On Monday 08 April 2013, Thomas Petazzoni wrote: > Dear Arnd Bergmann, > > On Mon, 8 Apr 2013 23:34:12 +0200, Arnd Bergmann wrote: > > > > No, I'm assuming PCIBIOS_MIN_IO is always 0. So presumarly, this should > > > be something like: > > > > > > pcie->realio.end = min(PCIBIOS_MIN_IO + > > > resource_size(&pcie->io), > > > IO_SPACE_LIMIT); > > > > > > > Normally PCIBIOS_MIN_IO is 0x1000, since the first 4096 ports are reserved > > for ISA and PCMCIA compatible drivers and should not be assigned to > > PCI devices. So the first port should get ports 0x1000 to 0xffff, later > > ones can used the entire 65536 ports e.g. 0x10000 to 0x1ffff. > > Then I guess it should work with the code I'm proposing here, no? > > Note: this pcie->realio region is global: it will be shared by all PCIe > interfaces. I think it's still wrong, unless you guarantee that resource_start(&pcie->io) is the same as PCIBIOS_MIN_IO. Why don't you just read the start and end values from the ranges property? I assume you want to run with io_offset=0, so you really need pcie->realio.type = IORESOURCE_IO; pcie->realio.start = max(PCIBIOS_MIN_IO, range->pci_addr); pcie->realio.end = max(IO_SPACE_LIMIT, range->pci_addr + range->size); Arnd -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
