On 03/12/2017 02:09 PM, Russell King - ARM Linux wrote:

On Sun, Mar 12, 2017 at 08:40:37PM +0000, Russell King - ARM Linux wrote:

quoted

On Sun, Mar 12, 2017 at 01:36:32PM -0700, Steve Longerbeam wrote:

quoted

But hold on, if my logic is correct, then why did the CSI power-off
get reached in your case, multiple times? Yes I think there is a bug,
link_notify() is not checking if the link has already been disabled.
I will fix this. But I'm surprised media core's link_notify handling
doesn't do this.

Well, I think there's something incredibly fishy going on here.  I
turned that dev_dbg() at the top of the function into a dev_info(),
and I get:

root at hbi2ex:~# dmesg |grep -A2 imx-ipuv3-csi
[   53.370949] imx-ipuv3-csi imx-ipuv3-csi.0: power OFF
[   53.371015] ------------[ cut here ]------------
[   53.371075] WARNING: CPU: 0 PID: 1515 at drivers/staging/media/imx/imx-media-csi.c:806 csi_s_power+0xb8/0xd0 [imx_media_csi]
--
[   53.372624] imx-ipuv3-csi imx-ipuv3-csi.0: power OFF
[   53.372637] ------------[ cut here ]------------
[   53.372663] WARNING: CPU: 0 PID: 1515 at drivers/staging/media/imx/imx-media-csi.c:806 csi_s_power+0xb8/0xd0 [imx_media_csi]

There isn't a power on event being generated before these two power
off events.  I don't see a power on event even when I attempt to
start streaming either (which fails due to the lack of bayer
support.)

Found it - my imx219 driver returns '1' from its s_power function when
powering up, which triggers a bug in your code - when imx_media_set_power()
fails to power up, you call imx_media_set_power() telling it to power
everything off - including devices that are already powered off.

Yep, there's a bug in the error cleanup in imx_media_pipeline_set_power().
On error, it needs to backout by calling s_power(off) as it is doing, 
but not
through the whole pipeline, but needs to stop at the subdev encountered
just before the subdev that failed. This was causing the s_power() 
imbalance.
I will fix.

This is really bad news - s_power() may be called via other paths,
such as when the subdev is opened.

I don't think that is a problem, as long as power_count is working
as it should, and the caller from the other paths has not created an
imbalance.

Steve