Re: [RFC] persistent store (version 2) (part 1 of 2)

[Andrew Morton]

On Wed, 1 Dec 2010 22:00:12 -0800 "Luck, Tony" <tony.luck@xxxxxxxxx> wrote:

> > filenames refer to files!  And lo, that's what we have here.  A
> > filesystem!  The files are created by the kernel and are read and
> > unlinked by userspace.
> >
> > So why not implement this whole thing as a proper filesystem?
> ...
> > Wait.  It *is* a filesystem.
> ...
> > So why can't I remove these "records" with rm?
> 
> Because I tried to use /sys for this and couldn't find a
> way to get notified about "unlink".  Alan Cox called this
> bit "daft" in v1 (and I agreed with him). Peter Anvin gave
> me some pointers on how easy this would be to do as a real
> filesystem ... so v3 will be out in a little while with
> this insanity removed.

OK.  Yes, the correct answer is usually "create a new filesystem
driver" ;)

<greps>

gad, there are over 200 register_filesystem() callsites.

One thing which does leap out of the v2 implementation is the hardwired
assumption that there is one store kernel-wide.  I suppose that's OK as
a version-1 implementation detail thing, but we should avoid hardwiring
that assumption into the presentation of v1's userspace interfaces.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/