You wrote: > While scanning through dmesg I found this line: > > i915 0000:00:02.0: BAR 6: can't assign [??? 0x00000000 flags 0x0] (bogus > alignment) > > I digged a bit and found that this happens when pci_map_rom() is called and > BAR6 is just 0. The question is: which's fault is this (if at all)? > > -the drivers fault: the driver should check if there is a ROM bar at all and > not call pci_map_rom() otherwise. In this case the warning message should > be more clear, like a WARN_ON or a message that this is a driver bug that > needs to be fixed. And the function doc needs to be updated. > > -pci_map_rom()'s, because it shoudl check itself. In this case the function > should check first if the resource has length 0 and then just return NULL > without any further message. The driver has to deal with the NULL case > anyway and can complain itself if that is unexpected. > > I'm fine either way, I just want to hear some opinions before doing a patch. > I would vote for #2. ping? Eike
Attachment:
signature.asc
Description: This is a digitally signed message part.
