On Wed, Jan 30, 2019 at 08:28:14AM -0800, Bart Van Assche wrote: > It would help if you could explain why you are so strongly opposed against > keeping BIDI support in the kernel. It creates a barely tested code path all over the SCSI and block layers, including a userspace attack vector. At the same time it sees zero real life usage - the only people touching it are less than a handful people just testing it. And yes, removing it also helps to shrink struct request, which is used by all block drivers, and removes additional code from the SCSI (not NVMe) as well as the block layer I/O path.
