On Tue, 12 May 2026 at 01:59, Joonas Lahtinen
<[email protected]> wrote:
>
> Commit is no-op as per discussion in the linked mail thread.
> The debated UAF scenario has not been possible in mainline kernel
> since 2021.
That was never the issue.
The issue was that the code used to do this:
if (likely(vma && vma->vm == vm))
vma = i915_vma_tryget(vma);
and the vma->vm == vm condition was nonsensical - and if it ever was
false, that code was all horribly buggy.
So the "since 2021" is complete nonsense.
It's only since April this year - commit a13edf9b92fc ("drm/i915/gem:
Drop check for changed VM in EXECBUF") that the "else" side is now
pointless and the revert makes sense.
Stop writing misleading commit logs and excuses for bad code. There is
no "since 2021".
Linus
