On 05/05/2011 04:10 PM, Jan Andersson wrote: > On 05/05/2011 04:04 PM, Alan Stern wrote: >> On Thu, 5 May 2011, Jan Andersson wrote: >> >>>> Something else to think about... The uhci_grlib_reset_hc() and >>>> uhci_grlib_check_and_reset_hc() routines are all generic -- there's >>>> nothing specific to the GRUSBHC about them. They could go into >>>> uhci-hcd.c. >>>> >>>> The reason they aren't there now is because they are part of the BIOS >>>> handoff needed by the early PCI code, even before uhci-hcd might be >>>> loaded. Therefore a copy had to go into pci-quirks.c, and there was >>>> no reason to duplicate it in uhci-hcd.c. >>>> >>>> Since you don't have to worry about taking control from the firmware, >>>> the routines don't need to be coded separately for each platform. You >>>> can put a generic version in uhci-hcd.c, and if the build is PCI-only >>>> then the generic routine can simply call the pci-quirks code. >>>> >>> >>> Thank you. I will create generic functions in uhci-hcd.c instead. >>> >>> I assumed that the pci_write_config_word()s in uhci_*reset_hc would be >>> needed in the case where GRUSBHC was used and there also was another >>> UCHI controller connected via PCI in the system. After looking at the >>> spec I suppose that it will work, as you say, without the writes to the >>> legacy support register. >> >> Unless somebody decides to run a non-PCI-UHCI-enabled kernel in an x86 >> PC. Maybe to be safe it would be best to rule out that combination in >> the Kconfig: make USB_UHCI_SUPPORT_NON_PCI_HC depend on (USB_UHCI_HCD >> && !X86). Would this be acceptable? >> > > I am fine with that. Too bad that I managed to send V2 before seeing > this message :-). I will wait for other comments on V2 before adding > this in V3. > Hmmm, after a bit more thought: I someone wants to run a non-PCI-UHCI-enabled kernel in an x86 PC we would need one more config option. USB_UHCI_SUPPORT_NON_PCI_HC is currently needed to include the necessary stuff for the non-PCI HC so that one must be set. One additional option could be introduced to still include the accesses to the legacy support register. Another way to solve this would be to add defined(CONFIG_X86) in uhci-pci.c (change to patch 6 in V2): drivers/usb/host/uhci-pci.c index c300bd2..0b7d347 100644 @@ -17,6 +17,14 @@ * (C) Copyright 2004-2007 Alan Stern, stern@xxxxxxxxxxxxxxxxxxx */ +/* + * If we are running on a system with keyboard and mouse legacy support we use + * the reset functions from pci-quirks. Otherwise we use generic functions from + * uhci-hcd.c + */ +#if !defined(CONFIG_USB_UHCI_SUPPORT_NON_PCI_HC) || defined(CONFIG_X86) +/* System that may require keyboard and mouse legacy support */ + #include "pci-quirks.h" /* @@ -55,6 +63,13 @@ static void uhci_pci_configure_hc(struct uhci_hcd *uhci) Best regards, 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
