On Tue, Jan 11 2022 at 3:34P -0500, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > On Tue, Dec 28, 2021 at 04:30:08PM -0500, Mike Snitzer wrote: > > Yeah, people use request-based for IO scheduling and more capable path > > selectors. Imposing bio-based would be a pretty jarring workaround for > > BLK_MQ_F_BLOCKING. request-based DM should properly support it. > > Given that nvme-tcp is the only blocking driver that has multipath > driver that driver explicitly does not intend to support dm-multipath > I'm absolutely against adding block layer cruft for this particular > use case. this diffstat amounts to what you call "cruft": block/blk-core.c | 2 +- block/blk-mq.c | 6 +++--- block/blk-mq.h | 2 +- block/blk-sysfs.c | 2 +- block/genhd.c | 5 +++-- drivers/md/dm-rq.c | 5 ++++- drivers/md/dm-rq.h | 3 ++- drivers/md/dm-table.c | 14 ++++++++++++++ drivers/md/dm.c | 5 +++-- drivers/md/dm.h | 1 + include/linux/blkdev.h | 5 +++-- include/linux/genhd.h | 12 ++++++++---- 12 files changed, 44 insertions(+), 18 deletions(-) > SCSI even has this: > > /* > * SCSI never enables blk-mq's BLK_MQ_F_BLOCKING flag so > * calling synchronize_rcu() once is enough. > */ > WARN_ON_ONCE(shost->tag_set.flags & BLK_MQ_F_BLOCKING); > Round and round we go.. Pretty tired of this. You are perfectly fine with incrementally compromising request-based DM's ability to evolve as block core does. Seriously, this patchset shouldn't warrant bickering: https://patchwork.kernel.org/project/dm-devel/list/?series=598823 Jens, this incremental weakening of what it is that DM is allowed to do is not something I can continue to work with (nor should Ming's or others' contributions be rejected for such reasons). This tribal war needs to stop. Mike
