nfs_serv.c revision 1.161.2.2 1 /* $NetBSD: nfs_serv.c,v 1.161.2.2 2012/10/30 17:22:51 yamt 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 1.161.2.2 2012/10/30 17:22:51 yamt 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 default:
148 return ENOTTY;
149 }
150 }
151
152 /*
153 * nfs v3 access service
154 */
155 int
156 nfsrv3_access(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
157 {
158 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
159 struct mbuf *nam = nfsd->nd_nam;
160 char *dpos = nfsd->nd_dpos;
161 kauth_cred_t cred = nfsd->nd_cr;
162 struct vnode *vp;
163 nfsrvfh_t nsfh;
164 u_int32_t *tl;
165 int32_t t1;
166 char *bpos;
167 int error = 0, rdonly, cache = 0, getret;
168 char *cp2;
169 struct mbuf *mb, *mreq;
170 struct vattr va;
171 u_long inmode, testmode, outmode;
172 u_quad_t frev;
173
174 nfsm_srvmtofh(&nsfh);
175 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
176 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly,
177 (nfsd->nd_flag & ND_KERBAUTH), false);
178 if (error) {
179 nfsm_reply(NFSX_UNSIGNED);
180 nfsm_srvpostop_attr(1, (struct vattr *)0);
181 return (0);
182 }
183 inmode = fxdr_unsigned(u_int32_t, *tl);
184 outmode = 0;
185 if ((inmode & NFSV3ACCESS_READ) &&
186 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0)
187 outmode |= NFSV3ACCESS_READ;
188 if (vp->v_type != VDIR) {
189 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
190 if (testmode &&
191 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0)
192 outmode |= testmode;
193 if ((inmode & NFSV3ACCESS_EXECUTE) &&
194 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0)
195 outmode |= NFSV3ACCESS_EXECUTE;
196 } else {
197 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
198 NFSV3ACCESS_DELETE);
199 if (testmode &&
200 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0)
201 outmode |= testmode;
202 if ((inmode & NFSV3ACCESS_LOOKUP) &&
203 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0)
204 outmode |= NFSV3ACCESS_LOOKUP;
205 }
206 getret = VOP_GETATTR(vp, &va, cred);
207 vput(vp);
208 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
209 nfsm_srvpostop_attr(getret, &va);
210 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
211 *tl = txdr_unsigned(outmode);
212 nfsm_srvdone;
213 }
214
215 /*
216 * nfs getattr service
217 */
218 int
219 nfsrv_getattr(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
220 {
221 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
222 struct mbuf *nam = nfsd->nd_nam;
223 char *dpos = nfsd->nd_dpos;
224 kauth_cred_t cred = nfsd->nd_cr;
225 struct nfs_fattr *fp;
226 struct vattr va;
227 struct vnode *vp;
228 nfsrvfh_t nsfh;
229 u_int32_t *tl;
230 int32_t t1;
231 char *bpos;
232 int error = 0, rdonly, cache = 0;
233 char *cp2;
234 struct mbuf *mb, *mreq;
235 u_quad_t frev;
236
237 nfsm_srvmtofh(&nsfh);
238 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly,
239 (nfsd->nd_flag & ND_KERBAUTH), false);
240 if (error) {
241 nfsm_reply(0);
242 return (0);
243 }
244 nqsrv_getl(vp, ND_READ);
245 error = VOP_GETATTR(vp, &va, cred);
246 vput(vp);
247 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
248 if (error)
249 return (0);
250 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
251 nfsm_srvfillattr(&va, fp);
252 nfsm_srvdone;
253 }
254
255 /*
256 * nfs setattr service
257 */
258 int
259 nfsrv_setattr(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
260 {
261 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
262 struct mbuf *nam = nfsd->nd_nam;
263 char *dpos = nfsd->nd_dpos;
264 kauth_cred_t cred = nfsd->nd_cr;
265 struct vattr va, preat;
266 struct nfsv2_sattr *sp;
267 struct nfs_fattr *fp;
268 struct vnode *vp;
269 nfsrvfh_t nsfh;
270 u_int32_t *tl;
271 int32_t t1;
272 char *bpos;
273 int error = 0, rdonly, cache = 0, preat_ret = 1, postat_ret = 1;
274 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
275 char *cp2;
276 struct mbuf *mb, *mreq;
277 u_quad_t frev;
278 struct timespec guard;
279
280 memset(&guard, 0, sizeof guard); /* XXX gcc */
281
282 nfsm_srvmtofh(&nsfh);
283 vattr_null(&va);
284 if (v3) {
285 nfsm_srvsattr(&va);
286 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
287 gcheck = fxdr_unsigned(int, *tl);
288 if (gcheck) {
289 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
290 fxdr_nfsv3time(tl, &guard);
291 }
292 } else {
293 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
294 /*
295 * Nah nah nah nah na nah
296 * There is a bug in the Sun client that puts 0xffff in the mode
297 * field of sattr when it should put in 0xffffffff. The u_short
298 * doesn't sign extend.
299 * --> check the low order 2 bytes for 0xffff
300 */
301 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
302 va.va_mode = nfstov_mode(sp->sa_mode);
303 if (sp->sa_uid != nfs_xdrneg1)
304 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
305 if (sp->sa_gid != nfs_xdrneg1)
306 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
307 if (sp->sa_size != nfs_xdrneg1)
308 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
309 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
310 #ifdef notyet
311 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime);
312 #else
313 va.va_atime.tv_sec =
314 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
315 va.va_atime.tv_nsec = 0;
316 #endif
317 }
318 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
319 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime);
320
321 }
322
323 /*
324 * Now that we have all the fields, lets do it.
325 */
326 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam, &rdonly,
327 (nfsd->nd_flag & ND_KERBAUTH), false);
328 if (error) {
329 nfsm_reply(2 * NFSX_UNSIGNED);
330 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
331 return (0);
332 }
333 nqsrv_getl(vp, ND_WRITE);
334 if (v3) {
335 error = preat_ret = VOP_GETATTR(vp, &preat, cred);
336 if (!error && gcheck &&
337 (preat.va_ctime.tv_sec != guard.tv_sec ||
338 preat.va_ctime.tv_nsec != guard.tv_nsec))
339 error = NFSERR_NOT_SYNC;
340 if (error) {
341 vput(vp);
342 nfsm_reply(NFSX_WCCDATA(v3));
343 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
344 return (0);
345 }
346 }
347
348 /*
349 * If the size is being changed write acces is required, otherwise
350 * just check for a read only file system.
351 */
352 if (va.va_size == ((u_quad_t)((quad_t) -1))) {
353 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
354 error = EROFS;
355 goto out;
356 }
357 } else {
358 if (vp->v_type == VDIR) {
359 error = EISDIR;
360 goto out;
361 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
362 lwp, 0)) != 0)
363 goto out;
364 }
365 error = VOP_SETATTR(vp, &va, cred);
366 postat_ret = VOP_GETATTR(vp, &va, cred);
367 if (!error)
368 error = postat_ret;
369 out:
370 vput(vp);
371 nfsm_reply(NFSX_WCCORFATTR(v3));
372 if (v3) {
373 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
374 return (0);
375 } else {
376 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
377 nfsm_srvfillattr(&va, fp);
378 }
379 nfsm_srvdone;
380 }
381
382 /*
383 * nfs lookup rpc
384 */
385 int
386 nfsrv_lookup(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
387 {
388 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
389 struct mbuf *nam = nfsd->nd_nam;
390 char *dpos = nfsd->nd_dpos;
391 kauth_cred_t cred = nfsd->nd_cr;
392 struct nfs_fattr *fp;
393 struct nameidata nd, ind, *ndp = &nd;
394 struct pathbuf *ipb = NULL;
395 struct vnode *vp, *dirp;
396 nfsrvfh_t nsfh;
397 char *cp;
398 u_int32_t *tl;
399 int32_t t1;
400 char *bpos;
401 int error = 0, cache = 0, dirattr_ret = 1;
402 uint32_t len;
403 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
404 char *cp2;
405 struct mbuf *mb, *mreq;
406 struct vattr va, dirattr;
407 u_quad_t frev;
408
409 nfsm_srvmtofh(&nsfh);
410 nfsm_srvnamesiz(len);
411
412 pubflag = nfs_ispublicfh(&nsfh);
413
414 nd.ni_cnd.cn_cred = cred;
415 nd.ni_cnd.cn_nameiop = LOOKUP;
416 nd.ni_cnd.cn_flags = LOCKLEAF;
417 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
418 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
419
420 if (!error && pubflag) {
421 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL &&
422 (ipb = pathbuf_create(nfs_pub.np_index)) != NULL) {
423 /*
424 * Setup call to lookup() to see if we can find
425 * the index file. Arguably, this doesn't belong
426 * in a kernel.. Ugh.
427 */
428 ind = nd;
429 VOP_UNLOCK(nd.ni_vp);
430 ind.ni_pathbuf = ipb;
431
432 error = lookup_for_nfsd_index(&ind, nd.ni_vp);
433 if (!error) {
434 /*
435 * Found an index file. Get rid of
436 * the old references.
437 */
438 if (dirp)
439 vrele(dirp);
440 dirp = nd.ni_vp;
441 ndp = &ind;
442 } else
443 error = 0;
444 }
445 /*
446 * If the public filehandle was used, check that this lookup
447 * didn't result in a filehandle outside the publicly exported
448 * filesystem.
449 */
450
451 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
452 vput(nd.ni_vp);
453 error = EPERM;
454 }
455 }
456
457 if (error) {
458 if (nd.ni_pathbuf != NULL) {
459 pathbuf_destroy(nd.ni_pathbuf);
460 }
461 if (ipb != NULL) {
462 pathbuf_destroy(ipb);
463 }
464 if (dirp) {
465 if (v3) {
466 vn_lock(dirp, LK_SHARED | LK_RETRY);
467 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred);
468 vput(dirp);
469 } else
470 vrele(dirp);
471 }
472 nfsm_reply(NFSX_POSTOPATTR(v3));
473 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
474 return (0);
475 }
476
477 nqsrv_getl(ndp->ni_startdir, ND_READ);
478 pathbuf_destroy(nd.ni_pathbuf);
479 if (ipb != NULL) {
480 pathbuf_destroy(ipb);
481 }
482 vp = ndp->ni_vp;
483 error = nfsrv_composefh(vp, &nsfh, v3);
484 if (!error)
485 error = VOP_GETATTR(vp, &va, cred);
486 vput(vp);
487 if (dirp) {
488 if (v3) {
489 vn_lock(dirp, LK_SHARED | LK_RETRY);
490 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred);
491 vput(dirp);
492 } else
493 vrele(dirp);
494 }
495 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPORFATTR(v3) +
496 NFSX_POSTOPATTR(v3));
497 if (error) {
498 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
499 return (0);
500 }
501 nfsm_srvfhtom(&nsfh, v3);
502 if (v3) {
503 nfsm_srvpostop_attr(0, &va);
504 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
505 } else {
506 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
507 nfsm_srvfillattr(&va, fp);
508 }
509 nfsm_srvdone;
510 }
511
512 /*
513 * nfs readlink service
514 */
515 int
516 nfsrv_readlink(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
517 {
518 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
519 struct mbuf *nam = nfsd->nd_nam;
520 char *dpos = nfsd->nd_dpos;
521 kauth_cred_t cred = nfsd->nd_cr;
522 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
523 struct iovec *ivp = iv;
524 struct mbuf *mp;
525 u_int32_t *tl;
526 int32_t t1;
527 char *bpos;
528 int error = 0, rdonly, cache = 0, i, padlen, getret;
529 uint32_t len;
530 int v3 = (nfsd->nd_flag & ND_NFSV3);
531 char *cp2;
532 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq;
533 struct vnode *vp;
534 struct vattr attr;
535 nfsrvfh_t nsfh;
536 struct uio io, *uiop = &io;
537 u_quad_t frev;
538
539 nfsm_srvmtofh(&nsfh);
540 len = 0;
541 i = 0;
542 while (len < NFS_MAXPATHLEN) {
543 mp = m_get(M_WAIT, MT_DATA);
544 MCLAIM(mp, &nfs_mowner);
545 m_clget(mp, M_WAIT);
546 mp->m_len = NFSMSIZ(mp);
547 if (len == 0)
548 mp3 = mp2 = mp;
549 else {
550 mp2->m_next = mp;
551 mp2 = mp;
552 }
553 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
554 mp->m_len = NFS_MAXPATHLEN-len;
555 len = NFS_MAXPATHLEN;
556 } else
557 len += mp->m_len;
558 ivp->iov_base = mtod(mp, void *);
559 ivp->iov_len = mp->m_len;
560 i++;
561 ivp++;
562 }
563 uiop->uio_iov = iv;
564 uiop->uio_iovcnt = i;
565 uiop->uio_offset = 0;
566 uiop->uio_resid = len;
567 uiop->uio_rw = UIO_READ;
568 UIO_SETUP_SYSSPACE(uiop);
569 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
570 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
571 if (error) {
572 m_freem(mp3);
573 nfsm_reply(2 * NFSX_UNSIGNED);
574 nfsm_srvpostop_attr(1, (struct vattr *)0);
575 return (0);
576 }
577 if (vp->v_type != VLNK) {
578 if (v3)
579 error = EINVAL;
580 else
581 error = ENXIO;
582 goto out;
583 }
584 nqsrv_getl(vp, ND_READ);
