nfs_serv.c revision 1.176 1 /* $NetBSD: nfs_serv.c,v 1.176 2019/02/03 03:19:28 mrg Exp $ */
2
3 /*
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Rick Macklem at The University of Guelph.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95
35 */
36
37 /*
38 * nfs version 2 and 3 server calls to vnode ops
39 * - these routines generally have 3 phases
40 * 1 - break down and validate rpc request in mbuf list
41 * 2 - do the vnode ops for the request
42 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c)
43 * 3 - build the rpc reply in an mbuf list
44 * nb:
45 * - do not mix the phases, since the nfsm_?? macros can return failures
46 * on a bad rpc or similar and do not do any vrele() or vput()'s
47 *
48 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs
49 * error number iff error != 0 whereas
50 * returning an error from the server function implies a fatal error
51 * such as a badly constructed rpc request that should be dropped without
52 * a reply.
53 * For Version 3, nfsm_reply() does not return for the error case, since
54 * most version 3 rpcs return more than the status for error cases.
55 */
56
57 #include <sys/cdefs.h>
58 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.176 2019/02/03 03:19:28 mrg Exp $");
59
60 #include <sys/param.h>
61 #include <sys/systm.h>
62 #include <sys/proc.h>
63 #include <sys/file.h>
64 #include <sys/namei.h>
65 #include <sys/vnode.h>
66 #include <sys/mount.h>
67 #include <sys/socket.h>
68 #include <sys/socketvar.h>
69 #include <sys/mbuf.h>
70 #include <sys/dirent.h>
71 #include <sys/stat.h>
72 #include <sys/kernel.h>
73 #include <sys/hash.h>
74 #include <sys/kauth.h>
75 #include <sys/module.h>
76 #include <sys/syscall.h>
77 #include <sys/syscallargs.h>
78 #include <sys/syscallvar.h>
79
80 #include <uvm/uvm.h>
81
82 #include <nfs/nfsproto.h>
83 #include <nfs/rpcv2.h>
84 #include <nfs/nfs.h>
85 #include <nfs/xdr_subs.h>
86 #include <nfs/nfsm_subs.h>
87 #include <nfs/nfs_var.h>
88
89 MODULE(MODULE_CLASS_MISC, nfsserver, "nfs");
90
91 /* Global vars */
92 extern u_int32_t nfs_xdrneg1;
93 extern u_int32_t nfs_false, nfs_true;
94 extern const enum vtype nv3tov_type[8];
95 extern struct nfsstats nfsstats;
96 extern const nfstype nfsv2_type[9];
97 extern const nfstype nfsv3_type[9];
98 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
99 bool nfsd_use_loan = true; /* use page-loan for READ OP */
100
101 #define nqsrv_getl(vp, rw) /* nothing */
102
103 static const struct syscall_package nfsserver_syscalls[] = {
104 { SYS_nfssvc, 0, (sy_call_t *)sys_nfssvc },
105 { 0, 0, NULL },
106 };
107
108 static int
109 nfsserver_modcmd(modcmd_t cmd, void *arg)
110 {
111 extern struct vfs_hooks nfs_export_hooks; /* XXX */
112 int error;
113
114 switch (cmd) {
115 case MODULE_CMD_INIT:
116 error = syscall_establish(NULL, nfsserver_syscalls);
117 if (error != 0) {
118 return error;
119 }
120 nfs_init(); /* XXX for monolithic kernel */
121 netexport_init();
122 nfsrv_initcache(); /* Init the server request cache */
123 nfsrv_init(0); /* Init server data structures */
124 vfs_hooks_attach(&nfs_export_hooks);
125 nfs_timer_srvinit(nfsrv_timer);
126 return 0;
127 case MODULE_CMD_FINI:
128 error = syscall_disestablish(NULL, nfsserver_syscalls);
129 if (error != 0) {
130 return error;
131 }
132 /*
133 * Kill export list before detaching VFS hooks, so we
134 * we don't leak state due to a concurrent umount().
135 */
136 netexport_fini();
137 vfs_hooks_detach(&nfs_export_hooks);
138
139 /* Kill timer before server goes away. */
140 nfs_timer_srvfini();
141 nfsrv_fini();
142
143 /* Server uses server cache, so kill cache last. */
144 nfsrv_finicache();
145 nfs_fini();
146 return 0;
147 case MODULE_CMD_AUTOUNLOAD:
148 if (netexport_hasexports())
149 return EBUSY;
150 /*FALLTHROUGH*/
151 default:
152 return ENOTTY;
153 }
154 }
155
156 /*
157 * nfs v3 access service
158 */
159 int
160 nfsrv3_access(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
161 {
162 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
163 struct mbuf *nam = nfsd->nd_nam;
164 char *dpos = nfsd->nd_dpos;
165 kauth_cred_t cred = nfsd->nd_cr;
166 struct vnode *vp;
167 nfsrvfh_t nsfh;
168 u_int32_t *tl;
169 int32_t t1;
170 char *bpos;
171 int error = 0, rdonly, cache = 0, getret;
172 char *cp2;
173 struct mbuf *mb, *mreq __unused;
174 struct vattr va;
175 u_long inmode, testmode, outmode;
176 u_quad_t frev;
177
178 nfsm_srvmtofh(&nsfh);
179 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
180 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly,
181 (nfsd->nd_flag & ND_KERBAUTH), false);
182 if (error) {
183 nfsm_reply(NFSX_UNSIGNED);
184 nfsm_srvpostop_attr(1, (struct vattr *)0);
185 return (0);
186 }
187 inmode = fxdr_unsigned(u_int32_t, *tl);
188 outmode = 0;
189 if ((inmode & NFSV3ACCESS_READ) &&
190 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0)
191 outmode |= NFSV3ACCESS_READ;
192 if (vp->v_type != VDIR) {
193 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
194 if (testmode &&
195 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0)
196 outmode |= testmode;
197 if ((inmode & NFSV3ACCESS_EXECUTE) &&
198 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0)
199 outmode |= NFSV3ACCESS_EXECUTE;
200 } else {
201 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
202 NFSV3ACCESS_DELETE);
203 if (testmode &&
204 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0)
205 outmode |= testmode;
206 if ((inmode & NFSV3ACCESS_LOOKUP) &&
207 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0)
208 outmode |= NFSV3ACCESS_LOOKUP;
209 }
210 getret = VOP_GETATTR(vp, &va, cred);
211 vput(vp);
212 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
213 nfsm_srvpostop_attr(getret, &va);
214 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
215 *tl = txdr_unsigned(outmode);
216 nfsm_srvdone;
217 }
218
219 /*
220 * nfs getattr service
221 */
222 int
223 nfsrv_getattr(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
224 {
225 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
226 struct mbuf *nam = nfsd->nd_nam;
227 char *dpos = nfsd->nd_dpos;
228 kauth_cred_t cred = nfsd->nd_cr;
229 struct nfs_fattr *fp;
230 struct vattr va;
231 struct vnode *vp;
232 nfsrvfh_t nsfh;
233 u_int32_t *tl;
234 int32_t t1;
235 char *bpos;
236 int error = 0, rdonly, cache = 0;
237 char *cp2;
238 struct mbuf *mb, *mreq __unused;
239 u_quad_t frev;
240
241 nfsm_srvmtofh(&nsfh);
242 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly,
243 (nfsd->nd_flag & ND_KERBAUTH), false);
244 if (error) {
245 nfsm_reply(0);
246 return (0);
247 }
248 nqsrv_getl(vp, ND_READ);
249 error = VOP_GETATTR(vp, &va, cred);
250 vput(vp);
251 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
252 if (error)
253 return (0);
254 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
255 nfsm_srvfillattr(&va, fp);
256 nfsm_srvdone;
257 }
258
259 /*
260 * nfs setattr service
261 */
262 int
263 nfsrv_setattr(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
264 {
265 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
266 struct mbuf *nam = nfsd->nd_nam;
267 char *dpos = nfsd->nd_dpos;
268 kauth_cred_t cred = nfsd->nd_cr;
269 struct vattr va, preat;
270 struct nfsv2_sattr *sp;
271 struct nfs_fattr *fp;
272 struct vnode *vp;
273 nfsrvfh_t nsfh;
274 u_int32_t *tl;
275 int32_t t1;
276 char *bpos;
277 int error = 0, rdonly, cache = 0, preat_ret = 1, postat_ret = 1;
278 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
279 char *cp2;
280 struct mbuf *mb, *mreq __unused;
281 u_quad_t frev;
282 struct timespec guard;
283
284 memset(&guard, 0, sizeof guard); /* XXX gcc */
285
286 nfsm_srvmtofh(&nsfh);
287 vattr_null(&va);
288 if (v3) {
289 nfsm_srvsattr(&va);
290 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
291 gcheck = fxdr_unsigned(int, *tl);
292 if (gcheck) {
293 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
294 fxdr_nfsv3time(tl, &guard);
295 }
296 } else {
297 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
298 /*
299 * Nah nah nah nah na nah
300 * There is a bug in the Sun client that puts 0xffff in the mode
301 * field of sattr when it should put in 0xffffffff. The u_short
302 * doesn't sign extend.
303 * --> check the low order 2 bytes for 0xffff
304 */
305 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
306 va.va_mode = nfstov_mode(sp->sa_mode);
307 if (sp->sa_uid != nfs_xdrneg1)
308 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
309 if (sp->sa_gid != nfs_xdrneg1)
310 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
311 if (sp->sa_size != nfs_xdrneg1)
312 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
313 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
314 #ifdef notyet
315 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime);
316 #else
317 va.va_atime.tv_sec =
318 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
319 va.va_atime.tv_nsec = 0;
320 #endif
321 }
322 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
323 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime);
324
325 }
326
327 /*
328 * Now that we have all the fields, lets do it.
329 */
330 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly,
331 (nfsd->nd_flag & ND_KERBAUTH), false);
332 if (error) {
333 nfsm_reply(2 * NFSX_UNSIGNED);
334 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
335 return (0);
336 }
337 nqsrv_getl(vp, ND_WRITE);
338 if (v3) {
339 error = preat_ret = VOP_GETATTR(vp, &preat, cred);
340 if (!error && gcheck &&
341 (preat.va_ctime.tv_sec != guard.tv_sec ||
342 preat.va_ctime.tv_nsec != guard.tv_nsec))
343 error = NFSERR_NOT_SYNC;
344 if (error) {
345 vput(vp);
346 nfsm_reply(NFSX_WCCDATA(v3));
347 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
348 return (0);
349 }
350 }
351
352 /*
353 * If the size is being changed write acces is required, otherwise
354 * just check for a read only file system.
355 */
356 if (va.va_size == ((u_quad_t)((quad_t) -1))) {
357 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
358 error = EROFS;
359 goto out;
360 }
361 } else {
362 if (vp->v_type == VDIR) {
363 error = EISDIR;
364 goto out;
365 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
366 lwp, 0)) != 0)
367 goto out;
368 }
369 error = VOP_SETATTR(vp, &va, cred);
370 postat_ret = VOP_GETATTR(vp, &va, cred);
371 if (!error)
372 error = postat_ret;
373 out:
374 vput(vp);
375 nfsm_reply(NFSX_WCCORFATTR(v3));
376 if (v3) {
377 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
378 return (0);
379 } else {
380 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
381 nfsm_srvfillattr(&va, fp);
382 }
383 nfsm_srvdone;
384 }
385
386 /*
387 * nfs lookup rpc
388 */
389 int
390 nfsrv_lookup(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
391 {
392 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
393 struct mbuf *nam = nfsd->nd_nam;
394 char *dpos = nfsd->nd_dpos;
395 kauth_cred_t cred = nfsd->nd_cr;
396 struct nfs_fattr *fp;
397 struct nameidata nd, ind, *ndp = &nd;
398 struct pathbuf *ipb = NULL;
399 struct vnode *vp, *dirp;
400 nfsrvfh_t nsfh;
401 char *cp;
402 u_int32_t *tl;
403 int32_t t1;
404 char *bpos;
405 int error = 0, cache = 0, dirattr_ret = 1;
406 uint32_t len;
407 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
408 char *cp2;
409 struct mbuf *mb, *mreq __unused;
410 struct vattr va, dirattr;
411 u_quad_t frev;
412
413 nfsm_srvmtofh(&nsfh);
414 nfsm_srvnamesiz(len);
415
416 pubflag = nfs_ispublicfh(&nsfh);
417
418 nd.ni_cnd.cn_cred = cred;
419 nd.ni_cnd.cn_nameiop = LOOKUP;
420 nd.ni_cnd.cn_flags = LOCKLEAF;
421 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
422 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
423
424 if (!error && pubflag) {
425 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL &&
426 (ipb = pathbuf_create(nfs_pub.np_index)) != NULL) {
427 /*
428 * Setup call to lookup() to see if we can find
429 * the index file. Arguably, this doesn't belong
430 * in a kernel.. Ugh.
431 */
432 ind = nd;
433 VOP_UNLOCK(nd.ni_vp);
434 ind.ni_pathbuf = ipb;
435
436 error = lookup_for_nfsd_index(&ind, nd.ni_vp);
437 if (!error) {
438 /*
439 * Found an index file. Get rid of
440 * the old references.
441 */
442 if (dirp)
443 vrele(dirp);
444 dirp = nd.ni_vp;
445 ndp = &ind;
446 } else
447 error = 0;
448 }
449 /*
450 * If the public filehandle was used, check that this lookup
451 * didn't result in a filehandle outside the publicly exported
452 * filesystem.
