1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 1990, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011 Bayard G. Bell. All rights reserved.
24 * Copyright (c) 2013 by Delphix. All rights reserved.
25 * Copyright (c) 2017 Joyent Inc
26 * Copyright 2019 Nexenta by DDN, Inc.
27 */
28
29 /*
30 * Copyright (c) 1983,1984,1985,1986,1987,1988,1989 AT&T.
31 * All rights reserved.
32 * Use is subject to license terms.
33 */
34
35 /*
36 * Copyright (c) 2011 Bayard G. Bell. All rights reserved.
37 * Copyright (c) 2013 by Delphix. All rights reserved.
38 * Copyright 2018 Nexenta Systems, Inc.
39 * Copyright (c) 2017 Joyent Inc
40 */
41
42 #include <sys/param.h>
43 #include <sys/types.h>
44 #include <sys/systm.h>
45 #include <sys/cred.h>
46 #include <sys/proc.h>
47 #include <sys/user.h>
48 #include <sys/buf.h>
49 #include <sys/vfs.h>
50 #include <sys/vnode.h>
51 #include <sys/pathname.h>
52 #include <sys/uio.h>
53 #include <sys/file.h>
54 #include <sys/stat.h>
55 #include <sys/errno.h>
56 #include <sys/socket.h>
57 #include <sys/sysmacros.h>
58 #include <sys/siginfo.h>
59 #include <sys/tiuser.h>
60 #include <sys/statvfs.h>
61 #include <sys/stream.h>
62 #include <sys/strsun.h>
63 #include <sys/strsubr.h>
64 #include <sys/stropts.h>
65 #include <sys/timod.h>
66 #include <sys/t_kuser.h>
67 #include <sys/kmem.h>
68 #include <sys/kstat.h>
69 #include <sys/dirent.h>
70 #include <sys/cmn_err.h>
71 #include <sys/debug.h>
72 #include <sys/unistd.h>
73 #include <sys/vtrace.h>
74 #include <sys/mode.h>
75 #include <sys/acl.h>
76 #include <sys/sdt.h>
77 #include <sys/debug.h>
78
79 #include <rpc/types.h>
80 #include <rpc/auth.h>
81 #include <rpc/auth_unix.h>
82 #include <rpc/auth_des.h>
83 #include <rpc/svc.h>
84 #include <rpc/xdr.h>
85 #include <rpc/rpc_rdma.h>
86
87 #include <nfs/nfs.h>
88 #include <nfs/export.h>
89 #include <nfs/nfssys.h>
90 #include <nfs/nfs_clnt.h>
91 #include <nfs/nfs_acl.h>
92 #include <nfs/nfs_log.h>
93 #include <nfs/lm.h>
94 #include <nfs/nfs_dispatch.h>
95 #include <nfs/nfs4_drc.h>
96
97 #include <sys/modctl.h>
98 #include <sys/cladm.h>
99 #include <sys/clconf.h>
100
101 #include <sys/tsol/label.h>
102
103 #define MAXHOST 32
104 const char *kinet_ntop6(uchar_t *, char *, size_t);
105
106 /*
107 * Module linkage information.
108 */
109
110 static struct modlmisc modlmisc = {
111 &mod_miscops, "NFS server module"
112 };
113
114 static struct modlinkage modlinkage = {
115 MODREV_1, (void *)&modlmisc, NULL
116 };
117
118 zone_key_t nfssrv_zone_key;
119 list_t nfssrv_globals_list;
120 krwlock_t nfssrv_globals_rwl;
121
122 kmem_cache_t *nfs_xuio_cache;
123 int nfs_loaned_buffers = 0;
124
125 int
126 _init(void)
127 {
128 int status;
129
130 nfs_srvinit();
131
132 status = mod_install((struct modlinkage *)&modlinkage);
133 if (status != 0) {
134 /*
135 * Could not load module, cleanup previous
136 * initialization work.
137 */
138 nfs_srvfini();
139
140 return (status);
141 }
142
143 /*
144 * Initialise some placeholders for nfssys() calls. These have
145 * to be declared by the nfs module, since that handles nfssys()
146 * calls - also used by NFS clients - but are provided by this
147 * nfssrv module. These also then serve as confirmation to the
148 * relevant code in nfs that nfssrv has been loaded, as they're
149 * initially NULL.
150 */
151 nfs_srv_quiesce_func = nfs_srv_quiesce_all;
152 nfs_srv_dss_func = rfs4_dss_setpaths;
153
154 /* setup DSS paths here; must be done before initial server startup */
155 rfs4_dss_paths = rfs4_dss_oldpaths = NULL;
156
157 /* initialize the copy reduction caches */
158
159 nfs_xuio_cache = kmem_cache_create("nfs_xuio_cache",
160 sizeof (nfs_xuio_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
161
162 return (status);
163 }
164
165 int
166 _fini()
167 {
168 return (EBUSY);
169 }
170
171 int
172 _info(struct modinfo *modinfop)
173 {
174 return (mod_info(&modlinkage, modinfop));
175 }
176
177 /*
178 * PUBLICFH_CHECK() checks if the dispatch routine supports
179 * RPC_PUBLICFH_OK, if the filesystem is exported public, and if the
180 * incoming request is using the public filehandle. The check duplicates
181 * the exportmatch() call done in checkexport(), and we should consider
182 * modifying those routines to avoid the duplication. For now, we optimize
183 * by calling exportmatch() only after checking that the dispatch routine
184 * supports RPC_PUBLICFH_OK, and if the filesystem is explicitly exported
185 * public (i.e., not the placeholder).
186 */
187 #define PUBLICFH_CHECK(ne, disp, exi, fsid, xfid) \
188 ((disp->dis_flags & RPC_PUBLICFH_OK) && \
189 ((exi->exi_export.ex_flags & EX_PUBLIC) || \
190 (exi == ne->exi_public && exportmatch(ne->exi_root, \
191 fsid, xfid))))
192
193 static void nfs_srv_shutdown_all(int);
194 static void rfs4_server_start(nfs_globals_t *, int);
195 static void nullfree(void);
196 static void rfs_dispatch(struct svc_req *, SVCXPRT *);
197 static void acl_dispatch(struct svc_req *, SVCXPRT *);
198 static void common_dispatch(struct svc_req *, SVCXPRT *,
199 rpcvers_t, rpcvers_t, char *,
200 struct rpc_disptable *);
201 static int checkauth(struct exportinfo *, struct svc_req *, cred_t *, int,
202 bool_t, bool_t *);
203 static char *client_name(struct svc_req *req);
204 static char *client_addr(struct svc_req *req, char *buf);
205 extern int sec_svc_getcred(struct svc_req *, cred_t *cr, char **, int *);
206 extern bool_t sec_svc_inrootlist(int, caddr_t, int, caddr_t *);
207 static void *nfs_server_zone_init(zoneid_t);
208 static void nfs_server_zone_fini(zoneid_t, void *);
209 static void nfs_server_zone_shutdown(zoneid_t, void *);
210
211 #define NFSLOG_COPY_NETBUF(exi, xprt, nb) { \
212 (nb)->maxlen = (xprt)->xp_rtaddr.maxlen; \
213 (nb)->len = (xprt)->xp_rtaddr.len; \
214 (nb)->buf = kmem_alloc((nb)->len, KM_SLEEP); \
215 bcopy((xprt)->xp_rtaddr.buf, (nb)->buf, (nb)->len); \
216 }
217
218 /*
219 * Public Filehandle common nfs routines
220 */
221 static int MCLpath(char **);
222 static void URLparse(char *);
223
224 /*
225 * NFS callout table.
226 * This table is used by svc_getreq() to dispatch a request with
227 * a given prog/vers pair to an appropriate service provider
228 * dispatch routine.
229 *
230 * NOTE: ordering is relied upon below when resetting the version min/max
231 * for NFS_PROGRAM. Careful, if this is ever changed.
232 */
233 static SVC_CALLOUT __nfs_sc_clts[] = {
234 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch },
235 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch }
236 };
237
238 static SVC_CALLOUT_TABLE nfs_sct_clts = {
239 sizeof (__nfs_sc_clts) / sizeof (__nfs_sc_clts[0]), FALSE,
240 __nfs_sc_clts
241 };
242
243 static SVC_CALLOUT __nfs_sc_cots[] = {
244 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch },
245 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch }
246 };
247
248 static SVC_CALLOUT_TABLE nfs_sct_cots = {
249 sizeof (__nfs_sc_cots) / sizeof (__nfs_sc_cots[0]), FALSE, __nfs_sc_cots
250 };
251
252 static SVC_CALLOUT __nfs_sc_rdma[] = {
253 { NFS_PROGRAM, NFS_VERSMIN, NFS_VERSMAX, rfs_dispatch },
254 { NFS_ACL_PROGRAM, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX, acl_dispatch }
255 };
256
257 static SVC_CALLOUT_TABLE nfs_sct_rdma = {
258 sizeof (__nfs_sc_rdma) / sizeof (__nfs_sc_rdma[0]), FALSE, __nfs_sc_rdma
259 };
260
261 /*
262 * DSS: distributed stable storage
263 * lists of all DSS paths: current, and before last warmstart
264 */
265 nvlist_t *rfs4_dss_paths, *rfs4_dss_oldpaths;
266
267 int rfs4_dispatch(struct rpcdisp *, struct svc_req *, SVCXPRT *, char *);
268 bool_t rfs4_minorvers_mismatch(struct svc_req *, SVCXPRT *, void *);
269
270 nfs_globals_t *
271 nfs_srv_getzg(void)
272 {
273 nfs_globals_t *ng;
274
275 ng = zone_getspecific(nfssrv_zone_key, curzone);
276 return (ng);
277 }
278
279 /*
280 * Will be called at the point the server pool is being unregistered
281 * from the pool list. From that point onwards, the pool is waiting
282 * to be drained and as such the server state is stale and pertains
283 * to the old instantiation of the NFS server pool.
284 */
285 void
286 nfs_srv_offline(void)
287 {
288 nfs_globals_t *ng;
289
290 ng = nfs_srv_getzg();
291
292 mutex_enter(&ng->nfs_server_upordown_lock);
293 if (ng->nfs_server_upordown == NFS_SERVER_RUNNING) {
294 ng->nfs_server_upordown = NFS_SERVER_OFFLINE;
295 }
296 mutex_exit(&ng->nfs_server_upordown_lock);
297 }
298
299 /*
300 * Will be called at the point the server pool is being destroyed so
301 * all transports have been closed and no service threads are in
302 * existence.
303 *
304 * If we quiesce the server, we're shutting it down without destroying the
305 * server state. This allows it to warm start subsequently.
306 */
307 void
308 nfs_srv_stop_all(void)
309 {
310 int quiesce = 0;
311 nfs_srv_shutdown_all(quiesce);
312 }
313
314 /*
315 * This alternative shutdown routine can be requested via nfssys()
316 */
317 void
318 nfs_srv_quiesce_all(void)
319 {
320 int quiesce = 1;
321 nfs_srv_shutdown_all(quiesce);
322 }
323
324 static void
325 nfs_srv_shutdown_all(int quiesce)
326 {
327 nfs_globals_t *ng = nfs_srv_getzg();
328
329 mutex_enter(&ng->nfs_server_upordown_lock);
330 if (quiesce) {
331 if (ng->nfs_server_upordown == NFS_SERVER_RUNNING ||
332 ng->nfs_server_upordown == NFS_SERVER_OFFLINE) {
333 ng->nfs_server_upordown = NFS_SERVER_QUIESCED;
334 cv_signal(&ng->nfs_server_upordown_cv);
335
336 /* reset DSS state */
337 rfs4_dss_numnewpaths = 0;
338 rfs4_dss_newpaths = NULL;
339
340 cmn_err(CE_NOTE, "nfs_server: server is now quiesced; "
341 "NFSv4 state has been preserved");
342 }
343 } else {
344 if (ng->nfs_server_upordown == NFS_SERVER_OFFLINE) {
345 ng->nfs_server_upordown = NFS_SERVER_STOPPING;
346 mutex_exit(&ng->nfs_server_upordown_lock);
347 rfs4_state_zone_fini();
348 rfs4_fini_drc();
349 mutex_enter(&ng->nfs_server_upordown_lock);
350 ng->nfs_server_upordown = NFS_SERVER_STOPPED;
351
352 /* reset DSS state */
353 rfs4_dss_numnewpaths = 0;
354 rfs4_dss_newpaths = NULL;
355
356 cv_signal(&ng->nfs_server_upordown_cv);
357 }
358 }
359 mutex_exit(&ng->nfs_server_upordown_lock);
360 }
361
362 static int
363 nfs_srv_set_sc_versions(struct file *fp, SVC_CALLOUT_TABLE **sctpp,
364 rpcvers_t versmin, rpcvers_t versmax)
365 {
366 struct strioctl strioc;
367 struct T_info_ack tinfo;
368 int error, retval;
369
370 /*
371 * Find out what type of transport this is.
372 */
373 strioc.ic_cmd = TI_GETINFO;
374 strioc.ic_timout = -1;
375 strioc.ic_len = sizeof (tinfo);
376 strioc.ic_dp = (char *)&tinfo;
377 tinfo.PRIM_type = T_INFO_REQ;
378
379 error = strioctl(fp->f_vnode, I_STR, (intptr_t)&strioc, 0, K_TO_K,
380 CRED(), &retval);
381 if (error || retval)
382 return (error);
383
384 /*
385 * Based on our query of the transport type...
386 *
387 * Reset the min/max versions based on the caller's request
388 * NOTE: This assumes that NFS_PROGRAM is first in the array!!
389 * And the second entry is the NFS_ACL_PROGRAM.
390 */
391 switch (tinfo.SERV_type) {
392 case T_CLTS:
393 if (versmax == NFS_V4)
394 return (EINVAL);
395 __nfs_sc_clts[0].sc_versmin = versmin;
396 __nfs_sc_clts[0].sc_versmax = versmax;
397 __nfs_sc_clts[1].sc_versmin = versmin;
398 __nfs_sc_clts[1].sc_versmax = versmax;
399 *sctpp = &nfs_sct_clts;
400 break;
401 case T_COTS:
402 case T_COTS_ORD:
403 __nfs_sc_cots[0].sc_versmin = versmin;
404 __nfs_sc_cots[0].sc_versmax = versmax;
405 /* For the NFS_ACL program, check the max version */
406 if (versmax > NFS_ACL_VERSMAX)
407 versmax = NFS_ACL_VERSMAX;
408 __nfs_sc_cots[1].sc_versmin = versmin;
409 __nfs_sc_cots[1].sc_versmax = versmax;
410 *sctpp = &nfs_sct_cots;
411 break;
412 default:
413 error = EINVAL;
414 }
415
416 return (error);
417 }
418
419 /*
420 * NFS Server system call.
421 * Does all of the work of running a NFS server.
422 * uap->fd is the fd of an open transport provider
423 */
424 int
425 nfs_svc(struct nfs_svc_args *arg, model_t model)
426 {
427 nfs_globals_t *ng;
428 file_t *fp;
429 SVCMASTERXPRT *xprt;
430 int error;
431 int readsize;
432 char buf[KNC_STRSIZE];
433 size_t len;
434 STRUCT_HANDLE(nfs_svc_args, uap);
435 struct netbuf addrmask;
436 SVC_CALLOUT_TABLE *sctp = NULL;
437
438 #ifdef lint
439 model = model; /* STRUCT macros don't always refer to it */
440 #endif
441
442 ng = nfs_srv_getzg();
443 STRUCT_SET_HANDLE(uap, model, arg);
444
445 /* Check privileges in nfssys() */
446
447 if ((fp = getf(STRUCT_FGET(uap, fd))) == NULL)
448 return (EBADF);
449
450 /* Setup global file handle in nfs_export */
451 if ((error = nfs_export_get_rootfh(ng)) != 0)
452 return (error);
453
454 /*
455 * Set read buffer size to rsize
456 * and add room for RPC headers.
457 */
458 readsize = nfs3tsize() + (RPC_MAXDATASIZE - NFS_MAXDATA);
459 if (readsize < RPC_MAXDATASIZE)
460 readsize = RPC_MAXDATASIZE;
461
462 error = copyinstr((const char *)STRUCT_FGETP(uap, netid), buf,
463 KNC_STRSIZE, &len);
464 if (error) {
465 releasef(STRUCT_FGET(uap, fd));
466 return (error);
467 }
468
469 addrmask.len = STRUCT_FGET(uap, addrmask.len);
470 addrmask.maxlen = STRUCT_FGET(uap, addrmask.maxlen);
471 addrmask.buf = kmem_alloc(addrmask.maxlen, KM_SLEEP);
472 error = copyin(STRUCT_FGETP(uap, addrmask.buf), addrmask.buf,
473 addrmask.len);
474 if (error) {
475 releasef(STRUCT_FGET(uap, fd));
476 kmem_free(addrmask.buf, addrmask.maxlen);
477 return (error);
478 }
479
480 ng->nfs_versmin = STRUCT_FGET(uap, versmin);
481 ng->nfs_versmax = STRUCT_FGET(uap, versmax);
482
483 /* Double check the vers min/max ranges */
484 if ((ng->nfs_versmin > ng->nfs_versmax) ||
485 (ng->nfs_versmin < NFS_VERSMIN) ||
486 (ng->nfs_versmax > NFS_VERSMAX)) {
487 ng->nfs_versmin = NFS_VERSMIN_DEFAULT;
488 ng->nfs_versmax = NFS_VERSMAX_DEFAULT;
489 }
490
491 if (error = nfs_srv_set_sc_versions(fp, &sctp, ng->nfs_versmin,
492 ng->nfs_versmax)) {
493 releasef(STRUCT_FGET(uap, fd));
494 kmem_free(addrmask.buf, addrmask.maxlen);
495 return (error);
496 }
497
498 /* Initialize nfsv4 server */
499 if (ng->nfs_versmax == (rpcvers_t)NFS_V4)
500 rfs4_server_start(ng, STRUCT_FGET(uap, delegation));
501
502 /* Create a transport handle. */
503 error = svc_tli_kcreate(fp, readsize, buf, &addrmask, &xprt,
504 sctp, NULL, NFS_SVCPOOL_ID, TRUE);
505
506 if (error)
507 kmem_free(addrmask.buf, addrmask.maxlen);
508
509 releasef(STRUCT_FGET(uap, fd));
510
511 /* HA-NFSv4: save the cluster nodeid */
512 if (cluster_bootflags & CLUSTER_BOOTED)
513 lm_global_nlmid = clconf_get_nodeid();
514
515 return (error);
516 }
517
518 static void
519 rfs4_server_start(nfs_globals_t *ng, int nfs4_srv_delegation)
520 {
521 /*
522 * Determine if the server has previously been "started" and
523 * if not, do the per instance initialization
524 */
525 mutex_enter(&ng->nfs_server_upordown_lock);
526
527 if (ng->nfs_server_upordown != NFS_SERVER_RUNNING) {
528 /* Do we need to stop and wait on the previous server? */
529 while (ng->nfs_server_upordown == NFS_SERVER_STOPPING ||
530 ng->nfs_server_upordown == NFS_SERVER_OFFLINE)
531 cv_wait(&ng->nfs_server_upordown_cv,
532 &ng->nfs_server_upordown_lock);
533
534 if (ng->nfs_server_upordown != NFS_SERVER_RUNNING) {
535 (void) svc_pool_control(NFS_SVCPOOL_ID,
536 SVCPSET_UNREGISTER_PROC, (void *)&nfs_srv_offline);
537 (void) svc_pool_control(NFS_SVCPOOL_ID,
538 SVCPSET_SHUTDOWN_PROC, (void *)&nfs_srv_stop_all);
539
540 rfs4_do_server_start(ng->nfs_server_upordown,
541 nfs4_srv_delegation,
542 cluster_bootflags & CLUSTER_BOOTED);
543
544 ng->nfs_server_upordown = NFS_SERVER_RUNNING;
545 }
546 cv_signal(&ng->nfs_server_upordown_cv);
547 }
548 mutex_exit(&ng->nfs_server_upordown_lock);
549 }
550
551 /*
552 * If RDMA device available,
553 * start RDMA listener.
554 */
555 int
556 rdma_start(struct rdma_svc_args *rsa)
557 {
558 nfs_globals_t *ng;
559 int error;
560 rdma_xprt_group_t started_rdma_xprts;
561 rdma_stat stat;
562 int svc_state = 0;
563
564 /* Double check the vers min/max ranges */
565 if ((rsa->nfs_versmin > rsa->nfs_versmax) ||
566 (rsa->nfs_versmin < NFS_VERSMIN) ||
567 (rsa->nfs_versmax > NFS_VERSMAX)) {
568 rsa->nfs_versmin = NFS_VERSMIN_DEFAULT;
569 rsa->nfs_versmax = NFS_VERSMAX_DEFAULT;
570 }
571
572 ng = nfs_srv_getzg();
573 ng->nfs_versmin = rsa->nfs_versmin;
574 ng->nfs_versmax = rsa->nfs_versmax;
575
576 /* Set the versions in the callout table */
577 __nfs_sc_rdma[0].sc_versmin = rsa->nfs_versmin;
578 __nfs_sc_rdma[0].sc_versmax = rsa->nfs_versmax;
579 /* For the NFS_ACL program, check the max version */
580 __nfs_sc_rdma[1].sc_versmin = rsa->nfs_versmin;
581 if (rsa->nfs_versmax > NFS_ACL_VERSMAX)
582 __nfs_sc_rdma[1].sc_versmax = NFS_ACL_VERSMAX;
583 else
584 __nfs_sc_rdma[1].sc_versmax = rsa->nfs_versmax;
585
586 /* Initialize nfsv4 server */
587 if (rsa->nfs_versmax == (rpcvers_t)NFS_V4)
588 rfs4_server_start(ng, rsa->delegation);
589
590 started_rdma_xprts.rtg_count = 0;
591 started_rdma_xprts.rtg_listhead = NULL;
592 started_rdma_xprts.rtg_poolid = rsa->poolid;
593
594 restart:
595 error = svc_rdma_kcreate(rsa->netid, &nfs_sct_rdma, rsa->poolid,
596 &started_rdma_xprts);
597
598 svc_state = !error;
599
600 while (!error) {
601
602 /*
603 * wait till either interrupted by a signal on
604 * nfs service stop/restart or signalled by a
605 * rdma attach/detatch.
606 */
607
608 stat = rdma_kwait();
609
610 /*
611 * stop services if running -- either on a HCA detach event
612 * or if the nfs service is stopped/restarted.
613 */
614
615 if ((stat == RDMA_HCA_DETACH || stat == RDMA_INTR) &&
616 svc_state) {
617 rdma_stop(&started_rdma_xprts);
618 svc_state = 0;
619 }
620
621 /*
622 * nfs service stop/restart, break out of the
623 * wait loop and return;
624 */
625 if (stat == RDMA_INTR)
626 return (0);
627
628 /*
629 * restart stopped services on a HCA attach event
630 * (if not already running)
631 */
632
633 if ((stat == RDMA_HCA_ATTACH) && (svc_state == 0))
634 goto restart;
635
636 /*
637 * loop until a nfs service stop/restart
638 */
639 }
640
641 return (error);
642 }
643
644 /* ARGSUSED */
645 void
646 rpc_null(caddr_t *argp, caddr_t *resp, struct exportinfo *exi,
647 struct svc_req *req, cred_t *cr, bool_t ro)
648 {
649 }
650
651 /* ARGSUSED */
652 void
653 rpc_null_v3(caddr_t *argp, caddr_t *resp, struct exportinfo *exi,
654 struct svc_req *req, cred_t *cr, bool_t ro)
655 {
656 DTRACE_NFSV3_4(op__null__start, struct svc_req *, req,
657 cred_t *, cr, vnode_t *, NULL, struct exportinfo *, exi);
658 DTRACE_NFSV3_4(op__null__done, struct svc_req *, req,
659 cred_t *, cr, vnode_t *, NULL, struct exportinfo *, exi);
660 }
661
662 /* ARGSUSED */
663 static void
664 rfs_error(caddr_t *argp, caddr_t *resp, struct exportinfo *exi,
665 struct svc_req *req, cred_t *cr, bool_t ro)
666 {
667 /* return (EOPNOTSUPP); */
668 }
669
670 static void
671 nullfree(void)
672 {
673 }
674
675 static char *rfscallnames_v2[] = {
676 "RFS2_NULL",
677 "RFS2_GETATTR",
678 "RFS2_SETATTR",
679 "RFS2_ROOT",
680 "RFS2_LOOKUP",
681 "RFS2_READLINK",
682 "RFS2_READ",
683 "RFS2_WRITECACHE",
684 "RFS2_WRITE",
685 "RFS2_CREATE",
686 "RFS2_REMOVE",
687 "RFS2_RENAME",
688 "RFS2_LINK",
689 "RFS2_SYMLINK",
690 "RFS2_MKDIR",
691 "RFS2_RMDIR",
692 "RFS2_READDIR",
693 "RFS2_STATFS"
694 };
695
696 static struct rpcdisp rfsdisptab_v2[] = {
697 /*
698 * NFS VERSION 2
699 */
700
701 /* RFS_NULL = 0 */
702 {rpc_null,
703 xdr_void, NULL_xdrproc_t, 0,
704 xdr_void, NULL_xdrproc_t, 0,
705 nullfree, RPC_IDEMPOTENT,
706 0},
707
708 /* RFS_GETATTR = 1 */
709 {rfs_getattr,
710 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
711 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
712 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
713 rfs_getattr_getfh},
714
715 /* RFS_SETATTR = 2 */
716 {rfs_setattr,
717 xdr_saargs, NULL_xdrproc_t, sizeof (struct nfssaargs),
718 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
719 nullfree, RPC_MAPRESP,
720 rfs_setattr_getfh},
721
722 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
723 {rfs_error,
724 xdr_void, NULL_xdrproc_t, 0,
725 xdr_void, NULL_xdrproc_t, 0,
726 nullfree, RPC_IDEMPOTENT,
727 0},
728
729 /* RFS_LOOKUP = 4 */
730 {rfs_lookup,
731 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
732 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
733 nullfree, RPC_IDEMPOTENT|RPC_MAPRESP|RPC_PUBLICFH_OK,
734 rfs_lookup_getfh},
735
736 /* RFS_READLINK = 5 */
737 {rfs_readlink,
738 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
739 xdr_rdlnres, NULL_xdrproc_t, sizeof (struct nfsrdlnres),
740 rfs_rlfree, RPC_IDEMPOTENT,
741 rfs_readlink_getfh},
742
743 /* RFS_READ = 6 */
744 {rfs_read,
745 xdr_readargs, NULL_xdrproc_t, sizeof (struct nfsreadargs),
746 xdr_rdresult, NULL_xdrproc_t, sizeof (struct nfsrdresult),
747 rfs_rdfree, RPC_IDEMPOTENT,
748 rfs_read_getfh},
749
750 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
751 {rfs_error,
752 xdr_void, NULL_xdrproc_t, 0,
753 xdr_void, NULL_xdrproc_t, 0,
754 nullfree, RPC_IDEMPOTENT,
755 0},
756
757 /* RFS_WRITE = 8 */
758 {rfs_write,
759 xdr_writeargs, NULL_xdrproc_t, sizeof (struct nfswriteargs),
760 xdr_attrstat, xdr_fastattrstat, sizeof (struct nfsattrstat),
761 nullfree, RPC_MAPRESP,
762 rfs_write_getfh},
763
764 /* RFS_CREATE = 9 */
765 {rfs_create,
766 xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs),
767 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
768 nullfree, RPC_MAPRESP,
769 rfs_create_getfh},
770
771 /* RFS_REMOVE = 10 */
772 {rfs_remove,
773 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
774 #ifdef _LITTLE_ENDIAN
775 xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
776 #else
777 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
778 #endif
779 nullfree, RPC_MAPRESP,
780 rfs_remove_getfh},
781
782 /* RFS_RENAME = 11 */
783 {rfs_rename,
784 xdr_rnmargs, NULL_xdrproc_t, sizeof (struct nfsrnmargs),
785 #ifdef _LITTLE_ENDIAN
786 xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
787 #else
788 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
789 #endif
790 nullfree, RPC_MAPRESP,
791 rfs_rename_getfh},
792
793 /* RFS_LINK = 12 */
794 {rfs_link,
795 xdr_linkargs, NULL_xdrproc_t, sizeof (struct nfslinkargs),
796 #ifdef _LITTLE_ENDIAN
797 xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
798 #else
799 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
800 #endif
801 nullfree, RPC_MAPRESP,
802 rfs_link_getfh},
803
804 /* RFS_SYMLINK = 13 */
805 {rfs_symlink,
806 xdr_slargs, NULL_xdrproc_t, sizeof (struct nfsslargs),
807 #ifdef _LITTLE_ENDIAN
808 xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
809 #else
810 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
811 #endif
812 nullfree, RPC_MAPRESP,
813 rfs_symlink_getfh},
814
815 /* RFS_MKDIR = 14 */
816 {rfs_mkdir,
817 xdr_creatargs, NULL_xdrproc_t, sizeof (struct nfscreatargs),
818 xdr_diropres, xdr_fastdiropres, sizeof (struct nfsdiropres),
819 nullfree, RPC_MAPRESP,
820 rfs_mkdir_getfh},
821
822 /* RFS_RMDIR = 15 */
823 {rfs_rmdir,
824 xdr_diropargs, NULL_xdrproc_t, sizeof (struct nfsdiropargs),
825 #ifdef _LITTLE_ENDIAN
826 xdr_enum, xdr_fastenum, sizeof (enum nfsstat),
827 #else
828 xdr_enum, NULL_xdrproc_t, sizeof (enum nfsstat),
829 #endif
830 nullfree, RPC_MAPRESP,
831 rfs_rmdir_getfh},
832
833 /* RFS_READDIR = 16 */
834 {rfs_readdir,
835 xdr_rddirargs, NULL_xdrproc_t, sizeof (struct nfsrddirargs),
836 xdr_putrddirres, NULL_xdrproc_t, sizeof (struct nfsrddirres),
837 rfs_rddirfree, RPC_IDEMPOTENT,
838 rfs_readdir_getfh},
839
840 /* RFS_STATFS = 17 */
841 {rfs_statfs,
842 xdr_fhandle, xdr_fastfhandle, sizeof (fhandle_t),
843 xdr_statfs, xdr_faststatfs, sizeof (struct nfsstatfs),
844 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
845 rfs_statfs_getfh},
846 };
847
848 static char *rfscallnames_v3[] = {
849 "RFS3_NULL",
850 "RFS3_GETATTR",
851 "RFS3_SETATTR",
852 "RFS3_LOOKUP",
853 "RFS3_ACCESS",
854 "RFS3_READLINK",
855 "RFS3_READ",
856 "RFS3_WRITE",
857 "RFS3_CREATE",
858 "RFS3_MKDIR",
859 "RFS3_SYMLINK",
860 "RFS3_MKNOD",
861 "RFS3_REMOVE",
862 "RFS3_RMDIR",
863 "RFS3_RENAME",
864 "RFS3_LINK",
865 "RFS3_READDIR",
866 "RFS3_READDIRPLUS",
867 "RFS3_FSSTAT",
868 "RFS3_FSINFO",
869 "RFS3_PATHCONF",
870 "RFS3_COMMIT"
871 };
872
873 static struct rpcdisp rfsdisptab_v3[] = {
874 /*
875 * NFS VERSION 3
876 */
877
878 /* RFS_NULL = 0 */
879 {rpc_null_v3,
880 xdr_void, NULL_xdrproc_t, 0,
881 xdr_void, NULL_xdrproc_t, 0,
882 nullfree, RPC_IDEMPOTENT,
883 0},
884
885 /* RFS3_GETATTR = 1 */
886 {rfs3_getattr,
