On 04/05/11 21:59, Alan Stern wrote:
> On Wed, 4 May 2011, Jan Andersson wrote:
>
>> This patch adds support for the UHCI part of the GRLIB GRUSBHC controller
>> found on some LEON/GRLIB SoCs.
>>
>> The UHCI HCD previously only supported controllers connected over PCI.
>> This patch adds support for the first non-PCI UHCI HC. I have tried to
>> replicate the solution used in ehci-hcd.c.
>>
>> This patch also extends the uhci_{read,write}* functions to allow accesses
>> to registers not mapped into PCI I/O space. This extension also includes
>> the addition of a void __iomem pointer to the uhci structure.
>
> I would prefer to see the new functions and the __iomem pointer added
> as a separate preliminary patch. They are logically separate from the
> GRUSBHC support.
>
>> Tested on GR-LEON4-ITX board (LEON4/GRLIB with GRUSBHC) and x86 with Intel
>> UHCI HC.
>>
>> Signed-off-by: Jan Andersson <jan@xxxxxxxxxxx>
>> ---
>>
>> Right now the choice of I/O function to use when supporting non-PCI is
>> (for instance):
>>
>> static inline u16 uhci_readw(struct uhci_hcd *uhci, int reg)
>> {
>> if (uhci_has_pci_registers(uhci))
>> return inw(uhci->io_addr + reg);
>> else
>> return readw(uhci->regs + reg);
>> }
>>
>> Another solution would be to use:
>>
>> static inline u16 uhci_readw(struct uhci_hcd *uhci, int reg)
>> {
>> return uhci->readw(uhci, reg);
>> }
>>
>> where uhci->readw is a function pointer that could be setup by
>> a probe function to either point at ehci_pciio_readw (which would
>> containt an 'inw') or ehci_memio_readw (which would contain a readw).
>
> "ehci"?
>
Typo, it should say uhci.
>> I did not implement this as it leads to more code being added. When/If
>> support for big endian registers is added then perhaps it would be better
>> to use function pointers when supporting non-PCI hosts.
>
> Seems like a reasonable choice.
Thanks. I plan to add support for big endian mmio and descriptors later
on and will change to function pointers then.
>
>> +config USB_UHCI_SUPPORT_NON_PCI_HC
>> + bool
>> + depends on USB_UHCI_HCD
>> + default y if SPARC_LEON
>> + default n
>
> This last line is not necessary. n is the "default" default. :-)
>
Great, I will remove the last line.
Thank you very much for reviewing. I try to address all of your and
Sergei's comments and send a V2 after I have slept and tested.
Thanks!
Jan
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
