I don’t think FreeBSD is ‘sticking with kqueue’. io_uring seems mostly well designed and I don’t think there would be any objection to someone adding it to FreeBSD, it just hasn’t been done yet. Patches welcome! There’s a bit less of a need for it, because the AIO subsystem on FreeBSD works pretty well (unlike Linux) and lio_listio is actually usable, but io_uring has the benefit of a couple of decades of learning from it, NetMap, DPDK, and so on. Now that the API is fairly stable, I don’t think there’s be any objections to adding it (and the Linux ABI layer will need it whether it’s added natively or not - I hope a native one won’t adopt the Linux mistake of using 64-bit integers for pointers in userspace ABIs).
I don’t think FreeBSD is ‘sticking with kqueue’. io_uring seems mostly well designed and I don’t think there would be any objection to someone adding it to FreeBSD, it just hasn’t been done yet. Patches welcome! There’s a bit less of a need for it, because the AIO subsystem on FreeBSD works pretty well (unlike Linux) and lio_listio is actually usable, but io_uring has the benefit of a couple of decades of learning from it, NetMap, DPDK, and so on. Now that the API is fairly stable, I don’t think there’s be any objections to adding it (and the Linux ABI layer will need it whether it’s added natively or not - I hope a native one won’t adopt the Linux mistake of using 64-bit integers for pointers in userspace ABIs).
Why FreeBSD’s linux compat layer is ‘dubious’ ?
This is so exciting. This is definitely the future of the linux syscalls system.