Re: [PATCH v7 2/5] media: i2c: imx355: Support devicetree and power management

[Richard Acayan <mailingradian@gmail.com>]

On Wed, Jan 28, 2026 at 09:59:20AM +0200, Sakari Ailus wrote:
> On Tue, Jan 27, 2026 at 09:53:38PM -0500, Richard Acayan wrote:
> > On Tue, Jan 20, 2026 at 04:49:21PM +0200, Sakari Ailus wrote:
> > > Hi Bryan, others,
> > > 
> > > On Tue, Jan 20, 2026 at 12:44:24PM +0000, Bryan O'Donoghue wrote:
> > > > I think reset should be asserted before regulators and power are switched
> > > > on. i.e. before you try to switch the chip on, you should establish that the
> > > > reset pin is in the state that the timing diagram calls for.
> > > 
> > > Indeed.
> > 
> > I think the discussion is more about whether there should be an assert
> > in the same function as the de-assert.
> > 
> > > The xshutdown pin, as it is typically called labelled as "reset" in this
> > > case, functions as both hardware reset and hardware standby mode control.
> > > It should be asserted (i.e. be set to low level) whenever the sensor is
> > > expected to be powered off. Typically deasserting it is the last step in
> > > the sensor's power-up sequence. This applies to nearly all CSI-2 and DVP
> > > (parallel) camera sensors. (There are some exceptions that use explicitly
> > > two GPIOs for similar functions but there are very few of them.)
> > 
> > This patch has the reset asserted by the time it gets to
> > imx355_power_on():
> > 
> > - when coming from runtime PM, the suspend callback asserted it
> > - when coming from probe, GPIOD_OUT_HIGH asserted it (considering that
> >   active-low also affects the initial output setting)
> > 
> > Should it be asserted again inside the function, or
> > should the initial `gpiod_set_value_cansleep()` be removed?
> 
> Please remove it as requested.

Ok, I will remove it.