nfs_serv.c revision 1.101.4.4 1 /* $NetBSD: nfs_serv.c,v 1.101.4.4 2006/04/19 05:06:37 elad 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 1.101.4.4 2006/04/19 05:06:37 elad 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
74 #include <uvm/uvm.h>
75
76 #include <nfs/nfsproto.h>
77 #include <nfs/rpcv2.h>
78 #include <nfs/nfs.h>
79 #include <nfs/xdr_subs.h>
80 #include <nfs/nfsm_subs.h>
81 #include <nfs/nqnfs.h>
82 #include <nfs/nfs_var.h>
83
84 /* Global vars */
85 extern u_int32_t nfs_xdrneg1;
86 extern u_int32_t nfs_false, nfs_true;
87 extern enum vtype nv3tov_type[8];
88 extern struct nfsstats nfsstats;
89 extern const nfstype nfsv2_type[9];
90 extern const nfstype nfsv3_type[9];
91 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
92 int nfsd_use_loan = 1; /* use page-loan for READ OP */
93
94 /*
95 * nfs v3 access service
96 */
97 int
98 nfsrv3_access(nfsd, slp, lwp, mrq)
99 struct nfsrv_descript *nfsd;
100 struct nfssvc_sock *slp;
101 struct lwp *lwp;
102 struct mbuf **mrq;
103 {
104 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
105 struct mbuf *nam = nfsd->nd_nam;
106 caddr_t dpos = nfsd->nd_dpos;
107 kauth_cred_t cred = nfsd->nd_cr;
108 struct vnode *vp;
109 nfsfh_t nfh;
110 fhandle_t *fhp;
111 u_int32_t *tl;
112 int32_t t1;
113 caddr_t bpos;
114 int error = 0, rdonly, cache = 0, getret;
115 char *cp2;
116 struct mbuf *mb, *mreq;
117 struct vattr va;
118 u_long inmode, testmode, outmode;
119 u_quad_t frev;
120
121 fhp = &nfh.fh_generic;
122 nfsm_srvmtofh(fhp);
123 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
124 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
125 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
126 if (error) {
127 nfsm_reply(NFSX_UNSIGNED);
128 nfsm_srvpostop_attr(1, (struct vattr *)0);
129 return (0);
130 }
131 inmode = fxdr_unsigned(u_int32_t, *tl);
132 outmode = 0;
133 if ((inmode & NFSV3ACCESS_READ) &&
134 nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0)
135 outmode |= NFSV3ACCESS_READ;
136 if (vp->v_type != VDIR) {
137 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
138 if (testmode &&
139 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0)
140 outmode |= testmode;
141 if ((inmode & NFSV3ACCESS_EXECUTE) &&
142 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0)
143 outmode |= NFSV3ACCESS_EXECUTE;
144 } else {
145 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
146 NFSV3ACCESS_DELETE);
147 if (testmode &&
148 nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0)
149 outmode |= testmode;
150 if ((inmode & NFSV3ACCESS_LOOKUP) &&
151 nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0)
152 outmode |= NFSV3ACCESS_LOOKUP;
153 }
154 getret = VOP_GETATTR(vp, &va, cred, lwp);
155 vput(vp);
156 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
157 nfsm_srvpostop_attr(getret, &va);
158 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
159 *tl = txdr_unsigned(outmode);
160 nfsm_srvdone;
161 }
162
163 /*
164 * nfs getattr service
165 */
166 int
167 nfsrv_getattr(nfsd, slp, lwp, mrq)
168 struct nfsrv_descript *nfsd;
169 struct nfssvc_sock *slp;
170 struct lwp *lwp;
171 struct mbuf **mrq;
172 {
173 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
174 struct mbuf *nam = nfsd->nd_nam;
175 caddr_t dpos = nfsd->nd_dpos;
176 kauth_cred_t cred = nfsd->nd_cr;
177 struct nfs_fattr *fp;
178 struct vattr va;
179 struct vnode *vp;
180 nfsfh_t nfh;
181 fhandle_t *fhp;
182 u_int32_t *tl;
183 int32_t t1;
184 caddr_t bpos;
185 int error = 0, rdonly, cache = 0;
186 char *cp2;
187 struct mbuf *mb, *mreq;
188 u_quad_t frev;
189
190 fhp = &nfh.fh_generic;
191 nfsm_srvmtofh(fhp);
192 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
193 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
194 if (error) {
195 nfsm_reply(0);
196 return (0);
197 }
198 nqsrv_getl(vp, ND_READ);
199 error = VOP_GETATTR(vp, &va, cred, lwp);
200 vput(vp);
201 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
202 if (error)
203 return (0);
204 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
205 nfsm_srvfillattr(&va, fp);
206 nfsm_srvdone;
207 }
208
209 /*
210 * nfs setattr service
211 */
212 int
213 nfsrv_setattr(nfsd, slp, lwp, mrq)
214 struct nfsrv_descript *nfsd;
215 struct nfssvc_sock *slp;
216 struct lwp *lwp;
217 struct mbuf **mrq;
218 {
219 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
220 struct mbuf *nam = nfsd->nd_nam;
221 caddr_t dpos = nfsd->nd_dpos;
222 kauth_cred_t cred = nfsd->nd_cr;
223 struct vattr va, preat;
224 struct nfsv2_sattr *sp;
225 struct nfs_fattr *fp;
226 struct vnode *vp;
227 nfsfh_t nfh;
228 fhandle_t *fhp;
229 u_int32_t *tl;
230 int32_t t1;
231 caddr_t bpos;
232 int error = 0, rdonly, cache = 0, preat_ret = 1, postat_ret = 1;
233 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
234 char *cp2;
235 struct mbuf *mb, *mreq;
236 u_quad_t frev;
237 struct timespec guard;
238 struct mount *mp = NULL;
239
240 fhp = &nfh.fh_generic;
241 nfsm_srvmtofh(fhp);
242 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
243 return (ESTALE);
244 vn_start_write(NULL, &mp, V_WAIT);
245 VATTR_NULL(&va);
246 if (v3) {
247 nfsm_srvsattr(&va);
248 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
249 gcheck = fxdr_unsigned(int, *tl);
250 if (gcheck) {
251 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
252 fxdr_nfsv3time(tl, &guard);
253 }
254 } else {
255 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
256 /*
257 * Nah nah nah nah na nah
258 * There is a bug in the Sun client that puts 0xffff in the mode
259 * field of sattr when it should put in 0xffffffff. The u_short
260 * doesn't sign extend.
261 * --> check the low order 2 bytes for 0xffff
262 */
263 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
264 va.va_mode = nfstov_mode(sp->sa_mode);
265 if (sp->sa_uid != nfs_xdrneg1)
266 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
267 if (sp->sa_gid != nfs_xdrneg1)
268 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
269 if (sp->sa_size != nfs_xdrneg1)
270 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
271 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
272 #ifdef notyet
273 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime);
274 #else
275 va.va_atime.tv_sec =
276 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
277 va.va_atime.tv_nsec = 0;
278 #endif
279 }
280 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
281 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime);
282
283 }
284
285 /*
286 * Now that we have all the fields, lets do it.
287 */
288 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
289 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
290 if (error) {
291 nfsm_reply(2 * NFSX_UNSIGNED);
292 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
293 vn_finished_write(mp, 0);
294 return (0);
295 }
296 nqsrv_getl(vp, ND_WRITE);
297 if (v3) {
298 error = preat_ret = VOP_GETATTR(vp, &preat, cred, lwp);
299 if (!error && gcheck &&
300 (preat.va_ctime.tv_sec != guard.tv_sec ||
301 preat.va_ctime.tv_nsec != guard.tv_nsec))
302 error = NFSERR_NOT_SYNC;
303 if (error) {
304 vput(vp);
305 nfsm_reply(NFSX_WCCDATA(v3));
306 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
307 vn_finished_write(mp, 0);
308 return (0);
309 }
310 }
311
312 /*
313 * If the size is being changed write acces is required, otherwise
314 * just check for a read only file system.
315 */
316 if (va.va_size == ((u_quad_t)((quad_t) -1))) {
317 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
318 error = EROFS;
319 goto out;
320 }
321 } else {
322 if (vp->v_type == VDIR) {
323 error = EISDIR;
324 goto out;
325 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
326 lwp, 0)) != 0)
327 goto out;
328 }
329 error = VOP_SETATTR(vp, &va, cred, lwp);
330 postat_ret = VOP_GETATTR(vp, &va, cred, lwp);
331 if (!error)
332 error = postat_ret;
333 out:
334 vput(vp);
335 vn_finished_write(mp, 0);
336 nfsm_reply(NFSX_WCCORFATTR(v3));
337 if (v3) {
338 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
339 return (0);
340 } else {
341 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
342 nfsm_srvfillattr(&va, fp);
343 }
344 nfsm_srvdone;
345 }
346
347 /*
348 * nfs lookup rpc
349 */
350 int
351 nfsrv_lookup(nfsd, slp, lwp, mrq)
352 struct nfsrv_descript *nfsd;
353 struct nfssvc_sock *slp;
354 struct lwp *lwp;
355 struct mbuf **mrq;
356 {
357 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
358 struct mbuf *nam = nfsd->nd_nam;
359 caddr_t dpos = nfsd->nd_dpos;
360 kauth_cred_t cred = nfsd->nd_cr;
361 struct nfs_fattr *fp;
362 struct nameidata nd, ind, *ndp = &nd;
363 struct vnode *vp, *dirp;
364 nfsfh_t nfh;
365 fhandle_t *fhp;
366 caddr_t cp;
367 u_int32_t *tl;
368 int32_t t1;
369 caddr_t bpos;
370 int error = 0, cache = 0, dirattr_ret = 1;
371 uint32_t len;
372 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
373 char *cp2;
374 struct mbuf *mb, *mreq;
375 struct vattr va, dirattr;
376 u_quad_t frev;
377
378 fhp = &nfh.fh_generic;
379 nfsm_srvmtofh(fhp);
380 nfsm_srvnamesiz(len);
381
382 pubflag = nfs_ispublicfh(fhp);
383
384 nd.ni_cnd.cn_cred = cred;
385 nd.ni_cnd.cn_nameiop = LOOKUP;
386 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
387 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
388 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
389
390 if (!error && pubflag) {
391 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
392 /*
393 * Setup call to lookup() to see if we can find
394 * the index file. Arguably, this doesn't belong
395 * in a kernel.. Ugh.
396 */
397 ind = nd;
398 VOP_UNLOCK(nd.ni_vp, 0);
399 ind.ni_pathlen = strlen(nfs_pub.np_index);
400 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
401 nfs_pub.np_index;
402 ind.ni_startdir = nd.ni_vp;
403 VREF(ind.ni_startdir);
404 error = lookup(&ind);
405 if (!error) {
406 /*
407 * Found an index file. Get rid of
408 * the old references.
409 */
410 if (dirp)
411 vrele(dirp);
412 dirp = nd.ni_vp;
413 vrele(nd.ni_startdir);
414 ndp = &ind;
415 } else
416 error = 0;
417 }
418 /*
419 * If the public filehandle was used, check that this lookup
420 * didn't result in a filehandle outside the publicly exported
421 * filesystem.
