From: Kanchan Joshi <joshi.k@samsung.com> To: axboe@kernel.dk, hch@lst.de, kbusch@kernel.org Cc: io-uring@vger.kernel.org, linux-nvme@lists.infradead.org, anuj20.g@samsung.com, javier.gonz@samsung.com, hare@suse.de, Kanchan Joshi <joshi.k@samsung.com> Subject: [RFC PATCH 0/6] Fixed-buffers io_uring passthrough over nvme-char Date: Thu, 5 Aug 2021 18:25:33 +0530 [thread overview] Message-ID: <20210805125539.66958-1-joshi.k@samsung.com> (raw) In-Reply-To: <CGME20210805125910epcas5p1100e7093dd2b1ac5bbb751331e2ded23@epcas5p1.samsung.com> Hi, This series takes a dig at expanding io_uring passthrough to have fixed-buffer support. For the NVMe side plumbing, it uses the per-namespace char-device as the backend. The work is on top of Jens' branch (which is 5.14-rc1 based): https://git.kernel.dk/cgit/linux-block/log/?h=io_uring-fops.v5 Testing is done by integrating user-space side with fio: https://github.com/joshkan/fio/commit/a8f40d4f7013ccd38fc51275d8b00fb4ce2404f5 Patches can be grouped into two parts - First part (patch 1 and 2) involves extending 'io_uring cmd' infra (for ioctl updates in task-context), and wire up nvme-passthrough on char-device (/dev/ngXnY). The second part (patch 4, 5, 6) introduces fixed-buffer variant of uring-cmd (IORING_OP_URING_CMD_FIXED), and new NVMe passthrough IOCTLs that operate on pre-registered buffers. The buffers are registered with io_uring in the usual fashion. And the buffer index is passed in command-SQE. This to be used along with two new fixed-buffer ioctl opcodes: NVME_IOCTL_IO_CMD_FIXED or NVME_IOCTL_IO64_CMD_FIXED. For existing passthrough ioctl, nvme-driver maps the buffers in I/O path for each command. For the new ioctl, driver supplies buffer-index to io_uring and gets to reuse pre-mapped buffers. If introducing new ioctls in nvme (for fixed-buffer) does not sound great, another option would be to use 'user flags' in the passthrough command. I look forward to the thoughts of the community on this. Jens, Command-SQE didn't have the field for buf_index. I'm torn among few options, but repurposed len for now. Please take a look. I'd appreciate the feedback/comments. Thanks, 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 (2): io_uring: mark iopoll not supported for uring-cmd io_uring: add support for uring_cmd with fixed-buffer Kanchan Joshi (4): io_uring: add infra for uring_cmd completion in submitter-task. nvme: wire-up support for async-passthru on char-device. io_uring: add helper for fixed-buffer uring-cmd nvme: enable passthrough with fixed-buffer drivers/nvme/host/core.c | 1 + drivers/nvme/host/ioctl.c | 245 +++++++++++++++++++++++++++++--- drivers/nvme/host/multipath.c | 1 + drivers/nvme/host/nvme.h | 5 + fs/io_uring.c | 69 ++++++++- include/linux/io_uring.h | 15 ++ include/uapi/linux/io_uring.h | 6 +- include/uapi/linux/nvme_ioctl.h | 2 + 8 files changed, 319 insertions(+), 25 deletions(-) -- 2.25.1
next parent reply other threads:[~2021-08-05 13:15 UTC|newest] Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top [not found] <CGME20210805125910epcas5p1100e7093dd2b1ac5bbb751331e2ded23@epcas5p1.samsung.com> 2021-08-05 12:55 ` Kanchan Joshi [this message] [not found] ` <CGME20210805125917epcas5p4f75c9423a7b886dc79500901cc8f55ab@epcas5p4.samsung.com> 2021-08-05 12:55 ` [RFC PATCH 1/6] io_uring: add infra for uring_cmd completion in submitter-task Kanchan Joshi [not found] ` <CGME20210805125923epcas5p10e6c1b95475440be68f58244d5a3cb9a@epcas5p1.samsung.com> 2021-08-05 12:55 ` [RFC PATCH 2/6] nvme: wire-up support for async-passthru on char-device Kanchan Joshi 2021-09-07 7:46 ` Christoph Hellwig 2021-09-07 16:20 ` Kanchan Joshi 2021-09-08 6:15 ` Christoph Hellwig 2021-09-22 7:19 ` Kanchan Joshi [not found] ` <CGME20210805125927epcas5p28f3413fe3d0a2baed37a05453df0d482@epcas5p2.samsung.com> 2021-08-05 12:55 ` [RFC PATCH 3/6] io_uring: mark iopoll not supported for uring-cmd Kanchan Joshi [not found] ` <CGME20210805125931epcas5p259fec172085ea34fdbf5a1c1f8da5e90@epcas5p2.samsung.com> 2021-08-05 12:55 ` [RFC PATCH 4/6] io_uring: add helper for fixed-buffer uring-cmd Kanchan Joshi 2021-09-07 7:47 ` Christoph Hellwig [not found] ` <CGME20210805125934epcas5p4ff88e95d558ad9f65d77a888a4211b18@epcas5p4.samsung.com> 2021-08-05 12:55 ` [RFC PATCH 5/6] io_uring: add support for uring_cmd with fixed-buffer Kanchan Joshi 2021-09-07 7:48 ` Christoph Hellwig 2021-09-07 16:29 ` Kanchan Joshi [not found] ` <CGME20210805125937epcas5p15667b460e28d87bd40400f69005aafe3@epcas5p1.samsung.com> 2021-08-05 12:55 ` [RFC PATCH 6/6] nvme: enable passthrough " Kanchan Joshi 2021-09-07 7:50 ` Christoph Hellwig 2021-09-07 16:47 ` Kanchan Joshi 2021-09-08 6:16 ` Christoph Hellwig 2021-09-07 7:10 ` [RFC PATCH 0/6] Fixed-buffers io_uring passthrough over nvme-char Christoph Hellwig 2021-09-07 12:38 ` Jens Axboe
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=20210805125539.66958-1-joshi.k@samsung.com \ --to=joshi.k@samsung.com \ --cc=anuj20.g@samsung.com \ --cc=axboe@kernel.dk \ --cc=hare@suse.de \ --cc=hch@lst.de \ --cc=io-uring@vger.kernel.org \ --cc=javier.gonz@samsung.com \ --cc=kbusch@kernel.org \ --cc=linux-nvme@lists.infradead.org \ --subject='Re: [RFC PATCH 0/6] Fixed-buffers io_uring passthrough over nvme-char' \ /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
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).