On Tue, 16 Jan 2024 10:21:49 -0800 Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: > Here's a clue: just fix your inode numbers. > > I can think of many ways to do it. Here's a couple: > > - use a fixed inode number for all inodes. It's fine. Really. You might > confuse some programs that still do getpwd() the legacy way, but hey, > nobody cares > > - just put the inode number in the same data structure everything else is > > > - make the inode number be a hash of the address of your data structure. > That's actually the closest to a traditional "real" inode number, which is > just an index to some on-disk thing > > I'm sure there are other *trivial* solutions. > > None of this is an excuse to misuse sentries. > > Try the first one - one single inode number - first. You shouldn't be doing > iget() anyway, so why do you care so deeply about a number that makes no > sense and nobody should care about? It was me being paranoid that using the same inode number would break user space. If that is not a concern, then I'm happy to just make it either the same, or maybe just hash the ei and name that it is associated with. If I do not fully understand how something is used, I try hard to make it act the same as it does for other use cases. That is, I did all this to keep inodes unique and consistent because I did not know if it would break something if I didn't. Removing that requirement does make it much easier to implement readdir. I think I'll do the hashing, just because I'm still paranoid that something might still break if they are all the same. -- Steve
