On 10/31/2017 05:21 PM, Dan Williams wrote: > DAX support in brd is awkward because its backing page frames are > distinct from the ones provided by pmem, dcssblk, or axonram. We need > pfn_t_devmap() entries to fully support DAX, and the limited DAX support > for pfn_t_special() page frames is not interesting for brd when pmem is > already a superset of brd. Lastly, brd is the only dax capable driver > that may sleep in its ->direct_access() implementation. So it causes a > global burden with no net gain of kernel functionality. > > For all these reasons, remove DAX support. Reviewed-by: Jens Axboe <axboe@xxxxxxxxx> -- Jens Axboe
