Hans Reiser wrote:
> I didn't say, do the meta-stat-data access with an iget, I meant let
> the FS decide how to do it. If it does it with iget, fine, but what I
> want to see is a general interface which can access arbitrary
> non-filebody data (which the FS can store where it chooses and fetch
> how it chooses).
Fine, but I think that's pretty separate from the d_type change which is
specifically "something returned by readdir".
> This way we can handle ACLs and other things as yet undreamt of in a
> symmetrical manner. We'll produce some code to do this sometime in
> the next year I think, and then put it out for comment and review.
Does it really make sense to avoid reading inodes for ACLs and other
file attributes? I think not -- that's exactly what inodes are for.
> You and James
That's "Jamie". Really, it's the name on my birth certificate.
> provided good answers to the issues I was concerned about, I guess I
> just lack enthusiasm for more changes to VFS interfaces while we are
> trying to port to the changes already made, and I am wrong in this.
The VFS interface change is very, very simple. Just #include
<linux/dirent.h> and insert DT_UNKNOWN as the second argument to
filldir() in your readdir() function.
You can do more, but it's not required. It might not even be worthwhile
with Reiserfs, depending on how the fs is laid out on disk.
If my patch goes into the kernel, I'll send you an update for Reiserfs
if you like.
-- Jamie
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Mar 07 2000 - 21:00:23 EST