platform/kernel/linux-starfive.git
19 months agomedia: mt9p031: Drop bogus v4l2_subdev_get_try_crop() call from mt9p031_init_cfg()
Marek Vasut [Sun, 20 Nov 2022 14:13:04 +0000 (14:13 +0000)]
media: mt9p031: Drop bogus v4l2_subdev_get_try_crop() call from mt9p031_init_cfg()

The mt9p031_init_cfg() already calls __mt9p031_get_pad_crop(), which
correctly calls v4l2_subdev_get_try_crop() on V4L2_SUBDEV_FORMAT_TRY
or returns &mt9p031->crop on V4L2_SUBDEV_FORMAT_ACTIVE. No need to
call v4l2_subdev_get_try_crop() in mt9p031_init_cfg() again in case
of both V4L2_SUBDEV_FORMAT_TRY and V4L2_SUBDEV_FORMAT_ACTIVE.

This also fixes a splat generated by this call since commit
2ba3e38517f5a ("media: v4l: subdev: Fail graciously when getting try data for NULL state")
because v4l2_subdev_get_try_crop() is called with sd_state = NULL
in mt9p031_init_cfg().

Fixes: 69681cd041648 ("media: mt9p031: Move open subdev op init code into init_cfg")
Signed-off-by: Marek Vasut <marex@denx.de>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: i2c: ov9282: Add support for regulators.
Dave Stevenson [Wed, 5 Oct 2022 15:20:18 +0000 (16:20 +0100)]
media: i2c: ov9282: Add support for regulators.

The sensor takes 3 supply rails - AVDD, DVDD, and DOVDD.

Add hooks into the regulator framework for each of these
regulators.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: dt-bindings: media: ovti,ov9282: Add optional regulators
Dave Stevenson [Wed, 5 Oct 2022 15:20:17 +0000 (16:20 +0100)]
media: dt-bindings: media: ovti,ov9282: Add optional regulators

The OV9282 image sensor takes 3 voltage supplies, so
define the relevant regulators.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: i2c: st-vgxy61: Fix regulator counter underflow
Benjamin Mugnier [Mon, 7 Nov 2022 15:00:00 +0000 (15:00 +0000)]
media: i2c: st-vgxy61: Fix regulator counter underflow

Previously regulators were enabled on probe and never again.
However, as regulators are disabled on power off. After a second power off
the regulator counter will underflow. Plus regulators are not required
for probing the sensor, but for streaming.
Fix this by enabling regulators on power on to balance regulator counter
properly.

Signed-off-by: Benjamin Mugnier <benjamin.mugnier@foss.st.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: ov2740: Use traditional pattern when checking error codes
Andy Shevchenko [Tue, 26 Jul 2022 12:05:56 +0000 (13:05 +0100)]
media: ov2740: Use traditional pattern when checking error codes

Instead of 'if (!ret)' switch to "check for the error first" rule.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: ov2740: Add missed \n to the end of the messages
Andy Shevchenko [Tue, 26 Jul 2022 12:05:55 +0000 (13:05 +0100)]
media: ov2740: Add missed \n to the end of the messages

Add missed \n to the end of the messages.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Bingbu Cao <bingbu.cao@intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: ov2740: Switch to use dev_err_probe()
Andy Shevchenko [Tue, 26 Jul 2022 12:05:54 +0000 (13:05 +0100)]
media: ov2740: Switch to use dev_err_probe()

Switch to use dev_err_probe() to simpify error path and unify message
template.

While at it, add missed \n to the end of the messages.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Bingbu Cao <bingbu.cao@intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: ov2740: Drop redundant assignments of ret = 0
Andy Shevchenko [Tue, 26 Jul 2022 12:05:53 +0000 (13:05 +0100)]
media: ov2740: Drop redundant assignments of ret = 0

In some cases it might hide real bugs, in most cases here it's just
redundant as it's being reassigned immediately after initial assignment.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: ov2740: Remove duplicate check for NULL fwnode
Andy Shevchenko [Tue, 26 Jul 2022 12:05:52 +0000 (13:05 +0100)]
media: ov2740: Remove duplicate check for NULL fwnode

fwnode API does proper checks and returns correct codes, no need
to repeat it in the caller.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Bingbu Cao <bingbu.cao@intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: ov2740: Switch from __maybe_unused to pm_sleep_ptr() etc
Andy Shevchenko [Tue, 26 Jul 2022 12:05:51 +0000 (13:05 +0100)]
media: ov2740: Switch from __maybe_unused to pm_sleep_ptr() etc

Letting the compiler remove these functions when the kernel is built
without CONFIG_PM_SLEEP support is simpler and less heavier for builds
than the use of __maybe_unused attributes.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: ov2740: Remove duplicative pointer in struct nvm_data
Andy Shevchenko [Tue, 26 Jul 2022 12:05:49 +0000 (13:05 +0100)]
media: ov2740: Remove duplicative pointer in struct nvm_data

The struct i2c_client pointer is used only to get driver data,
associated with a struct device or print messages on behalf.
Moreover, the very same pointer to a struct device is already
assigned by a regmap and can be retrieved from there.
No need to keep a duplicative pointer.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: tc358746: drop selecting COMMON_CLK
Marco Felsch [Mon, 7 Nov 2022 09:42:12 +0000 (09:42 +0000)]
media: tc358746: drop selecting COMMON_CLK

Still there are archs/platforms which do not support the common clk
framework. If such a platform is used in combination with the module
enabled the compiler will throw an error. Since the clock has stubs if
not selected we can drop it, so it is up to the arch/platform to select
the correct clock framework.

Fixes: 80a21da36051 ("media: tc358746: add Toshiba TC358746 Parallel to CSI-2 bridge driver")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Marco Felsch <m.felsch@pengutronix.de>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: i2c: aptina-pll: Remove a useless include
Christophe JAILLET [Sat, 12 Nov 2022 20:51:06 +0000 (20:51 +0000)]
media: i2c: aptina-pll: Remove a useless include

<linux/lcm.h> is not needed for this driver. Remove the corresponding
 #include.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: staging: omap4iss: remove cacheflush import
Ian Cowan [Mon, 14 Nov 2022 16:17:00 +0000 (16:17 +0000)]
media: staging: omap4iss: remove cacheflush import

The cacheflush import is never used, so it is safe to remove it as an
import.

Signed-off-by: Ian Cowan <ian@linux.cowan.aero>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: imx: Unstage the imx7-media-csi driver
Laurent Pinchart [Tue, 10 May 2022 18:37:05 +0000 (19:37 +0100)]
media: imx: Unstage the imx7-media-csi driver

The imx7-media-csi driver, currently in staging, is ready for
prime-time. The staging TODO file lists a few items specific to that
driver, that are already addressed (the "all of the above" part) or can
be addressed later:

- The frame interval monitoring support is a software mechanism to
  monitor the device for unexpected stalls, and should be part of the
  V4L2 core if desired.

- Restricting the support media bus formats based on the SoC integration
  only aims at reducing userspace confusion by not enumerating options
  that are known not to be possible, it won't cause regressions if
  handled later.

Move the description of the media bus format restriction TODO item to
the driver, drop the other TODO items, and move the driver out of
staging.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Rui Miguel Silva <rmfrfs@gmail.com>
Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: imx: Decouple imx8mq-mipi-csi2 from imx7-media-csi
Laurent Pinchart [Tue, 10 May 2022 15:46:33 +0000 (16:46 +0100)]
media: imx: Decouple imx8mq-mipi-csi2 from imx7-media-csi

The imx8mq-mipi-csi2 driver targets SoCs that also run the
imx7-media-csi driver, but they are distinct. Decouple them in Kconfig
to prepare for destaging of the imx7-media-csi driver.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: imx: imx7-media-csi: Clear BIT_MIPI_DOUBLE_CMPNT for <16b formats
Laurent Pinchart [Wed, 7 Sep 2022 18:42:16 +0000 (19:42 +0100)]
media: imx: imx7-media-csi: Clear BIT_MIPI_DOUBLE_CMPNT for <16b formats

Commit 9babbbaaeb87 ("media: imx: imx7-media-csi: Use dual sampling for
YUV 1X16") set BIT_MIPI_DOUBLE_CMPNT in the CR18 register for 16-bit YUV
formats in imx7_csi_configure(). The CR18 register is always updated
with read-modify-write cycles, so if a 16-bit YUV format is selected,
the bit will stay set forever, even if the format is changed. Fix it by
clearing the bit at the beginning of the imx7_csi_configure() function.

While at it, swap two of the bits being cleared to match the MSB to LSB
order. This doesn't cause any functional change.

Fixes: 9babbbaaeb87 ("media: imx: imx7-media-csi: Use dual sampling for YUV 1X16")
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Rui Miguel Silva <rmfrfs@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: imx: imx7-media-csi: Rename phys variables to dma_addr
Laurent Pinchart [Tue, 6 Sep 2022 23:04:01 +0000 (00:04 +0100)]
media: imx: imx7-media-csi: Rename phys variables to dma_addr

All the phys variables and structure fields store a DMA address, not a
physical address. Even if the two are effectively identical on all
platforms where this driver is used due to the lack of IOMMU, rename the
variables to dma_addr to make their usage clearer.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Rui Miguel Silva <rmfrfs@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: imx: imx7-media-csi: Move variable to loop scope
Laurent Pinchart [Tue, 6 Sep 2022 23:04:01 +0000 (00:04 +0100)]
media: imx: imx7-media-csi: Move variable to loop scope

The phys variable is only used as a local loop variable in
imx7_csi_setup_vb2_buf(), with each entry in the array being used in the
corresponding iteration of the loop only. Move it to loop scope,
simplifying the array to a single variable.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Rui Miguel Silva <rmfrfs@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: imx: imx7-media-csi: Add support for fast-tracking queued buffers
Paul Elder [Wed, 7 Sep 2022 11:47:37 +0000 (12:47 +0100)]
media: imx: imx7-media-csi: Add support for fast-tracking queued buffers

The CSI hardware compatible with this driver handles buffers using a
ping-pong mechanism with two sets of destination addresses. Normally,
when an interrupt comes in to signal the completion of one buffer, say
FB1, it assigns the next buffer in the queue to the next FB1, and the
hardware starts to capture into FB2 in the meantime.

In a buffer underrun situation, in the above example without loss of
generality, if a new buffer is queued before the interrupt for FB1 comes
in, we can program the buffer into FB2 (which is programmed with a dummy
buffer, as there is a buffer underrun).

This of course races with the interrupt that signals FB1 completion, as
once that interrupt comes in, we are no longer guaranteed that the
programming of FB2 was in time and must assume it was too late. This
race is resolved partly by locking the programming of FB2. If it came
after the interrupt for FB1, then the variable that is used to determine
which FB to program would have been swapped by the interrupt handler.

This alone isn't sufficient, however, because the interrupt could still
be generated (thus the hardware starts capturing into the other fb)
while the fast-tracking routine has the irq lock. Thus, after
programming the fb register to fast-track the buffer, the isr also must
be checked to confirm that an interrupt didn't come in the meantime. If
it has, we must assume that programming the register for the
fast-tracked buffer was not in time, and queue the buffer normally.

Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Acked-by: Rui Miguel Silva <rmfrfs@gmail.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: imx: imx7-media-csi: Remove a useless include
Christophe JAILLET [Sat, 12 Nov 2022 21:24:34 +0000 (21:24 +0000)]
media: imx: imx7-media-csi: Remove a useless include

<linux/gcd.h> is not needed for this driver. Remove the corresponding
 #include.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agoMerge tag 'br-v6.2f' of git://linuxtv.org/hverkuil/media_tree into media_stage
Mauro Carvalho Chehab [Fri, 25 Nov 2022 08:29:39 +0000 (08:29 +0000)]
Merge tag 'br-v6.2f' of git://linuxtv.org/hverkuil/media_tree into media_stage

Tag branch

* tag 'br-v6.2f' of git://linuxtv.org/hverkuil/media_tree: (26 commits)
  media: atmel: atmel-isc: move to staging
  media: microchip: microchip-isc: move media_pipeline_* to (un)prepare cb
  media: microchip: microchip-isc: implement media controller
  media: microchip: microchip-isc: prepare for media controller support
  media: microchip: add ISC driver as Microchip ISC
  media: atmel: move microchip_csi2dc to dedicated microchip platform
  vb2/au0828: move the v4l_vb2q_enable_media_source to the au0828 driver
  vb2: add (un)prepare_streaming queue ops
  media: admin-guide: cec.rst
  staging: media: sunxi: cedrus: make vb2_ops struct definition const
  media: amphion: Fix error handling in vpu_driver_init()
  media: platform: exynos4-is: Fix error handling in fimc_md_init()
  media: mtk-jpegdec: add missing destroy_workqueue()
  media: aspeed: Use v4l2_dbg to replace v4l2_warn to avoid log spam
  media: solo6x10: fix possible memory leak in solo_sysfs_init()
  media: cedrus: Relax HEVC SPS restrictions
  media: cedrus: h265: Support decoding 10-bit frames
  media: cedrus: Adjust buffer size based on codec
  media: s5p-mfc: Optimisation of code to remove error message
  media: s5p-mfc:fix usage of Block comment alignment
  ...

Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: atomisp: Make bds_factors_list be type of struct u32_fract
Andy Shevchenko [Fri, 18 Nov 2022 18:43:09 +0000 (18:43 +0000)]
media: atomisp: Make bds_factors_list be type of struct u32_fract

The list contains the Bayer scale index, and rational fraction of it.
The struct u32_fract is suitable type to hold that. Convert the driver
to use latter instead of former.

Signed-off-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>
19 months agomedia: atomisp_ov2680: Fix 1280x720 -> 1296x736 resolution
Hans de Goede [Sun, 23 Oct 2022 23:05:08 +0000 (00:05 +0100)]
media: atomisp_ov2680: Fix 1280x720 -> 1296x736 resolution

The ov2680_720p_30fps register init list used for the 1296x736 resolution
sets the hsize register to 1296 and the vsize register to 736.

This is actually the right thing to do when combined with the atomISP2
because the ISP requires 16 bytes padding leaving userspace to see
1280x720.

But the resolution list entries for this was incorrectly reporting
the resolution being send to the ISP as already being 1280x720,
leaving usespace to see 1274x704 as resolution.

Worse then userspace seeing a weird resolution selecting the
1280x720 sensor resolution (which in reality is sending 1296x736)
to the ISP was causing the ISP to hang on Cherry Trail based tablets
(Bay Trail works fine for some reason).

This commit also adds a bunch of comments annotating what
the various register writes the init lists are doing.

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>
19 months agomedia: atomisp: Remove atomisp_ospm_dphy_up()/_down() functions
Hans de Goede [Mon, 14 Nov 2022 16:07:45 +0000 (16:07 +0000)]
media: atomisp: Remove atomisp_ospm_dphy_up()/_down() functions

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>
19 months agomedia: atomisp: Remove atomisp_ospm_dphy_down() call from probe error path
Hans de Goede [Mon, 14 Nov 2022 16:00:51 +0000 (16:00 +0000)]
media: atomisp: Remove atomisp_ospm_dphy_down() call from probe error path

The only thing which atomisp_ospm_dphy_down() does is disable the CSI
pins, but if we failed to probe the ISP then these will never have
been enabled (because the ISP never started streaming).

So the atomisp_ospm_dphy_down() call in the probe error path is
unnecessary, remove it.

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>
19 months agomedia: atomisp: Move calling of css_[un]init() to power_on()/_off()
Hans de Goede [Mon, 14 Nov 2022 15:55:38 +0000 (15:55 +0000)]
media: atomisp: Move calling of css_[un]init() to power_on()/_off()

atomisp_css_init() is always called after calling atomisp_power_on()
either directly or through getting a runtime-pm reference.

Likewise atomisp_css_uninit() is always called after calling
atomisp_power_off().

Move the call site of these 2 functions to inside atomisp_power_on() /
atomisp_power_off() to make this more explicit.

Note this makes atomisp_reset() also set isp_fatal_error on
atomisp_power_on() errors, where as before it only did this on
atomisp_css_init() errors. This behavior change is for the better,
since power-on failing is pretty fatal too.

