Re: [PATCH v5 2/4] dt-bindings: watchdog: add new binding for meson secure watchdog

[Rob Herring]

On Wed, Dec 18, 2019 at 03:37:49PM +0800, Xingyu Chen wrote:
> Hi, Guenter Martin
> 
> On 2019/12/16 21:30, Guenter Roeck wrote:
> > On 12/15/19 10:03 PM, Xingyu Chen wrote:
> > > Hi, Martin
> > > 
> > > Sorry for the late reply.
> > > 
> > > On 2019/12/13 4:05, Martin Blumenstingl wrote:
> > > > Hi Xingyu and Rob,
> > > > 
> > > > On Thu, Dec 12, 2019 at 1:20 PM Xingyu Chen
> > > > <xingyu.chen@xxxxxxxxxxx> wrote:
> > > > [...]
> > > > > +examples:
> > > > > +Â - |
> > > > > +ÂÂÂ watchdog {
> > > > > +ÂÂÂÂÂÂÂÂÂ compatible = "amlogic,meson-sec-wdt";
> > > > > +ÂÂÂÂÂÂÂÂÂ timeout-sec = <60>;
> > > > > +ÂÂÂ };
> > > > in v3 of this patch Rob commented that there shouldn't be an OF node
> > > > if there are no additional properties
> > > > with timeout-sec there's now an additional property so my
> > > > understanding is that it's fine to have an OF node
> > > Your understanding is correct.
> > > > 
> > > > what I don't understand yet is where this node should be placed.
> > > > is it supposed to be a child node of the secure monitor node (for
> > > > which we already have a binding here:
> > > > Documentation/devicetree/bindings/firmware/meson/meson_sm.txt) or
> > > > where else would we place it inside the .dts?
> > > IMO, Although the watchdog node need to reference the meson_sm
> > > node, there is no
> > > bus-like dependencies between the devices which the two nodes
> > > corresponding to.
> > > so i think that the watchdog node as child node of meson_sm maybe
> > > not appropriate.
> > 
> > The watchdog driver needs the meson SM's dt node, and it depends on the
> > existence
> > of that node. That seems enough of a relationship to warrant having it
> > as child note.
> Thanks for your reply, if i take the wdt node as child of secure monitor
> (sm), how should
> i register or find the wdt device ?
> 
> I only think of the following three methods :
> 1). update the sm driverïand scan&register wdt device when the sm driver
> probes(It is like i2c), but there
> are too many changes involved.

Just add of_platform_default_populate() call and clean-up calls. That's 
not what I'd call 'too many changes'.

 
> 2). add "simple-bus" key string to compatible of sm node, and it will make
> the child node is registered as
> platform device, but it seems that the key string is not match current
> scene.

You previously said it's not a bus...

> 
> secure-monitor {
>  compatible = "amlogic,meson-gxbb-sm", "simple-bus";
> 
> ÂÂÂ watchdog {
> ÂÂÂ ÂÂÂ compatible = "amlogic,meson-sec-wdt";
> ÂÂÂ ÂÂÂ timeout-sec = <60>;
> ÂÂÂ }
> }
> 
> 3). don't register device, and find directly the watchdog node by using the
> of_* API in watchdog
> driver (Eg: linux-4.x/drivers/tee/optee/core.c)
> 
> secure-monitor {
> ÂÂÂ compatible = "amlogic,meson-gxbb-sm";
> 
> ÂÂÂ watchdog {
> ÂÂÂÂÂÂÂ compatible = "amlogic,meson-sec-wdt";
> ÂÂÂÂÂÂÂ timeout-sec = <60>;
> ÂÂÂ }
> }
> 
> The method 3 looks better for me, do you have a better suggestion ? Thanks
> 
> BR
> > 
> > Guenter
> > 
> > .
> >