585 error = VOP_READLINK(vp, uiop, cred);
586 out:
587 getret = VOP_GETATTR(vp, &attr, cred);
588 vput(vp);
589 if (error)
590 m_freem(mp3);
591 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
592 if (v3) {
593 nfsm_srvpostop_attr(getret, &attr);
594 if (error)
595 return (0);
596 }
597 len -= uiop->uio_resid;
598 padlen = nfsm_padlen(len);
599 if (uiop->uio_resid || padlen)
600 nfs_zeropad(mp3, uiop->uio_resid, padlen);
601 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
602 *tl = txdr_unsigned(len);
603 mb->m_next = mp3;
604 nfsm_srvdone;
605 }
606
607 /*
608 * nfs read service
609 */
610 int
611 nfsrv_read(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
612 {
613 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
614 struct mbuf *nam = nfsd->nd_nam;
615 char *dpos = nfsd->nd_dpos;
616 kauth_cred_t cred = nfsd->nd_cr;
617 struct mbuf *m;
618 struct nfs_fattr *fp;
619 u_int32_t *tl;
620 int32_t t1;
621 int i;
622 char *bpos;
623 int error = 0, rdonly, cache = 0, getret;
624 int v3 = (nfsd->nd_flag & ND_NFSV3);
625 uint32_t reqlen, len, cnt, left;
626 int padlen;
627 char *cp2;
628 struct mbuf *mb, *mreq;
629 struct vnode *vp;
630 nfsrvfh_t nsfh;
631 struct uio io, *uiop = &io;
632 struct vattr va;
633 off_t off;
634 u_quad_t frev;
635
636 nfsm_srvmtofh(&nsfh);
637 if (v3) {
638 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
639 off = fxdr_hyper(tl);
640 } else {
641 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
642 off = (off_t)fxdr_unsigned(u_int32_t, *tl);
643 }
644 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
645 reqlen = fxdr_unsigned(uint32_t, *tl);
646 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd));
647 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
648 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
649 if (error) {
650 nfsm_reply(2 * NFSX_UNSIGNED);
651 nfsm_srvpostop_attr(1, (struct vattr *)0);
652 return (0);
653 }
654 if (vp->v_type != VREG) {
655 if (v3)
656 error = EINVAL;
657 else
658 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
659 }
660 if (!error) {
661 nqsrv_getl(vp, ND_READ);
662 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0)
663 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1);
664 }
665 getret = VOP_GETATTR(vp, &va, cred);
666 if (!error)
667 error = getret;
668 if (error) {
669 vput(vp);
670 nfsm_reply(NFSX_POSTOPATTR(v3));
671 nfsm_srvpostop_attr(getret, &va);
672 return (0);
673 }
674 if (off >= va.va_size)
675 cnt = 0;
676 else if ((off + reqlen) > va.va_size)
677 cnt = va.va_size - off;
678 else
679 cnt = reqlen;
680 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
681 if (v3) {
682 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
683 *tl++ = nfs_true;
684 fp = (struct nfs_fattr *)tl;
685 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
686 } else {
687 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
688 fp = (struct nfs_fattr *)tl;
689 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
690 }
691 len = left = cnt;
692 if (cnt > 0) {
693 if (nfsd_use_loan) {
694 struct vm_page **pgpp;
695 voff_t pgoff = trunc_page(off);
696 int npages;
697 vaddr_t lva;
698
699 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT;
700 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */
701
702 /* allocate kva for mbuf data */
703 lva = sokvaalloc(pgoff, npages << PAGE_SHIFT,
704 slp->ns_so);
705 if (lva == 0) {
706 /* fall back to VOP_READ */
707 goto loan_fail;
708 }
709
710 /* allocate mbuf */
711 m = m_get(M_WAIT, MT_DATA);
712 MCLAIM(m, &nfs_mowner);
713 pgpp = m->m_ext.ext_pgs;
714
715 /* loan pages */
716 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages,
717 pgpp);
718 if (error) {
719 sokvafree(lva, npages << PAGE_SHIFT);
720 m_free(m);
721 if (error == EBUSY)
722 goto loan_fail;
723 goto read_error;
724 }
725
726 /* associate kva to mbuf */
727 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)),
728 cnt, M_MBUF, soloanfree, slp->ns_so);
729 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
730 m->m_len = cnt;
731
732 /* map pages */
733 for (i = 0; i < npages; i++) {
734 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]),
735 VM_PROT_READ, 0);
736 lva += PAGE_SIZE;
737 }
738
739 pmap_update(pmap_kernel());
740
741 mb->m_next = m;
742 mb = m;
743 error = 0;
744 uiop->uio_resid = 0;
745 } else {
746 struct iovec *iv;
747 struct iovec *iv2;
748 struct mbuf *m2;
749 int siz;
750 loan_fail:
751 /*
752 * Generate the mbuf list with the uio_iov ref. to it.
753 */
754 i = 0;
755 m = m2 = mb;
756 while (left > 0) {
757 siz = min(M_TRAILINGSPACE(m), left);
758 if (siz > 0) {
759 left -= siz;
760 i++;
761 }
762 if (left > 0) {
763 m = m_get(M_WAIT, MT_DATA);
764 MCLAIM(m, &nfs_mowner);
765 m_clget(m, M_WAIT);
766 m->m_len = 0;
767 m2->m_next = m;
768 m2 = m;
769 }
770 }
771 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK);
772 uiop->uio_iov = iv2 = iv;
773 m = mb;
774 left = cnt;
775 i = 0;
776 while (left > 0) {
777 if (m == NULL)
778 panic("nfsrv_read iov");
779 siz = min(M_TRAILINGSPACE(m), left);
780 if (siz > 0) {
781 iv->iov_base = mtod(m, char *) +
782 m->m_len;
783 iv->iov_len = siz;
784 m->m_len += siz;
785 left -= siz;
786 iv++;
787 i++;
788 }
789 m = m->m_next;
790 }
791 uiop->uio_iovcnt = i;
792 uiop->uio_offset = off;
793 uiop->uio_resid = cnt;
794 uiop->uio_rw = UIO_READ;
795 UIO_SETUP_SYSSPACE(uiop);
796 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
797 free((void *)iv2, M_TEMP);
798 }
799 read_error:
800 if (error || (getret = VOP_GETATTR(vp, &va, cred)) != 0){
801 if (!error)
802 error = getret;
803 m_freem(mreq);
804 vput(vp);
805 nfsm_reply(NFSX_POSTOPATTR(v3));
806 nfsm_srvpostop_attr(getret, &va);
807 return (0);
808 }
809 } else {
810 uiop->uio_resid = 0;
811 }
812 vput(vp);
813 nfsm_srvfillattr(&va, fp);
814 len -= uiop->uio_resid;
815 padlen = nfsm_padlen(len);
816 if (uiop->uio_resid || padlen)
817 nfs_zeropad(mb, uiop->uio_resid, padlen);
818 if (v3) {
819 /* count */
820 *tl++ = txdr_unsigned(len);
821 /* eof */
822 if (off + len >= va.va_size)
823 *tl++ = nfs_true;
824 else
825 *tl++ = nfs_false;
826 }
827 *tl = txdr_unsigned(len);
828 nfsm_srvdone;
829 }
830
831 /*
832 * nfs write service
833 */
834 int
835 nfsrv_write(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
836 {
837 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
838 struct mbuf *nam = nfsd->nd_nam;
839 char *dpos = nfsd->nd_dpos;
840 kauth_cred_t cred = nfsd->nd_cr;
841 struct iovec *ivp;
842 int i, cnt;
843 struct mbuf *mp;
844 struct nfs_fattr *fp;
845 struct iovec *iv;
846 struct vattr va, forat;
847 u_int32_t *tl;
848 int32_t t1;
849 char *bpos;
850 int error = 0, rdonly, cache = 0, len, forat_ret = 1;
851 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
852 int stable = NFSV3WRITE_FILESYNC;
853 int v3 = (nfsd->nd_flag & ND_NFSV3);
854 char *cp2;
855 struct mbuf *mb, *mreq;
856 struct vnode *vp;
857 nfsrvfh_t nsfh;
858 struct uio io, *uiop = &io;
859 off_t off;
860 u_quad_t frev;
861
862 if (mrep == NULL) {
863 *mrq = NULL;
864 return (0);
865 }
866 nfsm_srvmtofh(&nsfh);
867 if (v3) {
868 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
869 off = fxdr_hyper(tl);
870 tl += 3;
871 stable = fxdr_unsigned(int, *tl++);
872 } else {
873 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
874 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
875 tl += 2;
876 }
877 retlen = len = fxdr_unsigned(int32_t, *tl);
878 cnt = i = 0;
879
880 /*
881 * For NFS Version 2, it is not obvious what a write of zero length
882 * should do, but I might as well be consistent with Version 3,
883 * which is to return ok so long as there are no permission problems.
884 */
885 if (len > 0) {
886 zeroing = 1;
887 mp = mrep;
888 while (mp) {
889 if (mp == md) {
890 zeroing = 0;
891 adjust = dpos - mtod(mp, char *);
892 mp->m_len -= adjust;
893 if (mp->m_len > 0 && adjust > 0)
894 NFSMADV(mp, adjust);
895 }
896 if (zeroing)
897 mp->m_len = 0;
898 else if (mp->m_len > 0) {
899 i += mp->m_len;
900 if (i > len) {
901 mp->m_len -= (i - len);
902 zeroing = 1;
903 }
904 if (mp->m_len > 0)
905 cnt++;
906 }
907 mp = mp->m_next;
908 }
909 }
910 if (len > NFS_MAXDATA || len < 0 || i < len) {
911 error = EIO;
912 nfsm_reply(2 * NFSX_UNSIGNED);
913 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
914 return (0);
915 }
916 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
917 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
918 if (error) {
919 nfsm_reply(2 * NFSX_UNSIGNED);
920 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
921 return (0);
922 }
923 if (v3)
924 forat_ret = VOP_GETATTR(vp, &forat, cred);
925 if (vp->v_type != VREG) {
926 if (v3)
927 error = EINVAL;
928 else
929 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
930 }
931 if (!error) {
932 nqsrv_getl(vp, ND_WRITE);
933 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1);
934 }
935 if (error) {
936 vput(vp);
937 nfsm_reply(NFSX_WCCDATA(v3));
938 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
939 return (0);
940 }
941
942 if (len > 0) {
943 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK);
944 uiop->uio_iov = iv = ivp;
945 uiop->uio_iovcnt = cnt;
946 mp = mrep;
947 while (mp) {
948 if (mp->m_len > 0) {
949 ivp->iov_base = mtod(mp, void *);
950 ivp->iov_len = mp->m_len;
951 ivp++;
952 }
953 mp = mp->m_next;
954 }
955
956 /*
957 * XXX
958 * The IO_METASYNC flag indicates that all metadata (and not
959 * just enough to ensure data integrity) must be written to
960 * stable storage synchronously.
961 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
962 */
963 if (stable == NFSV3WRITE_UNSTABLE)
964 ioflags = IO_NODELOCKED;
965 else if (stable == NFSV3WRITE_DATASYNC)
966 ioflags = (IO_SYNC | IO_NODELOCKED);
967 else
968 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
969 uiop->uio_resid = len;
970 uiop->uio_rw = UIO_WRITE;
971 uiop->uio_offset = off;
972 UIO_SETUP_SYSSPACE(uiop);
973 error = VOP_WRITE(vp, uiop, ioflags, cred);
974 nfsstats.srvvop_writes++;
975 free(iv, M_TEMP);
976 }
977 aftat_ret = VOP_GETATTR(vp, &va, cred);
978 vput(vp);
979 if (!error)
980 error = aftat_ret;
981 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
982 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
983 if (v3) {
984 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
985 if (error)
986 return (0);
987 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
988 *tl++ = txdr_unsigned(retlen);
989 if (stable == NFSV3WRITE_UNSTABLE)
990 *tl++ = txdr_unsigned(stable);
991 else
992 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
993 /*
994 * Actually, there is no need to txdr these fields,
995 * but it may make the values more human readable,
996 * for debugging purposes.
997 */
998 *tl++ = txdr_unsigned(boottime.tv_sec);
999 *tl = txdr_unsigned(boottime.tv_nsec / 1000);
1000 } else {
1001 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1002 nfsm_srvfillattr(&va, fp);
1003 }
1004 nfsm_srvdone;
1005 }
1006
1007 /*
1008 * XXX elad: the original NFSW_SAMECRED() macro also made sure the
1009 * two nd_flag fields of the descriptors contained
1010 * ND_KERBAUTH.
1011 */
1012 static int
1013 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2)
1014 {
1015 int i, do_ngroups;
1016
1017 if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2))
1018 return (0);
1019 if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2))
1020 return (0);
1021 do_ngroups = kauth_cred_ngroups(cred1);
1022 for (i = 0; i < do_ngroups; i++)
1023 if (kauth_cred_group(cred1, i) !=
1024 kauth_cred_group(cred2, i))
1025 return (0);
1026
1027 return (1);
1028 }
1029
1030 static struct nfsrvw_delayhash *
1031 nfsrv_nwdelayhash(struct nfssvc_sock *slp, const nfsrvfh_t *nsfh)
1032 {
1033 uint32_t hash;
1034
1035 hash = hash32_buf(NFSRVFH_DATA(nsfh), NFSRVFH_SIZE(nsfh),
1036 HASH32_BUF_INIT);
1037 return &slp->ns_wdelayhashtbl[hash % NFS_WDELAYHASHSIZ];
1038 }
1039
1040 /*
1041 * NFS write service with write gathering support. Called when
1042 * nfsrvw_procrastinate > 0.
1043 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
1044 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
1045 * Jan. 1994.
