uipc_socket.c

来自「基于组件方式开发操作系统的OSKIT源代码」· C语言 代码 · 共 1,219 行 · 第 1/2 页

bad:
		if (m)
			m_freem(m);
		return (error);
	}
	if (mp)
		*mp = (struct mbuf *)0;
	if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
		(*pr->pr_usrreqs->pru_rcvd)(so, 0);

restart:
	error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
	if (error)
		return (error);
	s = splnet();

	m = so->so_rcv.sb_mb;
	/*
	 * If we have less data than requested, block awaiting more
	 * (subject to any timeout) if:
	 *   1. the current count is less than the low water mark, or
	 *   2. MSG_WAITALL is set, and it is possible to do the entire
	 *	receive operation at once if we block (resid <= hiwat).
	 *   3. MSG_DONTWAIT is not set
	 * If MSG_WAITALL is set but resid is larger than the receive buffer,
	 * we have to do the receive in sections, and thus risk returning
	 * a short count if a timeout or signal occurs after we start.
	 */
	if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
	    so->so_rcv.sb_cc < uio->uio_resid) &&
	    (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
	    ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
	    m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
		KASSERT(m != 0 || !so->so_rcv.sb_cc, ("receive 1"));
		if (so->so_error) {
			if (m)
				goto dontblock;
			error = so->so_error;
			if ((flags & MSG_PEEK) == 0)
				so->so_error = 0;
			goto release;
		}
		if (so->so_state & SS_CANTRCVMORE) {
			if (m)
				goto dontblock;
			else
				goto release;
		}
		for (; m; m = m->m_next)
			if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
				m = so->so_rcv.sb_mb;
				goto dontblock;
			}
		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
		    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
			error = ENOTCONN;
			goto release;
		}
		if (uio->uio_resid == 0)
			goto release;
		if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
			error = EWOULDBLOCK;
			goto release;
		}
		sbunlock(&so->so_rcv);
		error = sbwait(&so->so_rcv);
		splx(s);
		if (error)
			return (error);
		goto restart;
	}
dontblock:
	if (uio->uio_procp)
		uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
	nextrecord = m->m_nextpkt;
	if (pr->pr_flags & PR_ADDR) {
		KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
		orig_resid = 0;
		if (psa)
			*psa = dup_sockaddr(mtod(m, struct sockaddr *),
					    mp0 == 0);
		if (flags & MSG_PEEK) {
			m = m->m_next;
		} else {
			sbfree(&so->so_rcv, m);
			MFREE(m, so->so_rcv.sb_mb);
			m = so->so_rcv.sb_mb;
		}
	}
	while (m && m->m_type == MT_CONTROL && error == 0) {
		if (flags & MSG_PEEK) {
			if (controlp)
				*controlp = m_copy(m, 0, m->m_len);
			m = m->m_next;
		} else {
			sbfree(&so->so_rcv, m);
			if (controlp) {
				if (pr->pr_domain->dom_externalize &&
				    mtod(m, struct cmsghdr *)->cmsg_type ==
				    SCM_RIGHTS)
				   error = (*pr->pr_domain->dom_externalize)(m);
				*controlp = m;
				so->so_rcv.sb_mb = m->m_next;
				m->m_next = 0;
				m = so->so_rcv.sb_mb;
			} else {
				MFREE(m, so->so_rcv.sb_mb);
				m = so->so_rcv.sb_mb;
			}
		}
		if (controlp) {
			orig_resid = 0;
			controlp = &(*controlp)->m_next;
		}
	}
	if (m) {
		if ((flags & MSG_PEEK) == 0)
			m->m_nextpkt = nextrecord;
		type = m->m_type;
		if (type == MT_OOBDATA)
			flags |= MSG_OOB;
	}
	moff = 0;
	offset = 0;
	while (m && uio->uio_resid > 0 && error == 0) {
		if (m->m_type == MT_OOBDATA) {
			if (type != MT_OOBDATA)
				break;
		} else if (type == MT_OOBDATA)
			break;
		else
		    KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
			("receive 3"));
		so->so_state &= ~SS_RCVATMARK;
		len = uio->uio_resid;
		if (so->so_oobmark && len > so->so_oobmark - offset)
			len = so->so_oobmark - offset;
		if (len > m->m_len - moff)
			len = m->m_len - moff;
		/*
		 * If mp is set, just pass back the mbufs.
		 * Otherwise copy them out via the uio, then free.
		 * Sockbuf must be consistent here (points to current mbuf,
		 * it points to next record) when we drop priority;
		 * we must note any additions to the sockbuf when we
		 * block interrupts again.
		 */
		if (mp == 0) {
			splx(s);
			error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
			s = splnet();
			if (error)
				goto release;
		} else
			uio->uio_resid -= len;
		if (len == m->m_len - moff) {
			if (m->m_flags & M_EOR)
				flags |= MSG_EOR;
			if (flags & MSG_PEEK) {
				m = m->m_next;
				moff = 0;
			} else {
				nextrecord = m->m_nextpkt;
				sbfree(&so->so_rcv, m);
				if (mp) {
					*mp = m;
					mp = &m->m_next;
					so->so_rcv.sb_mb = m = m->m_next;
					*mp = (struct mbuf *)0;
				} else {
					MFREE(m, so->so_rcv.sb_mb);
					m = so->so_rcv.sb_mb;
				}
				if (m)
					m->m_nextpkt = nextrecord;
			}
		} else {
			if (flags & MSG_PEEK)
				moff += len;
			else {
				if (mp)
					*mp = m_copym(m, 0, len, M_WAIT);
				m->m_data += len;
				m->m_len -= len;
				so->so_rcv.sb_cc -= len;
			}
		}
		if (so->so_oobmark) {
			if ((flags & MSG_PEEK) == 0) {
				so->so_oobmark -= len;
				if (so->so_oobmark == 0) {
					so->so_state |= SS_RCVATMARK;
					break;
				}
			} else {
				offset += len;
				if (offset == so->so_oobmark)
					break;
			}
		}
		if (flags & MSG_EOR)
			break;
		/*
		 * If the MSG_WAITALL flag is set (for non-atomic socket),
		 * we must not quit until "uio->uio_resid == 0" or an error
		 * termination.  If a signal/timeout occurs, return
		 * with a short count but without error.
		 * Keep sockbuf locked against other readers.
		 */
		while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
		    !sosendallatonce(so) && !nextrecord) {
			if (so->so_error || so->so_state & SS_CANTRCVMORE)
				break;
			error = sbwait(&so->so_rcv);
			if (error) {
				sbunlock(&so->so_rcv);
				splx(s);
				return (0);
			}
			m = so->so_rcv.sb_mb;
			if (m)
				nextrecord = m->m_nextpkt;
		}
	}

