Russell King wrote:
>
> Philip Blundell said:
> > Again this sounds like it might be a DMA cache coherency problem. The 3c59x.c
> > driver doesn't seem to have had the appropriate surgery yet as far as I can
> > see. Did you try running with the D-cache disabled like Dave suggested?
>
> The 3c59x driver does not have any Dcache coherency problems - I've
> been running my 3c592 Vortex card with it for 6 months with no problems
> on my EBSA285.
Sorry, but It does, at least with 3c905B Cyclone
Or, maybe am I initialising it badly (in pcibios_fixup_ebsa285()):
#define ETH_IO_BASE 0x401
#define ETH_MEM_BASE 0x00800000
if
(dev->vendor==PCI_VENDOR_ID_3COM||dev->vendor==PCI_VENDOR_ID_AMD) {
pci_set_base_addr(dev, 0, ETH_IO_BASE);
pci_set_base_addr(dev, 1, ETH_MEM_BASE);
pci_set_cmd(dev, 0, PCI_COMMAND_MASTER |
PCI_COMMAND_MEMORY | PCI_COMMAND_IO);
/* Put the chip to sleep in case the driver
isn't loaded */
pci_write_config_dword(dev, 0x40, 0x80000000);
} else pci_set_cmd(dev, 0, PCI_COMMAND_MASTER);
unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]
