Re: [GIT PULL] io_uring fixes for 6.0-rc1

[Linus Torvalds <torvalds@linux-foundation.org>]

On Fri, Aug 12, 2022 at 3:24 PM Keith Busch <kbusch@kernel.org> wrote:
>
> I'd prefer if we can get away with forcing struct kiocb to not grow. The below
> should have the randomization move the smallest two fields together so we don't
> introduce more padding than necessary:

I like this concept, but I think you might hit issues on 32-bit, where
"loff_t" can be a 64-bit entity with 64-bit alignment, and then you
see the same "oops, now it grew because loff_t moved" there.

Note that you'd never see that on x86-32, because - for strange
historical reasons - even 64-bit fields only have 32-bit alignment,
but other 32-bit architectures don't act that way.

Honestly, I think maybe we should just stop randomizing kiocb.

Because it has six fields, and basically three of them we wouldn't
want to randomly move around.

At some point, randomization no longer even matters.

            Linus