Re: [PATCH 1/2] dt: watchdog: Add DT binding documentation for jz47xx watchdog timer

[Arnd Bergmann <arnd-r2nGTMty4D4@public.gmane.org>]

On Tuesday 27 January 2015 14:19:09 Guenter Roeck wrote:
> On Tue, Jan 27, 2015 at 10:29:49PM +0100, Arnd Bergmann wrote:
> > On Tuesday 27 January 2015 12:52:29 Guenter Roeck wrote:
> > > Driver does this (today):
> > > 
> > >          drvdata->rtc_clk = clk_get(&pdev->dev, "rtc");
> > > 
> > > Isn't that the name to use ? Just wondering.
> > 
> > Just because the driver uses it at the moment does not mean it's the name
> > that the IP block uses.
> > 
> > clk_get() has the unpleasant property of doing fuzzy matching
> > on the name that is passed. It first tries to use the string
> > as the name of the clock input of the device, but if that is
> > not there, it falls back to looking for a global clk with a con_id.
> > 
> > In DT, we only support the first kind, but if a driver currently
> > uses the second, you get the wrong name.
> > 
> > Looking at arch/mips/jz4740/clock.c now, this seems to be exactly
> > what is going on here: there is no clkdev_add call to associate
> > the device clocks, so it can only match a global clock entry. 
> > 
> Me confused :-(.
> 
> Does that mean the driver needs to be fixed, that the DT property
> needs to change (to what ?), or both ?

Both.

The jz47xx clock driver should register a clkdev lookup table with
proper clock input names for each clock that is referenced by a
device, and then the drivers can use the right names.

In a lot of cases, the best name for a clock is no name so you
just use an anonymous clock like

	clk = clk_get(dev, NULL);

but this still requires a clock lookup table.

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-watchdog" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html