xref: /freebsd/sys/rpc/clnt_bck.c (revision e9ac41698b2f322d55ccf9da50a3596edb2c1800) 
1 /*	$NetBSD: clnt_vc.c,v 1.4 2000/07/14 08:40:42 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 #include <sys/cdefs.h>
32 /*
33  * clnt_tcp.c, Implements a TCP/IP based, client side RPC.
34  *
35  * Copyright (C) 1984, Sun Microsystems, Inc.
36  *
37  * TCP based RPC supports 'batched calls'.
38  * A sequence of calls may be batched-up in a send buffer.  The rpc call
39  * return immediately to the client even though the call was not necessarily
40  * sent.  The batching occurs if the results' xdr routine is NULL (0) AND
41  * the rpc timeout value is zero (see clnt.h, rpc).
42  *
43  * Clients should NOT casually batch calls that in fact return results; that is,
44  * the server side should be aware that a call is batched and not produce any
45  * return message.  Batched calls that produce many result messages can
46  * deadlock (netlock) the client and the server....
47  *
48  * Now go hang yourself.
49  */
50 
51 /*
52  * This code handles the special case of a NFSv4.n backchannel for
53  * callback RPCs. It is similar to clnt_vc.c, but uses the TCP
54  * connection provided by the client to the server.
55  */
56 
57 #include "opt_kern_tls.h"
58 
59 #include <sys/param.h>
60 #include <sys/systm.h>
61 #include <sys/ktls.h>
62 #include <sys/lock.h>
63 #include <sys/malloc.h>
64 #include <sys/mbuf.h>
65 #include <sys/mutex.h>
66 #include <sys/pcpu.h>
67 #include <sys/proc.h>
68 #include <sys/protosw.h>
69 #include <sys/socket.h>
70 #include <sys/socketvar.h>
71 #include <sys/sx.h>
72 #include <sys/syslog.h>
73 #include <sys/time.h>
74 #include <sys/uio.h>
75 
76 #include <net/vnet.h>
77 
78 #include <netinet/tcp.h>
79 
80 #include <rpc/rpc.h>
81 #include <rpc/rpc_com.h>
82 #include <rpc/krpc.h>
83 #include <rpc/rpcsec_tls.h>
84 
85 struct cmessage {
86         struct cmsghdr cmsg;
87         struct cmsgcred cmcred;
88 };
89 
90 static void clnt_bck_geterr(CLIENT *, struct rpc_err *);
91 static bool_t clnt_bck_freeres(CLIENT *, xdrproc_t, void *);
92 static void clnt_bck_abort(CLIENT *);
93 static bool_t clnt_bck_control(CLIENT *, u_int, void *);
94 static void clnt_bck_close(CLIENT *);
95 static void clnt_bck_destroy(CLIENT *);
96 
97 static const struct clnt_ops clnt_bck_ops = {
98 	.cl_abort =	clnt_bck_abort,
99 	.cl_geterr =	clnt_bck_geterr,
100 	.cl_freeres =	clnt_bck_freeres,
101 	.cl_close =	clnt_bck_close,
102 	.cl_destroy =	clnt_bck_destroy,
103 	.cl_control =	clnt_bck_control
104 };
105 
106 /*
107  * Create a client handle for a connection.
108  * Default options are set, which the user can change using clnt_control()'s.
109  * This code handles the special case of an NFSv4.1 session backchannel
110  * call, which is sent on a TCP connection created against the server
111  * by a client.