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>
19 months agomedia: atomisp: Remove duplication between runtime-pm and normal-pm code
Hans de Goede [Mon, 14 Nov 2022 15:44:49 +0000 (15:44 +0000)]
media: atomisp: Remove duplication between runtime-pm and normal-pm code

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>
19 months agomedia: atomisp: Remove sw_contex.power_state checks
Hans de Goede [Mon, 14 Nov 2022 15:32:41 +0000 (15:32 +0000)]
media: atomisp: Remove sw_contex.power_state checks

Remove the unnecessary sw_contex.power_state checks:

1. atomisp_freq_scaling() and atomisp_stop_streaming() only run when the
   ISP is powered up through runtime-pm, so the checks are not necessary

2. atomisp_mrfld_pre_power_down() and atomisp_runtime_resume() are only
   called through the driver-core pm handling code which already
   guarantees they are not called when already powered down / up.

3. atomisp_isr() also checks isp->css_initialized which only gets set
   by atomisp_css_init() which runs *after* powering up the ISP and which
   gets cleared by atomisp_css_uninit() *before* powering down the ISP.

So all the checks are unnecessary, remove them as well as the
sw_contex.power_state field itself.

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>
19 months agomedia: atomisp: Remove atomisp_css_suspend()/_resume()
Hans de Goede [Mon, 14 Nov 2022 15:09:52 +0000 (15:09 +0000)]
media: atomisp: Remove atomisp_css_suspend()/_resume()

atomisp_css_suspend() is a 1:1 copy of atomisp_css_uninit() and
atomisp_css_resume() is a 1:1 copy of atomisp_css_init().

Remove the 2 copies and have their one caller just call
atomisp_css_uninit() / atomisp_css_init() instead.

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>
19 months agomedia: atomisp: Remove clearing of config from atomisp_css_uninit()
Hans de Goede [Mon, 14 Nov 2022 15:07:18 +0000 (15:07 +0000)]
media: atomisp: Remove clearing of config from atomisp_css_uninit()

atomisp_css_uninit() only runs when all streams are stopped and
atomisp_css_stop() already clears the config, so the clearing
of the config can be dropped from atomisp_css_uninit().

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>
19 months agomedia: atomisp: Remove atomisp_mrfld_power_down()/_up()
Hans de Goede [Fri, 11 Nov 2022 20:47:40 +0000 (20:47 +0000)]
media: atomisp: Remove atomisp_mrfld_power_down()/_up()

atomisp_mrfld_power_down()/_up() are unnecessary wrappers around
atomisp_mrfld_power() remove them and just call atomisp_mrfld_power()
directly.

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>
19 months agomedia: atomisp: Remove double atomisp_mrfld_power_down()/_up() calls from atomisp_reset()
Hans de Goede [Fri, 11 Nov 2022 17:48:14 +0000 (17:48 +0000)]
media: atomisp: Remove double atomisp_mrfld_power_down()/_up() calls from 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>
19 months agomedia: atomisp: Drop userptr support from hmm
Hans de Goede [Tue, 8 Nov 2022 19:34:56 +0000 (19:34 +0000)]
media: atomisp: Drop userptr support from hmm

After the conversion to videobuf2 userptr support is no longer needed,
drop it.

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>
19 months agomedia: atomisp: Remove unused ia_css_frame_*() functions
Hans de Goede [Tue, 8 Nov 2022 17:09:44 +0000 (17:09 +0000)]
media: atomisp: Remove unused ia_css_frame_*() functions

After the conversion to videobuf2 a bunch of ia_css_frame_*()
functions are unused, 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>
19 months agomedia: atomisp: Remove atomisp_css_yuvpp_configure_viewfinder() function
Hans de Goede [Tue, 25 Oct 2022 14:05:20 +0000 (15:05 +0100)]
media: atomisp: Remove atomisp_css_yuvpp_configure_viewfinder() function

Remove atomisp_css_yuvpp_configure_viewfinder(), it is not used anywhere.

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>
19 months agomedia: atomisp: Remove 2 unused accelerator mode related functions
Hans de Goede [Tue, 25 Oct 2022 14:02:15 +0000 (15:02 +0100)]
media: atomisp: Remove 2 unused accelerator mode related functions

Remove ia_css_pipe_get_acc_stage_desc() and sh_css_flush(),
after removing the accelerator /dev/video# node and related ioctls
these are no longer used.

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>
19 months agomedia: atomisp: Fix deadlock when the /dev/video# node is closed while still streaming
Hans de Goede [Sun, 23 Oct 2022 20:22:14 +0000 (21:22 +0100)]
media: atomisp: Fix deadlock when the /dev/video# node is closed while still streaming

atomisp_release() was taking pipe->vb_queue_mutex + isp->mutex at the
same time. But if the /dev/video# node is closed while still streaming
then vb2_queue_release() will call atomisp_stop_streaming() which takes
isp->mutex itself, leading to a deadlock.

To fix this only take isp->mutex after cleaning up the v4l2_fh /
the vb2_queue. While at it switch to vb2_fop_release() which will take
pipe->vb_queue_mutex for us, which also resolves a FIXME comment.

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>
19 months agomedia: atomisp: Log an error on failing to alloc private-mem
Hans de Goede [Sun, 23 Oct 2022 20:16:09 +0000 (21:16 +0100)]
media: atomisp: Log an error on failing to alloc private-mem

I managed to trigger an atomisp_css_start() error by pushing my test
system towards an OOM situation, this resulted in the following errors:

atomisp-isp2 0000:00:03.0: alloc pages err...
atomisp-isp2 0000:00:03.0: hmm_bo_alloc_pages failed.
atomisp-isp2 0000:00:03.0: stream[0] start error.

But it is not entirely clear what the root cause of
the "alloc pages err..." error is. I suspect the root cause is
alloc_pages_bulk_array() failing. Add a log message to make
the root cause more clear if this is hit again.

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>
19 months agomedia: atomisp: Flush queue on atomisp_css_start() error
Hans de Goede [Sun, 23 Oct 2022 20:13:01 +0000 (21:13 +0100)]
media: atomisp: Flush queue on atomisp_css_start() error

I managed to trigger an atomisp_css_start() error by pushing my test system
towards an OOM situation, this triggered the following WARN_ON() in
__vb2_queue_cancel() in videobuf2-core.c:

        /*
         * If you see this warning, then the driver isn't cleaning up properly
         * after a failed start_streaming(). See the start_streaming()
         * documentation in videobuf2-core.h for more information how buffers
         * should be returned to vb2 in start_streaming().
         */
        if (WARN_ON(atomic_read(&q->owned_by_drv_count))) {

Fix this by calling atomisp_flush_video_pipe() to return any queued buffers
back to the videobuf2-core on an atomisp_css_start() error.

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>
19 months agomedia: atomisp: Remove unused QOS defines / structure member
Hans de Goede [Sat, 22 Oct 2022 14:54:37 +0000 (15:54 +0100)]
media: atomisp: Remove unused QOS defines / structure member

With the accel code gone this is unused, remove it.

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>
19 months agomedia: atomisp: Remove accelerator pipe creation code
Hans de Goede [Thu, 20 Oct 2022 21:28:51 +0000 (22:28 +0100)]
media: atomisp: Remove accelerator pipe creation code

The ATOMISP_ACC_* custom ioctls and the ACC device node have been removed
in commit a5c17adbadcb ("media: atomisp: Remove the ACC device node").

This means that pipe_configs[pipe_id].acc_extension now never gets set
which causes atomisp_compat_css20.c: __create_pipe() to always skip
creation of pipes with a pipe_id of IA_CSS_PIPE_ID_ACC / a mode of
IA_CSS_PIPE_MODE_ACC.

This allows removing of the acc_pipe creation / handling code
from mainly sh_css.c and a bunch of other places.

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>
19 months agomedia: atomisp: Silence: 'atomisp_q_one_s3a_buffer: drop one s3a stat which has exp_i...
Hans de Goede [Thu, 20 Oct 2022 19:56:17 +0000 (20:56 +0100)]
media: atomisp: Silence: 'atomisp_q_one_s3a_buffer: drop one s3a stat which has exp_id xx' log messages

Standard v4l2 userspace apps do not consume the s3a statistics block
data. Until we have a userspace consumer for this (libcamera), which
might also involve changing the API for this, lower the log level
of these messages to dev_dbg() to avoid them filling up the logs.

