On Mon, 9 May 2011, Jan Andersson wrote:
> This patch changes the uhci_{read,write}* functions into using function
> pointers from the uhci_hcd struct when accessing registers. This is done in
> order to reduce the amount of code that need to be ifdef:ed around
> uhci_{read,write}* when later adding (dynamic) support for big endian mmio.
>
> This change leads to a bit more work for the processor when running with
> a PCI-only UHCI kernel. Considering how relatively slow PCI accesses are
> it should not matter much, and it saves us the trouble of having a special
> case for PCI-only configurations in uhci-hcd.h.
>
> Also added 'const' to uhci_hcd argument.
>
> Signed-off-by: Jan Andersson <jan@xxxxxxxxxxx>
> ---
>
> My guess is that the change in this patch may not be to everyones (anyones?)
> liking. Another solution would be to use someting more similar to the previous
> version, like:
>
> #if !defined(CONFIG_USB_UHCI_SUPPORT_NON_PCI_HC)
> static inline u32 uhci_readl(struct uhci_hcd *uhci, int reg)
> {
> return inl(uhci->io_addr + reg);
> }
> #else
> #define uhci_has_pci_registers(u) ((u)->io_addr != 0)
> static inline u32 uhci_readl(struct uhci_hcd *uhci, int reg)
> {
> if (uhci_has_pci_registers(uhci))
> return inl(uhci->io_addr + reg);
> else
> return uhci->ureadl(uhci->regs, reg);
> }
> #endif
>
> ... or more similar to the solution in used for the EHCI HCD (leads to the
> largest amount of special case ifdefs, at least in my attempts, but no extra
> function call):
>
> #if !defined(CONFIG_USB_UHCI_SUPPORT_NON_PCI_HC)
> static inline u32 uhci_readl(struct uhci_hcd *uhci, int reg)
> {
> return inl(uhci->io_addr + reg);
> }
> #else
> #define uhci_has_pci_registers(u) ((u)->io_addr != 0)
> static inline u32 uhci_readl(struct uhci_hcd *uhci, int reg)
> {
> if (uhci_has_pci_registers(uhci))
> return inl(uhci->io_addr + reg);
> #if defined(CONFIG_USB_UHCI_BIG_ENDIAN_MMIO)
> elsif uhci_big_endian_mmio(uhci)
> return readl_be(uhci->regs, reg);
> #endif
> else
> return readl(uhci->regs, reg);
> }
> #endif
I would really like to see no change to the object code in the PCI-only
case.
For the various non-PCI cases, imitating EHCI may be messy but it has
the advantage of following a well-established precedent.
Alan Stern
--
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