453 */
454
455 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
456 vput(nd.ni_vp);
457 error = EPERM;
458 }
459 }
460
461 if (error) {
462 if (nd.ni_pathbuf != NULL) {
463 pathbuf_destroy(nd.ni_pathbuf);
464 }
465 if (ipb != NULL) {
466 pathbuf_destroy(ipb);
467 }
468 if (dirp) {
469 if (v3) {
470 vn_lock(dirp, LK_SHARED | LK_RETRY);
471 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred);
472 vput(dirp);
473 } else
474 vrele(dirp);
475 }
476 nfsm_reply(NFSX_POSTOPATTR(v3));
477 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
478 return (0);
479 }
480
481 nqsrv_getl(ndp->ni_startdir, ND_READ);
482 pathbuf_destroy(nd.ni_pathbuf);
483 if (ipb != NULL) {
484 pathbuf_destroy(ipb);
485 }
486 vp = ndp->ni_vp;
487 error = nfsrv_composefh(vp, &nsfh, v3);
488 if (!error)
489 error = VOP_GETATTR(vp, &va, cred);
490 vput(vp);
491 if (dirp) {
492 if (v3) {
493 vn_lock(dirp, LK_SHARED | LK_RETRY);
494 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred);
495 vput(dirp);
496 } else
497 vrele(dirp);
498 }
499 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPORFATTR(v3) +
500 NFSX_POSTOPATTR(v3));
501 if (error) {
502 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
503 return (0);
504 }
505 nfsm_srvfhtom(&nsfh, v3);
506 if (v3) {
507 nfsm_srvpostop_attr(0, &va);
508 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
509 } else {
510 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
511 nfsm_srvfillattr(&va, fp);
512 }
513 nfsm_srvdone;
514 }
515
516 /*
517 * nfs readlink service
518 */
519 int
520 nfsrv_readlink(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
521 {
522 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
523 struct mbuf *nam = nfsd->nd_nam;
524 char *dpos = nfsd->nd_dpos;
525 kauth_cred_t cred = nfsd->nd_cr;
526 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
527 struct iovec *ivp = iv;
528 struct mbuf *mp;
529 u_int32_t *tl;
530 int32_t t1;
531 char *bpos;
532 int error = 0, rdonly, cache = 0, i, padlen, getret;
533 uint32_t len;
534 int v3 = (nfsd->nd_flag & ND_NFSV3);
535 char *cp2;
536 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq __unused;
537 struct vnode *vp;
538 struct vattr attr;
539 nfsrvfh_t nsfh;
540 struct uio io, *uiop = &io;
541 u_quad_t frev;
542
543 nfsm_srvmtofh(&nsfh);
544 len = 0;
545 i = 0;
546 while (len < NFS_MAXPATHLEN) {
547 mp = m_get(M_WAIT, MT_DATA);
548 MCLAIM(mp, &nfs_mowner);
549 m_clget(mp, M_WAIT);
550 mp->m_len = NFSMSIZ(mp);
551 if (len == 0)
552 mp3 = mp2 = mp;
553 else {
554 mp2->m_next = mp;
555 mp2 = mp;
556 }
557 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
558 mp->m_len = NFS_MAXPATHLEN-len;
559 len = NFS_MAXPATHLEN;
560 } else
561 len += mp->m_len;
562 ivp->iov_base = mtod(mp, void *);
563 ivp->iov_len = mp->m_len;
564 i++;
565 ivp++;
566 }
567 uiop->uio_iov = iv;
568 uiop->uio_iovcnt = i;
569 uiop->uio_offset = 0;
570 uiop->uio_resid = len;
571 uiop->uio_rw = UIO_READ;
572 UIO_SETUP_SYSSPACE(uiop);
573 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
574 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
575 if (error) {
576 m_freem(mp3);
577 nfsm_reply(2 * NFSX_UNSIGNED);
578 nfsm_srvpostop_attr(1, (struct vattr *)0);
579 return (0);
580 }
581 if (vp->v_type != VLNK) {
582 if (v3)
583 error = EINVAL;
584 else
585 error = ENXIO;
586 goto out;
587 }
588 nqsrv_getl(vp, ND_READ);
589 error = VOP_READLINK(vp, uiop, cred);
590 out:
591 getret = VOP_GETATTR(vp, &attr, cred);
592 vput(vp);
593 if (error)
594 m_freem(mp3);
595 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
596 if (v3) {
597 nfsm_srvpostop_attr(getret, &attr);
598 if (error)
599 return (0);
600 }
601 len -= uiop->uio_resid;
602 padlen = nfsm_padlen(len);
603 if (len == 0) {
604 m_freem(mp3);
605 mp3 = NULL;
606 } else if (uiop->uio_resid || padlen)
607 nfs_zeropad(mp3, uiop->uio_resid, padlen);
608 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
609 *tl = txdr_unsigned(len);
610 mb->m_next = mp3;
611 nfsm_srvdone;
612 }
613
614 /*
615 * nfs read service
616 */
617 int
618 nfsrv_read(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
619 {
620 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
621 struct mbuf *nam = nfsd->nd_nam;
622 char *dpos = nfsd->nd_dpos;
623 kauth_cred_t cred = nfsd->nd_cr;
624 struct mbuf *m;
625 struct nfs_fattr *fp;
626 u_int32_t *tl;
627 int32_t t1;
628 int i;
629 char *bpos;
630 int error = 0, rdonly, cache = 0, getret;
631 int v3 = (nfsd->nd_flag & ND_NFSV3);
632 uint32_t reqlen, len, cnt, left;
633 int padlen;
634 char *cp2;
635 struct mbuf *mb, *mreq;
636 struct vnode *vp;
637 nfsrvfh_t nsfh;
638 struct uio io, *uiop = &io;
639 struct vattr va;
640 off_t off;
641 u_quad_t frev;
642
643 nfsm_srvmtofh(&nsfh);
644 if (v3) {
645 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
646 off = fxdr_hyper(tl);
647 } else {
648 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
649 off = (off_t)fxdr_unsigned(u_int32_t, *tl);
650 }
651 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
652 reqlen = fxdr_unsigned(uint32_t, *tl);
653 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd));
654 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
655 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
656 if (error) {
657 nfsm_reply(2 * NFSX_UNSIGNED);
658 nfsm_srvpostop_attr(1, (struct vattr *)0);
659 return (0);
660 }
661 if (vp->v_type != VREG) {
662 if (v3)
663 error = EINVAL;
664 else
665 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
666 }
667 if (!error) {
668 nqsrv_getl(vp, ND_READ);
669 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0)
670 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1);
671 }
672 getret = VOP_GETATTR(vp, &va, cred);
673 if (!error)
674 error = getret;
675 if (error) {
676 vput(vp);
677 nfsm_reply(NFSX_POSTOPATTR(v3));
678 nfsm_srvpostop_attr(getret, &va);
679 return (0);
680 }
681 if (off >= va.va_size)
682 cnt = 0;
683 else if ((off + reqlen) > va.va_size)
684 cnt = va.va_size - off;
685 else
686 cnt = reqlen;
687 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
688 if (v3) {
689 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
690 *tl++ = nfs_true;
691 fp = (struct nfs_fattr *)tl;
692 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
693 } else {
694 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
695 fp = (struct nfs_fattr *)tl;
696 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
697 }
698 len = left = cnt;
699 if (cnt > 0) {
700 if (nfsd_use_loan) {
701 struct vm_page **pgpp;
702 voff_t pgoff = trunc_page(off);
703 int npages;
704 vaddr_t lva;
705
706 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT;
707 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */
708
709 /* allocate kva for mbuf data */
710 lva = sokvaalloc(pgoff, npages << PAGE_SHIFT,
711 slp->ns_so);
712 if (lva == 0) {
713 /* fall back to VOP_READ */
714 goto loan_fail;
715 }
716
717 /* allocate mbuf */
718 m = m_get(M_WAIT, MT_DATA);
719 MCLAIM(m, &nfs_mowner);
720 pgpp = m->m_ext.ext_pgs;
721
722 /* loan pages */
723 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages,
724 pgpp);
725 if (error) {
726 sokvafree(lva, npages << PAGE_SHIFT);
727 m_free(m);
728 if (error == EBUSY)
729 goto loan_fail;
730 goto read_error;
731 }
732
733 /* associate kva to mbuf */
734 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)),
735 cnt, M_MBUF, soloanfree, slp->ns_so);
736 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
737 m->m_len = cnt;
738
739 /* map pages */
740 for (i = 0; i < npages; i++) {
741 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]),
742 VM_PROT_READ, 0);
743 lva += PAGE_SIZE;
744 }
745
746 pmap_update(pmap_kernel());
747
748 mb->m_next = m;
749 mb = m;
750 error = 0;
751 uiop->uio_resid = 0;
752 } else {
753 struct iovec *iv;
754 struct iovec *iv2;
755 struct mbuf *m2;
756 int siz;
757 loan_fail:
758 /*
759 * Generate the mbuf list with the uio_iov ref. to it.
760 */
761 i = 0;
762 m = m2 = mb;
763 while (left > 0) {
764 siz = uimin(M_TRAILINGSPACE(m), left);
765 if (siz > 0) {
766 left -= siz;
767 i++;
768 }
769 if (left > 0) {
770 m = m_get(M_WAIT, MT_DATA);
771 MCLAIM(m, &nfs_mowner);
772 m_clget(m, M_WAIT);
773 m->m_len = 0;
774 m2->m_next = m;
775 m2 = m;
776 }
777 }
778 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK);
779 uiop->uio_iov = iv2 = iv;
780 m = mb;
781 left = cnt;
782 i = 0;
783 while (left > 0) {
784 if (m == NULL)
785 panic("nfsrv_read iov");
786 siz = uimin(M_TRAILINGSPACE(m), left);
787 if (siz > 0) {
788 iv->iov_base = mtod(m, char *) +
789 m->m_len;
790 iv->iov_len = siz;
791 m->m_len += siz;
792 left -= siz;
793 iv++;
794 i++;
795 }
796 m = m->m_next;
797 }
798 uiop->uio_iovcnt = i;
799 uiop->uio_offset = off;
800 uiop->uio_resid = cnt;
801 uiop->uio_rw = UIO_READ;
802 UIO_SETUP_SYSSPACE(uiop);
803 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
804 free((void *)iv2, M_TEMP);
805 }
806 read_error:
807 if (error || (getret = VOP_GETATTR(vp, &va, cred)) != 0){
808 if (!error)
809 error = getret;
810 m_freem(mreq);
811 vput(vp);
812 nfsm_reply(NFSX_POSTOPATTR(v3));
813 nfsm_srvpostop_attr(getret, &va);
814 return (0);
815 }
816 } else {
817 uiop->uio_resid = 0;
818 }
819 vput(vp);
820 nfsm_srvfillattr(&va, fp);
821 len -= uiop->uio_resid;
822 padlen = nfsm_padlen(len);
823 if (uiop->uio_resid || padlen)
824 nfs_zeropad(mb, uiop->uio_resid, padlen);
825 if (v3) {
826 /* count */
827 *tl++ = txdr_unsigned(len);
828 /* eof */
829 if (off + len >= va.va_size)
830 *tl++ = nfs_true;
831 else
832 *tl++ = nfs_false;
833 }
834 *tl = txdr_unsigned(len);
835 nfsm_srvdone;
836 }
837
838 /*
839 * nfs write service
840 */
841 int
842 nfsrv_write(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
843 {
844 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
845 struct mbuf *nam = nfsd->nd_nam;
846 char *dpos = nfsd->nd_dpos;
847 kauth_cred_t cred = nfsd->nd_cr;
848 struct iovec *ivp;
849 int i, cnt;
850 struct mbuf *mp;
851 struct nfs_fattr *fp;
852 struct iovec *iv;
853 struct vattr va, forat;
854 u_int32_t *tl;
855 int32_t t1;
856 char *bpos;
857 int error = 0, rdonly, cache = 0, len, forat_ret = 1;
858 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
859 int stable = NFSV3WRITE_FILESYNC;
860 int v3 = (nfsd->nd_flag & ND_NFSV3);
861 char *cp2;
862 struct mbuf *mb, *mreq __unused;
863 struct vnode *vp;
864 nfsrvfh_t nsfh;
865 struct uio io, *uiop = &io;
866 off_t off;
867 u_quad_t frev;
868
869 if (mrep == NULL) {
870 *mrq = NULL;
871 return (0);
872 }
873 nfsm_srvmtofh(&nsfh);
874 if (v3) {
875 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
876 off = fxdr_hyper(tl);
877 tl += 3;
878 stable = fxdr_unsigned(int, *tl++);
879 } else {
880 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
881 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
882 tl += 2;
883 }
884 retlen = len = fxdr_unsigned(int32_t, *tl);
885 cnt = i = 0;
886
887 /*
888 * For NFS Version 2, it is not obvious what a write of zero length
889 * should do, but I might as well be consistent with Version 3,
890 * which is to return ok so long as there are no permission problems.
891 */
892 if (len > 0) {
893 zeroing = 1;
894 mp = mrep;
895 while (mp) {
896 if (mp == md) {
897 zeroing = 0;
898 adjust = dpos - mtod(mp, char *);
899 mp->m_len -= adjust;
900 if (mp->m_len > 0 && adjust > 0)
901 NFSMADV(mp, adjust);
902 }
903 if (zeroing)
904 mp->m_len = 0;
905 else if (mp->m_len > 0) {
906 i += mp->m_len;
907 if (i > len) {
908 mp->m_len -= (i - len);
909 zeroing = 1;
910 }
911 if (mp->m_len > 0)
912 cnt++;
913 }
914 mp = mp->m_next;
915 }
916 }
917 if (len > NFS_MAXDATA || len < 0 || i < len) {
918 error = EIO;
919 nfsm_reply(2 * NFSX_UNSIGNED);
920 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
921 return (0);
922 }
923 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
924 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
925 if (error) {
926 nfsm_reply(2 * NFSX_UNSIGNED);
927 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
928 return (0);
929 }
930 if (v3)
931 forat_ret = VOP_GETATTR(vp, &forat, cred);
932 if (vp->v_type != VREG) {
933 if (v3)
934 error = EINVAL;
935 else
936 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
937 }
938 if (!error) {
939 nqsrv_getl(vp, ND_WRITE);
940 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1);
941 }
942 if (error) {
943 vput(vp);
944 nfsm_reply(NFSX_WCCDATA(v3));
945 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
946 return (0);
947 }
948
949 if (len > 0) {
950 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK);
951 uiop->uio_iov = iv = ivp;
952 uiop->uio_iovcnt = cnt;
953 mp = mrep;
954 while (mp) {
955 if (mp->m_len > 0) {
956 ivp->iov_base = mtod(mp, void *);
957 ivp->iov_len = mp->m_len;
958 ivp++;
959 }
960 mp = mp->m_next;
961 }
962
963 /*
964 * XXX
965 * The IO_METASYNC flag indicates that all metadata (and not
966 * just enough to ensure data integrity) must be written to
967 * stable storage synchronously.
968 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
969 */
970 if (stable == NFSV3WRITE_UNSTABLE)
971 ioflags = IO_NODELOCKED;
972 else if (stable == NFSV3WRITE_DATASYNC)
973 ioflags = (IO_SYNC | IO_NODELOCKED);
974 else
975 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
976 uiop->uio_resid = len;
977 uiop->uio_rw = UIO_WRITE;
978 uiop->uio_offset = off;
979 UIO_SETUP_SYSSPACE(uiop);
980 error = VOP_WRITE(vp, uiop, ioflags, cred);
981 nfsstats.srvvop_writes++;
982 free(iv, M_TEMP);
983 }
984 aftat_ret = VOP_GETATTR(vp, &va, cred);
985 vput(vp);
986 if (!error)
987 error = aftat_ret;
988 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
989 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
990 if (v3) {
991 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
992 if (error)
993 return (0);
994 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
995 *tl++ = txdr_unsigned(retlen);
996 if (stable == NFSV3WRITE_UNSTABLE)
997 *tl++ = txdr_unsigned(stable);
998 else
999 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
1000 /*
1001 * Actually, there is no need to txdr these fields,
1002 * but it may make the values more human readable,
1003 * for debugging purposes.
