Re: [PATCH v3 1/2] usb: musb: add support for JZ4740 usb device controller

[Lars-Peter Clausen <lars@metafoo.de>]

On 01/04/2014 12:11 PM, Richard Weinberger wrote:
> Am Samstag, 4. Januar 2014, 12:06:22 schrieb Apelete Seketeli:
>> On 04-Jan-14, Richard Weinberger wrote:
>>> On Thu, Dec 19, 2013 at 9:42 PM, Apelete Seketeli <apelete@seketeli.net>
> wrote:
>>>> Add support for Ingenic JZ4740 USB Device Controller through a
>>>> specific musb glue layer.
>>>>
>>>> JZ4740 UDC not being OTG compatible and missing some hardware
>>>> registers, this musb glue layer is written from scratch to be used in
>>>> gadget mode only and take silicon design specifics into account.
>>>>
>>>> Signed-off-by: Apelete Seketeli <apelete@seketeli.net>
>>>> Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
>>>> ---
>>>>
>>>>   drivers/usb/musb/Kconfig  |    8 +-
>>>>   drivers/usb/musb/Makefile |    1 +
>>>>   drivers/usb/musb/jz4740.c |  201
>>>>   +++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 209
>>>>   insertions(+), 1 deletion(-)
>>>>   create mode 100644 drivers/usb/musb/jz4740.c
>>>>
>>>> diff --git a/drivers/usb/musb/Kconfig b/drivers/usb/musb/Kconfig
>>>> index 57dfc0c..14d7e72 100644
>>>> --- a/drivers/usb/musb/Kconfig
>>>> +++ b/drivers/usb/musb/Kconfig
>>>> @@ -93,6 +93,12 @@ config USB_MUSB_BLACKFIN
>>>>
>>>>   config USB_MUSB_UX500
>>>>
>>>>          tristate "Ux500 platforms"
>>>>
>>>> +config USB_MUSB_JZ4740
>>>> +       tristate "JZ4740"
>>>> +       depends on MACH_JZ4740 || COMPILE_TEST
>>>> +       depends on USB_MUSB_GADGET
>>>> +       depends on USB_OTG_BLACKLIST_HUB
>>>> +
>>>>
>>>>   endchoice
>>>>
>>>>   config USB_MUSB_AM335X_CHILD
>>>>
>>>> @@ -100,7 +106,7 @@ config USB_MUSB_AM335X_CHILD
>>>>
>>>>   choice
>>>>
>>>>          prompt 'MUSB DMA mode'
>>>>
>>>> -       default MUSB_PIO_ONLY if ARCH_MULTIPLATFORM
>>>> +       default MUSB_PIO_ONLY if ARCH_MULTIPLATFORM || USB_MUSB_JZ4740
>>>
>>> Just out of curiosity, why can't we use DMA?
>>
>> We wrote the musb glue layer for the Ben Nanonote handheld computer
>> only to provide ethernet-over-usb functionality.
>>
>> I was thinking about adding DMA support later, but Lars-Peter Clausen
>> did experiment with DMA, and found out that all ethernet packets were
>> unaligned, while the DMA only works if they are aligned.
>> As a consequence, the DMA was bypassed all the time, making it useless
>> at least for the ethernet gadget we are interested in.
>>
>> That's why DMA didn't make it into the glue layer.
>
> DMA support would be nice to have.
> Especially to make the USB storage gadget work well with the Ben NanoNote.
>

It is not clear at this point how well the DMA works. While there seems support 
for DMA, there is no documentation about it in the datasheet, which might 
indicate that it is broken. So we first need to make sure that it works 
reliable before we can enable it.

- Lars