887 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (GETATTR3args),
888 xdr_GETATTR3res, NULL_xdrproc_t, sizeof (GETATTR3res),
889 nullfree, (RPC_IDEMPOTENT | RPC_ALLOWANON),
890 rfs3_getattr_getfh},
891
892 /* RFS3_SETATTR = 2 */
893 {rfs3_setattr,
894 xdr_SETATTR3args, NULL_xdrproc_t, sizeof (SETATTR3args),
895 xdr_SETATTR3res, NULL_xdrproc_t, sizeof (SETATTR3res),
896 nullfree, 0,
897 rfs3_setattr_getfh},
898
899 /* RFS3_LOOKUP = 3 */
900 {rfs3_lookup,
901 xdr_diropargs3, NULL_xdrproc_t, sizeof (LOOKUP3args),
902 xdr_LOOKUP3res, NULL_xdrproc_t, sizeof (LOOKUP3res),
903 nullfree, (RPC_IDEMPOTENT | RPC_PUBLICFH_OK),
904 rfs3_lookup_getfh},
905
906 /* RFS3_ACCESS = 4 */
907 {rfs3_access,
908 xdr_ACCESS3args, NULL_xdrproc_t, sizeof (ACCESS3args),
909 xdr_ACCESS3res, NULL_xdrproc_t, sizeof (ACCESS3res),
910 nullfree, RPC_IDEMPOTENT,
911 rfs3_access_getfh},
912
913 /* RFS3_READLINK = 5 */
914 {rfs3_readlink,
915 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (READLINK3args),
916 xdr_READLINK3res, NULL_xdrproc_t, sizeof (READLINK3res),
917 rfs3_readlink_free, RPC_IDEMPOTENT,
918 rfs3_readlink_getfh},
919
920 /* RFS3_READ = 6 */
921 {rfs3_read,
922 xdr_READ3args, NULL_xdrproc_t, sizeof (READ3args),
923 xdr_READ3res, NULL_xdrproc_t, sizeof (READ3res),
924 rfs3_read_free, RPC_IDEMPOTENT,
925 rfs3_read_getfh},
926
927 /* RFS3_WRITE = 7 */
928 {rfs3_write,
929 xdr_WRITE3args, NULL_xdrproc_t, sizeof (WRITE3args),
930 xdr_WRITE3res, NULL_xdrproc_t, sizeof (WRITE3res),
931 nullfree, 0,
932 rfs3_write_getfh},
933
934 /* RFS3_CREATE = 8 */
935 {rfs3_create,
936 xdr_CREATE3args, NULL_xdrproc_t, sizeof (CREATE3args),
937 xdr_CREATE3res, NULL_xdrproc_t, sizeof (CREATE3res),
938 nullfree, 0,
939 rfs3_create_getfh},
940
941 /* RFS3_MKDIR = 9 */
942 {rfs3_mkdir,
943 xdr_MKDIR3args, NULL_xdrproc_t, sizeof (MKDIR3args),
944 xdr_MKDIR3res, NULL_xdrproc_t, sizeof (MKDIR3res),
945 nullfree, 0,
946 rfs3_mkdir_getfh},
947
948 /* RFS3_SYMLINK = 10 */
949 {rfs3_symlink,
950 xdr_SYMLINK3args, NULL_xdrproc_t, sizeof (SYMLINK3args),
951 xdr_SYMLINK3res, NULL_xdrproc_t, sizeof (SYMLINK3res),
952 nullfree, 0,
953 rfs3_symlink_getfh},
954
955 /* RFS3_MKNOD = 11 */
956 {rfs3_mknod,
957 xdr_MKNOD3args, NULL_xdrproc_t, sizeof (MKNOD3args),
958 xdr_MKNOD3res, NULL_xdrproc_t, sizeof (MKNOD3res),
959 nullfree, 0,
960 rfs3_mknod_getfh},
961
962 /* RFS3_REMOVE = 12 */
963 {rfs3_remove,
964 xdr_diropargs3, NULL_xdrproc_t, sizeof (REMOVE3args),
965 xdr_REMOVE3res, NULL_xdrproc_t, sizeof (REMOVE3res),
966 nullfree, 0,
967 rfs3_remove_getfh},
968
969 /* RFS3_RMDIR = 13 */
970 {rfs3_rmdir,
971 xdr_diropargs3, NULL_xdrproc_t, sizeof (RMDIR3args),
972 xdr_RMDIR3res, NULL_xdrproc_t, sizeof (RMDIR3res),
973 nullfree, 0,
974 rfs3_rmdir_getfh},
975
976 /* RFS3_RENAME = 14 */
977 {rfs3_rename,
978 xdr_RENAME3args, NULL_xdrproc_t, sizeof (RENAME3args),
979 xdr_RENAME3res, NULL_xdrproc_t, sizeof (RENAME3res),
980 nullfree, 0,
981 rfs3_rename_getfh},
982
983 /* RFS3_LINK = 15 */
984 {rfs3_link,
985 xdr_LINK3args, NULL_xdrproc_t, sizeof (LINK3args),
986 xdr_LINK3res, NULL_xdrproc_t, sizeof (LINK3res),
987 nullfree, 0,
988 rfs3_link_getfh},
989
990 /* RFS3_READDIR = 16 */
991 {rfs3_readdir,
992 xdr_READDIR3args, NULL_xdrproc_t, sizeof (READDIR3args),
993 xdr_READDIR3res, NULL_xdrproc_t, sizeof (READDIR3res),
994 rfs3_readdir_free, RPC_IDEMPOTENT,
995 rfs3_readdir_getfh},
996
997 /* RFS3_READDIRPLUS = 17 */
998 {rfs3_readdirplus,
999 xdr_READDIRPLUS3args, NULL_xdrproc_t, sizeof (READDIRPLUS3args),
1000 xdr_READDIRPLUS3res, NULL_xdrproc_t, sizeof (READDIRPLUS3res),
1001 rfs3_readdirplus_free, RPC_AVOIDWORK,
1002 rfs3_readdirplus_getfh},
1003
1004 /* RFS3_FSSTAT = 18 */
1005 {rfs3_fsstat,
1006 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSSTAT3args),
1007 xdr_FSSTAT3res, NULL_xdrproc_t, sizeof (FSSTAT3res),
1008 nullfree, RPC_IDEMPOTENT,
1009 rfs3_fsstat_getfh},
1010
1011 /* RFS3_FSINFO = 19 */
1012 {rfs3_fsinfo,
1013 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (FSINFO3args),
1014 xdr_FSINFO3res, NULL_xdrproc_t, sizeof (FSINFO3res),
1015 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON,
1016 rfs3_fsinfo_getfh},
1017
1018 /* RFS3_PATHCONF = 20 */
1019 {rfs3_pathconf,
1020 xdr_nfs_fh3_server, NULL_xdrproc_t, sizeof (PATHCONF3args),
1021 xdr_PATHCONF3res, NULL_xdrproc_t, sizeof (PATHCONF3res),
1022 nullfree, RPC_IDEMPOTENT,
1023 rfs3_pathconf_getfh},
1024
1025 /* RFS3_COMMIT = 21 */
1026 {rfs3_commit,
1027 xdr_COMMIT3args, NULL_xdrproc_t, sizeof (COMMIT3args),
1028 xdr_COMMIT3res, NULL_xdrproc_t, sizeof (COMMIT3res),
1029 nullfree, RPC_IDEMPOTENT,
1030 rfs3_commit_getfh},
1031 };
1032
1033 static char *rfscallnames_v4[] = {
1034 "RFS4_NULL",
1035 "RFS4_COMPOUND",
1036 "RFS4_NULL",
1037 "RFS4_NULL",
1038 "RFS4_NULL",
1039 "RFS4_NULL",
1040 "RFS4_NULL",
1041 "RFS4_NULL",
1042 "RFS4_CREATE"
1043 };
1044
1045 static struct rpcdisp rfsdisptab_v4[] = {
1046 /*
1047 * NFS VERSION 4
1048 */
1049
1050 /* RFS_NULL = 0 */
1051 {rpc_null,
1052 xdr_void, NULL_xdrproc_t, 0,
1053 xdr_void, NULL_xdrproc_t, 0,
1054 nullfree, RPC_IDEMPOTENT, 0},
1055
1056 /* RFS4_compound = 1 */
1057 {rfs4_compound,
1058 xdr_COMPOUND4args_srv, NULL_xdrproc_t, sizeof (COMPOUND4args),
1059 xdr_COMPOUND4res_srv, NULL_xdrproc_t, sizeof (COMPOUND4res),
1060 rfs4_compound_free, 0, 0},
1061 };
1062
1063 union rfs_args {
1064 /*
1065 * NFS VERSION 2
1066 */
1067
1068 /* RFS_NULL = 0 */
1069
1070 /* RFS_GETATTR = 1 */
1071 fhandle_t nfs2_getattr_args;
1072
1073 /* RFS_SETATTR = 2 */
1074 struct nfssaargs nfs2_setattr_args;
1075
1076 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
1077
1078 /* RFS_LOOKUP = 4 */
1079 struct nfsdiropargs nfs2_lookup_args;
1080
1081 /* RFS_READLINK = 5 */
1082 fhandle_t nfs2_readlink_args;
1083
1084 /* RFS_READ = 6 */
1085 struct nfsreadargs nfs2_read_args;
1086
1087 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
1088
1089 /* RFS_WRITE = 8 */
1090 struct nfswriteargs nfs2_write_args;
1091
1092 /* RFS_CREATE = 9 */
1093 struct nfscreatargs nfs2_create_args;
1094
1095 /* RFS_REMOVE = 10 */
1096 struct nfsdiropargs nfs2_remove_args;
1097
1098 /* RFS_RENAME = 11 */
1099 struct nfsrnmargs nfs2_rename_args;
1100
1101 /* RFS_LINK = 12 */
1102 struct nfslinkargs nfs2_link_args;
1103
1104 /* RFS_SYMLINK = 13 */
1105 struct nfsslargs nfs2_symlink_args;
1106
1107 /* RFS_MKDIR = 14 */
1108 struct nfscreatargs nfs2_mkdir_args;
1109
1110 /* RFS_RMDIR = 15 */
1111 struct nfsdiropargs nfs2_rmdir_args;
1112
1113 /* RFS_READDIR = 16 */
1114 struct nfsrddirargs nfs2_readdir_args;
1115
1116 /* RFS_STATFS = 17 */
1117 fhandle_t nfs2_statfs_args;
1118
1119 /*
1120 * NFS VERSION 3
1121 */
1122
1123 /* RFS_NULL = 0 */
1124
1125 /* RFS3_GETATTR = 1 */
1126 GETATTR3args nfs3_getattr_args;
1127
1128 /* RFS3_SETATTR = 2 */
1129 SETATTR3args nfs3_setattr_args;
1130
1131 /* RFS3_LOOKUP = 3 */
1132 LOOKUP3args nfs3_lookup_args;
1133
1134 /* RFS3_ACCESS = 4 */
1135 ACCESS3args nfs3_access_args;
1136
1137 /* RFS3_READLINK = 5 */
1138 READLINK3args nfs3_readlink_args;
1139
1140 /* RFS3_READ = 6 */
1141 READ3args nfs3_read_args;
1142
1143 /* RFS3_WRITE = 7 */
1144 WRITE3args nfs3_write_args;
1145
1146 /* RFS3_CREATE = 8 */
1147 CREATE3args nfs3_create_args;
1148
1149 /* RFS3_MKDIR = 9 */
1150 MKDIR3args nfs3_mkdir_args;
1151
1152 /* RFS3_SYMLINK = 10 */
1153 SYMLINK3args nfs3_symlink_args;
1154
1155 /* RFS3_MKNOD = 11 */
1156 MKNOD3args nfs3_mknod_args;
1157
1158 /* RFS3_REMOVE = 12 */
1159 REMOVE3args nfs3_remove_args;
1160
1161 /* RFS3_RMDIR = 13 */
1162 RMDIR3args nfs3_rmdir_args;
1163
1164 /* RFS3_RENAME = 14 */
1165 RENAME3args nfs3_rename_args;
1166
1167 /* RFS3_LINK = 15 */
1168 LINK3args nfs3_link_args;
1169
1170 /* RFS3_READDIR = 16 */
1171 READDIR3args nfs3_readdir_args;
1172
1173 /* RFS3_READDIRPLUS = 17 */
1174 READDIRPLUS3args nfs3_readdirplus_args;
1175
1176 /* RFS3_FSSTAT = 18 */
1177 FSSTAT3args nfs3_fsstat_args;
1178
1179 /* RFS3_FSINFO = 19 */
1180 FSINFO3args nfs3_fsinfo_args;
1181
1182 /* RFS3_PATHCONF = 20 */
1183 PATHCONF3args nfs3_pathconf_args;
1184
1185 /* RFS3_COMMIT = 21 */
1186 COMMIT3args nfs3_commit_args;
1187
1188 /*
1189 * NFS VERSION 4
1190 */
1191
1192 /* RFS_NULL = 0 */
1193
1194 /* COMPUND = 1 */
1195 COMPOUND4args nfs4_compound_args;
1196 };
1197
1198 union rfs_res {
1199 /*
1200 * NFS VERSION 2
1201 */
1202
1203 /* RFS_NULL = 0 */
1204
1205 /* RFS_GETATTR = 1 */
1206 struct nfsattrstat nfs2_getattr_res;
1207
1208 /* RFS_SETATTR = 2 */
1209 struct nfsattrstat nfs2_setattr_res;
1210
1211 /* RFS_ROOT = 3 *** NO LONGER SUPPORTED *** */
1212
1213 /* RFS_LOOKUP = 4 */
1214 struct nfsdiropres nfs2_lookup_res;
1215
1216 /* RFS_READLINK = 5 */
1217 struct nfsrdlnres nfs2_readlink_res;
1218
1219 /* RFS_READ = 6 */
1220 struct nfsrdresult nfs2_read_res;
1221
1222 /* RFS_WRITECACHE = 7 *** NO LONGER SUPPORTED *** */
1223
1224 /* RFS_WRITE = 8 */
1225 struct nfsattrstat nfs2_write_res;
1226
1227 /* RFS_CREATE = 9 */
1228 struct nfsdiropres nfs2_create_res;
1229
1230 /* RFS_REMOVE = 10 */
1231 enum nfsstat nfs2_remove_res;
1232
1233 /* RFS_RENAME = 11 */
1234 enum nfsstat nfs2_rename_res;
1235
1236 /* RFS_LINK = 12 */
1237 enum nfsstat nfs2_link_res;
1238
1239 /* RFS_SYMLINK = 13 */
1240 enum nfsstat nfs2_symlink_res;
1241
1242 /* RFS_MKDIR = 14 */
1243 struct nfsdiropres nfs2_mkdir_res;
1244
1245 /* RFS_RMDIR = 15 */
1246 enum nfsstat nfs2_rmdir_res;
1247
1248 /* RFS_READDIR = 16 */
1249 struct nfsrddirres nfs2_readdir_res;
1250
1251 /* RFS_STATFS = 17 */
1252 struct nfsstatfs nfs2_statfs_res;
1253
1254 /*
1255 * NFS VERSION 3
1256 */
1257
1258 /* RFS_NULL = 0 */
1259
1260 /* RFS3_GETATTR = 1 */
1261 GETATTR3res nfs3_getattr_res;
1262
1263 /* RFS3_SETATTR = 2 */
1264 SETATTR3res nfs3_setattr_res;
1265
1266 /* RFS3_LOOKUP = 3 */
1267 LOOKUP3res nfs3_lookup_res;
1268
1269 /* RFS3_ACCESS = 4 */
1270 ACCESS3res nfs3_access_res;
1271
1272 /* RFS3_READLINK = 5 */
1273 READLINK3res nfs3_readlink_res;
1274
1275 /* RFS3_READ = 6 */
1276 READ3res nfs3_read_res;
1277
1278 /* RFS3_WRITE = 7 */
1279 WRITE3res nfs3_write_res;
1280
1281 /* RFS3_CREATE = 8 */
1282 CREATE3res nfs3_create_res;
1283
1284 /* RFS3_MKDIR = 9 */
1285 MKDIR3res nfs3_mkdir_res;
1286
1287 /* RFS3_SYMLINK = 10 */
1288 SYMLINK3res nfs3_symlink_res;
1289
1290 /* RFS3_MKNOD = 11 */
1291 MKNOD3res nfs3_mknod_res;
1292
1293 /* RFS3_REMOVE = 12 */
1294 REMOVE3res nfs3_remove_res;
1295
1296 /* RFS3_RMDIR = 13 */
1297 RMDIR3res nfs3_rmdir_res;
1298
1299 /* RFS3_RENAME = 14 */
1300 RENAME3res nfs3_rename_res;
1301
1302 /* RFS3_LINK = 15 */
1303 LINK3res nfs3_link_res;
1304
1305 /* RFS3_READDIR = 16 */
1306 READDIR3res nfs3_readdir_res;
1307
1308 /* RFS3_READDIRPLUS = 17 */
1309 READDIRPLUS3res nfs3_readdirplus_res;
1310
1311 /* RFS3_FSSTAT = 18 */
1312 FSSTAT3res nfs3_fsstat_res;
1313
1314 /* RFS3_FSINFO = 19 */
1315 FSINFO3res nfs3_fsinfo_res;
1316
1317 /* RFS3_PATHCONF = 20 */
1318 PATHCONF3res nfs3_pathconf_res;
1319
1320 /* RFS3_COMMIT = 21 */
1321 COMMIT3res nfs3_commit_res;
1322
1323 /*
1324 * NFS VERSION 4
1325 */
1326
1327 /* RFS_NULL = 0 */
1328
1329 /* RFS4_COMPOUND = 1 */
1330 COMPOUND4res nfs4_compound_res;
1331
1332 };
1333
1334 static struct rpc_disptable rfs_disptable[] = {
1335 {sizeof (rfsdisptab_v2) / sizeof (rfsdisptab_v2[0]),
1336 rfscallnames_v2,
1337 &rfsproccnt_v2_ptr, rfsdisptab_v2},
1338 {sizeof (rfsdisptab_v3) / sizeof (rfsdisptab_v3[0]),
1339 rfscallnames_v3,
1340 &rfsproccnt_v3_ptr, rfsdisptab_v3},
1341 {sizeof (rfsdisptab_v4) / sizeof (rfsdisptab_v4[0]),
1342 rfscallnames_v4,
1343 &rfsproccnt_v4_ptr, rfsdisptab_v4},
1344 };
1345
1346 /*
1347 * If nfs_portmon is set, then clients are required to use privileged
1348 * ports (ports < IPPORT_RESERVED) in order to get NFS services.
1349 *
1350 * N.B.: this attempt to carry forward the already ill-conceived notion
1351 * of privileged ports for TCP/UDP is really quite ineffectual. Not only
1352 * is it transport-dependent, it's laughably easy to spoof. If you're
1353 * really interested in security, you must start with secure RPC instead.
1354 */
1355 static int nfs_portmon = 0;
1356
1357 #ifdef DEBUG
1358 static int cred_hits = 0;
1359 static int cred_misses = 0;
1360 #endif
1361
1362 #ifdef DEBUG
1363 /*
1364 * Debug code to allow disabling of rfs_dispatch() use of
1365 * fastxdrargs() and fastxdrres() calls for testing purposes.
1366 */
1367 static int rfs_no_fast_xdrargs = 0;
1368 static int rfs_no_fast_xdrres = 0;
1369 #endif
1370
1371 union acl_args {
1372 /*
1373 * ACL VERSION 2
1374 */
1375
1376 /* ACL2_NULL = 0 */
1377
1378 /* ACL2_GETACL = 1 */
1379 GETACL2args acl2_getacl_args;
1380
1381 /* ACL2_SETACL = 2 */
1382 SETACL2args acl2_setacl_args;
1383
1384 /* ACL2_GETATTR = 3 */
1385 GETATTR2args acl2_getattr_args;
1386
1387 /* ACL2_ACCESS = 4 */
1388 ACCESS2args acl2_access_args;
1389
1390 /* ACL2_GETXATTRDIR = 5 */
1391 GETXATTRDIR2args acl2_getxattrdir_args;
1392
1393 /*
1394 * ACL VERSION 3
1395 */
1396
1397 /* ACL3_NULL = 0 */
1398
1399 /* ACL3_GETACL = 1 */
1400 GETACL3args acl3_getacl_args;
1401
1402 /* ACL3_SETACL = 2 */
1403 SETACL3args acl3_setacl;
1404
1405 /* ACL3_GETXATTRDIR = 3 */
1406 GETXATTRDIR3args acl3_getxattrdir_args;
1407
1408 };
1409
1410 union acl_res {
1411 /*
1412 * ACL VERSION 2
1413 */
1414
1415 /* ACL2_NULL = 0 */
1416
1417 /* ACL2_GETACL = 1 */
1418 GETACL2res acl2_getacl_res;
1419
1420 /* ACL2_SETACL = 2 */
1421 SETACL2res acl2_setacl_res;
1422
1423 /* ACL2_GETATTR = 3 */
1424 GETATTR2res acl2_getattr_res;
1425
1426 /* ACL2_ACCESS = 4 */
1427 ACCESS2res acl2_access_res;
1428
1429 /* ACL2_GETXATTRDIR = 5 */
1430 GETXATTRDIR2args acl2_getxattrdir_res;
1431
1432 /*
1433 * ACL VERSION 3
1434 */
1435
1436 /* ACL3_NULL = 0 */
1437
1438 /* ACL3_GETACL = 1 */
1439 GETACL3res acl3_getacl_res;
1440
1441 /* ACL3_SETACL = 2 */
1442 SETACL3res acl3_setacl_res;
1443
1444 /* ACL3_GETXATTRDIR = 3 */
1445 GETXATTRDIR3res acl3_getxattrdir_res;
1446
1447 };
1448
1449 static bool_t
1450 auth_tooweak(struct svc_req *req, char *res)
1451 {
1452
1453 if (req->rq_vers == NFS_VERSION && req->rq_proc == RFS_LOOKUP) {
1454 struct nfsdiropres *dr = (struct nfsdiropres *)res;
1455 if ((enum wnfsstat)dr->dr_status == WNFSERR_CLNT_FLAVOR)
1456 return (TRUE);
1457 } else if (req->rq_vers == NFS_V3 && req->rq_proc == NFSPROC3_LOOKUP) {
1458 LOOKUP3res *resp = (LOOKUP3res *)res;
1459 if ((enum wnfsstat)resp->status == WNFSERR_CLNT_FLAVOR)
1460 return (TRUE);
1461 }
1462 return (FALSE);
1463 }
1464
1465
1466 static void
1467 common_dispatch(struct svc_req *req, SVCXPRT *xprt, rpcvers_t min_vers,
1468 rpcvers_t max_vers, char *pgmname, struct rpc_disptable *disptable)
1469 {
1470 int which;
1471 rpcvers_t vers;
1472 char *args;
1473 union {
1474 union rfs_args ra;
1475 union acl_args aa;
1476 } args_buf;
1477 char *res;
1478 union {
1479 union rfs_res rr;
1480 union acl_res ar;
1481 } res_buf;
1482 struct rpcdisp *disp = NULL;
1483 int dis_flags = 0;
1484 cred_t *cr;
1485 int error = 0;
1486 int anon_ok;
1487 struct exportinfo *exi = NULL;
1488 unsigned int nfslog_rec_id;
1489 int dupstat;
1490 struct dupreq *dr;
1491 int authres;
1492 bool_t publicfh_ok = FALSE;
1493 enum_t auth_flavor;
1494 bool_t dupcached = FALSE;
1495 struct netbuf nb;
1496 bool_t logging_enabled = FALSE;
1497 struct exportinfo *nfslog_exi = NULL;
1498 char **procnames;
1499 char cbuf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */
1500 bool_t ro = FALSE;
1501 nfs_export_t *ne = nfs_get_export();
1502
1503 vers = req->rq_vers;
1504
1505 if (vers < min_vers || vers > max_vers) {
1506 svcerr_progvers(req->rq_xprt, min_vers, max_vers);
1507 error++;
1508 cmn_err(CE_NOTE, "%s: bad version number %u", pgmname, vers);
1509 goto done;
1510 }
1511 vers -= min_vers;
1512
1513 which = req->rq_proc;
1514 if (which < 0 || which >= disptable[(int)vers].dis_nprocs) {
1515 svcerr_noproc(req->rq_xprt);
1516 error++;
1517 goto done;
1518 }
1519
1520 (*(disptable[(int)vers].dis_proccntp))[which].value.ui64++;
1521
1522 disp = &disptable[(int)vers].dis_table[which];
1523 procnames = disptable[(int)vers].dis_procnames;
1524
1525 auth_flavor = req->rq_cred.oa_flavor;
1526
1527 /*
1528 * Deserialize into the args struct.
1529 */
1530 args = (char *)&args_buf;
1531
1532 #ifdef DEBUG
1533 if (rfs_no_fast_xdrargs || (auth_flavor == RPCSEC_GSS) ||
1534 disp->dis_fastxdrargs == NULL_xdrproc_t ||
1535 !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args))
1536 #else
1537 if ((auth_flavor == RPCSEC_GSS) ||
1538 disp->dis_fastxdrargs == NULL_xdrproc_t ||
1539 !SVC_GETARGS(xprt, disp->dis_fastxdrargs, (char *)&args))
1540 #endif
1541 {
1542 bzero(args, disp->dis_argsz);
1543 if (!SVC_GETARGS(xprt, disp->dis_xdrargs, args)) {
1544 error++;
1545 /*
1546 * Check if we are outside our capabilities.
1547 */
1548 if (rfs4_minorvers_mismatch(req, xprt, (void *)args))
1549 goto done;
1550
1551 svcerr_decode(xprt);
1552 cmn_err(CE_NOTE,
1553 "Failed to decode arguments for %s version %u "
1554 "procedure %s client %s%s",
1555 pgmname, vers + min_vers, procnames[which],
1556 client_name(req), client_addr(req, cbuf));
1557 goto done;
1558 }
1559 }
1560
1561 /*
1562 * If Version 4 use that specific dispatch function.
1563 */
1564 if (req->rq_vers == 4) {
1565 error += rfs4_dispatch(disp, req, xprt, args);
1566 goto done;
1567 }
1568
1569 dis_flags = disp->dis_flags;
1570
1571 /*
1572 * Find export information and check authentication,
1573 * setting the credential if everything is ok.
1574 */
1575 if (disp->dis_getfh != NULL) {
1576 void *fh;
1577 fsid_t *fsid;
1578 fid_t *fid, *xfid;
1579 fhandle_t *fh2;
1580 nfs_fh3 *fh3;
1581
1582 fh = (*disp->dis_getfh)(args);
1583 switch (req->rq_vers) {
1584 case NFS_VERSION:
1585 fh2 = (fhandle_t *)fh;
1586 fsid = &fh2->fh_fsid;
1587 fid = (fid_t *)&fh2->fh_len;
1588 xfid = (fid_t *)&fh2->fh_xlen;
1589 break;
1590 case NFS_V3:
1591 fh3 = (nfs_fh3 *)fh;
1592 fsid = &fh3->fh3_fsid;
1593 fid = FH3TOFIDP(fh3);
1594 xfid = FH3TOXFIDP(fh3);
1595 break;
1596 }
1597
1598 /*
1599 * Fix for bug 1038302 - corbin
1600 * There is a problem here if anonymous access is
1601 * disallowed. If the current request is part of the
1602 * client's mount process for the requested filesystem,
1603 * then it will carry root (uid 0) credentials on it, and
1604 * will be denied by checkauth if that client does not
1605 * have explicit root=0 permission. This will cause the
1606 * client's mount operation to fail. As a work-around,
1607 * we check here to see if the request is a getattr or
1608 * statfs operation on the exported vnode itself, and
1609 * pass a flag to checkauth with the result of this test.
1610 *
1611 * The filehandle refers to the mountpoint itself if
1612 * the fh_data and fh_xdata portions of the filehandle
1613 * are equal.
1614 *
1615 * Added anon_ok argument to checkauth().
1616 */
1617
1618 if ((dis_flags & RPC_ALLOWANON) && EQFID(fid, xfid))
1619 anon_ok = 1;
1620 else
1621 anon_ok = 0;
1622
1623 cr = xprt->xp_cred;
1624 ASSERT(cr != NULL);
1625 #ifdef DEBUG
1626 {
1627 if (crgetref(cr) != 1) {
1628 crfree(cr);
1629 cr = crget();
1630 xprt->xp_cred = cr;
1631 cred_misses++;
1632 } else
1633 cred_hits++;
1634 }
1635 #else
1636 if (crgetref(cr) != 1) {
1637 crfree(cr);
1638 cr = crget();
1639 xprt->xp_cred = cr;
1640 }
1641 #endif
1642
1643 exi = checkexport(fsid, xfid);
1644
1645 if (exi != NULL) {
1646 publicfh_ok = PUBLICFH_CHECK(ne, disp, exi, fsid, xfid);
1647
1648 /*
1649 * Don't allow non-V4 clients access
1650 * to pseudo exports
1651 */
1652 if (PSEUDO(exi)) {
1653 svcerr_weakauth(xprt);
1654 error++;
1655 goto done;
1656 }
1657
1658 authres = checkauth(exi, req, cr, anon_ok, publicfh_ok,
1659 &ro);
1660 /*
1661 * authres > 0: authentication OK - proceed
1662 * authres == 0: authentication weak - return error
1663 * authres < 0: authentication timeout - drop
1664 */
1665 if (authres <= 0) {
1666 if (authres == 0) {
1667 svcerr_weakauth(xprt);
1668 error++;
1669 }
1670 goto done;
1671 }
1672 }
1673 } else
1674 cr = NULL;
1675
1676 if ((dis_flags & RPC_MAPRESP) && (auth_flavor != RPCSEC_GSS)) {
1677 res = (char *)SVC_GETRES(xprt, disp->dis_ressz);
1678 if (res == NULL)
1679 res = (char *)&res_buf;
1680 } else
1681 res = (char *)&res_buf;
1682
1683 if (!(dis_flags & RPC_IDEMPOTENT)) {
1684 dupstat = SVC_DUP_EXT(xprt, req, res, disp->dis_ressz, &dr,
1685 &dupcached);
1686
1687 switch (dupstat) {
1688 case DUP_ERROR:
1689 svcerr_systemerr(xprt);
1690 error++;
1691 goto done;
1692 /* NOTREACHED */
1693 case DUP_INPROGRESS:
1694 if (res != (char *)&res_buf)
1695 SVC_FREERES(xprt);
1696 error++;
1697 goto done;
1698 /* NOTREACHED */
1699 case DUP_NEW:
1700 case DUP_DROP:
1701 curthread->t_flag |= T_DONTPEND;
1702
1703 (*disp->dis_proc)(args, res, exi, req, cr, ro);
1704
1705 curthread->t_flag &= ~T_DONTPEND;
1706 if (curthread->t_flag & T_WOULDBLOCK) {
1707 curthread->t_flag &= ~T_WOULDBLOCK;
1708 SVC_DUPDONE_EXT(xprt, dr, res, NULL,
1709 disp->dis_ressz, DUP_DROP);
1710 if (res != (char *)&res_buf)
1711 SVC_FREERES(xprt);
1712 error++;
1713 goto done;
1714 }
1715 if (dis_flags & RPC_AVOIDWORK) {
1716 SVC_DUPDONE_EXT(xprt, dr, res, NULL,
1717 disp->dis_ressz, DUP_DROP);
1718 } else {
1719 SVC_DUPDONE_EXT(xprt, dr, res,
1720 disp->dis_resfree == nullfree ? NULL :
1721 disp->dis_resfree,
1722 disp->dis_ressz, DUP_DONE);
1723 dupcached = TRUE;
1724 }
1725 break;
1726 case DUP_DONE:
1727 break;
1728 }
1729
1730 } else {
1731 curthread->t_flag |= T_DONTPEND;
1732
1733 (*disp->dis_proc)(args, res, exi, req, cr, ro);
1734
1735 curthread->t_flag &= ~T_DONTPEND;
1736 if (curthread->t_flag & T_WOULDBLOCK) {
1737 curthread->t_flag &= ~T_WOULDBLOCK;
1738 if (res != (char *)&res_buf)
1739 SVC_FREERES(xprt);
1740 error++;
1741 goto done;
1742 }
1743 }
1744
1745 if (auth_tooweak(req, res)) {
1746 svcerr_weakauth(xprt);
1747 error++;
1748 goto done;
1749 }
1750
1751 /*
1752 * Check to see if logging has been enabled on the server.
