1/*	$NetBSD: svc_vc.c,v 1.7 2000/08/03 00:01:53 fvdl Exp $	*/
2
3/*-
4 * Copyright (c) 2009, Sun Microsystems, Inc.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions are met:
9 * - Redistributions of source code must retain the above copyright notice,
10 *   this list of conditions and the following disclaimer.
11 * - Redistributions in binary form must reproduce the above copyright notice,
12 *   this list of conditions and the following disclaimer in the documentation
13 *   and/or other materials provided with the distribution.
14 * - Neither the name of Sun Microsystems, Inc. nor the names of its
15 *   contributors may be used to endorse or promote products derived
16 *   from this software without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
19 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
22 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28 * POSSIBILITY OF SUCH DAMAGE.
29 */
30
31#if defined(LIBC_SCCS) && !defined(lint)
32static char *sccsid2 = "@(#)svc_tcp.c 1.21 87/08/11 Copyr 1984 Sun Micro";
33static char *sccsid = "@(#)svc_tcp.c	2.2 88/08/01 4.0 RPCSRC";
34#endif
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: stable/10/sys/rpc/svc_vc.c 309503 2016-12-03 19:03:25Z ngie $");
37
38/*
39 * svc_vc.c, Server side for Connection Oriented based RPC.
40 *
41 * Actually implements two flavors of transporter -
42 * a tcp rendezvouser (a listner and connection establisher)
43 * and a record/tcp stream.
44 */
45
46#include <sys/param.h>
47#include <sys/lock.h>
48#include <sys/kernel.h>
49#include <sys/malloc.h>
50#include <sys/mbuf.h>
51#include <sys/mutex.h>
52#include <sys/proc.h>
53#include <sys/protosw.h>
54#include <sys/queue.h>
55#include <sys/socket.h>
56#include <sys/socketvar.h>
57#include <sys/sx.h>
58#include <sys/systm.h>
59#include <sys/uio.h>
60
61#include <net/vnet.h>
62
63#include <netinet/tcp.h>
64
65#include <rpc/rpc.h>
66
67#include <rpc/krpc.h>
68#include <rpc/rpc_com.h>
69
70#include <security/mac/mac_framework.h>
71
72static bool_t svc_vc_rendezvous_recv(SVCXPRT *, struct rpc_msg *,
73    struct sockaddr **, struct mbuf **);
74static enum xprt_stat svc_vc_rendezvous_stat(SVCXPRT *);
75static void svc_vc_rendezvous_destroy(SVCXPRT *);
76static bool_t svc_vc_null(void);
77static void svc_vc_destroy(SVCXPRT *);
78static enum xprt_stat svc_vc_stat(SVCXPRT *);
79static bool_t svc_vc_ack(SVCXPRT *, uint32_t *);
80static bool_t svc_vc_recv(SVCXPRT *, struct rpc_msg *,
81    struct sockaddr **, struct mbuf **);
82static bool_t svc_vc_reply(SVCXPRT *, struct rpc_msg *,
83    struct sockaddr *, struct mbuf *, uint32_t *seq);
84static bool_t svc_vc_control(SVCXPRT *xprt, const u_int rq, void *in);
85static bool_t svc_vc_rendezvous_control (SVCXPRT *xprt, const u_int rq,
86    void *in);
87static void svc_vc_backchannel_destroy(SVCXPRT *);
88static enum xprt_stat svc_vc_backchannel_stat(SVCXPRT *);
89static bool_t svc_vc_backchannel_recv(SVCXPRT *, struct rpc_msg *,
90    struct sockaddr **, struct mbuf **);
91static bool_t svc_vc_backchannel_reply(SVCXPRT *, struct rpc_msg *,
92    struct sockaddr *, struct mbuf *, uint32_t *);
93static bool_t svc_vc_backchannel_control(SVCXPRT *xprt, const u_int rq,
94    void *in);
95static SVCXPRT *svc_vc_create_conn(SVCPOOL *pool, struct socket *so,
96    struct sockaddr *raddr);
97static int svc_vc_accept(struct socket *head, struct socket **sop);
98static int svc_vc_soupcall(struct socket *so, void *arg, int waitflag);
99
100static struct xp_ops svc_vc_rendezvous_ops = {
101	.xp_recv =	svc_vc_rendezvous_recv,
102	.xp_stat =	svc_vc_rendezvous_stat,
103	.xp_reply =	(bool_t (*)(SVCXPRT *, struct rpc_msg *,
104		struct sockaddr *, struct mbuf *, uint32_t *))svc_vc_null,
105	.xp_destroy =	svc_vc_rendezvous_destroy,
106	.xp_control =	svc_vc_rendezvous_control
107};
108
109static struct xp_ops svc_vc_ops = {
110	.xp_recv =	svc_vc_recv,
111	.xp_stat =	svc_vc_stat,
112	.xp_ack =	svc_vc_ack,
113	.xp_reply =	svc_vc_reply,
114	.xp_destroy =	svc_vc_destroy,
115	.xp_control =	svc_vc_control
116};
117
118static struct xp_ops svc_vc_backchannel_ops = {
119	.xp_recv =	svc_vc_backchannel_recv,
120	.xp_stat =	svc_vc_backchannel_stat,
121	.xp_reply =	svc_vc_backchannel_reply,
122	.xp_destroy =	svc_vc_backchannel_destroy,
123	.xp_control =	svc_vc_backchannel_control
124};
125
126/*
127 * Usage:
128 *	xprt = svc_vc_create(sock, send_buf_size, recv_buf_size);
129 *
130 * Creates, registers, and returns a (rpc) tcp based transporter.
131 * Once *xprt is initialized, it is registered as a transporter
132 * see (svc.h, xprt_register).  This routine returns
133 * a NULL if a problem occurred.
134 *
135 * The filedescriptor passed in is expected to refer to a bound, but
136 * not yet connected socket.
137 *
138 * Since streams do buffered io similar to stdio, the caller can specify
139 * how big the send and receive buffers are via the second and third parms;
140 * 0 => use the system default.
141 */
142SVCXPRT *
143svc_vc_create(SVCPOOL *pool, struct socket *so, size_t sendsize,
144    size_t recvsize)
145{
146	SVCXPRT *xprt;
147	struct sockaddr* sa;
148	int error;
149
150	SOCK_LOCK(so);
151	if (so->so_state & (SS_ISCONNECTED|SS_ISDISCONNECTED)) {
152		SOCK_UNLOCK(so);
153		CURVNET_SET(so->so_vnet);
154		error = so->so_proto->pr_usrreqs->pru_peeraddr(so, &sa);
155		CURVNET_RESTORE();
156		if (error)
157			return (NULL);
158		xprt = svc_vc_create_conn(pool, so, sa);
159		free(sa, M_SONAME);
160		return (xprt);
161	}
162	SOCK_UNLOCK(so);
163
164	xprt = svc_xprt_alloc();
165	sx_init(&xprt->xp_lock, "xprt->xp_lock");
166	xprt->xp_pool = pool;
167	xprt->xp_socket = so;
168	xprt->xp_p1 = NULL;
169	xprt->xp_p2 = NULL;
170	xprt->xp_ops = &svc_vc_rendezvous_ops;
171
172	CURVNET_SET(so->so_vnet);
173	error = so->so_proto->pr_usrreqs->pru_sockaddr(so, &sa);
174	CURVNET_RESTORE();
175	if (error) {
176		goto cleanup_svc_vc_create;
177	}
178
179	memcpy(&xprt->xp_ltaddr, sa, sa->sa_len);
180	free(sa, M_SONAME);
181
182	xprt_register(xprt);
183
184	solisten(so, -1, curthread);
185
186	SOCKBUF_LOCK(&so->so_rcv);
187	xprt->xp_upcallset = 1;
188	soupcall_set(so, SO_RCV, svc_vc_soupcall, xprt);
189	SOCKBUF_UNLOCK(&so->so_rcv);
190
191	return (xprt);
192
193cleanup_svc_vc_create:
194	sx_destroy(&xprt->xp_lock);
195	svc_xprt_free(xprt);
196
197	return (NULL);
198}
199
200/*
201 * Create a new transport for a socket optained via soaccept().
202 */
203SVCXPRT *
204svc_vc_create_conn(SVCPOOL *pool, struct socket *so, struct sockaddr *raddr)
205{
206	SVCXPRT *xprt;
207	struct cf_conn *cd;
208	struct sockaddr* sa = NULL;
209	struct sockopt opt;
210	int one = 1;
211	int error;
212
213	bzero(&opt, sizeof(struct sockopt));
214	opt.sopt_dir = SOPT_SET;
215	opt.sopt_level = SOL_SOCKET;
216	opt.sopt_name = SO_KEEPALIVE;
217	opt.sopt_val = &one;
218	opt.sopt_valsize = sizeof(one);
219	error = sosetopt(so, &opt);
220	if (error) {
221		return (NULL);
222	}
223
224	if (so->so_proto->pr_protocol == IPPROTO_TCP) {
225		bzero(&opt, sizeof(struct sockopt));
226		opt.sopt_dir = SOPT_SET;
227		opt.sopt_level = IPPROTO_TCP;
228		opt.sopt_name = TCP_NODELAY;
229		opt.sopt_val = &one;
230		opt.sopt_valsize = sizeof(one);
231		error = sosetopt(so, &opt);
232		if (error) {
233			return (NULL);
234		}
235	}
236
237	cd = mem_alloc(sizeof(*cd));
238	cd->strm_stat = XPRT_IDLE;
239
240	xprt = svc_xprt_alloc();
241	sx_init(&xprt->xp_lock, "xprt->xp_lock");
242	xprt->xp_pool = pool;
243	xprt->xp_socket = so;
244	xprt->xp_p1 = cd;
245	xprt->xp_p2 = NULL;
246	xprt->xp_ops = &svc_vc_ops;
247
248	/*
249	 * See http://www.connectathon.org/talks96/nfstcp.pdf - client
250	 * has a 5 minute timer, server has a 6 minute timer.
251	 */
252	xprt->xp_idletimeout = 6 * 60;
253
254	memcpy(&xprt->xp_rtaddr, raddr, raddr->sa_len);
255
256	CURVNET_SET(so->so_vnet);
257	error = so->so_proto->pr_usrreqs->pru_sockaddr(so, &sa);
258	CURVNET_RESTORE();
259	if (error)
260		goto cleanup_svc_vc_create;
261
262	memcpy(&xprt->xp_ltaddr, sa, sa->sa_len);
263	free(sa, M_SONAME);
264
265	xprt_register(xprt);
266
267	SOCKBUF_LOCK(&so->so_rcv);
268	xprt->xp_upcallset = 1;
269	soupcall_set(so, SO_RCV, svc_vc_soupcall, xprt);
270	SOCKBUF_UNLOCK(&so->so_rcv);
271
272	/*
273	 * Throw the transport into the active list in case it already
274	 * has some data buffered.
275	 */
276	sx_xlock(&xprt->xp_lock);
277	xprt_active(xprt);
278	sx_xunlock(&xprt->xp_lock);
279
280	return (xprt);
281cleanup_svc_vc_create:
282	sx_destroy(&xprt->xp_lock);
283	svc_xprt_free(xprt);
284	mem_free(cd, sizeof(*cd));
285
286	return (NULL);
287}
288
289/*
290 * Create a new transport for a backchannel on a clnt_vc socket.
291 */
292SVCXPRT *
293svc_vc_create_backchannel(SVCPOOL *pool)
294{
295	SVCXPRT *xprt = NULL;
296	struct cf_conn *cd = NULL;
297
298	cd = mem_alloc(sizeof(*cd));
299	cd->strm_stat = XPRT_IDLE;
300
301	xprt = svc_xprt_alloc();
302	sx_init(&xprt->xp_lock, "xprt->xp_lock");
303	xprt->xp_pool = pool;
304	xprt->xp_socket = NULL;
305	xprt->xp_p1 = cd;
306	xprt->xp_p2 = NULL;
307	xprt->xp_ops = &svc_vc_backchannel_ops;
308	return (xprt);
309}
310
311/*
312 * This does all of the accept except the final call to soaccept. The
313 * caller will call soaccept after dropping its locks (soaccept may
314 * call malloc).
315 */
316int
317svc_vc_accept(struct socket *head, struct socket **sop)
318{
319	int error = 0;
320	struct socket *so;
321
322	if ((head->so_options & SO_ACCEPTCONN) == 0) {
323		error = EINVAL;
324		goto done;
325	}
326#ifdef MAC
327	error = mac_socket_check_accept(curthread->td_ucred, head);
328	if (error != 0)
329		goto done;
330#endif
331	ACCEPT_LOCK();
332	if (TAILQ_EMPTY(&head->so_comp)) {
333		ACCEPT_UNLOCK();
334		error = EWOULDBLOCK;
335		goto done;
336	}
337	so = TAILQ_FIRST(&head->so_comp);
338	KASSERT(!(so->so_qstate & SQ_INCOMP), ("svc_vc_accept: so SQ_INCOMP"));
339	KASSERT(so->so_qstate & SQ_COMP, ("svc_vc_accept: so not SQ_COMP"));
340
341	/*
342	 * Before changing the flags on the socket, we have to bump the
343	 * reference count.  Otherwise, if the protocol calls sofree(),
344	 * the socket will be released due to a zero refcount.