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>
19 months agomedia: atomisp: gc0310: Power on sensor from set_fmt() callback
Hans de Goede [Sun, 16 Oct 2022 19:49:43 +0000 (20:49 +0100)]
media: atomisp: gc0310: Power on sensor from set_fmt() callback

Depending on which order userspace makes various v4l2 calls, the sensor
might still be powered down when set_fmt is called.

What should really happen here is delay the writing of the mode-related
registers till streaming is started, but for now use the same quick fix
as the atomisp_ov2680 code and call power_up() from set_fmt() in
combination with keeping track of the power-state to avoid doing the
power-up sequence twice.

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>
19 months agomedia: atomisp: Remove __atomisp_get_pipe() helper
Hans de Goede [Sun, 16 Oct 2022 18:28:33 +0000 (19:28 +0100)]
media: atomisp: Remove __atomisp_get_pipe() helper

Remove the complicated __atomisp_get_pipe() helper, atomisp_buf_done()
only needs the pipe pointer in cases where it has a frame, so we can
simply get the pipe from the frame using the vb_to_pipe() helper.

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>
19 months agomedia: atomisp: Make atomisp_g_fmt_cap() default to YUV420
Hans de Goede [Sun, 16 Oct 2022 18:00:07 +0000 (19:00 +0100)]
media: atomisp: Make atomisp_g_fmt_cap() default to YUV420

Make atomisp_g_fmt_cap() default to YUV420 so that it matches with what
atomisp_try_fmt_cap() and atomisp_queue_setup() do when they need to
pick a default pixelformat.

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>
19 months agomedia: atomisp: Fix atomisp_try_fmt_cap() always returning YUV420 pixelformat
Hans de Goede [Sun, 16 Oct 2022 18:02:54 +0000 (19:02 +0100)]
media: atomisp: Fix atomisp_try_fmt_cap() always returning YUV420 pixelformat

The atomisp_try_fmt() call in atomisp_try_fmt_cap() replaces
the pixelformat passed by userspace with the sensors native pixelformat.

Which always gets replaced by V4L2_PIX_FMT_YUV420 by atomisp_adjust_fmt()
because raw sensor formats are not supported.

This needs to be fixed so that userspace which does a try_fmt call before
s_fmt does not end up always getting YUV420 even if it passed something
else into the try_fmt call.

To fix this restore the userspace requested pixelformat before
the atomisp_adjust_fmt() call. atomisp_adjust_fmt() will replace this
with V4L2_PIX_FMT_YUV420 in case an unsupported format is requested.

Note this relies on the "media: atomisp: Refactor atomisp_adjust_fmt()"
change, before that atomisp_adjust_fmt() would return -EINVAL for
unsupported pixelformats.

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>
19 months agomedia: atomisp: Refactor atomisp_adjust_fmt()
Hans de Goede [Sun, 16 Oct 2022 09:35:03 +0000 (10:35 +0100)]
media: atomisp: Refactor atomisp_adjust_fmt()

Refactor atomisp_adjust_fmt():

1. The block starting at "format_bridge = atomisp_get_format_bridge(...)"
and ending with "if (field == V4L2_FIELD_ANY) field = V4L2_FIELD_NONE;"
is duplicated. With only the second block:
a) Properly checking that format_bridge is not NULL; amd
b) Having the special handling for IA_CSS_FRAME_FORMAT_RAW

Remove the first block.

2. On a NULL return from atomisp_get_format_bridge(f->fmt.pix.pixelformat)
fall back to V4L2_PIX_FMT_YUV420 just like in the IA_CSS_FRAME_FORMAT_RAW
case. atomisp_adjust_fmt() is used in VIDIOC_TRY_FMT handling and that
should jusy pick a fmt rather then returning -EINVAL.

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>
19 months agomedia: atomisp: Fix VIDIOC_REQBUFS failing when VIDIOC_S_FMT has not been called yet
Hans de Goede [Sun, 16 Oct 2022 09:43:11 +0000 (10:43 +0100)]
media: atomisp: Fix VIDIOC_REQBUFS failing when VIDIOC_S_FMT has not been called yet

camorama calls VIDIOC_REQBUFS(V4L2_MEMORY_MMAP) to test if MMAP support
works (this was added specifically because of the previously broken
MMAP support in atomisp).

Currently this fails because atomisp_get_css_frame_info() fails in this
case. Although it is weird to call VIDIOC_REQBUFS before VIDIOC_S_FMT,
it is allowed to do this.

Fix this not working by doing a S_FMT to V4L2_PIX_FMT_YUV420 + the highest
supported resolution.

Note this will cause camorama to use mmap mode, which means it will also
use libv4l2 to do format conversion. libv4l2 will pick V4L2_PIX_FMT_RGB565
as input format and this will lead to a garbled video display. This is
a libv4lconvert bug, the RGB565 -> RGB24 path in libv4lconvert assumes
that stride == width which is not true on the atomisp.

I've already send out a libv4lconvert fix for this. Also this can be worked
around by passing --dont-use-libv4l2 as argument to camorama.

Link: https://git.linuxtv.org/v4l-utils.git/commit/?id=aecfcfccfc2f78d7531456ffa5465666c6bc641e
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>
19 months agomedia: atomisp: Make it possible to call atomisp_set_fmt() without a file handle
Hans de Goede [Sat, 15 Oct 2022 17:21:42 +0000 (18:21 +0100)]
media: atomisp: Make it possible to call atomisp_set_fmt() without a file handle

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>
19 months agomedia: atomisp: Convert to videobuf2
Hans de Goede [Sat, 8 Oct 2022 16:39:32 +0000 (17:39 +0100)]
media: atomisp: Convert to videobuf2

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>
19 months agomedia: atomisp: Add ia_css_frame_get_info() helper
Hans de Goede [Fri, 14 Oct 2022 16:13:41 +0000 (17:13 +0100)]
media: atomisp: Add ia_css_frame_get_info() helper

Several places rely on the [frame_]info member being the first member of
struct ia_css_frame, so that &frame->info will yield NULL when frame is
NULL (some places already explicitly check for a NULL frame pointer but
not nearly all).

For videobuf2 support the vb2_v4l2_buffer struct needs to be embedded
in the frame struct and it needs to be the first member. Breaking the
assumption that &frame->info will yield NULL when frame is NULL.

Add a ia_css_frame_get_info() helper to return either the ia_css_frame_info
struct embedded in the frame, or NULL when the frame pointer is NULL and
use this in places where a ia_css_frame_info ptr or NULL is expected.

To make sure that we catch all uses of the info field this patch also
renames the info field to frame_info.

This is a preparation patch for converting the driver to videobuf2.

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>
19 months agomedia: atomisp: Use new atomisp_flush_video_pipe() helper in atomisp_streamoff()
Hans de Goede [Sat, 8 Oct 2022 15:27:26 +0000 (16:27 +0100)]
media: atomisp: Use new atomisp_flush_video_pipe() helper in atomisp_streamoff()

Use the new atomisp_flush_video_pipe() helper instead of open-coding it.

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>
19 months agomedia: atomisp: Add an index helper variable to atomisp_buf_done()
Hans de Goede [Sat, 8 Oct 2022 15:20:09 +0000 (16:20 +0100)]
media: atomisp: Add an index helper variable to atomisp_buf_done()

With the videobuf2 conversion accessing the index of a frame is going
to become more involved and writing this out each time is undesired.

Add an 'i' helper variable for the index and assing this once as
preparation for the videobuf2 conversion.

This also makes use of the new rules wrt max line-lengths in the kernel
to avoid breaking up a bunch of lines. Not breaking these lines results
in better readable code (IMHO).

