On Mon, 7 Nov 2022 11:33:46 -0500 Brian Foster <bfoster@xxxxxxxxxx> wrote: > Perhaps a good starting point would be to wrap the check in > this patch with a WARN_ON_ONCE() and let it soak in -next for a while? > That would avoid excessive noise from repetitive callers [1] but still > allow those callsites to be identified/fixed. If there is some really > weird fdatawrite-only caller that conflicts, the change could always be > loosened up from there (as unlikely as that seems).. Hm? Sounds reasonable. Please let's be clear in the changelog why we're adding this. I mean, we could add a zillion checks everywhere for misbehaving callers. Why choose this one place in particular?