345	 * XXX might not need soref() since this is simpler than kern_accept.
346	 */
347	SOCK_LOCK(so);			/* soref() and so_state update */
348	soref(so);			/* file descriptor reference */
349
350	TAILQ_REMOVE(&head->so_comp, so, so_list);
351	head->so_qlen--;
352	so->so_state |= (head->so_state & SS_NBIO);
353	so->so_qstate &= ~SQ_COMP;
354	so->so_head = NULL;
355
356	SOCK_UNLOCK(so);
357	ACCEPT_UNLOCK();
358
359	*sop = so;
360
361	/* connection has been removed from the listen queue */
362	KNOTE_UNLOCKED(&head->so_rcv.sb_sel.si_note, 0);
363done:
364	return (error);
365}
366
367/*ARGSUSED*/
368static bool_t
369svc_vc_rendezvous_recv(SVCXPRT *xprt, struct rpc_msg *msg,
370    struct sockaddr **addrp, struct mbuf **mp)
371{
372	struct socket *so = NULL;
373	struct sockaddr *sa = NULL;
374	int error;
375	SVCXPRT *new_xprt;
376
377	/*
378	 * The socket upcall calls xprt_active() which will eventually
379	 * cause the server to call us here. We attempt to accept a
380	 * connection from the socket and turn it into a new
381	 * transport. If the accept fails, we have drained all pending
382	 * connections so we call xprt_inactive().
383	 */
384	sx_xlock(&xprt->xp_lock);
385
386	error = svc_vc_accept(xprt->xp_socket, &so);
387
388	if (error == EWOULDBLOCK) {
389		/*
390		 * We must re-test for new connections after taking
391		 * the lock to protect us in the case where a new
392		 * connection arrives after our call to accept fails
393		 * with EWOULDBLOCK.
394		 */
395		ACCEPT_LOCK();
396		if (TAILQ_EMPTY(&xprt->xp_socket->so_comp))
397			xprt_inactive_self(xprt);
398		ACCEPT_UNLOCK();
399		sx_xunlock(&xprt->xp_lock);
400		return (FALSE);
401	}
402
403	if (error) {
404		SOCKBUF_LOCK(&xprt->xp_socket->so_rcv);
405		if (xprt->xp_upcallset) {
406			xprt->xp_upcallset = 0;
407			soupcall_clear(xprt->xp_socket, SO_RCV);
408		}
409		SOCKBUF_UNLOCK(&xprt->xp_socket->so_rcv);
410		xprt_inactive_self(xprt);
411		sx_xunlock(&xprt->xp_lock);
412		return (FALSE);
413	}
414
415	sx_xunlock(&xprt->xp_lock);
416
417	sa = NULL;
418	error = soaccept(so, &sa);
419
420	if (error) {
421		/*
422		 * XXX not sure if I need to call sofree or soclose here.
423		 */
424		if (sa)
425			free(sa, M_SONAME);
426		return (FALSE);
427	}
428
429	/*
430	 * svc_vc_create_conn will call xprt_register - we don't need
431	 * to do anything with the new connection except derefence it.
432	 */
433	new_xprt = svc_vc_create_conn(xprt->xp_pool, so, sa);
434	if (!new_xprt) {
435		soclose(so);
436	} else {
437		SVC_RELEASE(new_xprt);
438	}
439
440	free(sa, M_SONAME);
441
442	return (FALSE); /* there is never an rpc msg to be processed */
443}
444
445/*ARGSUSED*/
446static enum xprt_stat
447svc_vc_rendezvous_stat(SVCXPRT *xprt)
448{
449
450	return (XPRT_IDLE);
451}
452
453static void
454svc_vc_destroy_common(SVCXPRT *xprt)
455{
456	SOCKBUF_LOCK(&xprt->xp_socket->so_rcv);
457	if (xprt->xp_upcallset) {
458		xprt->xp_upcallset = 0;
459		soupcall_clear(xprt->xp_socket, SO_RCV);
460	}
461	SOCKBUF_UNLOCK(&xprt->xp_socket->so_rcv);
462
463	if (xprt->xp_socket)
464		(void)soclose(xprt->xp_socket);
465
466	if (xprt->xp_netid)
467		(void) mem_free(xprt->xp_netid, strlen(xprt->xp_netid) + 1);
468	svc_xprt_free(xprt);
469}
470
471static void
472svc_vc_rendezvous_destroy(SVCXPRT *xprt)
473{
474
475	svc_vc_destroy_common(xprt);
476}
477
478static void
479svc_vc_destroy(SVCXPRT *xprt)
480{
481	struct cf_conn *cd = (struct cf_conn *)xprt->xp_p1;
482
483	svc_vc_destroy_common(xprt);
484
485	if (cd->mreq)
486		m_freem(cd->mreq);
487	if (cd->mpending)
488		m_freem(cd->mpending);
489	mem_free(cd, sizeof(*cd));
490}
491
492static void
493svc_vc_backchannel_destroy(SVCXPRT *xprt)
494{
495	struct cf_conn *cd = (struct cf_conn *)xprt->xp_p1;
496	struct mbuf *m, *m2;
497
498	svc_xprt_free(xprt);
499	m = cd->mreq;
500	while (m != NULL) {
501		m2 = m;
502		m = m->m_nextpkt;
503		m_freem(m2);
504	}
505	mem_free(cd, sizeof(*cd));
506}
507
508/*ARGSUSED*/
509static bool_t
510svc_vc_control(SVCXPRT *xprt, const u_int rq, void *in)
511{
512	return (FALSE);
513}
514
515static bool_t
516svc_vc_rendezvous_control(SVCXPRT *xprt, const u_int rq, void *in)
517{
518
519	return (FALSE);
520}
521
522static bool_t
523svc_vc_backchannel_control(SVCXPRT *xprt, const u_int rq, void *in)
524{
525
526	return (FALSE);
527}
528
529static enum xprt_stat
530svc_vc_stat(SVCXPRT *xprt)
531{
532	struct cf_conn *cd;
533
534	cd = (struct cf_conn *)(xprt->xp_p1);
535
536	if (cd->strm_stat == XPRT_DIED)
537		return (XPRT_DIED);
538
539	if (cd->mreq != NULL && cd->resid == 0 && cd->eor)
540		return (XPRT_MOREREQS);
541
542	if (soreadable(xprt->xp_socket))
543		return (XPRT_MOREREQS);
544
545	return (XPRT_IDLE);
546}
547
548static bool_t
549svc_vc_ack(SVCXPRT *xprt, uint32_t *ack)
550{
551
552	*ack = atomic_load_acq_32(&xprt->xp_snt_cnt);
553	*ack -= xprt->xp_socket->so_snd.sb_cc;
554	return (TRUE);
555}
556
557static enum xprt_stat
558svc_vc_backchannel_stat(SVCXPRT *xprt)
559{
560	struct cf_conn *cd;
561
562	cd = (struct cf_conn *)(xprt->xp_p1);
563
564	if (cd->mreq != NULL)
565		return (XPRT_MOREREQS);
566
567	return (XPRT_IDLE);
568}
569
570/*
571 * If we have an mbuf chain in cd->mpending, try to parse a record from it,
572 * leaving the result in cd->mreq. If we don't have a complete record, leave
573 * the partial result in cd->mreq and try to read more from the socket.
574 */
575static int
576svc_vc_process_pending(SVCXPRT *xprt)
577{
578	struct cf_conn *cd = (struct cf_conn *) xprt->xp_p1;
579	struct socket *so = xprt->xp_socket;
580	struct mbuf *m;
581
582	/*
583	 * If cd->resid is non-zero, we have part of the
584	 * record already, otherwise we are expecting a record
585	 * marker.
586	 */
587	if (!cd->resid && cd->mpending) {
588		/*
589		 * See if there is enough data buffered to
590		 * make up a record marker. Make sure we can
591		 * handle the case where the record marker is
592		 * split across more than one mbuf.