1004 */
1005 *tl++ = txdr_unsigned(boottime.tv_sec);
1006 *tl = txdr_unsigned(boottime.tv_nsec / 1000);
1007 } else {
1008 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1009 nfsm_srvfillattr(&va, fp);
1010 }
1011 nfsm_srvdone;
1012 }
1013
1014 /*
1015 * XXX elad: the original NFSW_SAMECRED() macro also made sure the
1016 * two nd_flag fields of the descriptors contained
1017 * ND_KERBAUTH.
1018 */
1019 static int
1020 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2)
1021 {
1022 int i, do_ngroups;
1023
1024 if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2))
1025 return (0);
1026 if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2))
1027 return (0);
1028 do_ngroups = kauth_cred_ngroups(cred1);
1029 for (i = 0; i < do_ngroups; i++)
1030 if (kauth_cred_group(cred1, i) !=
1031 kauth_cred_group(cred2, i))
1032 return (0);
1033
1034 return (1);
1035 }
1036
1037 static struct nfsrvw_delayhash *
1038 nfsrv_nwdelayhash(struct nfssvc_sock *slp, const nfsrvfh_t *nsfh)
1039 {
1040 uint32_t hash;
1041
1042 hash = hash32_buf(NFSRVFH_DATA(nsfh), NFSRVFH_SIZE(nsfh),
1043 HASH32_BUF_INIT);
1044 return &slp->ns_wdelayhashtbl[hash % NFS_WDELAYHASHSIZ];
1045 }
1046
1047 /*
1048 * NFS write service with write gathering support. Called when
1049 * nfsrvw_procrastinate > 0.
1050 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
1051 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
1052 * Jan. 1994.
1053 */
1054 int
1055 nfsrv_writegather(struct nfsrv_descript **ndp, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1056 {
1057 struct timeval now;
1058 struct iovec *ivp;
1059 struct mbuf *mp;
1060 struct nfsrv_descript *wp, *nfsd, *owp, *swp;
1061 struct nfs_fattr *fp;
1062 int i = 0;
1063 struct iovec *iov;
1064 struct nfsrvw_delayhash *wpp;
1065 kauth_cred_t cred;
1066 struct vattr va, forat;
1067 u_int32_t *tl;
1068 int32_t t1;
1069 char *bpos, *dpos;
1070 int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1;
1071 int ioflags, aftat_ret = 1, adjust, v3, zeroing;
1072 char *cp2;
1073 struct mbuf *mb, *mreq, *mrep, *md;
1074 struct vnode *vp;
1075 struct uio io, *uiop = &io;
1076 u_quad_t frev, cur_usec;
1077
1078 *mrq = NULL;
1079 if (*ndp) {
1080 nfsd = *ndp;
1081 *ndp = NULL;
1082 mrep = nfsd->nd_mrep;
1083 md = nfsd->nd_md;
1084 dpos = nfsd->nd_dpos;
1085 cred = nfsd->nd_cr;
1086 v3 = (nfsd->nd_flag & ND_NFSV3);
1087 LIST_INIT(&nfsd->nd_coalesce);
1088 nfsd->nd_mreq = NULL;
1089 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
1090 getmicrotime(&now);
1091 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec;
1092 nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
1093
1094 /*
1095 * Now, get the write header..
1096 */
1097 nfsm_srvmtofh(&nfsd->nd_fh);
1098 if (v3) {
1099 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1100 nfsd->nd_off = fxdr_hyper(tl);
1101 tl += 3;
1102 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
1103 } else {
1104 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1105 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
1106 tl += 2;
1107 }
1108 len = fxdr_unsigned(int32_t, *tl);
1109 nfsd->nd_len = len;
1110 nfsd->nd_eoff = nfsd->nd_off + len;
1111
1112 /*
1113 * Trim the header out of the mbuf list and trim off any trailing
1114 * junk so that the mbuf list has only the write data.
1115 */
1116 zeroing = 1;
1117 i = 0;
1118 mp = mrep;
1119 while (mp) {
1120 if (mp == md) {
1121 zeroing = 0;
1122 adjust = dpos - mtod(mp, char *);
1123 mp->m_len -= adjust;
1124 if (mp->m_len > 0 && adjust > 0)
1125 NFSMADV(mp, adjust);
1126 }
1127 if (zeroing)
1128 mp->m_len = 0;
1129 else {
1130 i += mp->m_len;
1131 if (i > len) {
1132 mp->m_len -= (i - len);
1133 zeroing = 1;
1134 }
1135 }
1136 mp = mp->m_next;
1137 }
1138 if (len > NFS_MAXDATA || len < 0 || i < len) {
1139 nfsmout:
1140 m_freem(mrep);
1141 error = EIO;
1142 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1143 if (v3)
1144 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1145 nfsd->nd_mreq = mreq;
1146 nfsd->nd_mrep = NULL;
1147 nfsd->nd_time = 0;
1148 }
1149
1150 /*
1151 * Add this entry to the hash and time queues.
1152 */
1153 owp = NULL;
1154 mutex_enter(&nfsd_lock);
1155 wp = LIST_FIRST(&slp->ns_tq);
1156 while (wp && wp->nd_time < nfsd->nd_time) {
1157 owp = wp;
1158 wp = LIST_NEXT(wp, nd_tq);
1159 }
1160 if (owp) {
1161 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1162 } else {
1163 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1164 }
1165 if (nfsd->nd_mrep) {
1166 wpp = nfsrv_nwdelayhash(slp, &nfsd->nd_fh);
1167 owp = NULL;
1168 wp = LIST_FIRST(wpp);
1169 while (wp && nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) {
1170 owp = wp;
1171 wp = LIST_NEXT(wp, nd_hash);
1172 }
1173 while (wp && wp->nd_off < nfsd->nd_off &&
1174 !nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) {
1175 owp = wp;
1176 wp = LIST_NEXT(wp, nd_hash);
1177 }
1178 if (owp) {
1179 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1180
1181 /*
1182 * Search the hash list for overlapping entries and
1183 * coalesce.
1184 */
1185 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1186 wp = LIST_NEXT(nfsd, nd_hash);
1187 if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr))
1188 nfsrvw_coalesce(owp, nfsd);
1189 }
1190 } else {
1191 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1192 }
1193 }
1194 mutex_exit(&nfsd_lock);
1195 }
1196
1197 /*
1198 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1199 * and generate the associated reply mbuf list(s).
1200 */
1201 loop1:
1202 getmicrotime(&now);
1203 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec;
1204 mutex_enter(&nfsd_lock);
1205 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) {
1206 owp = LIST_NEXT(nfsd, nd_tq);
1207 if (nfsd->nd_time > cur_usec)
1208 break;
1209 if (nfsd->nd_mreq)
1210 continue;
1211 LIST_REMOVE(nfsd, nd_tq);
1212 LIST_REMOVE(nfsd, nd_hash);
1213 mutex_exit(&nfsd_lock);
1214
1215 mrep = nfsd->nd_mrep;
1216 nfsd->nd_mrep = NULL;
1217 cred = nfsd->nd_cr;
1218 v3 = (nfsd->nd_flag & ND_NFSV3);
1219 forat_ret = aftat_ret = 1;
1220 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1221 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1222 false);
1223 if (!error) {
1224 if (v3)
1225 forat_ret = VOP_GETATTR(vp, &forat, cred);
1226 if (vp->v_type != VREG) {
1227 if (v3)
1228 error = EINVAL;
1229 else
1230 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1231 }
1232 } else
1233 vp = NULL;
1234 if (!error) {
1235 nqsrv_getl(vp, ND_WRITE);
1236 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1);
1237 }
1238
1239 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1240 ioflags = IO_NODELOCKED;
1241 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1242 ioflags = (IO_SYNC | IO_NODELOCKED);
1243 else
1244 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1245 uiop->uio_rw = UIO_WRITE;
1246 uiop->uio_offset = nfsd->nd_off;
1247 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1248 UIO_SETUP_SYSSPACE(uiop);
1249 if (uiop->uio_resid > 0) {
1250 mp = mrep;
1251 i = 0;
1252 while (mp) {
1253 if (mp->m_len > 0)
1254 i++;
1255 mp = mp->m_next;
1256 }
1257 uiop->uio_iovcnt = i;
1258 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
1259 uiop->uio_iov = ivp = iov;
1260 mp = mrep;
1261 while (mp) {
1262 if (mp->m_len > 0) {
1263 ivp->iov_base = mtod(mp, void *);
1264 ivp->iov_len = mp->m_len;
1265 ivp++;
1266 }
1267 mp = mp->m_next;
1268 }
1269 if (!error) {
1270 error = VOP_WRITE(vp, uiop, ioflags, cred);
1271 nfsstats.srvvop_writes++;
1272 }
1273 free((void *)iov, M_TEMP);
1274 }
1275 m_freem(mrep);
1276 if (vp) {
1277 aftat_ret = VOP_GETATTR(vp, &va, cred);
1278 vput(vp);
1279 }
1280
1281 /*
1282 * Loop around generating replies for all write rpcs that have
1283 * now been completed.
1284 */
1285 swp = nfsd;
1286 do {
1287 if (error) {
1288 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1289 if (v3) {
1290 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1291 }
1292 } else {
1293 nfsm_writereply(NFSX_PREOPATTR(v3) +
1294 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1295 NFSX_WRITEVERF(v3), v3);
1296 if (v3) {
1297 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1298 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1299 *tl++ = txdr_unsigned(nfsd->nd_len);
1300 *tl++ = txdr_unsigned(swp->nd_stable);
1301 /*
1302 * Actually, there is no need to txdr these fields,
1303 * but it may make the values more human readable,
1304 * for debugging purposes.
1305 */
1306 *tl++ = txdr_unsigned(boottime.tv_sec);
1307 *tl = txdr_unsigned(boottime.tv_nsec / 1000);
1308 } else {
1309 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1310 nfsm_srvfillattr(&va, fp);
1311 }
1312 }
1313 nfsd->nd_mreq = mreq;
1314 if (nfsd->nd_mrep)
1315 panic("nfsrv_write: nd_mrep not free");
1316
1317 /*
1318 * Done. Put it at the head of the timer queue so that
1319 * the final phase can return the reply.
1320 */
1321 mutex_enter(&nfsd_lock);
1322 if (nfsd != swp) {
1323 nfsd->nd_time = 0;
1324 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1325 }
1326 nfsd = LIST_FIRST(&swp->nd_coalesce);
1327 if (nfsd) {
1328 LIST_REMOVE(nfsd, nd_tq);
1329 }
1330 mutex_exit(&nfsd_lock);
1331 } while (nfsd);
1332 swp->nd_time = 0;
1333
1334 mutex_enter(&nfsd_lock);
1335 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1336 mutex_exit(&nfsd_lock);
1337 goto loop1;
1338 }
1339 mutex_exit(&nfsd_lock);
1340 nfs_timer_start();
1341
1342 /*
1343 * Search for a reply to return.
1344 */
1345 mutex_enter(&nfsd_lock);
1346 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) {
1347 if (nfsd->nd_mreq) {
1348 LIST_REMOVE(nfsd, nd_tq);
1349 *mrq = nfsd->nd_mreq;
1350 *ndp = nfsd;
1351 break;
1352 }
1353 }
1354 mutex_exit(&nfsd_lock);
1355 return (0);
1356 }
1357
1358 /*
1359 * Coalesce the write request nfsd into owp. To do this we must:
1360 * - remove nfsd from the queues
1361 * - merge nfsd->nd_mrep into owp->nd_mrep
1362 * - update the nd_eoff and nd_stable for owp
1363 * - put nfsd on owp's nd_coalesce list
1364 * NB: Must be called at splsoftclock().
1365 */
1366 void
1367 nfsrvw_coalesce(struct nfsrv_descript *owp, struct nfsrv_descript *nfsd)
1368 {
1369 int overlap;
1370 struct mbuf *mp;
1371 struct nfsrv_descript *m;
1372
1373 KASSERT(mutex_owned(&nfsd_lock));
1374
1375 LIST_REMOVE(nfsd, nd_hash);
1376 LIST_REMOVE(nfsd, nd_tq);
1377 if (owp->nd_eoff < nfsd->nd_eoff) {
1378 overlap = owp->nd_eoff - nfsd->nd_off;
1379 if (overlap < 0)
1380 panic("nfsrv_coalesce: bad off");
1381 if (overlap > 0)
1382 m_adj(nfsd->nd_mrep, overlap);
1383 mp = owp->nd_mrep;
1384 while (mp->m_next)
1385 mp = mp->m_next;
1386 mp->m_next = nfsd->nd_mrep;
1387 owp->nd_eoff = nfsd->nd_eoff;
1388 } else
1389 m_freem(nfsd->nd_mrep);
1390 nfsd->nd_mrep = NULL;
1391 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1392 owp->nd_stable = NFSV3WRITE_FILESYNC;
1393 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1394 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1395 owp->nd_stable = NFSV3WRITE_DATASYNC;
1396 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1397 /*
1398 * nfsd might hold coalesce elements! Move them to owp.
1399 * Otherwise, requests may be lost and clients will be stuck.
1400 */
1401 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) {
1402 LIST_REMOVE(m, nd_tq);
1403 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1404 }
1405 }
1406
1407 /*
1408 * nfs create service
1409 * now does a truncate to 0 length via. setattr if it already exists
1410 */
1411 int
1412 nfsrv_create(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1413 {
1414 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1415 struct mbuf *nam = nfsd->nd_nam;
1416 char *dpos = nfsd->nd_dpos;
1417 kauth_cred_t cred = nfsd->nd_cr;
1418 struct nfs_fattr *fp;
1419 struct vattr va, dirfor, diraft;
1420 struct nfsv2_sattr *sp;
1421 u_int32_t *tl;
1422 struct nameidata nd;
1423 char *cp;
1424 int32_t t1;
1425 char *bpos;
1426 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1427 int rdev = 0, abort = 0;
1428 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1429 char *cp2;
1430 struct mbuf *mb, *mreq __unused;
1431 struct vnode *vp = NULL, *dirp = NULL;
1432 nfsrvfh_t nsfh;
1433 u_quad_t frev, tempsize;
1434 u_char cverf[NFSX_V3CREATEVERF];
1435
1436 nd.ni_cnd.cn_nameiop = 0;
1437 nfsm_srvmtofh(&nsfh);
1438 nfsm_srvnamesiz(len);
1439 nd.ni_cnd.cn_cred = cred;
1440 nd.ni_cnd.cn_nameiop = CREATE;
1441 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1442 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
1443 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
1444 if (dirp && v3) {
1445 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
1446 }
1447 if (error) {
1448 nfsm_reply(NFSX_WCCDATA(v3));
1449 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1450 if (dirp)
1451 vrele(dirp);
1452 if (nd.ni_pathbuf != NULL) {
1453 pathbuf_destroy(nd.ni_pathbuf);
1454 nd.ni_pathbuf = NULL;
1455 }
1456 return (0);
1457 }
1458 abort = 1;
1459 vattr_null(&va);
1460 if (v3) {
1461 va.va_mode = 0;
1462 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1463 how = fxdr_unsigned(int, *tl);
1464 switch (how) {
1465 case NFSV3CREATE_GUARDED:
1466 if (nd.ni_vp) {
1467 error = EEXIST;
1468 break;
1469 }
1470 /* FALLTHROUGH */
1471 case NFSV3CREATE_UNCHECKED:
1472 nfsm_srvsattr(&va);
1473 break;
1474 case NFSV3CREATE_EXCLUSIVE:
1475 nfsm_dissect(cp, void *, NFSX_V3CREATEVERF);
1476 memcpy(cverf, cp, NFSX_V3CREATEVERF);
1477 exclusive_flag = 1;
1478 break;
1479 };
1480 va.va_type = VREG;
1481 } else {
1482 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1483 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1484 if (va.va_type == VNON)
1485 va.va_type = VREG;
1486 va.va_mode = nfstov_mode(sp->sa_mode);
1487 switch (va.va_type) {
1488 case VREG:
1489 tsize = fxdr_unsigned(int32_t, sp->sa_size);
1490 if (tsize != -1)
1491 va.va_size = (u_quad_t)tsize;
1492 break;
1493 case VCHR:
1494 case VBLK:
1495 case VFIFO:
1496 rdev = fxdr_unsigned(int32_t, sp->sa_size);
1497 break;
1498 default:
1499 break;
1500 };
1501 }
1502
1503 /*
1504 * Iff doesn't exist, create it
1505 * otherwise just truncate to 0 length
1506 * should I set the mode too ??
1507 */
1508 if (nd.ni_vp == NULL) {
1509 if (va.va_type == VREG || va.va_type == VSOCK) {
1510 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1511 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1512 if (!error) {
1513 vn_lock(nd.ni_vp, LK_EXCLUSIVE | LK_RETRY);
1514 if (exclusive_flag) {
1515 exclusive_flag = 0;
1516 vattr_null(&va);
1517 /*
1518 * XXX
1519 * assuming NFSX_V3CREATEVERF
1520 * == sizeof(nfstime3)
1521 */
1522 fxdr_nfsv3time(cverf, &va.va_atime);
1523 error = VOP_SETATTR(nd.ni_vp, &va,
1524 cred);
1525 }
1526 }
1527 } else if (va.va_type == VCHR || va.va_type == VBLK ||
1528 va.va_type == VFIFO) {
1529 if (va.va_type == VCHR && rdev == 0xffffffff)
1530 va.va_type = VFIFO;
1531 if (va.va_type != VFIFO &&
1532 (error = kauth_authorize_system(cred,
1533 KAUTH_SYSTEM_MKNOD, 0, NULL, NULL, NULL))) {
1534 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1535 vput(nd.ni_dvp);
1536 abort = 0;
1537 nfsm_reply(0);
1538 if (nd.ni_pathbuf != NULL) {
1539 pathbuf_destroy(nd.ni_pathbuf);
1540 nd.ni_pathbuf = NULL;
1541 }
1542 return (error);
1543 } else
1544 va.va_rdev = (dev_t)rdev;
1545 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1546 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1547 &va);
1548 if (!error) {
1549 vn_lock(nd.ni_vp, LK_EXCLUSIVE | LK_RETRY);
1550 } else {
1551 nfsm_reply(0);
1552 }
1553 } else {
1554 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1555 if (nd.ni_pathbuf != NULL) {
1556 pathbuf_destroy(nd.ni_pathbuf);
1557 nd.ni_pathbuf = NULL;
1558 }
1559 error = ENXIO;
1560 abort = 0;
1561 }
1562 vp = nd.ni_vp;
1563 } else {
1564 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1565 if (nd.ni_pathbuf != NULL) {
1566 pathbuf_destroy(nd.ni_pathbuf);
1567 nd.ni_pathbuf = NULL;
1568 }
1569 vp = nd.ni_vp;
1570 abort = 0;
1571 if (!error && va.va_size != -1) {
1572 error = nfsrv_access(vp, VWRITE, cred,
1573 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0);
1574 if (!error) {
1575 nqsrv_getl(vp, ND_WRITE);
1576 tempsize = va.va_size;
1577 vattr_null(&va);
1578 va.va_size = tempsize;
1579 error = VOP_SETATTR(vp, &va, cred);
1580 }
1581 }
1582 if (error)
1583 vput(vp);
1584 }
1585 if (!error) {
1586 error = nfsrv_composefh(vp, &nsfh, v3);
1587 if (!error)
1588 error = VOP_GETATTR(vp, &va, cred);
1589 vput(vp);
1590 }
1591 if (nd.ni_dvp == vp)
1592 vrele(nd.ni_dvp);
1593 else
1594 vput(nd.ni_dvp);
1595 if (v3) {
1596 if (exclusive_flag && !error) {
1597 /*
1598 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3)
1599 */
1600 char oldverf[NFSX_V3CREATEVERF];
1601
1602 txdr_nfsv3time(&va.va_atime, oldverf);
1603 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF))
1604 error = EEXIST;
1605 }
1606 if (dirp) {
1607 vn_lock(dirp, LK_SHARED | LK_RETRY);
1608 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
1609 VOP_UNLOCK(dirp);
1610 }
1611 }
1612 if (dirp) {
1613 vrele(dirp);
1614 dirp = NULL;
1615 }
1616 if (nd.ni_pathbuf != NULL) {
1617 pathbuf_destroy(nd.ni_pathbuf);
1618 nd.ni_pathbuf = NULL;
1619 }
1620 abort = 0;
1621 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1622 if (v3) {
1623 if (!error) {
1624 nfsm_srvpostop_fh(&nsfh);
1625 nfsm_srvpostop_attr(0, &va);
1626 }
1627 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1628 } else {
1629 nfsm_srvfhtom(&nsfh, v3);
1630 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1631 nfsm_srvfillattr(&va, fp);
1632 }
1633 return (0);
1634 nfsmout:
1635 if (dirp)
1636 vrele(dirp);
1637 if (abort) {
1638 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1639 if (nd.ni_dvp == nd.ni_vp)
1640 vrele(nd.ni_dvp);
1641 else
1642 vput(nd.ni_dvp);
1643 if (nd.ni_vp)
1644 vput(nd.ni_vp);
1645 }
1646 if (nd.ni_pathbuf != NULL) {
1647 pathbuf_destroy(nd.ni_pathbuf);
1648 nd.ni_pathbuf = NULL;
1649 }
1650 return (error);
1651 }
1652
1653 /*
1654 * nfs v3 mknod service
1655 */
1656 int
1657 nfsrv_mknod(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1658 {
1659 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1660 struct mbuf *nam = nfsd->nd_nam;
1661 char *dpos = nfsd->nd_dpos;
1662 kauth_cred_t cred = nfsd->nd_cr;
1663 struct vattr va, dirfor, diraft;
1664 u_int32_t *tl;
1665 struct nameidata nd;
1666 int32_t t1;
1667 char *bpos;
1668 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
1669 int abort = 0;
1670 u_int32_t major, minor;
1671 enum vtype vtyp;
1672 char *cp2;
1673 struct mbuf *mb, *mreq __unused;
1674 struct vnode *vp, *dirp = (struct vnode *)0;
1675 nfsrvfh_t nsfh;
1676 u_quad_t frev;
1677
1678 nd.ni_cnd.cn_nameiop = 0;
1679 nfsm_srvmtofh(&nsfh);
1680 nfsm_srvnamesiz(len);
1681 nd.ni_cnd.cn_cred = cred;
1682 nd.ni_cnd.cn_nameiop = CREATE;
1683 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1684 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
1685 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
1686 if (dirp)
1687 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
1688 if (error) {
1689 nfsm_reply(NFSX_WCCDATA(1));
1690 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1691 if (dirp)
1692 vrele(dirp);
1693 if (nd.ni_pathbuf != NULL) {
1694 pathbuf_destroy(nd.ni_pathbuf);
1695 nd.ni_pathbuf = NULL;
1696 }
1697 return (0);
1698 }
1699 abort = 1;
1700 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1701 vtyp = nfsv3tov_type(*tl);
1702 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1703 error = NFSERR_BADTYPE;
1704 goto abort;
1705 }
1706 vattr_null(&va);
1707 va.va_mode = 0;
1708 nfsm_srvsattr(&va);
1709 if (vtyp == VCHR || vtyp == VBLK) {
1710 dev_t rdev;
1711
1712 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1713 major = fxdr_unsigned(u_int32_t, *tl++);
1714 minor = fxdr_unsigned(u_int32_t, *tl);
1715 rdev = makedev(major, minor);
1716 if (major(rdev) != major || minor(rdev) != minor) {
1717 error = EINVAL;
1718 goto abort;
1719 }
1720 va.va_rdev = rdev;
1721 }
1722
1723 /*
1724 * Iff doesn't exist, create it.
1725 */
1726 if (nd.ni_vp) {
1727 error = EEXIST;
1728 abort:
1729 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1730 if (nd.ni_vp)
1731 vput(nd.ni_vp);
1732 if (nd.ni_pathbuf != NULL) {
1733 pathbuf_destroy(nd.ni_pathbuf);
1734 nd.ni_pathbuf = NULL;
1735 }
1736 goto out;
1737 }
1738 va.va_type = vtyp;
1739 if (vtyp == VSOCK) {
1740 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1741 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1742 } else {
1743 if (va.va_type != VFIFO &&
1744 (error = kauth_authorize_system(cred,
1745 KAUTH_SYSTEM_MKNOD, 0, NULL, NULL, NULL))) {
1746 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1747 vput(nd.ni_dvp);
1748 goto out;
1749 }
1750 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1751 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1752 if (error)
1753 goto out;
1754 }
1755 if (!error) {
1756 vn_lock(nd.ni_vp, LK_EXCLUSIVE | LK_RETRY);
1757 }
1758 out:
1759 vp = nd.ni_vp;
1760 if (!error) {
1761 error = nfsrv_composefh(vp, &nsfh, true);
1762 if (!error)
1763 error = VOP_GETATTR(vp, &va, cred);
1764 vput(vp);
1765 }
1766 if (nd.ni_dvp == nd.ni_vp)
1767 vrele(nd.ni_dvp);
1768 else
1769 vput(nd.ni_dvp);
1770 if (dirp) {
1771 vn_lock(dirp, LK_SHARED | LK_RETRY);
1772 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
1773 VOP_UNLOCK(dirp);
1774 vrele(dirp);
1775 dirp = NULL;
1776 }
1777 if (nd.ni_pathbuf != NULL) {
1778 pathbuf_destroy(nd.ni_pathbuf);
1779 nd.ni_pathbuf = NULL;
1780 }
1781 abort = 0;
1782 nfsm_reply(NFSX_SRVFH(&nsfh, true) + NFSX_POSTOPATTR(1) +
1783 NFSX_WCCDATA(1));
1784 if (!error) {
1785 nfsm_srvpostop_fh(&nsfh);
1786 nfsm_srvpostop_attr(0, &va);
1787 }
1788 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1789 return (0);
1790 nfsmout:
1791 if (abort) {
1792 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1793 if (nd.ni_dvp == nd.ni_vp)
1794 vrele(nd.ni_dvp);
1795 else
1796 vput(nd.ni_dvp);
1797 if (nd.ni_vp)
1798 vput(nd.ni_vp);
1799 }
1800 if (nd.ni_pathbuf != NULL) {
1801 pathbuf_destroy(nd.ni_pathbuf);
1802 nd.ni_pathbuf = NULL;
1803 }
1804 if (dirp)
1805 vrele(dirp);
1806 return (error);
1807 }
1808
1809 /*
1810 * nfs remove service
1811 */
1812 int
1813 nfsrv_remove(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1814 {
1815 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1816 struct mbuf *nam = nfsd->nd_nam;
1817 char *dpos = nfsd->nd_dpos;
1818 kauth_cred_t cred = nfsd->nd_cr;
1819 struct nameidata nd;
1820 u_int32_t *tl;
1821 int32_t t1;
1822 char *bpos;
1823 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
1824 int v3 = (nfsd->nd_flag & ND_NFSV3);
1825 char *cp2;
1826 struct mbuf *mb, *mreq __unused;
1827 struct vnode *vp, *dirp;
1828 struct vattr dirfor, diraft;
1829 nfsrvfh_t nsfh;
1830 u_quad_t frev;
1831
1832 #ifndef nolint
1833 vp = (struct vnode *)0;
1834 #endif
1835 nfsm_srvmtofh(&nsfh);
1836 nfsm_srvnamesiz(len);
1837 nd.ni_cnd.cn_cred = cred;
1838 nd.ni_cnd.cn_nameiop = DELETE;
1839 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1840 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
1841 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
1842 if (error == 0 && dirp && v3) {
1843 if (nd.ni_dvp == nd.ni_vp)
1844 vn_lock(dirp, LK_SHARED | LK_RETRY);
1845 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
1846 if (nd.ni_dvp == nd.ni_vp)
1847 VOP_UNLOCK(dirp);
1848 }
1849 if (!error) {
1850 vp = nd.ni_vp;
1851 if (vp->v_type == VDIR) {
1852 error = EPERM;
1853 goto out;
1854 }
1855 /*
1856 * The root of a mounted filesystem cannot be deleted.
