On Thu, Dec 12, 2013 at 02:36:01PM -0500, Santosh Shilimkar wrote:
> On Thursday 12 December 2013 12:48 PM, Felipe Balbi wrote:
> > Hi,
> >
> > On Tue, Dec 10, 2013 at 10:11:36AM -0500, Santosh Shilimkar wrote:
> >>>> +static int kdwc3_remove(struct platform_device *pdev)
> >>>> +{
> >>>> + struct dwc3_keystone *kdwc = platform_get_drvdata(pdev);
> >>>> +
> >>>> + kdwc3_disable_irqs(kdwc);
> >>>> + clk_disable_unprepare(kdwc->clk);
> >>>
> >>> I hope the clock isn't shared between core and wrapper, otherwise you
> >>> could run into some troubles here. Can you confirm ?
> >>>
> >> Yes. the clock isn't shared. Thanks for taking care of other parts.
> >
> > so clock for core is always running too ?
> >
> I take that back. The clock is actually common so we should disable
> it after removing the kdwc3_remove_core() as you suggested.
>
> You won't see issue since the kdwc3_remove_core() not doing
> any register access but moving the clock disable after
> the core remove is right thing to do.
the problem is not kdwc3_remove_core() accessing registers, but
dwc3_remove() _does_ access registers during remove. If you just
mopdrobe -r dwc3-keystone without removing dwc3.ko first, then
kdwc3_remove_core() will cause dwc3.ko to be removed (because of
platform_driver_unregister()) and, since clocks have already been
disabled, then we'd die :-)
cheers
--
balbi
Attachment:
signature.asc
Description: Digital signature
