Linux RDMA and InfiniBand development
 help / color / mirror / Atom feed
From: Jason Gunthorpe <jgg@nvidia.com>
To: Leon Romanovsky <leon@kernel.org>, linux-rdma@vger.kernel.org
Cc: Jiri Pirko <jiri@resnulli.us>, patches@lists.linux.dev
Subject: Re: [PATCH v3 0/6] Remove driver dependencies on ib_uverbs.ko
Date: Tue, 26 May 2026 10:21:56 -0300	[thread overview]
Message-ID: <20260526132156.GA3221769@nvidia.com> (raw)
In-Reply-To: <0-v3-43aba1969751+1988-ib_uverbs_support_ko_jgg@nvidia.com>

On Mon, May 25, 2026 at 10:22:36PM -0300, Jason Gunthorpe wrote:
> The original design was for user facing modules like ib_uverbs to be
> independently loadable, if the user didn't want to have those char devs
> then they could block the module.
> 
> This has slowly gotten degraded over time and right now every driver is
> depending on ib_uverbs.ko. Fixup everything except
> rdma_user_mmap_disassociate() in hns by moving code around and adding a
> new module ib_uverbs_support.ko to hold the driver functions without any
> of the uverbs cdev code.
> 
> After this series mlx5_ib and bnxt_re will use ib_uverbs_support.ko.
> 
> v3:
>   - Handle core=y uverbs=m cases using a new kconfig
>   - Remove doubled rdma_user_mmap_disassociate() inline
>   - Add missing _exit
> v2: https://patch.msgid.link/r/0-v2-4a21959414f2+3d7-ib_uverbs_support_ko_jgg@nvidia.com
>   - Rebase on the rc branch
>   - Fix ucaps module mistakes
>   - Add a patch to not build ib_core_uverbs.o without
>     CONFIG_INFINIBAND_USER_ACCESS
> v1: https://patch.msgid.link/r/0-v1-045258567bd6+9fe-ib_uverbs_support_ko_jgg@nvidia.com
> 
> Jason Gunthorpe (6):
>   RDMA/core: Do not compile ib_core_uverbs without USER_ACCESS
>   RDMA/core: Move many of the little EXPORTs from uverbs_ioctl into
>     ib_core_uverbs
>   RDMA/core: Remove uverbs_async_event_release()
>   RDMA/core: Make a new module for the uverbs components needed by
>     drivers
>   RDMA/core: Move ucaps into ib_uverbs_support.ko
>   RDMA/core: Move flow related functions to ib_uverbs_support.ko

Applied to for-next, I fixed the two new minor Sashiko remarks about a stale
comment and a READ_ONCE in uverbs_uobject_fd_release().

Jason

      parent reply	other threads:[~2026-05-26 13:22 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-26  1:22 [PATCH v3 0/6] Remove driver dependencies on ib_uverbs.ko Jason Gunthorpe
2026-05-26  1:22 ` [PATCH v3 1/6] RDMA/core: Do not compile ib_core_uverbs without USER_ACCESS Jason Gunthorpe
2026-05-26  1:22 ` [PATCH v3 2/6] RDMA/core: Move many of the little EXPORTs from uverbs_ioctl into ib_core_uverbs Jason Gunthorpe
2026-05-26  1:22 ` [PATCH v3 3/6] RDMA/core: Remove uverbs_async_event_release() Jason Gunthorpe
2026-05-26  1:22 ` [PATCH v3 4/6] RDMA/core: Make a new module for the uverbs components needed by drivers Jason Gunthorpe
2026-05-26  1:22 ` [PATCH v3 5/6] RDMA/core: Move ucaps into ib_uverbs_support.ko Jason Gunthorpe
2026-05-26  1:22 ` [PATCH v3 6/6] RDMA/core: Move flow related functions to ib_uverbs_support.ko Jason Gunthorpe
2026-05-26 13:21 ` Jason Gunthorpe [this message]

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=20260526132156.GA3221769@nvidia.com \
    --to=jgg@nvidia.com \
    --cc=jiri@resnulli.us \
    --cc=leon@kernel.org \
    --cc=linux-rdma@vger.kernel.org \
    --cc=patches@lists.linux.dev \
    /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