1857 */
1858 if (vp->v_vflag & VV_ROOT) {
1859 error = EBUSY;
1860 }
1861 out:
1862 if (!error) {
1863 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1864 nqsrv_getl(vp, ND_WRITE);
1865 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1866 vput(nd.ni_dvp);
1867 } else {
1868 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1869 if (nd.ni_dvp == vp)
1870 vrele(nd.ni_dvp);
1871 else
1872 vput(nd.ni_dvp);
1873 vput(vp);
1874 }
1875 }
1876 if (nd.ni_pathbuf != NULL) {
1877 pathbuf_destroy(nd.ni_pathbuf);
1878 nd.ni_pathbuf = NULL;
1879 }
1880 if (dirp) {
1881 if (v3) {
1882 vn_lock(dirp, LK_SHARED | LK_RETRY);
1883 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
1884 VOP_UNLOCK(dirp);
1885 }
1886 vrele(dirp);
1887 }
1888 nfsm_reply(NFSX_WCCDATA(v3));
1889 if (v3) {
1890 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1891 return (0);
1892 }
1893 nfsm_srvdone;
1894 }
1895
1896 /*
1897 * nfs rename service
1898 */
1899 int
1900 nfsrv_rename(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1901 {
1902 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1903 struct mbuf *nam = nfsd->nd_nam;
1904 char *dpos = nfsd->nd_dpos;
1905 kauth_cred_t cred = nfsd->nd_cr;
1906 u_int32_t *tl;
1907 int32_t t1;
1908 char *bpos;
1909 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1;
1910 uint32_t len, len2;
1911 int tdirfor_ret = 1, tdiraft_ret = 1;
1912 int v3 = (nfsd->nd_flag & ND_NFSV3);
1913 char *cp2;
1914 struct mbuf *mb, *mreq __unused;
1915 struct nameidata fromnd, tond;
1916 struct vnode *fvp, *tvp, *tdvp;
1917 struct vnode *fdirp = NULL, *tdirp = NULL;
1918 struct mount *localfs = NULL;
1919 struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1920 nfsrvfh_t fnsfh, tnsfh;
1921 u_quad_t frev;
1922 uid_t saved_uid;
1923
1924 #ifndef nolint
1925 fvp = (struct vnode *)0;
1926 #endif
1927 fromnd.ni_cnd.cn_nameiop = 0;
1928 tond.ni_cnd.cn_nameiop = 0;
1929 nfsm_srvmtofh(&fnsfh);
1930 nfsm_srvnamesiz(len);
1931 /*
1932 * Remember our original uid so that we can reset cr_uid before
1933 * the second nfs_namei() call, in case it is remapped.
1934 */
1935 saved_uid = kauth_cred_geteuid(cred);
1936 fromnd.ni_cnd.cn_cred = cred;
1937 fromnd.ni_cnd.cn_nameiop = DELETE;
1938 fromnd.ni_cnd.cn_flags = LOCKPARENT;
1939 error = nfs_namei(&fromnd, &fnsfh, len, slp, nam, &md,
1940 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
1941 if (error == 0 && fdirp && v3) {
1942 if (fromnd.ni_dvp == fromnd.ni_vp)
1943 vn_lock(fdirp, LK_SHARED | LK_RETRY);
1944 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred);
1945 if (fromnd.ni_dvp == fromnd.ni_vp)
1946 VOP_UNLOCK(fdirp);
1947 }
1948 if (error) {
1949 fromnd.ni_cnd.cn_nameiop = 0;
1950 nfsm_reply(2 * NFSX_WCCDATA(v3));
1951 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1952 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1953 if (fdirp)
1954 vrele(fdirp);
1955 if (fromnd.ni_pathbuf != NULL) {
1956 pathbuf_destroy(fromnd.ni_pathbuf);
1957 }
1958 return (0);
1959 }
1960 if (fromnd.ni_dvp != fromnd.ni_vp) {
1961 VOP_UNLOCK(fromnd.ni_dvp);
1962 }
1963 fvp = fromnd.ni_vp;
1964
1965 localfs = fvp->v_mount;
1966 error = VFS_RENAMELOCK_ENTER(localfs);
1967 if (error) {
1968 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1969 vrele(fromnd.ni_dvp);
1970 vrele(fvp);
1971 goto out1;
1972 }
1973
1974 /* Copied, regrettably, from vfs_syscalls.c (q.v.) */
1975 vrele(fvp);
1976 if ((fromnd.ni_cnd.cn_namelen == 1 &&
1977 fromnd.ni_cnd.cn_nameptr[0] == '.') ||
1978 (fromnd.ni_cnd.cn_namelen == 2 &&
1979 fromnd.ni_cnd.cn_nameptr[0] == '.' &&
1980 fromnd.ni_cnd.cn_nameptr[1] == '.')) {
1981 error = EINVAL;
1982 VFS_RENAMELOCK_EXIT(localfs);
1983 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1984 vrele(fromnd.ni_dvp);
1985 goto out1;
1986 }
1987 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
1988 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0);
1989 if (error) {
1990 VOP_UNLOCK(fromnd.ni_dvp);
1991 VFS_RENAMELOCK_EXIT(localfs);
1992 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1993 vrele(fromnd.ni_dvp);
1994 goto out1;
1995 }
1996 VOP_UNLOCK(fromnd.ni_vp);
1997 if (fromnd.ni_dvp != fromnd.ni_vp)
1998 VOP_UNLOCK(fromnd.ni_dvp);
1999 fvp = fromnd.ni_vp;
2000
2001 nfsm_srvmtofh(&tnsfh);
2002 if (v3) {
2003 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
2004 len2 = fxdr_unsigned(uint32_t, *tl);
2005 /* len2 will be checked by nfs_namei */
2006 }
2007 else {
2008 /* NFSv2 */
2009 nfsm_strsiz(len2, NFS_MAXNAMLEN);
2010 }
2011 kauth_cred_seteuid(cred, saved_uid);
2012 tond.ni_cnd.cn_cred = cred;
2013 tond.ni_cnd.cn_nameiop = RENAME;
2014 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE;
2015 error = nfs_namei(&tond, &tnsfh, len2, slp, nam, &md,
2016 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2017 if (tdirp && v3) {
2018 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred);
2019 }
2020 if (error) {
2021 VFS_RENAMELOCK_EXIT(localfs);
2022 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2023 vrele(fromnd.ni_dvp);
2024 vrele(fvp);
2025 goto out1;
2026 }
2027 tdvp = tond.ni_dvp;
2028 tvp = tond.ni_vp;
2029 if (tvp != NULL) {
2030 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2031 if (v3)
2032 error = EEXIST;
2033 else
2034 error = EISDIR;
2035 goto out;
2036 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2037 if (v3)
2038 error = EEXIST;
2039 else
2040 error = ENOTDIR;
2041 goto out;
2042 }
2043 if (tvp->v_type == VDIR && tvp->v_mountedhere) {
2044 if (v3)
2045 error = EXDEV;
2046 else
2047 error = ENOTEMPTY;
2048 goto out;
2049 }
2050 }
2051 if (fvp->v_type == VDIR && fvp->v_mountedhere) {
2052 if (v3)
2053 error = EXDEV;
2054 else
2055 error = ENOTEMPTY;
2056 goto out;
2057 }
2058 if (fvp->v_mount != tdvp->v_mount) {
2059 if (v3)
2060 error = EXDEV;
2061 else
2062 error = ENOTEMPTY;
2063 goto out;
2064 }
2065 if (fvp == tdvp) {
2066 if (v3)
2067 error = EINVAL;
2068 else
2069 error = ENOTEMPTY;
2070 }
2071 /*
2072 * If source is the same as the destination (that is the
2073 * same vnode with the same name in the same directory),
2074 * then there is nothing to do.
2075 */
2076 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
2077 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2078 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
2079 fromnd.ni_cnd.cn_namelen))
2080 error = -1;
2081 out:
2082 if (!error) {
2083 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
2084 nqsrv_getl(tdvp, ND_WRITE);
2085 if (tvp) {
2086 nqsrv_getl(tvp, ND_WRITE);
2087 }
2088 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2089 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2090 VFS_RENAMELOCK_EXIT(localfs);
2091 } else {
2092 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2093 if (tdvp == tvp)
2094 vrele(tdvp);
2095 else
2096 vput(tdvp);
2097 if (tvp)
2098 vput(tvp);
2099 VFS_RENAMELOCK_EXIT(localfs);
2100 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2101 vrele(fromnd.ni_dvp);
2102 vrele(fvp);
2103 if (error == -1)
2104 error = 0;
2105 }
2106 if (tond.ni_pathbuf != NULL) {
2107 pathbuf_destroy(tond.ni_pathbuf);
2108 tond.ni_pathbuf = NULL;
2109 }
2110 tond.ni_cnd.cn_nameiop = 0;
2111 out1:
2112 if (fdirp) {
2113 if (v3) {
2114 vn_lock(fdirp, LK_SHARED | LK_RETRY);
2115 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred);
2116 VOP_UNLOCK(fdirp);
2117 }
2118 vrele(fdirp);
2119 fdirp = NULL;
2120 }
2121 if (tdirp) {
2122 if (v3) {
2123 vn_lock(tdirp, LK_SHARED | LK_RETRY);
2124 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred);
2125 VOP_UNLOCK(tdirp);
2126 }
2127 vrele(tdirp);
2128 tdirp = NULL;
2129 }
2130 pathbuf_destroy(fromnd.ni_pathbuf);
2131 fromnd.ni_pathbuf = NULL;
2132 fromnd.ni_cnd.cn_nameiop = 0;
2133 localfs = NULL;
2134 nfsm_reply(2 * NFSX_WCCDATA(v3));
2135 if (v3) {
2136 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
2137 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
2138 }
2139 return (0);
2140
2141 nfsmout:
2142 if (fdirp)
2143 vrele(fdirp);
2144 #ifdef notdef
2145 if (tdirp)
2146 vrele(tdirp);
2147 #endif
2148 if (tond.ni_cnd.cn_nameiop) {
2149 if (tond.ni_pathbuf != NULL) {
2150 pathbuf_destroy(tond.ni_pathbuf);
2151 tond.ni_pathbuf = NULL;
2152 }
2153 }
2154 if (localfs) {
2155 VFS_RENAMELOCK_EXIT(localfs);
2156 }
2157 if (fromnd.ni_cnd.cn_nameiop) {
2158 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2159 if (fromnd.ni_pathbuf != NULL) {
2160 pathbuf_destroy(fromnd.ni_pathbuf);
2161 fromnd.ni_pathbuf = NULL;
2162 }
2163 vrele(fromnd.ni_dvp);
2164 vrele(fvp);
2165 }
2166 return (error);
2167 }
2168
2169 /*
2170 * nfs link service
2171 */
2172 int
2173 nfsrv_link(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2174 {
2175 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2176 struct mbuf *nam = nfsd->nd_nam;
2177 char *dpos = nfsd->nd_dpos;
2178 kauth_cred_t cred = nfsd->nd_cr;
2179 struct nameidata nd;
2180 u_int32_t *tl;
2181 int32_t t1;
2182 char *bpos;
2183 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2184 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
2185 char *cp2;
2186 struct mbuf *mb, *mreq __unused;
2187 struct vnode *vp, *xp, *dirp = (struct vnode *)0;
2188 struct vattr dirfor, diraft, at;
2189 nfsrvfh_t nsfh, dnsfh;
2190 u_quad_t frev;
2191
2192 nfsm_srvmtofh(&nsfh);
2193 nfsm_srvmtofh(&dnsfh);
2194 nfsm_srvnamesiz(len);
2195 error = nfsrv_fhtovp(&nsfh, false, &vp, cred, slp, nam,
2196 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
2197 if (error) {
2198 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2199 nfsm_srvpostop_attr(getret, &at);
2200 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2201 return (0);
2202 }
2203 if (vp->v_type == VDIR) {
2204 error = EPERM;
2205 goto out1;
2206 }
2207 nd.ni_cnd.cn_cred = cred;
2208 nd.ni_cnd.cn_nameiop = CREATE;
2209 nd.ni_cnd.cn_flags = LOCKPARENT;
2210 error = nfs_namei(&nd, &dnsfh, len, slp, nam, &md, &dpos,
2211 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2212 if (dirp && v3) {
2213 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
2214 }
2215 if (error)
2216 goto out1;
2217 xp = nd.ni_vp;
2218 if (xp != NULL) {
2219 error = EEXIST;
2220 goto out;
2221 }
2222 xp = nd.ni_dvp;
2223 if (vp->v_mount != xp->v_mount)
2224 error = EXDEV;
2225 out:
2226 if (!error) {
2227 nqsrv_getl(vp, ND_WRITE);
2228 nqsrv_getl(xp, ND_WRITE);
2229 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2230 if (nd.ni_dvp != nd.ni_vp)
2231 VOP_UNLOCK(nd.ni_dvp);
2232 vrele(nd.ni_dvp);
2233 } else {
2234 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2235 if (nd.ni_dvp == nd.ni_vp)
2236 vrele(nd.ni_dvp);
2237 else
2238 vput(nd.ni_dvp);
2239 if (nd.ni_vp)
2240 vrele(nd.ni_vp);
2241 }
2242 out1:
2243 if (v3) {
2244 vn_lock(vp, LK_SHARED | LK_RETRY);
2245 getret = VOP_GETATTR(vp, &at, cred);
2246 VOP_UNLOCK(vp);
2247 }
2248 if (dirp) {
2249 if (v3) {
2250 vn_lock(dirp, LK_SHARED | LK_RETRY);
2251 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
2252 VOP_UNLOCK(dirp);
2253 }
2254 vrele(dirp);
2255 }
2256 vrele(vp);
2257 if (nd.ni_pathbuf != NULL) {
2258 pathbuf_destroy(nd.ni_pathbuf);
2259 nd.ni_pathbuf = NULL;
2260 }
2261 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2262 if (v3) {
2263 nfsm_srvpostop_attr(getret, &at);
2264 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2265 return (0);
2266 }
2267 nfsm_srvdone;
2268 }
2269
2270 /*
2271 * nfs symbolic link service
2272 */
2273 int
2274 nfsrv_symlink(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2275 {
2276 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2277 struct mbuf *nam = nfsd->nd_nam;
2278 char *dpos = nfsd->nd_dpos;
2279 kauth_cred_t cred = nfsd->nd_cr;
2280 struct vattr va, dirfor, diraft;
2281 struct nameidata nd;
2282 u_int32_t *tl;
2283 int32_t t1;
2284 struct nfsv2_sattr *sp;
2285 char *bpos, *pathcp = NULL, *cp2;
2286 struct uio io;
2287 struct iovec iv;
2288 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1, abort = 0;
2289 uint32_t len, len2;
2290 int v3 = (nfsd->nd_flag & ND_NFSV3);
2291 struct mbuf *mb, *mreq __unused;
2292 struct vnode *dirp = (struct vnode *)0;
2293 nfsrvfh_t nsfh;
2294 u_quad_t frev;
2295
2296 nd.ni_cnd.cn_nameiop = 0;
2297 nfsm_srvmtofh(&nsfh);
2298 nfsm_srvnamesiz(len);
2299 nd.ni_cnd.cn_cred = cred;
2300 nd.ni_cnd.cn_nameiop = CREATE;
2301 nd.ni_cnd.cn_flags = LOCKPARENT;
2302 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
2303 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2304 if (dirp && v3) {