593		 */
594		size_t n = 0;
595		uint32_t header;
596
597		m = cd->mpending;
598		while (n < sizeof(uint32_t) && m) {
599			n += m->m_len;
600			m = m->m_next;
601		}
602		if (n < sizeof(uint32_t)) {
603			so->so_rcv.sb_lowat = sizeof(uint32_t) - n;
604			return (FALSE);
605		}
606		m_copydata(cd->mpending, 0, sizeof(header),
607		    (char *)&header);
608		header = ntohl(header);
609		cd->eor = (header & 0x80000000) != 0;
610		cd->resid = header & 0x7fffffff;
611		m_adj(cd->mpending, sizeof(uint32_t));
612	}
613
614	/*
615	 * Start pulling off mbufs from cd->mpending
616	 * until we either have a complete record or
617	 * we run out of data. We use m_split to pull
618	 * data - it will pull as much as possible and
619	 * split the last mbuf if necessary.
620	 */
621	while (cd->mpending && cd->resid) {
622		m = cd->mpending;
623		if (cd->mpending->m_next
624		    || cd->mpending->m_len > cd->resid)
625			cd->mpending = m_split(cd->mpending,
626			    cd->resid, M_WAITOK);
627		else
628			cd->mpending = NULL;
629		if (cd->mreq)
630			m_last(cd->mreq)->m_next = m;
631		else
632			cd->mreq = m;
633		while (m) {
634			cd->resid -= m->m_len;
635			m = m->m_next;
636		}
637	}
638
639	/*
640	 * Block receive upcalls if we have more data pending,
641	 * otherwise report our need.
642	 */
643	if (cd->mpending)
644		so->so_rcv.sb_lowat = INT_MAX;
645	else
646		so->so_rcv.sb_lowat =
647		    imax(1, imin(cd->resid, so->so_rcv.sb_hiwat / 2));
648	return (TRUE);
649}
650
651static bool_t
652svc_vc_recv(SVCXPRT *xprt, struct rpc_msg *msg,
653    struct sockaddr **addrp, struct mbuf **mp)
654{
655	struct cf_conn *cd = (struct cf_conn *) xprt->xp_p1;
656	struct uio uio;
657	struct mbuf *m;
658	struct socket* so = xprt->xp_socket;
659	XDR xdrs;
660	int error, rcvflag;
661	uint32_t xid_plus_direction[2];
662
663	/*
664	 * Serialise access to the socket and our own record parsing
665	 * state.
666	 */
667	sx_xlock(&xprt->xp_lock);
668
669	for (;;) {
670		/* If we have no request ready, check pending queue. */
671		while (cd->mpending &&
672		    (cd->mreq == NULL || cd->resid != 0 || !cd->eor)) {
673			if (!svc_vc_process_pending(xprt))
674				break;
675		}
676
677		/* Process and return complete request in cd->mreq. */
678		if (cd->mreq != NULL && cd->resid == 0 && cd->eor) {
679
680			/*
681			 * Now, check for a backchannel reply.
682			 * The XID is in the first uint32_t of the reply
683			 * and the message direction is the second one.
684			 */
685			if ((cd->mreq->m_len >= sizeof(xid_plus_direction) ||
686			    m_length(cd->mreq, NULL) >=
687			    sizeof(xid_plus_direction)) &&
688			    xprt->xp_p2 != NULL) {
689				m_copydata(cd->mreq, 0,
690				    sizeof(xid_plus_direction),
691				    (char *)xid_plus_direction);
692				xid_plus_direction[0] =
693				    ntohl(xid_plus_direction[0]);
694				xid_plus_direction[1] =
695				    ntohl(xid_plus_direction[1]);
696				/* Check message direction. */
697				if (xid_plus_direction[1] == REPLY) {
698					clnt_bck_svccall(xprt->xp_p2,
699					    cd->mreq,
700					    xid_plus_direction[0]);
701					cd->mreq = NULL;
702					continue;
703				}
704			}
705
706			xdrmbuf_create(&xdrs, cd->mreq, XDR_DECODE);
707			cd->mreq = NULL;
708
709			/* Check for next request in a pending queue. */
710			svc_vc_process_pending(xprt);
711			if (cd->mreq == NULL || cd->resid != 0) {
712				SOCKBUF_LOCK(&so->so_rcv);
713				if (!soreadable(so))
714					xprt_inactive_self(xprt);
715				SOCKBUF_UNLOCK(&so->so_rcv);
716			}
717
718			sx_xunlock(&xprt->xp_lock);
719
720			if (! xdr_callmsg(&xdrs, msg)) {
721				XDR_DESTROY(&xdrs);
722				return (FALSE);
723			}
724
725			*addrp = NULL;
726			*mp = xdrmbuf_getall(&xdrs);
727			XDR_DESTROY(&xdrs);
728
729			return (TRUE);
730		}
731
732		/*
733		 * The socket upcall calls xprt_active() which will eventually
734		 * cause the server to call us here. We attempt to
735		 * read as much as possible from the socket and put
736		 * the result in cd->mpending. If the read fails,
737		 * we have drained both cd->mpending and the socket so
738		 * we can call xprt_inactive().
739		 */
740		uio.uio_resid = 1000000000;
741		uio.uio_td = curthread;
742		m = NULL;
743		rcvflag = MSG_DONTWAIT;
744		error = soreceive(so, NULL, &uio, &m, NULL, &rcvflag);
745
746		if (error == EWOULDBLOCK) {
747			/*
748			 * We must re-test for readability after
749			 * taking the lock to protect us in the case
750			 * where a new packet arrives on the socket
751			 * after our call to soreceive fails with
752			 * EWOULDBLOCK.
753			 */
754			SOCKBUF_LOCK(&so->so_rcv);
755			if (!soreadable(so))
756				xprt_inactive_self(xprt);
757			SOCKBUF_UNLOCK(&so->so_rcv);
758			sx_xunlock(&xprt->xp_lock);
759			return (FALSE);
760		}
761
762		if (error) {
763			SOCKBUF_LOCK(&so->so_rcv);
764			if (xprt->xp_upcallset) {
765				xprt->xp_upcallset = 0;
766				soupcall_clear(so, SO_RCV);
767			}
768			SOCKBUF_UNLOCK(&so->so_rcv);
769			xprt_inactive_self(xprt);
770			cd->strm_stat = XPRT_DIED;
771			sx_xunlock(&xprt->xp_lock);
772			return (FALSE);
773		}
774
775		if (!m) {
776			/*
777			 * EOF - the other end has closed the socket.
778			 */
779			xprt_inactive_self(xprt);
780			cd->strm_stat = XPRT_DIED;
781			sx_xunlock(&xprt->xp_lock);
782			return (FALSE);
783		}
784
785		if (cd->mpending)
786			m_last(cd->mpending)->m_next = m;
787		else
788			cd->mpending = m;
789	}
790}
791
792static bool_t
793svc_vc_backchannel_recv(SVCXPRT *xprt, struct rpc_msg *msg,
794    struct sockaddr **addrp, struct mbuf **mp)
795{
796	struct cf_conn *cd = (struct cf_conn *) xprt->xp_p1;
797	struct ct_data *ct;
798	struct mbuf *m;
799	XDR xdrs;
800
801	sx_xlock(&xprt->xp_lock);
802	ct = (struct ct_data *)xprt->xp_p2;
803	if (ct == NULL) {
804		sx_xunlock(&xprt->xp_lock);
805		return (FALSE);
806	}
807	mtx_lock(&ct->ct_lock);
808	m = cd->mreq;
809	if (m == NULL) {
810		xprt_inactive_self(xprt);
811		mtx_unlock(&ct->ct_lock);
812		sx_xunlock(&xprt->xp_lock);
813		return (FALSE);
814	}
815	cd->mreq = m->m_nextpkt;
816	mtx_unlock(&ct->ct_lock);
817	sx_xunlock(&xprt->xp_lock);
818
819	xdrmbuf_create(&xdrs, m, XDR_DECODE);
820	if (! xdr_callmsg(&xdrs, msg)) {
821		XDR_DESTROY(&xdrs);
822		return (FALSE);
823	}
824	*addrp = NULL;
825	*mp = xdrmbuf_getall(&xdrs);
826	XDR_DESTROY(&xdrs);
827	return (TRUE);
828}
829
830static bool_t
831svc_vc_reply(SVCXPRT *xprt, struct rpc_msg *msg,
832    struct sockaddr *addr, struct mbuf *m, uint32_t *seq)
833{
834	XDR xdrs;
835	struct mbuf *mrep;
836	bool_t stat = TRUE;
837	int error, len;
838
839	/*
840	 * Leave space for record mark.
841	 */
842	mrep = m_gethdr(M_WAITOK, MT_DATA);
843	mrep->m_data += sizeof(uint32_t);
844
845	xdrmbuf_create(&xdrs, mrep, XDR_ENCODE);
846
847	if (msg->rm_reply.rp_stat == MSG_ACCEPTED &&
848	    msg->rm_reply.rp_acpt.ar_stat == SUCCESS) {
849		if (!xdr_replymsg(&xdrs, msg))
850			stat = FALSE;
851		else
852			xdrmbuf_append(&xdrs, m);
853	} else {
854		stat = xdr_replymsg(&xdrs, msg);
855	}
856
857	if (stat) {
858		m_fixhdr(mrep);
859
860		/*
861		 * Prepend a record marker containing the reply length.