1753 * If so, then obtain the export info struct to be used for
1754 * the later writing of the log record. This is done for
1755 * the case that a lookup is done across a non-logged public
1756 * file system.
1757 */
1758 if (nfslog_buffer_list != NULL) {
1759 nfslog_exi = nfslog_get_exi(ne, exi, req, res, &nfslog_rec_id);
1760 /*
1761 * Is logging enabled?
1762 */
1763 logging_enabled = (nfslog_exi != NULL);
1764
1765 /*
1766 * Copy the netbuf for logging purposes, before it is
1767 * freed by svc_sendreply().
1768 */
1769 if (logging_enabled) {
1770 NFSLOG_COPY_NETBUF(nfslog_exi, xprt, &nb);
1771 /*
1772 * If RPC_MAPRESP flag set (i.e. in V2 ops) the
1773 * res gets copied directly into the mbuf and
1774 * may be freed soon after the sendreply. So we
1775 * must copy it here to a safe place...
1776 */
1777 if (res != (char *)&res_buf) {
1778 bcopy(res, (char *)&res_buf, disp->dis_ressz);
1779 }
1780 }
1781 }
1782
1783 /*
1784 * Serialize and send results struct
1785 */
1786 #ifdef DEBUG
1787 if (rfs_no_fast_xdrres == 0 && res != (char *)&res_buf)
1788 #else
1789 if (res != (char *)&res_buf)
1790 #endif
1791 {
1792 if (!svc_sendreply(xprt, disp->dis_fastxdrres, res)) {
1793 cmn_err(CE_NOTE, "%s: bad sendreply", pgmname);
1794 svcerr_systemerr(xprt);
1795 error++;
1796 }
1797 } else {
1798 if (!svc_sendreply(xprt, disp->dis_xdrres, res)) {
1799 cmn_err(CE_NOTE, "%s: bad sendreply", pgmname);
1800 svcerr_systemerr(xprt);
1801 error++;
1802 }
1803 }
1804
1805 /*
1806 * Log if needed
1807 */
1808 if (logging_enabled) {
1809 nfslog_write_record(nfslog_exi, req, args, (char *)&res_buf,
1810 cr, &nb, nfslog_rec_id, NFSLOG_ONE_BUFFER);
1811 exi_rele(nfslog_exi);
1812 kmem_free((&nb)->buf, (&nb)->len);
1813 }
1814
1815 /*
1816 * Free results struct. With the addition of NFS V4 we can
1817 * have non-idempotent procedures with functions.
1818 */
1819 if (disp->dis_resfree != nullfree && dupcached == FALSE) {
1820 (*disp->dis_resfree)(res);
1821 }
1822
1823 done:
1824 /*
1825 * Free arguments struct
1826 */
1827 if (disp) {
1828 if (!SVC_FREEARGS(xprt, disp->dis_xdrargs, args)) {
1829 cmn_err(CE_NOTE, "%s: bad freeargs", pgmname);
1830 error++;
1831 }
1832 } else {
1833 if (!SVC_FREEARGS(xprt, (xdrproc_t)0, (caddr_t)0)) {
1834 cmn_err(CE_NOTE, "%s: bad freeargs", pgmname);
1835 error++;
1836 }
1837 }
1838
1839 if (exi != NULL)
1840 exi_rele(exi);
1841
1842 global_svstat_ptr[req->rq_vers][NFS_BADCALLS].value.ui64 += error;
1843
1844 global_svstat_ptr[req->rq_vers][NFS_CALLS].value.ui64++;
1845 }
1846
1847 static void
1848 rfs_dispatch(struct svc_req *req, SVCXPRT *xprt)
1849 {
1850 common_dispatch(req, xprt, NFS_VERSMIN, NFS_VERSMAX,
1851 "NFS", rfs_disptable);
1852 }
1853
1854 static char *aclcallnames_v2[] = {
1855 "ACL2_NULL",
1856 "ACL2_GETACL",
1857 "ACL2_SETACL",
1858 "ACL2_GETATTR",
1859 "ACL2_ACCESS",
1860 "ACL2_GETXATTRDIR"
1861 };
1862
1863 static struct rpcdisp acldisptab_v2[] = {
1864 /*
1865 * ACL VERSION 2
1866 */
1867
1868 /* ACL2_NULL = 0 */
1869 {rpc_null,
1870 xdr_void, NULL_xdrproc_t, 0,
1871 xdr_void, NULL_xdrproc_t, 0,
1872 nullfree, RPC_IDEMPOTENT,
1873 0},
1874
1875 /* ACL2_GETACL = 1 */
1876 {acl2_getacl,
1877 xdr_GETACL2args, xdr_fastGETACL2args, sizeof (GETACL2args),
1878 xdr_GETACL2res, NULL_xdrproc_t, sizeof (GETACL2res),
1879 acl2_getacl_free, RPC_IDEMPOTENT,
1880 acl2_getacl_getfh},
1881
1882 /* ACL2_SETACL = 2 */
1883 {acl2_setacl,
1884 xdr_SETACL2args, NULL_xdrproc_t, sizeof (SETACL2args),
1885 #ifdef _LITTLE_ENDIAN
1886 xdr_SETACL2res, xdr_fastSETACL2res, sizeof (SETACL2res),
1887 #else
1888 xdr_SETACL2res, NULL_xdrproc_t, sizeof (SETACL2res),
1889 #endif
1890 nullfree, RPC_MAPRESP,
1891 acl2_setacl_getfh},
1892
1893 /* ACL2_GETATTR = 3 */
1894 {acl2_getattr,
1895 xdr_GETATTR2args, xdr_fastGETATTR2args, sizeof (GETATTR2args),
1896 #ifdef _LITTLE_ENDIAN
1897 xdr_GETATTR2res, xdr_fastGETATTR2res, sizeof (GETATTR2res),
1898 #else
1899 xdr_GETATTR2res, NULL_xdrproc_t, sizeof (GETATTR2res),
1900 #endif
1901 nullfree, RPC_IDEMPOTENT|RPC_ALLOWANON|RPC_MAPRESP,
1902 acl2_getattr_getfh},
1903
1904 /* ACL2_ACCESS = 4 */
1905 {acl2_access,
1906 xdr_ACCESS2args, xdr_fastACCESS2args, sizeof (ACCESS2args),
1907 #ifdef _LITTLE_ENDIAN
1908 xdr_ACCESS2res, xdr_fastACCESS2res, sizeof (ACCESS2res),
1909 #else
1910 xdr_ACCESS2res, NULL_xdrproc_t, sizeof (ACCESS2res),
1911 #endif
1912 nullfree, RPC_IDEMPOTENT|RPC_MAPRESP,
1913 acl2_access_getfh},
1914
1915 /* ACL2_GETXATTRDIR = 5 */
1916 {acl2_getxattrdir,
1917 xdr_GETXATTRDIR2args, NULL_xdrproc_t, sizeof (GETXATTRDIR2args),
1918 xdr_GETXATTRDIR2res, NULL_xdrproc_t, sizeof (GETXATTRDIR2res),
1919 nullfree, RPC_IDEMPOTENT,
1920 acl2_getxattrdir_getfh},
1921 };
1922
1923 static char *aclcallnames_v3[] = {
1924 "ACL3_NULL",
1925 "ACL3_GETACL",
1926 "ACL3_SETACL",
1927 "ACL3_GETXATTRDIR"
1928 };
1929
1930 static struct rpcdisp acldisptab_v3[] = {
1931 /*
1932 * ACL VERSION 3
1933 */
1934
1935 /* ACL3_NULL = 0 */
1936 {rpc_null,
1937 xdr_void, NULL_xdrproc_t, 0,
1938 xdr_void, NULL_xdrproc_t, 0,
1939 nullfree, RPC_IDEMPOTENT,
1940 0},
1941
1942 /* ACL3_GETACL = 1 */
1943 {acl3_getacl,
1944 xdr_GETACL3args, NULL_xdrproc_t, sizeof (GETACL3args),
1945 xdr_GETACL3res, NULL_xdrproc_t, sizeof (GETACL3res),
1946 acl3_getacl_free, RPC_IDEMPOTENT,
1947 acl3_getacl_getfh},
1948
1949 /* ACL3_SETACL = 2 */
1950 {acl3_setacl,
1951 xdr_SETACL3args, NULL_xdrproc_t, sizeof (SETACL3args),
1952 xdr_SETACL3res, NULL_xdrproc_t, sizeof (SETACL3res),
1953 nullfree, 0,
1954 acl3_setacl_getfh},
1955
1956 /* ACL3_GETXATTRDIR = 3 */
1957 {acl3_getxattrdir,
1958 xdr_GETXATTRDIR3args, NULL_xdrproc_t, sizeof (GETXATTRDIR3args),
1959 xdr_GETXATTRDIR3res, NULL_xdrproc_t, sizeof (GETXATTRDIR3res),
1960 nullfree, RPC_IDEMPOTENT,
1961 acl3_getxattrdir_getfh},
1962 };
1963
1964 static struct rpc_disptable acl_disptable[] = {
1965 {sizeof (acldisptab_v2) / sizeof (acldisptab_v2[0]),
1966 aclcallnames_v2,
1967 &aclproccnt_v2_ptr, acldisptab_v2},
1968 {sizeof (acldisptab_v3) / sizeof (acldisptab_v3[0]),
1969 aclcallnames_v3,
1970 &aclproccnt_v3_ptr, acldisptab_v3},
1971 };
1972
1973 static void
1974 acl_dispatch(struct svc_req *req, SVCXPRT *xprt)
1975 {
1976 common_dispatch(req, xprt, NFS_ACL_VERSMIN, NFS_ACL_VERSMAX,
1977 "ACL", acl_disptable);
1978 }
1979
1980 int
1981 checkwin(int flavor, int window, struct svc_req *req)
1982 {
1983 struct authdes_cred *adc;
1984
1985 switch (flavor) {
1986 case AUTH_DES:
1987 adc = (struct authdes_cred *)req->rq_clntcred;
1988 CTASSERT(sizeof (struct authdes_cred) <= RQCRED_SIZE);
1989 if (adc->adc_fullname.window > window)
1990 return (0);
1991 break;
1992
1993 default:
1994 break;
1995 }
1996 return (1);
1997 }
1998
1999
2000 /*
2001 * checkauth() will check the access permission against the export
2002 * information. Then map root uid/gid to appropriate uid/gid.
2003 *
2004 * This routine is used by NFS V3 and V2 code.
2005 */
2006 static int
2007 checkauth(struct exportinfo *exi, struct svc_req *req, cred_t *cr, int anon_ok,
2008 bool_t publicfh_ok, bool_t *ro)
2009 {
2010 int i, nfsflavor, rpcflavor, stat, access;
2011 struct secinfo *secp;
2012 caddr_t principal;
2013 char buf[INET6_ADDRSTRLEN]; /* to hold both IPv4 and IPv6 addr */
2014 int anon_res = 0;
2015
2016 uid_t uid;
2017 gid_t gid;
2018 uint_t ngids;
2019 gid_t *gids;
2020
2021 /*
2022 * Check for privileged port number
2023 * N.B.: this assumes that we know the format of a netbuf.
2024 */
2025 if (nfs_portmon) {
2026 struct sockaddr *ca;
2027 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2028
2029 if (ca == NULL)
2030 return (0);
2031
2032 if ((ca->sa_family == AF_INET &&
2033 ntohs(((struct sockaddr_in *)ca)->sin_port) >=
2034 IPPORT_RESERVED) ||
2035 (ca->sa_family == AF_INET6 &&
2036 ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >=
2037 IPPORT_RESERVED)) {
2038 cmn_err(CE_NOTE,
2039 "nfs_server: client %s%ssent NFS request from "
2040 "unprivileged port",
2041 client_name(req), client_addr(req, buf));
2042 return (0);
2043 }
2044 }
2045
2046 /*
2047 * return 1 on success or 0 on failure
2048 */
2049 stat = sec_svc_getcred(req, cr, &principal, &nfsflavor);
2050
2051 /*
2052 * A failed AUTH_UNIX sec_svc_getcred() implies we couldn't set
2053 * the credentials; below we map that to anonymous.
2054 */
2055 if (!stat && nfsflavor != AUTH_UNIX) {
2056 cmn_err(CE_NOTE,
2057 "nfs_server: couldn't get unix cred for %s",
2058 client_name(req));
2059 return (0);
2060 }
2061
2062 /*
2063 * Short circuit checkauth() on operations that support the
2064 * public filehandle, and if the request for that operation
2065 * is using the public filehandle. Note that we must call
2066 * sec_svc_getcred() first so that xp_cookie is set to the
2067 * right value. Normally xp_cookie is just the RPC flavor
2068 * of the the request, but in the case of RPCSEC_GSS it
2069 * could be a pseudo flavor.
2070 */
2071 if (publicfh_ok)
2072 return (1);
2073
2074 rpcflavor = req->rq_cred.oa_flavor;
2075 /*
2076 * Check if the auth flavor is valid for this export
2077 */
2078 access = nfsauth_access(exi, req, cr, &uid, &gid, &ngids, &gids);
2079 if (access & NFSAUTH_DROP)
2080 return (-1); /* drop the request */
2081
2082 if (access & NFSAUTH_RO)
2083 *ro = TRUE;
2084
2085 if (access & NFSAUTH_DENIED) {
2086 /*
2087 * If anon_ok == 1 and we got NFSAUTH_DENIED, it was
2088 * probably due to the flavor not matching during
2089 * the mount attempt. So map the flavor to AUTH_NONE
2090 * so that the credentials get mapped to the anonymous
2091 * user.
