On Tue, 2013-10-22 at 14:34 +0100, Mark Rutland wrote: > > If the patch you referenced changed the layout without maintainng backwards > compatibility, it sounds like it is broken and has caused a regression. > > This is a work around, not a fix. Kind of. Yes, the best answer right now at -rc6 might be to just revert the functional change and *always* use legacy_set_geometry(). And the name of the proposed new DT property could certainly do with some improvement. But fundamentally, the "Linux-specific" bit was already there as an implicit assumption in the bindings, because there was a de facto standard for how you lay out the ECC "as legacy_set_geometry() does it". Now we want to set an option to indicate that it should be automatically calculated from the properties of the chip (as set_geometry_by_ecc_info does it), rather than done the old way. And it's not unreasonable to have that as a DT property. We're no *more* Linux-specific than we were before. But perhaps it's too late for 3.12 to do that? -- David Woodhouse Open Source Technology Centre David.Woodhouse@xxxxxxxxx Intel Corporation
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
