> When a PCI device header type field bit 7 is set, the device is a multi-
> function device and all the functions must be scanned as well. I see that
> the pci.c code does attempt to do that, apparently, but it doesn't seem
> to work.
It seems there is a minor bug in the PCI scanning code in 2.1 causing
it to fail when different functions of a single device have different
header types. Please try the following patch:
--- drivers/pci/pci.c.mj Thu Mar 19 14:09:37 1998
+++ drivers/pci/pci.c Thu Mar 19 14:14:32 1998
@@ -108,7 +108,7 @@
__initfunc(unsigned int pci_scan_bus(struct pci_bus *bus, unsigned long *mem_startp))
{
unsigned int devfn, l, max, class;
- unsigned char cmd, irq, tmp, hdr_type = 0;
+ unsigned char cmd, irq, tmp, hdr_type, is_multi = 0;
struct pci_dev *dev;
struct pci_bus *child;
int reg;
@@ -119,12 +119,13 @@
max = bus->secondary;
for (devfn = 0; devfn < 0xff; ++devfn) {
- if (PCI_FUNC(devfn) == 0) {
- pcibios_read_config_byte(bus->number, devfn, PCI_HEADER_TYPE, &hdr_type);
- } else if (!(hdr_type & 0x80)) {
+ if (PCI_FUNC(devfn) && !is_multi) {
/* not a multi-function device */
continue;
}
+ pcibios_read_config_byte(bus->number, devfn, PCI_HEADER_TYPE, &hdr_type);
+ if (!PCI_FUNC(devfn))
+ is_multi = hdr_type & 0x80;
pcibios_read_config_dword(bus->number, devfn, PCI_VENDOR_ID, &l);
/* some broken boards return 0 if a slot is empty: */
Have a nice fortnight
-- Martin `MJ' Mares <mj@ucw.cz> http://atrey.karlin.mff.cuni.cz/~mj/ Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth "Black holes are where God divided by zero."- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu