[PATCH v3 00/20] USB support for Armada 38x and Armada 375

[Hans de Goede]

Hi,

On 05/07/2014 11:21 AM, Sebastian Hesselbarth wrote:
> [Adding back the original Cc-List]
> 
> On 05/07/2014 10:33 AM, Hans de Goede wrote:
>> Hi Gregory,
>>
>> Gregory CLEMENT <gregory.clement <at> free-electrons.com> writes:
>>
>>> This patch set adds the USB support for the Armada 38x Armada
>>> 375. These SoCs use an xHCI but still need specific initialization,
>>> mainly to setup the windows memory on the mbus. They also use the same
>>> controller that the other mvebu SoC for EHCI.
>>
>> <snip>
>>
>>> Gregory CLEMENT (20):
>>>    usb: ehci-orion: Fix clock reference leaking
>>>    usb: ehci-orion: Add the optional PHY support
>>
>> If you're using a proper phy driver, do you really need a soc specific ehci
>> driver? I've spend a lot of time during the last release cycle to make
>> ehci-platform a lot more generic, including adding (optional) phy support to
>> it. It would be great if the new Armada SoCs could use ehci-platform instead
>> of extending ehci-orion to cover more models (ideally most ehci-foo drivers
>> would go away completely). I've no idea how feasible it is to use
>> ehci-platform in your case, but IMHO it should at least be considered.
> 
> Hans,
> 
> unfortunately we currently have no USB-PHY driver for the other mvebu
> SoCs. I have started some code grabbing for barebox bootloader and plan
> to add proper USB-PHYs for Linux later.
> 
> There is one thing very specific to mvebu SoCs, which is the mbus:
> you need to setup upstream memory windows for the usb bus master.
> From my current understanding, that will require at least a minimal
> stub to remain for ehci-orion.
> 
> I do think that most of the stuff in ehci-orion can (and should) be
> replaced with ehci-platform then. So, we definitely consider reusing
> the generic foo-platform drivers on all opportunities but IMHO we are
> not yet ready for it.

Ok, thanks for explaining.

Regards,

Hans