Manos Pitsidianakis
This previous weekend I wrote and Submitted Rust bindings for virtio drivers in the Linux kernel mailing list including a WIP sample driver for virtio-rtc device https://lore.kernel.org/all/20260505-r[email protected]/
#Virtualization #RustForLinux #kernel #RustLang #virtio
[PATCH RFC 0/6] Add Rust virtio bindings and sample device - Manos Pitsidianakis

May 7 at 7:25amWeb
Show threadManos PitsidianakisMay 17
With a few changes, a complete virtio-rng rust driver currently requires only **one** `unsafe` use, which can potentially be eliminated too. The virtio bindings have to use unsafe, but its consumers do not need to worry about it.
Show threadManos PitsidianakisMay 17

Also, the Rust virtio-rng performance looks to be completely on par with the C one, when using rngtest:

C:

rngtest: input channel speed: (min=544.957; avg=7819.738; max=19073.486)Mibits/s
rngtest: FIPS tests speed: (min=55.446; avg=107.397; max=111.541)Mibits/s

Rust:

rngtest: input channel speed: (min=476.837; avg=7909.806; max=19073.486)Mibits/s
rngtest: FIPS tests speed: (min=48.165; avg=107.335; max=111.541)Mibits/s

Show threadManos PitsidianakisMay 17
Caveat: the virtio-rng device is one of the simplest in functionality, even if it can potentially do a lot of I/O, so this was expected.
Show threadнабMay 17
@epilys have you reproed the ~12% lower min?
Show threadManos PitsidianakisMay 17
@nabijaczleweli the min/max results do not seem to be stable, it probably depends on available host entropy, and the same behavior appears in both C and Rust