, card and bus_info fields.
Before this patch:
Driver Info (not using libv4l2):
Driver name : 13e00000.video-
Card type : 13e00000.video-scaler
Bus info : platform:13e00000.video-scaler
Driver version: 4.7.0
After this patch:
Driver Info (not using libv4l2):
Driver name : exynos-gsc
Card type : exynos-gsc gscaler
Bus info : platform:13e00000.video-scaler
Driver version: 4.7.0
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com> Acked-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
[media] s5p-jpeg: only fill driver's name in capabilities driver field
The driver fills in both the struct v4l2_capability driver and card fields
the same values, that is the driver's name plus the information if the dev
is a decoder or an encoder.
But the driver field has a fixed length of 16 bytes so the filled data is
truncated:
Driver Info (not using libv4l2):
Driver name : s5p-jpeg decode
Card type : s5p-jpeg decoder
Bus info : platform:11f50000.jpeg
Driver version: 4.7.0
Also, this field should only contain the driver's name so use just that.
The information if the device is a decoder or an encoder is in the card
type field anyways.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com> Acked-by: Jacek Anaszewski <j.anaszewski@samsung.com> Acked-by: Hans Verkuil <hans.verkuil@cisco.com> Reviewed-by: Shuah Khan <shuahkh@osg.samsung.com> Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The same struct device_node * is used for looking up the I2C sensor, OF
graph endpoint and port. So the reference count is incremented but not
decremented for the endpoint and port nodes.
Fix this by having separate pointers for each node looked up.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com> Tested-by: Nicolas Dufresne <nicoas.dufresne@collabora.com> Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Remove those two vars that aren't used, as reported by smatch:
drivers/media/pci/tw5864/tw5864-video.c: In function 'tw5864_prepare_frame_headers':
drivers/media/pci/tw5864/tw5864-video.c:1219:16: warning: variable 'space_before_sl_hdr' set but not used [-Wunused-but-set-variable]
unsigned long space_before_sl_hdr;
^~~~~~~~~~~~~~~~~~~
drivers/media/pci/tw5864/tw5864-video.c:1218:6: warning: variable 'sl_hdr' set but not used [-Wunused-but-set-variable]
u8 *sl_hdr;
^~~~~~
Hans Verkuil [Thu, 21 Jul 2016 12:14:03 +0000 (09:14 -0300)]
[media] vb2: add WARN_ONs checking if a valid struct device was passed
The dma-contig and dma-sg memops require a valid struct device for
the DMA to be handled correctly.
Call WARN_ON and return ERR_PTR(-EINVAL) if it was NULL.
Setting the correct device pointer was forgotten in several new driver
submissions. This was caught during code review, but it really should be
caught in the code.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
[media] vb2: remove TODO comment for dma-buf in QBUF
There is a TODO comment about the dma-buf being mapped in VIDIOC_QBUF
instead of doing it closer to when the actual DMA is going to happen
when the buffers are queued in the driver (i.e: __enqueue_in_driver).
But there is a reason to do it earlier in QBUF, and is that userspace
has no way to know if a exported dma-buf can be imported successfully
and so relies on QBUF succeeding as indication that the dma-buf mapped.
If QBUF fails, the application can fallback to another streaming I/O
method. But moving the dma-buf mapping later when queueing the buffers
can be too late for userspace to recover, since it may had dropped the
buffer(s) already when it knows that the dma-buf mapping failed.
So remove the TODO instead and change the comment to explain this.
Suggested-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Hans Verkuil [Wed, 20 Jul 2016 12:31:25 +0000 (09:31 -0300)]
[media] vivid: return -ENODATA if the current input doesn't support g/s_selection
Returning -EINVAL indicates wrong arguments, but that's not the case
here.
Returning -ENOTTY is also no option, since the ioctl is implemented, but
it just is not valid for this input.
So use -ENODATA instead. This is also used elsewhere when an ioctl isn't
valid for a specific input.
In this case G/S_SELECTION returned -EINVAL for the webcam input. That
input doesn't support cropping, instead it uses ENUM_FRAMESIZES to
enumerate a list of discrete frame sizes.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The workqueue work_queue is involved in EDID (Extended Display
Identification Data) handling.
It has a single work item(&state->edid_handler) and hence
doesn't require ordering. It is not being used on a memory reclaim path.
Hence, the singlethreaded workqueue has been replaced with
the use of system_wq.
