>>>Seems following might be the required change - can you try it out?
>>	your patch does not seem to change the situation, unfortunately.
>>
>>	if i "break" instead of "continue" on the 2nd if statement,
>>	most of the pci devices (ethernet chips) are skipped but i can go to
>>	multiuser.  so there has to be something to do with the 2nd loop.
>
>	with this diff i finally got to multiuser with pci devices.
>	ok to commit? (does anyone knkow the name for 0x100b/0x0510?)
>
>	the panic occurred when we try to access id/csr/class on
>	function 1 of vendor 0x100b/prod 0x0510 (which claims to be
>	multifunction).  adding quirk table to make it non-multifunction
>	workaround the situation.

	documentation says it is indeed multifunciton device, so it may not be
	the right idea...  any ideas are welcome.

	function 0 (100b/0510): Bridge configuration for GPIO and LPC support
	function 1 (100b/0511): SMI status and ACPI support
	function 2 (100b/0512): IDE controller
	function 3 (100b/0513): XpressAUDIO
	function 4 (100b/????): reserved
	function 5 (100b/0515): X-Bus expansion support

itojun