1046 */
1047 int
1048 nfsrv_writegather(struct nfsrv_descript **ndp, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1049 {
1050 struct timeval now;
1051 struct iovec *ivp;
1052 struct mbuf *mp;
1053 struct nfsrv_descript *wp, *nfsd, *owp, *swp;
1054 struct nfs_fattr *fp;
1055 int i = 0;
1056 struct iovec *iov;
1057 struct nfsrvw_delayhash *wpp;
1058 kauth_cred_t cred;
1059 struct vattr va, forat;
1060 u_int32_t *tl;
1061 int32_t t1;
1062 char *bpos, *dpos;
1063 int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1;
1064 int ioflags, aftat_ret = 1, adjust, v3, zeroing;
1065 char *cp2;
1066 struct mbuf *mb, *mreq, *mrep, *md;
1067 struct vnode *vp;
1068 struct uio io, *uiop = &io;
1069 u_quad_t frev, cur_usec;
1070
1071 *mrq = NULL;
1072 if (*ndp) {
1073 nfsd = *ndp;
1074 *ndp = NULL;
1075 mrep = nfsd->nd_mrep;
1076 md = nfsd->nd_md;
1077 dpos = nfsd->nd_dpos;
1078 cred = nfsd->nd_cr;
1079 v3 = (nfsd->nd_flag & ND_NFSV3);
1080 LIST_INIT(&nfsd->nd_coalesce);
1081 nfsd->nd_mreq = NULL;
1082 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
1083 getmicrotime(&now);
1084 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec;
1085 nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
1086
1087 /*
1088 * Now, get the write header..
1089 */
1090 nfsm_srvmtofh(&nfsd->nd_fh);
1091 if (v3) {
1092 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1093 nfsd->nd_off = fxdr_hyper(tl);
1094 tl += 3;
1095 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
1096 } else {
1097 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1098 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
1099 tl += 2;
1100 }
1101 len = fxdr_unsigned(int32_t, *tl);
1102 nfsd->nd_len = len;
1103 nfsd->nd_eoff = nfsd->nd_off + len;
1104
1105 /*
1106 * Trim the header out of the mbuf list and trim off any trailing
1107 * junk so that the mbuf list has only the write data.
1108 */
1109 zeroing = 1;
1110 i = 0;
1111 mp = mrep;
1112 while (mp) {
1113 if (mp == md) {
1114 zeroing = 0;
1115 adjust = dpos - mtod(mp, char *);
1116 mp->m_len -= adjust;
1117 if (mp->m_len > 0 && adjust > 0)
1118 NFSMADV(mp, adjust);
1119 }
1120 if (zeroing)
1121 mp->m_len = 0;
1122 else {
1123 i += mp->m_len;
1124 if (i > len) {
1125 mp->m_len -= (i - len);
1126 zeroing = 1;
1127 }
1128 }
1129 mp = mp->m_next;
1130 }
1131 if (len > NFS_MAXDATA || len < 0 || i < len) {
1132 nfsmout:
1133 m_freem(mrep);
1134 error = EIO;
1135 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1136 if (v3)
1137 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1138 nfsd->nd_mreq = mreq;
1139 nfsd->nd_mrep = NULL;
1140 nfsd->nd_time = 0;
1141 }
1142
1143 /*
1144 * Add this entry to the hash and time queues.
1145 */
1146 owp = NULL;
1147 mutex_enter(&nfsd_lock);
1148 wp = LIST_FIRST(&slp->ns_tq);
1149 while (wp && wp->nd_time < nfsd->nd_time) {
1150 owp = wp;
1151 wp = LIST_NEXT(wp, nd_tq);
1152 }
1153 if (owp) {
1154 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1155 } else {
1156 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1157 }
1158 if (nfsd->nd_mrep) {
1159 wpp = nfsrv_nwdelayhash(slp, &nfsd->nd_fh);
1160 owp = NULL;
1161 wp = LIST_FIRST(wpp);
1162 while (wp && nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) {
1163 owp = wp;
1164 wp = LIST_NEXT(wp, nd_hash);
1165 }
1166 while (wp && wp->nd_off < nfsd->nd_off &&
1167 !nfsrv_comparefh(&nfsd->nd_fh, &wp->nd_fh)) {
1168 owp = wp;
1169 wp = LIST_NEXT(wp, nd_hash);
1170 }
1171 if (owp) {
1172 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1173
1174 /*
1175 * Search the hash list for overlapping entries and
1176 * coalesce.
1177 */
1178 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1179 wp = LIST_NEXT(nfsd, nd_hash);
1180 if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr))
1181 nfsrvw_coalesce(owp, nfsd);
1182 }
1183 } else {
1184 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1185 }
1186 }
1187 mutex_exit(&nfsd_lock);
1188 }
1189
1190 /*
1191 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1192 * and generate the associated reply mbuf list(s).
1193 */
1194 loop1:
1195 getmicrotime(&now);
1196 cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec;
1197 mutex_enter(&nfsd_lock);
1198 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) {
1199 owp = LIST_NEXT(nfsd, nd_tq);
1200 if (nfsd->nd_time > cur_usec)
1201 break;
1202 if (nfsd->nd_mreq)
1203 continue;
1204 LIST_REMOVE(nfsd, nd_tq);
1205 LIST_REMOVE(nfsd, nd_hash);
1206 mutex_exit(&nfsd_lock);
1207
1208 mrep = nfsd->nd_mrep;
1209 nfsd->nd_mrep = NULL;
1210 cred = nfsd->nd_cr;
1211 v3 = (nfsd->nd_flag & ND_NFSV3);
1212 forat_ret = aftat_ret = 1;
1213 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1214 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1215 false);
1216 if (!error) {
1217 if (v3)
1218 forat_ret = VOP_GETATTR(vp, &forat, cred);
1219 if (vp->v_type != VREG) {
1220 if (v3)
1221 error = EINVAL;
1222 else
1223 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1224 }
1225 } else
1226 vp = NULL;
1227 if (!error) {
1228 nqsrv_getl(vp, ND_WRITE);
1229 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1);
1230 }
1231
1232 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1233 ioflags = IO_NODELOCKED;
1234 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1235 ioflags = (IO_SYNC | IO_NODELOCKED);
1236 else
1237 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1238 uiop->uio_rw = UIO_WRITE;
1239 uiop->uio_offset = nfsd->nd_off;
1240 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1241 UIO_SETUP_SYSSPACE(uiop);
1242 if (uiop->uio_resid > 0) {
1243 mp = mrep;
1244 i = 0;
1245 while (mp) {
1246 if (mp->m_len > 0)
1247 i++;
1248 mp = mp->m_next;
1249 }
1250 uiop->uio_iovcnt = i;
1251 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
1252 uiop->uio_iov = ivp = iov;
1253 mp = mrep;
1254 while (mp) {
1255 if (mp->m_len > 0) {
1256 ivp->iov_base = mtod(mp, void *);
1257 ivp->iov_len = mp->m_len;
1258 ivp++;
1259 }
1260 mp = mp->m_next;
1261 }
1262 if (!error) {
1263 error = VOP_WRITE(vp, uiop, ioflags, cred);
1264 nfsstats.srvvop_writes++;
1265 }
1266 free((void *)iov, M_TEMP);
1267 }
1268 m_freem(mrep);
1269 if (vp) {
1270 aftat_ret = VOP_GETATTR(vp, &va, cred);
1271 vput(vp);
1272 }
1273
1274 /*
1275 * Loop around generating replies for all write rpcs that have
1276 * now been completed.
1277 */
1278 swp = nfsd;
1279 do {
1280 if (error) {
1281 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1282 if (v3) {
1283 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1284 }
1285 } else {
1286 nfsm_writereply(NFSX_PREOPATTR(v3) +
1287 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1288 NFSX_WRITEVERF(v3), v3);
1289 if (v3) {
1290 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1291 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1292 *tl++ = txdr_unsigned(nfsd->nd_len);
1293 *tl++ = txdr_unsigned(swp->nd_stable);
1294 /*
1295 * Actually, there is no need to txdr these fields,
1296 * but it may make the values more human readable,
1297 * for debugging purposes.
1298 */
1299 *tl++ = txdr_unsigned(boottime.tv_sec);
1300 *tl = txdr_unsigned(boottime.tv_nsec / 1000);
1301 } else {
1302 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1303 nfsm_srvfillattr(&va, fp);
1304 }
1305 }
1306 nfsd->nd_mreq = mreq;
1307 if (nfsd->nd_mrep)
1308 panic("nfsrv_write: nd_mrep not free");
1309
1310 /*
1311 * Done. Put it at the head of the timer queue so that
1312 * the final phase can return the reply.
1313 */
1314 mutex_enter(&nfsd_lock);
1315 if (nfsd != swp) {
1316 nfsd->nd_time = 0;
1317 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1318 }
1319 nfsd = LIST_FIRST(&swp->nd_coalesce);
1320 if (nfsd) {
1321 LIST_REMOVE(nfsd, nd_tq);
1322 }
1323 mutex_exit(&nfsd_lock);
1324 } while (nfsd);
1325 swp->nd_time = 0;
1326
1327 mutex_enter(&nfsd_lock);
1328 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1329 mutex_exit(&nfsd_lock);
1330 goto loop1;
1331 }
1332 mutex_exit(&nfsd_lock);
1333 nfs_timer_start();
1334
1335 /*
1336 * Search for a reply to return.
1337 */
1338 mutex_enter(&nfsd_lock);
1339 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) {
1340 if (nfsd->nd_mreq) {
1341 LIST_REMOVE(nfsd, nd_tq);
1342 *mrq = nfsd->nd_mreq;
1343 *ndp = nfsd;
1344 break;
1345 }
1346 }
1347 mutex_exit(&nfsd_lock);
1348 return (0);
1349 }
1350
1351 /*
1352 * Coalesce the write request nfsd into owp. To do this we must:
1353 * - remove nfsd from the queues
1354 * - merge nfsd->nd_mrep into owp->nd_mrep
1355 * - update the nd_eoff and nd_stable for owp
1356 * - put nfsd on owp's nd_coalesce list
1357 * NB: Must be called at splsoftclock().
1358 */
1359 void
1360 nfsrvw_coalesce(struct nfsrv_descript *owp, struct nfsrv_descript *nfsd)
1361 {
1362 int overlap;
1363 struct mbuf *mp;
1364 struct nfsrv_descript *m;
1365
1366 KASSERT(mutex_owned(&nfsd_lock));
1367
1368 LIST_REMOVE(nfsd, nd_hash);
1369 LIST_REMOVE(nfsd, nd_tq);
1370 if (owp->nd_eoff < nfsd->nd_eoff) {
1371 overlap = owp->nd_eoff - nfsd->nd_off;
1372 if (overlap < 0)
1373 panic("nfsrv_coalesce: bad off");
1374 if (overlap > 0)
1375 m_adj(nfsd->nd_mrep, overlap);
1376 mp = owp->nd_mrep;
1377 while (mp->m_next)
1378 mp = mp->m_next;
1379 mp->m_next = nfsd->nd_mrep;
1380 owp->nd_eoff = nfsd->nd_eoff;
1381 } else
1382 m_freem(nfsd->nd_mrep);
1383 nfsd->nd_mrep = NULL;
1384 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1385 owp->nd_stable = NFSV3WRITE_FILESYNC;
1386 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1387 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1388 owp->nd_stable = NFSV3WRITE_DATASYNC;
1389 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1390 /*
1391 * nfsd might hold coalesce elements! Move them to owp.
1392 * Otherwise, requests may be lost and clients will be stuck.
1393 */
1394 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) {
1395 LIST_REMOVE(m, nd_tq);
1396 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1397 }
1398 }
1399
1400 /*
1401 * nfs create service
1402 * now does a truncate to 0 length via. setattr if it already exists
1403 */
1404 int
1405 nfsrv_create(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1406 {
1407 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1408 struct mbuf *nam = nfsd->nd_nam;
1409 char *dpos = nfsd->nd_dpos;
1410 kauth_cred_t cred = nfsd->nd_cr;
1411 struct nfs_fattr *fp;
1412 struct vattr va, dirfor, diraft;
1413 struct nfsv2_sattr *sp;
1414 u_int32_t *tl;
1415 struct nameidata nd;
1416 char *cp;
1417 int32_t t1;
1418 char *bpos;
1419 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1420 int rdev = 0, abort = 0;
1421 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1422 char *cp2;
1423 struct mbuf *mb, *mreq;
1424 struct vnode *vp = NULL, *dirp = NULL;
1425 nfsrvfh_t nsfh;
1426 u_quad_t frev, tempsize;
1427 u_char cverf[NFSX_V3CREATEVERF];
1428
1429 nd.ni_cnd.cn_nameiop = 0;
1430 nfsm_srvmtofh(&nsfh);
1431 nfsm_srvnamesiz(len);
1432 nd.ni_cnd.cn_cred = cred;
1433 nd.ni_cnd.cn_nameiop = CREATE;
1434 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1435 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
1436 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
1437 if (dirp && v3) {
1438 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
1439 }
1440 if (error) {
1441 nfsm_reply(NFSX_WCCDATA(v3));
1442 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1443 if (dirp)
1444 vrele(dirp);
1445 if (nd.ni_pathbuf != NULL) {
1446 pathbuf_destroy(nd.ni_pathbuf);
1447 nd.ni_pathbuf = NULL;
1448 }
1449 return (0);
1450 }
1451 abort = 1;
1452 vattr_null(&va);
1453 if (v3) {
1454 va.va_mode = 0;
1455 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1456 how = fxdr_unsigned(int, *tl);
1457 switch (how) {
1458 case NFSV3CREATE_GUARDED:
1459 if (nd.ni_vp) {
1460 error = EEXIST;
1461 break;
1462 }
1463 case NFSV3CREATE_UNCHECKED:
1464 nfsm_srvsattr(&va);
1465 break;
1466 case NFSV3CREATE_EXCLUSIVE:
1467 nfsm_dissect(cp, void *, NFSX_V3CREATEVERF);
1468 memcpy(cverf, cp, NFSX_V3CREATEVERF);
1469 exclusive_flag = 1;
1470 break;
1471 };
1472 va.va_type = VREG;
1473 } else {
1474 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1475 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1476 if (va.va_type == VNON)
1477 va.va_type = VREG;
1478 va.va_mode = nfstov_mode(sp->sa_mode);
1479 switch (va.va_type) {
1480 case VREG:
1481 tsize = fxdr_unsigned(int32_t, sp->sa_size);
1482 if (tsize != -1)
1483 va.va_size = (u_quad_t)tsize;
1484 break;
1485 case VCHR:
1486 case VBLK:
1487 case VFIFO:
1488 rdev = fxdr_unsigned(int32_t, sp->sa_size);
1489 break;
1490 default:
1491 break;
1492 };
1493 }
1494
1495 /*
1496 * Iff doesn't exist, create it
1497 * otherwise just truncate to 0 length
1498 * should I set the mode too ??
