On Fri, Nov 25, 2011 at 09:27:43AM +0100, Paolo Bonzini wrote: > No, it adds support for 512b logical block sizes running on a host with 4k > logical block size and cache=none. This is suboptimal as it requires > bounce buffers, but it can happen with migration and until libvirt provides > a knob for the guest's logical block size.
It's also highly dangerous becausw the fs may assume a sector is written atomically, but now it isn't. If at all I would recommend to only support this case read only.
