Re: [CFT][PATCH] ramfs/tmpfs readdir()

[Alexander Viro]

On Sun, 4 Nov 2001, Linus Torvalds wrote:

> This is better than what we have, though, so I wouldn't object to the
> patch. I wonder why you export the internal dcache functions, though? The
> only thing that _should_ need exporting is "dcache_dir_ops", no? We don't
> want other filesystems mucking with the internals of this, as far as I can
> tell.

I would love to.  But autofs4 doesn't allow that, what with ioctls on
root directory...  It's the only place that uses individual functions,
though - the rest just uses dcache_dir_ops.

> Admittedly linear traversal is the _common_ case, and arguably the much
> more important of the two. However, right is right, and a true quality
> implementation gets seekdir/telldir right too.
> 
> Have you looked at how nasty the d_offset thing would be?

I'll look into that, but frankly, I'm not too optimistic.