From: Joanne Koong <joannelkoong@gmail.com>
To: axboe@kernel.dk, miklos@szeredi.hu
Cc: bschubert@ddn.com, csander@purestorage.com, krisman@suse.de,
io-uring@vger.kernel.org, asml.silence@gmail.com,
xiaobing.li@samsung.com, safinaskar@gmail.com,
linux-fsdevel@vger.kernel.org
Subject: Re: [PATCH v4 00/25] fuse/io-uring: add kernel-managed buffer rings and zero-copy
Date: Tue, 27 Jan 2026 12:12:11 -0800 [thread overview]
Message-ID: <CAJnrk1Z-9rsP86Fc=57P9gy=vFjfjT8nuAgE2_snL3_vfbbBmg@mail.gmail.com> (raw)
In-Reply-To: <20260116233044.1532965-1-joannelkoong@gmail.com>
On Fri, Jan 16, 2026 at 3:31 PM Joanne Koong <joannelkoong@gmail.com> wrote:
>
> This series adds buffer ring and zero-copy capabilities to fuse over io-uring.
> This requires adding a new kernel-managed buf (kmbuf) ring type to io-uring
> where the buffers are provided and managed by the kernel instead of by
> userspace.
>
> On the io-uring side, the kmbuf interface is basically identical to pbufs.
> They differ mostly in how the memory region is set up and whether it is
> userspace or kernel that recycles back the buffer. Internally, the
> IOBL_KERNEL_MANAGED flag is used to mark the buffer ring as kernel-managed.
>
> The zero-copy work builds on top of the infrastructure added for
> kernel-managed buffer rings (the bulk of which is in patch 19: "fuse: add
> io-uring kernel-managed buffer ring") and that informs some of the design
> choices for how fuse uses the kernel-managed buffer ring without zero-copy.
Could anyone on the fuse side review the fuse changes in patches 19 and 24?
Thanks,
Joanne
>
> There was a previous submission for supporting registered buffers in fuse [1]
> but that was abandoned in favor of using kernel-managed buffer rings, which,
> once incremental buffer consumption is added in a later patchset, gives
> significant memory usage advantages in allowing the full buffer capacity to be
> utilized across multiple requests, as well as offers more flexibility for
> future additions. As well, it also makes the userspace side setup simpler.
> The relevant refactoring fuse patches from the previous submission are carried
> over into this one.
>
> Benchmarks for zero-copy (patch 24) show approximately the following
> differences in throughput for bs=1M:
>
> direct randreads: ~20% increase (~2100 MB/s -> ~2600 MB/s)
> buffered randreads: ~25% increase (~1900 MB/s -> 2400 MB/s)
> direct randwrites: no difference (~750 MB/s)
> buffered randwrites: ~10% increase (950 MB/s -> 1050 MB/s)
>
> The benchmark was run using fio on the passthrough_hp server:
> fio --name=test_run --ioengine=sync --rw=rand{read,write} --bs=1M
> --size=1G --numjobs=2 --ramp_time=30 --group_reporting=1
>
> This series is on top of commit b71e635feefc in the io-uring tree.
>
> The libfuse changes can be found in [2]. This has a dependency on the liburing
> changes in [3]. To test the server, you can run it with:
> sudo ~/libfuse/build/example/passthrough_hp ~/src ~/mounts/tmp
> --nopassthrough -o io_uring_zero_copy -o io_uring_q_depth=8
>
> Thanks,
> Joanne
>
> [1] https://lore.kernel.org/linux-fsdevel/20251027222808.2332692-1-joannelkoong@gmail.com/
> [2] https://github.com/joannekoong/libfuse/tree/zero_copy
> [3] https://github.com/joannekoong/liburing/tree/kmbuf
>
> v3: https://lore.kernel.org/linux-fsdevel/20251223003522.3055912-1-joannelkoong@gmail.com/
> v3 -> v4:
> * Get rid of likely()s and get rid of going through cmd interface layer (Gabriel)
> * Fix io_uring_cmd_fixed_index_get() to return back the node pointer (Caleb)
> * Add documentation for io_buffer_register_bvec (Caleb)
> * Remove WARN_ON_ONCE() for io_buffer_unregister() call (Caleb)
>
> v2: https://lore.kernel.org/linux-fsdevel/20251218083319.3485503-1-joannelkoong@gmail.com/
> v2 -> v3:
> * fix casting between void * and u64 for 32-bit architectures (kernel test robot)
> * add newline for documentation bullet points (kernel test robot)
> * fix unrecognized "boolean" (kernel test robot), switch it to a flag (me)
>
> v1: https://lore.kernel.org/linux-fsdevel/20251203003526.2889477-1-joannelkoong@gmail.com/
> v1 -> v2:
> * drop fuse buffer cleanup on ring death, which makes things a lot simpler (Jens)
> - this drops a lot of things (eg needing ring_ctx tracking, needing callback
> for ring death, etc)
> * drop fixed buffer pinning altogether and just do lookup every time (Jens)
> (didn't significantly affect the benchmark results seen)
> * fix spelling mistake in docs (Askar)
> * use -EALREADY for pinning already pinned bufring, return PTR_ERR for
> registration instead of err, move initializations to outside locks (Caleb)
> * drop fuse patches for zero-ed out headers (me)
>
> Joanne Koong (25):
> io_uring/kbuf: refactor io_buf_pbuf_register() logic into generic
> helpers
> io_uring/kbuf: rename io_unregister_pbuf_ring() to
> io_unregister_buf_ring()
> io_uring/kbuf: add support for kernel-managed buffer rings
> io_uring/kbuf: add mmap support for kernel-managed buffer rings
> io_uring/kbuf: support kernel-managed buffer rings in buffer selection
> io_uring/kbuf: add buffer ring pinning/unpinning
> io_uring/kbuf: add recycling for kernel managed buffer rings
> io_uring: add io_uring_fixed_index_get() and
> io_uring_fixed_index_put()
> io_uring/kbuf: add io_uring_is_kmbuf_ring()
> io_uring/kbuf: export io_ring_buffer_select()
> io_uring/kbuf: return buffer id in buffer selection
> io_uring/cmd: set selected buffer index in __io_uring_cmd_done()
> fuse: refactor io-uring logic for getting next fuse request
> fuse: refactor io-uring header copying to ring
> fuse: refactor io-uring header copying from ring
> fuse: use enum types for header copying
> fuse: refactor setting up copy state for payload copying
> fuse: support buffer copying for kernel addresses
> fuse: add io-uring kernel-managed buffer ring
> io_uring/rsrc: rename
> io_buffer_register_bvec()/io_buffer_unregister_bvec()
> io_uring/rsrc: split io_buffer_register_request() logic
> io_uring/rsrc: Allow buffer release callback to be optional
> io_uring/rsrc: add io_buffer_register_bvec()
> fuse: add zero-copy over io-uring
> docs: fuse: add io-uring bufring and zero-copy documentation
>
> Documentation/block/ublk.rst | 14 +-
> .../filesystems/fuse/fuse-io-uring.rst | 59 +-
> drivers/block/ublk_drv.c | 18 +-
> fs/fuse/dev.c | 30 +-
> fs/fuse/dev_uring.c | 692 ++++++++++++++----
> fs/fuse/dev_uring_i.h | 42 +-
> fs/fuse/fuse_dev_i.h | 8 +-
> include/linux/io_uring/buf.h | 25 +
> include/linux/io_uring/cmd.h | 97 ++-
> include/linux/io_uring_types.h | 10 +-
> include/uapi/linux/fuse.h | 17 +-
> include/uapi/linux/io_uring.h | 17 +-
> io_uring/kbuf.c | 355 +++++++--
> io_uring/kbuf.h | 19 +-
> io_uring/memmap.c | 117 ++-
> io_uring/memmap.h | 4 +
> io_uring/register.c | 9 +-
> io_uring/rsrc.c | 183 ++++-
> io_uring/uring_cmd.c | 6 +-
> 19 files changed, 1447 insertions(+), 275 deletions(-)
> create mode 100644 include/linux/io_uring/buf.h
>
> --
> 2.47.3
>
next prev parent reply other threads:[~2026-01-27 20:12 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-01-16 23:30 [PATCH v4 00/25] fuse/io-uring: add kernel-managed buffer rings and zero-copy Joanne Koong
2026-01-16 23:30 ` [PATCH v4 01/25] io_uring/kbuf: refactor io_buf_pbuf_register() logic into generic helpers Joanne Koong
2026-01-16 23:30 ` [PATCH v4 02/25] io_uring/kbuf: rename io_unregister_pbuf_ring() to io_unregister_buf_ring() Joanne Koong
2026-01-16 23:30 ` [PATCH v4 03/25] io_uring/kbuf: add support for kernel-managed buffer rings Joanne Koong
2026-02-06 7:53 ` Christoph Hellwig
2026-02-07 0:59 ` Joanne Koong