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>
19 months agomedia: atomisp: Also track buffers in a list when submitted to the ISP
Hans de Goede [Sat, 8 Oct 2022 15:08:08 +0000 (16:08 +0100)]
media: atomisp: Also track buffers in a list when submitted to the ISP

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>
19 months agomedia: atomisp: Remove unused atomisp_buffers_queued[_pipe] functions
Hans de Goede [Fri, 7 Oct 2022 18:27:09 +0000 (19:27 +0100)]
media: atomisp: Remove unused atomisp_buffers_queued[_pipe] functions

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>
19 months agomedia: atomisp: On streamoff wait for buffers owned by the CSS to be given back
Hans de Goede [Fri, 30 Sep 2022 19:05:44 +0000 (20:05 +0100)]
media: atomisp: On streamoff wait for buffers owned by the CSS to be given back

There is no guarantee that when we stop the pipeline all buffers owned
by the CSS are cleanly returned to the videobuf queue.

This is a problem with videobuf2 which will complain loudly when not
all buffers have been returned after the streamoff() queue op has
returned.

And this also allows removing a WARN() in the continuous mode path.

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>
19 months agomedia: atomisp: Make atomisp_q_video_buffers_to_css() static
Hans de Goede [Fri, 30 Sep 2022 15:54:27 +0000 (16:54 +0100)]
media: atomisp: Make atomisp_q_video_buffers_to_css() static

atomisp_q_video_buffers_to_css() is only used insidd atomisp_fops.c, make
it static.

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>
19 months agomedia: atomisp: Add ia_css_frame_init_from_info() function
Hans de Goede [Mon, 26 Sep 2022 19:43:54 +0000 (20:43 +0100)]
media: atomisp: Add ia_css_frame_init_from_info() function

Add a function to initialize (rather then alloc/create) a
ia_css_frame struct based on an ia_css_frame_info struct.

This is a preparation patch for adding videobuf2 support.

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>
19 months agomedia: atomisp: Add hmm_create_from_vmalloc_buf() function
Hans de Goede [Mon, 26 Sep 2022 17:45:46 +0000 (18:45 +0100)]
media: atomisp: Add hmm_create_from_vmalloc_buf() function

Add a new hmm creating function to create a vmm object from a vmalloc-ed
kernel buffer. This is a preparation patch for adding videobuf2 (and
working MMAP mode) support.

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>
19 months agomedia: imon: fix a race condition in send_packet()
Gautam Menghani [Wed, 19 Oct 2022 05:02:14 +0000 (06:02 +0100)]
media: imon: fix a race condition in send_packet()

The function send_packet() has a race condition as follows:

func send_packet()
{
    // do work
    call usb_submit_urb()
    mutex_unlock()
    wait_for_event_interruptible()  <-- lock gone
    mutex_lock()
}

func vfd_write()
{
    mutex_lock()
    call send_packet()  <- prev call is not completed
    mutex_unlock()
}

When the mutex is unlocked and the function send_packet() waits for the
call to complete, vfd_write() can start another call, which leads to the
"URB submitted while active" warning in usb_submit_urb().
Fix this by removing the mutex_unlock() call in send_packet() and using
mutex_lock_interruptible().

Link: https://syzkaller.appspot.com/bug?id=e378e6a51fbe6c5cc43e34f131cc9a315ef0337e
Fixes: 21677cfc562a ("V4L/DVB: ir-core: add imon driver")
Reported-by: syzbot+0c3cb6dc05fbbdc3ad66@syzkaller.appspotmail.com
Signed-off-by: Gautam Menghani <gautammenghani201@gmail.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: ir-spi: silence no spi_device_id warnings
Wei Yongjun [Thu, 22 Sep 2022 08:37:03 +0000 (09:37 +0100)]
media: ir-spi: silence no spi_device_id warnings

SPI devices use the spi_device_id for module autoloading even on
systems using device tree, after commit 5fa6863ba692 ("spi: Check
we have a spi_device_id for each DT compatible"), kernel warns as
follows since the spi_device_id is missing:

SPI driver ir-spi has no spi_device_id for ir-spi-led

Add spi_device_id entries to silence the warning, and ensure driver
module autoloading works.

Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agoMerge tag 'br-v6.2g' of git://linuxtv.org/hverkuil/media_tree into media_stage
Mauro Carvalho Chehab [Fri, 25 Nov 2022 07:55:03 +0000 (07:55 +0000)]
Merge tag 'br-v6.2g' of git://linuxtv.org/hverkuil/media_tree into media_stage

Despite being big, the changes here are trivial: they just change the
probing method for I2C. So, apply it as a merge.

Tag branch

* tag 'br-v6.2g' of git://linuxtv.org/hverkuil/media_tree: (96 commits)
  media: v4l2-core/tuner-core: Convert to i2c's .probe_new()
  media: usb: go7007: s2250-board: Convert to i2c's .probe_new()
  media: tuners/tua9001: Convert to i2c's .probe_new()
  media: tuners/tda18250: Convert to i2c's .probe_new()
  media: tuners/tda18212: Convert to i2c's .probe_new()
  media: tuners/qm1d1c0042: Convert to i2c's .probe_new()
  media: tuners/qm1d1b0004: Convert to i2c's .probe_new()
  media: tuners/mxl301rf: Convert to i2c's .probe_new()
  media: tuners/mt2060: Convert to i2c's .probe_new()
  media: tuners/m88rs6000t: Convert to i2c's .probe_new()
  media: tuners/fc2580: Convert to i2c's .probe_new()
  media: tuners/e4000: Convert to i2c's .probe_new()
  media: vidtv: Convert to i2c's .probe_new()
  media: radio/tef6862: Convert to i2c's .probe_new()
  media: radio/saa7706h: Convert to i2c's .probe_new()
  media: radio/radio-tea5764: Convert to i2c's .probe_new()
  media: i2c/wm8775: Convert to i2c's .probe_new()
  media: i2c/wm8739: Convert to i2c's .probe_new()
  media: i2c/vs6624: Convert to i2c's .probe_new()
  media: i2c/vpx3220: Convert to i2c's .probe_new()
  ...

Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: atmel: atmel-isc: move to staging
Eugen Hristev [Mon, 7 Nov 2022 14:18:18 +0000 (14:18 +0000)]
media: atmel: atmel-isc: move to staging

The Atmel ISC driver is not compliant with media controller specification.
In order to evolve this driver, it has to move to media controller, to
support enhanced features and future products which embed it.
The move to media controller involves several changes which are
not backwards compatible with the current usability of the driver.

The best example is the way the format is propagated from the top video
driver /dev/videoX down to the sensor.

In a simple configuration sensor ==> isc , the isc just calls subdev s_fmt
and controls the sensor directly. This is achieved by having a lot of code
inside the driver that will query the subdev at probe time and make a list
of formats which are usable.
Basically the user has nothing to configure, as the isc will handle
everything at the top level. This is an easy way to capture, but also comes
with the drawback of lack of flexibility.
In a more complicated pipeline
sensor ==> controller 1 ==> controller 2 ==> isc
this will not be achievable, as controller 1 and controller 2 might be
media-controller configurable, and will not propagate the formats down to
the sensor.

After discussions with the media maintainers, the decision is to move
Atmel ISC to staging as-is, to keep the Kconfig symbols and the users
to the driver in staging. Thus, all the existing users of the non
media-controller paradigm will continue to be happy and use the old config
way.

The new driver was added in the media subsystem with a different
symbol, with the conversion to media controller done, and new users
of the driver will be able to use all the new features.

This patch is merely a file move to staging, not affecting any of the
users.

The exported symbols had to be renamed to atmel_* to avoid duplication with
the new Microchip ISC driver.

Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: microchip: microchip-isc: move media_pipeline_* to (un)prepare cb
Eugen Hristev [Mon, 7 Nov 2022 14:18:17 +0000 (14:18 +0000)]
media: microchip: microchip-isc: move media_pipeline_* to (un)prepare cb

Move the media_pipeline_start/stop calls from start/stop streaming to
the new prepare_streaming and unprepare_streaming callbacks.

Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: microchip: microchip-isc: implement media controller
Eugen Hristev [Mon, 7 Nov 2022 14:18:16 +0000 (14:18 +0000)]
media: microchip: microchip-isc: implement media controller

As a top MC video driver, the microchip-isc should not propagate the format
to the subdevice, it should rather check at start_streaming() time if the
subdev is properly configured with a compatible format.
Removed the whole format finding logic, and reworked the format
verification at start_streaming time, such that the ISC will return an
error if the subdevice is not properly configured.
To achieve this, media_pipeline_start is called and a link_validate
callback is created to check the formats.
With this being done, the module parameter 'sensor_preferred' makes no
sense anymore. The ISC should not decide which format the sensor is using.
The ISC should only cope with the situation and inform userspace if the
streaming is possible in the current configuration.
The redesign of the format propagation has also risen the question of the
enumfmt callback. If enumfmt is called with an mbus_code, the enumfmt
handler should only return the formats that are supported for this
mbus_code. Otherwise, the enumfmt will report all the formats that the ISC
could output.
With this rework, the dynamic list of user formats is removed. It makes no
more sense to identify at complete time which formats the sensor could
emit, and add those into a separate dynamic list.
The ISC will start with a simple preconfigured default format, and at
link validate time, decide whether it can use the format that is
configured on the sink or not.
>From now on, the driver also advertises the IO_MC capability.

Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: microchip: microchip-isc: prepare for media controller support
Eugen Hristev [Mon, 7 Nov 2022 14:18:15 +0000 (14:18 +0000)]
media: microchip: microchip-isc: prepare for media controller support

Prepare the support for media-controller.
This means that the capabilities of the driver have changed and now it's
capable of media controller operations.
The driver will register its media device, and add the video entity to this
media device. The subdevices are registered to the same media device.
The ISC will have a base entity which is auto-detected as
microchip_isc_base.
It will also register a subdevice that allows cropping of the incoming
frame to the maximum frame size supported by the ISC.
The ISC will create a link between the subdevice that is asynchronously
registered and the microchip_isc_scaler entity.
Then, the microchip_isc_scaler and microchip_isc_base are connected
through another link.
This patch does not change the previous capability of the driver, the
fact that the format is still being propagated from the top video node
down to the sensor.

Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Reviewed-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: microchip: add ISC driver as Microchip ISC
Eugen Hristev [Mon, 7 Nov 2022 14:18:14 +0000 (14:18 +0000)]
media: microchip: add ISC driver as Microchip ISC

The Atmel ISC driver will be moved to staging to support old users that
are not using the media controller paradigm.
The ISC driver was converted to media controller in the public patch
series:

https://lore.kernel.org/lkml/20220503095127.48710-1-eugen.hristev@microchip.com/T/#m2c320fa8153c01379a1c35b1d90a00903949513a

However the conversion cannot be done directly as it would affect existing
users by breaking the old way of configuration for sama5d2 platforms.

After discussions with the media maintainers, the decision is to move
Atmel ISC to staging as-is, to keep the Kconfig symbols and the users
to the driver in staging. Thus, all the existing users of the non
media-controller paradigm will continue to be happy and use the old config
way.

The converted driver would support both sama5d2 and sama7g5 platforms with
media controller paradigm, but it requires userspace configuration of the
pipeline for all the pipeline modules.

In a simple configuration sensor ==> isc , the old isc driver used to call
subdev s_fmt and control the sensor directly.
This was achieved by having a lot of code inside the driver that was
querying the subdev at probe time and made a list of formats which are
usable.
Basically the user had nothing to configure, as the isc would handle
everything at the top level. This was an easy way to capture, but also came
with the drawback of lack of flexibility.
In a more complicated pipeline
 sensor ==> controller 1 ==> controller 2 ==> isc
this would not be achievable, as controller 1 and controller 2 might be
media-controller configurable, and will not propagate the formats down to
the sensor.
The new driver Microchip ISC would solve all these problems and exposes pads
entities and links to userspace.
For the ease of tracking, the patches that convert to media controller come
on top of this patch that simply readds the driver to the new location under
the new Kconfig symbols.

To differentiate between the old driver and the new driver, I have renamed
the new driver to Microchip ISC, renaming the Kconfig symbols as well, and
all the mentions inside the driver.
The only thing that remains common is the file
include/linux/atmel-isc-media.h which is the ABI for the v4l2 custom
controls that the ISC exposes.
This file is used by both driver, so I kept it as-is.
To further avoid confusion all files have been renamed and all functions
named isc_* as well.
The exported symbols have been renamed with added microchip_ prefix, to
avoid symbol duplication with the old driver, and to avoid confusion.
Other than that, I have fixed small checkpatch issues when readding the
driver.

Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: atmel: move microchip_csi2dc to dedicated microchip platform
Eugen Hristev [Mon, 7 Nov 2022 14:18:13 +0000 (14:18 +0000)]
media: atmel: move microchip_csi2dc to dedicated microchip platform

The Atmel ISC will be moved to staging thus the atmel platform will only
have the ISI driver.
The new media-controller converted ISC driver will be placed inside a
dedicated microchip platform directory.
It is then natural to have the microchip-csi2dc moved to this new platform
directory.
The next step is to add the Microchip ISC driver to the new platform
directory and reside together with the Microchip CSI2DC driver.

Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: vb2/au0828: move the v4l_vb2q_enable_media_source to the au0828 driver
Hans Verkuil [Wed, 22 Jun 2022 09:31:45 +0000 (10:31 +0100)]
media: vb2/au0828: move the v4l_vb2q_enable_media_source to the au0828 driver

With the new prepare_streaming op it is possible to move the ugly
v4l_vb2q_enable_media_source() call in vb2_core_streamon to the
driver. It was called incorrectly in vb2 as well: it was only
called if sufficient buffers were queued at STREAMON time, but not
if more buffers were queued later. This was not an issue with the
au0828 driver since it never set min_buffers_needed.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: vb2: add (un)prepare_streaming queue ops
Hans Verkuil [Wed, 22 Jun 2022 09:31:44 +0000 (10:31 +0100)]
media: vb2: add (un)prepare_streaming queue ops

When userspace called VIDIOC_STREAMON, then you want to claim any streaming
resources needed and validate the video pipeline. Waiting for
start_streaming to be called is too late, since that can be postponed
until the required minimum of buffers is queued.

So add a prepare_streaming op (optional) that can be used for that
purpose, and a matching unprepare_streaming op (optional) that can
release any claimed resources. The unprepare_streaming op is called
when VIDIOC_STREAMOFF is called and q->streaming is 1, or when the
filehandle is closed while q->streaming is 1.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: admin-guide: cec.rst
Hans Verkuil [Tue, 27 Sep 2022 08:43:11 +0000 (09:43 +0100)]
media: admin-guide: cec.rst

Document administration details about CEC devices. This was formerly
documented in a cec-status.txt I kept on my website, but this really
belongs here as an admin guide.

Updated the original cec-status.txt, and converted it to .rst.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: staging: media: sunxi: cedrus: make vb2_ops struct definition const
Ian Cowan [Sun, 13 Nov 2022 01:47:03 +0000 (01:47 +0000)]
media: staging: media: sunxi: cedrus: make vb2_ops struct definition const

This cleans up a checkstyle warning where the vb2_ops struct definitions
should be const.

Signed-off-by: Ian Cowan <ian@linux.cowan.aero>
Reviewed-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: amphion: Fix error handling in vpu_driver_init()
Yuan Can [Fri, 11 Nov 2022 06:09:26 +0000 (06:09 +0000)]
media: amphion: Fix error handling in vpu_driver_init()

A problem about modprobe amphion-vpu failed is triggered with the
following log given:

 [ 2208.634841] Error: Driver 'amphion-vpu' is already registered, aborting...
 modprobe: ERROR: could not insert 'amphion_vpu': Device or resource busy

The reason is that vpu_driver_init() returns vpu_core_driver_init()
directly without checking its return value, if vpu_core_driver_init()
failed, it returns without unregister amphion_vpu_driver, resulting the
amphion-vpu can never be installed later.
A simple call graph is shown as below:

 vpu_driver_init()
   platform_driver_register() # register amphion_vpu_driver
   vpu_core_driver_init()
     platform_driver_register()
       driver_register()
         bus_add_driver()
           dev = kzalloc(...) # OOM happened
   # return without unregister amphion_vpu_driver

Fix by unregister amphion_vpu_driver when vpu_core_driver_init() returns
error.