2092 */
2093 if (anon_ok == 1)
2094 rpcflavor = AUTH_NONE;
2095 else
2096 return (0); /* deny access */
2097
2098 } else if (access & NFSAUTH_MAPNONE) {
2099 /*
2100 * Access was granted even though the flavor mismatched
2101 * because AUTH_NONE was one of the exported flavors.
2102 */
2103 rpcflavor = AUTH_NONE;
2104
2105 } else if (access & NFSAUTH_WRONGSEC) {
2106 /*
2107 * NFSAUTH_WRONGSEC is used for NFSv4. If we get here,
2108 * it means a client ignored the list of allowed flavors
2109 * returned via the MOUNT protocol. So we just disallow it!
2110 */
2111 return (0);
2112 }
2113
2114 if (rpcflavor != AUTH_SYS)
2115 kmem_free(gids, ngids * sizeof (gid_t));
2116
2117 switch (rpcflavor) {
2118 case AUTH_NONE:
2119 anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2120 exi->exi_export.ex_anon);
2121 (void) crsetgroups(cr, 0, NULL);
2122 break;
2123
2124 case AUTH_UNIX:
2125 if (!stat || crgetuid(cr) == 0 && !(access & NFSAUTH_UIDMAP)) {
2126 anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2127 exi->exi_export.ex_anon);
2128 (void) crsetgroups(cr, 0, NULL);
2129 } else if (crgetuid(cr) == 0 && access & NFSAUTH_ROOT) {
2130 /*
2131 * It is root, so apply rootid to get real UID
2132 * Find the secinfo structure. We should be able
2133 * to find it by the time we reach here.
2134 * nfsauth_access() has done the checking.
2135 */
2136 secp = NULL;
2137 for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2138 struct secinfo *sptr;
2139 sptr = &exi->exi_export.ex_secinfo[i];
2140 if (sptr->s_secinfo.sc_nfsnum == nfsflavor) {
2141 secp = sptr;
2142 break;
2143 }
2144 }
2145 if (secp != NULL) {
2146 (void) crsetugid(cr, secp->s_rootid,
2147 secp->s_rootid);
2148 (void) crsetgroups(cr, 0, NULL);
2149 }
2150 } else if (crgetuid(cr) != uid || crgetgid(cr) != gid) {
2151 if (crsetugid(cr, uid, gid) != 0)
2152 anon_res = crsetugid(cr,
2153 exi->exi_export.ex_anon,
2154 exi->exi_export.ex_anon);
2155 (void) crsetgroups(cr, 0, NULL);
2156 } else if (access & NFSAUTH_GROUPS) {
2157 (void) crsetgroups(cr, ngids, gids);
2158 }
2159
2160 kmem_free(gids, ngids * sizeof (gid_t));
2161
2162 break;
2163
2164 case AUTH_DES:
2165 case RPCSEC_GSS:
2166 /*
2167 * Find the secinfo structure. We should be able
2168 * to find it by the time we reach here.
2169 * nfsauth_access() has done the checking.
2170 */
2171 secp = NULL;
2172 for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2173 if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum ==
2174 nfsflavor) {
2175 secp = &exi->exi_export.ex_secinfo[i];
2176 break;
2177 }
2178 }
2179
2180 if (!secp) {
2181 cmn_err(CE_NOTE, "nfs_server: client %s%shad "
2182 "no secinfo data for flavor %d",
2183 client_name(req), client_addr(req, buf),
2184 nfsflavor);
2185 return (0);
2186 }
2187
2188 if (!checkwin(rpcflavor, secp->s_window, req)) {
2189 cmn_err(CE_NOTE,
2190 "nfs_server: client %s%sused invalid "
2191 "auth window value",
2192 client_name(req), client_addr(req, buf));
2193 return (0);
2194 }
2195
2196 /*
2197 * Map root principals listed in the share's root= list to root,
2198 * and map any others principals that were mapped to root by RPC
2199 * to anon.
2200 */
2201 if (principal && sec_svc_inrootlist(rpcflavor, principal,
2202 secp->s_rootcnt, secp->s_rootnames)) {
2203 if (crgetuid(cr) == 0 && secp->s_rootid == 0)
2204 return (1);
2205
2206
2207 (void) crsetugid(cr, secp->s_rootid, secp->s_rootid);
2208
2209 /*
2210 * NOTE: If and when kernel-land privilege tracing is
2211 * added this may have to be replaced with code that
2212 * retrieves root's supplementary groups (e.g., using
2213 * kgss_get_group_info(). In the meantime principals
2214 * mapped to uid 0 get all privileges, so setting cr's
2215 * supplementary groups for them does nothing.
2216 */
2217 (void) crsetgroups(cr, 0, NULL);
2218
2219 return (1);
2220 }
2221
2222 /*
2223 * Not a root princ, or not in root list, map UID 0/nobody to
2224 * the anon ID for the share. (RPC sets cr's UIDs and GIDs to
2225 * UID_NOBODY and GID_NOBODY, respectively.)
2226 */
2227 if (crgetuid(cr) != 0 &&
2228 (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY))
2229 return (1);
2230
2231 anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2232 exi->exi_export.ex_anon);
2233 (void) crsetgroups(cr, 0, NULL);
2234 break;
2235 default:
2236 return (0);
2237 } /* switch on rpcflavor */
2238
2239 /*
2240 * Even if anon access is disallowed via ex_anon == -1, we allow
2241 * this access if anon_ok is set. So set creds to the default
2242 * "nobody" id.
2243 */
2244 if (anon_res != 0) {
2245 if (anon_ok == 0) {
2246 cmn_err(CE_NOTE,
2247 "nfs_server: client %s%ssent wrong "
2248 "authentication for %s",
2249 client_name(req), client_addr(req, buf),
2250 exi->exi_export.ex_path ?
2251 exi->exi_export.ex_path : "?");
2252 return (0);
2253 }
2254
2255 if (crsetugid(cr, UID_NOBODY, GID_NOBODY) != 0)
2256 return (0);
2257 }
2258
2259 return (1);
2260 }
2261
2262 /*
2263 * returns 0 on failure, -1 on a drop, -2 on wrong security flavor,
2264 * and 1 on success
2265 */
2266 int
2267 checkauth4(struct compound_state *cs, struct svc_req *req)
2268 {
2269 int i, rpcflavor, access;
2270 struct secinfo *secp;
2271 char buf[MAXHOST + 1];
2272 int anon_res = 0, nfsflavor;
2273 struct exportinfo *exi;
2274 cred_t *cr;
2275 caddr_t principal;
2276
2277 uid_t uid;
2278 gid_t gid;
2279 uint_t ngids;
2280 gid_t *gids;
2281
2282 exi = cs->exi;
2283 cr = cs->cr;
2284 principal = cs->principal;
2285 nfsflavor = cs->nfsflavor;
2286
2287 ASSERT(cr != NULL);
2288
2289 rpcflavor = req->rq_cred.oa_flavor;
2290 cs->access &= ~CS_ACCESS_LIMITED;
2291
2292 /*
2293 * Check for privileged port number
2294 * N.B.: this assumes that we know the format of a netbuf.
2295 */
2296 if (nfs_portmon) {
2297 struct sockaddr *ca;
2298 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2299
2300 if (ca == NULL)
2301 return (0);
2302
2303 if ((ca->sa_family == AF_INET &&
2304 ntohs(((struct sockaddr_in *)ca)->sin_port) >=
2305 IPPORT_RESERVED) ||
2306 (ca->sa_family == AF_INET6 &&
2307 ntohs(((struct sockaddr_in6 *)ca)->sin6_port) >=
2308 IPPORT_RESERVED)) {
2309 cmn_err(CE_NOTE,
2310 "nfs_server: client %s%ssent NFSv4 request from "
2311 "unprivileged port",
2312 client_name(req), client_addr(req, buf));
2313 return (0);
2314 }
2315 }
2316
2317 /*
2318 * Check the access right per auth flavor on the vnode of
2319 * this export for the given request.
2320 */
2321 access = nfsauth4_access(cs->exi, cs->vp, req, cr, &uid, &gid, &ngids,
2322 &gids);
2323
2324 if (access & NFSAUTH_WRONGSEC)
2325 return (-2); /* no access for this security flavor */
2326
2327 if (access & NFSAUTH_DROP)
2328 return (-1); /* drop the request */
2329
2330 if (access & NFSAUTH_DENIED) {
2331
2332 if (exi->exi_export.ex_seccnt > 0)
2333 return (0); /* deny access */
2334
2335 } else if (access & NFSAUTH_LIMITED) {
2336
2337 cs->access |= CS_ACCESS_LIMITED;
2338
2339 } else if (access & NFSAUTH_MAPNONE) {
2340 /*
2341 * Access was granted even though the flavor mismatched
2342 * because AUTH_NONE was one of the exported flavors.
2343 */
2344 rpcflavor = AUTH_NONE;
2345 }
2346
2347 /*
2348 * XXX probably need to redo some of it for nfsv4?
2349 * return 1 on success or 0 on failure
2350 */
2351
2352 if (rpcflavor != AUTH_SYS)
2353 kmem_free(gids, ngids * sizeof (gid_t));
2354
2355 switch (rpcflavor) {
2356 case AUTH_NONE:
2357 anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2358 exi->exi_export.ex_anon);
2359 (void) crsetgroups(cr, 0, NULL);
2360 break;
2361
2362 case AUTH_UNIX:
2363 if (crgetuid(cr) == 0 && !(access & NFSAUTH_UIDMAP)) {
2364 anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2365 exi->exi_export.ex_anon);
2366 (void) crsetgroups(cr, 0, NULL);
2367 } else if (crgetuid(cr) == 0 && access & NFSAUTH_ROOT) {
2368 /*
2369 * It is root, so apply rootid to get real UID
2370 * Find the secinfo structure. We should be able
2371 * to find it by the time we reach here.
2372 * nfsauth_access() has done the checking.
2373 */
2374 secp = NULL;
2375 for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2376 struct secinfo *sptr;
2377 sptr = &exi->exi_export.ex_secinfo[i];
2378 if (sptr->s_secinfo.sc_nfsnum == nfsflavor) {
2379 secp = &exi->exi_export.ex_secinfo[i];
2380 break;
2381 }
2382 }
2383 if (secp != NULL) {
2384 (void) crsetugid(cr, secp->s_rootid,
2385 secp->s_rootid);
2386 (void) crsetgroups(cr, 0, NULL);
2387 }
2388 } else if (crgetuid(cr) != uid || crgetgid(cr) != gid) {
2389 if (crsetugid(cr, uid, gid) != 0)
2390 anon_res = crsetugid(cr,
2391 exi->exi_export.ex_anon,
2392 exi->exi_export.ex_anon);
2393 (void) crsetgroups(cr, 0, NULL);
2394 } if (access & NFSAUTH_GROUPS) {
2395 (void) crsetgroups(cr, ngids, gids);
2396 }
2397
2398 kmem_free(gids, ngids * sizeof (gid_t));
2399
2400 break;
2401
2402 default:
2403 /*
2404 * Find the secinfo structure. We should be able
2405 * to find it by the time we reach here.
2406 * nfsauth_access() has done the checking.
2407 */
2408 secp = NULL;
2409 for (i = 0; i < exi->exi_export.ex_seccnt; i++) {
2410 if (exi->exi_export.ex_secinfo[i].s_secinfo.sc_nfsnum ==
2411 nfsflavor) {
2412 secp = &exi->exi_export.ex_secinfo[i];
2413 break;
2414 }
2415 }
2416
2417 if (!secp) {
2418 cmn_err(CE_NOTE, "nfs_server: client %s%shad "
2419 "no secinfo data for flavor %d",
2420 client_name(req), client_addr(req, buf),
2421 nfsflavor);
2422 return (0);
2423 }
2424
2425 if (!checkwin(rpcflavor, secp->s_window, req)) {
2426 cmn_err(CE_NOTE,
2427 "nfs_server: client %s%sused invalid "
2428 "auth window value",
2429 client_name(req), client_addr(req, buf));
2430 return (0);
2431 }
2432
2433 /*
2434 * Map root principals listed in the share's root= list to root,
2435 * and map any others principals that were mapped to root by RPC
2436 * to anon. If not going to anon, set to rootid (root_mapping).
2437 */
2438 if (principal && sec_svc_inrootlist(rpcflavor, principal,
2439 secp->s_rootcnt, secp->s_rootnames)) {
2440 if (crgetuid(cr) == 0 && secp->s_rootid == 0)
2441 return (1);
2442
2443 (void) crsetugid(cr, secp->s_rootid, secp->s_rootid);
2444
2445 /*
2446 * NOTE: If and when kernel-land privilege tracing is
2447 * added this may have to be replaced with code that
2448 * retrieves root's supplementary groups (e.g., using
2449 * kgss_get_group_info(). In the meantime principals
2450 * mapped to uid 0 get all privileges, so setting cr's
2451 * supplementary groups for them does nothing.
2452 */
2453 (void) crsetgroups(cr, 0, NULL);
2454
2455 return (1);
2456 }
2457
2458 /*
2459 * Not a root princ, or not in root list, map UID 0/nobody to
2460 * the anon ID for the share. (RPC sets cr's UIDs and GIDs to
2461 * UID_NOBODY and GID_NOBODY, respectively.)
2462 */
2463 if (crgetuid(cr) != 0 &&
2464 (crgetuid(cr) != UID_NOBODY || crgetgid(cr) != GID_NOBODY))
2465 return (1);
2466
2467 anon_res = crsetugid(cr, exi->exi_export.ex_anon,
2468 exi->exi_export.ex_anon);
2469 (void) crsetgroups(cr, 0, NULL);
2470 break;
2471 } /* switch on rpcflavor */
2472
2473 /*
2474 * Even if anon access is disallowed via ex_anon == -1, we allow
2475 * this access if anon_ok is set. So set creds to the default
2476 * "nobody" id.
2477 */
2478
2479 if (anon_res != 0) {
2480 cmn_err(CE_NOTE,
2481 "nfs_server: client %s%ssent wrong "
2482 "authentication for %s",
2483 client_name(req), client_addr(req, buf),
2484 exi->exi_export.ex_path ?
2485 exi->exi_export.ex_path : "?");
2486 return (0);
2487 }
2488
2489 return (1);
2490 }
2491
2492
2493 static char *
2494 client_name(struct svc_req *req)
2495 {
2496 char *hostname = NULL;
2497
2498 /*
2499 * If it's a Unix cred then use the
2500 * hostname from the credential.
2501 */
2502 if (req->rq_cred.oa_flavor == AUTH_UNIX) {
2503 hostname = ((struct authunix_parms *)
2504 req->rq_clntcred)->aup_machname;
2505 }
2506 if (hostname == NULL)
2507 hostname = "";
2508
2509 return (hostname);
2510 }
2511
2512 static char *
2513 client_addr(struct svc_req *req, char *buf)
2514 {
2515 struct sockaddr *ca;
2516 uchar_t *b;
2517 char *frontspace = "";
2518
2519 /*
2520 * We assume we are called in tandem with client_name and the
2521 * format string looks like "...client %s%sblah blah..."
2522 *
2523 * If it's a Unix cred then client_name returned
2524 * a host name, so we need insert a space between host name
2525 * and IP address.
