I seized on Mikulas's noclone patches and took it upon myself to add noclone support to bio-based dm-multipath. Pretty happy with how things turned out but also open to all review/feedback. I've verified that all mptest tests pass (with tcmloop, mptest's nvmefcloop based test setup panics with 5.0-rc6.. I should probably report and/or fix that). I've yet to do the before vs after multipath benchmarking to see if there is a performance win but I'm betting there will be (otherwise I wouldn't have done this work). I'll backfill benchmarking results later in the week. Mike Snitzer (6): dm: update dm_process_bio() to split bio if in ->make_request_fn() dm: eliminate 'split_discard_bios' flag from DM target interface dm: improve noclone bio support dm: add the ability to attach per-bio-data to dm_noclone bio dm: improve noclone_endio() to support multipath target dm mpath: add support for dm_noclone and its per-bio-data Mikulas Patocka (2): dm: refactor start_io_acct and end_io_acct dm: implement noclone optimization for bio-based drivers/md/dm-cache-target.c | 1 - drivers/md/dm-core.h | 1 + drivers/md/dm-linear.c | 1 + drivers/md/dm-mpath.c | 65 ++++++-- drivers/md/dm-raid.c | 14 +- drivers/md/dm-stripe.c | 1 + drivers/md/dm-table.c | 11 ++ drivers/md/dm-thin.c | 1 - drivers/md/dm-zero.c | 1 + drivers/md/dm-zoned-target.c | 1 - drivers/md/dm.c | 357 +++++++++++++++++++++++++++++------------- drivers/md/dm.h | 1 + include/linux/device-mapper.h | 22 ++- 13 files changed, 346 insertions(+), 131 deletions(-) -- 2.15.0 -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
