On Wednesday, November 05, 2014 at 07:00:32 AM, Peng Fan wrote: > 在 11/5/2014 1:33 AM, Marek Vasut 写道: > > On Tuesday, November 04, 2014 at 02:29:56 PM, Peng Fan wrote: > >> Hi Marek, > >> > >> 在 11/4/2014 7:01 PM, Marek Vasut 写道: > >>> On Tuesday, November 04, 2014 at 11:50:29 AM, Peng Fan wrote: > >>>> 在 11/4/2014 6:33 PM, Marek Vasut 写道: > >>>>> On Tuesday, November 04, 2014 at 08:50:00 AM, Peng Fan wrote: > >>>>>> Include a weak function board_ehci_usb_mode to gives board code > >>>>>> a choice. > >>>>> > >>>>> What choice? > >>>>> > >>>>>> If the board want the otg port work in host mode but not > >>>>>> device mode, this should be handled. > >>>>> > >>>>> How? > >>>>> > >>>>> Also, isn't usb_phy_enable() supposed to do exactly this kind of > >>>>> selection between device and host mode ? > >>>> > >>>> In mx6sxsabresd board, there are two usb port, one used for otg, the > >>>> other used for host. However they are connected to SOC USB controller > >>>> otg1 core and otg2 core respectively. Like following: > >>>> > >>>> OTG1 CORE <----> board otg port > >>>> OTG2 CORE <----> board host port > >>>> > >>>> However the board do not have ID pin set for board host port. If just > >>>> use usb_phy_enable, the board host port will not work, because > >>>> "type = usb_phy_enable(index, ehci) ? USB_INIT_DEVICE : > >>>> USB_INIT_HOST;" will always set type with USB_INIT_DEVICE. > >>>> > >>>> Because i did not find way to handle this situation in > >>>> board/freescale/mx6sxsabresd/mx6sxsabresd.c, add this function to let > >>>> board level code handle handle 'type', if board level code want to set > >>>> it's own 'type'. > >>> > >>> This part in usb_phy_enable() > >>> > >>> 163 return val & USBPHY_CTRL_OTG_ID; > >>> > >>> should be replaced by some kind of a board-specific callback then, with > >>> default implmentation being the above (reading the phy ctrl register). > >> > >> How about using the following piece of code? > >> in ehci-mx6.c > >> > >> unsigned int __weak board_usb_phy_mode(int index, unsigned int val) > >> { > >> > >> return val & USBPHY_CTRL_OTG_ID; > >> > >> } > >> > >> replace "return val & USBPHY_CTRL_OTG_ID;" using " > >> return board_usb_phy_mode(index, val);" > >> > >> In board file, > >> unsigned int board_usb_phy_mode(int index, unsigned int val) > > > > Why not pass in full struct usb_ehci * instead ? Passing some ad-hoc $val > > into the function doesn't seem like a scalable future-proof solution. > > [...] > > Passing struct usb_ehci * to board code seems exports ehci register > definition to board layer.
Yeah. > How about just use > "int board_usb_phy_mode(int index)" without using 'val' or 'struct > usb_ehci *ehci'. The board part might need to read the EHCI registers though. How would the board part be able to do it if you didn't pass the *ehci in ? _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot