Re: aio_read2() and aio_write2()

On Sat, Jan 13, 2024 at 5:51 PM Vinícius dos Santos Oliveira
<[email protected]> wrote:
>
> Em sáb., 13 de jan. de 2024 às 21:35, Konstantin Belousov
> <[email protected]> escreveu:
> > In principle the function of the new flag could be done like
> >         aio.aio_offset = lseek(fd, 0, SEEK_CUR);
> >         aio_read(&aio);
> > with the dis-advantage of requiring two syscalls instead of one.
>
> There are more disadvantages besides the extra syscall.
>
> I've built a green threading framework for Lua on top of Boost.Asio.
> However I also added support for sandboxed actors (capsicum and jails
> on FreeBSD). For file descriptors received through SCM_RIGHTS from
> sandboxed processes, I really don't want to risk blocking the thread
> (DoS) going through O_NONBLOCK (file IO is always ready, and that's
> why AIO exists). So, for these file descriptors (one can never know
> their "type"... whether they refer to pipes and O_NONBLOCK is enough,
> or they refer to files and AIO should be used), I'll use AIO as well.
> On Linux I can get away by just using io_uring for everything.

I don't understand.  What are you worried about that would block?
lseek never does.  It never actually goes to disk; it just retrieves a
variable from within the kernel.  So while the expense of a syscall is
non-zero, you don't have to worry about lseek blocking.