862		 */
863		M_PREPEND(mrep, sizeof(uint32_t), M_WAITOK);
864		len = mrep->m_pkthdr.len;
865		*mtod(mrep, uint32_t *) =
866			htonl(0x80000000 | (len - sizeof(uint32_t)));
867		atomic_add_acq_32(&xprt->xp_snd_cnt, len);
868		error = sosend(xprt->xp_socket, NULL, NULL, mrep, NULL,
869		    0, curthread);
870		if (!error) {
871			atomic_add_rel_32(&xprt->xp_snt_cnt, len);
872			if (seq)
873				*seq = xprt->xp_snd_cnt;
874			stat = TRUE;
875		} else
876			atomic_subtract_32(&xprt->xp_snd_cnt, len);
877	} else {
878		m_freem(mrep);
879	}
880
881	XDR_DESTROY(&xdrs);
882
883	return (stat);
884}
885
886static bool_t
887svc_vc_backchannel_reply(SVCXPRT *xprt, struct rpc_msg *msg,
888    struct sockaddr *addr, struct mbuf *m, uint32_t *seq)
889{
890	struct ct_data *ct;
891	XDR xdrs;
892	struct mbuf *mrep;
893	bool_t stat = TRUE;
894	int error;
895
896	/*
897	 * Leave space for record mark.
898	 */
899	mrep = m_gethdr(M_WAITOK, MT_DATA);
900	mrep->m_data += sizeof(uint32_t);
901
902	xdrmbuf_create(&xdrs, mrep, XDR_ENCODE);
903
904	if (msg->rm_reply.rp_stat == MSG_ACCEPTED &&
905	    msg->rm_reply.rp_acpt.ar_stat == SUCCESS) {
906		if (!xdr_replymsg(&xdrs, msg))
907			stat = FALSE;
908		else
909			xdrmbuf_append(&xdrs, m);
910	} else {
911		stat = xdr_replymsg(&xdrs, msg);
912	}
913
914	if (stat) {
915		m_fixhdr(mrep);
916
917		/*
918		 * Prepend a record marker containing the reply length.
919		 */
920		M_PREPEND(mrep, sizeof(uint32_t), M_WAITOK);
921		*mtod(mrep, uint32_t *) =
922			htonl(0x80000000 | (mrep->m_pkthdr.len
923				- sizeof(uint32_t)));
924		sx_xlock(&xprt->xp_lock);
925		ct = (struct ct_data *)xprt->xp_p2;
926		if (ct != NULL)
927			error = sosend(ct->ct_socket, NULL, NULL, mrep, NULL,
928			    0, curthread);
929		else
930			error = EPIPE;
931		sx_xunlock(&xprt->xp_lock);
932		if (!error) {
933			stat = TRUE;
934		}
935	} else {
936		m_freem(mrep);
937	}
938
939	XDR_DESTROY(&xdrs);
940
941	return (stat);
942}
943
944static bool_t
945svc_vc_null()
946{
947
948	return (FALSE);
949}
950
951static int
952svc_vc_soupcall(struct socket *so, void *arg, int waitflag)
953{
954	SVCXPRT *xprt = (SVCXPRT *) arg;
955
956	if (soreadable(xprt->xp_socket))
957		xprt_active(xprt);
958	return (SU_OK);
959}
960
961#if 0
962/*
963 * Get the effective UID of the sending process. Used by rpcbind, keyserv
964 * and rpc.yppasswdd on AF_LOCAL.
965 */
966int
967__rpc_get_local_uid(SVCXPRT *transp, uid_t *uid) {
968	int sock, ret;
969	gid_t egid;
970	uid_t euid;
971	struct sockaddr *sa;
972
973	sock = transp->xp_fd;
974	sa = (struct sockaddr *)transp->xp_rtaddr;
975	if (sa->sa_family == AF_LOCAL) {
976		ret = getpeereid(sock, &euid, &egid);
977		if (ret == 0)
978			*uid = euid;
979		return (ret);
980	} else
981		return (-1);
982}
983#endif
984