Re: file metadata via fs API

[Linus Torvalds]

On Mon, Aug 17, 2020 at 10:15 AM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> So it has this very complex "random structures of random things"
> implementation. It's a huge sign of over-design and "I don't know what
> the hell I want to expose, so I'll make this generic thing that can
> expose anything, and then I start adding random fields".

You can see the overdesign in other places too: that "time
granularity" is some very odd stuff. It doesn't actually even match
the kernel granularity rules, so that fsinfo interface is basically
exporting random crap that doesn't match reality.

In the kernel, we give the granularity in nsec, but for some reason
that fsinfo stuff gives it in some hand-written pseudo-floating-point
format. Why? Don't ask me.

And do we really want to have that whole odd Nth/Mth thing?
Considering that it cannot be consistent or atomic, and the complaint
against the /proc interfaces have been about that part, it really
smells completely bogus.

So please. Can we just make a simple extended statfs() and be done
with it, instead of this hugely complex thing that does five different
things with the same interface and makes it really odd as a result?

                  Linus
So honestly,  there's a