[Intel-gfx] [PATCH 00/21] i915/display: split and constify vtable

[Jani Nikula]

On Wed, 08 Sep 2021, Dave Airlie <airlied at gmail.com> wrote:
> This is orthogonal to my display ptr refactoring and should probably
> be applied first.

Yeah, overall nice cleanups, and a much easier bandwagon to jump onto
than the other one. ;)

Nothing too bad, a few bugs had crept in, and I had some nitpicks.

> The display funcs vtable was a bit of mess, lots of intermixing of
> internal display functionality and interfaces to watermarks/irqs.
>
> It's also considered not great security practice to leave writeable
> function pointers around for exploits to get into.

On the one hand I get this, but on the other hand the pointers to the
structs do remain writable. I suppose it increases the complexity of an
exploit by some margin?

In any case, I think this is cleaner in general, and that's enough merit
for the change, regardless of the security aspect.

BR,
Jani.

>
> This series attempts to address both problems, first there are a
> few cleanups, then it splits the function table into multiple pieces.
> Some of the splits might be bikesheds but I think we should apply first
> and merge things later if there is good reason.
>
> The second half converts all the vtables to static const structs,
> I've used macros in some of them to make it less messy, the cdclk
> one is probably the worst one.
>
> Dave.
>
>

-- 
Jani Nikula, Intel Open Source Graphics Center