112  */
113 void *
114 clnt_bck_create(
115 	struct socket *so,		/* Server transport socket. */
116 	const rpcprog_t prog,		/* program number */
117 	const rpcvers_t vers)		/* version number */
118 {
119 	CLIENT *cl;			/* client handle */
120 	struct ct_data *ct = NULL;	/* client handle */
121 	struct timeval now;
122 	struct rpc_msg call_msg;
123 	static uint32_t disrupt;
124 	XDR xdrs;
125 
126 	if (disrupt == 0)
127 		disrupt = (uint32_t)(long)so;
128 
129 	cl = (CLIENT *)mem_alloc(sizeof (*cl));
130 	ct = (struct ct_data *)mem_alloc(sizeof (*ct));
131 
132 	mtx_init(&ct->ct_lock, "ct->ct_lock", NULL, MTX_DEF);
133 	ct->ct_threads = 0;
134 	ct->ct_closing = FALSE;
135 	ct->ct_closed = FALSE;
136 	ct->ct_upcallrefs = 0;
137 	ct->ct_closeit = FALSE;
138 
139 	/*
140 	 * Set up private data struct
141 	 */
142 	ct->ct_wait.tv_sec = -1;
143 	ct->ct_wait.tv_usec = -1;
144 
145 	/*
146 	 * Initialize call message
147 	 */
148 	getmicrotime(&now);
149 	ct->ct_xid = ((uint32_t)++disrupt) ^ __RPC_GETXID(&now);
150 	call_msg.rm_xid = ct->ct_xid;
151 	call_msg.rm_direction = CALL;
152 	call_msg.rm_call.cb_rpcvers = RPC_MSG_VERSION;
153 	call_msg.rm_call.cb_prog = (uint32_t)prog;
154 	call_msg.rm_call.cb_vers = (uint32_t)vers;
155 
156 	/*
157 	 * pre-serialize the static part of the call msg and stash it away
158 	 */
159 	xdrmem_create(&xdrs, ct->ct_mcallc, MCALL_MSG_SIZE,
160 	    XDR_ENCODE);
161 	if (!xdr_callhdr(&xdrs, &call_msg))
162 		goto err;
163 	ct->ct_mpos = XDR_GETPOS(&xdrs);
164 	XDR_DESTROY(&xdrs);
165 	ct->ct_waitchan = "rpcbck";
166 	ct->ct_waitflag = 0;
167 	cl->cl_refs = 1;
168 	cl->cl_ops = &clnt_bck_ops;
169 	cl->cl_private = ct;
170 	cl->cl_auth = authnone_create();
171 	TAILQ_INIT(&ct->ct_pending);
172 	return (cl);
173 
174 err:
175 	mtx_destroy(&ct->ct_lock);
176 	mem_free(ct, sizeof (struct ct_data));
177 	mem_free(cl, sizeof (CLIENT));
178 	return (NULL);
179 }
180 
181 enum clnt_stat
182 clnt_bck_call(
183 	CLIENT		*cl,		/* client handle */
184 	struct rpc_callextra *ext,	/* call metadata */
185 	rpcproc_t	proc,		/* procedure number */
186 	struct mbuf	*args,		/* pointer to args */
187 	struct mbuf	**resultsp,	/* pointer to results */
188 	struct timeval	utimeout,
189 	SVCXPRT		*xprt)
190 {
191 	struct ct_data *ct = (struct ct_data *) cl->cl_private;
192 	AUTH *auth;
193 	struct rpc_err *errp;
194 	enum clnt_stat stat;
195 	XDR xdrs;
196 	struct rpc_msg reply_msg;
197 	bool_t ok;
198 	int nrefreshes = 2;		/* number of times to refresh cred */
199 	struct timeval timeout;
200 	uint32_t xid;
201 	struct mbuf *mreq = NULL, *results;
202 	struct ct_request *cr;
203 	int error, maxextsiz;
204 #ifdef KERN_TLS
205 	u_int maxlen;
206 #endif
207 
208 	cr = malloc(sizeof(struct ct_request), M_RPC, M_WAITOK);
209 
210 	mtx_lock(&ct->ct_lock);
211 
212 	if (ct->ct_closing || ct->ct_closed) {
213 		mtx_unlock(&ct->ct_lock);
214 		free(cr, M_RPC);
215 		return (RPC_CANTSEND);
216 	}
217 	ct->ct_threads++;
218 
219 	if (ext) {
220 		auth = ext->rc_auth;
221 		errp = &ext->rc_err;
222 	} else {
223 		auth = cl->cl_auth;
224 		errp = &ct->ct_error;
225 	}
226 
227 	cr->cr_mrep = NULL;
228 	cr->cr_error = 0;
229 
230 	if (ct->ct_wait.tv_usec == -1)
231 		timeout = utimeout;	/* use supplied timeout */
232 	else
233 		timeout = ct->ct_wait;	/* use default timeout */
234 
235 call_again:
236 	mtx_assert(&ct->ct_lock, MA_OWNED);
237 
238 	ct->ct_xid++;
239 	xid = ct->ct_xid;
240 
241 	mtx_unlock(&ct->ct_lock);
242 
243 	/*
244 	 * Leave space to pre-pend the record mark.
245 	 */
246 	mreq = m_gethdr(M_WAITOK, MT_DATA);
247 	mreq->m_data += sizeof(uint32_t);
248 	KASSERT(ct->ct_mpos + sizeof(uint32_t) <= MHLEN,
249 	    ("RPC header too big"));
250 	bcopy(ct->ct_mcallc, mreq->m_data, ct->ct_mpos);
251 	mreq->m_len = ct->ct_mpos;
252 
253 	/*
254 	 * The XID is the first thing in the request.