422 */
423
424 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
425 vput(nd.ni_vp);
426 error = EPERM;
427 }
428 }
429
430 if (dirp) {
431 if (v3)
432 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp);
433 vrele(dirp);
434 }
435
436 if (error) {
437 nfsm_reply(NFSX_POSTOPATTR(v3));
438 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
439 return (0);
440 }
441
442 nqsrv_getl(ndp->ni_startdir, ND_READ);
443 vrele(ndp->ni_startdir);
444 PNBUF_PUT(nd.ni_cnd.cn_pnbuf);
445 vp = ndp->ni_vp;
446 memset((caddr_t)fhp, 0, sizeof(nfh));
447 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
448 error = VFS_VPTOFH(vp, &fhp->fh_fid);
449 if (!error)
450 error = VOP_GETATTR(vp, &va, cred, lwp);
451 vput(vp);
452 KASSERT(fhp->fh_fid.fid_len <= _VFS_MAXFIDSZ);
453 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
454 if (error) {
455 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
456 return (0);
457 }
458 nfsm_srvfhtom(fhp, v3);
459 if (v3) {
460 nfsm_srvpostop_attr(0, &va);
461 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
462 } else {
463 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
464 nfsm_srvfillattr(&va, fp);
465 }
466 nfsm_srvdone;
467 }
468
469 /*
470 * nfs readlink service
471 */
472 int
473 nfsrv_readlink(nfsd, slp, lwp, mrq)
474 struct nfsrv_descript *nfsd;
475 struct nfssvc_sock *slp;
476 struct lwp *lwp;
477 struct mbuf **mrq;
478 {
479 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
480 struct mbuf *nam = nfsd->nd_nam;
481 caddr_t dpos = nfsd->nd_dpos;
482 kauth_cred_t cred = nfsd->nd_cr;
483 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
484 struct iovec *ivp = iv;
485 struct mbuf *mp;
486 u_int32_t *tl;
487 int32_t t1;
488 caddr_t bpos;
489 int error = 0, rdonly, cache = 0, i, padlen, getret;
490 uint32_t len;
491 int v3 = (nfsd->nd_flag & ND_NFSV3);
492 char *cp2;
493 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq;
494 struct vnode *vp;
495 struct vattr attr;
496 nfsfh_t nfh;
497 fhandle_t *fhp;
498 struct uio io, *uiop = &io;
499 u_quad_t frev;
500
501 fhp = &nfh.fh_generic;
502 nfsm_srvmtofh(fhp);
503 len = 0;
504 i = 0;
505 while (len < NFS_MAXPATHLEN) {
506 mp = m_get(M_WAIT, MT_DATA);
507 MCLAIM(mp, &nfs_mowner);
508 m_clget(mp, M_WAIT);
509 mp->m_len = NFSMSIZ(mp);
510 if (len == 0)
511 mp3 = mp2 = mp;
512 else {
513 mp2->m_next = mp;
514 mp2 = mp;
515 }
516 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
517 mp->m_len = NFS_MAXPATHLEN-len;
518 len = NFS_MAXPATHLEN;
519 } else
520 len += mp->m_len;
521 ivp->iov_base = mtod(mp, caddr_t);
522 ivp->iov_len = mp->m_len;
523 i++;
524 ivp++;
525 }
526 uiop->uio_iov = iv;
527 uiop->uio_iovcnt = i;
528 uiop->uio_offset = 0;
529 uiop->uio_resid = len;
530 uiop->uio_rw = UIO_READ;
531 UIO_SETUP_SYSSPACE(uiop);
532 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
533 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
534 if (error) {
535 m_freem(mp3);
536 nfsm_reply(2 * NFSX_UNSIGNED);
537 nfsm_srvpostop_attr(1, (struct vattr *)0);
538 return (0);
539 }
540 if (vp->v_type != VLNK) {
541 if (v3)
542 error = EINVAL;
543 else
544 error = ENXIO;
545 goto out;
546 }
547 nqsrv_getl(vp, ND_READ);
548 error = VOP_READLINK(vp, uiop, cred);
549 out:
550 getret = VOP_GETATTR(vp, &attr, cred, lwp);
551 vput(vp);
552 if (error)
553 m_freem(mp3);
554 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
555 if (v3) {
556 nfsm_srvpostop_attr(getret, &attr);
557 if (error)
558 return (0);
559 }
560 len -= uiop->uio_resid;
561 padlen = nfsm_padlen(len);
562 if (uiop->uio_resid || padlen)
563 nfs_zeropad(mp3, uiop->uio_resid, padlen);
564 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
565 *tl = txdr_unsigned(len);
566 mb->m_next = mp3;
567 nfsm_srvdone;
568 }
569
570 /*
571 * nfs read service
572 */
573 int
574 nfsrv_read(nfsd, slp, lwp, mrq)
575 struct nfsrv_descript *nfsd;
576 struct nfssvc_sock *slp;
577 struct lwp *lwp;
578 struct mbuf **mrq;
579 {
580 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
581 struct mbuf *nam = nfsd->nd_nam;
582 caddr_t dpos = nfsd->nd_dpos;
583 kauth_cred_t cred = nfsd->nd_cr;
584 struct mbuf *m;
585 struct nfs_fattr *fp;
586 u_int32_t *tl;
587 int32_t t1;
588 int i;
589 caddr_t bpos;
590 int error = 0, rdonly, cache = 0, getret;
591 int v3 = (nfsd->nd_flag & ND_NFSV3);
592 uint32_t reqlen, len, cnt, left;
593 int padlen;
594 char *cp2;
595 struct mbuf *mb, *mreq;
596 struct vnode *vp;
597 nfsfh_t nfh;
598 fhandle_t *fhp;
599 struct uio io, *uiop = &io;
600 struct vattr va;
601 off_t off;
602 u_quad_t frev;
603
604 fhp = &nfh.fh_generic;
605 nfsm_srvmtofh(fhp);
606 if (v3) {
607 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
608 off = fxdr_hyper(tl);
609 } else {
610 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
611 off = (off_t)fxdr_unsigned(u_int32_t, *tl);
612 }
613 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
614 reqlen = fxdr_unsigned(uint32_t, *tl);
615 reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd));
616 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
617 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
618 if (error) {
619 nfsm_reply(2 * NFSX_UNSIGNED);
620 nfsm_srvpostop_attr(1, (struct vattr *)0);
621 return (0);
622 }
623 if (vp->v_type != VREG) {
624 if (v3)
625 error = EINVAL;
626 else
627 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
628 }
629 if (!error) {
630 nqsrv_getl(vp, ND_READ);
631 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0)
632 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1);
633 }
634 getret = VOP_GETATTR(vp, &va, cred, lwp);
635 if (!error)
636 error = getret;
637 if (error) {
638 vput(vp);
639 nfsm_reply(NFSX_POSTOPATTR(v3));
640 nfsm_srvpostop_attr(getret, &va);
641 return (0);
642 }
643 if (off >= va.va_size)
644 cnt = 0;
645 else if ((off + reqlen) > va.va_size)
646 cnt = va.va_size - off;
647 else
648 cnt = reqlen;
649 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
650 if (v3) {
651 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
652 *tl++ = nfs_true;
653 fp = (struct nfs_fattr *)tl;
654 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
655 } else {
656 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
657 fp = (struct nfs_fattr *)tl;
658 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
659 }
660 len = left = cnt;
661 if (cnt > 0) {
662 if (nfsd_use_loan) {
663 struct vm_page **pgpp;
664 voff_t pgoff = trunc_page(off);
665 int npages;
666 vaddr_t lva;
667
668 npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT;
669 KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */
670
671 /* allocate kva for mbuf data */
672 lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so);
673 if (lva == 0) {
674 /* fall back to VOP_READ */
675 goto loan_fail;
676 }
677
678 /* allocate mbuf */
679 m = m_get(M_WAIT, MT_DATA);
680 MCLAIM(m, &nfs_mowner);
681 pgpp = m->m_ext.ext_pgs;
682
683 /* loan pages */
684 error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages,
685 pgpp);
686 if (error) {
687 sokvafree(lva, npages << PAGE_SHIFT);
688 m_free(m);
689 if (error == EBUSY)
690 goto loan_fail;
691 goto read_error;
692 }
693
694 /* associate kva to mbuf */
695 MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)),
696 cnt, M_MBUF, soloanfree, slp->ns_so);
697 m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
698 m->m_len = cnt;
699
700 /* map pages */
701 for (i = 0; i < npages; i++) {
702 pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]),
703 VM_PROT_READ);
704 lva += PAGE_SIZE;
705 }
706
707 pmap_update(pmap_kernel());
708
709 mb->m_next = m;
710 mb = m;
711 error = 0;
712 uiop->uio_resid = 0;
713 } else {
714 struct iovec *iv;
715 struct iovec *iv2;
716 struct mbuf *m2;
717 int siz;
718 loan_fail:
719 /*
720 * Generate the mbuf list with the uio_iov ref. to it.
721 */
722 i = 0;
723 m = m2 = mb;
724 while (left > 0) {
725 siz = min(M_TRAILINGSPACE(m), left);
726 if (siz > 0) {
727 left -= siz;
728 i++;
729 }
730 if (left > 0) {
731 m = m_get(M_WAIT, MT_DATA);
732 MCLAIM(m, &nfs_mowner);
733 m_clget(m, M_WAIT);
734 m->m_len = 0;
735 m2->m_next = m;
736 m2 = m;
737 }
738 }
739 iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK);
740 uiop->uio_iov = iv2 = iv;
741 m = mb;
742 left = cnt;
743 i = 0;
744 while (left > 0) {
745 if (m == NULL)
746 panic("nfsrv_read iov");
747 siz = min(M_TRAILINGSPACE(m), left);
748 if (siz > 0) {
749 iv->iov_base = mtod(m, caddr_t) +
750 m->m_len;
751 iv->iov_len = siz;
752 m->m_len += siz;
753 left -= siz;
754 iv++;
755 i++;
756 }
757 m = m->m_next;
758 }
759 uiop->uio_iovcnt = i;
760 uiop->uio_offset = off;
761 uiop->uio_resid = cnt;
762 uiop->uio_rw = UIO_READ;
763 UIO_SETUP_SYSSPACE(uiop);
764 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
765 free((caddr_t)iv2, M_TEMP);
766 }
767 read_error:
768 if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){
769 if (!error)
770 error = getret;
771 m_freem(mreq);
772 vput(vp);
773 nfsm_reply(NFSX_POSTOPATTR(v3));
774 nfsm_srvpostop_attr(getret, &va);
775 return (0);
776 }
777 } else {
778 uiop->uio_resid = 0;
779 }
780 vput(vp);
781 nfsm_srvfillattr(&va, fp);
782 len -= uiop->uio_resid;
783 padlen = nfsm_padlen(len);
784 if (uiop->uio_resid || padlen)
785 nfs_zeropad(mb, uiop->uio_resid, padlen);
786 if (v3) {
787 /* count */
788 *tl++ = txdr_unsigned(len);
789 /* eof */
790 if (off + len >= va.va_size)
791 *tl++ = nfs_true;
792 else
793 *tl++ = nfs_false;
794 }
795 *tl = txdr_unsigned(len);
796 nfsm_srvdone;
797 }
798
799 /*
800 * nfs write service
801 */
802 int
803 nfsrv_write(nfsd, slp, lwp, mrq)
804 struct nfsrv_descript *nfsd;
805 struct nfssvc_sock *slp;
806 struct lwp *lwp;
807 struct mbuf **mrq;
808 {
809 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
810 struct mbuf *nam = nfsd->nd_nam;
811 caddr_t dpos = nfsd->nd_dpos;
812 kauth_cred_t cred = nfsd->nd_cr;
813 struct iovec *ivp;
814 int i, cnt;
815 struct mbuf *mp;
816 struct nfs_fattr *fp;
817 struct iovec *iv;
818 struct vattr va, forat;
819 u_int32_t *tl;
820 int32_t t1;
821 caddr_t bpos;
822 int error = 0, rdonly, cache = 0, len, forat_ret = 1;
823 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
824 int stable = NFSV3WRITE_FILESYNC;
825 int v3 = (nfsd->nd_flag & ND_NFSV3);
826 char *cp2;
827 struct mbuf *mb, *mreq;
828 struct vnode *vp;
829 nfsfh_t nfh;
830 fhandle_t *fhp;
831 struct uio io, *uiop = &io;
832 off_t off;
833 u_quad_t frev;
834 struct mount *mntp = NULL;
835
836 if (mrep == NULL) {
837 *mrq = NULL;
838 return (0);
839 }
840 fhp = &nfh.fh_generic;
841 nfsm_srvmtofh(fhp);
842 if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
843 return (ESTALE);
844 vn_start_write(NULL, &mntp, V_WAIT);
845 if (v3) {
846 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
847 off = fxdr_hyper(tl);
848 tl += 3;
849 stable = fxdr_unsigned(int, *tl++);
850 } else {
851 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
852 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
853 tl += 2;
854 }
855 retlen = len = fxdr_unsigned(int32_t, *tl);
856 cnt = i = 0;
857
858 /*
859 * For NFS Version 2, it is not obvious what a write of zero length
860 * should do, but I might as well be consistent with Version 3,
861 * which is to return ok so long as there are no permission problems.
862 */
863 if (len > 0) {
864 zeroing = 1;
865 mp = mrep;
866 while (mp) {
867 if (mp == md) {
868 zeroing = 0;
869 adjust = dpos - mtod(mp, caddr_t);
870 mp->m_len -= adjust;
871 if (mp->m_len > 0 && adjust > 0)
872 NFSMADV(mp, adjust);
873 }
874 if (zeroing)
875 mp->m_len = 0;
876 else if (mp->m_len > 0) {
877 i += mp->m_len;
878 if (i > len) {
879 mp->m_len -= (i - len);
880 zeroing = 1;
881 }
882 if (mp->m_len > 0)
883 cnt++;
884 }
885 mp = mp->m_next;
886 }
887 }
888 if (len > NFS_MAXDATA || len < 0 || i < len) {
889 error = EIO;
890 nfsm_reply(2 * NFSX_UNSIGNED);
891 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
892 vn_finished_write(mntp, 0);
893 return (0);
894 }
895 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
896 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
897 if (error) {
898 nfsm_reply(2 * NFSX_UNSIGNED);
899 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
900 vn_finished_write(mntp, 0);
901 return (0);
902 }
903 if (v3)
904 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp);
905 if (vp->v_type != VREG) {
906 if (v3)
907 error = EINVAL;
908 else
909 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
910 }
911 if (!error) {
912 nqsrv_getl(vp, ND_WRITE);
913 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1);
914 }
915 if (error) {
916 vput(vp);
917 nfsm_reply(NFSX_WCCDATA(v3));
918 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
919 vn_finished_write(mntp, 0);
920 return (0);
921 }
922
923 if (len > 0) {
924 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK);
925 uiop->uio_iov = iv = ivp;
926 uiop->uio_iovcnt = cnt;
927 mp = mrep;
928 while (mp) {
929 if (mp->m_len > 0) {
930 ivp->iov_base = mtod(mp, caddr_t);
931 ivp->iov_len = mp->m_len;
932 ivp++;
933 }
934 mp = mp->m_next;
935 }
936
937 /*
938 * XXX
939 * The IO_METASYNC flag indicates that all metadata (and not
940 * just enough to ensure data integrity) must be written to
941 * stable storage synchronously.
942 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
943 */
944 if (stable == NFSV3WRITE_UNSTABLE)
945 ioflags = IO_NODELOCKED;
946 else if (stable == NFSV3WRITE_DATASYNC)
947 ioflags = (IO_SYNC | IO_NODELOCKED);
948 else
949 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
950 uiop->uio_resid = len;
951 uiop->uio_rw = UIO_WRITE;
952 uiop->uio_offset = off;
953 UIO_SETUP_SYSSPACE(uiop);
954 error = VOP_WRITE(vp, uiop, ioflags, cred);
955 nfsstats.srvvop_writes++;
956 free(iv, M_TEMP);
957 }
958 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp);
959 vput(vp);
960 vn_finished_write(mntp, 0);
961 if (!error)
962 error = aftat_ret;
963 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
964 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
965 if (v3) {
966 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
967 if (error)
968 return (0);
969 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
970 *tl++ = txdr_unsigned(retlen);
971 if (stable == NFSV3WRITE_UNSTABLE)
972 *tl++ = txdr_unsigned(stable);
973 else
974 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
975 /*
976 * Actually, there is no need to txdr these fields,
977 * but it may make the values more human readable,
978 * for debugging purposes.