2305 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
2306 }
2307 if (error)
2308 goto out;
2309 abort = 1;
2310 vattr_null(&va);
2311 va.va_type = VLNK;
2312 if (v3) {
2313 va.va_mode = 0;
2314 nfsm_srvsattr(&va);
2315 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
2316 len2 = fxdr_unsigned(uint32_t, *tl);
2317 if (len2 > PATH_MAX) {
2318 /* XXX should check _PC_NO_TRUNC */
2319 error = ENAMETOOLONG;
2320 goto abortop;
2321 }
2322 }
2323 else {
2324 /* NFSv2 */
2325 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2326 }
2327 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK);
2328 iv.iov_base = pathcp;
2329 iv.iov_len = len2;
2330 io.uio_resid = len2;
2331 io.uio_offset = 0;
2332 io.uio_iov = &iv;
2333 io.uio_iovcnt = 1;
2334 io.uio_rw = UIO_READ;
2335 UIO_SETUP_SYSSPACE(&io);
2336 nfsm_mtouio(&io, len2);
2337 if (!v3) {
2338 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2339 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2340 }
2341 *(pathcp + len2) = '\0';
2342 if (nd.ni_vp) {
2343 error = EEXIST;
2344 abortop:
2345 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2346 if (nd.ni_dvp == nd.ni_vp)
2347 vrele(nd.ni_dvp);
2348 else
2349 vput(nd.ni_dvp);
2350 if (nd.ni_vp)
2351 vrele(nd.ni_vp);
2352 goto out;
2353 }
2354 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2355 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2356 if (!error) {
2357 if (v3) {
2358 vn_lock(nd.ni_vp, LK_SHARED | LK_RETRY);
2359 error = nfsrv_composefh(nd.ni_vp, &nsfh, v3);
2360 if (!error)
2361 error = VOP_GETATTR(nd.ni_vp, &va, cred);
2362 vput(nd.ni_vp);
2363 } else {
2364 vrele(nd.ni_vp);
2365 }
2366 }
2367 vput(nd.ni_dvp);
2368 out:
2369 if (pathcp)
2370 free(pathcp, M_TEMP);
2371 if (dirp) {
2372 if (v3) {
2373 vn_lock(dirp, LK_SHARED | LK_RETRY);
2374 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
2375 VOP_UNLOCK(dirp);
2376 }
2377 vrele(dirp);
2378 dirp = NULL;
2379 }
2380 if (nd.ni_pathbuf != NULL) {
2381 pathbuf_destroy(nd.ni_pathbuf);
2382 nd.ni_pathbuf = NULL;
2383 }
2384 abort = 0;
2385 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) +
2386 NFSX_WCCDATA(v3));
2387 if (v3) {
2388 if (!error) {
2389 nfsm_srvpostop_fh(&nsfh);
2390 nfsm_srvpostop_attr(0, &va);
2391 }
2392 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2393 }
2394 return (0);
2395 nfsmout:
2396 if (abort) {
2397 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2398 if (nd.ni_dvp == nd.ni_vp)
2399 vrele(nd.ni_dvp);
2400 else
2401 vput(nd.ni_dvp);
2402 if (nd.ni_vp)
2403 vrele(nd.ni_vp);
2404 if (nd.ni_pathbuf != NULL) {
2405 pathbuf_destroy(nd.ni_pathbuf);
2406 nd.ni_pathbuf = NULL;
2407 }
2408 }
2409 if (dirp)
2410 vrele(dirp);
2411 if (pathcp)
2412 free(pathcp, M_TEMP);
2413 return (error);
2414 }
2415
2416 /*
2417 * nfs mkdir service
2418 */
2419 int
2420 nfsrv_mkdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2421 {
2422 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2423 struct mbuf *nam = nfsd->nd_nam;
2424 char *dpos = nfsd->nd_dpos;
2425 kauth_cred_t cred = nfsd->nd_cr;
2426 struct vattr va, dirfor, diraft;
2427 struct nfs_fattr *fp;
2428 struct nameidata nd;
2429 char *cp;
2430 u_int32_t *tl;
2431 int32_t t1;
2432 char *bpos;
2433 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2434 int abort = 0;
2435 int v3 = (nfsd->nd_flag & ND_NFSV3);
2436 char *cp2;
2437 struct mbuf *mb, *mreq __unused;
2438 struct vnode *vp, *dirp = (struct vnode *)0;
2439 nfsrvfh_t nsfh;
2440 u_quad_t frev;
2441
2442 nfsm_srvmtofh(&nsfh);
2443 nfsm_srvnamesiz(len);
2444 nd.ni_cnd.cn_cred = cred;
2445 nd.ni_cnd.cn_nameiop = CREATE;
2446 nd.ni_cnd.cn_flags = LOCKPARENT;
2447 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
2448 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2449 if (dirp && v3) {
2450 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
2451 }
2452 if (error) {
2453 if (nd.ni_pathbuf != NULL) {
2454 pathbuf_destroy(nd.ni_pathbuf);
2455 nd.ni_pathbuf = NULL;
2456 }
2457 nfsm_reply(NFSX_WCCDATA(v3));
2458 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2459 if (dirp)
2460 vrele(dirp);
2461 return (0);
2462 }
2463 abort = 1;
2464 vattr_null(&va);
2465 if (v3) {
2466 va.va_mode = 0;
2467 nfsm_srvsattr(&va);
2468 } else {
2469 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2470 va.va_mode = nfstov_mode(*tl++);
2471 }
2472 va.va_type = VDIR;
2473 vp = nd.ni_vp;
2474 if (vp != NULL) {
2475 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2476 if (nd.ni_dvp == vp)
2477 vrele(nd.ni_dvp);
2478 else
2479 vput(nd.ni_dvp);
2480 vrele(vp);
2481 error = EEXIST;
2482 goto out;
2483 }
2484 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2485 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2486 if (!error) {
2487 vp = nd.ni_vp;
2488 vn_lock(vp, LK_SHARED | LK_RETRY);
2489 error = nfsrv_composefh(vp, &nsfh, v3);
2490 if (!error)
2491 error = VOP_GETATTR(vp, &va, cred);
2492 vput(vp);
2493 }
2494 vput(nd.ni_dvp);
2495 out:
2496 if (dirp) {
2497 if (v3) {
2498 vn_lock(dirp, LK_SHARED | LK_RETRY);
2499 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
2500 VOP_UNLOCK(dirp);
2501 }
2502 vrele(dirp);
2503 dirp = NULL;
2504 }
2505 if (nd.ni_pathbuf != NULL) {
2506 pathbuf_destroy(nd.ni_pathbuf);
2507 nd.ni_pathbuf = NULL;
2508 }
2509 abort = 0;
2510 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) +
2511 NFSX_WCCDATA(v3));
2512 if (v3) {
2513 if (!error) {
2514 nfsm_srvpostop_fh(&nsfh);
2515 nfsm_srvpostop_attr(0, &va);
2516 }
2517 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2518 } else {
2519 nfsm_srvfhtom(&nsfh, v3);
2520 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2521 nfsm_srvfillattr(&va, fp);
2522 }
2523 return (0);
2524 nfsmout:
2525 if (abort) {
2526 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2527 if (nd.ni_dvp == nd.ni_vp)
2528 vrele(nd.ni_dvp);
2529 else
2530 vput(nd.ni_dvp);
2531 if (nd.ni_vp)
2532 vrele(nd.ni_vp);
2533 if (nd.ni_pathbuf != NULL) {
2534 pathbuf_destroy(nd.ni_pathbuf);
2535 nd.ni_pathbuf = NULL;
2536 }
2537 }
2538 if (dirp)
2539 vrele(dirp);
2540 return (error);
2541 }
2542
2543 /*
2544 * nfs rmdir service
2545 */
2546 int
2547 nfsrv_rmdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2548 {
2549 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2550 struct mbuf *nam = nfsd->nd_nam;
2551 char *dpos = nfsd->nd_dpos;
2552 kauth_cred_t cred = nfsd->nd_cr;
2553 u_int32_t *tl;
2554 int32_t t1;
2555 char *bpos;
2556 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2557 int v3 = (nfsd->nd_flag & ND_NFSV3);
2558 char *cp2;
2559 struct mbuf *mb, *mreq __unused;
2560 struct vnode *vp, *dirp = (struct vnode *)0;
2561 struct vattr dirfor, diraft;
2562 nfsrvfh_t nsfh;
2563 struct nameidata nd;
2564 u_quad_t frev;
2565
2566 nfsm_srvmtofh(&nsfh);
2567 nfsm_srvnamesiz(len);
2568 nd.ni_cnd.cn_cred = cred;
2569 nd.ni_cnd.cn_nameiop = DELETE;
2570 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2571 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
2572 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2573 if (dirp && v3) {
2574 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
2575 }
2576 if (error) {
2577 if (nd.ni_pathbuf != NULL) {
2578 pathbuf_destroy(nd.ni_pathbuf);
2579 nd.ni_pathbuf = NULL;
2580 }
2581 nfsm_reply(NFSX_WCCDATA(v3));
2582 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2583 if (dirp)
2584 vrele(dirp);
2585 return (0);
2586 }
2587 vp = nd.ni_vp;
2588 if (vp->v_type != VDIR) {
2589 error = ENOTDIR;
2590 goto out;
2591 }
2592 /*
2593 * No rmdir "." please.
2594 */
2595 if (nd.ni_dvp == vp) {
2596 error = EINVAL;
2597 goto out;
2598 }
2599 /*
2600 * The root of a mounted filesystem cannot be deleted.
2601 */
2602 if (vp->v_vflag & VV_ROOT)
2603 error = EBUSY;
2604 out:
2605 if (!error) {
2606 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2607 nqsrv_getl(vp, ND_WRITE);
2608 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2609 vput(nd.ni_dvp);
2610 } else {
2611 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2612 if (nd.ni_dvp == nd.ni_vp)
2613 vrele(nd.ni_dvp);
2614 else
2615 vput(nd.ni_dvp);
2616 vput(vp);
2617 }
2618 if (nd.ni_pathbuf != NULL) {
2619 pathbuf_destroy(nd.ni_pathbuf);
2620 nd.ni_pathbuf = NULL;
2621 }
2622 if (dirp) {
2623 if (v3) {
2624 vn_lock(dirp, LK_SHARED | LK_RETRY);
2625 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
2626 VOP_UNLOCK(dirp);
2627 }
2628 vrele(dirp);
2629 }
2630 nfsm_reply(NFSX_WCCDATA(v3));
2631 if (v3) {
2632 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2633 return (0);
2634 }
2635 nfsm_srvdone;
2636 }
2637
2638 /*
2639 * nfs readdir service
2640 * - mallocs what it thinks is enough to read
2641 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR
2642 * - calls VOP_READDIR()
2643 * - loops around building the reply
2644 * if the output generated exceeds count break out of loop
2645 * The nfsm_clget macro is used here so that the reply will be packed
2646 * tightly in mbuf clusters.
2647 * - it only knows that it has encountered eof when the VOP_READDIR()
2648 * reads nothing
2649 * - as such one readdir rpc will return eof false although you are there
2650 * and then the next will return eof
2651 * - it trims out records with d_fileno == 0
2652 * this doesn't matter for Unix clients, but they might confuse clients
2653 * for other os'.
2654 * - it trims out records with d_type == DT_WHT
2655 * these cannot be seen through NFS (unless we extend the protocol)
2656 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2657 * than requested, but this may not apply to all filesystems. For
2658 * example, client NFS does not { although it is never remote mounted
2659 * anyhow }
2660 * The alternate call nfsrv_readdirplus() does lookups as well.
2661 * PS: The NFS protocol spec. does not clarify what the "count" byte
2662 * argument is a count of.. just name strings and file id's or the
2663 * entire reply rpc or ...
2664 * I tried just file name and id sizes and it confused the Sun client,
2665 * so I am using the full rpc size now. The "paranoia.." comment refers
2666 * to including the status longwords that are not a part of the dir.
2667 * "entry" structures, but are in the rpc.
2668 */
2669
2670 #define NFS_SRVDIRBLKSIZ 1024
2671
2672 struct flrep {
2673 nfsuint64 fl_off;
2674 u_int32_t fl_postopok;
2675 struct nfs_fattr fl_fattr; /* XXX: must be of fattr3 size */
2676 u_int32_t fl_fhok;
2677 u_int32_t fl_fhsize;
2678 /* handle comes here, filled in dynamically */
2679 };
2680
2681 int
2682 nfsrv_readdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2683 {
2684 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2685 struct mbuf *nam = nfsd->nd_nam;
2686 char *dpos = nfsd->nd_dpos;
2687 kauth_cred_t cred = nfsd->nd_cr;
2688 char *bp, *be;
2689 struct mbuf *mp;
2690 struct dirent *dp;
2691 char *cp;
2692 u_int32_t *tl;
2693 int32_t t1;
2694 char *bpos;
2695 struct mbuf *mb, *mreq __unused, *mp2;
2696 char *cpos, *cend, *cp2, *rbuf;
2697 struct vnode *vp;
2698 struct vattr at;
2699 nfsrvfh_t nsfh;
2700 struct uio io;
2701 struct iovec iv;
2702 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2703 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies;
2704 int v3 = (nfsd->nd_flag & ND_NFSV3);
2705 u_quad_t frev, off, toff;
2706 #ifdef NFS3_STRICTVERF
2707 u_quad_t verf;
2708 #endif
2709 off_t *cookies = NULL, *cookiep;
2710 nfsuint64 jar;
2711
2712 nfsm_srvmtofh(&nsfh);
2713 if (v3) {
2714 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2715 toff = fxdr_hyper(tl);
2716 tl += 2;
2717 #ifdef NFS3_STRICTVERF
2718 verf = fxdr_hyper(tl);
2719 #endif
2720 tl += 2;
2721 } else {
2722 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2723 toff = fxdr_unsigned(u_quad_t, *tl++);
2724 }
2725 off = toff;
2726 cnt = fxdr_unsigned(int, *tl);
2727 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1));
2728 xfer = NFS_SRVMAXDATA(nfsd);
2729 if (siz > xfer)
2730 siz = xfer;
2731 fullsiz = siz;
2732 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
2733 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
2734 if (!error && vp->v_type != VDIR) {
2735 error = ENOTDIR;
2736 vput(vp);
2737 }
2738 if (error) {
2739 nfsm_reply(NFSX_UNSIGNED);
2740 nfsm_srvpostop_attr(getret, &at);
2741 return (0);
2742 }
2743 nqsrv_getl(vp, ND_READ);
2744 if (v3) {
2745 error = getret = VOP_GETATTR(vp, &at, cred);
2746 #ifdef NFS3_STRICTVERF
2747 /*
2748 * XXX This check is too strict for Solaris 2.5 clients.
