On Mon, 2004-09-13 at 22:51, Lothar Wassmann wrote: > The PXA processor doesn't have something like I/O address space. > As far as I understand your device is a real PCI device. So you > probably wouldn't use a platform_device as base, but a pci_dev. > You should have a look at drivers/usb/core/hcd-pci.c. That's where > part of my code in ohci-sl811.c comes from. ;)
Yes it is, though I'll try to make it as platform indipendent as possible. I'm keeping the pci-specific code separate just like ohci-pci.c. Apart from that, I do have a real PCI device, that implements a USB HC with isp1160 and a PLX PCI bridge, as well as an Altera FPGA, that I don't know it's purpose... Anyway, I've studied already hcd-pci.c and I use it's functions in my isp1160-pci.c code (that is #included in a #IFDEF PCI). But my question persists: in hcd-pci.c lies a function named usb_hcd_pci_probe(struct pci_dev, const struct pci_device_id) that I use for probing the pci device. So I create a pci_dev from my hc_driver struct, a proper pci_device_id and call the function. Inside this function there is a branch that checks the driver flags for bit HCD_MEMORY (0x0001) which states that the HC regs use memory (else I/O). If flags&HCD_MEMORY==1, as in the case of OHCI, the function calls request_mem_region and the ioremap_nocache for the region that was requested. Else, as in the case of UHCI, the function searches the resources of the PCI device for regions that have the flag IORESOURCE_IO (as opposed to the previous case where the flag is IORESOURCE_MEM, all defined in include/linux/ioport.h) and allocates the first matching region. No ioremap or something. I obviously use the second case (UHCI-like) since my command and data ports are accessed through IO ports... So, to end this long description, if I don't have a "struct regs" that is mapped to memory and accessed via read"x"/write"x", but rather strict IO ports accesses via in"x"_p/out"x"_p (where "x" stands for b,l,w etc.), ------> how can I use the same method with you? <------ I mean, up until now, I have defined all the register addresses and functions like [read/write]_reg[16/32](reg_address, [value]) to use inw_p and outw_p to get or set the appropriate value, and this differentiates a lot my work from OHCI, and is my main problem up until now. Is there a way to map IO ports to memory and issue commands through readw/writew fucntions??? I really want to work again for my driver and this keeps me from going. Thanks, Dimitris ------------------------------------------------------- This SF.Net email is sponsored by: thawte's Crypto Challenge Vl Crack the code and win a Sony DCRHC40 MiniDV Digital Handycam Camcorder. More prizes in the weekly Lunch Hour Challenge. Sign up NOW http://ad.doubleclick.net/clk;10740251;10262165;m _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