979 */
980 *tl++ = txdr_unsigned(boottime.tv_sec);
981 *tl = txdr_unsigned(boottime.tv_usec);
982 } else {
983 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
984 nfsm_srvfillattr(&va, fp);
985 }
986 nfsm_srvdone;
987 }
988
989 /*
990 * XXX elad: the original NFSW_SAMECRED() macro also made sure the
991 * two nd_flag fields of the descriptors contained
992 * ND_KERBAUTH.
993 */
994 static int
995 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2)
996 {
997 int i, do_ngroups;
998
999 if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2))
1000 return (0);
1001 if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2))
1002 return (0);
1003 do_ngroups = kauth_cred_ngroups(cred1);
1004 for (i = 0; i < do_ngroups; i++)
1005 if (kauth_cred_group(cred1, i) !=
1006 kauth_cred_group(cred2, i))
1007 return (0);
1008
1009 return (1);
1010 }
1011
1012 /*
1013 * NFS write service with write gathering support. Called when
1014 * nfsrvw_procrastinate > 0.
1015 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
1016 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
1017 * Jan. 1994.
1018 */
1019 int
1020 nfsrv_writegather(ndp, slp, lwp, mrq)
1021 struct nfsrv_descript **ndp;
1022 struct nfssvc_sock *slp;
1023 struct lwp *lwp;
1024 struct mbuf **mrq;
1025 {
1026 struct iovec *ivp;
1027 struct mbuf *mp;
1028 struct nfsrv_descript *wp, *nfsd, *owp, *swp;
1029 struct nfs_fattr *fp;
1030 int i = 0;
1031 struct iovec *iov;
1032 struct nfsrvw_delayhash *wpp;
1033 kauth_cred_t cred;
1034 struct vattr va, forat;
1035 u_int32_t *tl;
1036 int32_t t1;
1037 caddr_t bpos, dpos;
1038 int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1;
1039 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
1040 char *cp2;
1041 struct mbuf *mb, *mreq, *mrep, *md;
1042 struct vnode *vp;
1043 struct uio io, *uiop = &io;
1044 u_quad_t frev, cur_usec;
1045 struct mount *mntp = NULL;
1046
1047 *mrq = NULL;
1048 if (*ndp) {
1049 nfsd = *ndp;
1050 *ndp = NULL;
1051 mrep = nfsd->nd_mrep;
1052 md = nfsd->nd_md;
1053 dpos = nfsd->nd_dpos;
1054 cred = nfsd->nd_cr;
1055 v3 = (nfsd->nd_flag & ND_NFSV3);
1056 LIST_INIT(&nfsd->nd_coalesce);
1057 nfsd->nd_mreq = NULL;
1058 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
1059 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1060 nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
1061
1062 /*
1063 * Now, get the write header..
1064 */
1065 nfsm_srvmtofh(&nfsd->nd_fh);
1066 if (v3) {
1067 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1068 nfsd->nd_off = fxdr_hyper(tl);
1069 tl += 3;
1070 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
1071 } else {
1072 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1073 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
1074 tl += 2;
1075 }
1076 len = fxdr_unsigned(int32_t, *tl);
1077 nfsd->nd_len = len;
1078 nfsd->nd_eoff = nfsd->nd_off + len;
1079
1080 /*
1081 * Trim the header out of the mbuf list and trim off any trailing
1082 * junk so that the mbuf list has only the write data.
1083 */
1084 zeroing = 1;
1085 i = 0;
1086 mp = mrep;
1087 while (mp) {
1088 if (mp == md) {
1089 zeroing = 0;
1090 adjust = dpos - mtod(mp, caddr_t);
1091 mp->m_len -= adjust;
1092 if (mp->m_len > 0 && adjust > 0)
1093 NFSMADV(mp, adjust);
1094 }
1095 if (zeroing)
1096 mp->m_len = 0;
1097 else {
1098 i += mp->m_len;
1099 if (i > len) {
1100 mp->m_len -= (i - len);
1101 zeroing = 1;
1102 }
1103 }
1104 mp = mp->m_next;
1105 }
1106 if (len > NFS_MAXDATA || len < 0 || i < len) {
1107 nfsmout:
1108 m_freem(mrep);
1109 error = EIO;
1110 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1111 if (v3)
1112 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1113 nfsd->nd_mreq = mreq;
1114 nfsd->nd_mrep = NULL;
1115 nfsd->nd_time = 0;
1116 }
1117
1118 /*
1119 * Add this entry to the hash and time queues.
1120 */
1121 s = splsoftclock();
1122 owp = NULL;
1123 wp = LIST_FIRST(&slp->ns_tq);
1124 while (wp && wp->nd_time < nfsd->nd_time) {
1125 owp = wp;
1126 wp = LIST_NEXT(wp, nd_tq);
1127 }
1128 if (owp) {
1129 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1130 } else {
1131 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1132 }
1133 if (nfsd->nd_mrep) {
1134 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1135 owp = NULL;
1136 wp = LIST_FIRST(wpp);
1137 while (wp &&
1138 memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) {
1139 owp = wp;
1140 wp = LIST_NEXT(wp, nd_hash);
1141 }
1142 while (wp && wp->nd_off < nfsd->nd_off &&
1143 !memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) {
1144 owp = wp;
1145 wp = LIST_NEXT(wp, nd_hash);
1146 }
1147 if (owp) {
1148 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1149
1150 /*
1151 * Search the hash list for overlapping entries and
1152 * coalesce.
1153 */
1154 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1155 wp = LIST_NEXT(nfsd, nd_hash);
1156 if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr))
1157 nfsrvw_coalesce(owp, nfsd);
1158 }
1159 } else {
1160 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1161 }
1162 }
1163 splx(s);
1164 }
1165
1166 /*
1167 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1168 * and generate the associated reply mbuf list(s).
1169 */
1170 loop1:
1171 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1172 s = splsoftclock();
1173 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) {
1174 owp = LIST_NEXT(nfsd, nd_tq);
1175 if (nfsd->nd_time > cur_usec)
1176 break;
1177 if (nfsd->nd_mreq)
1178 continue;
1179 LIST_REMOVE(nfsd, nd_tq);
1180 LIST_REMOVE(nfsd, nd_hash);
1181 splx(s);
1182 mrep = nfsd->nd_mrep;
1183 nfsd->nd_mrep = NULL;
1184 cred = nfsd->nd_cr;
1185 v3 = (nfsd->nd_flag & ND_NFSV3);
1186 forat_ret = aftat_ret = 1;
1187 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1188 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1189 FALSE);
1190 if (!error) {
1191 if (v3)
1192 forat_ret = VOP_GETATTR(vp, &forat, cred, lwp);
1193 if (vp->v_type != VREG) {
1194 if (v3)
1195 error = EINVAL;
1196 else
1197 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1198 }
1199 } else
1200 vp = NULL;
1201 if (!error) {
1202 nqsrv_getl(vp, ND_WRITE);
1203 error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1);
1204 }
1205
1206 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1207 ioflags = IO_NODELOCKED;
1208 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1209 ioflags = (IO_SYNC | IO_NODELOCKED);
1210 else
1211 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1212 uiop->uio_rw = UIO_WRITE;
1213 uiop->uio_offset = nfsd->nd_off;
1214 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1215 UIO_SETUP_SYSSPACE(uiop);
1216 if (uiop->uio_resid > 0) {
1217 mp = mrep;
1218 i = 0;
1219 while (mp) {
1220 if (mp->m_len > 0)
1221 i++;
1222 mp = mp->m_next;
1223 }
1224 uiop->uio_iovcnt = i;
1225 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
1226 uiop->uio_iov = ivp = iov;
1227 mp = mrep;
1228 while (mp) {
1229 if (mp->m_len > 0) {
1230 ivp->iov_base = mtod(mp, caddr_t);
1231 ivp->iov_len = mp->m_len;
1232 ivp++;
1233 }
1234 mp = mp->m_next;
1235 }
1236 if (!error) {
1237 if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) {
1238 VOP_UNLOCK(vp, 0);
1239 vn_start_write(NULL, &mntp, V_WAIT);
1240 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1241 }
1242 if (!error) {
1243 error = VOP_WRITE(vp, uiop, ioflags, cred);
1244 nfsstats.srvvop_writes++;
1245 vn_finished_write(mntp, 0);
1246 }
1247 }
1248 free((caddr_t)iov, M_TEMP);
1249 }
1250 m_freem(mrep);
1251 if (vp) {
1252 aftat_ret = VOP_GETATTR(vp, &va, cred, lwp);
1253 vput(vp);
1254 }
1255
1256 /*
1257 * Loop around generating replies for all write rpcs that have
1258 * now been completed.
1259 */
1260 swp = nfsd;
1261 do {
1262 if (error) {
1263 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1264 if (v3) {
1265 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1266 }
1267 } else {
1268 nfsm_writereply(NFSX_PREOPATTR(v3) +
1269 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1270 NFSX_WRITEVERF(v3), v3);
1271 if (v3) {
1272 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1273 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1274 *tl++ = txdr_unsigned(nfsd->nd_len);
1275 *tl++ = txdr_unsigned(swp->nd_stable);
1276 /*
1277 * Actually, there is no need to txdr these fields,
1278 * but it may make the values more human readable,
1279 * for debugging purposes.
1280 */
1281 *tl++ = txdr_unsigned(boottime.tv_sec);
1282 *tl = txdr_unsigned(boottime.tv_usec);
1283 } else {
1284 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1285 nfsm_srvfillattr(&va, fp);
1286 }
1287 }
1288 nfsd->nd_mreq = mreq;
1289 if (nfsd->nd_mrep)
1290 panic("nfsrv_write: nd_mrep not free");
1291
1292 /*
1293 * Done. Put it at the head of the timer queue so that
1294 * the final phase can return the reply.
1295 */
1296 s = splsoftclock();
1297 if (nfsd != swp) {
1298 nfsd->nd_time = 0;
1299 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1300 }
1301 nfsd = LIST_FIRST(&swp->nd_coalesce);
1302 if (nfsd) {
1303 LIST_REMOVE(nfsd, nd_tq);
1304 }
1305 splx(s);
1306 } while (nfsd);
1307 s = splsoftclock();
1308 swp->nd_time = 0;
1309 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1310 splx(s);
1311 goto loop1;
1312 }
1313 splx(s);
1314
1315 /*
1316 * Search for a reply to return.
1317 */
1318 s = splsoftclock();
1319 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) {
1320 if (nfsd->nd_mreq) {
1321 LIST_REMOVE(nfsd, nd_tq);
1322 *mrq = nfsd->nd_mreq;
1323 *ndp = nfsd;
1324 break;
1325 }
1326 }
1327 splx(s);
1328 return (0);
1329 }
1330
1331 /*
1332 * Coalesce the write request nfsd into owp. To do this we must:
1333 * - remove nfsd from the queues
1334 * - merge nfsd->nd_mrep into owp->nd_mrep
1335 * - update the nd_eoff and nd_stable for owp
1336 * - put nfsd on owp's nd_coalesce list
1337 * NB: Must be called at splsoftclock().
1338 */
1339 void
1340 nfsrvw_coalesce(owp, nfsd)
1341 struct nfsrv_descript *owp;
1342 struct nfsrv_descript *nfsd;
1343 {
1344 int overlap;
1345 struct mbuf *mp;
1346 struct nfsrv_descript *m;
1347
1348 LIST_REMOVE(nfsd, nd_hash);
1349 LIST_REMOVE(nfsd, nd_tq);
1350 if (owp->nd_eoff < nfsd->nd_eoff) {
1351 overlap = owp->nd_eoff - nfsd->nd_off;
1352 if (overlap < 0)
1353 panic("nfsrv_coalesce: bad off");
1354 if (overlap > 0)
1355 m_adj(nfsd->nd_mrep, overlap);
1356 mp = owp->nd_mrep;
1357 while (mp->m_next)
1358 mp = mp->m_next;
1359 mp->m_next = nfsd->nd_mrep;
1360 owp->nd_eoff = nfsd->nd_eoff;
1361 } else
1362 m_freem(nfsd->nd_mrep);
1363 nfsd->nd_mrep = NULL;
1364 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1365 owp->nd_stable = NFSV3WRITE_FILESYNC;
1366 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1367 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1368 owp->nd_stable = NFSV3WRITE_DATASYNC;
1369 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1370 /*
1371 * nfsd might hold coalesce elements! Move them to owp.
1372 * Otherwise, requests may be lost and clients will be stuck.
