On Sunday, May 20, 2007, Pete Zaitcev wrote: > On Fri, 18 May 2007 14:23:43 -0700, "Mark Miesfeld" <[EMAIL PROTECTED]> wrote: > > > There is a published errata for the PPC440EPX, USBH_23: EHCI and OHCI > > Linux module contention. > > So, you're set for the classic (and unsoluble) inter-module problem, > but only if you attempt to keep parts accessing OHCI inside > the ohci-hcd. > > In short, this is the part which knows when to poke OHCI: > > > > > - } else > > +#ifdef CONFIG_USB_PPC440EPX_USBH_23_ERRATA > > + /* ensure 440EPX ohci controller state is operational */ > > + ohci_ppc_set_hcfs(HCFS_OPERATIONAL); > > +#endif > > And this is the part which implements poking: > > > + > > + /* write the new state and flush the write. */ > > + ohci_writel(cached_ohci, hc_control, > &cached_ohci->regs->control); > > + (void) ohci_readl(cached_ohci, &cached_ohci->regs->control); > > It looks to me that the second part does not have to be located in > ohci-hcd. On PPC, ohci_writel only uses the ohci private struct > to determine the endianness, presumably known in your case. So, > plain readl can be used. This only leaves the address for readl, > which comes from ioremap. If PPC allowed aliased ioremaps, you'd > be golden: just do a second ioremap. Can you do that? Please give > it a thought.
Yes, that would be much better. I'll look into doing that. If I can't, I'll use your idea below. > > +struct ohci_hcd *cached_ohci = NULL; > > This is just nasty. > > If aliased ioremaps are not possible, you better create a global > symbol for this in the hcd.c or other core part. Fill the word > when ohci probes, let ehci use it. Then, see above. Thanks for the good suggestions. -- Mark ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ linux-usb-devel@lists.sourceforge.net To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