2749 */
2750 if (!error && toff && verf != at.va_filerev)
2751 error = NFSERR_BAD_COOKIE;
2752 #endif
2753 }
2754 if (!error)
2755 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0);
2756 if (error) {
2757 vput(vp);
2758 nfsm_reply(NFSX_POSTOPATTR(v3));
2759 nfsm_srvpostop_attr(getret, &at);
2760 return (0);
2761 }
2762 VOP_UNLOCK(vp);
2763 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2764 again:
2765 iv.iov_base = rbuf;
2766 iv.iov_len = fullsiz;
2767 io.uio_iov = &iv;
2768 io.uio_iovcnt = 1;
2769 io.uio_offset = (off_t)off;
2770 io.uio_resid = fullsiz;
2771 io.uio_rw = UIO_READ;
2772 UIO_SETUP_SYSSPACE(&io);
2773 eofflag = 0;
2774 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2775
2776 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2777
2778 off = (off_t)io.uio_offset;
2779 if (!cookies && !error)
2780 error = NFSERR_PERM;
2781 if (v3) {
2782 getret = VOP_GETATTR(vp, &at, cred);
2783 if (!error)
2784 error = getret;
2785 }
2786
2787 VOP_UNLOCK(vp);
2788 if (error) {
2789 vrele(vp);
2790 free((void *)rbuf, M_TEMP);
2791 if (cookies)
2792 free((void *)cookies, M_TEMP);
2793 nfsm_reply(NFSX_POSTOPATTR(v3));
2794 nfsm_srvpostop_attr(getret, &at);
2795 return (0);
2796 }
2797 if (io.uio_resid) {
2798 siz -= io.uio_resid;
2799
2800 /*
2801 * If nothing read, return eof
2802 * rpc reply
2803 */
2804 if (siz == 0) {
2805 vrele(vp);
2806 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2807 2 * NFSX_UNSIGNED);
2808 if (v3) {
2809 nfsm_srvpostop_attr(getret, &at);
2810 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2811 txdr_hyper(at.va_filerev, tl);
2812 tl += 2;
2813 } else
2814 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2815 *tl++ = nfs_false;
2816 *tl = nfs_true;
2817 free((void *)rbuf, M_TEMP);
2818 free((void *)cookies, M_TEMP);
2819 return (0);
2820 }
2821 }
2822
2823 /*
2824 * Check for degenerate cases of nothing useful read.
2825 * If so go try again
2826 */
2827 cpos = rbuf;
2828 cend = rbuf + siz;
2829 dp = (struct dirent *)cpos;
2830 cookiep = cookies;
2831
2832 while (cpos < cend && ncookies > 0 &&
2833 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2834 cpos += dp->d_reclen;
2835 dp = (struct dirent *)cpos;
2836 cookiep++;
2837 ncookies--;
2838 }
2839 if (cpos >= cend || ncookies == 0) {
2840 toff = off;
2841 siz = fullsiz;
2842 free(cookies, M_TEMP);
2843 cookies = NULL;
2844 goto again;
2845 }
2846
2847 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
2848 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2849 if (v3) {
2850 nfsm_srvpostop_attr(getret, &at);
2851 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2852 txdr_hyper(at.va_filerev, tl);
2853 }
2854 mp = mp2 = mb;
2855 bp = bpos;
2856 be = bp + M_TRAILINGSPACE(mp);
2857
2858 /* Loop through the records and build reply */
2859 while (cpos < cend && ncookies > 0) {
2860 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2861 nlen = dp->d_namlen;
2862 rem = nfsm_rndup(nlen)-nlen;
2863 len += (4 * NFSX_UNSIGNED + nlen + rem);
2864 if (v3)
2865 len += 2 * NFSX_UNSIGNED;
2866 if (len > cnt) {
2867 eofflag = 0;
2868 break;
2869 }
2870 /*
2871 * Build the directory record xdr from
2872 * the dirent entry.
2873 */
2874 nfsm_clget;
2875 *tl = nfs_true;
2876 bp += NFSX_UNSIGNED;
2877 if (v3) {
2878 nfsm_clget;
2879 *tl = txdr_unsigned(dp->d_fileno >> 32);
2880 bp += NFSX_UNSIGNED;
2881 }
2882 nfsm_clget;
2883 *tl = txdr_unsigned(dp->d_fileno);
2884 bp += NFSX_UNSIGNED;
2885 nfsm_clget;
2886 *tl = txdr_unsigned(nlen);
2887 bp += NFSX_UNSIGNED;
2888
2889 /* And loop around copying the name */
2890 xfer = nlen;
2891 cp = dp->d_name;
2892 while (xfer > 0) {
2893 nfsm_clget;
2894 if ((bp+xfer) > be)
2895 tsiz = be-bp;
2896 else
2897 tsiz = xfer;
2898 memcpy(bp, cp, tsiz);
2899 bp += tsiz;
2900 xfer -= tsiz;
2901 if (xfer > 0)
2902 cp += tsiz;
2903 }
2904 /* And null pad to an int32_t boundary */
2905 for (i = 0; i < rem; i++)
2906 *bp++ = '\0';
2907 nfsm_clget;
2908
2909 /* Finish off the record */
2910 txdr_hyper(*cookiep, &jar);
2911 if (v3) {
2912 *tl = jar.nfsuquad[0];
2913 bp += NFSX_UNSIGNED;
2914 nfsm_clget;
2915 }
2916 *tl = jar.nfsuquad[1];
2917 bp += NFSX_UNSIGNED;
2918 }
2919 cpos += dp->d_reclen;
2920 dp = (struct dirent *)cpos;
2921 cookiep++;
2922 ncookies--;
2923 }
2924 vrele(vp);
2925 nfsm_clget;
2926 *tl = nfs_false;
2927 bp += NFSX_UNSIGNED;
2928 nfsm_clget;
2929 if (eofflag)
2930 *tl = nfs_true;
2931 else
2932 *tl = nfs_false;
2933 bp += NFSX_UNSIGNED;
2934 if (mp != mb) {
2935 if (bp < be)
2936 mp->m_len = bp - mtod(mp, char *);
2937 } else
2938 mp->m_len += bp - bpos;
2939 free((void *)rbuf, M_TEMP);
2940 free((void *)cookies, M_TEMP);
2941 nfsm_srvdone;
2942 }
2943
2944 int
2945 nfsrv_readdirplus(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2946 {
2947 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2948 struct mbuf *nam = nfsd->nd_nam;
2949 char *dpos = nfsd->nd_dpos;
2950 kauth_cred_t cred = nfsd->nd_cr;
2951 char *bp, *be;
2952 struct mbuf *mp;
2953 struct dirent *dp;
2954 char *cp;
2955 u_int32_t *tl;
2956 int32_t t1;
2957 char *bpos;
2958 struct mbuf *mb, *mreq __unused, *mp2;
2959 char *cpos, *cend, *cp2, *rbuf;
2960 struct vnode *vp, *nvp;
2961 struct flrep fl;
2962 nfsrvfh_t nsfh;
2963 struct uio io;
2964 struct iovec iv;
2965 struct vattr va, at, *vap = &va;
2966 struct nfs_fattr *fp;
2967 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2968 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies;
2969 u_quad_t frev, off, toff;
2970 #ifdef NFS3_STRICTVERF
2971 u_quad_t verf;
2972 #endif
2973 off_t *cookies = NULL, *cookiep;
2974
2975 nfsm_srvmtofh(&nsfh);
2976 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2977 toff = fxdr_hyper(tl);
2978 tl += 2;
2979 #ifdef NFS3_STRICTVERF
2980 verf = fxdr_hyper(tl);
2981 #endif
2982 tl += 2;
2983 siz = fxdr_unsigned(int, *tl++);
2984 cnt = fxdr_unsigned(int, *tl);
2985 off = toff;
2986 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1));
2987 xfer = NFS_SRVMAXDATA(nfsd);
2988 if (siz > xfer)
2989 siz = xfer;
2990 fullsiz = siz;
2991 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
2992 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
2993 if (!error && vp->v_type != VDIR) {
2994 error = ENOTDIR;
2995 vput(vp);
2996 }
2997 if (error) {
2998 nfsm_reply(NFSX_UNSIGNED);
2999 nfsm_srvpostop_attr(getret, &at);
3000 return (0);
3001 }
3002 error = getret = VOP_GETATTR(vp, &at, cred);
3003 #ifdef NFS3_STRICTVERF
3004 /*
3005 * XXX This check is too strict for Solaris 2.5 clients.
3006 */
3007 if (!error && toff && verf != at.va_filerev)
3008 error = NFSERR_BAD_COOKIE;
3009 #endif
3010 if (!error) {
3011 nqsrv_getl(vp, ND_READ);
3012 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0);
3013 }
3014 if (error) {
3015 vput(vp);
3016 nfsm_reply(NFSX_V3POSTOPATTR);
3017 nfsm_srvpostop_attr(getret, &at);
3018 return (0);
3019 }
3020 VOP_UNLOCK(vp);
3021
3022 rbuf = malloc(siz, M_TEMP, M_WAITOK);
3023 again:
3024 iv.iov_base = rbuf;
3025 iv.iov_len = fullsiz;
3026 io.uio_iov = &iv;
3027 io.uio_iovcnt = 1;
3028 io.uio_offset = (off_t)off;
3029 io.uio_resid = fullsiz;
3030 io.uio_rw = UIO_READ;
3031 UIO_SETUP_SYSSPACE(&io);
3032 eofflag = 0;
3033
3034 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3035
3036 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
3037
3038 off = (u_quad_t)io.uio_offset;
3039 getret = VOP_GETATTR(vp, &at, cred);
3040
3041 VOP_UNLOCK(vp);
3042
3043 /*
3044 * If the VGET operation doesn't work for this filesystem,
3045 * we can't support readdirplus. Returning NOTSUPP should
3046 * make clients fall back to plain readdir.
3047 * There's no need to check for VPTOFH as well, we wouldn't
3048 * even be here otherwise.
3049 */
3050 if (!getret) {
3051 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
3052 getret = (getret == EOPNOTSUPP) ?
3053 NFSERR_NOTSUPP : NFSERR_IO;
3054 else
3055 vput(nvp);
3056 }
3057
3058 if (!cookies && !error)
3059 error = NFSERR_PERM;
3060 if (!error)
3061 error = getret;
3062 if (error) {
3063 vrele(vp);
3064 if (cookies)
3065 free((void *)cookies, M_TEMP);
3066 free((void *)rbuf, M_TEMP);
3067 nfsm_reply(NFSX_V3POSTOPATTR);
3068 nfsm_srvpostop_attr(getret, &at);
3069 return (0);
3070 }
3071 if (io.uio_resid) {
3072 siz -= io.uio_resid;
3073
3074 /*
3075 * If nothing read, return eof
3076 * rpc reply
3077 */
3078 if (siz == 0) {
3079 vrele(vp);
3080 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
3081 2 * NFSX_UNSIGNED);
3082 nfsm_srvpostop_attr(getret, &at);
3083 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3084 txdr_hyper(at.va_filerev, tl);
3085 tl += 2;
3086 *tl++ = nfs_false;
3087 *tl = nfs_true;
3088 free((void *)cookies, M_TEMP);
3089 free((void *)rbuf, M_TEMP);
3090 return (0);
3091 }
3092 }
3093
3094 /*
3095 * Check for degenerate cases of nothing useful read.
3096 * If so go try again
3097 */
3098 cpos = rbuf;
3099 cend = rbuf + siz;
3100 dp = (struct dirent *)cpos;
3101 cookiep = cookies;
3102
3103 while (cpos < cend && ncookies > 0 &&
3104 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
3105 cpos += dp->d_reclen;
3106 dp = (struct dirent *)cpos;
3107 cookiep++;
3108 ncookies--;
3109 }
3110 if (cpos >= cend || ncookies == 0) {
3111 toff = off;
3112 siz = fullsiz;
3113 free(cookies, M_TEMP);
3114 cookies = NULL;
3115 goto again;
3116 }
3117
3118 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
3119 nfsm_reply(cnt);
3120 nfsm_srvpostop_attr(getret, &at);
3121 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3122 txdr_hyper(at.va_filerev, tl);
3123 mp = mp2 = mb;
3124 bp = bpos;
3125 be = bp + M_TRAILINGSPACE(mp);
3126
3127 /* Loop through the records and build reply */
3128 while (cpos < cend && ncookies > 0) {
3129 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
3130 nfsrvfh_t nnsfh;
3131
3132 nlen = dp->d_namlen;
3133 rem = nfsm_rndup(nlen)-nlen;
3134
3135 /*
3136 * For readdir_and_lookup get the vnode using
3137 * the file number.
3138 */
3139 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
3140 goto invalid;
3141 if (nfsrv_composefh(nvp, &nnsfh, true)) {
3142 vput(nvp);
3143 goto invalid;
3144 }
3145 if (VOP_GETATTR(nvp, vap, cred)) {
3146 vput(nvp);
3147 goto invalid;
3148 }
3149 vput(nvp);
3150
3151 /*
3152 * If either the dircount or maxcount will be
3153 * exceeded, get out now. Both of these lengths
3154 * are calculated conservatively, including all
3155 * XDR overheads.
3156 */
3157 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
3158 NFSX_V3POSTOPATTR);
3159 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
3160 if (len > cnt || dirlen > fullsiz) {
3161 eofflag = 0;
3162 break;
3163 }
3164
3165 /*
3166 * Build the directory record xdr from
3167 * the dirent entry.
