Re: [PATCH 03/11] fs: add new read_uptr and write_uptr file operations

[Christoph Hellwig]

On Mon, Jun 29, 2020 at 11:29:22AM -0700, Linus Torvalds wrote:
> I didn't look at the compat cases, but if anything I'd expect those to
> become simpler by having kernel pointers. And there doesn't actually
> seem to be that many of them (possibly because the "int" case si so
> common that it all ends up being the same?)

Having resurrect my work there really are tons of int cases.  Which
makes me thing that splitting out a setsockopt_int method which gets
passed value instead of a pointer, then converting all the simple cases
to that first and then doing the real shit later sounds like a promÑsing
idea.  Let me think a bit more about that.

And yes, a lot of the common methods have tons of cases and
sub-dispatchers and everything else you'd expect from an ioctl-like
interface..