Re: [PATCH 1/4] mm: introduce 'encoded' page pointers with embedded extra bits

[Linus Torvalds]

On Wed, Nov 9, 2022 at 10:00 AM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> But while I think a BUILD_BUG_ON() would be the right thing to do
> here, I do not feel confident enough to really put that to the test.

Oh, what the hell.

Just writing that whole explanation out made me just go "let's try to
re-organize it a bit so that we *can* inline everything, and see how
well it works".

And it does actually work to use BUILD_BUG_ON(), both with gcc and clang.

At least that's the case with the versions of gcc and clang _I_ use,
and in the configurations I tested.

So now I have a slightly massaged version of the patches (I did have
to move the 'encode_page()' around a bit), which has that
BUILD_BUG_ON() in it, and it passes for me.

And I find that I really like seeing that whole page pointer encoding
be so obviously much stricter. That was obviously the point of the
whole separate type system checking, now it does bit value validity
checking too.

So I'll walk through my patches one more time to check for it, but
I'll post it as a git branch and send out a new series (and do it in a
separate thread with a cover letter, to not confuse the little mind of
'b4' again).

If it turns out that some other compiler version or configuration
doesn't deal with the BUILD_BUG_ON() gracefully, it's easy enough to
remove, and it will hopefully show up in linux-next when Andrew picks
it up.

                  Linus