On Sun, 7 May 2000, Linus Torvalds wrote:
> Oh.
>
> If anybody whips out a patch to do the "struct file" move, and not have a
> "void *", then I'll apply that. No problem. I agree 100% with the fact
> that "struct file" is the fundamental IO entity - even if it turns out
> that most users don't actually need either the file nor the dentry to
> actually do the physical IO (ie all "unix-like" filesystems will always be
> happy with just the inode - that is, to some degree, what "unix-like"
> really means).
Interesting... What, in your opinion, makes pagecache unsuitable for stuff
like inode table (done the same way as regular files on quite a few filesystems)
or EAs on NTFS/HPFS/whatnot or forks on HFS, etc.? If it _is_ suitable - OK,
which struct file might be naturally associated with these objects? Before
you say that these case are pathologies consider directories on good ol'
ext2. No opened files around in that case.
That's what I find ugly about struct file * - yes, void * is a cop-out and
we might need to rethink the separation of code in mm/filemap.c. But IMO
requiring struct file * for every pagecache is not a good idea.
Cheers,
Al
-
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 : Sun May 07 2000 - 21:00:21 EST