1373 */
1374 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) {
1375 LIST_REMOVE(m, nd_tq);
1376 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1377 }
1378 }
1379
1380 /*
1381 * nfs create service
1382 * now does a truncate to 0 length via. setattr if it already exists
1383 */
1384 int
1385 nfsrv_create(nfsd, slp, lwp, mrq)
1386 struct nfsrv_descript *nfsd;
1387 struct nfssvc_sock *slp;
1388 struct lwp *lwp;
1389 struct mbuf **mrq;
1390 {
1391 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1392 struct mbuf *nam = nfsd->nd_nam;
1393 caddr_t dpos = nfsd->nd_dpos;
1394 kauth_cred_t cred = nfsd->nd_cr;
1395 struct nfs_fattr *fp;
1396 struct vattr va, dirfor, diraft;
1397 struct nfsv2_sattr *sp;
1398 u_int32_t *tl;
1399 struct nameidata nd;
1400 caddr_t cp;
1401 int32_t t1;
1402 caddr_t bpos;
1403 int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1404 int rdev = 0;
1405 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1406 char *cp2;
1407 struct mbuf *mb, *mreq;
1408 struct vnode *vp = NULL, *dirp = NULL;
1409 nfsfh_t nfh;
1410 fhandle_t *fhp;
1411 u_quad_t frev, tempsize;
1412 u_char cverf[NFSX_V3CREATEVERF];
1413 struct mount *mp = NULL;
1414
1415 nd.ni_cnd.cn_nameiop = 0;
1416 fhp = &nfh.fh_generic;
1417 nfsm_srvmtofh(fhp);
1418 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1419 return (ESTALE);
1420 vn_start_write(NULL, &mp, V_WAIT);
1421 nfsm_srvnamesiz(len);
1422 nd.ni_cnd.cn_cred = cred;
1423 nd.ni_cnd.cn_nameiop = CREATE;
1424 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1425 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1426 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1427 if (dirp) {
1428 if (v3)
1429 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
1430 else {
1431 vrele(dirp);
1432 dirp = (struct vnode *)0;
1433 }
1434 }
1435 if (error) {
1436 nfsm_reply(NFSX_WCCDATA(v3));
1437 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1438 if (dirp)
1439 vrele(dirp);
1440 vn_finished_write(mp, 0);
1441 return (0);
1442 }
1443 VATTR_NULL(&va);
1444 if (v3) {
1445 va.va_mode = 0;
1446 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1447 how = fxdr_unsigned(int, *tl);
1448 switch (how) {
1449 case NFSV3CREATE_GUARDED:
1450 if (nd.ni_vp) {
1451 error = EEXIST;
1452 break;
1453 }
1454 case NFSV3CREATE_UNCHECKED:
1455 nfsm_srvsattr(&va);
1456 break;
1457 case NFSV3CREATE_EXCLUSIVE:
1458 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
1459 memcpy(cverf, cp, NFSX_V3CREATEVERF);
1460 exclusive_flag = 1;
1461 break;
1462 };
1463 va.va_type = VREG;
1464 } else {
1465 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1466 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1467 if (va.va_type == VNON)
1468 va.va_type = VREG;
1469 va.va_mode = nfstov_mode(sp->sa_mode);
1470 switch (va.va_type) {
1471 case VREG:
1472 tsize = fxdr_unsigned(int32_t, sp->sa_size);
1473 if (tsize != -1)
1474 va.va_size = (u_quad_t)tsize;
1475 break;
1476 case VCHR:
1477 case VBLK:
1478 case VFIFO:
1479 rdev = fxdr_unsigned(int32_t, sp->sa_size);
1480 break;
1481 default:
1482 break;
1483 };
1484 }
1485
1486 /*
1487 * Iff doesn't exist, create it
1488 * otherwise just truncate to 0 length
1489 * should I set the mode too ??
1490 */
1491 if (nd.ni_vp == NULL) {
1492 if (va.va_type == VREG || va.va_type == VSOCK) {
1493 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1494 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1495 if (!error) {
1496 if (exclusive_flag) {
1497 exclusive_flag = 0;
1498 VATTR_NULL(&va);
1499 /*
1500 * XXX
1501 * assuming NFSX_V3CREATEVERF
1502 * == sizeof(nfstime3)
1503 */
1504 fxdr_nfsv3time(cverf, &va.va_atime);
1505 error = VOP_SETATTR(nd.ni_vp, &va, cred,
1506 lwp);
1507 }
1508 }
1509 } else if (va.va_type == VCHR || va.va_type == VBLK ||
1510 va.va_type == VFIFO) {
1511 if (va.va_type == VCHR && rdev == 0xffffffff)
1512 va.va_type = VFIFO;
1513 if (va.va_type != VFIFO &&
1514 (error = kauth_authorize_generic(cred,
1515 KAUTH_GENERIC_ISSUSER, (u_short *)0))) {
1516 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1517 vput(nd.ni_dvp);
1518 nfsm_reply(0);
1519 vn_finished_write(mp, 0);
1520 return (error);
1521 } else
1522 va.va_rdev = (dev_t)rdev;
1523 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1524 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1525 &va);
1526 if (error) {
1527 nfsm_reply(0);
1528 }
1529 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1530 vrele(nd.ni_dvp);
1531 vput(nd.ni_vp);
1532 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1533 error = EINVAL;
1534 nfsm_reply(0);
1535 }
1536 } else {
1537 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1538 vput(nd.ni_dvp);
1539 error = ENXIO;
1540 }
1541 vp = nd.ni_vp;
1542 } else {
1543 vp = nd.ni_vp;
1544 if (nd.ni_dvp == vp)
1545 vrele(nd.ni_dvp);
1546 else
1547 vput(nd.ni_dvp);
1548 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1549 if (!error && va.va_size != -1) {
1550 error = nfsrv_access(vp, VWRITE, cred,
1551 (nd.ni_cnd.cn_flags & RDONLY), lwp, 0);
1552 if (!error) {
1553 nqsrv_getl(vp, ND_WRITE);
1554 tempsize = va.va_size;
1555 VATTR_NULL(&va);
1556 va.va_size = tempsize;
1557 error = VOP_SETATTR(vp, &va, cred, lwp);
1558 }
1559 }
1560 if (error)
1561 vput(vp);
1562 }
1563 if (!error) {
1564 memset((caddr_t)fhp, 0, sizeof(nfh));
1565 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1566 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1567 if (!error)
1568 error = VOP_GETATTR(vp, &va, cred, lwp);
1569 vput(vp);
1570 KASSERT(fhp->fh_fid.fid_len <= _VFS_MAXFIDSZ);
1571 }
1572 if (v3) {
1573 if (exclusive_flag && !error) {
1574 /*
1575 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3)
1576 */
1577 char oldverf[NFSX_V3CREATEVERF];
1578
1579 txdr_nfsv3time(&va.va_atime, oldverf);
1580 if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF))
1581 error = EEXIST;
1582 }
1583 if (dirp) {
1584 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
1585 vrele(dirp);
1586 }
1587 }
1588 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1589 if (v3) {
1590 if (!error) {
1591 nfsm_srvpostop_fh(fhp);
1592 nfsm_srvpostop_attr(0, &va);
1593 }
1594 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1595 } else {
1596 nfsm_srvfhtom(fhp, v3);
1597 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1598 nfsm_srvfillattr(&va, fp);
1599 }
1600 vn_finished_write(mp, 0);
1601 return (0);
1602 nfsmout:
1603 if (dirp)
1604 vrele(dirp);
1605 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1606 if (nd.ni_dvp == nd.ni_vp)
1607 vrele(nd.ni_dvp);
1608 else
1609 vput(nd.ni_dvp);
1610 if (nd.ni_vp)
1611 vput(nd.ni_vp);
1612 vn_finished_write(mp, 0);
1613 return (error);
1614 }
1615
1616 /*
1617 * nfs v3 mknod service
1618 */
1619 int
1620 nfsrv_mknod(nfsd, slp, lwp, mrq)
1621 struct nfsrv_descript *nfsd;
1622 struct nfssvc_sock *slp;
1623 struct lwp *lwp;
1624 struct mbuf **mrq;
1625 {
1626 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1627 struct mbuf *nam = nfsd->nd_nam;
1628 caddr_t dpos = nfsd->nd_dpos;
1629 kauth_cred_t cred = nfsd->nd_cr;
1630 struct vattr va, dirfor, diraft;
1631 u_int32_t *tl;
1632 struct nameidata nd;
1633 int32_t t1;
1634 caddr_t bpos;
1635 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
1636 u_int32_t major, minor;
1637 enum vtype vtyp;
1638 char *cp2;
1639 struct mbuf *mb, *mreq;
1640 struct vnode *vp, *dirp = (struct vnode *)0;
1641 nfsfh_t nfh;
1642 fhandle_t *fhp;
1643 u_quad_t frev;
1644 struct mount *mp = NULL;
1645
1646 nd.ni_cnd.cn_nameiop = 0;
1647 fhp = &nfh.fh_generic;
1648 nfsm_srvmtofh(fhp);
1649 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1650 return (ESTALE);
1651 vn_start_write(NULL, &mp, V_WAIT);
1652 nfsm_srvnamesiz(len);
1653 nd.ni_cnd.cn_cred = cred;
1654 nd.ni_cnd.cn_nameiop = CREATE;
1655 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1656 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1657 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1658 if (dirp)
1659 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
1660 if (error) {
1661 nfsm_reply(NFSX_WCCDATA(1));
1662 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1663 if (dirp)
1664 vrele(dirp);
1665 vn_finished_write(mp, 0);
1666 return (0);
1667 }
1668 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1669 vtyp = nfsv3tov_type(*tl);
1670 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1671 error = NFSERR_BADTYPE;
1672 goto abort;
1673 }
1674 VATTR_NULL(&va);
1675 va.va_mode = 0;
1676 nfsm_srvsattr(&va);
1677 if (vtyp == VCHR || vtyp == VBLK) {
1678 dev_t rdev;
1679
1680 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1681 major = fxdr_unsigned(u_int32_t, *tl++);
1682 minor = fxdr_unsigned(u_int32_t, *tl);
1683 rdev = makedev(major, minor);
1684 if (major(rdev) != major || minor(rdev) != minor) {
1685 error = EINVAL;
1686 goto abort;
1687 }
1688 va.va_rdev = rdev;
1689 }
1690
1691 /*
1692 * Iff doesn't exist, create it.
1693 */
1694 if (nd.ni_vp) {
1695 error = EEXIST;
1696 abort:
1697 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1698 if (nd.ni_dvp == nd.ni_vp)
1699 vrele(nd.ni_dvp);
1700 else
1701 vput(nd.ni_dvp);
1702 if (nd.ni_vp)
1703 vput(nd.ni_vp);
1704 goto out;
1705 }
1706 va.va_type = vtyp;
1707 if (vtyp == VSOCK) {
1708 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1709 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1710 } else {
1711 if (va.va_type != VFIFO &&
1712 (error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
1713 (u_short *)0))) {
1714 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1715 vput(nd.ni_dvp);
1716 goto out;
1717 }
1718 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1719 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1720 if (error)
1721 goto out;
1722 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1723 vput(nd.ni_vp);
1724 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1725 error = EINVAL;
1726 }
1727 }
1728 out:
1729 vp = nd.ni_vp;
1730 if (!error) {
1731 memset((caddr_t)fhp, 0, sizeof(nfh));
1732 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1733 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1734 if (!error)
1735 error = VOP_GETATTR(vp, &va, cred, lwp);
1736 vput(vp);
1737 KASSERT(fhp->fh_fid.fid_len <= _VFS_MAXFIDSZ);
1738 }
1739 if (dirp) {
1740 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
1741 vrele(dirp);
1742 }
1743 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
1744 if (!error) {
1745 nfsm_srvpostop_fh(fhp);
1746 nfsm_srvpostop_attr(0, &va);
1747 }
1748 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1749 vn_finished_write(mp, 0);
1750 return (0);
1751 nfsmout:
1752 if (dirp)
1753 vrele(dirp);
1754 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1755 if (nd.ni_dvp == nd.ni_vp)
1756 vrele(nd.ni_dvp);
1757 else
1758 vput(nd.ni_dvp);
1759 if (nd.ni_vp)
1760 vput(nd.ni_vp);
1761 vn_finished_write(mp, 0);
1762 return (error);
1763 }
1764
1765 /*
1766 * nfs remove service
1767 */
1768 int
1769 nfsrv_remove(nfsd, slp, lwp, mrq)
1770 struct nfsrv_descript *nfsd;
1771 struct nfssvc_sock *slp;
1772 struct lwp *lwp;
1773 struct mbuf **mrq;
1774 {
1775 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1776 struct mbuf *nam = nfsd->nd_nam;
1777 caddr_t dpos = nfsd->nd_dpos;
1778 kauth_cred_t cred = nfsd->nd_cr;
1779 struct nameidata nd;
1780 u_int32_t *tl;
1781 int32_t t1;
1782 caddr_t bpos;
1783 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
1784 int v3 = (nfsd->nd_flag & ND_NFSV3);
1785 char *cp2;
1786 struct mbuf *mb, *mreq;
1787 struct vnode *vp, *dirp;
1788 struct vattr dirfor, diraft;
1789 nfsfh_t nfh;
1790 fhandle_t *fhp;
1791 u_quad_t frev;
1792 struct mount *mp = NULL;
1793
1794 #ifndef nolint
1795 vp = (struct vnode *)0;
1796 #endif
1797 fhp = &nfh.fh_generic;
1798 nfsm_srvmtofh(fhp);
1799 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1800 return (ESTALE);
1801 vn_start_write(NULL, &mp, V_WAIT);
1802 nfsm_srvnamesiz(len);
1803 nd.ni_cnd.cn_cred = cred;
1804 nd.ni_cnd.cn_nameiop = DELETE;
1805 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1806 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1807 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1808 if (dirp) {
1809 if (v3)
1810 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
1811 else
1812 vrele(dirp);
1813 }
1814 if (!error) {
1815 vp = nd.ni_vp;
1816 if (vp->v_type == VDIR &&
1817 (error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
1818 (u_short *)0)) != 0)
1819 goto out;
1820 /*
1821 * The root of a mounted filesystem cannot be deleted.
1822 */
1823 if (vp->v_flag & VROOT) {
1824 error = EBUSY;
1825 goto out;
1826 }
1827 out:
1828 if (!error) {
1829 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1830 nqsrv_getl(vp, ND_WRITE);
1831 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1832 } else {
1833 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1834 if (nd.ni_dvp == vp)
1835 vrele(nd.ni_dvp);
1836 else
1837 vput(nd.ni_dvp);
1838 vput(vp);
1839 }
1840 }
1841 if (dirp && v3) {
1842 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
1843 vrele(dirp);
1844 }
1845 nfsm_reply(NFSX_WCCDATA(v3));
1846 if (v3) {
1847 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1848 vn_finished_write(mp, 0);
1849 return (0);
1850 }
1851 vn_finished_write(mp, 0);
1852 nfsm_srvdone;
1853 }
1854
1855 /*
1856 * nfs rename service
1857 */
1858 int
1859 nfsrv_rename(nfsd, slp, lwp, mrq)
1860 struct nfsrv_descript *nfsd;
1861 struct nfssvc_sock *slp;
1862 struct lwp *lwp;
1863 struct mbuf **mrq;
1864 {
1865 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1866 struct mbuf *nam = nfsd->nd_nam;
1867 caddr_t dpos = nfsd->nd_dpos;
1868 kauth_cred_t cred = nfsd->nd_cr;
1869 u_int32_t *tl;
1870 int32_t t1;
1871 caddr_t bpos;
1872 int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1;
1873 uint32_t len, len2;
1874 int tdirfor_ret = 1, tdiraft_ret = 1;
1875 int v3 = (nfsd->nd_flag & ND_NFSV3);
1876 char *cp2;
1877 struct mbuf *mb, *mreq;
1878 struct nameidata fromnd, tond;
1879 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
1880 struct vnode *tdirp = (struct vnode *)0;
1881 struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1882 nfsfh_t fnfh, tnfh;
1883 fhandle_t *ffhp, *tfhp;
1884 u_quad_t frev;
1885 uid_t saved_uid;
1886 struct mount *mp = NULL;
1887
1888 #ifndef nolint
1889 fvp = (struct vnode *)0;
1890 #endif
1891 ffhp = &fnfh.fh_generic;
1892 tfhp = &tnfh.fh_generic;
1893 fromnd.ni_cnd.cn_nameiop = 0;
1894 tond.ni_cnd.cn_nameiop = 0;
1895 nfsm_srvmtofh(ffhp);
1896 if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL)
1897 return (ESTALE);
1898 vn_start_write(NULL, &mp, V_WAIT);
1899 nfsm_srvnamesiz(len);
1900 /*
1901 * Remember our original uid so that we can reset cr_uid before
1902 * the second nfs_namei() call, in case it is remapped.
1903 */
1904 saved_uid = kauth_cred_geteuid(cred);
1905 fromnd.ni_cnd.cn_cred = cred;
1906 fromnd.ni_cnd.cn_nameiop = DELETE;
1907 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART;
1908 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
1909 &dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1910 if (fdirp) {
1911 if (v3)
1912 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp);
1913 else {
1914 vrele(fdirp);
1915 fdirp = (struct vnode *)0;
1916 }
1917 }
1918 if (error) {
1919 nfsm_reply(2 * NFSX_WCCDATA(v3));
1920 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1921 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1922 if (fdirp)
1923 vrele(fdirp);
1924 vn_finished_write(mp, 0);
1925 return (0);
1926 }
1927 fvp = fromnd.ni_vp;
1928 nfsm_srvmtofh(tfhp);
1929 if (v3) {
1930 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
1931 len2 = fxdr_unsigned(uint32_t, *tl);
1932 /* len2 will be checked by nfs_namei */
1933 }
1934 else {
1935 /* NFSv2 */
1936 nfsm_strsiz(len2, NFS_MAXNAMLEN);
1937 }
1938 kauth_cred_seteuid(cred, saved_uid);
1939 tond.ni_cnd.cn_cred = cred;
1940 tond.ni_cnd.cn_nameiop = RENAME;
1941 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
1942 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
1943 &dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1944 if (tdirp) {
1945 if (v3)
1946 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp);
1947 else {
1948 vrele(tdirp);
1949 tdirp = (struct vnode *)0;
1950 }
1951 }
1952 if (error) {
1953 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1954 vrele(fromnd.ni_dvp);
1955 vrele(fvp);
1956 goto out1;
1957 }
1958 tdvp = tond.ni_dvp;
1959 tvp = tond.ni_vp;
1960 if (tvp != NULL) {
1961 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1962 if (v3)
1963 error = EEXIST;
1964 else
1965 error = EISDIR;
1966 goto out;
1967 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1968 if (v3)
1969 error = EEXIST;
1970 else
1971 error = ENOTDIR;
1972 goto out;
1973 }
1974 if (tvp->v_type == VDIR && tvp->v_mountedhere) {
1975 if (v3)
1976 error = EXDEV;
1977 else
1978 error = ENOTEMPTY;
1979 goto out;
1980 }
1981 }
1982 if (fvp->v_type == VDIR && fvp->v_mountedhere) {
1983 if (v3)
1984 error = EXDEV;
1985 else
1986 error = ENOTEMPTY;
1987 goto out;
1988 }
1989 if (fvp->v_mount != tdvp->v_mount) {
1990 if (v3)
1991 error = EXDEV;
1992 else
1993 error = ENOTEMPTY;
1994 goto out;
1995 }
1996 if (fvp == tdvp) {
1997 if (v3)
1998 error = EINVAL;
1999 else
2000 error = ENOTEMPTY;
2001 }
2002 /*
2003 * If source is the same as the destination (that is the
2004 * same vnode with the same name in the same directory),
2005 * then there is nothing to do.
2006 */
2007 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
2008 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2009 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
2010 fromnd.ni_cnd.cn_namelen))
2011 error = -1;
2012 out:
2013 if (!error) {
2014 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
2015 nqsrv_getl(tdvp, ND_WRITE);
2016 if (tvp) {
2017 nqsrv_getl(tvp, ND_WRITE);
2018 }
2019 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2020 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2021 } else {
2022 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2023 if (tdvp == tvp)
2024 vrele(tdvp);
2025 else
2026 vput(tdvp);
2027 if (tvp)
2028 vput(tvp);
2029 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2030 vrele(fromnd.ni_dvp);
2031 vrele(fvp);
2032 if (error == -1)
2033 error = 0;
2034 }
2035 vrele(tond.ni_startdir);
2036 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2037 out1:
2038 if (fdirp) {
2039 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp);
2040 vrele(fdirp);
2041 }
2042 if (tdirp) {
2043 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp);
2044 vrele(tdirp);
2045 }
2046 vrele(fromnd.ni_startdir);
2047 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2048 nfsm_reply(2 * NFSX_WCCDATA(v3));
2049 if (v3) {
2050 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
2051 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
2052 }
2053 vn_finished_write(mp, 0);
2054 return (0);
2055
2056 nfsmout:
2057 if (fdirp)
2058 vrele(fdirp);
2059 #ifdef notdef
2060 if (tdirp)
2061 vrele(tdirp);
2062 #endif
2063 if (tond.ni_cnd.cn_nameiop) {
2064 vrele(tond.ni_startdir);
2065 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2066 }
2067 if (fromnd.ni_cnd.cn_nameiop) {
2068 vrele(fromnd.ni_startdir);
2069 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2070 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2071 vrele(fromnd.ni_dvp);
2072 vrele(fvp);
2073 }
2074 return (error);
2075 }
2076
2077 /*
2078 * nfs link service
2079 */
2080 int
2081 nfsrv_link(nfsd, slp, lwp, mrq)
2082 struct nfsrv_descript *nfsd;
2083 struct nfssvc_sock *slp;
2084 struct lwp *lwp;
2085 struct mbuf **mrq;
2086 {
2087 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2088 struct mbuf *nam = nfsd->nd_nam;
2089 caddr_t dpos = nfsd->nd_dpos;
2090 kauth_cred_t cred = nfsd->nd_cr;
2091 struct nameidata nd;
2092 u_int32_t *tl;
2093 int32_t t1;
2094 caddr_t bpos;
2095 int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2096 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
2097 char *cp2;
2098 struct mbuf *mb, *mreq;
2099 struct vnode *vp, *xp, *dirp = (struct vnode *)0;
2100 struct vattr dirfor, diraft, at;
2101 nfsfh_t nfh, dnfh;
2102 fhandle_t *fhp, *dfhp;
2103 u_quad_t frev;
2104 struct mount *mp = NULL;
2105
2106 fhp = &nfh.fh_generic;
2107 dfhp = &dnfh.fh_generic;
2108 nfsm_srvmtofh(fhp);
2109 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2110 return (ESTALE);
2111 vn_start_write(NULL, &mp, V_WAIT);
2112 nfsm_srvmtofh(dfhp);
2113 nfsm_srvnamesiz(len);
2114 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
2115 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2116 if (error) {
2117 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2118 nfsm_srvpostop_attr(getret, &at);
2119 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2120 vn_finished_write(mp, 0);
2121 return (0);
2122 }
2123 if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred,
2124 KAUTH_GENERIC_ISSUSER, (u_short *)0)) != 0)
2125 goto out1;
2126 nd.ni_cnd.cn_cred = cred;
2127 nd.ni_cnd.cn_nameiop = CREATE;
2128 nd.ni_cnd.cn_flags = LOCKPARENT;
2129 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
2130 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2131 if (dirp) {
2132 if (v3)
2133 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
2134 else {
2135 vrele(dirp);
2136 dirp = (struct vnode *)0;
2137 }
2138 }
2139 if (error)
2140 goto out1;
2141 xp = nd.ni_vp;
2142 if (xp != NULL) {
2143 error = EEXIST;
2144 goto out;
2145 }
2146 xp = nd.ni_dvp;
2147 if (vp->v_mount != xp->v_mount)
2148 error = EXDEV;
2149 out:
2150 if (!error) {
2151 nqsrv_getl(vp, ND_WRITE);
2152 nqsrv_getl(xp, ND_WRITE);
2153 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2154 } else {
2155 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2156 if (nd.ni_dvp == nd.ni_vp)
2157 vrele(nd.ni_dvp);
2158 else
2159 vput(nd.ni_dvp);
2160 if (nd.ni_vp)
2161 vrele(nd.ni_vp);
2162 }
2163 out1:
2164 if (v3)
2165 getret = VOP_GETATTR(vp, &at, cred, lwp);
2166 if (dirp) {
2167 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
2168 vrele(dirp);
2169 }
2170 vrele(vp);
2171 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2172 if (v3) {
2173 nfsm_srvpostop_attr(getret, &at);
2174 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2175 vn_finished_write(mp, 0);
2176 return (0);
2177 }
2178 vn_finished_write(mp, 0);
2179 nfsm_srvdone;
2180 }
2181
2182 /*
2183 * nfs symbolic link service
2184 */
2185 int
2186 nfsrv_symlink(nfsd, slp, lwp, mrq)
2187 struct nfsrv_descript *nfsd;
2188 struct nfssvc_sock *slp;
2189 struct lwp *lwp;
2190 struct mbuf **mrq;
2191 {
2192 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2193 struct mbuf *nam = nfsd->nd_nam;
2194 caddr_t dpos = nfsd->nd_dpos;
2195 kauth_cred_t cred = nfsd->nd_cr;
2196 struct vattr va, dirfor, diraft;
2197 struct nameidata nd;
2198 u_int32_t *tl;
2199 int32_t t1;
2200 struct nfsv2_sattr *sp;
2201 char *bpos, *pathcp = NULL, *cp2;
2202 struct uio io;
2203 struct iovec iv;
2204 int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1;
2205 uint32_t len, len2;
2206 int v3 = (nfsd->nd_flag & ND_NFSV3);
2207 struct mbuf *mb, *mreq;
2208 struct vnode *dirp = (struct vnode *)0;
2209 nfsfh_t nfh;
2210 fhandle_t *fhp;
2211 u_quad_t frev;
2212 struct mount *mp = NULL;
2213
2214 nd.ni_cnd.cn_nameiop = 0;
2215 fhp = &nfh.fh_generic;
2216 nfsm_srvmtofh(fhp);
2217 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2218 return (ESTALE);
2219 vn_start_write(NULL, &mp, V_WAIT);
2220 nfsm_srvnamesiz(len);
2221 nd.ni_cnd.cn_cred = cred;
2222 nd.ni_cnd.cn_nameiop = CREATE;
2223 nd.ni_cnd.cn_flags = LOCKPARENT;
2224 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2225 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2226 if (dirp) {
2227 if (v3)
2228 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
2229 else {
2230 vrele(dirp);
2231 dirp = (struct vnode *)0;
2232 }
2233 }
2234 if (error)
2235 goto out;
2236 VATTR_NULL(&va);
2237 va.va_type = VLNK;
2238 if (v3) {
2239 va.va_mode = 0;
2240 nfsm_srvsattr(&va);
2241 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
2242 len2 = fxdr_unsigned(uint32_t, *tl);
2243 if (len2 > PATH_MAX) {
2244 /* XXX should check _PC_NO_TRUNC */
2245 error = ENAMETOOLONG;
2246 goto abortop;
2247 }
2248 }
2249 else {
2250 /* NFSv2 */
2251 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2252 }
2253 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK);
2254 iv.iov_base = pathcp;
2255 iv.iov_len = len2;
2256 io.uio_resid = len2;
2257 io.uio_offset = 0;
2258 io.uio_iov = &iv;
2259 io.uio_iovcnt = 1;
2260 io.uio_rw = UIO_READ;
2261 UIO_SETUP_SYSSPACE(&io);
2262 nfsm_mtouio(&io, len2);
2263 if (!v3) {
2264 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2265 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2266 }
2267 *(pathcp + len2) = '\0';
2268 if (nd.ni_vp) {
2269 error = EEXIST;
2270 abortop:
2271 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2272 if (nd.ni_dvp == nd.ni_vp)
2273 vrele(nd.ni_dvp);
2274 else
2275 vput(nd.ni_dvp);
2276 if (nd.ni_vp)
2277 vrele(nd.ni_vp);
2278 goto out;
2279 }
2280 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2281 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2282 if (!error) {
2283 if (v3) {
2284 memset((caddr_t)fhp, 0, sizeof(nfh));
2285 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx;
