Age | Commit message (Collapse) | Author | Files | Lines |
|
atomisp_ospm_dphy_up() is an empty function now and
atomisp_ospm_dphy_down() contains a couple of checks + goto done
statements which don't matter since the function always ends up at
the done label regardless and then it does 1 pci-config write.
Move the single pci-config write directly to atomisp_power_off()
and remove the atomisp_ospm_dphy_up()/_down() functions.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
atomisp_suspend() contains a 1:1 copy of atomisp_runtime_suspend() and
the same goes for the resume() functions.
Rename the runtime functions to atomisp_power_on()/_off() and use these
as runtime-pm handlers as well as helper in other places to remove
the code duplication.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
atomisp_reset()
atomisp_reset() calls atomisp_mrfld_power_down() after calling
atomisp_runtime_suspend(), which already calls
atomisp_mrfld_power_down() itself.
And the some goes for atomisp_runtime_resume() / atomisp_mrfld_power_up().
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
To fix atomisp_queue_setup() sometimes failing it needs to be able to call
atomisp_set_fmt(), but atomisp_queue_setup() (VIDIOC_REQBUFS) does not get
passed a file handle by the videobuf2 core.
Partly revert commit b3be98f984d4 ("media: atomisp: Remove a couple of not
useful function wrappers") so that atomisp_set_fmt() can be used
without a file handle.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
Convert atomisp to use videobuf2.
This fixes mmap not working and in general moving over to
the more modern videobuf2 is a good plan.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
Instead of using an integer to keep count of how many buffers have
been handed over to the ISP (buffers_in_css) move buffers handed
over to the ISP to a new buffers_in_css list_head so that we can
easily loop over them.
This removes the need for atomisp_flush_video_pipe() to loop over
all buffers and then (ab)use the state to figure out if they
were handed over to the ISP.
Since the buffers are now always on a list when owned by the driver
this also allows the buffer_done path on flush vs normal completion
to be unified (both now need a list_del()) and this common code can
now be factored out into a new atomisp_buffer_done() helper.
This is a preparation patch for moving the driver over to
the videobuf2 framework.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
The atomisp_buffers_queued[_pipe] functions are not used anywhere,
remove them.
Reviewed-by: Andy Shevchenko <andy@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
atomisp_source_pad_to_stream_id() returns ATOMISP_INPUT_STREAM_GENERAL
unconditionally now. Drop it and directly use ATOMISP_INPUT_STREAM_GENERAL
in its callers.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
atomisp_set_raw_buffer_bitmap() is only used in atomisp_cmd.c,
make it static. Unfortunately this still requires a forward
declaration (the function cannot be moved easily).
Still this will at least make it obvious to anyone reading the code
that the function is not used elsewhere.
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
The __atomisp_reqbufs(), __atomisp_streamoff() are 1:1 wrappers for
the non __ prefixed functions now, drop these wrappers.
The atomisp_s_fmt_cap() wrapper is almost a 1:1 wrapper for
atomisp_set_fmt() adjust the latter to have the right function
prototype and drop the wrapper.
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
The watchdog timer code to recover from the ISP getting stuck has several
major issues:
1. There is no way to do fault injection and normally the ISP does not
get stuck, so is it is impossible to test it.
2. It in essence just stops all streams, resets the ISP and then brings
everything back up. Userspace can easily do this itself by using a
timeout on dqbuf and then closing (which causes a poweroff) +
re-opening the device. Doing this in userspace (if it ever turns out
to be necessary) greatly simplifies the kernel code and in general
will be a more robust solution.
Even just a quick look at the code finds several more issues:
3. The need to sync-cancel the timers + work on streamoff requires
isp->mutex to be dropped halfway during the ioctl opening all sorts of
races.
4. The atomisp code supports setting up 2 pipelines, streaming from
two sensors at the same time. But there is only a single wdt_work
and stopping one of the 2 streams will cancel the timers + work,
stopping the wdt even though the other stream might still be running.
5. In case atomisp_css_flush() the sync cancel is done while keeping
isp->mutex locked, causing a deadlock when racing with wdt_work which
also takes isp->mutex.
6. Even though the watchdog is purely a software/driver thing which
just checkes that new frames keep coming in, there are 2 completely
different implementations for the ISP2400/ISP2401 which is not
necessary at all.
So all in all I believe that it is better to just remove the current
watchdog implementation. Fixing all the issues with the current
implementation will be so much work, that if it turns out that we do
need something like this then doing a clean re-implementation from
scratch will be better anyways.
wdt_work was also (ab)used to reset the ISP after the firmware signalled
an fw-assert error through the irq, add a new assert_recover_work to
replace this.
