Re: [PATCH 1/5] pinctrl: core: Use delayed work for hogs

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




* Linus Walleij <linus.walleij@xxxxxxxxxx> [161111 12:17]:
> On Tue, Oct 25, 2016 at 11:02 PM, Tony Lindgren <tony@xxxxxxxxxxx> wrote:
> 
> > Having the pin control framework call pin controller functions
> > before it's probe has finished is not nice as the pin controller
> > device driver does not yet have struct pinctrl_dev handle.
> >
> > Let's fix this issue by adding deferred work for hogs. This is
> > needed to be able to add pinctrl generic helper functions.
> >
> > Note that the pinctrl functions already take care of the necessary
> > locking.
> >
> > Signed-off-by: Tony Lindgren <tony@xxxxxxxxxxx>
> 
> I don't see why this is necessary?

It's needed because the pin controller driver has not yet
finished it's probe at this point. We end up calling functions
in the device driver where no struct pinctrl_dev is yet known
to the driver. Asking a device driver to do something before
it's probe is done does not quite follow the Linux driver model :)

> The hogging was placed inside pinctrl_register() so that any hogs
> would be taken before it returns, so nothing else can take it
> before the controller itself has the first chance. This semantic
> needs to be preserved I think.
> 
> > +       schedule_delayed_work(&pctldev->hog_work,
> > +                                     msecs_to_jiffies(100));
> 
> If we arbitrarily delay, something else can go in and take the
> pins used by the hogs before the pinctrl core? That is what
> we want to avoid.
> 
> Hm, 100ms seems arbitrarily chosen BTW. Can it be 1 ms?
> 1 ns?

Yeah well seems like it should not matter but the race we need
to remove somehow.

> I'm pretty sure that whatever it is that needs to happen before
> the hog work runs can race with this delayed work under
> some circumstances (such as slow external expanders
> on i2c). It should be impossible for that to happen
> and I don't think it is?

Yes it's totally possible even with delay set to 0.

Maybe we could add some trigger on the first consumer request
and if that does not happen use the timer?

Regards,

Tony

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]
  Powered by Linux