	if (m && pr->pr_flags & PR_ATOMIC) {
		flags |= MSG_TRUNC;
		if ((flags & MSG_PEEK) == 0)
			(void) sbdroprecord(&so->so_rcv);
	}
	if ((flags & MSG_PEEK) == 0) {
		if (m == 0)
			so->so_rcv.sb_mb = nextrecord;
		if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
			(*pr->pr_usrreqs->pru_rcvd)(so, flags);
	}
	if (orig_resid == uio->uio_resid && orig_resid &&
	    (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
		sbunlock(&so->so_rcv);
		splx(s);
		goto restart;
	}

	if (flagsp)
		*flagsp |= flags;
release:
	sbunlock(&so->so_rcv);
	splx(s);
	return (error);
}

int
soshutdown(so, how)
	register struct socket *so;
	register int how;
{
	register struct protosw *pr = so->so_proto;

	how++;
	if (how & FREAD)
		sorflush(so);
	if (how & FWRITE)
		return ((*pr->pr_usrreqs->pru_shutdown)(so));
	return (0);
}

void
sorflush(so)
	register struct socket *so;
{
	register struct sockbuf *sb = &so->so_rcv;
	register struct protosw *pr = so->so_proto;
	register int s;
	struct sockbuf asb;

	sb->sb_flags |= SB_NOINTR;
	(void) sblock(sb, M_WAITOK);
	s = splimp();
	socantrcvmore(so);
	sbunlock(sb);
	asb = *sb;
	bzero((caddr_t)sb, sizeof (*sb));
	splx(s);
	if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
		(*pr->pr_domain->dom_dispose)(asb.sb_mb);
	sbrelease(&asb);
}

