PCI changes and fixes for 2.5.72
Quoted 11 lines · Greg
[PATCH] PCI: Unconfuse /proc
If we are to cope with multiple domains with clashing PCI bus numbers, we must refrain from creating two directories of the same name in
/proc/bus/pci. This is one solution to the problem; busses with a non-zero domain number get it prepended.Alternative solutions include cowardly refusing to create non-domain-zero bus directories, refusing to create directories with clashing names, and sticking our heads in the sand and pretending the problem doesn't exist.
Please don't do this. It's gross. As long as we have the bus number mangling, stuff can stay as it is. When the bus number mangling goes, the old-style entries can go as well. I'm working on a patch that makes the old-style entries be symlinks like this:
../../pci%d/bus%d/dev%d/fn%d/config-space
That solves the problem for good, in the right way. It allows for migration to something sane.
Quoted 14 lines · Albert Cahalan
Quoted 11 lines · Greg
[PATCH] PCI: Unconfuse /proc
If we are to cope with multiple domains with clashing PCI bus numbers, we must refrain from creating two directories of the same name in
/proc/bus/pci. This is one solution to the problem; busses with a non-zero domain number get it prepended.Alternative solutions include cowardly refusing to create non-domain-zero bus directories, refusing to create directories with clashing names, and sticking our heads in the sand and pretending the problem doesn't exist.
Quoted 7 lines
Please don't do this. It's gross. As long as we have the bus number mangling, stuff can stay as it is. When the bus number mangling goes, the old-style entries can go as well. I'm working on a patch that makes the old-style entries be symlinks like this:
../../pci%d/bus%d/dev%d/fn%d/config-space
Symlinks from what to that new file? Have the result from 'tree' to show what you are considering?
And until we have such a change, the patch from Matthew is needed, so it should stay.
thanks,
greg k-h