1499 */
1500 if (nd.ni_vp == NULL) {
1501 if (va.va_type == VREG || va.va_type == VSOCK) {
1502 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1503 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1504 if (!error) {
1505 if (exclusive_flag) {
1506 exclusive_flag = 0;
1507 vattr_null(&va);
1508 /*
1509 * XXX
1510 * assuming NFSX_V3CREATEVERF
1511 * == sizeof(nfstime3)
1512 */
1513 fxdr_nfsv3time(cverf, &va.va_atime);
1514 error = VOP_SETATTR(nd.ni_vp, &va,
1515 cred);
1516 }
1517 }
1518 } else if (va.va_type == VCHR || va.va_type == VBLK ||
1519 va.va_type == VFIFO) {
1520 if (va.va_type == VCHR && rdev == 0xffffffff)
1521 va.va_type = VFIFO;
1522 if (va.va_type != VFIFO &&
1523 (error = kauth_authorize_system(cred,
1524 KAUTH_SYSTEM_MKNOD, 0, NULL, NULL, NULL))) {
1525 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1526 vput(nd.ni_dvp);
1527 abort = 0;
1528 nfsm_reply(0);
1529 if (nd.ni_pathbuf != NULL) {
1530 pathbuf_destroy(nd.ni_pathbuf);
1531 nd.ni_pathbuf = NULL;
1532 }
1533 return (error);
1534 } else
1535 va.va_rdev = (dev_t)rdev;
1536 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1537 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1538 &va);
1539 if (error) {
1540 nfsm_reply(0);
1541 }
1542 } else {
1543 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1544 if (nd.ni_pathbuf != NULL) {
1545 pathbuf_destroy(nd.ni_pathbuf);
1546 nd.ni_pathbuf = NULL;
1547 }
1548 vput(nd.ni_dvp);
1549 error = ENXIO;
1550 abort = 0;
1551 }
1552 vp = nd.ni_vp;
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 vp = nd.ni_vp;
1560 if (nd.ni_dvp == vp)
1561 vrele(nd.ni_dvp);
1562 else
1563 vput(nd.ni_dvp);
1564 abort = 0;
1565 if (!error && va.va_size != -1) {
1566 error = nfsrv_access(vp, VWRITE, cred,
1567 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0);
1568 if (!error) {
1569 nqsrv_getl(vp, ND_WRITE);
1570 tempsize = va.va_size;
1571 vattr_null(&va);
1572 va.va_size = tempsize;
1573 error = VOP_SETATTR(vp, &va, cred);
1574 }
1575 }
1576 if (error)
1577 vput(vp);
1578 }
1579 if (!error) {
1580 error = nfsrv_composefh(vp, &nsfh, v3);
1581 if (!error)
1582 error = VOP_GETATTR(vp, &va, cred);
1583 vput(vp);
1584 }
1585 if (v3) {
1586 if (exclusive_flag && !error) {
1587 /*
1588 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3)
1589 */
1590 char oldverf[NFSX_V3CREATEVERF];
1591
1592 txdr_nfsv3time(&va.va_atime, oldverf);
1593 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF))
1594 error = EEXIST;
1595 }
1596 if (dirp) {
1597 vn_lock(dirp, LK_SHARED | LK_RETRY);
1598 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
1599 VOP_UNLOCK(dirp);
1600 }
1601 }
1602 if (dirp) {
1603 vrele(dirp);
1604 dirp = NULL;
1605 }
1606 if (nd.ni_pathbuf != NULL) {
1607 pathbuf_destroy(nd.ni_pathbuf);
1608 nd.ni_pathbuf = NULL;
1609 }
1610 abort = 0;
1611 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1612 if (v3) {
1613 if (!error) {
1614 nfsm_srvpostop_fh(&nsfh);
1615 nfsm_srvpostop_attr(0, &va);
1616 }
1617 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1618 } else {
1619 nfsm_srvfhtom(&nsfh, v3);
1620 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1621 nfsm_srvfillattr(&va, fp);
1622 }
1623 return (0);
1624 nfsmout:
1625 if (dirp)
1626 vrele(dirp);
1627 if (abort) {
1628 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1629 if (nd.ni_dvp == nd.ni_vp)
1630 vrele(nd.ni_dvp);
1631 else
1632 vput(nd.ni_dvp);
1633 if (nd.ni_vp)
1634 vput(nd.ni_vp);
1635 }
1636 if (nd.ni_pathbuf != NULL) {
1637 pathbuf_destroy(nd.ni_pathbuf);
1638 nd.ni_pathbuf = NULL;
1639 }
1640 return (error);
1641 }
1642
1643 /*
1644 * nfs v3 mknod service
1645 */
1646 int
1647 nfsrv_mknod(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1648 {
1649 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1650 struct mbuf *nam = nfsd->nd_nam;
1651 char *dpos = nfsd->nd_dpos;
1652 kauth_cred_t cred = nfsd->nd_cr;
1653 struct vattr va, dirfor, diraft;
1654 u_int32_t *tl;
1655 struct nameidata nd;
1656 int32_t t1;
1657 char *bpos;
1658 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
1659 int abort = 0;
1660 u_int32_t major, minor;
1661 enum vtype vtyp;
1662 char *cp2;
1663 struct mbuf *mb, *mreq;
1664 struct vnode *vp, *dirp = (struct vnode *)0;
1665 nfsrvfh_t nsfh;
1666 u_quad_t frev;
1667
1668 nd.ni_cnd.cn_nameiop = 0;
1669 nfsm_srvmtofh(&nsfh);
1670 nfsm_srvnamesiz(len);
1671 nd.ni_cnd.cn_cred = cred;
1672 nd.ni_cnd.cn_nameiop = CREATE;
1673 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1674 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
1675 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
1676 if (dirp)
1677 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
1678 if (error) {
1679 nfsm_reply(NFSX_WCCDATA(1));
1680 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1681 if (dirp)
1682 vrele(dirp);
1683 if (nd.ni_pathbuf != NULL) {
1684 pathbuf_destroy(nd.ni_pathbuf);
1685 nd.ni_pathbuf = NULL;
1686 }
1687 return (0);
1688 }
1689 abort = 1;
1690 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1691 vtyp = nfsv3tov_type(*tl);
1692 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1693 error = NFSERR_BADTYPE;
1694 goto abort;
1695 }
1696 vattr_null(&va);
1697 va.va_mode = 0;
1698 nfsm_srvsattr(&va);
1699 if (vtyp == VCHR || vtyp == VBLK) {
1700 dev_t rdev;
1701
1702 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1703 major = fxdr_unsigned(u_int32_t, *tl++);
1704 minor = fxdr_unsigned(u_int32_t, *tl);
1705 rdev = makedev(major, minor);
1706 if (major(rdev) != major || minor(rdev) != minor) {
1707 error = EINVAL;
1708 goto abort;
1709 }
1710 va.va_rdev = rdev;
1711 }
1712
1713 /*
1714 * Iff doesn't exist, create it.
1715 */
1716 if (nd.ni_vp) {
1717 error = EEXIST;
1718 abort:
1719 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1720 if (nd.ni_dvp == nd.ni_vp)
1721 vrele(nd.ni_dvp);
1722 else
1723 vput(nd.ni_dvp);
1724 if (nd.ni_vp)
1725 vput(nd.ni_vp);
1726 if (nd.ni_pathbuf != NULL) {
1727 pathbuf_destroy(nd.ni_pathbuf);
1728 nd.ni_pathbuf = NULL;
1729 }
1730 goto out;
1731 }
1732 va.va_type = vtyp;
1733 if (vtyp == VSOCK) {
1734 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1735 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1736 } else {
1737 if (va.va_type != VFIFO &&
1738 (error = kauth_authorize_system(cred,
1739 KAUTH_SYSTEM_MKNOD, 0, NULL, NULL, NULL))) {
1740 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1741 vput(nd.ni_dvp);
1742 goto out;
1743 }
1744 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1745 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1746 if (error)
1747 goto out;
1748 }
1749 out:
1750 vp = nd.ni_vp;
1751 if (!error) {
1752 error = nfsrv_composefh(vp, &nsfh, true);
1753 if (!error)
1754 error = VOP_GETATTR(vp, &va, cred);
1755 vput(vp);
1756 }
1757 if (dirp) {
1758 vn_lock(dirp, LK_SHARED | LK_RETRY);
1759 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
1760 VOP_UNLOCK(dirp);
1761 vrele(dirp);
1762 dirp = NULL;
1763 }
1764 if (nd.ni_pathbuf != NULL) {
1765 pathbuf_destroy(nd.ni_pathbuf);
1766 nd.ni_pathbuf = NULL;
1767 }
1768 abort = 0;
1769 nfsm_reply(NFSX_SRVFH(&nsfh, true) + NFSX_POSTOPATTR(1) +
1770 NFSX_WCCDATA(1));
1771 if (!error) {
1772 nfsm_srvpostop_fh(&nsfh);
1773 nfsm_srvpostop_attr(0, &va);
1774 }
1775 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1776 return (0);
1777 nfsmout:
1778 if (abort) {
1779 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1780 if (nd.ni_dvp == nd.ni_vp)
1781 vrele(nd.ni_dvp);
1782 else
1783 vput(nd.ni_dvp);
1784 if (nd.ni_vp)
1785 vput(nd.ni_vp);
1786 }
1787 if (nd.ni_pathbuf != NULL) {
1788 pathbuf_destroy(nd.ni_pathbuf);
1789 nd.ni_pathbuf = NULL;
1790 }
1791 if (dirp)
1792 vrele(dirp);
1793 return (error);
1794 }
1795
1796 /*
1797 * nfs remove service
1798 */
1799 int
1800 nfsrv_remove(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1801 {
1802 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1803 struct mbuf *nam = nfsd->nd_nam;
1804 char *dpos = nfsd->nd_dpos;
1805 kauth_cred_t cred = nfsd->nd_cr;
1806 struct nameidata nd;
1807 u_int32_t *tl;
1808 int32_t t1;
1809 char *bpos;
1810 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
1811 int v3 = (nfsd->nd_flag & ND_NFSV3);
1812 char *cp2;
1813 struct mbuf *mb, *mreq;
1814 struct vnode *vp, *dirp;
1815 struct vattr dirfor, diraft;
1816 nfsrvfh_t nsfh;
1817 u_quad_t frev;
1818
1819 #ifndef nolint
1820 vp = (struct vnode *)0;
1821 #endif
1822 nfsm_srvmtofh(&nsfh);
1823 nfsm_srvnamesiz(len);
1824 nd.ni_cnd.cn_cred = cred;
1825 nd.ni_cnd.cn_nameiop = DELETE;
1826 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1827 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
1828 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
1829 if (error == 0 && dirp && v3) {
1830 if (nd.ni_dvp == nd.ni_vp)
1831 vn_lock(dirp, LK_SHARED | LK_RETRY);
1832 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
1833 if (nd.ni_dvp == nd.ni_vp)
1834 VOP_UNLOCK(dirp);
1835 }
1836 if (!error) {
1837 vp = nd.ni_vp;
1838 if (vp->v_type == VDIR) {
1839 error = EPERM;
1840 goto out;
1841 }
1842 /*
1843 * The root of a mounted filesystem cannot be deleted.
1844 */
1845 if (vp->v_vflag & VV_ROOT) {
1846 error = EBUSY;
1847 }
1848 out:
1849 if (!error) {
1850 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1851 nqsrv_getl(vp, ND_WRITE);
1852 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1853 } else {
1854 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1855 if (nd.ni_dvp == vp)
1856 vrele(nd.ni_dvp);
1857 else
1858 vput(nd.ni_dvp);
1859 vput(vp);
1860 }
1861 }
1862 if (nd.ni_pathbuf != NULL) {
1863 pathbuf_destroy(nd.ni_pathbuf);
1864 nd.ni_pathbuf = NULL;
1865 }
1866 if (dirp) {
1867 if (v3) {
1868 vn_lock(dirp, LK_SHARED | LK_RETRY);
1869 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
1870 VOP_UNLOCK(dirp);
1871 }
1872 vrele(dirp);
1873 }
1874 nfsm_reply(NFSX_WCCDATA(v3));
1875 if (v3) {
1876 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1877 return (0);
1878 }
1879 nfsm_srvdone;
1880 }
1881
1882 /*
1883 * nfs rename service
1884 */
1885 int
1886 nfsrv_rename(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
1887 {
1888 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1889 struct mbuf *nam = nfsd->nd_nam;
1890 char *dpos = nfsd->nd_dpos;
1891 kauth_cred_t cred = nfsd->nd_cr;
1892 u_int32_t *tl;
1893 int32_t t1;
1894 char *bpos;
1895 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1;
1896 uint32_t len, len2;
1897 int tdirfor_ret = 1, tdiraft_ret = 1;
1898 int v3 = (nfsd->nd_flag & ND_NFSV3);
1899 char *cp2;
1900 struct mbuf *mb, *mreq;
1901 struct nameidata fromnd, tond;
1902 struct vnode *fvp, *tvp, *tdvp;
1903 struct vnode *fdirp = NULL, *tdirp = NULL;
1904 struct mount *localfs = NULL;
1905 struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1906 nfsrvfh_t fnsfh, tnsfh;
1907 u_quad_t frev;
1908 uid_t saved_uid;
1909
1910 #ifndef nolint
1911 fvp = (struct vnode *)0;
1912 #endif
1913 fromnd.ni_cnd.cn_nameiop = 0;
1914 tond.ni_cnd.cn_nameiop = 0;
1915 nfsm_srvmtofh(&fnsfh);
1916 nfsm_srvnamesiz(len);
1917 /*
1918 * Remember our original uid so that we can reset cr_uid before
1919 * the second nfs_namei() call, in case it is remapped.