/*
 * Perhaps this routine, and sooptcopyout(), below, ought to come in
 * an additional variant to handle the case where the option value needs
 * to be some kind of integer, but not a specific size.
 * In addition to their use here, these functions are also called by the
 * protocol-level pr_ctloutput() routines.
 */
int
sooptcopyin(sopt, buf, len, minlen)
	struct	sockopt *sopt;
	void	*buf;
	size_t	len;
	size_t	minlen;
{
	size_t	valsize;

	/*
	 * If the user gives us more than we wanted, we ignore it,
	 * but if we don't get the minimum length the caller
	 * wants, we return EINVAL.  On success, sopt->sopt_valsize
	 * is set to however much we actually retrieved.
	 */
	if ((valsize = sopt->sopt_valsize) < minlen)
		return EINVAL;
	if (valsize > len)
		sopt->sopt_valsize = valsize = len;

	if (sopt->sopt_p != 0)
		return (copyin(sopt->sopt_val, buf, valsize));

	bcopy(sopt->sopt_val, buf, valsize);
	return 0;
}

int
sosetopt(so, sopt)
	struct socket *so;
	struct sockopt *sopt;
{
	int	error, optval;
	struct	linger l;
	struct	timeval tv;
	short	val;

	error = 0;
	if (sopt->sopt_level != SOL_SOCKET) {
		if (so->so_proto && so->so_proto->pr_ctloutput)
			return ((*so->so_proto->pr_ctloutput)
				  (so, sopt));
		error = ENOPROTOOPT;
	} else {
		switch (sopt->sopt_name) {
		case SO_LINGER:
			error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
			if (error)
				goto bad;

			so->so_linger = l.l_linger;
			if (l.l_onoff)
				so->so_options |= SO_LINGER;
			else
				so->so_options &= ~SO_LINGER;
			break;

		case SO_DEBUG:
		case SO_KEEPALIVE:
		case SO_DONTROUTE:
		case SO_USELOOPBACK:
		case SO_BROADCAST:
		case SO_REUSEADDR:
		case SO_REUSEPORT:
		case SO_OOBINLINE:
		case SO_TIMESTAMP:
			error = sooptcopyin(sopt, &optval, sizeof optval,
					    sizeof optval);
			if (error)
				goto bad;
			if (optval)
				so->so_options |= sopt->sopt_name;
			else
				so->so_options &= ~sopt->sopt_name;
			break;

		case SO_SNDBUF:
		case SO_RCVBUF:
		case SO_SNDLOWAT:
		case SO_RCVLOWAT:
			error = sooptcopyin(sopt, &optval, sizeof optval,
					    sizeof optval);
			if (error)
				goto bad;

			/*
			 * Values < 1 make no sense for any of these
			 * options, so disallow them.
			 */
			if (optval < 1) {
				error = EINVAL;
				goto bad;
			}

			switch (sopt->sopt_name) {
			case SO_SNDBUF:
			case SO_RCVBUF:
				if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
					      &so->so_snd : &so->so_rcv,
					      (u_long) optval) == 0) {
					error = ENOBUFS;
					goto bad;
				}
				break;

			/*
			 * Make sure the low-water is never greater than
			 * the high-water.
			 */
			case SO_SNDLOWAT:
				so->so_snd.sb_lowat =
				    (optval > so->so_snd.sb_hiwat) ?
				    so->so_snd.sb_hiwat : optval;
				break;
			case SO_RCVLOWAT:
				so->so_rcv.sb_lowat =
				    (optval > so->so_rcv.sb_hiwat) ?
				    so->so_rcv.sb_hiwat : optval;
				break;
			}
			break;

		case SO_SNDTIMEO:
		case SO_RCVTIMEO:
			error = sooptcopyin(sopt, &tv, sizeof tv,
					    sizeof tv);
			if (error)
				goto bad;

			if (tv.tv_sec > SHRT_MAX / hz - hz) {
				error = EDOM;
				goto bad;
			}
			val = tv.tv_sec * hz + tv.tv_usec / tick;

			switch (sopt->sopt_name) {
			case SO_SNDTIMEO:
				so->so_snd.sb_timeo = val;
				break;
			case SO_RCVTIMEO:
				so->so_rcv.sb_timeo = val;
				break;
			}
			break;

		default:
			error = ENOPROTOOPT;
			break;
		}
		if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
			(void) ((*so->so_proto->pr_ctloutput)
				  (so, sopt));
		}
	}
bad:
	return (error);
}