2026-02-06 13:39 ` Askar Safin
2026-02-07 1:22 ` Joanne Koong
2026-02-07 16:13 ` Askar Safin
2026-02-09 17:31 ` Joanne Koong
2026-01-16 23:30 ` [PATCH v4 04/25] io_uring/kbuf: add mmap " Joanne Koong
2026-01-16 23:30 ` [PATCH v4 05/25] io_uring/kbuf: support kernel-managed buffer rings in buffer selection Joanne Koong
2026-01-16 23:30 ` [PATCH v4 06/25] io_uring/kbuf: add buffer ring pinning/unpinning Joanne Koong
2026-02-03 18:52 ` Bernd Schubert
2026-02-05 19:37 ` Joanne Koong
2026-01-16 23:30 ` [PATCH v4 07/25] io_uring/kbuf: add recycling for kernel managed buffer rings Joanne Koong
2026-02-03 18:44 ` Bernd Schubert
2026-02-05 19:47 ` Joanne Koong
2026-02-05 20:42 ` Bernd Schubert
2026-01-16 23:30 ` [PATCH v4 08/25] io_uring: add io_uring_fixed_index_get() and io_uring_fixed_index_put() Joanne Koong
2026-01-22 21:02 ` Caleb Sander Mateos
2026-01-27 20:05 ` Joanne Koong
2026-01-16 23:30 ` [PATCH v4 09/25] io_uring/kbuf: add io_uring_is_kmbuf_ring() Joanne Koong
2026-01-16 23:30 ` [PATCH v4 10/25] io_uring/kbuf: export io_ring_buffer_select() Joanne Koong
2026-01-16 23:30 ` [PATCH v4 11/25] io_uring/kbuf: return buffer id in buffer selection Joanne Koong
2026-01-16 23:30 ` [PATCH v4 12/25] io_uring/cmd: set selected buffer index in __io_uring_cmd_done() Joanne Koong
2026-01-16 23:30 ` [PATCH v4 13/25] fuse: refactor io-uring logic for getting next fuse request Joanne Koong
2026-01-16 23:30 ` [PATCH v4 14/25] fuse: refactor io-uring header copying to ring Joanne Koong
2026-01-16 23:30 ` [PATCH v4 15/25] fuse: refactor io-uring header copying from ring Joanne Koong
2026-01-27 23:11 ` Bernd Schubert
2026-01-16 23:30 ` [PATCH v4 16/25] fuse: use enum types for header copying Joanne Koong
2026-01-16 23:30 ` [PATCH v4 17/25] fuse: refactor setting up copy state for payload copying Joanne Koong
2026-01-16 23:30 ` [PATCH v4 18/25] fuse: support buffer copying for kernel addresses Joanne Koong
2026-01-27 23:39 ` Bernd Schubert
2026-01-28 0:23 ` Joanne Koong
2026-01-28 21:14 ` Bernd Schubert
2026-01-29 1:16 ` Joanne Koong
2026-01-16 23:30 ` [PATCH v4 19/25] fuse: add io-uring kernel-managed buffer ring Joanne Koong
2026-01-17 5:28 ` kernel test robot
2026-01-28 21:44 ` Bernd Schubert
2026-01-29 1:15 ` Joanne Koong
2026-02-25 17:55 ` Bernd Schubert
2026-02-25 23:42 ` Joanne Koong
2026-02-26 18:21 ` Bernd Schubert
2026-02-26 20:58 ` Joanne Koong
2026-01-16 23:30 ` [PATCH v4 20/25] io_uring/rsrc: rename io_buffer_register_bvec()/io_buffer_unregister_bvec() Joanne Koong
2026-01-16 23:30 ` [PATCH v4 21/25] io_uring/rsrc: split io_buffer_register_request() logic Joanne Koong
2026-01-16 23:30 ` [PATCH v4 22/25] io_uring/rsrc: Allow buffer release callback to be optional Joanne Koong
2026-01-16 23:30 ` [PATCH v4 23/25] io_uring/rsrc: add io_buffer_register_bvec() Joanne Koong
2026-01-16 23:30 ` [PATCH v4 24/25] fuse: add zero-copy over io-uring Joanne Koong
2026-01-16 23:30 ` [PATCH v4 25/25] docs: fuse: add io-uring bufring and zero-copy documentation Joanne Koong
2026-02-03 18:56 ` Bernd Schubert
2026-02-05 20:30 ` Joanne Koong
2026-01-27 20:12 ` Joanne Koong [this message]
2026-01-27 22:44 ` [PATCH v4 00/25] fuse/io-uring: add kernel-managed buffer rings and zero-copy Bernd Schubert
2026-01-27 23:27 ` Joanne Koong
2026-02-06 7:42 ` Christoph Hellwig
2026-02-07 0:24 ` Joanne Koong
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAJnrk1Z-9rsP86Fc=57P9gy=vFjfjT8nuAgE2_snL3_vfbbBmg@mail.gmail.com' \
--to=joannelkoong@gmail.com \
--cc=asml.silence@gmail.com \
--cc=axboe@kernel.dk \
--cc=bschubert@ddn.com \
--cc=csander@purestorage.com \
--cc=io-uring@vger.kernel.org \
--cc=krisman@suse.de \
--cc=linux-fsdevel@vger.kernel.org \
--cc=miklos@szeredi.hu \
--cc=safinaskar@gmail.com \
--cc=xiaobing.li@samsung.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox