(In reply to Jochen Hansper from comment #23) > > Sigh, hopefully not a case where added debugging obscures the bug. > > If you want to try something else, I'll try to test it. Thanks a lot for > looking into this. You could try building your own kernel, but without my patches, and run without "nogbpages", which can rule out the "self-compilation obscures the bug" case. > If I understand correctly, then setting nogbpages might work around the bug > on an unpatched kernel, too? I highly doubt that, but let's go step by step.