Request-based DM's blk-mq support (dm-mq) was reported to be 50% slower
than if an underlying null_blk device were used directly.  One of the
reasons for this drop in performance is that blk_insert_clone_request()
was calling blk_mq_insert_request() with @async=true.  This forced the
use of kblockd_schedule_delayed_work_on() to run the blk-mq hw queues
which ushered in ping-ponging between process context (fio in this case)
and kblockd's kworker to submit the cloned request.  The ftrace
function_graph tracer showed:

  kworker-2013  =>   fio-12190
  fio-12190    =>  kworker-2013
  ...
  kworker-2013  =>   fio-12190
  fio-12190    =>  kworker-2013
  ...

Fixing blk_insert_clone_request()'s blk_mq_insert_request() call to
_not_ use kblockd to submit the cloned requests isn't enough to
eliminate the observed context switches.

In addition to this dm-mq specific blk-core fix, there are 2 DM core
fixes to dm-mq that (when paired with the blk-core fix) completely
eliminate the observed context switching:

1)  don't blk_mq_run_hw_queues in blk-mq request completion

    Motivated by desire to reduce overhead of dm-mq, punting to kblockd
    just increases context switches.

    In my testing against a really fast null_blk device there was no benefit
    to running blk_mq_run_hw_queues() on completion (and no other blk-mq
    driver does this).  So hopefully this change doesn't induce the need for
    yet another revert like commit 621739b0 !

2)  use blk_mq_complete_request() in dm_complete_request()

    blk_complete_request() doesn't offer the traditional q->mq_ops vs
    .request_fn branching pattern that other historic block interfaces
    do (e.g. blk_get_request).  Using blk_mq_complete_request() for
    blk-mq requests is important for performance.  It should be noted
    that, like blk_complete_request(), blk_mq_complete_request() doesn't
    natively handle partial completions -- but the request-based
    DM-multipath target does provide the required partial completion
    support by dm.c:end_clone_bio() triggering requeueing of the request
    via dm-mpath.c:multipath_end_io()'s return of DM_ENDIO_REQUEUE.

dm-mq fix #2 is _much_ more important than #1 for eliminating the
context switches.
Before: cpu          : usr=15.10%, sys=59.39%, ctx=7905181, majf=0, minf=475
After:  cpu          : usr=20.60%, sys=79.35%, ctx=2008, majf=0, minf=472

With these changes multithreaded async read IOPs improved from ~950K
to ~1350K for this dm-mq stacked on null_blk test-case.  The raw read
IOPs of the underlying null_blk device for the same workload is ~1950K.

Fixes: 7fb4898e ("block: add blk-mq support to blk_insert_cloned_request()")
Fixes: bfebd1cd

 ("dm: add full blk-mq support to request-based DM")
Cc: stable@vger.kernel.org # 4.1+
Reported-by: Sagi Grimberg <sagig@dev.mellanox.co.il>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Jens Axboe <axboe@kernel.dk>