&state->edid_handler is a self requeueing work item and it has been
been sync cancelled in ad9389b_remove() to ensure that nothing is
pending when the driver is disconnected.
The workqueue "work_thread" is involved in streaming the camera data.
It has a single work item(&dev->work_struct) and hence doesn't require
ordering. Also, it is not being used on a memory reclaim path.
Hence, the singlethreaded workqueue has been replaced with the use of
system_wq.
System workqueues have been able to handle high level of concurrency
for a long time now and hence it's not required to have a singlethreaded
workqueue just to gain concurrency. Unlike a dedicated per-cpu workqueue
created with create_singlethread_workqueue(), system_wq allows multiple
work items to overlap executions even on the same CPU; however, a
per-cpu workqueue doesn't have any CPU locality or global ordering
guarantee unless the target CPU is explicitly specified and thus the
increase of local concurrency shouldn't make any difference.
Work item has been flushed in sd_stop0() to ensure that there are no
pending tasks while disconnecting the driver.
The workqueue "work_thread" is involved in streaming the camera data.
It has a single work item(&sd->work_struct) and hence doesn't require
ordering. Also, it is not being used on a memory reclaim path.
Hence, the singlethreaded workqueue has been replaced with the use of
system_wq.
System workqueues have been able to handle high level of concurrency
for a long time now and hence it's not required to have a singlethreaded
workqueue just to gain concurrency. Unlike a dedicated per-cpu workqueue
created with create_singlethread_workqueue(), system_wq allows multiple
work items to overlap executions even on the same CPU; however, a
per-cpu workqueue doesn't have any CPU locality or global ordering
guarantee unless the target CPU is explicitly specified and thus the
increase of local concurrency shouldn't make any difference.
Work item has been flushed in sd_stop0() to ensure that there are no
pending tasks while disconnecting the driver.
The workqueue "work_thread" is involved in streaming the camera data.
It has a single work item(&sd->work_struct) and hence doesn't require
ordering. Also, it is not being used on a memory reclaim path.
Hence, the singlethreaded workqueue has been replaced with the use of
system_wq.
System workqueues have been able to handle high level of concurrency
for a long time now and hence it's not required to have a singlethreaded
workqueue just to gain concurrency. Unlike a dedicated per-cpu workqueue
created with create_singlethread_workqueue(), system_wq allows multiple
work items to overlap executions even on the same CPU; however, a
per-cpu workqueue doesn't have any CPU locality or global ordering
guarantee unless the target CPU is explicitly specified and thus the
increase of local concurrency shouldn't make any difference.
Work item has been flushed in sd_stop0() to ensure that there are no
pending tasks while disconnecting the driver.
The workqueue "work_thread" is involved in updating the JPEG quality
of the gspca_dev. It has a single work item(&sd->work) and hence doesn't
require ordering. Also, it is not being used on a memory reclaim path.
Hence, the singlethreaded workqueue has been replaced with the use of
system_wq.
System workqueues have been able to handle high level of concurrency
for a long time now and hence it's not required to have a singlethreaded
workqueue just to gain concurrency. Unlike a dedicated per-cpu workqueue
created with create_singlethread_workqueue(), system_wq allows multiple
work items to overlap executions even on the same CPU; however, a
per-cpu workqueue doesn't have any CPU locality or global ordering
guarantee unless the target CPU is explicitly specified and thus the
increase of local concurrency shouldn't make any difference.
Work item has been flushed in sd_stop0() to ensure that there are no
pending tasks while disconnecting the driver.
The workqueue "workqueue" is involved in polling the pvrusb2 hardware
(pvr2_hdw).
It has a single work item(&hdw->workpoll) and hence doesn't require
ordering. Also, it is not being used on a memory reclaim path.
Hence, the singlethreaded workqueue has been replaced with the use of
system_wq.
System workqueues have been able to handle high level of concurrency
for a long time now and hence it's not required to have a singlethreaded
workqueue just to gain concurrency. Unlike a dedicated per-cpu workqueue
created with create_singlethread_workqueue(), system_wq allows multiple
work items to overlap executions even on the same CPU; however, a
per-cpu workqueue doesn't have any CPU locality or global ordering
guarantee unless the target CPU is explicitly specified and thus the
increase of local concurrency shouldn't make any difference.
Work item has been flushed in pvr2_hdw_destroy to ensure that there are no
pending tasks while disconnecting the driver.
projects / karo-tx-linux.git / log