marzen pinctl conflict for renesas-next-20130515v2 + v3.10-rc1

marzen pinctl conflict for renesas-next-20130515v2 + v3.10-rc1

[Simon Horman]

Hi Laurent,

I am preparing to rebase the branches of the renesas tree which
are targeted for v3.11 on top of v3.10-rc1. In the course of
doing so I have noticed a problem on the marzen board which appears
to be related to pinmux.

On boot, using the default config I see the following:

sh-pfc pfc-r8a7779: pin GP_4_22 already requested by sh-hspi.0; cannot claim for ehci-platform.0
sh-pfc pfc-r8a7779: pin-150 (ehci-platform.0) status -22
sh-pfc pfc-r8a7779: could not request pin 150 on device sh-pfc
ehci-platform ehci-platform.0: Error applying setting, reverse things back
Unable to handle kernel NULL pointer dereference at virtual address 00000036

Looking at pfc-r8a7779 and the SoC documentation it seems
that HSPI0 and USB0 do indeed use different functions of
GPIO pin 4 22 (and other pins too).

I am unsure how this should be managed by either the pinmux driver
or the board code.

For reference I have pushed a merge of renesas-next-20130515v2 and
v3.10-rc1 as renesas-next-20130515v2+v3.10-rc1. If you have a marzen
board you should be able to reproduce the problem with that tree.

Re: marzen pinctl conflict for renesas-next-20130515v2 + v3.10-rc1

[Sergei Shtylyov]

Hello.

On 15-05-2013 12:46, Simon Horman wrote:

> Hi Laurent,

> I am preparing to rebase the branches of the renesas tree which
> are targeted for v3.11 on top of v3.10-rc1. In the course of
> doing so I have noticed a problem on the marzen board which appears
> to be related to pinmux.

> On boot, using the default config I see the following:

> sh-pfc pfc-r8a7779: pin GP_4_22 already requested by sh-hspi.0; cannot claim for ehci-platform.0
> sh-pfc pfc-r8a7779: pin-150 (ehci-platform.0) status -22
> sh-pfc pfc-r8a7779: could not request pin 150 on device sh-pfc

    I have already reported this. Laurent prepared a fix, which I 
somewhat criticized, but he never published a second version.

> ehci-platform ehci-platform.0: Error applying setting, reverse things back
> Unable to handle kernel NULL pointer dereference at virtual address 00000036

    These are new to me.

> Looking at pfc-r8a7779 and the SoC documentation it seems
> that HSPI0 and USB0 do indeed use different functions of
> GPIO pin 4 22 (and other pins too).

    No, only this pin.

> I am unsure how this should be managed by either the pinmux driver
> or the board code.

    PENC0 and USB_OVC0 were incorrectly grouped together, the fix is to 
ungroup them. USB_OVC0 which causes the conflict is not actually used.

WBR, Sergei

Re: marzen pinctl conflict for renesas-next-20130515v2 + v3.10-rc1

[Laurent Pinchart]

Hi Sergei,

On Wednesday 15 May 2013 16:36:58 Sergei Shtylyov wrote:
> On 15-05-2013 12:46, Simon Horman wrote:
> > Hi Laurent,
> > 
> > I am preparing to rebase the branches of the renesas tree which
> > are targeted for v3.11 on top of v3.10-rc1. In the course of
> > doing so I have noticed a problem on the marzen board which appears
> > to be related to pinmux.
> > 
> > On boot, using the default config I see the following:
> > 
> > sh-pfc pfc-r8a7779: pin GP_4_22 already requested by sh-hspi.0; cannot
> > claim for ehci-platform.0 sh-pfc pfc-r8a7779: pin-150 (ehci-platform.0)
> > status -22
> > sh-pfc pfc-r8a7779: could not request pin 150 on device sh-pfc
> 
> I have already reported this. Laurent prepared a fix, which I somewhat
> criticized, but he never published a second version.

I've posted v2 earlier today.

> > ehci-platform ehci-platform.0: Error applying setting, reverse things back
> > Unable to handle kernel NULL pointer dereference at virtual address
> > 00000036
>
> These are new to me.
> 
> > Looking at pfc-r8a7779 and the SoC documentation it seems
> > that HSPI0 and USB0 do indeed use different functions of
> > GPIO pin 4 22 (and other pins too).
> 
> No, only this pin.
> 
> > I am unsure how this should be managed by either the pinmux driver
> > or the board code.
> 
> PENC0 and USB_OVC0 were incorrectly grouped together, the fix is to ungroup
> them. USB_OVC0 which causes the conflict is not actually used.

-- 
Regards,

Laurent Pinchart

Re: marzen pinctl conflict for renesas-next-20130515v2 + v3.10-rc1

[Sergei Shtylyov]

Hello.

On 15-05-2013 16:38, Laurent Pinchart wrote:

>>> I am preparing to rebase the branches of the renesas tree which
>>> are targeted for v3.11 on top of v3.10-rc1. In the course of
>>> doing so I have noticed a problem on the marzen board which appears
>>> to be related to pinmux.

>>> On boot, using the default config I see the following:

>>> sh-pfc pfc-r8a7779: pin GP_4_22 already requested by sh-hspi.0; cannot
>>> claim for ehci-platform.0 sh-pfc pfc-r8a7779: pin-150 (ehci-platform.0)
>>> status -22
>>> sh-pfc pfc-r8a7779: could not request pin 150 on device sh-pfc

>> I have already reported this. Laurent prepared a fix, which I somewhat
>> criticized, but he never published a second version.

> I've posted v2 earlier today.

    Saw it after replying to Simon, thanks.

WBR, Sergei