On Wed, Nov 03, 2010 at 05:28:26PM +0100, Valentin Longchamp wrote: > Hi, > > I am working on a driver for a device using a PCI interface on some > embedded hardware. > > Our hardware is no x86 and thus we have no BIOS to setup the PCI config > registers (let's take BAR0 for instance). On our devboard, these > registers are configured though values contained in a serial EEPROM that > is read by the device at power-up. > > Now I would like to be able to modify some of these registers so that I > can map other physical memory regions than the ones setup by the EEPROM. > > When I perform these writes (with pci_write_config_xxx), the physical > registers in the PCI config space is updated (at least I read the > corresponding values with pci_read_config_xxx). However, the value > returned by the call to pci_resource_start for the corresponding BAR is > not updated and still is the initial value written from the EEPROM. I > guess that this is due to the fact that the value read by > pci_resource_start is only read in hardware once, at device > enumaration/discovery on the bus. > > How can my driver write new values so that they are the ones actually > used by the kernel pci code ? Or how can I trigger a "refresh" of all > these values in the corresponding structs for a given device ? Couple of quick thoughts on this: 1) PCI subsystem was designed with the idea that BARs are "input" to the drivers and not something a PCI driver should be messing with. The "BIOS" is just a conceptual idea that some machine specific code knows where stuff can/should be mapped physically. SPARC, PARISC, Alpha, Arm, etc all have no "BIOS" - they have the equivalent though. Such code is NOT in any "PCI device driver" since those are by design expected to be platform/machine agnostic. 2) Look at PCI or PCMCIA hotplug support and you'll find support to dynamically reassign MMIO/IO Port resources. You might have user space generate a "fake" a hotplug event in order to get the resource assignment "fixed up" properly without having such code in your device driver. hth, grant > Thank you for your help > > Valentin Longchamp > -- > 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 -- 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
