[RFC 00/13] uring-passthru for nvme

[Kanchan Joshi]

Here is a revamped series on uring-passthru which is on top of Jens
"nvme-passthru-wip.2" branch.
https://git.kernel.dk/cgit/linux-block/commit/?h=nvme-passthru-wip.2

This scales much better than before with the addition of following:
- plugging
- passthru polling (sync and async; sync part comes from a patch that
  Keith did earlier)
- bio-cache (this is regardless of irq/polling since we submit/complete in
  task-contex anyway. Currently kicks in when fixed-buffer option is
also passed, but that's primarily to keep the plumbing simple)

Also the feedback from Christoph (previous fixed-buffer series) is in
which has streamlined the plumbing.

I look forward to further feedback/comments.

KIOPS(512b) on P5800x looked like this:

QD    uring    pt    uring-poll    pt-poll
8      538     589      831         902
64     967     1131     1351        1378
256    1043    1230     1376        1429

Here uring is operating on block-interface (nvme0n1) while 'pt' refers
to uring-passthru operating on char-interface (ng0n1).

Perf/testing is with this custom fio that turnes regular io into
passthru on supplying "uring_cmd=1" option.
https://github.com/joshkan/fio/tree/nvme-passthru-wip-polling
Example command-line:
fio -iodepth=256 -rw=randread -ioengine=io_uring -bs=512 -numjobs=1 -runtime=60 -group_reporting -iodepth_batch_submit=64 -iodepth_batch_complete_min=1 -iodepth_batch_complete_max=64 -fixedbufs=1 -hipri=1 -sqthread_poll=0 -filename=/dev/ng0n1 -name=io_uring_256 -uring_cmd=1

background/context:
https://linuxplumbersconf.org/event/11/contributions/989/attachments/747/1723/lpc-2021-building-a-fast-passthru.pdf

Changes from v5:
https://lore.kernel.org/linux-nvme/20210805125539.66958-1-joshi.k@samsung.com/
1. Fixed-buffer passthru with same ioctl code + other feedback from hch
2. Plugging (from Jens)
3. Sync polling (from Keith)
3. Async polling via io_uring
4. Enable bio-cache for fixed-buffer passthru

Changes from v4:
https://lore.kernel.org/linux-nvme/20210325170540.59619-1-joshi.k@samsung.com/
1. Moved to v5 branch of Jens, adapted to task-work changes in io_uring
2. Removed support for block-passthrough (over nvme0n1) for now
3. Added support for char-passthrough (over ng0n1)
4. Added fixed-buffer passthrough in io_uring and nvme plumbing


Anuj Gupta (3):
  io_uring: mark iopoll not supported for uring-cmd
  io_uring: modify unused field in io_uring_cmd to store flags
  io_uring: add support for uring_cmd with fixed-buffer

Jens Axboe (2):
  io_uring: plug for async bypass
  block: wire-up support for plugging

Kanchan Joshi (6):
  io_uring: add infra for uring_cmd completion in submitter-task
  nvme: wire-up support for async-passthru on char-device.
  io_uring: add flag and helper for fixed-buffer uring-cmd
  nvme: enable passthrough with fixed-buffer
  block: factor out helper for bio allocation from cache
  nvme: enable bio-cache for fixed-buffer passthru

Keith Busch (1):
  nvme: allow user passthrough commands to poll

Pankaj Raghav (1):
  nvme: Add async passthru polling support

 block/bio.c                     |  43 +++--
 block/blk-map.c                 |  46 ++++++
 block/blk-mq.c                  |  93 +++++------
 drivers/nvme/host/core.c        |  21 ++-
 drivers/nvme/host/ioctl.c       | 271 ++++++++++++++++++++++++++++----
 drivers/nvme/host/multipath.c   |   2 +
 drivers/nvme/host/nvme.h        |  13 +-
 drivers/nvme/host/pci.c         |   4 +-
 drivers/nvme/target/passthru.c  |   2 +-
 fs/io_uring.c                   | 113 +++++++++++--
 include/linux/bio.h             |   1 +
 include/linux/blk-mq.h          |   4 +
 include/linux/io_uring.h        |  26 ++-
 include/uapi/linux/io_uring.h   |   6 +-
 include/uapi/linux/nvme_ioctl.h |   4 +
 15 files changed, 542 insertions(+), 107 deletions(-)

-- 
2.25.1