1920 */
1921 saved_uid = kauth_cred_geteuid(cred);
1922 fromnd.ni_cnd.cn_cred = cred;
1923 fromnd.ni_cnd.cn_nameiop = DELETE;
1924 fromnd.ni_cnd.cn_flags = LOCKPARENT | INRENAME;
1925 error = nfs_namei(&fromnd, &fnsfh, len, slp, nam, &md,
1926 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
1927 if (error == 0 && fdirp && v3) {
1928 if (fromnd.ni_dvp == fromnd.ni_vp)
1929 vn_lock(fdirp, LK_SHARED | LK_RETRY);
1930 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred);
1931 if (fromnd.ni_dvp == fromnd.ni_vp)
1932 VOP_UNLOCK(fdirp);
1933 }
1934 if (error) {
1935 fromnd.ni_cnd.cn_nameiop = 0;
1936 nfsm_reply(2 * NFSX_WCCDATA(v3));
1937 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1938 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1939 if (fdirp)
1940 vrele(fdirp);
1941 if (fromnd.ni_pathbuf != NULL) {
1942 pathbuf_destroy(fromnd.ni_pathbuf);
1943 }
1944 return (0);
1945 }
1946 if (fromnd.ni_dvp != fromnd.ni_vp) {
1947 VOP_UNLOCK(fromnd.ni_dvp);
1948 }
1949 fvp = fromnd.ni_vp;
1950
1951 localfs = fvp->v_mount;
1952 error = VFS_RENAMELOCK_ENTER(localfs);
1953 if (error) {
1954 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1955 vrele(fromnd.ni_dvp);
1956 vrele(fvp);
1957 goto out1;
1958 }
1959
1960 /* Copied, regrettably, from vfs_syscalls.c (q.v.) */
1961 vrele(fvp);
1962 if ((fromnd.ni_cnd.cn_namelen == 1 &&
1963 fromnd.ni_cnd.cn_nameptr[0] == '.') ||
1964 (fromnd.ni_cnd.cn_namelen == 2 &&
1965 fromnd.ni_cnd.cn_nameptr[0] == '.' &&
1966 fromnd.ni_cnd.cn_nameptr[1] == '.')) {
1967 error = EINVAL;
1968 VFS_RENAMELOCK_EXIT(localfs);
1969 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1970 vrele(fromnd.ni_dvp);
1971 goto out1;
1972 }
1973 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
1974 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0);
1975 if (error) {
1976 VOP_UNLOCK(fromnd.ni_dvp);
1977 VFS_RENAMELOCK_EXIT(localfs);
1978 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1979 vrele(fromnd.ni_dvp);
1980 goto out1;
1981 }
1982 VOP_UNLOCK(fromnd.ni_vp);
1983 if (fromnd.ni_dvp != fromnd.ni_vp)
1984 VOP_UNLOCK(fromnd.ni_dvp);
1985 fvp = fromnd.ni_vp;
1986
1987 nfsm_srvmtofh(&tnsfh);
1988 if (v3) {
1989 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
1990 len2 = fxdr_unsigned(uint32_t, *tl);
1991 /* len2 will be checked by nfs_namei */
1992 }
1993 else {
1994 /* NFSv2 */
1995 nfsm_strsiz(len2, NFS_MAXNAMLEN);
1996 }
1997 kauth_cred_seteuid(cred, saved_uid);
1998 tond.ni_cnd.cn_cred = cred;
1999 tond.ni_cnd.cn_nameiop = RENAME;
2000 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | INRENAME;
2001 error = nfs_namei(&tond, &tnsfh, len2, slp, nam, &md,
2002 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2003 if (tdirp && v3) {
2004 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred);
2005 }
2006 if (error) {
2007 VFS_RENAMELOCK_EXIT(localfs);
2008 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2009 vrele(fromnd.ni_dvp);
2010 vrele(fvp);
2011 goto out1;
2012 }
2013 tdvp = tond.ni_dvp;
2014 tvp = tond.ni_vp;
2015 if (tvp != NULL) {
2016 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2017 if (v3)
2018 error = EEXIST;
2019 else
2020 error = EISDIR;
2021 goto out;
2022 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2023 if (v3)
2024 error = EEXIST;
2025 else
2026 error = ENOTDIR;
2027 goto out;
2028 }
2029 if (tvp->v_type == VDIR && tvp->v_mountedhere) {
2030 if (v3)
2031 error = EXDEV;
2032 else
2033 error = ENOTEMPTY;
2034 goto out;
2035 }
2036 }
2037 if (fvp->v_type == VDIR && fvp->v_mountedhere) {
2038 if (v3)
2039 error = EXDEV;
2040 else
2041 error = ENOTEMPTY;
2042 goto out;
2043 }
2044 if (fvp->v_mount != tdvp->v_mount) {
2045 if (v3)
2046 error = EXDEV;
2047 else
2048 error = ENOTEMPTY;
2049 goto out;
2050 }
2051 if (fvp == tdvp) {
2052 if (v3)
2053 error = EINVAL;
2054 else
2055 error = ENOTEMPTY;
2056 }
2057 /*
2058 * If source is the same as the destination (that is the
2059 * same vnode with the same name in the same directory),
2060 * then there is nothing to do.
2061 */
2062 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
2063 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2064 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
2065 fromnd.ni_cnd.cn_namelen))
2066 error = -1;
2067 out:
2068 if (!error) {
2069 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
2070 nqsrv_getl(tdvp, ND_WRITE);
2071 if (tvp) {
2072 nqsrv_getl(tvp, ND_WRITE);
2073 }
2074 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2075 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2076 VFS_RENAMELOCK_EXIT(localfs);
2077 } else {
2078 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2079 if (tdvp == tvp)
2080 vrele(tdvp);
2081 else
2082 vput(tdvp);
2083 if (tvp)
2084 vput(tvp);
2085 VFS_RENAMELOCK_EXIT(localfs);
2086 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2087 vrele(fromnd.ni_dvp);
2088 vrele(fvp);
2089 if (error == -1)
2090 error = 0;
2091 }
2092 if (tond.ni_pathbuf != NULL) {
2093 pathbuf_destroy(tond.ni_pathbuf);
2094 tond.ni_pathbuf = NULL;
2095 }
2096 tond.ni_cnd.cn_nameiop = 0;
2097 out1:
2098 if (fdirp) {
2099 if (v3) {
2100 vn_lock(fdirp, LK_SHARED | LK_RETRY);
2101 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred);
2102 VOP_UNLOCK(fdirp);
2103 }
2104 vrele(fdirp);
2105 fdirp = NULL;
2106 }
2107 if (tdirp) {
2108 if (v3) {
2109 vn_lock(tdirp, LK_SHARED | LK_RETRY);
2110 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred);
2111 VOP_UNLOCK(tdirp);
2112 }
2113 vrele(tdirp);
2114 tdirp = NULL;
2115 }
2116 pathbuf_destroy(fromnd.ni_pathbuf);
2117 fromnd.ni_pathbuf = NULL;
2118 fromnd.ni_cnd.cn_nameiop = 0;
2119 localfs = NULL;
2120 nfsm_reply(2 * NFSX_WCCDATA(v3));
2121 if (v3) {
2122 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
2123 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
2124 }
2125 return (0);
2126
2127 nfsmout:
2128 if (fdirp)
2129 vrele(fdirp);
2130 #ifdef notdef
2131 if (tdirp)
2132 vrele(tdirp);
2133 #endif
2134 if (tond.ni_cnd.cn_nameiop) {
2135 if (tond.ni_pathbuf != NULL) {
2136 pathbuf_destroy(tond.ni_pathbuf);
2137 tond.ni_pathbuf = NULL;
2138 }
2139 }
2140 if (localfs) {
2141 VFS_RENAMELOCK_EXIT(localfs);
2142 }
2143 if (fromnd.ni_cnd.cn_nameiop) {
2144 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2145 if (fromnd.ni_pathbuf != NULL) {
2146 pathbuf_destroy(fromnd.ni_pathbuf);
2147 fromnd.ni_pathbuf = NULL;
2148 }
2149 vrele(fromnd.ni_dvp);
2150 vrele(fvp);
2151 }
2152 return (error);
2153 }
2154
2155 /*
2156 * nfs link service
2157 */
2158 int
2159 nfsrv_link(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2160 {
2161 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2162 struct mbuf *nam = nfsd->nd_nam;
2163 char *dpos = nfsd->nd_dpos;
2164 kauth_cred_t cred = nfsd->nd_cr;
2165 struct nameidata nd;
2166 u_int32_t *tl;
2167 int32_t t1;
2168 char *bpos;
2169 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2170 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
2171 char *cp2;
2172 struct mbuf *mb, *mreq;
2173 struct vnode *vp, *xp, *dirp = (struct vnode *)0;
2174 struct vattr dirfor, diraft, at;
2175 nfsrvfh_t nsfh, dnsfh;
2176 u_quad_t frev;
2177
2178 nfsm_srvmtofh(&nsfh);
2179 nfsm_srvmtofh(&dnsfh);
2180 nfsm_srvnamesiz(len);
2181 error = nfsrv_fhtovp(&nsfh, false, &vp, cred, slp, nam,
2182 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
2183 if (error) {
2184 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2185 nfsm_srvpostop_attr(getret, &at);
2186 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2187 return (0);
2188 }
2189 if (vp->v_type == VDIR) {
2190 error = EPERM;
2191 goto out1;
2192 }
2193 nd.ni_cnd.cn_cred = cred;
2194 nd.ni_cnd.cn_nameiop = CREATE;
2195 nd.ni_cnd.cn_flags = LOCKPARENT;
2196 error = nfs_namei(&nd, &dnsfh, len, slp, nam, &md, &dpos,
2197 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2198 if (dirp && v3) {
2199 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
2200 }
2201 if (error)
2202 goto out1;
2203 xp = nd.ni_vp;
2204 if (xp != NULL) {
2205 error = EEXIST;
2206 goto out;
2207 }
2208 xp = nd.ni_dvp;
2209 if (vp->v_mount != xp->v_mount)
2210 error = EXDEV;
2211 out:
2212 if (!error) {
2213 nqsrv_getl(vp, ND_WRITE);
2214 nqsrv_getl(xp, ND_WRITE);
2215 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2216 } else {
2217 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2218 if (nd.ni_dvp == nd.ni_vp)
2219 vrele(nd.ni_dvp);
2220 else
2221 vput(nd.ni_dvp);
2222 if (nd.ni_vp)
2223 vrele(nd.ni_vp);
2224 }
2225 out1:
2226 if (v3) {
2227 vn_lock(vp, LK_SHARED | LK_RETRY);
2228 getret = VOP_GETATTR(vp, &at, cred);
2229 VOP_UNLOCK(vp);
2230 }
2231 if (dirp) {
2232 if (v3) {
2233 vn_lock(dirp, LK_SHARED | LK_RETRY);
2234 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
2235 VOP_UNLOCK(dirp);
2236 }
2237 vrele(dirp);
2238 }
2239 vrele(vp);
2240 if (nd.ni_pathbuf != NULL) {
2241 pathbuf_destroy(nd.ni_pathbuf);
2242 nd.ni_pathbuf = NULL;
2243 }
2244 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2245 if (v3) {
2246 nfsm_srvpostop_attr(getret, &at);
2247 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2248 return (0);
2249 }
2250 nfsm_srvdone;
2251 }
2252
2253 /*
2254 * nfs symbolic link service
2255 */
2256 int
2257 nfsrv_symlink(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2258 {
2259 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2260 struct mbuf *nam = nfsd->nd_nam;
2261 char *dpos = nfsd->nd_dpos;
2262 kauth_cred_t cred = nfsd->nd_cr;
2263 struct vattr va, dirfor, diraft;
2264 struct nameidata nd;
2265 u_int32_t *tl;
2266 int32_t t1;
2267 struct nfsv2_sattr *sp;
2268 char *bpos, *pathcp = NULL, *cp2;
2269 struct uio io;
2270 struct iovec iv;
2271 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1, abort = 0;
2272 uint32_t len, len2;
2273 int v3 = (nfsd->nd_flag & ND_NFSV3);
2274 struct mbuf *mb, *mreq;
2275 struct vnode *dirp = (struct vnode *)0;
2276 nfsrvfh_t nsfh;
2277 u_quad_t frev;
2278
2279 nd.ni_cnd.cn_nameiop = 0;
2280 nfsm_srvmtofh(&nsfh);
2281 nfsm_srvnamesiz(len);
2282 nd.ni_cnd.cn_cred = cred;
2283 nd.ni_cnd.cn_nameiop = CREATE;
2284 nd.ni_cnd.cn_flags = LOCKPARENT;
2285 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
2286 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2287 if (dirp && v3) {
2288 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
2289 }
2290 if (error)
2291 goto out;
2292 abort = 1;
2293 vattr_null(&va);
2294 va.va_type = VLNK;
2295 if (v3) {
2296 va.va_mode = 0;
2297 nfsm_srvsattr(&va);
2298 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
2299 len2 = fxdr_unsigned(uint32_t, *tl);
2300 if (len2 > PATH_MAX) {
2301 /* XXX should check _PC_NO_TRUNC */
2302 error = ENAMETOOLONG;
2303 goto abortop;
2304 }
2305 }
2306 else {
2307 /* NFSv2 */
2308 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2309 }
2310 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK);
2311 iv.iov_base = pathcp;
2312 iv.iov_len = len2;
2313 io.uio_resid = len2;
2314 io.uio_offset = 0;
2315 io.uio_iov = &iv;
2316 io.uio_iovcnt = 1;
2317 io.uio_rw = UIO_READ;
2318 UIO_SETUP_SYSSPACE(&io);
2319 nfsm_mtouio(&io, len2);
2320 if (!v3) {
2321 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2322 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2323 }
2324 *(pathcp + len2) = '\0';
2325 if (nd.ni_vp) {
2326 error = EEXIST;
2327 abortop:
2328 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2329 if (nd.ni_dvp == nd.ni_vp)
2330 vrele(nd.ni_dvp);
2331 else
2332 vput(nd.ni_dvp);
2333 if (nd.ni_vp)
2334 vrele(nd.ni_vp);
2335 goto out;
2336 }
2337 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2338 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2339 if (!error) {
2340 if (v3) {
2341 error = nfsrv_composefh(nd.ni_vp, &nsfh, v3);
2342 if (!error)
2343 error = VOP_GETATTR(nd.ni_vp, &va, cred);
2344 vput(nd.ni_vp);
2345 } else {
2346 vput(nd.ni_vp);
2347 }
2348 }
2349 out:
2350 if (pathcp)
2351 free(pathcp, M_TEMP);
2352 if (dirp) {
2353 if (v3) {
2354 vn_lock(dirp, LK_SHARED | LK_RETRY);
2355 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
2356 VOP_UNLOCK(dirp);
2357 }
2358 vrele(dirp);
2359 dirp = NULL;
2360 }
2361 if (nd.ni_pathbuf != NULL) {
2362 pathbuf_destroy(nd.ni_pathbuf);
2363 nd.ni_pathbuf = NULL;
2364 }
2365 abort = 0;
2366 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) +
2367 NFSX_WCCDATA(v3));
2368 if (v3) {
2369 if (!error) {
2370 nfsm_srvpostop_fh(&nsfh);
2371 nfsm_srvpostop_attr(0, &va);
2372 }
2373 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2374 }
2375 return (0);
2376 nfsmout:
2377 if (abort) {
2378 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2379 if (nd.ni_dvp == nd.ni_vp)
2380 vrele(nd.ni_dvp);
2381 else
2382 vput(nd.ni_dvp);
2383 if (nd.ni_vp)
2384 vrele(nd.ni_vp);
2385 if (nd.ni_pathbuf != NULL) {
2386 pathbuf_destroy(nd.ni_pathbuf);
2387 nd.ni_pathbuf = NULL;
2388 }
2389 }
2390 if (dirp)
2391 vrele(dirp);
2392 if (pathcp)
2393 free(pathcp, M_TEMP);
2394 return (error);
2395 }
2396
2397 /*
2398 * nfs mkdir service
2399 */
2400 int
2401 nfsrv_mkdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2402 {
2403 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2404 struct mbuf *nam = nfsd->nd_nam;
2405 char *dpos = nfsd->nd_dpos;
2406 kauth_cred_t cred = nfsd->nd_cr;
2407 struct vattr va, dirfor, diraft;
2408 struct nfs_fattr *fp;
2409 struct nameidata nd;
2410 char *cp;
2411 u_int32_t *tl;
2412 int32_t t1;
2413 char *bpos;
2414 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2415 int abort = 0;
2416 int v3 = (nfsd->nd_flag & ND_NFSV3);
2417 char *cp2;
2418 struct mbuf *mb, *mreq;
2419 struct vnode *vp, *dirp = (struct vnode *)0;
2420 nfsrvfh_t nsfh;
2421 u_quad_t frev;
2422
2423 nfsm_srvmtofh(&nsfh);
2424 nfsm_srvnamesiz(len);
2425 nd.ni_cnd.cn_cred = cred;
2426 nd.ni_cnd.cn_nameiop = CREATE;
2427 nd.ni_cnd.cn_flags = LOCKPARENT;
2428 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
2429 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2430 if (dirp && v3) {
2431 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
2432 }
2433 if (error) {
2434 if (nd.ni_pathbuf != NULL) {
2435 pathbuf_destroy(nd.ni_pathbuf);
2436 nd.ni_pathbuf = NULL;
2437 }
2438 nfsm_reply(NFSX_WCCDATA(v3));
2439 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2440 if (dirp)
2441 vrele(dirp);
2442 return (0);
2443 }
2444 abort = 1;
2445 vattr_null(&va);
2446 if (v3) {
2447 va.va_mode = 0;
2448 nfsm_srvsattr(&va);
2449 } else {
2450 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2451 va.va_mode = nfstov_mode(*tl++);
2452 }
2453 va.va_type = VDIR;
2454 vp = nd.ni_vp;
2455 if (vp != NULL) {
2456 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2457 if (nd.ni_dvp == vp)
2458 vrele(nd.ni_dvp);
2459 else
2460 vput(nd.ni_dvp);
2461 vrele(vp);
2462 error = EEXIST;
2463 goto out;
2464 }
2465 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2466 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2467 if (!error) {
2468 vp = nd.ni_vp;
2469 error = nfsrv_composefh(vp, &nsfh, v3);
2470 if (!error)
2471 error = VOP_GETATTR(vp, &va, cred);
2472 vput(vp);
2473 }
2474 out:
2475 if (dirp) {
2476 if (v3) {
2477 vn_lock(dirp, LK_SHARED | LK_RETRY);
2478 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
2479 VOP_UNLOCK(dirp);
2480 }
2481 vrele(dirp);
2482 dirp = NULL;
2483 }
2484 if (nd.ni_pathbuf != NULL) {
2485 pathbuf_destroy(nd.ni_pathbuf);
2486 nd.ni_pathbuf = NULL;
2487 }
2488 abort = 0;
2489 nfsm_reply(NFSX_SRVFH(&nsfh, v3) + NFSX_POSTOPATTR(v3) +
2490 NFSX_WCCDATA(v3));
2491 if (v3) {
2492 if (!error) {
2493 nfsm_srvpostop_fh(&nsfh);
2494 nfsm_srvpostop_attr(0, &va);
2495 }
2496 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2497 } else {
2498 nfsm_srvfhtom(&nsfh, v3);
2499 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2500 nfsm_srvfillattr(&va, fp);
2501 }
2502 return (0);
2503 nfsmout:
2504 if (abort) {
2505 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2506 if (nd.ni_dvp == nd.ni_vp)
2507 vrele(nd.ni_dvp);
2508 else
2509 vput(nd.ni_dvp);
2510 if (nd.ni_vp)
2511 vrele(nd.ni_vp);
2512 if (nd.ni_pathbuf != NULL) {
2513 pathbuf_destroy(nd.ni_pathbuf);
2514 nd.ni_pathbuf = NULL;
2515 }
2516 }
2517 if (dirp)
2518 vrele(dirp);
2519 return (error);
2520 }
2521
2522 /*
2523 * nfs rmdir service
2524 */
2525 int
2526 nfsrv_rmdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2527 {
2528 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2529 struct mbuf *nam = nfsd->nd_nam;
2530 char *dpos = nfsd->nd_dpos;
2531 kauth_cred_t cred = nfsd->nd_cr;
2532 u_int32_t *tl;
2533 int32_t t1;
2534 char *bpos;
2535 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2536 int v3 = (nfsd->nd_flag & ND_NFSV3);
2537 char *cp2;
2538 struct mbuf *mb, *mreq;
2539 struct vnode *vp, *dirp = (struct vnode *)0;
2540 struct vattr dirfor, diraft;
2541 nfsrvfh_t nsfh;
2542 struct nameidata nd;
2543 u_quad_t frev;
2544
2545 nfsm_srvmtofh(&nsfh);
2546 nfsm_srvnamesiz(len);
2547 nd.ni_cnd.cn_cred = cred;
2548 nd.ni_cnd.cn_nameiop = DELETE;
2549 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2550 error = nfs_namei(&nd, &nsfh, len, slp, nam, &md, &dpos,
2551 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), false);
2552 if (dirp && v3) {
2553 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred);
2554 }
2555 if (error) {
2556 if (nd.ni_pathbuf != NULL) {
2557 pathbuf_destroy(nd.ni_pathbuf);
2558 nd.ni_pathbuf = NULL;
2559 }
2560 nfsm_reply(NFSX_WCCDATA(v3));
2561 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2562 if (dirp)
2563 vrele(dirp);
2564 return (0);
2565 }
2566 vp = nd.ni_vp;
2567 if (vp->v_type != VDIR) {
2568 error = ENOTDIR;
2569 goto out;
2570 }
2571 /*
2572 * No rmdir "." please.
2573 */
2574 if (nd.ni_dvp == vp) {
2575 error = EINVAL;
2576 goto out;
2577 }
2578 /*
2579 * The root of a mounted filesystem cannot be deleted.
2580 */
2581 if (vp->v_vflag & VV_ROOT)
2582 error = EBUSY;
2583 out:
2584 if (!error) {
2585 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2586 nqsrv_getl(vp, ND_WRITE);
2587 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2588 } else {
2589 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2590 if (nd.ni_dvp == nd.ni_vp)
2591 vrele(nd.ni_dvp);
2592 else
2593 vput(nd.ni_dvp);
2594 vput(vp);
2595 }
2596 if (nd.ni_pathbuf != NULL) {
2597 pathbuf_destroy(nd.ni_pathbuf);
2598 nd.ni_pathbuf = NULL;
2599 }
2600 if (dirp) {
2601 if (v3) {
2602 vn_lock(dirp, LK_SHARED | LK_RETRY);
2603 diraft_ret = VOP_GETATTR(dirp, &diraft, cred);
2604 VOP_UNLOCK(dirp);
2605 }
2606 vrele(dirp);
2607 }
2608 nfsm_reply(NFSX_WCCDATA(v3));
2609 if (v3) {
2610 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2611 return (0);
2612 }
2613 nfsm_srvdone;
2614 }
2615
2616 /*
2617 * nfs readdir service
2618 * - mallocs what it thinks is enough to read
2619 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR
2620 * - calls VOP_READDIR()
2621 * - loops around building the reply
2622 * if the output generated exceeds count break out of loop
2623 * The nfsm_clget macro is used here so that the reply will be packed
2624 * tightly in mbuf clusters.
2625 * - it only knows that it has encountered eof when the VOP_READDIR()
2626 * reads nothing
2627 * - as such one readdir rpc will return eof false although you are there
2628 * and then the next will return eof
2629 * - it trims out records with d_fileno == 0
2630 * this doesn't matter for Unix clients, but they might confuse clients
2631 * for other os'.
2632 * - it trims out records with d_type == DT_WHT
2633 * these cannot be seen through NFS (unless we extend the protocol)
2634 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2635 * than requested, but this may not apply to all filesystems. For
2636 * example, client NFS does not { although it is never remote mounted
2637 * anyhow }
2638 * The alternate call nfsrv_readdirplus() does lookups as well.
2639 * PS: The NFS protocol spec. does not clarify what the "count" byte
2640 * argument is a count of.. just name strings and file id's or the
2641 * entire reply rpc or ...
2642 * I tried just file name and id sizes and it confused the Sun client,
2643 * so I am using the full rpc size now. The "paranoia.." comment refers
2644 * to including the status longwords that are not a part of the dir.
2645 * "entry" structures, but are in the rpc.
2646 */
2647
2648 #define NFS_SRVDIRBLKSIZ 1024
2649
2650 struct flrep {
2651 nfsuint64 fl_off;
2652 u_int32_t fl_postopok;
2653 struct nfs_fattr fl_fattr; /* XXX: must be of fattr3 size */
2654 u_int32_t fl_fhok;
2655 u_int32_t fl_fhsize;
2656 /* handle comes here, filled in dynamically */
2657 };
2658
2659 int
2660 nfsrv_readdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2661 {
2662 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2663 struct mbuf *nam = nfsd->nd_nam;
2664 char *dpos = nfsd->nd_dpos;
2665 kauth_cred_t cred = nfsd->nd_cr;
2666 char *bp, *be;
2667 struct mbuf *mp;
2668 struct dirent *dp;
2669 char *cp;
2670 u_int32_t *tl;
2671 int32_t t1;
2672 char *bpos;
2673 struct mbuf *mb, *mreq, *mp2;
2674 char *cpos, *cend, *cp2, *rbuf;
2675 struct vnode *vp;
2676 struct vattr at;
2677 nfsrvfh_t nsfh;
2678 struct uio io;
2679 struct iovec iv;
2680 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2681 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies;
2682 int v3 = (nfsd->nd_flag & ND_NFSV3);
2683 u_quad_t frev, off, toff, verf;
2684 off_t *cookies = NULL, *cookiep;
2685 nfsuint64 jar;
2686
2687 nfsm_srvmtofh(&nsfh);
2688 if (v3) {
2689 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2690 toff = fxdr_hyper(tl);
2691 tl += 2;
2692 verf = fxdr_hyper(tl);
2693 tl += 2;
2694 } else {
2695 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2696 toff = fxdr_unsigned(u_quad_t, *tl++);
2697 }
2698 off = toff;
2699 cnt = fxdr_unsigned(int, *tl);
2700 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1));
2701 xfer = NFS_SRVMAXDATA(nfsd);
2702 if (siz > xfer)
2703 siz = xfer;
2704 fullsiz = siz;
2705 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
2706 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
2707 if (!error && vp->v_type != VDIR) {
2708 error = ENOTDIR;
2709 vput(vp);
2710 }
2711 if (error) {
2712 nfsm_reply(NFSX_UNSIGNED);
2713 nfsm_srvpostop_attr(getret, &at);
2714 return (0);
2715 }
2716 nqsrv_getl(vp, ND_READ);
2717 if (v3) {
2718 error = getret = VOP_GETATTR(vp, &at, cred);
2719 #ifdef NFS3_STRICTVERF
2720 /*
2721 * XXX This check is too strict for Solaris 2.5 clients.
