On Mon, 2016-04-18 at 23:35 +0200, Jan Kara wrote: > <> Hi Jan, I've noticed that patches 1 through 12 of your series are relatively independent, and are probably more stable than the remaining part of the series that actually changes locking. My dax error handling series also depends on the patches that change zeroing in DAX (patches 5, 6, 9). To allow the error handling stuff to move faster, can we split these into two patchsets? I was hoping to send the dax error handling series through the nvdimm tree, and if you'd like, I can also prepend your patches 1-12 with my series. Let me know your preference. Thanks, -Vishal��.n��������+%������w��{.n�����{���)��jg��������ݢj����G�������j:+v���w�m������w�������h�����٥