255 	 */
256 	*mtod(mreq, uint32_t *) = htonl(xid);
257 
258 	xdrmbuf_create(&xdrs, mreq, XDR_ENCODE);
259 
260 	errp->re_status = stat = RPC_SUCCESS;
261 
262 	if ((!XDR_PUTINT32(&xdrs, &proc)) ||
263 	    (!AUTH_MARSHALL(auth, xid, &xdrs,
264 	     m_copym(args, 0, M_COPYALL, M_WAITOK)))) {
265 		errp->re_status = stat = RPC_CANTENCODEARGS;
266 		mtx_lock(&ct->ct_lock);
267 		goto out;
268 	}
269 	mreq->m_pkthdr.len = m_length(mreq, NULL);
270 
271 	/*
272 	 * Prepend a record marker containing the packet length.
273 	 */
274 	M_PREPEND(mreq, sizeof(uint32_t), M_WAITOK);
275 	*mtod(mreq, uint32_t *) =
276 	    htonl(0x80000000 | (mreq->m_pkthdr.len - sizeof(uint32_t)));
277 
278 	cr->cr_xid = xid;
279 	mtx_lock(&ct->ct_lock);
280 	/*
281 	 * Check to see if the client end has already started to close down
282 	 * the connection. The svc code will have set ct_error.re_status
283 	 * to RPC_CANTRECV if this is the case.
284 	 * If the client starts to close down the connection after this
285 	 * point, it will be detected later when cr_error is checked,
286 	 * since the request is in the ct_pending queue.
287 	 */
288 	if (ct->ct_error.re_status == RPC_CANTRECV) {
289 		if (errp != &ct->ct_error) {
290 			errp->re_errno = ct->ct_error.re_errno;
291 			errp->re_status = RPC_CANTRECV;
292 		}
293 		stat = RPC_CANTRECV;
294 		goto out;
295 	}
296 	TAILQ_INSERT_TAIL(&ct->ct_pending, cr, cr_link);
297 	mtx_unlock(&ct->ct_lock);
298 
299 	/* For RPC-over-TLS, copy mrep to a chain of ext_pgs. */
300 	if ((xprt->xp_tls & RPCTLS_FLAGS_HANDSHAKE) != 0) {
301 		/*
302 		 * Copy the mbuf chain to a chain of
303 		 * ext_pgs mbuf(s) as required by KERN_TLS.
304 		 */
305 		maxextsiz = TLS_MAX_MSG_SIZE_V10_2;
306 #ifdef KERN_TLS
307 		if (rpctls_getinfo(&maxlen, false, false))
308 			maxextsiz = min(maxextsiz, maxlen);
309 #endif
310 		mreq = _rpc_copym_into_ext_pgs(mreq, maxextsiz);
311 	}
312 	/*
313 	 * sosend consumes mreq.
314 	 */
315 	sx_xlock(&xprt->xp_lock);
316 	error = sosend(xprt->xp_socket, NULL, NULL, mreq, NULL, 0, curthread);
317 if (error != 0) printf("sosend=%d\n", error);
318 	mreq = NULL;
319 	if (error == EMSGSIZE) {
320 printf("emsgsize\n");
321 		SOCK_SENDBUF_LOCK(xprt->xp_socket);
322 		sbwait(xprt->xp_socket, SO_SND);
323 		SOCK_SENDBUF_UNLOCK(xprt->xp_socket);
324 		sx_xunlock(&xprt->xp_lock);
325 		AUTH_VALIDATE(auth, xid, NULL, NULL);
326 		mtx_lock(&ct->ct_lock);
327 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
328 		goto call_again;
329 	}
330 	sx_xunlock(&xprt->xp_lock);
331 
332 	reply_msg.acpted_rply.ar_verf.oa_flavor = AUTH_NULL;
333 	reply_msg.acpted_rply.ar_verf.oa_base = cr->cr_verf;
334 	reply_msg.acpted_rply.ar_verf.oa_length = 0;
335 	reply_msg.acpted_rply.ar_results.where = NULL;
336 	reply_msg.acpted_rply.ar_results.proc = (xdrproc_t)xdr_void;
337 
338 	mtx_lock(&ct->ct_lock);
339 	if (error) {
340 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
341 		errp->re_errno = error;
342 		errp->re_status = stat = RPC_CANTSEND;
343 		goto out;
344 	}
345 
346 	/*
347 	 * Check to see if we got an upcall while waiting for the
348 	 * lock. In both these cases, the request has been removed
349 	 * from ct->ct_pending.
350 	 */
351 	if (cr->cr_error) {
352 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
353 		errp->re_errno = cr->cr_error;
354 		errp->re_status = stat = RPC_CANTRECV;
355 		goto out;
356 	}
357 	if (cr->cr_mrep) {
358 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
359 		goto got_reply;
360 	}
361 
362 	/*
363 	 * Hack to provide rpc-based message passing
364 	 */
365 	if (timeout.tv_sec == 0 && timeout.tv_usec == 0) {
366 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
367 		errp->re_status = stat = RPC_TIMEDOUT;
368 		goto out;
369 	}
370 
371 	error = msleep(cr, &ct->ct_lock, ct->ct_waitflag, ct->ct_waitchan,
372 	    tvtohz(&timeout));
373 
374 	TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
375 
376 	if (error) {
377 		/*
378 		 * The sleep returned an error so our request is still
379 		 * on the list. Turn the error code into an
380 		 * appropriate client status.
381 		 */
382 		errp->re_errno = error;
383 		switch (error) {
384 		case EINTR:
385 			stat = RPC_INTR;
386 			break;
387 		case EWOULDBLOCK:
388 			stat = RPC_TIMEDOUT;
389 			break;
390 		default:
391 			stat = RPC_CANTRECV;
392 		}
393 		errp->re_status = stat;
394 		goto out;
395 	} else {
396 		/*
397 		 * We were woken up by the svc thread.  If the
398 		 * upcall had a receive error, report that,
399 		 * otherwise we have a reply.
400 		 */
401 		if (cr->cr_error) {
402 			errp->re_errno = cr->cr_error;
403 			errp->re_status = stat = RPC_CANTRECV;
404 			goto out;
405 		}
406 	}
407 
408 got_reply:
409 	/*
410 	 * Now decode and validate the response. We need to drop the
411 	 * lock since xdr_replymsg may end up sleeping in malloc.
412 	 */
413 	mtx_unlock(&ct->ct_lock);
414 
415 	if (ext && ext->rc_feedback)
416 		ext->rc_feedback(FEEDBACK_OK, proc, ext->rc_feedback_arg);
417 
418 	xdrmbuf_create(&xdrs, cr->cr_mrep, XDR_DECODE);
419 	ok = xdr_replymsg(&xdrs, &reply_msg);
420 	cr->cr_mrep = NULL;
421 
422 	if (ok) {
423 		if ((reply_msg.rm_reply.rp_stat == MSG_ACCEPTED) &&
424 		    (reply_msg.acpted_rply.ar_stat == SUCCESS))
425 			errp->re_status = stat = RPC_SUCCESS;
426 		else
427 			stat = _seterr_reply(&reply_msg, errp);
428 
429 		if (stat == RPC_SUCCESS) {
430 			results = xdrmbuf_getall(&xdrs);
431 			if (!AUTH_VALIDATE(auth, xid,
432 			    &reply_msg.acpted_rply.ar_verf, &results)) {
433 				errp->re_status = stat = RPC_AUTHERROR;
434 				errp->re_why = AUTH_INVALIDRESP;
435 			} else {
436 				KASSERT(results,
437 				    ("auth validated but no result"));
438 				*resultsp = results;
439 			}
440 		}		/* end successful completion */
441 		/*
442 		 * If unsuccessful AND error is an authentication error
443 		 * then refresh credentials and try again, else break
444 		 */
445 		else if (stat == RPC_AUTHERROR)
446 			/* maybe our credentials need to be refreshed ... */
447 			if (nrefreshes > 0 && AUTH_REFRESH(auth, &reply_msg)) {
448 				nrefreshes--;
449 				XDR_DESTROY(&xdrs);
450 				mtx_lock(&ct->ct_lock);
451 				goto call_again;
452 			}
453 			/* end of unsuccessful completion */
454 		/* end of valid reply message */
455 	} else
456 		errp->re_status = stat = RPC_CANTDECODERES;
457 	XDR_DESTROY(&xdrs);
458 	mtx_lock(&ct->ct_lock);
459 out:
460 	mtx_assert(&ct->ct_lock, MA_OWNED);
461 
462 	KASSERT(stat != RPC_SUCCESS || *resultsp,
463 	    ("RPC_SUCCESS without reply"));
464 
465 	if (mreq != NULL)
466 		m_freem(mreq);
467 	if (cr->cr_mrep != NULL)