2526 */
2527 if (req->rq_cred.oa_flavor == AUTH_UNIX)
2528 frontspace = " ";
2529
2530 /*
2531 * Convert the caller's IP address to a dotted string
2532 */
2533 ca = (struct sockaddr *)svc_getrpccaller(req->rq_xprt)->buf;
2534
2535 if (ca->sa_family == AF_INET) {
2536 b = (uchar_t *)&((struct sockaddr_in *)ca)->sin_addr;
2537 (void) sprintf(buf, "%s(%d.%d.%d.%d) ", frontspace,
2538 b[0] & 0xFF, b[1] & 0xFF, b[2] & 0xFF, b[3] & 0xFF);
2539 } else if (ca->sa_family == AF_INET6) {
2540 struct sockaddr_in6 *sin6;
2541 sin6 = (struct sockaddr_in6 *)ca;
2542 (void) kinet_ntop6((uchar_t *)&sin6->sin6_addr,
2543 buf, INET6_ADDRSTRLEN);
2544
2545 } else {
2546
2547 /*
2548 * No IP address to print. If there was a host name
2549 * printed, then we print a space.
2550 */
2551 (void) sprintf(buf, frontspace);
2552 }
2553
2554 return (buf);
2555 }
2556
2557 /*
2558 * NFS Server initialization routine. This routine should only be called
2559 * once. It performs the following tasks:
2560 * - Call sub-initialization routines (localize access to variables)
2561 * - Initialize all locks
2562 * - initialize the version 3 write verifier
2563 */
2564 void
2565 nfs_srvinit(void)
2566 {
2567
2568 /* Truly global stuff in this module (not per zone) */
2569 rw_init(&nfssrv_globals_rwl, NULL, RW_DEFAULT, NULL);
2570 list_create(&nfssrv_globals_list, sizeof (nfs_globals_t),
2571 offsetof (nfs_globals_t, nfs_g_link));
2572
2573 /* The order here is important */
2574 nfs_exportinit();
2575 rfs_srvrinit();
2576 rfs3_srvrinit();
2577 rfs4_srvrinit();
2578 nfsauth_init();
2579
2580 /*
2581 * NFS server zone-specific global variables
2582 * Note the zone_init is called for the GZ here.
2583 */
2584 zone_key_create(&nfssrv_zone_key, nfs_server_zone_init,
2585 nfs_server_zone_shutdown, nfs_server_zone_fini);
2586 }
2587
2588 /*
2589 * NFS Server finalization routine. This routine is called to cleanup the
2590 * initialization work previously performed if the NFS server module could
2591 * not be loaded correctly.
2592 */
2593 void
2594 nfs_srvfini(void)
2595 {
2596
2597 /*
2598 * NFS server zone-specific global variables
2599 * Note the zone_fini is called for the GZ here.
2600 */
2601 (void) zone_key_delete(nfssrv_zone_key);
2602
2603 /* The order here is important (reverse of init) */
2604 nfsauth_fini();
2605 rfs4_srvrfini();
2606 rfs3_srvrfini();
2607 rfs_srvrfini();
2608 nfs_exportfini();
2609
2610 /* Truly global stuff in this module (not per zone) */
2611 list_destroy(&nfssrv_globals_list);
2612 rw_destroy(&nfssrv_globals_rwl);
2613 }
2614
2615 /*
2616 * Zone init, shutdown, fini functions for the NFS server
2617 *
2618 * This design is careful to create the entire hierarhcy of
2619 * NFS server "globals" (including those created by various
2620 * per-module *_zone_init functions, etc.) so that all these
2621 * objects have exactly the same lifetime.
2622 *
2623 * These objects are also kept on a list for two reasons:
2624 * 1: It makes finding these in mdb _much_ easier.
2625 * 2: It allows operating across all zone globals for
2626 * functions like nfs_auth.c:exi_cache_reclaim
2627 */
2628 static void *
2629 nfs_server_zone_init(zoneid_t zoneid)
2630 {
2631 nfs_globals_t *ng;
2632
2633 ng = kmem_zalloc(sizeof (*ng), KM_SLEEP);
2634
2635 ng->nfs_versmin = NFS_VERSMIN_DEFAULT;
2636 ng->nfs_versmax = NFS_VERSMAX_DEFAULT;
2637
2638 /* Init the stuff to control start/stop */
2639 ng->nfs_server_upordown = NFS_SERVER_STOPPED;
2640 mutex_init(&ng->nfs_server_upordown_lock, NULL, MUTEX_DEFAULT, NULL);
2641 cv_init(&ng->nfs_server_upordown_cv, NULL, CV_DEFAULT, NULL);
2642 mutex_init(&ng->rdma_wait_mutex, NULL, MUTEX_DEFAULT, NULL);
2643 cv_init(&ng->rdma_wait_cv, NULL, CV_DEFAULT, NULL);
2644
2645 ng->nfs_zoneid = zoneid;
2646
2647 /*
2648 * Order here is important.
2649 * export init must precede srv init calls.
2650 */
2651 nfs_export_zone_init(ng);
2652 rfs_srv_zone_init(ng);
2653 rfs3_srv_zone_init(ng);
2654 rfs4_srv_zone_init(ng);
2655 nfsauth_zone_init(ng);
2656
2657 rw_enter(&nfssrv_globals_rwl, RW_WRITER);
2658 list_insert_tail(&nfssrv_globals_list, ng);
2659 rw_exit(&nfssrv_globals_rwl);
2660
2661 return (ng);
2662 }
2663
2664 /* ARGSUSED */
2665 static void
2666 nfs_server_zone_shutdown(zoneid_t zoneid, void *data)
2667 {
2668 nfs_globals_t *ng;
2669
2670 ng = (nfs_globals_t *)data;
2671
2672 /*
2673 * Order is like _fini, but only
2674 * some modules need this hook.
2675 */
2676 nfsauth_zone_shutdown(ng);
2677 nfs_export_zone_shutdown(ng);
2678 }
2679
2680 /* ARGSUSED */
2681 static void
2682 nfs_server_zone_fini(zoneid_t zoneid, void *data)
2683 {
2684 nfs_globals_t *ng;
2685
2686 ng = (nfs_globals_t *)data;
2687
2688 rw_enter(&nfssrv_globals_rwl, RW_WRITER);
2689 list_remove(&nfssrv_globals_list, ng);
2690 rw_exit(&nfssrv_globals_rwl);
2691
2692 /*
2693 * Order here is important.
2694 * reverse order from init
2695 */
2696 nfsauth_zone_fini(ng);
2697 rfs4_srv_zone_fini(ng);
2698 rfs3_srv_zone_fini(ng);
2699 rfs_srv_zone_fini(ng);
2700 nfs_export_zone_fini(ng);
2701
2702 mutex_destroy(&ng->nfs_server_upordown_lock);
2703 cv_destroy(&ng->nfs_server_upordown_cv);
2704 mutex_destroy(&ng->rdma_wait_mutex);
2705 cv_destroy(&ng->rdma_wait_cv);
2706
2707 kmem_free(ng, sizeof (*ng));
2708 }
2709
2710 /*
2711 * Set up an iovec array of up to cnt pointers.
2712 */
2713 void
2714 mblk_to_iov(mblk_t *m, int cnt, struct iovec *iovp)
2715 {
2716 while (m != NULL && cnt-- > 0) {
2717 iovp->iov_base = (caddr_t)m->b_rptr;
2718 iovp->iov_len = (m->b_wptr - m->b_rptr);
2719 iovp++;
2720 m = m->b_cont;
2721 }
2722 }
2723
2724 /*
2725 * Common code between NFS Version 2 and NFS Version 3 for the public
2726 * filehandle multicomponent lookups.
2727 */
2728
2729 /*
2730 * Public filehandle evaluation of a multi-component lookup, following
2731 * symbolic links, if necessary. This may result in a vnode in another
2732 * filesystem, which is OK as long as the other filesystem is exported.
2733 *
2734 * Note that the exi will be set either to NULL or a new reference to the
2735 * exportinfo struct that corresponds to the vnode of the multi-component path.
2736 * It is the callers responsibility to release this reference.
2737 */
2738 int
2739 rfs_publicfh_mclookup(char *p, vnode_t *dvp, cred_t *cr, vnode_t **vpp,
2740 struct exportinfo **exi, struct sec_ol *sec)
2741 {
2742 int pathflag;
2743 vnode_t *mc_dvp = NULL;
2744 vnode_t *realvp;
2745 int error;
2746
2747 *exi = NULL;
2748
2749 /*
2750 * check if the given path is a url or native path. Since p is
2751 * modified by MCLpath(), it may be empty after returning from
2752 * there, and should be checked.
2753 */
2754 if ((pathflag = MCLpath(&p)) == -1)
2755 return (EIO);
2756
2757 /*
2758 * If pathflag is SECURITY_QUERY, turn the SEC_QUERY bit
2759 * on in sec->sec_flags. This bit will later serve as an
2760 * indication in makefh_ol() or makefh3_ol() to overload the
2761 * filehandle to contain the sec modes used by the server for
2762 * the path.
2763 */
2764 if (pathflag == SECURITY_QUERY) {
2765 if ((sec->sec_index = (uint_t)(*p)) > 0) {
2766 sec->sec_flags |= SEC_QUERY;
2767 p++;
2768 if ((pathflag = MCLpath(&p)) == -1)
2769 return (EIO);
2770 } else {
2771 cmn_err(CE_NOTE,
2772 "nfs_server: invalid security index %d, "
2773 "violating WebNFS SNEGO protocol.", sec->sec_index);
2774 return (EIO);
2775 }
2776 }
2777
2778 if (p[0] == '\0') {
2779 error = ENOENT;
2780 goto publicfh_done;
2781 }
2782
2783 error = rfs_pathname(p, &mc_dvp, vpp, dvp, cr, pathflag);
2784
2785 /*
2786 * If name resolves to "/" we get EINVAL since we asked for
2787 * the vnode of the directory that the file is in. Try again
2788 * with NULL directory vnode.
2789 */
2790 if (error == EINVAL) {
2791 error = rfs_pathname(p, NULL, vpp, dvp, cr, pathflag);
2792 if (!error) {
2793 ASSERT(*vpp != NULL);
2794 if ((*vpp)->v_type == VDIR) {
2795 VN_HOLD(*vpp);
2796 mc_dvp = *vpp;
2797 } else {
2798 /*
2799 * This should not happen, the filesystem is
2800 * in an inconsistent state. Fail the lookup
2801 * at this point.
2802 */
2803 VN_RELE(*vpp);
2804 error = EINVAL;
2805 }
2806 }
2807 }
2808
2809 if (error)
2810 goto publicfh_done;
2811
2812 if (*vpp == NULL) {
2813 error = ENOENT;
2814 goto publicfh_done;
2815 }
2816
2817 ASSERT(mc_dvp != NULL);
2818 ASSERT(*vpp != NULL);
2819
2820 if ((*vpp)->v_type == VDIR) {
2821 do {
2822 /*
2823 * *vpp may be an AutoFS node, so we perform
2824 * a VOP_ACCESS() to trigger the mount of the intended
2825 * filesystem, so we can perform the lookup in the
2826 * intended filesystem.
2827 */
2828 (void) VOP_ACCESS(*vpp, 0, 0, cr, NULL);
2829
2830 /*
2831 * If vnode is covered, get the
2832 * the topmost vnode.
2833 */
2834 if (vn_mountedvfs(*vpp) != NULL) {
2835 error = traverse(vpp);
2836 if (error) {
2837 VN_RELE(*vpp);
2838 goto publicfh_done;
2839 }
2840 }
2841
2842 if (VOP_REALVP(*vpp, &realvp, NULL) == 0 &&
2843 realvp != *vpp) {
2844 /*
2845 * If realvp is different from *vpp
2846 * then release our reference on *vpp, so that
2847 * the export access check be performed on the
2848 * real filesystem instead.
2849 */
2850 VN_HOLD(realvp);
2851 VN_RELE(*vpp);
2852 *vpp = realvp;
2853 } else {
2854 break;
2855 }
2856 /* LINTED */
2857 } while (TRUE);
2858
2859 /*
2860 * Let nfs_vptexi() figure what the real parent is.
2861 */
2862 VN_RELE(mc_dvp);
2863 mc_dvp = NULL;
2864
2865 } else {
2866 /*
2867 * If vnode is covered, get the
2868 * the topmost vnode.
2869 */
2870 if (vn_mountedvfs(mc_dvp) != NULL) {
2871 error = traverse(&mc_dvp);
2872 if (error) {
2873 VN_RELE(*vpp);
2874 goto publicfh_done;
2875 }
2876 }
2877
2878 if (VOP_REALVP(mc_dvp, &realvp, NULL) == 0 &&
2879 realvp != mc_dvp) {
2880 /*
2881 * *vpp is a file, obtain realvp of the parent
2882 * directory vnode.
2883 */
2884 VN_HOLD(realvp);
2885 VN_RELE(mc_dvp);
2886 mc_dvp = realvp;
2887 }
2888 }
2889
2890 /*
2891 * The pathname may take us from the public filesystem to another.
2892 * If that's the case then just set the exportinfo to the new export
2893 * and build filehandle for it. Thanks to per-access checking there's
2894 * no security issues with doing this. If the client is not allowed
2895 * access to this new export then it will get an access error when it
2896 * tries to use the filehandle
2897 */
2898 if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) {
2899 VN_RELE(*vpp);
2900 goto publicfh_done;
2901 }
2902
2903 /*
2904 * Not allowed access to pseudo exports.
2905 */
2906 if (PSEUDO(*exi)) {
2907 error = ENOENT;
2908 VN_RELE(*vpp);
2909 goto publicfh_done;
2910 }
2911
2912 /*
2913 * Do a lookup for the index file. We know the index option doesn't
2914 * allow paths through handling in the share command, so mc_dvp will
2915 * be the parent for the index file vnode, if its present. Use
2916 * temporary pointers to preserve and reuse the vnode pointers of the
2917 * original directory in case there's no index file. Note that the
2918 * index file is a native path, and should not be interpreted by
2919 * the URL parser in rfs_pathname()
2920 */
2921 if (((*exi)->exi_export.ex_flags & EX_INDEX) &&
2922 ((*vpp)->v_type == VDIR) && (pathflag == URLPATH)) {
2923 vnode_t *tvp, *tmc_dvp; /* temporary vnode pointers */
2924
2925 tmc_dvp = mc_dvp;
2926 mc_dvp = tvp = *vpp;
2927
2928 error = rfs_pathname((*exi)->exi_export.ex_index, NULL, vpp,
2929 mc_dvp, cr, NATIVEPATH);
2930
2931 if (error == ENOENT) {
2932 *vpp = tvp;
2933 mc_dvp = tmc_dvp;
2934 error = 0;
2935 } else { /* ok or error other than ENOENT */
2936 if (tmc_dvp)
2937 VN_RELE(tmc_dvp);
2938 if (error)
2939 goto publicfh_done;
2940
2941 /*
2942 * Found a valid vp for index "filename". Sanity check
2943 * for odd case where a directory is provided as index
2944 * option argument and leads us to another filesystem
2945 */
2946
2947 /* Release the reference on the old exi value */
2948 ASSERT(*exi != NULL);
2949 exi_rele(*exi);
2950
2951 if (error = nfs_check_vpexi(mc_dvp, *vpp, kcred, exi)) {
2952 VN_RELE(*vpp);
2953 goto publicfh_done;
2954 }
2955 }
2956 }
2957
2958 publicfh_done:
2959 if (mc_dvp)
2960 VN_RELE(mc_dvp);
2961
2962 return (error);
2963 }
2964
2965 /*
2966 * Evaluate a multi-component path
2967 */
2968 int
2969 rfs_pathname(
2970 char *path, /* pathname to evaluate */
2971 vnode_t **dirvpp, /* ret for ptr to parent dir vnode */
2972 vnode_t **compvpp, /* ret for ptr to component vnode */
2973 vnode_t *startdvp, /* starting vnode */
2974 cred_t *cr, /* user's credential */
2975 int pathflag) /* flag to identify path, e.g. URL */
2976 {
2977 char namebuf[TYPICALMAXPATHLEN];
2978 struct pathname pn;
2979 int error;
2980
2981 /*
2982 * If pathname starts with '/', then set startdvp to root.
