Re: [PATCH v4 5/9] ARM: dts: omap3-overo: Add HSUSB PHY

[Florian Vaussard <florian.vaussard@epfl.ch>]

On 03/10/2014 07:03 PM, Tony Lindgren wrote:
> * Florian Vaussard <florian.vaussard@epfl.ch> [140310 08:17]:
>> On 03/10/2014 11:30 AM, Roger Quadros wrote:
>>>
>>> If you don't use the OMAP3x30_CORE2_IOPAD() macro then these definitions can fit in omap3-overo-base.dtsi.
>>> The offsets will be taken care of if you place them within &omap3_pmx_core2 {} . e.g.
>>>
>>> +                       0x50 (PIN_OUTPUT | MUX_MODE3)           /* etk_d10.hsusb2_clk */
>>> +                       0x52 (PIN_OUTPUT | MUX_MODE3)           /* etk_d11.hsusb2_stp */
>>> +                       0x54 (PIN_INPUT_PULLDOWN | MUX_MODE3)   /* etk_d12.hsusb2_dir */
>>> +                       0x56 (PIN_INPUT_PULLDOWN | MUX_MODE3)   /* etk_d13.hsusb2_nxt */
>>> +                       0x58 (PIN_INPUT_PULLDOWN | MUX_MODE3)   /* etk_d14.hsusb2_data0 */
>>> +                       0x5a (PIN_INPUT_PULLDOWN | MUX_MODE3)   /* etk_d15.hsusb2_data1 */
>>>
>>> Do you think this is better?
>>>
>>
>> I do not think that this could work. Let's take for example hsusb2_clk
>> et let's do the math. Here, 0x50 is the offset inside omap3_pmx_core2
>> pinctrl, so:
>>
>>         hsusb2_clk  omap3_pmx_core2      hsusb2_clk
>>          offset        base addr.       PADCONF addr.
>>           ----        ----------         ----------
>> OMAP34xx: 0x50        0x480025d8         0x48002628  *NO*
>> OMAP36xx: 0x50        0x480025a0         0x480025F0  *ok*
>>
>> Looking at the TRM, the PADCONF address for hsusb2_clk is 0x480025F0 in
>> both cases. Thus we will be wrong on OMAP34xx (offset should be 0x18
>> instead of 0x50). Thus so far, I do not see any magical solution apart
>> putting the omap3_pmx_core2 pins in a SoC-dependant .dtsi file.
>>
>> I already expressed my concerns [1] about having a different base
>> address for omap3_pmx_core2 depending on the SoC (although I do
>> understand the rational behind). This makes cross-SoC platforms more
>> difficult to maintain.
> 
> Yeah the muxing should always be possible to do at board level.
> 
> Just the use of external pulls for some boards will make it hard
> to have anything common. If you want something common, you can
> have something like omap36xx-usb-xyz.dtsi as long as it's always
> wired up the same way.
> 

This is why I put the muxing in a file that will be shared by all Overo
expansion boards. Looking at the other .dts, only omap3-beagle.dts /
omap3-beagle-xm.dts define the pinmux for hsusb2_*, so for now I do not
think it is worth the effort to create a pinmux file common to all omap3
.dts.

Regards,
Florian