Re: [PATCH v4 1/1] clkdev: change prototype of clk_register_clkdev()

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Wed, 2015-05-06 at 13:51 +0100, Russell King - ARM Linux wrote:
> On Wed, May 06, 2015 at 03:34:09PM +0300, Andy Shevchenko wrote:
> > On Wed, 2015-05-06 at 12:09 +0100, Russell King - ARM Linux wrote:
> > > On Wed, May 06, 2015 at 01:24:12PM +0300, Andy Shevchenko wrote:
> > > > Since clk_register_clkdev() is exported for modules the caller should get a
> > > > pointer to the allocated resources. Otherwise the memory leak is guaranteed on
> > > > the ->remove() stage.
> > > 
> > > clk_register_clkdev() is there to assist mass clock registrations, which
> > > typically happen in platform code.  It's there to simplify the "I need
> > > to register this clock which I'm not going to release again".
> > > 
> > > I don't see any of these locations trying to unregister their clk from
> > > clkdev, so I doubt this patch is needed.
> > 
> > We are doing the driver which will use this (as I mentioned in the cover
> > letter).
> > 
> > > 
> > > Where a module wants to remove its clk from clkdev, it should register
> > > its clk with clkdev_create() and remove it with clkdev_drop().
> > > 
> > 
> > You are talking about something in the wild? I can't find
> > clkdev_create() neither in current clk.git nor in linux-next.git.
> 
> It'll be in linux-next RSN.
> 
> Search lakml for "Fix fallout from per-user struct clk patches"
> 

Thanks, we will use it, though the documentation might say that there is
a risk of memory leak in case of usage clk_register_clkdev().

-- 
Andy Shevchenko <andriy.shevchenko@intel.com>
Intel Finland Oy