468 		m_freem(cr->cr_mrep);
469 
470 	ct->ct_threads--;
471 	if (ct->ct_closing)
472 		wakeup(ct);
473 
474 	mtx_unlock(&ct->ct_lock);
475 
476 	if (auth && stat != RPC_SUCCESS)
477 		AUTH_VALIDATE(auth, xid, NULL, NULL);
478 
479 	free(cr, M_RPC);
480 
481 	return (stat);
482 }
483 
484 static void
485 clnt_bck_geterr(CLIENT *cl, struct rpc_err *errp)
486 {
487 	struct ct_data *ct = (struct ct_data *) cl->cl_private;
488 
489 	*errp = ct->ct_error;
490 }
491 
492 static bool_t
493 clnt_bck_freeres(CLIENT *cl, xdrproc_t xdr_res, void *res_ptr)
494 {
495 	XDR xdrs;
496 	bool_t dummy;
497 
498 	xdrs.x_op = XDR_FREE;
499 	dummy = (*xdr_res)(&xdrs, res_ptr);
500 
501 	return (dummy);
502 }
503 
504 /*ARGSUSED*/
505 static void
506 clnt_bck_abort(CLIENT *cl)
507 {
508 }
509 
510 static bool_t
511 clnt_bck_control(CLIENT *cl, u_int request, void *info)
512 {
513 
514 	return (TRUE);
515 }
516 
517 static void
518 clnt_bck_close(CLIENT *cl)
519 {
520 	struct ct_data *ct = (struct ct_data *) cl->cl_private;
521 
522 	mtx_lock(&ct->ct_lock);
523 
524 	if (ct->ct_closed) {
525 		mtx_unlock(&ct->ct_lock);
526 		return;
527 	}
528 
529 	if (ct->ct_closing) {
530 		while (ct->ct_closing)
531 			msleep(ct, &ct->ct_lock, 0, "rpcclose", 0);
532 		KASSERT(ct->ct_closed, ("client should be closed"));
533 		mtx_unlock(&ct->ct_lock);
534 		return;
535 	}
536 
537 	ct->ct_closing = FALSE;
538 	ct->ct_closed = TRUE;
539 	mtx_unlock(&ct->ct_lock);
540 	wakeup(ct);
541 }
542 
543 static void
544 clnt_bck_destroy(CLIENT *cl)
545 {
546 	struct ct_data *ct = (struct ct_data *) cl->cl_private;
547 
548 	clnt_bck_close(cl);
549 
550 	mtx_destroy(&ct->ct_lock);
551 	mem_free(ct, sizeof(struct ct_data));
552 	if (cl->cl_netid && cl->cl_netid[0])
553 		mem_free(cl->cl_netid, strlen(cl->cl_netid) +1);
554 	if (cl->cl_tp && cl->cl_tp[0])
555 		mem_free(cl->cl_tp, strlen(cl->cl_tp) +1);
556 	mem_free(cl, sizeof(CLIENT));
557 }
558 
559 /*
560  * This call is done by the svc code when a backchannel RPC reply is
561  * received.
562  * For the server end, where callback RPCs to the client are performed,
563  * xp_p2 points to the "CLIENT" and not the associated "struct ct_data"
564  * so that svc_vc_destroy() can CLNT_RELEASE() the reference count on it.
565  */
566 void
567 clnt_bck_svccall(void *arg, struct mbuf *mrep, uint32_t xid)
568 {
569 	CLIENT *cl = (CLIENT *)arg;
570 	struct ct_data *ct;
571 	struct ct_request *cr;
572 	int foundreq;
573 
574 	ct = (struct ct_data *)cl->cl_private;
575 	mtx_lock(&ct->ct_lock);
576 	if (ct->ct_closing || ct->ct_closed) {
577 		mtx_unlock(&ct->ct_lock);
578 		m_freem(mrep);
579 		return;
580 	}
581 
582 	ct->ct_upcallrefs++;
583 	/*
584 	 * See if we can match this reply to a request.
585 	 */
586 	foundreq = 0;
587 	TAILQ_FOREACH(cr, &ct->ct_pending, cr_link) {
588 		if (cr->cr_xid == xid) {
589 			/*
590 			 * This one matches. We leave the reply mbuf list in
591 			 * cr->cr_mrep. Set the XID to zero so that we will
592 			 * ignore any duplicated replies.
593 			 */
594 			cr->cr_xid = 0;
595 			cr->cr_mrep = mrep;
596 			cr->cr_error = 0;
597 			foundreq = 1;
598 			wakeup(cr);
599 			break;
600 		}
601 	}
602 
603 	ct->ct_upcallrefs--;
604 	if (ct->ct_upcallrefs < 0)
605 		panic("rpcvc svccall refcnt");
606 	if (ct->ct_upcallrefs == 0)
607 		wakeup(&ct->ct_upcallrefs);
608 	mtx_unlock(&ct->ct_lock);
609 	if (foundreq == 0)
610 		m_freem(mrep);
611 }
612 
613