2722 */
2723 if (!error && toff && verf != at.va_filerev)
2724 error = NFSERR_BAD_COOKIE;
2725 #endif
2726 }
2727 if (!error)
2728 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0);
2729 if (error) {
2730 vput(vp);
2731 nfsm_reply(NFSX_POSTOPATTR(v3));
2732 nfsm_srvpostop_attr(getret, &at);
2733 return (0);
2734 }
2735 VOP_UNLOCK(vp);
2736 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2737 again:
2738 iv.iov_base = rbuf;
2739 iv.iov_len = fullsiz;
2740 io.uio_iov = &iv;
2741 io.uio_iovcnt = 1;
2742 io.uio_offset = (off_t)off;
2743 io.uio_resid = fullsiz;
2744 io.uio_rw = UIO_READ;
2745 UIO_SETUP_SYSSPACE(&io);
2746 eofflag = 0;
2747 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2748
2749 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2750
2751 off = (off_t)io.uio_offset;
2752 if (!cookies && !error)
2753 error = NFSERR_PERM;
2754 if (v3) {
2755 getret = VOP_GETATTR(vp, &at, cred);
2756 if (!error)
2757 error = getret;
2758 }
2759
2760 VOP_UNLOCK(vp);
2761 if (error) {
2762 vrele(vp);
2763 free((void *)rbuf, M_TEMP);
2764 if (cookies)
2765 free((void *)cookies, M_TEMP);
2766 nfsm_reply(NFSX_POSTOPATTR(v3));
2767 nfsm_srvpostop_attr(getret, &at);
2768 return (0);
2769 }
2770 if (io.uio_resid) {
2771 siz -= io.uio_resid;
2772
2773 /*
2774 * If nothing read, return eof
2775 * rpc reply
2776 */
2777 if (siz == 0) {
2778 vrele(vp);
2779 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2780 2 * NFSX_UNSIGNED);
2781 if (v3) {
2782 nfsm_srvpostop_attr(getret, &at);
2783 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2784 txdr_hyper(at.va_filerev, tl);
2785 tl += 2;
2786 } else
2787 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2788 *tl++ = nfs_false;
2789 *tl = nfs_true;
2790 free((void *)rbuf, M_TEMP);
2791 free((void *)cookies, M_TEMP);
2792 return (0);
2793 }
2794 }
2795
2796 /*
2797 * Check for degenerate cases of nothing useful read.
2798 * If so go try again
2799 */
2800 cpos = rbuf;
2801 cend = rbuf + siz;
2802 dp = (struct dirent *)cpos;
2803 cookiep = cookies;
2804
2805 while (cpos < cend && ncookies > 0 &&
2806 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2807 cpos += dp->d_reclen;
2808 dp = (struct dirent *)cpos;
2809 cookiep++;
2810 ncookies--;
2811 }
2812 if (cpos >= cend || ncookies == 0) {
2813 toff = off;
2814 siz = fullsiz;
2815 free(cookies, M_TEMP);
2816 cookies = NULL;
2817 goto again;
2818 }
2819
2820 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
2821 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2822 if (v3) {
2823 nfsm_srvpostop_attr(getret, &at);
2824 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2825 txdr_hyper(at.va_filerev, tl);
2826 }
2827 mp = mp2 = mb;
2828 bp = bpos;
2829 be = bp + M_TRAILINGSPACE(mp);
2830
2831 /* Loop through the records and build reply */
2832 while (cpos < cend && ncookies > 0) {
2833 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2834 nlen = dp->d_namlen;
2835 rem = nfsm_rndup(nlen)-nlen;
2836 len += (4 * NFSX_UNSIGNED + nlen + rem);
2837 if (v3)
2838 len += 2 * NFSX_UNSIGNED;
2839 if (len > cnt) {
2840 eofflag = 0;
2841 break;
2842 }
2843 /*
2844 * Build the directory record xdr from
2845 * the dirent entry.
2846 */
2847 nfsm_clget;
2848 *tl = nfs_true;
2849 bp += NFSX_UNSIGNED;
2850 if (v3) {
2851 nfsm_clget;
2852 *tl = txdr_unsigned(dp->d_fileno >> 32);
2853 bp += NFSX_UNSIGNED;
2854 }
2855 nfsm_clget;
2856 *tl = txdr_unsigned(dp->d_fileno);
2857 bp += NFSX_UNSIGNED;
2858 nfsm_clget;
2859 *tl = txdr_unsigned(nlen);
2860 bp += NFSX_UNSIGNED;
2861
2862 /* And loop around copying the name */
2863 xfer = nlen;
2864 cp = dp->d_name;
2865 while (xfer > 0) {
2866 nfsm_clget;
2867 if ((bp+xfer) > be)
2868 tsiz = be-bp;
2869 else
2870 tsiz = xfer;
2871 memcpy(bp, cp, tsiz);
2872 bp += tsiz;
2873 xfer -= tsiz;
2874 if (xfer > 0)
2875 cp += tsiz;
2876 }
2877 /* And null pad to an int32_t boundary */
2878 for (i = 0; i < rem; i++)
2879 *bp++ = '\0';
2880 nfsm_clget;
2881
2882 /* Finish off the record */
2883 txdr_hyper(*cookiep, &jar);
2884 if (v3) {
2885 *tl = jar.nfsuquad[0];
2886 bp += NFSX_UNSIGNED;
2887 nfsm_clget;
2888 }
2889 *tl = jar.nfsuquad[1];
2890 bp += NFSX_UNSIGNED;
2891 }
2892 cpos += dp->d_reclen;
2893 dp = (struct dirent *)cpos;
2894 cookiep++;
2895 ncookies--;
2896 }
2897 vrele(vp);
2898 nfsm_clget;
2899 *tl = nfs_false;
2900 bp += NFSX_UNSIGNED;
2901 nfsm_clget;
2902 if (eofflag)
2903 *tl = nfs_true;
2904 else
2905 *tl = nfs_false;
2906 bp += NFSX_UNSIGNED;
2907 if (mp != mb) {
2908 if (bp < be)
2909 mp->m_len = bp - mtod(mp, char *);
2910 } else
2911 mp->m_len += bp - bpos;
2912 free((void *)rbuf, M_TEMP);
2913 free((void *)cookies, M_TEMP);
2914 nfsm_srvdone;
2915 }
2916
2917 int
2918 nfsrv_readdirplus(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
2919 {
2920 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2921 struct mbuf *nam = nfsd->nd_nam;
2922 char *dpos = nfsd->nd_dpos;
2923 kauth_cred_t cred = nfsd->nd_cr;
2924 char *bp, *be;
2925 struct mbuf *mp;
2926 struct dirent *dp;
2927 char *cp;
2928 u_int32_t *tl;
2929 int32_t t1;
2930 char *bpos;
2931 struct mbuf *mb, *mreq, *mp2;
2932 char *cpos, *cend, *cp2, *rbuf;
2933 struct vnode *vp, *nvp;
2934 struct flrep fl;
2935 nfsrvfh_t nsfh;
2936 struct uio io;
2937 struct iovec iv;
2938 struct vattr va, at, *vap = &va;
2939 struct nfs_fattr *fp;
2940 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2941 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies;
2942 u_quad_t frev, off, toff, verf;
2943 off_t *cookies = NULL, *cookiep;
2944
2945 nfsm_srvmtofh(&nsfh);
2946 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2947 toff = fxdr_hyper(tl);
2948 tl += 2;
2949 verf = fxdr_hyper(tl);
2950 tl += 2;
2951 siz = fxdr_unsigned(int, *tl++);
2952 cnt = fxdr_unsigned(int, *tl);
2953 off = toff;
2954 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1));
2955 xfer = NFS_SRVMAXDATA(nfsd);
2956 if (siz > xfer)
2957 siz = xfer;
2958 fullsiz = siz;
2959 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
2960 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
2961 if (!error && vp->v_type != VDIR) {
2962 error = ENOTDIR;
2963 vput(vp);
2964 }
2965 if (error) {
2966 nfsm_reply(NFSX_UNSIGNED);
2967 nfsm_srvpostop_attr(getret, &at);
2968 return (0);
2969 }
2970 error = getret = VOP_GETATTR(vp, &at, cred);
2971 #ifdef NFS3_STRICTVERF
2972 /*
2973 * XXX This check is too strict for Solaris 2.5 clients.
2974 */
2975 if (!error && toff && verf != at.va_filerev)
2976 error = NFSERR_BAD_COOKIE;
2977 #endif
2978 if (!error) {
2979 nqsrv_getl(vp, ND_READ);
2980 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0);
2981 }
2982 if (error) {
2983 vput(vp);
2984 nfsm_reply(NFSX_V3POSTOPATTR);
2985 nfsm_srvpostop_attr(getret, &at);
2986 return (0);
2987 }
2988 VOP_UNLOCK(vp);
2989
2990 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2991 again:
2992 iv.iov_base = rbuf;
2993 iv.iov_len = fullsiz;
2994 io.uio_iov = &iv;
2995 io.uio_iovcnt = 1;
2996 io.uio_offset = (off_t)off;
2997 io.uio_resid = fullsiz;
2998 io.uio_rw = UIO_READ;
2999 UIO_SETUP_SYSSPACE(&io);
3000 eofflag = 0;
3001
3002 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
3003
3004 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
3005
3006 off = (u_quad_t)io.uio_offset;
3007 getret = VOP_GETATTR(vp, &at, cred);
3008
3009 VOP_UNLOCK(vp);
3010
3011 /*
3012 * If the VGET operation doesn't work for this filesystem,
3013 * we can't support readdirplus. Returning NOTSUPP should
3014 * make clients fall back to plain readdir.
3015 * There's no need to check for VPTOFH as well, we wouldn't
3016 * even be here otherwise.
3017 */
3018 if (!getret) {
3019 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
3020 getret = (getret == EOPNOTSUPP) ?
3021 NFSERR_NOTSUPP : NFSERR_IO;
3022 else
3023 vput(nvp);
3024 }
3025
3026 if (!cookies && !error)
3027 error = NFSERR_PERM;
3028 if (!error)
3029 error = getret;
3030 if (error) {
3031 vrele(vp);
3032 if (cookies)
3033 free((void *)cookies, M_TEMP);
3034 free((void *)rbuf, M_TEMP);
3035 nfsm_reply(NFSX_V3POSTOPATTR);
3036 nfsm_srvpostop_attr(getret, &at);
3037 return (0);
3038 }
3039 if (io.uio_resid) {
3040 siz -= io.uio_resid;
3041
3042 /*
3043 * If nothing read, return eof
3044 * rpc reply
3045 */
3046 if (siz == 0) {
3047 vrele(vp);
3048 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
3049 2 * NFSX_UNSIGNED);
3050 nfsm_srvpostop_attr(getret, &at);
3051 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3052 txdr_hyper(at.va_filerev, tl);
3053 tl += 2;
3054 *tl++ = nfs_false;
3055 *tl = nfs_true;
3056 free((void *)cookies, M_TEMP);
3057 free((void *)rbuf, M_TEMP);
3058 return (0);
3059 }
3060 }
3061
3062 /*
3063 * Check for degenerate cases of nothing useful read.
3064 * If so go try again
3065 */
3066 cpos = rbuf;
3067 cend = rbuf + siz;
3068 dp = (struct dirent *)cpos;
3069 cookiep = cookies;
3070
3071 while (cpos < cend && ncookies > 0 &&
3072 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
3073 cpos += dp->d_reclen;
3074 dp = (struct dirent *)cpos;
3075 cookiep++;
3076 ncookies--;
3077 }
3078 if (cpos >= cend || ncookies == 0) {
3079 toff = off;
3080 siz = fullsiz;
3081 free(cookies, M_TEMP);
3082 cookies = NULL;
3083 goto again;
3084 }
3085
3086 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
3087 nfsm_reply(cnt);
3088 nfsm_srvpostop_attr(getret, &at);
3089 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3090 txdr_hyper(at.va_filerev, tl);
3091 mp = mp2 = mb;
3092 bp = bpos;
3093 be = bp + M_TRAILINGSPACE(mp);
3094
3095 /* Loop through the records and build reply */
3096 while (cpos < cend && ncookies > 0) {
3097 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
3098 nfsrvfh_t nnsfh;
3099
3100 nlen = dp->d_namlen;
3101 rem = nfsm_rndup(nlen)-nlen;
3102
3103 /*
3104 * For readdir_and_lookup get the vnode using
3105 * the file number.
3106 */
3107 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
3108 goto invalid;
3109 if (nfsrv_composefh(nvp, &nnsfh, true)) {
3110 vput(nvp);
3111 goto invalid;
3112 }
3113 if (VOP_GETATTR(nvp, vap, cred)) {
3114 vput(nvp);
3115 goto invalid;
3116 }
3117 vput(nvp);
3118
3119 /*
3120 * If either the dircount or maxcount will be
3121 * exceeded, get out now. Both of these lengths
3122 * are calculated conservatively, including all
3123 * XDR overheads.
3124 */
3125 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
3126 NFSX_V3POSTOPATTR);
3127 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
3128 if (len > cnt || dirlen > fullsiz) {
3129 eofflag = 0;
3130 break;
3131 }
3132
3133 /*
3134 * Build the directory record xdr from
3135 * the dirent entry.
3136 */
3137 fp = (struct nfs_fattr *)&fl.fl_fattr;
3138 nfsm_srvfillattr(vap, fp);
3139 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
3140 fl.fl_fhok = nfs_true;
3141 fl.fl_postopok = nfs_true;
3142 txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
3143
3144 nfsm_clget;
3145 *tl = nfs_true;
3146 bp += NFSX_UNSIGNED;
3147 nfsm_clget;
3148 *tl = txdr_unsigned(dp->d_fileno >> 32);
3149 bp += NFSX_UNSIGNED;
3150 nfsm_clget;
3151 *tl = txdr_unsigned(dp->d_fileno);
3152 bp += NFSX_UNSIGNED;
3153 nfsm_clget;
3154 *tl = txdr_unsigned(nlen);
3155 bp += NFSX_UNSIGNED;
3156
3157 /* And loop around copying the name */
3158 xfer = nlen;
3159 cp = dp->d_name;
3160 while (xfer > 0) {
3161 nfsm_clget;
3162 if ((bp + xfer) > be)
3163 tsiz = be - bp;
3164 else
3165 tsiz = xfer;
3166 memcpy(bp, cp, tsiz);
3167 bp += tsiz;
3168 xfer -= tsiz;
3169 if (xfer > 0)
3170 cp += tsiz;
3171 }
3172 /* And null pad to an int32_t boundary */
3173 for (i = 0; i < rem; i++)
3174 *bp++ = '\0';
3175
3176 /*
3177 * Now copy the flrep structure out.
