Yes, I mentioned this to Ulrich as one of the things that would make
sense.
Uli doesn't care that much about async syscalls, but I think that from a
kernel standpoint, we'd want to use this same indirect call for async
scheduling, rather than have two separate interfaces (because async
scheduling will want to have all the same flags intefaces for
open/socket/etc *too* and a doubly-indirect setup would be insanity!)
And it definitely fits the bill as a really simple syslet model for the
trivial case of doing just single system calls asynchronously.
Linus
-
