On Tue, Sep 05, 2017 at 10:40:07AM -0400, Brian Foster wrote: > Thanks.. I agree that a rework of the optimization can come later now > that the bug is fixed. I agree in principle, but I don't think the patch in for-next is enough, given that it doesn't take ic_refcnt into account. > Christoph, are you planning to continue with your flushseq based patch? I think it is fundamentally the right thing to do, but given how much I've got on my plate I wouldn't mind someone else looking into it :)