3168 */
3169 fp = (struct nfs_fattr *)&fl.fl_fattr;
3170 nfsm_srvfillattr(vap, fp);
3171 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
3172 fl.fl_fhok = nfs_true;
3173 fl.fl_postopok = nfs_true;
3174 txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
3175
3176 nfsm_clget;
3177 *tl = nfs_true;
3178 bp += NFSX_UNSIGNED;
3179 nfsm_clget;
3180 *tl = txdr_unsigned(dp->d_fileno >> 32);
3181 bp += NFSX_UNSIGNED;
3182 nfsm_clget;
3183 *tl = txdr_unsigned(dp->d_fileno);
3184 bp += NFSX_UNSIGNED;
3185 nfsm_clget;
3186 *tl = txdr_unsigned(nlen);
3187 bp += NFSX_UNSIGNED;
3188
3189 /* And loop around copying the name */
3190 xfer = nlen;
3191 cp = dp->d_name;
3192 while (xfer > 0) {
3193 nfsm_clget;
3194 if ((bp + xfer) > be)
3195 tsiz = be - bp;
3196 else
3197 tsiz = xfer;
3198 memcpy(bp, cp, tsiz);
3199 bp += tsiz;
3200 xfer -= tsiz;
3201 if (xfer > 0)
3202 cp += tsiz;
3203 }
3204 /* And null pad to an int32_t boundary */
3205 for (i = 0; i < rem; i++)
3206 *bp++ = '\0';
3207
3208 /*
3209 * Now copy the flrep structure out.
3210 */
3211 xfer = sizeof(struct flrep);
3212 cp = (void *)&fl;
3213 while (xfer > 0) {
3214 nfsm_clget;
3215 if ((bp + xfer) > be)
3216 tsiz = be - bp;
3217 else
3218 tsiz = xfer;
3219 memcpy(bp, cp, tsiz);
3220 bp += tsiz;
3221 xfer -= tsiz;
3222 if (xfer > 0)
3223 cp += tsiz;
3224 }
3225
3226 /*
3227 * ... and filehandle.
3228 */
3229 xfer = NFSRVFH_SIZE(&nnsfh);
3230 cp = NFSRVFH_DATA(&nnsfh);
3231 while (xfer > 0) {
3232 nfsm_clget;
3233 if ((bp + xfer) > be)
3234 tsiz = be - bp;
3235 else
3236 tsiz = xfer;
3237 memcpy(bp, cp, tsiz);
3238 bp += tsiz;
3239 xfer -= tsiz;
3240 if (xfer > 0)
3241 cp += tsiz;
3242 }
3243 }
3244 invalid:
3245 cpos += dp->d_reclen;
3246 dp = (struct dirent *)cpos;
3247 cookiep++;
3248 ncookies--;
3249 }
3250 vrele(vp);
3251 nfsm_clget;
3252 *tl = nfs_false;
3253 bp += NFSX_UNSIGNED;
3254 nfsm_clget;
3255 if (eofflag)
3256 *tl = nfs_true;
3257 else
3258 *tl = nfs_false;
3259 bp += NFSX_UNSIGNED;
3260 if (mp != mb) {
3261 if (bp < be)
3262 mp->m_len = bp - mtod(mp, char *);
3263 } else
3264 mp->m_len += bp - bpos;
3265 free((void *)cookies, M_TEMP);
3266 free((void *)rbuf, M_TEMP);
3267 nfsm_srvdone;
3268 }
3269
3270 /*
3271 * nfs commit service
3272 */
3273 int
3274 nfsrv_commit(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
3275 {
3276 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3277 struct mbuf *nam = nfsd->nd_nam;
3278 char *dpos = nfsd->nd_dpos;
3279 kauth_cred_t cred = nfsd->nd_cr;
3280 struct vattr bfor, aft;
3281 struct vnode *vp;
3282 nfsrvfh_t nsfh;
3283 u_int32_t *tl;
3284 int32_t t1;
3285 char *bpos;
3286 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0;
3287 uint32_t cnt;
3288 char *cp2;
3289 struct mbuf *mb, *mreq __unused;
3290 u_quad_t frev, off, end;
3291
3292 nfsm_srvmtofh(&nsfh);
3293 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3294
3295 off = fxdr_hyper(tl);
3296 tl += 2;
3297 cnt = fxdr_unsigned(uint32_t, *tl);
3298 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
3299 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
3300 if (error) {
3301 nfsm_reply(2 * NFSX_UNSIGNED);
3302 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3303 return (0);
3304 }
3305 for_ret = VOP_GETATTR(vp, &bfor, cred);
3306 end = (cnt > 0) ? off + cnt : vp->v_size;
3307 if (end < off || end > vp->v_size)
3308 end = vp->v_size;
3309 if (off < vp->v_size)
3310 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end);
3311 /* else error == 0, from nfsrv_fhtovp() */
3312 aft_ret = VOP_GETATTR(vp, &aft, cred);
3313 vput(vp);
3314 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3315 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3316 if (!error) {
3317 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3318 *tl++ = txdr_unsigned(boottime.tv_sec);
3319 *tl = txdr_unsigned(boottime.tv_nsec / 1000);
3320 } else {
3321 return (0);
3322 }
3323 nfsm_srvdone;
3324 }
3325
3326 /*
3327 * nfs statfs service
3328 */
3329 int
3330 nfsrv_statfs(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
3331 {
3332 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3333 struct mbuf *nam = nfsd->nd_nam;
3334 char *dpos = nfsd->nd_dpos;
3335 kauth_cred_t cred = nfsd->nd_cr;
3336 struct statvfs *sf = NULL;
3337 struct nfs_statfs *sfp;
3338 u_int32_t *tl;
3339 int32_t t1;
3340 char *bpos;
3341 int error = 0, rdonly, cache = 0, getret = 1;
3342 int v3 = (nfsd->nd_flag & ND_NFSV3);
3343 char *cp2;
3344 struct mbuf *mb, *mreq __unused;
3345 struct vnode *vp;
3346 struct vattr at;
3347 nfsrvfh_t nsfh;
3348 u_quad_t frev, tval;
3349
3350 nfsm_srvmtofh(&nsfh);
3351 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
3352 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
3353 if (error) {
3354 nfsm_reply(NFSX_UNSIGNED);
3355 nfsm_srvpostop_attr(getret, &at);
3356 return (0);
3357 }
3358 sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK);
3359 error = VFS_STATVFS(vp->v_mount, sf);
3360 getret = VOP_GETATTR(vp, &at, cred);
3361 vput(vp);
3362 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3363 if (v3)
3364 nfsm_srvpostop_attr(getret, &at);
3365 if (error) {
3366 free(sf, M_TEMP);
3367 return (0);
3368 }
3369 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3370 if (v3) {
3371 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize);
3372 txdr_hyper(tval, &sfp->sf_tbytes);
3373 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize);
3374 txdr_hyper(tval, &sfp->sf_fbytes);
3375 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize);
3376 txdr_hyper(tval, &sfp->sf_abytes);
3377 tval = (u_quad_t)sf->f_files;
3378 txdr_hyper(tval, &sfp->sf_tfiles);
3379 tval = (u_quad_t)sf->f_ffree;
3380 txdr_hyper(tval, &sfp->sf_ffiles);
3381 txdr_hyper(tval, &sfp->sf_afiles);
3382 sfp->sf_invarsec = 0;
3383 } else {
3384 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3385 sfp->sf_bsize = txdr_unsigned(sf->f_frsize);
3386 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3387 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3388 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3389 }
3390 nfsmout:
3391 if (sf)
3392 free(sf, M_TEMP);
3393 return error;
3394 }
3395
3396 /*
3397 * nfs fsinfo service
3398 */
3399 int
3400 nfsrv_fsinfo(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
3401 {
3402 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3403 struct mbuf *nam = nfsd->nd_nam;
3404 char *dpos = nfsd->nd_dpos;
3405 kauth_cred_t cred = nfsd->nd_cr;
3406 u_int32_t *tl;
3407 struct nfsv3_fsinfo *sip;
3408 int32_t t1;
3409 char *bpos;
3410 int error = 0, rdonly, cache = 0, getret = 1;
3411 uint32_t maxdata;
3412 char *cp2;
3413 struct mbuf *mb, *mreq __unused;
3414 struct vnode *vp;
3415 struct vattr at;
3416 nfsrvfh_t nsfh;
3417 u_quad_t frev, maxfsize;
3418 struct statvfs *sb;
3419
3420 nfsm_srvmtofh(&nsfh);
3421 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
3422 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
3423 if (error) {
3424 nfsm_reply(NFSX_UNSIGNED);
3425 nfsm_srvpostop_attr(getret, &at);
3426 return (0);
3427 }
3428
3429 /* XXX Try to make a guess on the max file size. */
3430 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK);
3431 VFS_STATVFS(vp->v_mount, sb);
3432 maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1;
3433 free(sb, M_TEMP);
3434
3435 getret = VOP_GETATTR(vp, &at, cred);
3436 vput(vp);
3437 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3438 nfsm_srvpostop_attr(getret, &at);
3439 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3440
3441 /*
3442 * XXX
3443 * There should be file system VFS OP(s) to get this information.
3444 * For now, assume ufs.
3445 */
3446 if (slp->ns_so->so_type == SOCK_DGRAM)
3447 maxdata = NFS_MAXDGRAMDATA;
3448 else
3449 maxdata = NFS_MAXDATA;
3450 sip->fs_rtmax = txdr_unsigned(maxdata);
3451 sip->fs_rtpref = txdr_unsigned(maxdata);
3452 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3453 sip->fs_wtmax = txdr_unsigned(maxdata);
3454 sip->fs_wtpref = txdr_unsigned(maxdata);
3455 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3456 sip->fs_dtpref = txdr_unsigned(maxdata);
3457 txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3458 sip->fs_timedelta.nfsv3_sec = 0;
3459 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3460 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3461 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3462 NFSV3FSINFO_CANSETTIME);
3463 nfsm_srvdone;
3464 }
3465
3466 /*
3467 * nfs pathconf service
3468 */
3469 int
3470 nfsrv_pathconf(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
3471 {
3472 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3473 struct mbuf *nam = nfsd->nd_nam;
3474 char *dpos = nfsd->nd_dpos;
3475 kauth_cred_t cred = nfsd->nd_cr;
3476 u_int32_t *tl;
3477 struct nfsv3_pathconf *pc;
3478 int32_t t1;
3479 char *bpos;
3480 int error = 0, rdonly, cache = 0, getret = 1;
3481 register_t linkmax, namemax, chownres, notrunc;
3482 char *cp2;
3483 struct mbuf *mb, *mreq __unused;
3484 struct vnode *vp;
3485 struct vattr at;
3486 nfsrvfh_t nsfh;
3487 u_quad_t frev;
3488
3489 nfsm_srvmtofh(&nsfh);
3490 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
3491 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
3492 if (error) {
3493 nfsm_reply(NFSX_UNSIGNED);
3494 nfsm_srvpostop_attr(getret, &at);
3495 return (0);
3496 }
3497 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3498 if (!error)
3499 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3500 if (!error)
3501 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3502 if (!error)
3503 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc);
3504 getret = VOP_GETATTR(vp, &at, cred);
3505 vput(vp);
3506 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3507 nfsm_srvpostop_attr(getret, &at);
3508 if (error)
3509 return (0);
3510 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3511
3512 pc->pc_linkmax = txdr_unsigned(linkmax);
3513 pc->pc_namemax = txdr_unsigned(namemax);
3514 pc->pc_notrunc = txdr_unsigned(notrunc);
3515 pc->pc_chownrestricted = txdr_unsigned(chownres);
3516
3517 /*
3518 * These should probably be supported by VOP_PATHCONF(), but
3519 * until msdosfs is exportable (why would you want to?), the
3520 * Unix defaults should be ok.
3521 */
3522 pc->pc_caseinsensitive = nfs_false;
3523 pc->pc_casepreserving = nfs_true;
3524 nfsm_srvdone;
3525 }
3526
3527 /*
3528 * Null operation, used by clients to ping server
3529 */
3530 /* ARGSUSED */
3531 int
3532 nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3533 struct lwp *lwp, struct mbuf **mrq)
3534 {
3535 struct mbuf *mrep = nfsd->nd_mrep;
3536 char *bpos;
3537 int error = NFSERR_RETVOID, cache = 0;
3538 struct mbuf *mb, *mreq __unused;
3539 u_quad_t frev;
3540
3541 nfsm_reply(0);
3542 nfsmout:
3543 return (0);
3544 }
3545
3546 /*
3547 * No operation, used for obsolete procedures
3548 */
3549 /* ARGSUSED */
3550 int
3551 nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3552 struct lwp *lwp, struct mbuf **mrq)
3553 {
3554 struct mbuf *mrep = nfsd->nd_mrep;
3555 char *bpos;
3556 int error, cache = 0;
3557 struct mbuf *mb, *mreq __unused;
3558 u_quad_t frev;
3559
3560 if (nfsd->nd_repstat)
3561 error = nfsd->nd_repstat;
3562 else
3563 error = EPROCUNAVAIL;
3564 nfsm_reply(0);
3565 nfsmout:
3566 return (0);
3567 }
3568
3569 /*
3570 * Perform access checking for vnodes obtained from file handles that would
3571 * refer to files already opened by a Unix client. You cannot just use
3572 * vn_writechk() and VOP_ACCESS() for two reasons.
3573 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3574 * 2 - The owner is to be given access irrespective of mode bits for some
3575 * operations, so that processes that chmod after opening a file don't
3576 * break. I don't like this because it opens a security hole, but since
3577 * the nfs server opens a security hole the size of a barn door anyhow,
3578 * what the heck.
3579 *
3580 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3581 * will return EPERM instead of EACCES. EPERM is always an error.
3582 */
3583 int
3584 nfsrv_access(struct vnode *vp, int flags, kauth_cred_t cred, int rdonly, struct lwp *lwp, int override)
3585 {
3586 struct vattr vattr;
3587 int error;
3588 if (flags & VWRITE) {
3589 /* Just vn_writechk() changed to check rdonly */
3590 /*
3591 * Disallow write attempts on read-only file systems;
3592 * unless the file is a socket or a block or character
3593 * device resident on the file system.
3594 */
3595 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3596 switch (vp->v_type) {
3597 case VREG:
3598 case VDIR:
3599 case VLNK:
3600 return (EROFS);
3601 default:
3602 break;
3603 }
3604 }
3605
3606 /*
3607 * If the vnode is in use as a process's text,
3608 * we can't allow writing.
3609 */
3610 if (vp->v_iflag & VI_TEXT)
3611 return (ETXTBSY);
3612 }
3613 error = VOP_GETATTR(vp, &vattr, cred);
3614 if (error)
3615 return (error);
3616 error = VOP_ACCESS(vp, flags, cred);
3617 /*
3618 * Allow certain operations for the owner (reads and writes
3619 * on files that are already open).
3620 */
3621 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid)
3622 error = 0;
3623 return error;
3624 }
3625