On Wed, Mar 02, 2022 at 11:52:49AM +0000, Catalin Marinas wrote:
On Wed, Mar 02, 2022 at 10:12:09AM +0000, Marc Zyngier wrote:
I don't think that's on purpose, and we should definitely address this. Maybe we should have a warning if we spot an occurrence of .width being 0.
We should indeed have a check. Alternatively, assume the default to be 4 and convert all 0s to 4 during boot (less patch churn).
That's got the issue that it becomes very easy to end up defaulting to a width of 4 in the case where the field is narrower which isn't going to be easily visible in testing until it encounters a system where it misfires in a way that breaks things.