On Tuesday 28 February 2017 04:30 PM, Bartosz Golaszewski wrote:

2017-02-28 6:36 GMT+01:00 Sekhar Nori [off-list ref]:

quoted

On Tuesday 28 February 2017 04:22 AM, Rob Herring wrote:

quoted

On Wed, Feb 22, 2017 at 02:43:47PM +0100, Bartosz Golaszewski wrote:

quoted

Add an optional property - enable-gpios - which can be used to specify
the GPIOs that must be requested to select the vpif functionality.

Signed-off-by: Bartosz Golaszewski <redacted>
---
 Documentation/devicetree/bindings/media/ti,da850-vpif.txt | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/Documentation/devicetree/bindings/media/ti,da850-vpif.txt b/Documentation/devicetree/bindings/media/ti,da850-vpif.txt
index df7182a..23c5405 100644
--- a/Documentation/devicetree/bindings/media/ti,da850-vpif.txt
+++ b/Documentation/devicetree/bindings/media/ti,da850-vpif.txt

@@ -13,6 +13,9 @@ Required properties:
 - reg: physical base address and length of the registers set for the device;
 - interrupts: should contain IRQ line for the VPIF

+Optional properties:
+- enable-gpios: phandle of the GPIOs used to select the vpif functionality

What does this control exactly? The GPIOs belong in the node they are
connected to and having GPIOs routed to this block seems strange.

The DA850 EVM board implements on-board muxing which lets the video
input (via VPIF) to be routed to a TVP5147 (video decoder) for composite
input or to a camera header. There are other mux options which use the
same VPIF SoC pins (RMII ethernet or character LCD).

There is a three-to-eight demux on the board which drives enable signals
to buffers letting you choose from these options.

From your response, it looks like you want the enable-gpios property to
be in the TVP5147 or camera node. That does make sense to me.

Wouldn't that cause races again? The TVP5147 ADC is probed from within
the vpif driver as a media subdev (after initialize_vpif(),
v4l2_device_register(), irq requests etc.).

When TVP5147 probe fails with -EPROBE_DEFER, I assume the error can
propagate back to vpif_probe() which would then de-register the V4L2
device as part of probe error unwinding. That should then allow the
probe to be retried at a later time.

Thanks,
Sekhar

Thanks,
Bartosz