Messages in this thread | | | Date | Tue, 29 Jul 1997 16:38:28 -0700 (PDT) | From | Linus Torvalds <> | Subject | Re: Msdos patch for aliased names |
| |
On Wed, 30 Jul 1997, Alain Knaff wrote: > > In which way does it slow down the critical path? Is it the > kmalloc() in msdos_canonize that you're worried about, or is it > something else?
No, I'm not worried about what happens inside the call, I'm worried about the testing whether we should do the call at all. And what I'm more worried about is actually the fact that the call can change the dentry outside the VFS layer - something I don't like from a conceptual standpoint.
I'd like _all_ name lookup logic to be in the VFS layer so that there aren't any surprises. And I do see that people would prefer for the msdos filesystem to be case-insensitive..
I certainly agree that my stupid "name_translate_char()" is by no means very good, and as you correctly point out we should actually change the character itself, not just the hash value. And we can't really do that in the place I envisioned for it, because the buffer that the name is in can be a symbolic link or similar, and we obviously shouldn't change that..
Hmm..
I'd like this al to be done in some sane place - and your patch works but doesn't count as "sane" in my book. It looks like a hack (having people malloc memory and exchange it for non-malloced stuff, having somebody _else_ free it etc). The iso9660 filesystem used to do a lot of ugly stuff like that for directory handling (happily cleaned up these days), and I let things like that pass, but I don't like it when it's in the kernel proper..
Linus
| |