On Tue, Aug 30 2022, Moritz Buhl <[email protected]> wrote:
> Hi tech@,
>
> the following diff only contains recvmmsg which should be the more useful
> syscall of the two.
>
> I implemented some minor feedback regarding the man page and attaching
> an error from recvit to the socket in case some messages were
> received before.
>
> I am also looking into passing the timeout through recvit and
> soreceive in order to not block indefinetly on a blocking socket
> as the other implementations do:
>
> BUGS
>
> The timeout argument does not work as intended. The timeout is
> checked only after the receipt of each datagram, so that if up to
> vlen-1 datagrams are received before the timeout expires, but
> then no further datagrams are received, the call will block
> forever.
> https://www.man7.org/linux/man-pages/man2/recvmmsg.2.html
>
> But I would prefer doing this in another change.
As discussed with Moritz, ports/net/knot (which I co-maintain) expects
to be able to use both recvmmsg and sendmmsg. I didn't check other
ports but from the code we read, using both recvmmsg and sendmmsg looks
legit (useful).
And since Moritz insists, please find my bikeshedding below. :)
tl;dr nothing looks wrong to me but I haven't tested it yet.
[...]
> Index: sys/kern/uipc_syscalls.c
> ===================================================================
> RCS file: /cvs/src/sys/kern/uipc_syscalls.c,v
> retrieving revision 1.201
> diff -u -p -r1.201 uipc_syscalls.c
> --- sys/kern/uipc_syscalls.c 14 Aug 2022 01:58:28 -0000 1.201
> +++ sys/kern/uipc_syscalls.c 30 Aug 2022 17:03:09 -0000
> @@ -805,6 +805,135 @@ done:
> }
>
> int
> +sys_recvmmsg(struct proc *p, void *v, register_t *retval)
> +{
> + struct sys_recvmmsg_args /* {
> + syscallarg(int) s;
> + syscallarg(struct mmsghdr *) mmsg;
> + syscallarg(unsigned int) vlen;
> + syscallarg(unsigned int) flags;
> + syscallarg(struct timespec *) timeout;
> + } */ *uap = v;
> + struct mmsghdr mmsg;
> + struct timespec ts, now;
> + struct iovec aiov[UIO_SMALLIOV], *uiov, *iov = aiov;
> + struct file *fp;
> + struct socket *so;
> + struct timespec *timeout;
> + unsigned int vlen, dg;
"dg" is a weird name. At first it looks like an index, I would name it
idx but if you want to keep the original name maybe put it on its own
line and add a comment like /* number of datagrams received */?
> + int error = 0, flags, s;
> +
> + timeout = SCARG(uap, timeout);
> + if (timeout != NULL) {
> + error = copyin(SCARG(uap, timeout), &ts, sizeof(ts));
^^^^^^^^^^^^^^^^^^^ could reuse the
timeout local variable
> + if (error != 0)
The local idiom is
if (error)
> + return error;
There are other occurrences below which you may or may not address if
you prefer to limit the differences with NetBSD.
> + getnanotime(&now);
> + timespecadd(&now, &ts, &ts);
> + }
> +
> + s = SCARG(uap, s);
> + if ((error = getsock(p, s, &fp)) != 0)
> + return (error);
> + so = (struct socket *)fp->f_data;
> +
> + flags = SCARG(uap, flags);
> +
> + vlen = SCARG(uap, vlen);
Maybe add a comment?
/* Arbitrarily capped at 1024 items. */
> + if (vlen > 1024)
> + vlen = 1024;
> +
> + for (dg = 0; dg < vlen;) {
> + error = copyin(SCARG(uap, mmsg) + dg, &mmsg, sizeof(mmsg));
We could have a struct mmsghdr *mmsgp local pointer and use mmsgp[idx]
instead of a mix of a macro and pointer arithmetics.
> + if (error != 0)
> + break;
> +
> + if (mmsg.msg_hdr.msg_iovlen > IOV_MAX) {
> + error = EMSGSIZE;
> + break;
> + }
> +
> + if (mmsg.msg_hdr.msg_iovlen > UIO_SMALLIOV)
> + iov = mallocarray(mmsg.msg_hdr.msg_iovlen,
> + sizeof(struct iovec), M_IOV, M_WAITOK);
> + else
> + iov = aiov;
> +
> + if (mmsg.msg_hdr.msg_iovlen > 0) {
> + error = copyin(mmsg.msg_hdr.msg_iov, iov,
> + mmsg.msg_hdr.msg_iovlen * sizeof(struct iovec));
> + if (error)
> + break;
> + }
> +
> + uiov = mmsg.msg_hdr.msg_iov;
> + mmsg.msg_hdr.msg_iov = iov;
> + mmsg.msg_hdr.msg_flags = flags;
> +
> + error = recvit(p, s, &mmsg.msg_hdr, NULL, retval);
> + if (error != 0) {
> + if (error == EAGAIN && dg > 0)
> + error = 0;
> + break;
> + }
> +
> + if (dg == 0 && flags & MSG_WAITFORONE) {
> + flags &= ~MSG_WAITFORONE;
> + flags |= MSG_DONTWAIT;
> + }
> +
> + mmsg.msg_hdr.msg_iov = uiov;
> + mmsg.msg_len = *retval;
> +#ifdef KTRACE
> + if (KTRPOINT(p, KTR_STRUCT)) {
> + ktrmsghdr(p, &mmsg.msg_hdr);
> + if (mmsg.msg_hdr.msg_iovlen)
> + ktriovec(p, iov, mmsg.msg_hdr.msg_iovlen);
> + }
> +#endif
> +
> + error = copyout(&mmsg, SCARG(uap, mmsg) + dg, sizeof(mmsg));
Same array indexing hidden behind pointer arithmetics.
> + if (error != 0)
> + break;
> +
> + if (iov != aiov) {
> + free(iov, M_IOV, sizeof(struct iovec) *
> + mmsg.msg_hdr.msg_iovlen);
> + iov = aiov;
> + }
> +
> + dg++;
> + if (mmsg.msg_hdr.msg_flags & MSG_OOB)
> + break;
> +
> + if (timeout != NULL) {
> + getnanotime(&now);
> + timespecsub(&now, &ts, &now);
> + if (now.tv_sec > 0)
> + break;
> + }
> + }
> +
> + *retval = dg;
> +
> + /*
> + * If we succeeded at least once, return 0, hopefully so->so_error
> + * will catch it next time.
> + */
> + if (error && dg > 0) {
> + so->so_error = error;
> + error = 0;
> + }
> +
> + if (iov != aiov)
> + free(iov, M_IOV,
> + sizeof(struct iovec) * mmsg.msg_hdr.msg_iovlen);
> +
> + FRELE(fp, p);
> + return (error);
> +}
> +
> +int
> recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
> register_t *retsize)
> {
--
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE