In theory FreeBSD supports a sane aio_read/write/fsync() implementation by extending sigevent to include the generation of kqueue events, but there are no aio_o

In theory FreeBSD supports a sane aio_read/write/fsync() implementation by extending sigevent to include the generation of kqueue events, but there are no aio_open/close/fstat/rename() syscalls and if you need a thread pool to deal with those remaining blocking file system calls there are few advantages to using the in-kernel thread pool implementing the aio_*() syscalls. An other problem is that filesystem aio calls are limited to local file systems, because accesses to network file systems could block indefinitely starving local file system accesses by blocking all threads in the pool.