Fixes: b50a64fc54af ("media: amphion: add amphion vpu device driver")
Signed-off-by: Yuan Can <yuancan@huawei.com>
Reviewed-by: ming_qian <ming.qian@nxp.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: platform: exynos4-is: Fix error handling in fimc_md_init()
Yuan Can [Fri, 11 Nov 2022 06:08:53 +0000 (06:08 +0000)]
media: platform: exynos4-is: Fix error handling in fimc_md_init()

A problem about modprobe s5p_fimc failed is triggered with the
following log given:

 [  272.075275] Error: Driver 'exynos4-fimc' is already registered, aborting...
 modprobe: ERROR: could not insert 's5p_fimc': Device or resource busy

The reason is that fimc_md_init() returns platform_driver_register()
directly without checking its return value, if platform_driver_register()
failed, it returns without unregister fimc_driver, resulting the
s5p_fimc can never be installed later.
A simple call graph is shown as below:

 fimc_md_init()
   fimc_register_driver() # register fimc_driver
   platform_driver_register()
     platform_driver_register()
       driver_register()
         bus_add_driver()
           dev = kzalloc(...) # OOM happened
   # return without unregister fimc_driver

Fix by unregister fimc_driver when platform_driver_register() returns
error.

Fixes: d3953223b090 ("[media] s5p-fimc: Add the media device driver")
Signed-off-by: Yuan Can <yuancan@huawei.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: mtk-jpegdec: add missing destroy_workqueue()
Yang Yingliang [Thu, 10 Nov 2022 15:05:59 +0000 (15:05 +0000)]
media: mtk-jpegdec: add missing destroy_workqueue()

destroy_workqueue() needs be called to when driver is
unloading, fix it by using devm_add_action_or_reset()
to make workqueuedevice-managed.

Fixes: dedc21500334 ("media: mtk-jpegdec: add jpeg decode worker interface")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: aspeed: Use v4l2_dbg to replace v4l2_warn to avoid log spam
Jammy Huang [Thu, 10 Nov 2022 09:56:11 +0000 (09:56 +0000)]
media: aspeed: Use v4l2_dbg to replace v4l2_warn to avoid log spam

If the host is powered off, there will be many warning log. To avoid the
log spam in this condition, replace v4l2_warn with v4l2_dbg.

Signed-off-by: Jammy Huang <jammy_huang@aspeedtech.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: solo6x10: fix possible memory leak in solo_sysfs_init()
Yang Yingliang [Thu, 10 Nov 2022 08:24:23 +0000 (08:24 +0000)]
media: solo6x10: fix possible memory leak in solo_sysfs_init()

If device_register() returns error in solo_sysfs_init(), the
name allocated by dev_set_name() need be freed. As comment of
device_register() says, it should use put_device() to give up
the reference in the error path. So fix this by calling
put_device(), then the name can be freed in kobject_cleanup().

Fixes: dcae5dacbce5 ("[media] solo6x10: sync to latest code from Bluecherry's git repo")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: cedrus: Relax HEVC SPS restrictions
Jernej Skrabec [Wed, 9 Nov 2022 18:23:09 +0000 (18:23 +0000)]
media: cedrus: Relax HEVC SPS restrictions

Testing reference video TSUNEQBD_A_MAIN10_Technicolor_2 has show that
Cedrus is capable of decoding frames with different chroma and luma bit
depths.

Relax restrictions so only highest depth is checked.

Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: cedrus: h265: Support decoding 10-bit frames
Jernej Skrabec [Wed, 9 Nov 2022 18:23:08 +0000 (18:23 +0000)]
media: cedrus: h265: Support decoding 10-bit frames

10-bit frames needs extra buffer space when 8-bit capture format is
used. Use previously prepared infrastructure to adjust buffer size.

Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: cedrus: Adjust buffer size based on codec
Jernej Skrabec [Wed, 9 Nov 2022 18:23:07 +0000 (18:23 +0000)]
media: cedrus: Adjust buffer size based on codec

In some cases decoding engine needs extra space in capture buffers. This
is the case for decoding 10-bit HEVC frames into 8-bit capture format.
This commit only adds infrastructure for such cases.

Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: s5p-mfc: Optimisation of code to remove error message
Aakarsh Jain [Wed, 9 Nov 2022 03:53:48 +0000 (03:53 +0000)]
media: s5p-mfc: Optimisation of code to remove error message

Already error number provision is present for block failing, while
requesting for DMA consistent memory allocation.
So removing error message line from the block as recommended
by scripts/checkpatch.pl.

Signed-off-by: Aakarsh Jain <aakarsh.jain@samsung.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: s5p-mfc:fix usage of Block comment alignment
Aakarsh Jain [Wed, 9 Nov 2022 03:53:47 +0000 (03:53 +0000)]
media: s5p-mfc:fix usage of Block comment alignment

Fix usage of block comment alignment in drivers/media/platform/samsung/
s5p-mfc/s5p_mfc.c as recommended by scripts/checkpatch.pl.

Signed-off-by: Aakarsh Jain <aakarsh.jain@samsung.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: s5p-mfc: fix usage of symbolic permissions to octal
Aakarsh Jain [Wed, 9 Nov 2022 03:53:46 +0000 (03:53 +0000)]
media: s5p-mfc: fix usage of symbolic permissions to octal

Change symbolic permissions to octal equivalent as recommended
by scripts/checkpatch.pl in drivers/media/platform/samsung/
s5p-mfc/s5p_mfc.c.

Signed-off-by: Aakarsh Jain <aakarsh.jain@samsung.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: vidtv: Fix use-after-free in vidtv_bridge_dvb_init()
Chen Zhongjin [Tue, 8 Nov 2022 07:06:30 +0000 (07:06 +0000)]
media: vidtv: Fix use-after-free in vidtv_bridge_dvb_init()

KASAN reports a use-after-free:
BUG: KASAN: use-after-free in dvb_dmxdev_release+0x4d5/0x5d0 [dvb_core]
Call Trace:
 ...
 dvb_dmxdev_release+0x4d5/0x5d0 [dvb_core]
 vidtv_bridge_probe+0x7bf/0xa40 [dvb_vidtv_bridge]
 platform_probe+0xb6/0x170
 ...
Allocated by task 1238:
 ...
 dvb_register_device+0x1a7/0xa70 [dvb_core]
 dvb_dmxdev_init+0x2af/0x4a0 [dvb_core]
 vidtv_bridge_probe+0x766/0xa40 [dvb_vidtv_bridge]
 ...
Freed by task 1238:
 dvb_register_device+0x6d2/0xa70 [dvb_core]
 dvb_dmxdev_init+0x2af/0x4a0 [dvb_core]
 vidtv_bridge_probe+0x766/0xa40 [dvb_vidtv_bridge]
 ...

It is because the error handling in vidtv_bridge_dvb_init() is wrong.

First, vidtv_bridge_dmx(dev)_init() will clean themselves when fail, but
goto fail_dmx(_dev): calls release functions again, which causes
use-after-free.

Also, in fail_fe, fail_tuner_probe and fail_demod_probe, j = i will cause
out-of-bound when i finished its loop (i == NUM_FE). And the loop
releasing is wrong, although now NUM_FE is 1 so it won't cause problem.

Fix this by correctly releasing everything.

Fixes: f90cf6079bf6 ("media: vidtv: add a bridge driver")
Signed-off-by: Chen Zhongjin <chenzhongjin@huawei.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: staging: media: meson: vdec: use min() for comparison and assignment
Deepak R Varma [Mon, 7 Nov 2022 19:11:21 +0000 (19:11 +0000)]
media: staging: media: meson: vdec: use min() for comparison and assignment

Use of standard min() helper macro is preferred over using ternary
operator for logical evaluation and value assignment. This issue is
identified by coccicheck using the minmax.cocci file.

Signed-off-by: Deepak R Varma <drv@mailo.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: visl: add virtual stateless decoder driver
Daniel Almeida [Sat, 5 Nov 2022 13:58:42 +0000 (13:58 +0000)]
media: visl: add virtual stateless decoder driver

A virtual stateless device for stateless uAPI development purposes.

This tool's objective is to help the development and testing of
userspace applications that use the V4L2 stateless API to decode media.

A userspace implementation can use visl to run a decoding loop even when
no hardware is available or when the kernel uAPI for the codec has not
been upstreamed yet. This can reveal bugs at an early stage.

This driver can also trace the contents of the V4L2 controls submitted
to it.  It can also dump the contents of the vb2 buffers through a
debugfs interface. This is in many ways similar to the tracing
infrastructure available for other popular encode/decode APIs out there
and can help develop a userspace application by using another (working)
one as a reference.

