Christoph, > Now that we are using REQ_OP_WRITE_ZEROES for all zeroing needs in the > kernel there is very little use left for REQ_OP_WRITE_SAME. We only > have two callers left, and both just export optional protocol features > to remote systems: DRBD and the target code. While I'm not particularly married to WRITE SAME, I do think it's a shame that the RAID5/6 code never started using it. It does make a difference when initializing RAID sets. The other thing that keeps me a bit on the fence is that a bunch of the plumbing to handle a bio with a payload different from bi_size is needed for the copy offload token. I'm hoping to have those patches ready for 4.13. Right now there are a bunch of places where handling of REQ_OP_COPY_IN and REQ_OP_COPY_OUT share conditionals with WRITE SAME. So I suggest postponing the decision about whether to rip out WRITE SAME until I finish the token stuff. -- Martin K. Petersen Oracle Linux Engineering