3178 */
3179 xfer = sizeof(struct flrep);
3180 cp = (void *)&fl;
3181 while (xfer > 0) {
3182 nfsm_clget;
3183 if ((bp + xfer) > be)
3184 tsiz = be - bp;
3185 else
3186 tsiz = xfer;
3187 memcpy(bp, cp, tsiz);
3188 bp += tsiz;
3189 xfer -= tsiz;
3190 if (xfer > 0)
3191 cp += tsiz;
3192 }
3193
3194 /*
3195 * ... and filehandle.
3196 */
3197 xfer = NFSRVFH_SIZE(&nnsfh);
3198 cp = NFSRVFH_DATA(&nnsfh);
3199 while (xfer > 0) {
3200 nfsm_clget;
3201 if ((bp + xfer) > be)
3202 tsiz = be - bp;
3203 else
3204 tsiz = xfer;
3205 memcpy(bp, cp, tsiz);
3206 bp += tsiz;
3207 xfer -= tsiz;
3208 if (xfer > 0)
3209 cp += tsiz;
3210 }
3211 }
3212 invalid:
3213 cpos += dp->d_reclen;
3214 dp = (struct dirent *)cpos;
3215 cookiep++;
3216 ncookies--;
3217 }
3218 vrele(vp);
3219 nfsm_clget;
3220 *tl = nfs_false;
3221 bp += NFSX_UNSIGNED;
3222 nfsm_clget;
3223 if (eofflag)
3224 *tl = nfs_true;
3225 else
3226 *tl = nfs_false;
3227 bp += NFSX_UNSIGNED;
3228 if (mp != mb) {
3229 if (bp < be)
3230 mp->m_len = bp - mtod(mp, char *);
3231 } else
3232 mp->m_len += bp - bpos;
3233 free((void *)cookies, M_TEMP);
3234 free((void *)rbuf, M_TEMP);
3235 nfsm_srvdone;
3236 }
3237
3238 /*
3239 * nfs commit service
3240 */
3241 int
3242 nfsrv_commit(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
3243 {
3244 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3245 struct mbuf *nam = nfsd->nd_nam;
3246 char *dpos = nfsd->nd_dpos;
3247 kauth_cred_t cred = nfsd->nd_cr;
3248 struct vattr bfor, aft;
3249 struct vnode *vp;
3250 nfsrvfh_t nsfh;
3251 u_int32_t *tl;
3252 int32_t t1;
3253 char *bpos;
3254 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0;
3255 uint32_t cnt;
3256 char *cp2;
3257 struct mbuf *mb, *mreq;
3258 u_quad_t frev, off, end;
3259
3260 nfsm_srvmtofh(&nsfh);
3261 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3262
3263 off = fxdr_hyper(tl);
3264 tl += 2;
3265 cnt = fxdr_unsigned(uint32_t, *tl);
3266 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
3267 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
3268 if (error) {
3269 nfsm_reply(2 * NFSX_UNSIGNED);
3270 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3271 return (0);
3272 }
3273 for_ret = VOP_GETATTR(vp, &bfor, cred);
3274 end = (cnt > 0) ? off + cnt : vp->v_size;
3275 if (end < off || end > vp->v_size)
3276 end = vp->v_size;
3277 if (off < vp->v_size)
3278 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end);
3279 /* else error == 0, from nfsrv_fhtovp() */
3280 aft_ret = VOP_GETATTR(vp, &aft, cred);
3281 vput(vp);
3282 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3283 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3284 if (!error) {
3285 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3286 *tl++ = txdr_unsigned(boottime.tv_sec);
3287 *tl = txdr_unsigned(boottime.tv_nsec / 1000);
3288 } else {
3289 return (0);
3290 }
3291 nfsm_srvdone;
3292 }
3293
3294 /*
3295 * nfs statfs service
3296 */
3297 int
3298 nfsrv_statfs(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
3299 {
3300 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3301 struct mbuf *nam = nfsd->nd_nam;
3302 char *dpos = nfsd->nd_dpos;
3303 kauth_cred_t cred = nfsd->nd_cr;
3304 struct statvfs *sf = NULL;
3305 struct nfs_statfs *sfp;
3306 u_int32_t *tl;
3307 int32_t t1;
3308 char *bpos;
3309 int error = 0, rdonly, cache = 0, getret = 1;
3310 int v3 = (nfsd->nd_flag & ND_NFSV3);
3311 char *cp2;
3312 struct mbuf *mb, *mreq;
3313 struct vnode *vp;
3314 struct vattr at;
3315 nfsrvfh_t nsfh;
3316 u_quad_t frev, tval;
3317
3318 nfsm_srvmtofh(&nsfh);
3319 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
3320 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
3321 if (error) {
3322 nfsm_reply(NFSX_UNSIGNED);
3323 nfsm_srvpostop_attr(getret, &at);
3324 return (0);
3325 }
3326 sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK);
3327 error = VFS_STATVFS(vp->v_mount, sf);
3328 getret = VOP_GETATTR(vp, &at, cred);
3329 vput(vp);
3330 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3331 if (v3)
3332 nfsm_srvpostop_attr(getret, &at);
3333 if (error) {
3334 free(sf, M_TEMP);
3335 return (0);
3336 }
3337 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3338 if (v3) {
3339 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize);
3340 txdr_hyper(tval, &sfp->sf_tbytes);
3341 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize);
3342 txdr_hyper(tval, &sfp->sf_fbytes);
3343 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize);
3344 txdr_hyper(tval, &sfp->sf_abytes);
3345 tval = (u_quad_t)sf->f_files;
3346 txdr_hyper(tval, &sfp->sf_tfiles);
3347 tval = (u_quad_t)sf->f_ffree;
3348 txdr_hyper(tval, &sfp->sf_ffiles);
3349 txdr_hyper(tval, &sfp->sf_afiles);
3350 sfp->sf_invarsec = 0;
3351 } else {
3352 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3353 sfp->sf_bsize = txdr_unsigned(sf->f_frsize);
3354 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3355 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3356 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3357 }
3358 nfsmout:
3359 if (sf)
3360 free(sf, M_TEMP);
3361 return error;
3362 }
3363
3364 /*
3365 * nfs fsinfo service
3366 */
3367 int
3368 nfsrv_fsinfo(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
3369 {
3370 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3371 struct mbuf *nam = nfsd->nd_nam;
3372 char *dpos = nfsd->nd_dpos;
3373 kauth_cred_t cred = nfsd->nd_cr;
3374 u_int32_t *tl;
3375 struct nfsv3_fsinfo *sip;
3376 int32_t t1;
3377 char *bpos;
3378 int error = 0, rdonly, cache = 0, getret = 1;
3379 uint32_t maxdata;
3380 char *cp2;
3381 struct mbuf *mb, *mreq;
3382 struct vnode *vp;
3383 struct vattr at;
3384 nfsrvfh_t nsfh;
3385 u_quad_t frev, maxfsize;
3386 struct statvfs *sb;
3387
3388 nfsm_srvmtofh(&nsfh);
3389 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
3390 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
3391 if (error) {
3392 nfsm_reply(NFSX_UNSIGNED);
3393 nfsm_srvpostop_attr(getret, &at);
3394 return (0);
3395 }
3396
3397 /* XXX Try to make a guess on the max file size. */
3398 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK);
3399 VFS_STATVFS(vp->v_mount, sb);
3400 maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1;
3401 free(sb, M_TEMP);
3402
3403 getret = VOP_GETATTR(vp, &at, cred);
3404 vput(vp);
3405 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3406 nfsm_srvpostop_attr(getret, &at);
3407 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3408
3409 /*
3410 * XXX
3411 * There should be file system VFS OP(s) to get this information.
3412 * For now, assume ufs.
3413 */
3414 if (slp->ns_so->so_type == SOCK_DGRAM)
3415 maxdata = NFS_MAXDGRAMDATA;
3416 else
3417 maxdata = NFS_MAXDATA;
3418 sip->fs_rtmax = txdr_unsigned(maxdata);
3419 sip->fs_rtpref = txdr_unsigned(maxdata);
3420 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3421 sip->fs_wtmax = txdr_unsigned(maxdata);
3422 sip->fs_wtpref = txdr_unsigned(maxdata);
3423 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3424 sip->fs_dtpref = txdr_unsigned(maxdata);
3425 txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3426 sip->fs_timedelta.nfsv3_sec = 0;
3427 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3428 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3429 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3430 NFSV3FSINFO_CANSETTIME);
3431 nfsm_srvdone;
3432 }
3433
3434 /*
3435 * nfs pathconf service
3436 */
3437 int
3438 nfsrv_pathconf(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct lwp *lwp, struct mbuf **mrq)
3439 {
3440 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3441 struct mbuf *nam = nfsd->nd_nam;
3442 char *dpos = nfsd->nd_dpos;
3443 kauth_cred_t cred = nfsd->nd_cr;
3444 u_int32_t *tl;
3445 struct nfsv3_pathconf *pc;
3446 int32_t t1;
3447 char *bpos;
3448 int error = 0, rdonly, cache = 0, getret = 1;
3449 register_t linkmax, namemax, chownres, notrunc;
3450 char *cp2;
3451 struct mbuf *mb, *mreq;
3452 struct vnode *vp;
3453 struct vattr at;
3454 nfsrvfh_t nsfh;
3455 u_quad_t frev;
3456
3457 nfsm_srvmtofh(&nsfh);
3458 error = nfsrv_fhtovp(&nsfh, 1, &vp, cred, slp, nam,
3459 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), false);
3460 if (error) {
3461 nfsm_reply(NFSX_UNSIGNED);
3462 nfsm_srvpostop_attr(getret, &at);
3463 return (0);
3464 }
3465 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3466 if (!error)
3467 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3468 if (!error)
3469 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3470 if (!error)
3471 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc);
3472 getret = VOP_GETATTR(vp, &at, cred);
3473 vput(vp);
3474 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3475 nfsm_srvpostop_attr(getret, &at);
3476 if (error)
3477 return (0);
3478 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3479
3480 pc->pc_linkmax = txdr_unsigned(linkmax);
3481 pc->pc_namemax = txdr_unsigned(namemax);
3482 pc->pc_notrunc = txdr_unsigned(notrunc);
3483 pc->pc_chownrestricted = txdr_unsigned(chownres);
3484
3485 /*
3486 * These should probably be supported by VOP_PATHCONF(), but
3487 * until msdosfs is exportable (why would you want to?), the
3488 * Unix defaults should be ok.
3489 */
3490 pc->pc_caseinsensitive = nfs_false;
3491 pc->pc_casepreserving = nfs_true;
3492 nfsm_srvdone;
3493 }
3494
3495 /*
3496 * Null operation, used by clients to ping server
3497 */
3498 /* ARGSUSED */
3499 int
3500 nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3501 struct lwp *lwp, struct mbuf **mrq)
3502 {
3503 struct mbuf *mrep = nfsd->nd_mrep;
3504 char *bpos;
3505 int error = NFSERR_RETVOID, cache = 0;
3506 struct mbuf *mb, *mreq;
3507 u_quad_t frev;
3508
3509 nfsm_reply(0);
3510 nfsmout:
3511 return (0);
3512 }
3513
3514 /*
3515 * No operation, used for obsolete procedures
3516 */
3517 /* ARGSUSED */
3518 int
3519 nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3520 struct lwp *lwp, struct mbuf **mrq)
3521 {
3522 struct mbuf *mrep = nfsd->nd_mrep;
3523 char *bpos;
3524 int error, cache = 0;
3525 struct mbuf *mb, *mreq;
3526 u_quad_t frev;
3527
3528 if (nfsd->nd_repstat)
3529 error = nfsd->nd_repstat;
3530 else
3531 error = EPROCUNAVAIL;
3532 nfsm_reply(0);
3533 nfsmout:
3534 return (0);
3535 }
3536
3537 /*
3538 * Perform access checking for vnodes obtained from file handles that would
3539 * refer to files already opened by a Unix client. You cannot just use
3540 * vn_writechk() and VOP_ACCESS() for two reasons.
3541 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3542 * 2 - The owner is to be given access irrespective of mode bits for some
3543 * operations, so that processes that chmod after opening a file don't
3544 * break. I don't like this because it opens a security hole, but since
3545 * the nfs server opens a security hole the size of a barn door anyhow,
3546 * what the heck.
3547 *
3548 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3549 * will return EPERM instead of EACCES. EPERM is always an error.
3550 */
3551 int
3552 nfsrv_access(struct vnode *vp, int flags, kauth_cred_t cred, int rdonly, struct lwp *lwp, int override)
3553 {
3554 struct vattr vattr;
3555 int error;
3556 if (flags & VWRITE) {
3557 /* Just vn_writechk() changed to check rdonly */
3558 /*
3559 * Disallow write attempts on read-only file systems;
3560 * unless the file is a socket or a block or character
3561 * device resident on the file system.
3562 */
3563 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3564 switch (vp->v_type) {
3565 case VREG:
3566 case VDIR:
3567 case VLNK:
3568 return (EROFS);
3569 default:
3570 break;
3571 }
3572 }
3573
3574 /*
3575 * If the vnode is in use as a process's text,
3576 * we can't allow writing.
3577 */
3578 if (vp->v_iflag & VI_TEXT)
3579 return (ETXTBSY);
3580 }
3581 error = VOP_GETATTR(vp, &vattr, cred);
3582 if (error)
3583 return (error);
3584 error = VOP_ACCESS(vp, flags, cred);
3585 /*
3586 * Allow certain operations for the owner (reads and writes
3587 * on files that are already open).
3588 */
3589 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid)
3590 error = 0;
3591 return error;
3592 }
3593