Hi, I do actually have a work-in-progress NVMe driver for xNVMe based on Driverkit. At the moment, it's only doing one command at a time (queue depth = 1), and blocks until the command returns. It's plenty fast for prototyping, but not for any high performance work. I hope to make it async and much, much faster in the future, but this is at least a proof-of-concept. I need to fix up the memory management before I feel confident in releasing the first beta version. I'll take a look at it tomorrow. What is your use-case? Maybe you can come to the xNVMe Discord server and we can discuss what you are looking for.

Hi, I do actually have a work-in-progress NVMe driver for xNVMe based on Driverkit. At the moment, it's only doing one command at a time (queue depth = 1), and blocks until the command returns. It's plenty fast for prototyping, but not for any high performance work. I hope to make it async and much, much faster in the future, but this is at least a proof-of-concept. I'm now in the process of releasing the code as the first beta. What is your use-case? Maybe you can come to the xNVMe Discord server and we can discuss what you are looking for?

I've posted the same question on StackOverflow. Together with user pmdj, we've concluded that it isn't available in the API, and reported the issue to Apple. https://stackoverflow.com/questions/77840032/sharing-a-ring-buffer-between-driver-and-user-space-in-driverkit

The NVMe driver is now released, and supports larger queue depths and is much, much faster now (benchmarking 1000k IOPS) https://github.com/OpenMPDK/MacVFN