https://bugzilla.kernel.org/show_bug.cgi?id=198123
--- Comment #12 from Daniel Vetter (daniel@ffwll.ch) --- Created attachment 273553 --> https://bugzilla.kernel.org/attachment.cgi?id=273553&action=edit fix of-by-one, or what looks like one
This is another theory that crossed my mind. Please test both this patch here (on latest upstream) and the previous patch on the first bad commit, I'd like to know the outcome of both experiments.