Ok, at worst we'll have to revert it, but before doing that, can you set
up a bugzilla entry with a before-and-after version of "lspci -vv", full
dmesg, and the output of /proc/iomem and /proc/ioports?
I suspect there wasn't a _real_ collision there, but the allocation failed
because of the alignment bits not being set up right for cardbus bridges,
but I'm not seeing the bug right now.
Ivan?
Linus
--