2286 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2287 if (!error)
2288 error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp);
2289 vput(nd.ni_vp);
2290 KASSERT(fhp->fh_fid.fid_len <= _VFS_MAXFIDSZ);
2291 } else {
2292 vput(nd.ni_vp);
2293 }
2294 }
2295 out:
2296 if (pathcp)
2297 free(pathcp, M_TEMP);
2298 if (dirp) {
2299 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
2300 vrele(dirp);
2301 }
2302 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2303 if (v3) {
2304 if (!error) {
2305 nfsm_srvpostop_fh(fhp);
2306 nfsm_srvpostop_attr(0, &va);
2307 }
2308 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2309 }
2310 vn_finished_write(mp, 0);
2311 return (0);
2312 nfsmout:
2313 if (dirp)
2314 vrele(dirp);
2315 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2316 if (nd.ni_dvp == nd.ni_vp)
2317 vrele(nd.ni_dvp);
2318 else
2319 vput(nd.ni_dvp);
2320 if (nd.ni_vp)
2321 vrele(nd.ni_vp);
2322 if (pathcp)
2323 free(pathcp, M_TEMP);
2324 vn_finished_write(mp, 0);
2325 return (error);
2326 }
2327
2328 /*
2329 * nfs mkdir service
2330 */
2331 int
2332 nfsrv_mkdir(nfsd, slp, lwp, mrq)
2333 struct nfsrv_descript *nfsd;
2334 struct nfssvc_sock *slp;
2335 struct lwp *lwp;
2336 struct mbuf **mrq;
2337 {
2338 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2339 struct mbuf *nam = nfsd->nd_nam;
2340 caddr_t dpos = nfsd->nd_dpos;
2341 kauth_cred_t cred = nfsd->nd_cr;
2342 struct vattr va, dirfor, diraft;
2343 struct nfs_fattr *fp;
2344 struct nameidata nd;
2345 caddr_t cp;
2346 u_int32_t *tl;
2347 int32_t t1;
2348 caddr_t bpos;
2349 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2350 int v3 = (nfsd->nd_flag & ND_NFSV3);
2351 char *cp2;
2352 struct mbuf *mb, *mreq;
2353 struct vnode *vp, *dirp = (struct vnode *)0;
2354 nfsfh_t nfh;
2355 fhandle_t *fhp;
2356 u_quad_t frev;
2357 struct mount *mp = NULL;
2358
2359 fhp = &nfh.fh_generic;
2360 nfsm_srvmtofh(fhp);
2361 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2362 return (ESTALE);
2363 vn_start_write(NULL, &mp, V_WAIT);
2364 nfsm_srvnamesiz(len);
2365 nd.ni_cnd.cn_cred = cred;
2366 nd.ni_cnd.cn_nameiop = CREATE;
2367 nd.ni_cnd.cn_flags = LOCKPARENT;
2368 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2369 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2370 if (dirp) {
2371 if (v3)
2372 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
2373 else {
2374 vrele(dirp);
2375 dirp = (struct vnode *)0;
2376 }
2377 }
2378 if (error) {
2379 nfsm_reply(NFSX_WCCDATA(v3));
2380 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2381 if (dirp)
2382 vrele(dirp);
2383 vn_finished_write(mp, 0);
2384 return (0);
2385 }
2386 VATTR_NULL(&va);
2387 if (v3) {
2388 va.va_mode = 0;
2389 nfsm_srvsattr(&va);
2390 } else {
2391 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2392 va.va_mode = nfstov_mode(*tl++);
2393 }
2394 va.va_type = VDIR;
2395 vp = nd.ni_vp;
2396 if (vp != NULL) {
2397 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2398 if (nd.ni_dvp == vp)
2399 vrele(nd.ni_dvp);
2400 else
2401 vput(nd.ni_dvp);
2402 vrele(vp);
2403 error = EEXIST;
2404 goto out;
2405 }
2406 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2407 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2408 if (!error) {
2409 vp = nd.ni_vp;
2410 memset((caddr_t)fhp, 0, sizeof(nfh));
2411 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
2412 error = VFS_VPTOFH(vp, &fhp->fh_fid);
2413 if (!error)
2414 error = VOP_GETATTR(vp, &va, cred, lwp);
2415 vput(vp);
2416 KASSERT(fhp->fh_fid.fid_len <= _VFS_MAXFIDSZ);
2417 }
2418 out:
2419 if (dirp) {
2420 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
2421 vrele(dirp);
2422 }
2423 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2424 if (v3) {
2425 if (!error) {
2426 nfsm_srvpostop_fh(fhp);
2427 nfsm_srvpostop_attr(0, &va);
2428 }
2429 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2430 } else {
2431 nfsm_srvfhtom(fhp, v3);
2432 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2433 nfsm_srvfillattr(&va, fp);
2434 }
2435 vn_finished_write(mp, 0);
2436 return (0);
2437 nfsmout:
2438 if (dirp)
2439 vrele(dirp);
2440 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2441 if (nd.ni_dvp == nd.ni_vp)
2442 vrele(nd.ni_dvp);
2443 else
2444 vput(nd.ni_dvp);
2445 if (nd.ni_vp)
2446 vrele(nd.ni_vp);
2447 vn_finished_write(mp, 0);
2448 return (error);
2449 }
2450
2451 /*
2452 * nfs rmdir service
2453 */
2454 int
2455 nfsrv_rmdir(nfsd, slp, lwp, mrq)
2456 struct nfsrv_descript *nfsd;
2457 struct nfssvc_sock *slp;
2458 struct lwp *lwp;
2459 struct mbuf **mrq;
2460 {
2461 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2462 struct mbuf *nam = nfsd->nd_nam;
2463 caddr_t dpos = nfsd->nd_dpos;
2464 kauth_cred_t cred = nfsd->nd_cr;
2465 u_int32_t *tl;
2466 int32_t t1;
2467 caddr_t bpos;
2468 int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2469 int v3 = (nfsd->nd_flag & ND_NFSV3);
2470 char *cp2;
2471 struct mbuf *mb, *mreq;
2472 struct vnode *vp, *dirp = (struct vnode *)0;
2473 struct vattr dirfor, diraft;
2474 nfsfh_t nfh;
2475 fhandle_t *fhp;
2476 struct nameidata nd;
2477 u_quad_t frev;
2478 struct mount *mp = NULL;
2479
2480 fhp = &nfh.fh_generic;
2481 nfsm_srvmtofh(fhp);
2482 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2483 return (ESTALE);
2484 vn_start_write(NULL, &mp, V_WAIT);
2485 nfsm_srvnamesiz(len);
2486 nd.ni_cnd.cn_cred = cred;
2487 nd.ni_cnd.cn_nameiop = DELETE;
2488 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2489 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2490 &dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2491 if (dirp) {
2492 if (v3)
2493 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
2494 else {
2495 vrele(dirp);
2496 dirp = (struct vnode *)0;
2497 }
2498 }
2499 if (error) {
2500 nfsm_reply(NFSX_WCCDATA(v3));
2501 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2502 if (dirp)
2503 vrele(dirp);
2504 vn_finished_write(mp, 0);
2505 return (0);
2506 }
2507 vp = nd.ni_vp;
2508 if (vp->v_type != VDIR) {
2509 error = ENOTDIR;
2510 goto out;
2511 }
2512 /*
2513 * No rmdir "." please.
2514 */
2515 if (nd.ni_dvp == vp) {
2516 error = EINVAL;
2517 goto out;
2518 }
2519 /*
2520 * The root of a mounted filesystem cannot be deleted.
2521 */
2522 if (vp->v_flag & VROOT)
2523 error = EBUSY;
2524 out:
2525 if (!error) {
2526 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2527 nqsrv_getl(vp, ND_WRITE);
2528 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2529 } else {
2530 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2531 if (nd.ni_dvp == nd.ni_vp)
2532 vrele(nd.ni_dvp);
2533 else
2534 vput(nd.ni_dvp);
2535 vput(vp);
2536 }
2537 if (dirp) {
2538 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
2539 vrele(dirp);
2540 }
2541 nfsm_reply(NFSX_WCCDATA(v3));
2542 if (v3) {
2543 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2544 vn_finished_write(mp, 0);
2545 return (0);
2546 }
2547 vn_finished_write(mp, 0);
2548 nfsm_srvdone;
2549 }
2550
2551 /*
2552 * nfs readdir service
2553 * - mallocs what it thinks is enough to read
2554 * count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR
2555 * - calls VOP_READDIR()
2556 * - loops around building the reply
2557 * if the output generated exceeds count break out of loop
2558 * The nfsm_clget macro is used here so that the reply will be packed
2559 * tightly in mbuf clusters.
2560 * - it only knows that it has encountered eof when the VOP_READDIR()
2561 * reads nothing
2562 * - as such one readdir rpc will return eof false although you are there
2563 * and then the next will return eof
2564 * - it trims out records with d_fileno == 0
2565 * this doesn't matter for Unix clients, but they might confuse clients
2566 * for other os'.
2567 * - it trims out records with d_type == DT_WHT
2568 * these cannot be seen through NFS (unless we extend the protocol)
2569 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2570 * than requested, but this may not apply to all filesystems. For
2571 * example, client NFS does not { although it is never remote mounted
2572 * anyhow }
2573 * The alternate call nfsrv_readdirplus() does lookups as well.
2574 * PS: The NFS protocol spec. does not clarify what the "count" byte
2575 * argument is a count of.. just name strings and file id's or the
2576 * entire reply rpc or ...
2577 * I tried just file name and id sizes and it confused the Sun client,
2578 * so I am using the full rpc size now. The "paranoia.." comment refers
2579 * to including the status longwords that are not a part of the dir.
2580 * "entry" structures, but are in the rpc.
2581 */
2582
2583 #define NFS_SRVDIRBLKSIZ 1024
2584
2585 struct flrep {
2586 nfsuint64 fl_off;
2587 u_int32_t fl_postopok;
2588 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2589 u_int32_t fl_fhok;
2590 u_int32_t fl_fhsize;
2591 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2592 };
2593
2594 int
2595 nfsrv_readdir(nfsd, slp, lwp, mrq)
2596 struct nfsrv_descript *nfsd;
2597 struct nfssvc_sock *slp;
2598 struct lwp *lwp;
2599 struct mbuf **mrq;
2600 {
2601 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2602 struct mbuf *nam = nfsd->nd_nam;
2603 caddr_t dpos = nfsd->nd_dpos;
2604 kauth_cred_t cred = nfsd->nd_cr;
2605 char *bp, *be;
2606 struct mbuf *mp;
2607 struct dirent *dp;
2608 caddr_t cp;
2609 u_int32_t *tl;
2610 int32_t t1;
2611 caddr_t bpos;
2612 struct mbuf *mb, *mreq, *mp2;
2613 char *cpos, *cend, *cp2, *rbuf;
2614 struct vnode *vp;
2615 struct vattr at;
2616 nfsfh_t nfh;
2617 fhandle_t *fhp;
2618 struct uio io;
2619 struct iovec iv;
2620 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2621 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies;
2622 int v3 = (nfsd->nd_flag & ND_NFSV3);
2623 u_quad_t frev, off, toff, verf;
2624 off_t *cookies = NULL, *cookiep;
2625 nfsuint64 jar;
2626
2627 fhp = &nfh.fh_generic;
2628 nfsm_srvmtofh(fhp);
2629 if (v3) {
2630 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2631 toff = fxdr_hyper(tl);
2632 tl += 2;
2633 verf = fxdr_hyper(tl);
2634 tl += 2;
2635 } else {
2636 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2637 toff = fxdr_unsigned(u_quad_t, *tl++);
2638 }
2639 off = toff;
2640 cnt = fxdr_unsigned(int, *tl);
2641 siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1));
2642 xfer = NFS_SRVMAXDATA(nfsd);
2643 if (siz > xfer)
2644 siz = xfer;
2645 fullsiz = siz;
2646 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2647 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2648 if (!error && vp->v_type != VDIR) {
2649 error = ENOTDIR;
2650 vput(vp);
2651 }
2652 if (error) {
2653 nfsm_reply(NFSX_UNSIGNED);
2654 nfsm_srvpostop_attr(getret, &at);
2655 return (0);
2656 }
2657 nqsrv_getl(vp, ND_READ);
2658 if (v3) {
2659 error = getret = VOP_GETATTR(vp, &at, cred, lwp);
2660 #ifdef NFS3_STRICTVERF
2661 /*
2662 * XXX This check is too strict for Solaris 2.5 clients.
2663 */
2664 if (!error && toff && verf != at.va_filerev)
2665 error = NFSERR_BAD_COOKIE;
2666 #endif
2667 }
2668 if (!error)
2669 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0);
2670 if (error) {
2671 vput(vp);
2672 nfsm_reply(NFSX_POSTOPATTR(v3));
2673 nfsm_srvpostop_attr(getret, &at);
2674 return (0);
2675 }
2676 VOP_UNLOCK(vp, 0);
2677 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2678 again:
2679 iv.iov_base = rbuf;
2680 iv.iov_len = fullsiz;
2681 io.uio_iov = &iv;
2682 io.uio_iovcnt = 1;
2683 io.uio_offset = (off_t)off;
2684 io.uio_resid = fullsiz;
2685 io.uio_rw = UIO_READ;
2686 UIO_SETUP_SYSSPACE(&io);
2687 eofflag = 0;
2688 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2689
2690 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2691
2692 off = (off_t)io.uio_offset;
2693 if (!cookies && !error)
2694 error = NFSERR_PERM;
2695 if (v3) {
2696 getret = VOP_GETATTR(vp, &at, cred, lwp);
2697 if (!error)
2698 error = getret;
2699 }
2700
2701 VOP_UNLOCK(vp, 0);
2702 if (error) {
2703 vrele(vp);
2704 free((caddr_t)rbuf, M_TEMP);
2705 if (cookies)
2706 free((caddr_t)cookies, M_TEMP);
2707 nfsm_reply(NFSX_POSTOPATTR(v3));
2708 nfsm_srvpostop_attr(getret, &at);
2709 return (0);
2710 }
2711 if (io.uio_resid) {
2712 siz -= io.uio_resid;
2713
2714 /*
2715 * If nothing read, return eof
2716 * rpc reply
2717 */
2718 if (siz == 0) {
2719 vrele(vp);
2720 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2721 2 * NFSX_UNSIGNED);
2722 if (v3) {
2723 nfsm_srvpostop_attr(getret, &at);
2724 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2725 txdr_hyper(at.va_filerev, tl);
2726 tl += 2;
2727 } else
2728 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2729 *tl++ = nfs_false;
2730 *tl = nfs_true;
2731 free((caddr_t)rbuf, M_TEMP);
2732 free((caddr_t)cookies, M_TEMP);
2733 return (0);
2734 }
2735 }
2736
2737 /*
2738 * Check for degenerate cases of nothing useful read.
