On Wed, 2009-02-18 at 19:34 +0300, Ivan Kokshaysky wrote: > On Wed, Feb 18, 2009 at 01:56:44AM +0800, Yu Zhao wrote: > > + max = pci_resource_limit(dev, resno); > > + > > /* First, try exact prefetching match.. */ > > - ret = pci_bus_alloc_resource(bus, res, size, align, min, > > + ret = pci_bus_alloc_resource(bus, res, size, align, min, max, > > IORESOURCE_PREFETCH, > > pcibios_align_resource, dev); > > This breaks resource allocation on alpha and probably other > arches with multiple pci domains. You cannot assume that resource > start is equal to the bus address of the respective BAR. > Even 32-bit BARs can have a resource->start well above 4G. Than the previous patch was wrong too. Perhaps pci_resource_limit() should translate its result to the address as seen by the CPU. It should walk the bridge tree to the root and apply all conversions made by the bridges. If all conversions are just adding the upper address lines, then max could be just moved to the same 4G area as min: max += (min & ((u64)-1 - max)) -- Regards, Pavel Roskin -- 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