/* Helper routine for getsockopt */
int
sooptcopyout(sopt, buf, len)
	struct	sockopt *sopt;
	void	*buf;
	size_t	len;
{
	int	error;
	size_t	valsize;

	error = 0;

	/*
	 * Documented get behavior is that we always return a value,
	 * possibly truncated to fit in the user's buffer.
	 * Traditional behavior is that we always tell the user
	 * precisely how much we copied, rather than something useful
	 * like the total amount we had available for her.
	 * Note that this interface is not idempotent; the entire answer must
	 * generated ahead of time.
	 */
	valsize = min(len, sopt->sopt_valsize);
	sopt->sopt_valsize = valsize;
	if (sopt->sopt_val != 0) {
		if (sopt->sopt_p != 0)
			error = copyout(buf, sopt->sopt_val, valsize);
		else
			bcopy(buf, sopt->sopt_val, valsize);
	}
	return error;
}

int
sogetopt(so, sopt)
	struct socket *so;
	struct sockopt *sopt;
{
	int	error, optval;
	struct	linger l;
	struct	timeval tv;

	error = 0;
	if (sopt->sopt_level != SOL_SOCKET) {
		if (so->so_proto && so->so_proto->pr_ctloutput) {
			return ((*so->so_proto->pr_ctloutput)
				  (so, sopt));
		} else
			return (ENOPROTOOPT);
	} else {
		switch (sopt->sopt_name) {
		case SO_LINGER:
			l.l_onoff = so->so_options & SO_LINGER;
			l.l_linger = so->so_linger;
			error = sooptcopyout(sopt, &l, sizeof l);
			break;

		case SO_USELOOPBACK:
		case SO_DONTROUTE:
		case SO_DEBUG:
		case SO_KEEPALIVE:
		case SO_REUSEADDR:
		case SO_REUSEPORT:
		case SO_BROADCAST:
		case SO_OOBINLINE:
		case SO_TIMESTAMP:
			optval = so->so_options & sopt->sopt_name;
integer:
			error = sooptcopyout(sopt, &optval, sizeof optval);
			break;

		case SO_TYPE:
			optval = so->so_type;
			goto integer;

		case SO_ERROR:
			optval = so->so_error;
			so->so_error = 0;
			goto integer;

		case SO_SNDBUF:
			optval = so->so_snd.sb_hiwat;
			goto integer;

		case SO_RCVBUF:
			optval = so->so_rcv.sb_hiwat;
			goto integer;

		case SO_SNDLOWAT:
			optval = so->so_snd.sb_lowat;
			goto integer;

		case SO_RCVLOWAT:
			optval = so->so_rcv.sb_lowat;
			goto integer;

		case SO_SNDTIMEO:
		case SO_RCVTIMEO:
			optval = (sopt->sopt_name == SO_SNDTIMEO ?
				  so->so_snd.sb_timeo : so->so_rcv.sb_timeo);

			tv.tv_sec = optval / hz;
			tv.tv_usec = (optval % hz) * tick;
			error = sooptcopyout(sopt, &tv, sizeof tv);
			break;			

		default:
			error = ENOPROTOOPT;
			break;
		}
		return (error);
	}
}

void
sohasoutofband(so)
	register struct socket *so;
{
	if (so->so_sigio != NULL)
		pgsigio(so->so_sigio, SIGURG, 0);
	selwakeup(&so->so_rcv.sb_sel);
}

int
sopoll(struct socket *so, int events, struct ucred *cred, struct proc *p)
{
	int revents = 0;
	int s = splnet();

	if (events & (POLLIN | POLLRDNORM))
		if (soreadable(so))
			revents |= events & (POLLIN | POLLRDNORM);

	if (events & (POLLOUT | POLLWRNORM))
		if (sowriteable(so))
			revents |= events & (POLLOUT | POLLWRNORM);

	if (events & (POLLPRI | POLLRDBAND))
		if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
			revents |= events & (POLLPRI | POLLRDBAND);

	if (revents == 0) {
		if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
			selrecord(p, &so->so_rcv.sb_sel);
			so->so_rcv.sb_flags |= SB_SEL;
		}

		if (events & (POLLOUT | POLLWRNORM)) {
			selrecord(p, &so->so_snd.sb_sel);
			so->so_snd.sb_flags |= SB_SEL;
		}
	}

	splx(s);
	return (revents);
}