Re: [Avf-fuse-dev] userland filesystem inode numbers

[Al A. <al_...@ma...>]

Hi Miklos!

Sorry for the delayed reply.  It's been hectic.


>> Yes and no.  On the kernel interface level it does, but on the library
side it currently doesn't.  If you are willing to use the kernel API
then you can implement full unix file semantics, including files kept
open after unlink, etc...

I see.  I tried modifying fusexmp and filling in st_ino, but down deeper
it ended up ignoring it with a call to next_ino() anyway.  I guess that
explains it.

So I guess the question is, would there be a way to have fuse not generate
i-nums for my filesystem, but still work with other fuse filesystems that
require this?  Perhaps a flag that leaves i-num generation to the userland
FS?


>> You can also look at FunFS by Michael Grigoriev
(http://www.luminal.org/wiki/index.php/FunFS/FunFS), which I think
uses the kernel API directly and has a C++ library interface.  The
author can probably tell you much more.  Michael?

I checked that link out, and downloaded FunFS.  Interesting stuff.
Skimming through it, though, I couldn't find where FunFS interfaces with
FUSE, which is probably the part that I need to look at.  Michael? :)


>> Can I ask what is HSM?

Sure.  Hierarchical Storage Management.  Basically, the concept of
caching on a networked filesystem level.  It seems to me that this would
be a breeze to develop using FUSE.


Regards,
-Al

-- 
___________________________________________________________
Sign-up for Ads Free at Mail.com
http://promo.mail.com/adsfreejump.htm