Note that no actual decoding of video frames is performed by visl. The
V4L2 test pattern generator is used to write various debug information
to the capture buffers instead.

Signed-off-by: Daniel Almeida <daniel.almeida@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: amphion: apply vb2_queue_error instead of setting manually
Ming Qian [Tue, 1 Nov 2022 08:04:53 +0000 (08:04 +0000)]
media: amphion: apply vb2_queue_error instead of setting manually

vb2_queue_error is help to set the error of vb2_queue,
don't need to set it manually

Fixes: 3cd084519c6f ("media: amphion: add vpu v4l2 m2m support")
Signed-off-by: Ming Qian <ming.qian@nxp.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: amphion: add lock around vdec_g_fmt
Ming Qian [Tue, 1 Nov 2022 08:04:20 +0000 (08:04 +0000)]
media: amphion: add lock around vdec_g_fmt

the capture format may be changed when
sequence header is parsed,
it may be read and write in the same time,
add lock around vdec_g_fmt to synchronize it

Fixes: 6de8d628df6e ("media: amphion: add v4l2 m2m vpu decoder stateful driver")
Signed-off-by: Ming Qian <ming.qian@nxp.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: vimc: Fix wrong function called when vimc_init() fails
Chen Zhongjin [Thu, 10 Nov 2022 07:26:33 +0000 (07:26 +0000)]
media: vimc: Fix wrong function called when vimc_init() fails

In vimc_init(), when platform_driver_register(&vimc_pdrv) fails,
platform_driver_unregister(&vimc_pdrv) is wrongly called rather than
platform_device_unregister(&vimc_pdev), which causes kernel warning:

 Unexpected driver unregister!
 WARNING: CPU: 1 PID: 14517 at drivers/base/driver.c:270 driver_unregister+0x8f/0xb0
 RIP: 0010:driver_unregister+0x8f/0xb0
 Call Trace:
  <TASK>
  vimc_init+0x7d/0x1000 [vimc]
  do_one_initcall+0xd0/0x4e0
  do_init_module+0x1cf/0x6b0
  load_module+0x65c2/0x7820

Fixes: 4a29b7090749 ("[media] vimc: Subdevices as modules")
Signed-off-by: Chen Zhongjin <chenzhongjin@huawei.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: sun6i-csi: Add support for hooking to the isp devices
Paul Kocialkowski [Thu, 3 Nov 2022 16:37:17 +0000 (16:37 +0000)]
media: sun6i-csi: Add support for hooking to the isp devices

In order to use the isp and csi together, both devices need to be
parented to the same v4l2 and media devices. We use the isp as
top-level device and let the csi code hook to its v4l2 and media
devices when async subdev registration takes place.

As a result v4l2/media device setup is only called when the ISP
is missing and the capture device is registered after the devices
are hooked. The bridge subdev and its notifier are registered
without any device when the ISP is available. Top-level pointers
for the devices are introduced to either redirect to the hooked ones
(isp available) or the registered ones (isp missing).

Also keep track of whether the capture node was setup or not to
avoid cleaning up resources when it wasn't.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: sun6i-csi: Detect the availability of the ISP
Paul Kocialkowski [Thu, 3 Nov 2022 16:37:16 +0000 (16:37 +0000)]
media: sun6i-csi: Detect the availability of the ISP

Add a helper to detect whether the ISP is available and connected
and store the indication in the driver-specific device structure.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: MAINTAINERS: Add entry for the Allwinner A31 ISP driver
Paul Kocialkowski [Thu, 3 Nov 2022 16:37:15 +0000 (16:37 +0000)]
media: MAINTAINERS: Add entry for the Allwinner A31 ISP driver

Add myself as maintainer of the Allwinner A31 ISP media driver.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: staging: media: Add support for the Allwinner A31 ISP
Paul Kocialkowski [Thu, 3 Nov 2022 16:37:14 +0000 (16:37 +0000)]
media: staging: media: Add support for the Allwinner A31 ISP

Some Allwinner platforms come with an Image Signal Processor, which
supports various features in order to enhance and transform data
received by image sensors into good-looking pictures. In most cases,
the data is raw bayer, which gets internally converted to RGB and
finally YUV, which is what the hardware produces.

This driver supports ISPs that are similar to the A31 ISP, which was
the first standalone ISP found in Allwinner platforms. Simpler ISP
blocks were found in the A10 and A20, where they are tied to a CSI
controller. Newer generations of Allwinner SoCs (starting with the
H6, H616, etc) come with a new camera subsystem and revised ISP.
Even though these previous and next-generation ISPs are somewhat
similar to the A31 ISP, they have enough significant differences to
be out of the scope of this driver.

While the ISP supports many features, including 3A and many
enhancement blocks, this implementation is limited to the following:
- V3s (V3/S3) platform support;
- Bayer media bus formats as input;
- Semi-planar YUV (NV12/NV21) as output;
- Debayering with per-component gain and offset configuration;
- 2D noise filtering with configurable coefficients.

Since many features are missing from the associated uAPI, the driver
is aimed to integrate staging until all features are properly
described.

On the technical side, it uses the v4l2 and media controller APIs,
with a video node for capture, a processor subdev and a video node
for parameters submission. A specific uAPI structure and associated
v4l2 meta format are used to configure parameters of the supported
modules.

One particular thing about the hardware is that configuration for
module registers needs to be stored in a DMA buffer and gets copied
to actual registers by the hardware at the next vsync, when instructed
by a flag. This is handled by the "state" mechanism in the driver.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: dt-bindings: media: sun6i-a31-csi: Add internal output port to the ISP
Paul Kocialkowski [Thu, 3 Nov 2022 16:37:13 +0000 (16:37 +0000)]
media: dt-bindings: media: sun6i-a31-csi: Add internal output port to the ISP

Some Allwinner devices come with an Image Signal Processor (ISP) that
allows processing camera data to produce good-looking images,
especially from raw bayer representations.

The ISP does not have a dedicated capture path: it is fed directly by
one of the CSI controllers, which can be selected at run-time.

Represent this possibility as a graph connection between the CSI
controller and the ISP in the device-tree bindings.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: dt-bindings: media: Add Allwinner A31 ISP bindings documentation
Paul Kocialkowski [Thu, 3 Nov 2022 16:37:12 +0000 (16:37 +0000)]
media: dt-bindings: media: Add Allwinner A31 ISP bindings documentation

This introduces YAML bindings documentation for the Allwinner A31 Image
Signal Processor (ISP).

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: MAINTAINERS: Add myself as sun6i-csi maintainer and rename/move entry
Paul Kocialkowski [Thu, 3 Nov 2022 16:31:13 +0000 (16:31 +0000)]
media: MAINTAINERS: Add myself as sun6i-csi maintainer and rename/move entry

Given the substantial rework of the driver that I carried out and the
knowledge acquired about the hardware along the way, make myself a
maintainer of the sun6i-csi driver.

Also rename and move the entry while at it since the driver is not
specific to the V3s.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: sun6i-csi: Request a shared interrupt
Paul Kocialkowski [Thu, 3 Nov 2022 16:31:12 +0000 (16:31 +0000)]
media: sun6i-csi: Request a shared interrupt

Request our interrupt shared since it is typically shared with the isp
block. The interrupt routine looks good to go for shared irq.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: sun6i-csi: Add extra checks to the interrupt routine
Paul Kocialkowski [Thu, 3 Nov 2022 16:31:11 +0000 (16:31 +0000)]
media: sun6i-csi: Add extra checks to the interrupt routine

Check against the enabled bits and make sure capture is running before
serving an interrupt, to add extra safety in the process.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
19 months agomedia: sun6i-csi: Only configure capture when streaming
Paul Kocialkowski [Thu, 3 Nov 2022 16:31:10 +0000 (16:31 +0000)]
media: sun6i-csi: Only configure capture when streaming

Add a streaming element to the capture state structure to know if the
capture device is used or not. Only configure things related to output
when streaming, including the output format, irq, state (dma buffer)
and window configuration registers.

After this change, it becomes possible to use the bridge without the
capture device, which will be the case in the isp media flow.

Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>