Bill Hawes (
Wed, 19 Nov 1997 09:24:35 -0500

Andrew E. Mileski wrote:
> The unique ID number is completely unrelated to block number.
> On an non-native UDF disk, the value could be anything, so this
> case has to be supported.
> I'm getting real tempted to throw together a VFS (not fs) patch
> for 64-bit inode numbers. It ought to be pretty simple, and completely
> compatible with the current implementation. You'd have to wait for a
> filesystem (say ext3) that supports them in order to make use of the
> extra elbow room.

There was a discussion a while back on the issue of 64-bit inode
numbers, and I recall it sounding like a can of worms. There seem to be
quite a few places where POSIX-mandated structures return 32-bit inode
numbers, so it's not something that can be changed lightly. Library
support and binary compatibility is an issue.

Perhaps you could just use the block number on the volume as an inode
number? You really want to go with something useful, and block numbers
will probably stay in the 32 bit range for a while.