2983 */
2984 if (*path == '/') {
2985 while (*path == '/')
2986 path++;
2987
2988 startdvp = ZONE_ROOTVP();
2989 }
2990
2991 error = pn_get_buf(path, UIO_SYSSPACE, &pn, namebuf, sizeof (namebuf));
2992 if (error == 0) {
2993 /*
2994 * Call the URL parser for URL paths to modify the original
2995 * string to handle any '%' encoded characters that exist.
2996 * Done here to avoid an extra bcopy in the lookup.
2997 * We need to be careful about pathlen's. We know that
2998 * rfs_pathname() is called with a non-empty path. However,
2999 * it could be emptied due to the path simply being all /'s,
3000 * which is valid to proceed with the lookup, or due to the
3001 * URL parser finding an encoded null character at the
3002 * beginning of path which should not proceed with the lookup.
3003 */
3004 if (pn.pn_pathlen != 0 && pathflag == URLPATH) {
3005 URLparse(pn.pn_path);
3006 if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0)
3007 return (ENOENT);
3008 }
3009 VN_HOLD(startdvp);
3010 error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp,
3011 ZONE_ROOTVP(), startdvp, cr);
3012 }
3013 if (error == ENAMETOOLONG) {
3014 /*
3015 * This thread used a pathname > TYPICALMAXPATHLEN bytes long.
3016 */
3017 if (error = pn_get(path, UIO_SYSSPACE, &pn))
3018 return (error);
3019 if (pn.pn_pathlen != 0 && pathflag == URLPATH) {
3020 URLparse(pn.pn_path);
3021 if ((pn.pn_pathlen = strlen(pn.pn_path)) == 0) {
3022 pn_free(&pn);
3023 return (ENOENT);
3024 }
3025 }
3026 VN_HOLD(startdvp);
3027 error = lookuppnvp(&pn, NULL, NO_FOLLOW, dirvpp, compvpp,
3028 ZONE_ROOTVP(), startdvp, cr);
3029 pn_free(&pn);
3030 }
3031
3032 return (error);
3033 }
3034
3035 /*
3036 * Adapt the multicomponent lookup path depending on the pathtype
3037 */
3038 static int
3039 MCLpath(char **path)
3040 {
3041 unsigned char c = (unsigned char)**path;
3042
3043 /*
3044 * If the MCL path is between 0x20 and 0x7E (graphic printable
3045 * character of the US-ASCII coded character set), its a URL path,
3046 * per RFC 1738.
3047 */
3048 if (c >= 0x20 && c <= 0x7E)
3049 return (URLPATH);
3050
3051 /*
3052 * If the first octet of the MCL path is not an ASCII character
3053 * then it must be interpreted as a tag value that describes the
3054 * format of the remaining octets of the MCL path.
3055 *
3056 * If the first octet of the MCL path is 0x81 it is a query
3057 * for the security info.
3058 */
3059 switch (c) {
3060 case 0x80: /* native path, i.e. MCL via mount protocol */
3061 (*path)++;
3062 return (NATIVEPATH);
3063 case 0x81: /* security query */
3064 (*path)++;
3065 return (SECURITY_QUERY);
3066 default:
3067 return (-1);
3068 }
3069 }
3070
3071 #define fromhex(c) ((c >= '0' && c <= '9') ? (c - '0') : \
3072 ((c >= 'A' && c <= 'F') ? (c - 'A' + 10) :\
3073 ((c >= 'a' && c <= 'f') ? (c - 'a' + 10) : 0)))
3074
3075 /*
3076 * The implementation of URLparse guarantees that the final string will
3077 * fit in the original one. Replaces '%' occurrences followed by 2 characters
3078 * with its corresponding hexadecimal character.
3079 */
3080 static void
3081 URLparse(char *str)
3082 {
3083 char *p, *q;
3084
3085 p = q = str;
3086 while (*p) {
3087 *q = *p;
3088 if (*p++ == '%') {
3089 if (*p) {
3090 *q = fromhex(*p) * 16;
3091 p++;
3092 if (*p) {
3093 *q += fromhex(*p);
3094 p++;
3095 }
3096 }
3097 }
3098 q++;
3099 }
3100 *q = '\0';
3101 }
3102
3103
3104 /*
3105 * Get the export information for the lookup vnode, and verify its
3106 * useable.
3107 */
3108 int
3109 nfs_check_vpexi(vnode_t *mc_dvp, vnode_t *vp, cred_t *cr,
3110 struct exportinfo **exi)
3111 {
3112 int walk;
3113 int error = 0;
3114
3115 *exi = nfs_vptoexi(mc_dvp, vp, cr, &walk, NULL, FALSE);
3116 if (*exi == NULL)
3117 error = EACCES;
3118 else {
3119 /*
3120 * If nosub is set for this export then
3121 * a lookup relative to the public fh
3122 * must not terminate below the
3123 * exported directory.
3124 */
3125 if ((*exi)->exi_export.ex_flags & EX_NOSUB && walk > 0)
3126 error = EACCES;
3127 }
3128
3129 return (error);
3130 }
3131
3132 /*
3133 * Used by NFSv3 and NFSv4 server to query label of
3134 * a pathname component during lookup/access ops.
3135 */
3136 ts_label_t *
3137 nfs_getflabel(vnode_t *vp, struct exportinfo *exi)
3138 {
3139 zone_t *zone;
3140 ts_label_t *zone_label;
3141 char *path;
3142
3143 mutex_enter(&vp->v_lock);
3144 if (vp->v_path != vn_vpath_empty) {
3145 zone = zone_find_by_any_path(vp->v_path, B_FALSE);
3146 mutex_exit(&vp->v_lock);
3147 } else {
3148 /*
3149 * v_path not cached. Fall back on pathname of exported
3150 * file system as we rely on pathname from which we can
3151 * derive a label. The exported file system portion of
3152 * path is sufficient to obtain a label.
3153 */
3154 path = exi->exi_export.ex_path;
3155 if (path == NULL) {
3156 mutex_exit(&vp->v_lock);
3157 return (NULL);
3158 }
3159 zone = zone_find_by_any_path(path, B_FALSE);
3160 mutex_exit(&vp->v_lock);
3161 }
3162 /*
3163 * Caller has verified that the file is either
3164 * exported or visible. So if the path falls in
3165 * global zone, admin_low is returned; otherwise
3166 * the zone's label is returned.
3167 */
3168 zone_label = zone->zone_slabel;
3169 label_hold(zone_label);
3170 zone_rele(zone);
3171 return (zone_label);
3172 }
3173
3174 /*
3175 * TX NFS routine used by NFSv3 and NFSv4 to do label check
3176 * on client label and server's file object lable.
3177 */
3178 boolean_t
3179 do_rfs_label_check(bslabel_t *clabel, vnode_t *vp, int flag,
3180 struct exportinfo *exi)
3181 {
3182 bslabel_t *slabel;
3183 ts_label_t *tslabel;
3184 boolean_t result;
3185
3186 if ((tslabel = nfs_getflabel(vp, exi)) == NULL) {
3187 return (B_FALSE);
3188 }
3189 slabel = label2bslabel(tslabel);
3190 DTRACE_PROBE4(tx__rfs__log__info__labelcheck, char *,
3191 "comparing server's file label(1) with client label(2) (vp(3))",
3192 bslabel_t *, slabel, bslabel_t *, clabel, vnode_t *, vp);
3193
3194 if (flag == EQUALITY_CHECK)
3195 result = blequal(clabel, slabel);
3196 else
3197 result = bldominates(clabel, slabel);
3198 label_rele(tslabel);
3199 return (result);
3200 }
3201
3202 /*
3203 * Callback function to return the loaned buffers.
3204 * Calls VOP_RETZCBUF() only after all uio_iov[]
3205 * buffers are returned. nu_ref maintains the count.
3206 */
3207 void
3208 rfs_free_xuio(void *free_arg)
3209 {
3210 uint_t ref;
3211 nfs_xuio_t *nfsuiop = (nfs_xuio_t *)free_arg;
3212
3213 ref = atomic_dec_uint_nv(&nfsuiop->nu_ref);
3214
3215 /*
3216 * Call VOP_RETZCBUF() only when all the iov buffers
3217 * are sent OTW.
3218 */
3219 if (ref != 0)
3220 return;
3221
3222 if (((uio_t *)nfsuiop)->uio_extflg & UIO_XUIO) {
3223 (void) VOP_RETZCBUF(nfsuiop->nu_vp, (xuio_t *)free_arg, NULL,
3224 NULL);
3225 VN_RELE(nfsuiop->nu_vp);
3226 }
3227
3228 kmem_cache_free(nfs_xuio_cache, free_arg);
3229 }
3230
3231 xuio_t *
3232 rfs_setup_xuio(vnode_t *vp)
3233 {
3234 nfs_xuio_t *nfsuiop;
3235
3236 nfsuiop = kmem_cache_alloc(nfs_xuio_cache, KM_SLEEP);
3237
3238 bzero(nfsuiop, sizeof (nfs_xuio_t));
3239 nfsuiop->nu_vp = vp;
3240
3241 /*
3242 * ref count set to 1. more may be added
3243 * if multiple mblks refer to multiple iov's.
3244 * This is done in uio_to_mblk().
3245 */
3246
3247 nfsuiop->nu_ref = 1;
3248
3249 nfsuiop->nu_frtn.free_func = rfs_free_xuio;
3250 nfsuiop->nu_frtn.free_arg = (char *)nfsuiop;
3251
3252 nfsuiop->nu_uio.xu_type = UIOTYPE_ZEROCOPY;
3253
3254 return (&nfsuiop->nu_uio);
3255 }
3256
3257 mblk_t *
3258 uio_to_mblk(uio_t *uiop)
3259 {
3260 struct iovec *iovp;
3261 int i;
3262 mblk_t *mp, *mp1;
3263 nfs_xuio_t *nfsuiop = (nfs_xuio_t *)uiop;
3264
3265 if (uiop->uio_iovcnt == 0)
3266 return (NULL);
3267
3268 iovp = uiop->uio_iov;
3269 mp = mp1 = esballoca((uchar_t *)iovp->iov_base, iovp->iov_len,
3270 BPRI_MED, &nfsuiop->nu_frtn);
3271 ASSERT(mp != NULL);
3272
3273 mp->b_wptr += iovp->iov_len;
3274 mp->b_datap->db_type = M_DATA;
3275
3276 for (i = 1; i < uiop->uio_iovcnt; i++) {
3277 iovp = (uiop->uio_iov + i);
3278
3279 mp1->b_cont = esballoca(
3280 (uchar_t *)iovp->iov_base, iovp->iov_len, BPRI_MED,
3281 &nfsuiop->nu_frtn);
3282
3283 mp1 = mp1->b_cont;
3284 ASSERT(mp1 != NULL);
3285 mp1->b_wptr += iovp->iov_len;
3286 mp1->b_datap->db_type = M_DATA;
3287 }
3288
3289 nfsuiop->nu_ref = uiop->uio_iovcnt;
3290
3291 return (mp);
3292 }
3293
3294 /*
3295 * Allocate memory to hold data for a read request of len bytes.
3296 *
3297 * We don't allocate buffers greater than kmem_max_cached in size to avoid
3298 * allocating memory from the kmem_oversized arena. If we allocate oversized
3299 * buffers, we incur heavy cross-call activity when freeing these large buffers
3300 * in the TCP receive path. Note that we can't set b_wptr here since the
3301 * length of the data returned may differ from the length requested when
3302 * reading the end of a file; we set b_wptr in rfs_rndup_mblks() once the
3303 * length of the read is known.
3304 */
3305 mblk_t *
3306 rfs_read_alloc(uint_t len, struct iovec **iov, int *iovcnt)
3307 {
3308 struct iovec *iovarr;
3309 mblk_t *mp, **mpp = ∓
3310 size_t mpsize;
3311 uint_t remain = len;
3312 int i, err = 0;
3313
3314 *iovcnt = howmany(len, kmem_max_cached);
3315
3316 iovarr = kmem_alloc(*iovcnt * sizeof (struct iovec), KM_SLEEP);
3317 *iov = iovarr;
3318
3319 for (i = 0; i < *iovcnt; remain -= mpsize, i++) {
3320 ASSERT(remain <= len);
3321 /*
3322 * We roundup the size we allocate to a multiple of
3323 * BYTES_PER_XDR_UNIT (4 bytes) so that the call to
3324 * xdrmblk_putmblk() never fails.
3325 */
3326 ASSERT(kmem_max_cached % BYTES_PER_XDR_UNIT == 0);
3327 mpsize = MIN(kmem_max_cached, remain);
3328 *mpp = allocb_wait(RNDUP(mpsize), BPRI_MED, STR_NOSIG, &err);
3329 ASSERT(*mpp != NULL);
3330 ASSERT(err == 0);
3331
3332 iovarr[i].iov_base = (caddr_t)(*mpp)->b_rptr;
3333 iovarr[i].iov_len = mpsize;
3334 mpp = &(*mpp)->b_cont;
3335 }
3336 return (mp);
3337 }
3338
3339 void
3340 rfs_rndup_mblks(mblk_t *mp, uint_t len, int buf_loaned)
3341 {
3342 int i;
3343 int alloc_err = 0;
3344 mblk_t *rmp;
3345 uint_t mpsize, remainder;
3346
3347 remainder = P2NPHASE(len, BYTES_PER_XDR_UNIT);
3348
3349 /*
3350 * Non copy-reduction case. This function assumes that blocks were
3351 * allocated in multiples of BYTES_PER_XDR_UNIT bytes, which makes this
3352 * padding safe without bounds checking.
3353 */
3354 if (!buf_loaned) {
3355 /*
3356 * Set the size of each mblk in the chain until we've consumed
3357 * the specified length for all but the last one.
3358 */
3359 while ((mpsize = MBLKSIZE(mp)) < len) {
3360 ASSERT(mpsize % BYTES_PER_XDR_UNIT == 0);
3361 mp->b_wptr += mpsize;
3362 len -= mpsize;
3363 mp = mp->b_cont;
3364 ASSERT(mp != NULL);
3365 }
3366
3367 ASSERT(len + remainder <= mpsize);
3368 mp->b_wptr += len;
3369 for (i = 0; i < remainder; i++)
3370 *mp->b_wptr++ = '\0';
3371 return;
3372 }
3373
3374 /*
3375 * No remainder mblk required.
3376 */
3377 if (remainder == 0)
3378 return;
3379
3380 /*
3381 * Get to the last mblk in the chain.
3382 */
3383 while (mp->b_cont != NULL)
3384 mp = mp->b_cont;
3385
3386 /*
3387 * In case of copy-reduction mblks, the size of the mblks are fixed
3388 * and are of the size of the loaned buffers. Allocate a remainder
3389 * mblk and chain it to the data buffers. This is sub-optimal, but not
3390 * expected to happen commonly.
3391 */
3392 rmp = allocb_wait(remainder, BPRI_MED, STR_NOSIG, &alloc_err);
3393 ASSERT(rmp != NULL);
3394 ASSERT(alloc_err == 0);
3395
3396 for (i = 0; i < remainder; i++)
3397 *rmp->b_wptr++ = '\0';
3398
3399 rmp->b_datap->db_type = M_DATA;
3400 mp->b_cont = rmp;
3401 }