Re: [RFC PATCH 0/4] Read/Write with meta buffer

[Pavel Begunkov <asml.silence@gmail.com>]

On 3/22/24 18:50, Kanchan Joshi wrote:
> This patchset is aimed at getting the feedback on a new io_uring
> interface that userspace can use to exchange meta buffer along with
> read/write.
> 
> Two new opcodes for that: IORING_OP_READ_META and IORING_OP_WRITE_META.
> The leftover space in the SQE is used to send meta buffer pointer
> and its length. Patch #2 for this.

I do remember there were back and forth design discussions about that
back when some other guy attempted to implement it, but have you tried
to do it not as a separate opcode?

It reads like all read/write opcodes might benefit from it, and it'd
be unfortunate to then be adding IORING_OP_READ_META_FIXED and
multiplicatively all other variants.

> The interface is supported for block direct IO. Patch #4 for this.
> Other two are prep patches.
> 
> It has been tried not to touch the hot read/write path, as much as
> possible. Performance for non-meta IO is same after the patches [2].
> There is some code in the cold path (worker-based async)
> though.
> 
> Moderately tested by modifying the fio [1] to use this interface
> (only for NVMe block devices)
> 
> [1] https://github.com/OpenMPDK/fio/tree/feat/test-meta
> 
> [2]
> without this:
> 
> taskset -c 2,5 t/io_uring -b512 -d128 -c32 -s32 -p1 -F1 -B1 -n2 -r4 /dev/nvme0n1 /dev/nvme1n1
> submitter=1, tid=2453, file=/dev/nvme1n1, node=-1
> submitter=0, tid=2452, file=/dev/nvme0n1, node=-1
> polled=1, fixedbufs=1, register_files=1, buffered=0, QD=128
> Engine=io_uring, sq_ring=128, cq_ring=128
> IOPS=10.02M, BW=4.89GiB/s, IOS/call=31/31
> IOPS=10.04M, BW=4.90GiB/s, IOS/call=31/31
> 
> With this:
> taskset -c 2,5 t/io_uring -b512 -d128 -c32 -s32 -p1 -F1 -B1 -n2 -r4 /dev/nvme0n1 /dev/nvme1n1
> submitter=1, tid=2453, file=/dev/nvme1n1, node=-1
> submitter=0, tid=2452, file=/dev/nvme0n1, node=-1
> polled=1, fixedbufs=1, register_files=1, buffered=0, QD=128
> Engine=io_uring, sq_ring=128, cq_ring=128
> IOPS=10.02M, BW=4.89GiB/s, IOS/call=31/31
> IOPS=10.04M, BW=4.90GiB/s, IOS/call=31/31
> 
> Anuj Gupta (3):
>    io_uring/rw: Get rid of flags field in struct io_rw
>    io_uring/rw: support read/write with metadata
>    block: modify bio_integrity_map_user to accept iov_iter as argument
> 
> Kanchan Joshi (1):
>    block: add support to pass the meta buffer
> 
>   block/bio-integrity.c         |  27 ++++++---
>   block/fops.c                  |   9 +++
>   block/t10-pi.c                |   6 ++
>   drivers/nvme/host/ioctl.c     |  11 +++-
>   include/linux/bio.h           |  13 +++-
>   include/linux/fs.h            |   1 +
>   include/uapi/linux/io_uring.h |   6 ++
>   io_uring/io_uring.c           |   2 +
>   io_uring/opdef.c              |  29 +++++++++
>   io_uring/rw.c                 | 108 +++++++++++++++++++++++++++++-----
>   io_uring/rw.h                 |   8 +++
>   11 files changed, 193 insertions(+), 27 deletions(-)
> 
> 
> base-commit: 6f0974eccbf78baead1735722c4f1ee3eb9422cd

-- 
Pavel Begunkov