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
The ACC /dev/video# device node uses a struct video_device embedded in
an atomisp_acc_pipe struct instead of in an atomisp_video_pipe struct.
Yet it uses the same file-ops and ioctl-ops even though it does not have
a videobuf queue, which makes e.g. the mmap fop nonsense.
Worse the only file-ops / ioctls which differentiate between
the 2 types and correctly do container_of on the right type
are the open/release fops and the vidioc_default handler.
The mmap and poll fops and *all* other ioctl handlers unconditionally
do container_of on the passed in struct video_device blindly assuming
they are dealing with the one embedded in the atomisp_video_pipe struct.
This makes it trivial for userspace to cause all sort of undefined behavior
by calling mmap, poll or the other ioctls on the ACC device node!
Presumably the use of the ACC device node was to allow making the special
ioctls to load custom programs while the other /dev/video# nodes were
already open, since the /dev/video# nodes can currently all be opened
only once (which needs to be fixed).
commit 4bbca788b6eb ("media: atomisp: remove private acceleration ioctls")
has removed the custom ATOMISP_ACC_* ioctls, so there no longer is any
reason to keep the ACC device node.
As explained above its presence can easily cause the kernel to crash,
so remove the ACC device node and the code for handling it.
Reviewed-by: Andy Shevchenko <andriy.shevchenko@intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
After the file-injection support removal, atomisp_video_pipe->type never
is V4L2_BUF_TYPE_VIDEO_OUTPUT anymore, so the V4L2_BUF_TYPE_VIDEO_OUTPUT
support path in atomisp_video_init() is never hit and this path is
the only user of atomisp_file_fops and atomisp_file_ioctl_ops.
Remove atomisp_file_fops and atomisp_file_ioctl_ops and all of
the functions which are only referenced by these ops structs.
Reviewed-by: Andy Shevchenko <andriy.shevchenko@intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
|
|
Simplify the code by getting rid of this function, as it ends
being just a single line of code.
Link: https://lore.kernel.org/linux-media/80ea920d14379124ba92aab2e6a6d12a92d79b2b.1636544620.git.mchehab+huawei@kernel.org
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
The current logic only initializes pipe->pix after setting up a
pipeline - e. g. after start streaming.
While it makes sense to get the format of the pipeline, when
it is set, this breaks support for generic applications, as they
rely on getting the current sensor format (which is usually the
highest resolution format).
So, implement a call to the sensor's get_fmt, when this is called
before setting up a pipeline.
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
coding-style in file atomisp_cmd.h
Format all comments according to the coding-style.
Issue detected by checkpatch.pl.
Signed-off-by: Aline Santana Cordeiro <alinesantanacordeiro@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
The atomisp_set_fmt() function has multiple copies of the large
v4l2_format structure on its stack, resulting in a warning about
excessive stack usage in some rare randconfig builds.
drivers/staging/media/atomisp/pci/atomisp_cmd.c:5600:5: error: stack frame size of 1084 bytes in function 'atomisp_set_fmt' [-Werror,-Wframe-larger-than=]
Of this structure, only three members in the 'fmt.pix' member are
used, so simplify it by using the smaller v4l2_pix_format structure
directly. This reduces the stack usage to 612 bytes, and it could
be reduced further by only storing the three members that are used.
Link: https://lore.kernel.org/linux-media/20210226140524.3478346-1-arnd@kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
struct atomisp device has struct device and struct pci_dev pointers
which are basically duplicates of each other. Drop the latter
in favour of the former.
While here, unify pdev to be pointer to struct pci_device and reindent
some (touched) lines for better readability.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
The hive_isp_css_custom_host_hrt.h code, together
with atomisp_helper.h, provides an abstraction layer for
some functions inside atomisp_compat_css20.c and atomisp_cmd.c.
There's no good reason for that. In a matter of fact, after
removing the abstraction, the code looked a lot cleaner
and easier to understand.
So, get rid of them.
While here, get rid also of the udelay(1) abstraction code.
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
This driver is licensed under GPL 2.0, as stated inside their
headers.
Add the proper tag there. We should probably latter cleanup
the reduntant licensing text, but this could be done later,
after we get rid of other abstraction layers.
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
The very same macros are defined as CSS_foo and IA_CSS_foo.
Remove this abstraction, as it just make things confusing,
for no good reason.
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
This driver has very long directories without a good
reason (IMHO). Let's drop two directories from such hierarchy,
in order to simplify things a little bit and make the dir
output a bit more readable.
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|