[LSF/MM/BPF TOPIC] Rust block device driver APIs

[LSF/MM/BPF TOPIC] Rust block device driver APIs

[Andreas Hindborg (Samsung)]

Hi All,

I would like to propose a session on the Rust block device driver APIs.

I submitted the APIs along with a simple null block driver as an RFC last year
[1]. Since then I have kept the code in sync with latest mainline release [2],
cleaned up the code, and added a few features.

After talking to some of you at various meetups over the past year, I think we
have reached a point where we can potentially agree on merging initial Rust
block layer support, along with the null block driver. To that end, I plan to
send a few iterations of the patch set before LSF in May, so that we can use the
session to discuss any remaining details.

Since Kent has also proposed a dedicated Rust session, we might find some
synergy with this topic [3].

I also maintain an NVMe driver based on the Rust block APIs [4]. Due to
community feedback, I have no plans for upstreaming this driver at the moment.
However, it is a valuable tool for designing a sensible Rust block device API
that is suitable real hardware.

Part of the NVMe patches are abstractions for PCI. Other users (drm) have
expressed interest in these, so I plan to separate these in their own tree to
make them easier to pick up for those users.

As a last note, I have recently become aware of ongoing work on
implementing nbd in Rust. The work looks promising, and I hope the
author will decide to send the patches, when they are ready to be
shared.

Best regards
Andreas Hindborg


[1] https://lore.kernel.org/all/20230503090708.2524310-1-nmi@metaspace.dk/
[2] https://rust-for-linux.com/null-block-driver
[3] https://lore.kernel.org/all/wjtuw2m3ojn7m6gx2ozyqtvlsetzwxv5hdhg2hocal3gyou3ue@34e37oox4d5m/
[4] https://rust-for-linux.com/nvme-driver

Re: [LSF/MM/BPF TOPIC] Rust block device driver APIs

[Hannes Reinecke]

On 1/29/24 14:17, Andreas Hindborg (Samsung) wrote:
> Hi All,
> 
> I would like to propose a session on the Rust block device driver APIs.
> 
> I submitted the APIs along with a simple null block driver as an RFC last year
> [1]. Since then I have kept the code in sync with latest mainline release [2],
> cleaned up the code, and added a few features.
> 
> After talking to some of you at various meetups over the past year, I think we
> have reached a point where we can potentially agree on merging initial Rust
> block layer support, along with the null block driver. To that end, I plan to
> send a few iterations of the patch set before LSF in May, so that we can use the
> session to discuss any remaining details.
> 
> Since Kent has also proposed a dedicated Rust session, we might find some
> synergy with this topic [3].
> 
> I also maintain an NVMe driver based on the Rust block APIs [4]. Due to
> community feedback, I have no plans for upstreaming this driver at the moment.
> However, it is a valuable tool for designing a sensible Rust block device API
> that is suitable real hardware.
> 
> Part of the NVMe patches are abstractions for PCI. Other users (drm) have
> expressed interest in these, so I plan to separate these in their own tree to
> make them easier to pick up for those users.
> 
> As a last note, I have recently become aware of ongoing work on
> implementing nbd in Rust. The work looks promising, and I hope the
> author will decide to send the patches, when they are ready to be
> shared.
> 
And I had an intern converting nbd to rust, which I could present as a 
proof-of-concept.
Or, rather, as a showcase how a (pretty basic) Rust driver would look 
like. Idea is to have a side-by-side comparison and figure out if
a) writing driver in Rust would simplify the code
b) existing/requiring functionality can be matched
or, in short, if Rust lives up to its promises.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                Kernel Storage Architect
hare@suse.de                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Ivo Totev, Andrew McDonald,
Werner Knoblich