[RFC PATCH 00/10] Tegra XHCI support

[Andrew Bresticker]

On Thu, May 15, 2014 at 12:33 PM, Stephen Warren <swarren at wwwdotorg.org> wrote:
> On 05/14/2014 06:32 PM, Andrew Bresticker wrote:
>> This is a first pass at the host and PHY drivers necessary for USB3.0
>> support on Tegra114 and Tegra124.  The Tegra XHCI host controller requires
>> external firmware [1] which must be loaded before using any USB ports owned
>> by the controller.  The XUSB PHY driver handles mapping and enabling of
>> the UTMI, HSIC, and SuperSpeed pads to the XHCI controller.
>>
>> Tested on a Venice2 with a variety of USB2.0 and USB3.0 memory sticks
>> and ethernet dongles.
>>
>> Notes:
>>  - I've included support for Tegra114, but since I don't have Tegra114-based
>>    hardware, it is completely untested.
>>  - The PCIe and SATA PHYs also are programmed using the XUSB_PADCTL space
>>    as well.  At least some of the code can be re-used, specifically with
>>    respect to lane programming.  I believe Thierry is working on the PCIe
>>    parts of this.
>
> If I understand the HW correctly, there's a separate "pad control" HW
> block that provides routing/sharing of signals from USB2(?), USB3, SATA,
> and PCIe to the pads.
>
> I believe Thierry is working on exposing this block as a pinctrl driver,
> or at least something that the other drivers can call into in order to
> configure that block. It'd be good if you can co-ordinate with him to
> rebase this driver on top of that, rather than (I assume; haven't read
> the code yet...) directly manipulating the padctrl registers inside each
> of the different drivers.

Yes, ideally i'd like to rebase this on top of Theirry's series once
it's ready.  It sounds like there will be a pinctrl driver which
handles the lane muxing between XUSB, PCIe, and SATA and a generic PHY
driver will deal wit the actual PHY programming.  I'm not sure what
the PCIe and SATA PHY drivers will look like, but the XUSB one
requires quite a bit of USB-specific programming of the padctl
registers, so I'm not sure we'll be able to fit all three PHY types
into a single PHY provider driver.  Either way, there's probably a
good deal of PHY code that could be broken out and shared among the
three.