2739 * If so go try again
2740 */
2741 cpos = rbuf;
2742 cend = rbuf + siz;
2743 dp = (struct dirent *)cpos;
2744 cookiep = cookies;
2745
2746 while (cpos < cend && ncookies > 0 &&
2747 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2748 cpos += dp->d_reclen;
2749 dp = (struct dirent *)cpos;
2750 cookiep++;
2751 ncookies--;
2752 }
2753 if (cpos >= cend || ncookies == 0) {
2754 toff = off;
2755 siz = fullsiz;
2756 goto again;
2757 }
2758
2759 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
2760 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2761 if (v3) {
2762 nfsm_srvpostop_attr(getret, &at);
2763 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2764 txdr_hyper(at.va_filerev, tl);
2765 }
2766 mp = mp2 = mb;
2767 bp = bpos;
2768 be = bp + M_TRAILINGSPACE(mp);
2769
2770 /* Loop through the records and build reply */
2771 while (cpos < cend && ncookies > 0) {
2772 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2773 nlen = dp->d_namlen;
2774 rem = nfsm_rndup(nlen)-nlen;
2775 len += (4 * NFSX_UNSIGNED + nlen + rem);
2776 if (v3)
2777 len += 2 * NFSX_UNSIGNED;
2778 if (len > cnt) {
2779 eofflag = 0;
2780 break;
2781 }
2782 /*
2783 * Build the directory record xdr from
2784 * the dirent entry.
2785 */
2786 nfsm_clget;
2787 *tl = nfs_true;
2788 bp += NFSX_UNSIGNED;
2789 if (v3) {
2790 nfsm_clget;
2791 *tl = txdr_unsigned(dp->d_fileno >> 32);
2792 bp += NFSX_UNSIGNED;
2793 }
2794 nfsm_clget;
2795 *tl = txdr_unsigned(dp->d_fileno);
2796 bp += NFSX_UNSIGNED;
2797 nfsm_clget;
2798 *tl = txdr_unsigned(nlen);
2799 bp += NFSX_UNSIGNED;
2800
2801 /* And loop around copying the name */
2802 xfer = nlen;
2803 cp = dp->d_name;
2804 while (xfer > 0) {
2805 nfsm_clget;
2806 if ((bp+xfer) > be)
2807 tsiz = be-bp;
2808 else
2809 tsiz = xfer;
2810 memcpy(bp, cp, tsiz);
2811 bp += tsiz;
2812 xfer -= tsiz;
2813 if (xfer > 0)
2814 cp += tsiz;
2815 }
2816 /* And null pad to an int32_t boundary */
2817 for (i = 0; i < rem; i++)
2818 *bp++ = '\0';
2819 nfsm_clget;
2820
2821 /* Finish off the record */
2822 txdr_hyper(*cookiep, &jar);
2823 if (v3) {
2824 *tl = jar.nfsuquad[0];
2825 bp += NFSX_UNSIGNED;
2826 nfsm_clget;
2827 }
2828 *tl = jar.nfsuquad[1];
2829 bp += NFSX_UNSIGNED;
2830 }
2831 cpos += dp->d_reclen;
2832 dp = (struct dirent *)cpos;
2833 cookiep++;
2834 ncookies--;
2835 }
2836 vrele(vp);
2837 nfsm_clget;
2838 *tl = nfs_false;
2839 bp += NFSX_UNSIGNED;
2840 nfsm_clget;
2841 if (eofflag)
2842 *tl = nfs_true;
2843 else
2844 *tl = nfs_false;
2845 bp += NFSX_UNSIGNED;
2846 if (mp != mb) {
2847 if (bp < be)
2848 mp->m_len = bp - mtod(mp, caddr_t);
2849 } else
2850 mp->m_len += bp - bpos;
2851 free((caddr_t)rbuf, M_TEMP);
2852 free((caddr_t)cookies, M_TEMP);
2853 nfsm_srvdone;
2854 }
2855
2856 int
2857 nfsrv_readdirplus(nfsd, slp, lwp, mrq)
2858 struct nfsrv_descript *nfsd;
2859 struct nfssvc_sock *slp;
2860 struct lwp *lwp;
2861 struct mbuf **mrq;
2862 {
2863 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2864 struct mbuf *nam = nfsd->nd_nam;
2865 caddr_t dpos = nfsd->nd_dpos;
2866 kauth_cred_t cred = nfsd->nd_cr;
2867 char *bp, *be;
2868 struct mbuf *mp;
2869 struct dirent *dp;
2870 caddr_t cp;
2871 u_int32_t *tl;
2872 int32_t t1;
2873 caddr_t bpos;
2874 struct mbuf *mb, *mreq, *mp2;
2875 char *cpos, *cend, *cp2, *rbuf;
2876 struct vnode *vp, *nvp;
2877 struct flrep fl;
2878 nfsfh_t nfh;
2879 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
2880 struct uio io;
2881 struct iovec iv;
2882 struct vattr va, at, *vap = &va;
2883 struct nfs_fattr *fp;
2884 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2885 int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies;
2886 u_quad_t frev, off, toff, verf;
2887 off_t *cookies = NULL, *cookiep;
2888
2889 fhp = &nfh.fh_generic;
2890 nfsm_srvmtofh(fhp);
2891 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2892 toff = fxdr_hyper(tl);
2893 tl += 2;
2894 verf = fxdr_hyper(tl);
2895 tl += 2;
2896 siz = fxdr_unsigned(int, *tl++);
2897 cnt = fxdr_unsigned(int, *tl);
2898 off = toff;
2899 siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1));
2900 xfer = NFS_SRVMAXDATA(nfsd);
2901 if (siz > xfer)
2902 siz = xfer;
2903 fullsiz = siz;
2904 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2905 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2906 if (!error && vp->v_type != VDIR) {
2907 error = ENOTDIR;
2908 vput(vp);
2909 }
2910 if (error) {
2911 nfsm_reply(NFSX_UNSIGNED);
2912 nfsm_srvpostop_attr(getret, &at);
2913 return (0);
2914 }
2915 error = getret = VOP_GETATTR(vp, &at, cred, lwp);
2916 #ifdef NFS3_STRICTVERF
2917 /*
2918 * XXX This check is too strict for Solaris 2.5 clients.
2919 */
2920 if (!error && toff && verf != at.va_filerev)
2921 error = NFSERR_BAD_COOKIE;
2922 #endif
2923 if (!error) {
2924 nqsrv_getl(vp, ND_READ);
2925 error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0);
2926 }
2927 if (error) {
2928 vput(vp);
2929 nfsm_reply(NFSX_V3POSTOPATTR);
2930 nfsm_srvpostop_attr(getret, &at);
2931 return (0);
2932 }
2933 VOP_UNLOCK(vp, 0);
2934
2935 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2936 again:
2937 iv.iov_base = rbuf;
2938 iv.iov_len = fullsiz;
2939 io.uio_iov = &iv;
2940 io.uio_iovcnt = 1;
2941 io.uio_offset = (off_t)off;
2942 io.uio_resid = fullsiz;
2943 io.uio_rw = UIO_READ;
2944 UIO_SETUP_SYSSPACE(&io);
2945 eofflag = 0;
2946
2947 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2948
2949 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2950
2951 off = (u_quad_t)io.uio_offset;
2952 getret = VOP_GETATTR(vp, &at, cred, lwp);
2953
2954 VOP_UNLOCK(vp, 0);
2955
2956 /*
2957 * If the VGET operation doesn't work for this filesystem,
2958 * we can't support readdirplus. Returning NOTSUPP should
2959 * make clients fall back to plain readdir.
2960 * There's no need to check for VPTOFH as well, we wouldn't
2961 * even be here otherwise.
2962 */
2963 if (!getret) {
2964 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
2965 getret = (getret == EOPNOTSUPP) ?
2966 NFSERR_NOTSUPP : NFSERR_IO;
2967 else
2968 vput(nvp);
2969 }
2970
2971 if (!cookies && !error)
2972 error = NFSERR_PERM;
2973 if (!error)
2974 error = getret;
2975 if (error) {
2976 vrele(vp);
2977 if (cookies)
2978 free((caddr_t)cookies, M_TEMP);
2979 free((caddr_t)rbuf, M_TEMP);
2980 nfsm_reply(NFSX_V3POSTOPATTR);
2981 nfsm_srvpostop_attr(getret, &at);
2982 return (0);
2983 }
2984 if (io.uio_resid) {
2985 siz -= io.uio_resid;
2986
2987 /*
2988 * If nothing read, return eof
2989 * rpc reply
2990 */
2991 if (siz == 0) {
2992 vrele(vp);
2993 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
2994 2 * NFSX_UNSIGNED);
2995 nfsm_srvpostop_attr(getret, &at);
2996 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2997 txdr_hyper(at.va_filerev, tl);
2998 tl += 2;
2999 *tl++ = nfs_false;
3000 *tl = nfs_true;
3001 free((caddr_t)cookies, M_TEMP);
3002 free((caddr_t)rbuf, M_TEMP);
3003 return (0);
3004 }
3005 }
3006
3007 /*
3008 * Check for degenerate cases of nothing useful read.
3009 * If so go try again
3010 */
3011 cpos = rbuf;
3012 cend = rbuf + siz;
3013 dp = (struct dirent *)cpos;
3014 cookiep = cookies;
3015
3016 while (cpos < cend && ncookies > 0 &&
3017 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
3018 cpos += dp->d_reclen;
3019 dp = (struct dirent *)cpos;
3020 cookiep++;
3021 ncookies--;
3022 }
3023 if (cpos >= cend || ncookies == 0) {
3024 toff = off;
3025 siz = fullsiz;
3026 goto again;
3027 }
3028
3029 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
3030 nfsm_reply(cnt);
3031 nfsm_srvpostop_attr(getret, &at);
3032 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3033 txdr_hyper(at.va_filerev, tl);
3034 mp = mp2 = mb;
3035 bp = bpos;
3036 be = bp + M_TRAILINGSPACE(mp);
3037
3038 /* Loop through the records and build reply */
3039 while (cpos < cend && ncookies > 0) {
3040 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
3041 nlen = dp->d_namlen;
3042 rem = nfsm_rndup(nlen)-nlen;
3043
3044 /*
3045 * For readdir_and_lookup get the vnode using
3046 * the file number.
3047 */
3048 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
3049 goto invalid;
3050 memset((caddr_t)nfhp, 0, NFSX_V3FH);
3051 nfhp->fh_fsid =
3052 nvp->v_mount->mnt_stat.f_fsidx;
3053 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
3054 vput(nvp);
3055 goto invalid;
3056 }
3057 if (VOP_GETATTR(nvp, vap, cred, lwp)) {
3058 vput(nvp);
3059 goto invalid;
3060 }
3061 vput(nvp);
3062 KASSERT(fhp->fh_fid.fid_len <= _VFS_MAXFIDSZ);
3063
3064 /*
3065 * If either the dircount or maxcount will be
3066 * exceeded, get out now. Both of these lengths
3067 * are calculated conservatively, including all
3068 * XDR overheads.
3069 */
3070 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
3071 NFSX_V3POSTOPATTR);
3072 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
3073 if (len > cnt || dirlen > fullsiz) {
3074 eofflag = 0;
3075 break;
3076 }
3077
3078 /*
3079 * Build the directory record xdr from
3080 * the dirent entry.
3081 */
3082 fp = (struct nfs_fattr *)&fl.fl_fattr;
3083 nfsm_srvfillattr(vap, fp);
3084 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
3085 fl.fl_fhok = nfs_true;
3086 fl.fl_postopok = nfs_true;
3087 txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
3088
3089 nfsm_clget;
3090 *tl = nfs_true;
3091 bp += NFSX_UNSIGNED;
3092 nfsm_clget;
3093 *tl = txdr_unsigned(dp->d_fileno >> 32);
3094 bp += NFSX_UNSIGNED;
3095 nfsm_clget;
3096 *tl = txdr_unsigned(dp->d_fileno);
3097 bp += NFSX_UNSIGNED;
3098 nfsm_clget;
3099 *tl = txdr_unsigned(nlen);
3100 bp += NFSX_UNSIGNED;
3101
3102 /* And loop around copying the name */
3103 xfer = nlen;
3104 cp = dp->d_name;
3105 while (xfer > 0) {
3106 nfsm_clget;
3107 if ((bp + xfer) > be)
3108 tsiz = be - bp;
3109 else
3110 tsiz = xfer;
3111 memcpy(bp, cp, tsiz);
3112 bp += tsiz;
3113 xfer -= tsiz;
3114 if (xfer > 0)
3115 cp += tsiz;
3116 }
3117 /* And null pad to an int32_t boundary */
3118 for (i = 0; i < rem; i++)
3119 *bp++ = '\0';
3120
3121 /*
3122 * Now copy the flrep structure out.
3123 */
3124 xfer = sizeof (struct flrep);
3125 cp = (caddr_t)&fl;
3126 while (xfer > 0) {
3127 nfsm_clget;
3128 if ((bp + xfer) > be)
3129 tsiz = be - bp;
3130 else
3131 tsiz = xfer;
3132 memcpy(bp, cp, tsiz);
3133 bp += tsiz;
3134 xfer -= tsiz;
3135 if (xfer > 0)
3136 cp += tsiz;
3137 }
3138 }
3139 invalid:
3140 cpos += dp->d_reclen;
3141 dp = (struct dirent *)cpos;
3142 cookiep++;
3143 ncookies--;
3144 }
3145 vrele(vp);
3146 nfsm_clget;
3147 *tl = nfs_false;
3148 bp += NFSX_UNSIGNED;
3149 nfsm_clget;
3150 if (eofflag)
3151 *tl = nfs_true;
3152 else
3153 *tl = nfs_false;
3154 bp += NFSX_UNSIGNED;
3155 if (mp != mb) {
3156 if (bp < be)
3157 mp->m_len = bp - mtod(mp, caddr_t);
3158 } else
3159 mp->m_len += bp - bpos;
3160 free((caddr_t)cookies, M_TEMP);
3161 free((caddr_t)rbuf, M_TEMP);
3162 nfsm_srvdone;
3163 }
3164
3165 /*
3166 * nfs commit service
3167 */
3168 int
3169 nfsrv_commit(nfsd, slp, lwp, mrq)
3170 struct nfsrv_descript *nfsd;
3171 struct nfssvc_sock *slp;
3172 struct lwp *lwp;
3173 struct mbuf **mrq;
3174 {
3175 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3176 struct mbuf *nam = nfsd->nd_nam;
3177 caddr_t dpos = nfsd->nd_dpos;
3178 kauth_cred_t cred = nfsd->nd_cr;
3179 struct vattr bfor, aft;
3180 struct vnode *vp;
3181 nfsfh_t nfh;
3182 fhandle_t *fhp;
3183 u_int32_t *tl;
3184 int32_t t1;
3185 caddr_t bpos;
3186 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0;
3187 uint32_t cnt;
3188 char *cp2;
3189 struct mbuf *mb, *mreq;
3190 u_quad_t frev, off, end;
3191 struct mount *mp = NULL;
3192
3193 fhp = &nfh.fh_generic;
3194 nfsm_srvmtofh(fhp);
3195 if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
3196 return (ESTALE);
3197 vn_start_write(NULL, &mp, V_WAIT);
3198 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3199
3200 off = fxdr_hyper(tl);
3201 tl += 2;
3202 cnt = fxdr_unsigned(uint32_t, *tl);
3203 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3204 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3205 if (error) {
3206 nfsm_reply(2 * NFSX_UNSIGNED);
3207 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3208 vn_finished_write(mp, 0);
3209 return (0);
3210 }
3211 for_ret = VOP_GETATTR(vp, &bfor, cred, lwp);
3212 end = (cnt > 0) ? off + cnt : vp->v_size;
3213 if (end < off || end > vp->v_size)
3214 end = vp->v_size;
3215 if (off < vp->v_size)
3216 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp);
3217 /* else error == 0, from nfsrv_fhtovp() */
3218 aft_ret = VOP_GETATTR(vp, &aft, cred, lwp);
3219 vput(vp);
3220 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3221 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3222 if (!error) {
3223 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3224 *tl++ = txdr_unsigned(boottime.tv_sec);
3225 *tl = txdr_unsigned(boottime.tv_usec);
3226 } else {
3227 vn_finished_write(mp, 0);
3228 return (0);
3229 }
3230 vn_finished_write(mp, 0);
3231 nfsm_srvdone;
3232 }
3233
3234 /*
3235 * nfs statfs service
3236 */
3237 int
3238 nfsrv_statfs(nfsd, slp, lwp, mrq)
3239 struct nfsrv_descript *nfsd;
3240 struct nfssvc_sock *slp;
3241 struct lwp *lwp;
3242 struct mbuf **mrq;
3243 {
3244 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3245 struct mbuf *nam = nfsd->nd_nam;
3246 caddr_t dpos = nfsd->nd_dpos;
3247 kauth_cred_t cred = nfsd->nd_cr;
3248 struct statvfs *sf;
3249 struct nfs_statfs *sfp;
3250 u_int32_t *tl;
3251 int32_t t1;
3252 caddr_t bpos;
3253 int error = 0, rdonly, cache = 0, getret = 1;
3254 int v3 = (nfsd->nd_flag & ND_NFSV3);
3255 char *cp2;
3256 struct mbuf *mb, *mreq;
3257 struct vnode *vp;
3258 struct vattr at;
3259 nfsfh_t nfh;
3260 fhandle_t *fhp;
3261 struct statvfs statvfs;
3262 u_quad_t frev, tval;
3263
3264 fhp = &nfh.fh_generic;
3265 nfsm_srvmtofh(fhp);
3266 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3267 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3268 if (error) {
3269 nfsm_reply(NFSX_UNSIGNED);
3270 nfsm_srvpostop_attr(getret, &at);
3271 return (0);
3272 }
3273 sf = &statvfs;
3274 error = VFS_STATVFS(vp->v_mount, sf, lwp);
3275 getret = VOP_GETATTR(vp, &at, cred, lwp);
3276 vput(vp);
3277 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3278 if (v3)
3279 nfsm_srvpostop_attr(getret, &at);
3280 if (error)
3281 return (0);
3282 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3283 if (v3) {
3284 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize);
3285 txdr_hyper(tval, &sfp->sf_tbytes);
3286 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize);
3287 txdr_hyper(tval, &sfp->sf_fbytes);
3288 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize);
3289 txdr_hyper(tval, &sfp->sf_abytes);
3290 tval = (u_quad_t)sf->f_files;
3291 txdr_hyper(tval, &sfp->sf_tfiles);
3292 tval = (u_quad_t)sf->f_ffree;
3293 txdr_hyper(tval, &sfp->sf_ffiles);
3294 txdr_hyper(tval, &sfp->sf_afiles);
3295 sfp->sf_invarsec = 0;
3296 } else {
3297 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3298 sfp->sf_bsize = txdr_unsigned(sf->f_frsize);
3299 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3300 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3301 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3302 }
3303 nfsm_srvdone;
3304 }
3305
3306 /*
3307 * nfs fsinfo service
3308 */
3309 int
3310 nfsrv_fsinfo(nfsd, slp, lwp, mrq)
3311 struct nfsrv_descript *nfsd;
3312 struct nfssvc_sock *slp;
3313 struct lwp *lwp;
3314 struct mbuf **mrq;
3315 {
3316 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3317 struct mbuf *nam = nfsd->nd_nam;
3318 caddr_t dpos = nfsd->nd_dpos;
3319 kauth_cred_t cred = nfsd->nd_cr;
3320 u_int32_t *tl;
3321 struct nfsv3_fsinfo *sip;
3322 int32_t t1;
3323 caddr_t bpos;
3324 int error = 0, rdonly, cache = 0, getret = 1;
3325 uint32_t maxdata;
3326 char *cp2;
3327 struct mbuf *mb, *mreq;
3328 struct vnode *vp;
3329 struct vattr at;
3330 nfsfh_t nfh;
3331 fhandle_t *fhp;
3332 u_quad_t frev, maxfsize;
3333 struct statvfs sb;
3334
3335 fhp = &nfh.fh_generic;
3336 nfsm_srvmtofh(fhp);
3337 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3338 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3339 if (error) {
3340 nfsm_reply(NFSX_UNSIGNED);
3341 nfsm_srvpostop_attr(getret, &at);
3342 return (0);
3343 }
3344
3345 /* XXX Try to make a guess on the max file size. */
3346 VFS_STATVFS(vp->v_mount, &sb, (struct lwp *)0);
3347 maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1;
3348
3349 getret = VOP_GETATTR(vp, &at, cred, lwp);
3350 vput(vp);
3351 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3352 nfsm_srvpostop_attr(getret, &at);
3353 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3354
3355 /*
3356 * XXX
3357 * There should be file system VFS OP(s) to get this information.
3358 * For now, assume ufs.
3359 */
3360 if (slp->ns_so->so_type == SOCK_DGRAM)
3361 maxdata = NFS_MAXDGRAMDATA;
3362 else
3363 maxdata = NFS_MAXDATA;
3364 sip->fs_rtmax = txdr_unsigned(maxdata);
3365 sip->fs_rtpref = txdr_unsigned(maxdata);
3366 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3367 sip->fs_wtmax = txdr_unsigned(maxdata);
3368 sip->fs_wtpref = txdr_unsigned(maxdata);
3369 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3370 sip->fs_dtpref = txdr_unsigned(maxdata);
3371 txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3372 sip->fs_timedelta.nfsv3_sec = 0;
3373 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3374 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3375 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3376 NFSV3FSINFO_CANSETTIME);
3377 nfsm_srvdone;
3378 }
3379
3380 /*
3381 * nfs pathconf service
3382 */
3383 int
3384 nfsrv_pathconf(nfsd, slp, lwp, mrq)
3385 struct nfsrv_descript *nfsd;
3386 struct nfssvc_sock *slp;
3387 struct lwp *lwp;
3388 struct mbuf **mrq;
3389 {
3390 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3391 struct mbuf *nam = nfsd->nd_nam;
3392 caddr_t dpos = nfsd->nd_dpos;
3393 kauth_cred_t cred = nfsd->nd_cr;
3394 u_int32_t *tl;
3395 struct nfsv3_pathconf *pc;
3396 int32_t t1;
3397 caddr_t bpos;
3398 int error = 0, rdonly, cache = 0, getret = 1;
3399 register_t linkmax, namemax, chownres, notrunc;
3400 char *cp2;
3401 struct mbuf *mb, *mreq;
3402 struct vnode *vp;
3403 struct vattr at;
3404 nfsfh_t nfh;
3405 fhandle_t *fhp;
3406 u_quad_t frev;
3407
3408 fhp = &nfh.fh_generic;
3409 nfsm_srvmtofh(fhp);
3410 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3411 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3412 if (error) {
3413 nfsm_reply(NFSX_UNSIGNED);
3414 nfsm_srvpostop_attr(getret, &at);
3415 return (0);
3416 }
3417 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3418 if (!error)
3419 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3420 if (!error)
3421 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3422 if (!error)
3423 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc);
3424 getret = VOP_GETATTR(vp, &at, cred, lwp);
3425 vput(vp);
3426 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3427 nfsm_srvpostop_attr(getret, &at);
3428 if (error)
3429 return (0);
3430 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3431
3432 pc->pc_linkmax = txdr_unsigned(linkmax);
3433 pc->pc_namemax = txdr_unsigned(namemax);
3434 pc->pc_notrunc = txdr_unsigned(notrunc);
3435 pc->pc_chownrestricted = txdr_unsigned(chownres);
3436
3437 /*
3438 * These should probably be supported by VOP_PATHCONF(), but
3439 * until msdosfs is exportable (why would you want to?), the
3440 * Unix defaults should be ok.
3441 */
3442 pc->pc_caseinsensitive = nfs_false;
3443 pc->pc_casepreserving = nfs_true;
3444 nfsm_srvdone;
3445 }
3446
3447 /*
3448 * Null operation, used by clients to ping server
3449 */
3450 /* ARGSUSED */
3451 int
3452 nfsrv_null(nfsd, slp, lwp, mrq)
3453 struct nfsrv_descript *nfsd;
3454 struct nfssvc_sock *slp;
3455 struct lwp *lwp;
3456 struct mbuf **mrq;
3457 {
3458 struct mbuf *mrep = nfsd->nd_mrep;
3459 caddr_t bpos;
3460 int error = NFSERR_RETVOID, cache = 0;
3461 struct mbuf *mb, *mreq;
3462 u_quad_t frev;
3463
3464 nfsm_reply(0);
3465 return (0);
3466 }
3467
3468 /*
3469 * No operation, used for obsolete procedures
3470 */
3471 /* ARGSUSED */
3472 int
3473 nfsrv_noop(nfsd, slp, lwp, mrq)
3474 struct nfsrv_descript *nfsd;
3475 struct nfssvc_sock *slp;
3476 struct lwp *lwp;
3477 struct mbuf **mrq;
3478 {
3479 struct mbuf *mrep = nfsd->nd_mrep;
3480 caddr_t bpos;
3481 int error, cache = 0;
3482 struct mbuf *mb, *mreq;
3483 u_quad_t frev;
3484
3485 if (nfsd->nd_repstat)
3486 error = nfsd->nd_repstat;
3487 else
3488 error = EPROCUNAVAIL;
3489 nfsm_reply(0);
3490 return (0);
3491 }
3492
3493 /*
3494 * Perform access checking for vnodes obtained from file handles that would
3495 * refer to files already opened by a Unix client. You cannot just use
3496 * vn_writechk() and VOP_ACCESS() for two reasons.
3497 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3498 * 2 - The owner is to be given access irrespective of mode bits for some
3499 * operations, so that processes that chmod after opening a file don't
3500 * break. I don't like this because it opens a security hole, but since
3501 * the nfs server opens a security hole the size of a barn door anyhow,
3502 * what the heck.
3503 *
3504 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3505 * will return EPERM instead of EACCESS. EPERM is always an error.
3506 */
3507 int
3508 nfsrv_access(vp, flags, cred, rdonly, lwp, override)
3509 struct vnode *vp;
3510 int flags;
3511 kauth_cred_t cred;
3512 int rdonly;
3513 struct lwp *lwp;
3514 {
3515 struct vattr vattr;
3516 int error;
3517 if (flags & VWRITE) {
3518 /* Just vn_writechk() changed to check rdonly */
3519 /*
3520 * Disallow write attempts on read-only file systems;
3521 * unless the file is a socket or a block or character
3522 * device resident on the file system.
3523 */
3524 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3525 switch (vp->v_type) {
3526 case VREG:
3527 case VDIR:
3528 case VLNK:
3529 return (EROFS);
3530 default:
3531 break;
3532 }
3533 }
3534
3535 /*
3536 * If the vnode is in use as a process's text,
3537 * we can't allow writing.
3538 */
3539 if (vp->v_flag & VTEXT)
3540 return (ETXTBSY);
3541 }
3542 error = VOP_GETATTR(vp, &vattr, cred, lwp);
3543 if (error)
3544 return (error);
3545 error = VOP_ACCESS(vp, flags, cred, lwp);
3546 /*
3547 * Allow certain operations for the owner (reads and writes
3548 * on files that are already open).
3549 */
3550 if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid)
3551 error = 0;
3552 return error;
3553 }
3554