From: Ming Lei <ming.lei@redhat.com>
To: Gabriel Krisman Bertazi <krisman@suse.de>
Cc: linux-block@vger.kernel.org, io-uring <io-uring@vger.kernel.org>,
Andreas Hindborg <nmi@metaspace.dk>,
Shinichiro Kawasaki <shinichiro.kawasaki@wdc.com>,
German Maglione <gmaglione@redhat.com>,
Stefano Garzarella <sgarzare@redhat.com>,
Joe Thornber <ethornbe@redhat.com>,
ming.lei@redhat.com
Subject: Re: Libublk-rs v0.1.0
Date: Tue, 15 Aug 2023 09:17:41 +0800 [thread overview]
Message-ID: <ZNrSNeCSg5qsOu61@fedora> (raw)
In-Reply-To: <87o7j95vql.fsf@suse.de>
Hi Gabriel,
On Mon, Aug 14, 2023 at 02:23:14PM -0400, Gabriel Krisman Bertazi wrote:
> Ming Lei <ming.lei@redhat.com> writes:
>
> > Hello,
> >
> > Libublk-rs(Rust)[1][2] 0.1.0 is released.
>
> Hi Ming,
>
> Do you intend to effectively deprecate the code in ubdsrv in favor of
> libublk-rs or do you intend to keep the C library? I'm asking because
> I'm looking into how to enable ublk in distributions.
So far, there are users of C ubdsrv library, and both are maintained
and co-exist, so C library is still kept.
Thanks
Ming
>
> Thanks,
>
> >
> > The original idea is to use Rust to write ublk target for covering all
> > kinds of block queue limits/parameters combination easily when talking
> > with Andreas and Shinichiro about blktests in LSFMM/BPF 2023.
> >
> > Finally it is evolved into one generic library. Attributed to Rust's
> > some modern language features, libublk interfaces are pretty simple:
> >
> > - one closure(tgt_init) for user to customize device by providing all
> > kind of parameter
> >
> > - the other closure(io handling) for user to handling IO which is
> > completely io_uring CQE driven: a) IO command CQE from ublk driver,
> > b) target IO CQE originated from target io handling code, c) eventfd
> > CQE if IO is offloaded to other context
> >
> > With low level APIs, <50 LoC can build one ublk-null, and if high level
> > APIs are used, 30 LoC is enough.
> >
> > Performance is basically aligned with pure C ublk implementation[3].
> >
> > The library has been verified on null, ramdisk, loop and zoned target.
> > The plan is to support async/await in 0.2 or 0.3 so that libublk can
> > be used to build complicated target easily and efficiently.
> >
> > Thanks Andreas for reviewing and providing lots of good ideas for
> > improvement & cleanup. Thanks German Maglione for some suggestions, such
> > as eventfd support. Thanks Joe for providing excellent Rust programming
> > guide.
> >
> > Any feedback is welcome!
> >
> > [1] https://crates.io/crates/libublk
> > [2] https://github.com/ming1/libublk-rs
> > [3] https://github.com/osandov/blktests/blob/master/src/miniublk.c
> >
> > Thanks,
> > Ming
> >
>
> --
> Gabriel Krisman Bertazi
>
--
Ming
prev parent reply other threads:[~2023-08-15 1:19 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-08-14 2:56 Libublk-rs v0.1.0 Ming Lei
2023-08-14 18:23 ` Gabriel Krisman Bertazi
2023-08-15 1:17 ` Ming Lei [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=ZNrSNeCSg5qsOu61@fedora \
--to=ming.lei@redhat.com \
--cc=ethornbe@redhat.com \
--cc=gmaglione@redhat.com \
--cc=io-uring@vger.kernel.org \
--cc=krisman@suse.de \
--cc=linux-block@vger.kernel.org \
--cc=nmi@metaspace.dk \
--cc=sgarzare@redhat.com \
--cc=shinichiro.kawasaki@wdc.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.