kernel.git/drivers/media/platform/davinci, branch linux-5.4.y

media: davinci: vpif: fix unbalanced runtime PM get

2022-04-15T12:18:03Z

commit 4a321de239213300a714fa0353a5f1272d381a44 upstream. Make sure to balance the runtime PM usage counter on driver unbind. Fixes: 407ccc65bfd2 ("[media] davinci: vpif: add pm_runtime support") Cc: stable@vger.kernel.org # 3.9 Cc: Lad, Prabhakar Signed-off-by: Johan Hovold Reviewed-by: Lad Prabhakar Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab Signed-off-by: Greg Kroah-Hartman

media: subdev: disallow ioctl for saa6588/davinci

2021-07-19T06:53:17Z

commit 0a7790be182d32b9b332a37cb4206e24fe94b728 upstream. The saa6588_ioctl() function expects to get called from other kernel functions with a 'saa6588_command' pointer, but I found nothing stops it from getting called from user space instead, which seems rather dangerous. The same thing happens in the davinci vpbe driver with its VENC_GET_FLD command. As a quick fix, add a separate .command() callback pointer for this driver and change the two callers over to that. This change can easily get backported to stable kernels if necessary, but since there are only two drivers, we may want to eventually replace this with a set of more specialized callbacks in the long run. Fixes: c3fda7f835b0 ("V4L/DVB (10537): saa6588: convert to v4l2_subdev.") Cc: stable@vger.kernel.org Signed-off-by: Arnd Bergmann Reviewed-by: Laurent Pinchart Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab Signed-off-by: Greg Kroah-Hartman

media: davinci: vpif_capture: fix potential double free

2020-09-03T09:26:48Z

[ Upstream commit 602649eadaa0c977e362e641f51ec306bc1d365d ] In case of errors vpif_probe_complete() releases memory for vpif_obj.sd and unregisters the V4L2 device. But then this is done again by vpif_probe() itself. The patch removes the cleaning from vpif_probe_complete(). Found by Linux Driver Verification project (linuxtesting.org). Signed-off-by: Evgeny Novikov Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab Signed-off-by: Sasha Levin

media: vpss: clean up resources in init

2020-08-26T08:40:54Z

[ Upstream commit 9c487b0b0ea7ff22127fe99a7f67657d8730ff94 ] If platform_driver_register() fails within vpss_init() resources are not cleaned up. The patch fixes this issue by introducing the corresponding error handling. Found by Linux Driver Verification project (linuxtesting.org). Signed-off-by: Evgeny Novikov Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab Signed-off-by: Sasha Levin

media: use the BIT() macro

2019-08-26T17:08:50Z

As warned by cppcheck: [drivers/media/dvb-frontends/cx24123.c:434]: (error) Shifting signed 32-bit value by 31 bits is undefined behaviour [drivers/media/pci/bt8xx/bttv-input.c:87]: (error) Shifting signed 32-bit value by 31 bits is undefined behaviour [drivers/media/pci/bt8xx/bttv-input.c:98]: (error) Shifting signed 32-bit value by 31 bits is undefined behaviour ... [drivers/media/v4l2-core/v4l2-ioctl.c:1391]: (error) Shifting signed 32-bit value by 31 bits is undefined behaviour There are lots of places where we're doing 1 << 31. That's bad, as, depending on the architecture, this has an undefined behavior. The BIT() macro is already prepared to handle this, so, let's just switch all "1 << number" macros by BIT(number) at the header files with has 1 << 31. Reviewed-by: Sylwester Nawrocki # exynos4-is and s3c-camif Reviewed-by: Laurent Pinchart # omap3isp, vsp1, xilinx, wl128x and ipu3 Reviewed-by: Benoit Parrot # am437x and ti-vpe Signed-off-by: Mauro Carvalho Chehab

media: v4l2-async: Get fwnode reference when putting it to the notifier's list

2019-07-25T12:01:43Z

The v4l2_async_notifier_add_fwnode_subdev() did not take a reference of the added fwnode, relying on the caller to handle that instead, in essence putting the fwnode to be added if there was an error. As the reference is eventually released during the notifier cleanup, this is not intuitive nor logical. Improve this by always getting a reference when the function succeeds, and the caller releasing the reference when it does not *itself* need it anymore. Luckily, perhaps, there were just a handful of callers using the function. Signed-off-by: Sakari Ailus Reviewed-by: Jacopo Mondi Tested-by: Niklas Söderlund Signed-off-by: Mauro Carvalho Chehab

media: davinci-vpif: Don't dereference endpoint after putting it, fix refcounting

2019-07-25T12:01:10Z

The davinci-vpif driver dereferences its local endpoints after releasing the reference to them. The driver also puts its endpoints explicitly while the of_graph_get_next_endpoint() does that, too, leading to obtaining a reference once and releasing it twice. Both are fixed by this patch. Signed-off-by: Sakari Ailus Signed-off-by: Mauro Carvalho Chehab

media: davinci/vpfe_capture.c: drop unused format descriptions

2019-07-25T10:30:00Z

Simplify vpfe_pixel_format to just contain the pixelformat and bpp fields. All others are unused. Signed-off-by: Hans Verkuil Acked-by: Lad, Prabhakar Signed-off-by: Mauro Carvalho Chehab

media: davinci-vpbe: remove obsolete includes

2019-07-23T12:50:41Z

The driver builds fine without these, and they cause build problems once davinci multiplatform support is enabled. Signed-off-by: Arnd Bergmann Acked-by: Lad, Prabhakar Acked-by: Sekhar Nori Signed-off-by: Hans Verkuil Signed-off-by: Mauro Carvalho Chehab

media: drivers/media: don't set pix->priv = 0

2019-07-23T12:48:33Z

The priv field of struct v4l2_pix_format shouldn't be set by drivers, it's set by the v4l2 core instead to V4L2_PIX_FMT_PRIV_MAGIC. Drop this from the few media drivers that still do this. Note that the gspca patch is slightly more involved since some of the sub-gspca drivers use the priv field internally. Signed-off-by: Hans Verkuil Reviewed-by: Laurent Pinchart [hverkuil-cisco@xs4all.nl: fix clash in gspca between priv arg and priv variable] Signed-off-by: Mauro Carvalho Chehab