And how do you think posix_spawn is implemented?
replyThis is an oft-overlooked point. An obvious place to look for improving fork+execve is to see whether posix_spawn can be given more efficient kernel mechanisms to be based upon.
replyAnd of course that has already been done. On NetBSD, posix_spawn() is a fully-fledged system call and much of the work is done in kernel mode.
* https://blog.netbsd.org/tnf/entry/posix_spawn_syscall_added
This is literally discussed in the article this post links to.
replyNot really. They didn't get anywhere near as far as noticing the prior art of NetBSD, not even on the mailing list discussion behind that article.
replyposix_spawn addresses the need from userspace. Under the hood, it's still doing more or less a fork/exec, with the baggage that comes with it. A syscall would be nicer.
reply