Improve iomap/xfs async dio write performance
iomap always punts async dio write completions to a workqueue, which has
a cost in terms of efficiency (now you need an unrelated worker to
process it) and latency (now you're bouncing a completion through an
async worker, which is a classic slowdown scenario).
io_uring handles IRQ completions via task_work, and for writes that
don't need to do extra IO at completion time, we can safely complete
them inline from that. This patchset adds IOCB_DIO_CALLER_COMP, which an
IO issuer can set to inform the completion side that any extra work that
needs doing for that completion can be punted to a safe task context.
The iomap dio completion will happen in hard/soft irq context, and we
need a saner context to process these completions. IOCB_DIO_CALLER_COMP
is added, which can be set in a struct kiocb->ki_flags by the issuer. If
the completion side of the iocb handling understands this flag, it can
choose to set a kiocb->dio_complete() handler and just call ki_complete
from IRQ context. The issuer must then ensure that this callback is
processed from a task. io_uring punts IRQ completions to task_work
already, so it's trivial wire it up to run more of the completion before
posting a CQE. This is good for up to a 37% improvement in
throughput/latency for low queue depth IO, patch 5 has the details.
If we need to do real work at completion time, iomap will clear the
IOMAP_DIO_CALLER_COMP flag.
This work came about when Andres tested low queue depth dio writes for
postgres and compared it to doing sync dio writes, showing that the
async processing slows us down a lot.
iomap: support IOCB_DIO_CALLER_COMP
If IOCB_DIO_CALLER_COMP is set, utilize that to set kiocb->dio_complete
handler and data for that callback. Rather than punt the completion to a
workqueue, we pass back the handler and data to the issuer and will get
a callback from a safe task context.
Using the following fio job to randomly dio write 4k blocks at
queue depths of 1..16:
fio --name=dio-write --filename=/data1/file --time_based=1 \
--runtime=10 --bs=4096 --rw=randwrite --norandommap --buffered=0 \
--cpus_allowed=4 --ioengine=io_uring --iodepth=$depth
shows the following results before and after this patch:
Stock Patched Diff
=======================================
QD1 155K 162K + 4.5%
QD2 290K 313K + 7.9%
QD4 533K 597K +12.0%
QD8 604K 827K +36.9%
QD16 615K 845K +37.4%
which shows nice wins all around. If we factored in per-IOP efficiency,
the wins look even nicer. This becomes apparent as queue depth rises,
as the offloaded workqueue completions runs out of steam.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
1 file changed
