nfs_serv.c revision 1.70 1 /* $NetBSD: nfs_serv.c,v 1.70 2003/04/02 15:14:20 yamt Exp $ */
2
3 /*
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Rick Macklem at The University of Guelph.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95
39 */
40
41 /*
42 * nfs version 2 and 3 server calls to vnode ops
43 * - these routines generally have 3 phases
44 * 1 - break down and validate rpc request in mbuf list
45 * 2 - do the vnode ops for the request
46 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c)
47 * 3 - build the rpc reply in an mbuf list
48 * nb:
49 * - do not mix the phases, since the nfsm_?? macros can return failures
50 * on a bad rpc or similar and do not do any vrele() or vput()'s
51 *
52 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs
53 * error number iff error != 0 whereas
54 * returning an error from the server function implies a fatal error
55 * such as a badly constructed rpc request that should be dropped without
56 * a reply.
57 * For Version 3, nfsm_reply() does not return for the error case, since
58 * most version 3 rpcs return more than the status for error cases.
59 */
60
61 #include <sys/cdefs.h>
62 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.70 2003/04/02 15:14:20 yamt Exp $");
63
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/proc.h>
67 #include <sys/file.h>
68 #include <sys/namei.h>
69 #include <sys/vnode.h>
70 #include <sys/mount.h>
71 #include <sys/socket.h>
72 #include <sys/socketvar.h>
73 #include <sys/mbuf.h>
74 #include <sys/dirent.h>
75 #include <sys/stat.h>
76 #include <sys/kernel.h>
77 #include <ufs/ufs/dir.h>
78
79 #include <uvm/uvm_extern.h>
80
81 #include <nfs/nfsproto.h>
82 #include <nfs/rpcv2.h>
83 #include <nfs/nfs.h>
84 #include <nfs/xdr_subs.h>
85 #include <nfs/nfsm_subs.h>
86 #include <nfs/nqnfs.h>
87 #include <nfs/nfs_var.h>
88
89 /* Global vars */
90 extern u_int32_t nfs_xdrneg1;
91 extern u_int32_t nfs_false, nfs_true;
92 extern enum vtype nv3tov_type[8];
93 extern struct nfsstats nfsstats;
94 extern const nfstype nfsv2_type[9];
95 extern const nfstype nfsv3_type[9];
96 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
97
98 /*
99 * nfs v3 access service
100 */
101 int
102 nfsrv3_access(nfsd, slp, procp, mrq)
103 struct nfsrv_descript *nfsd;
104 struct nfssvc_sock *slp;
105 struct proc *procp;
106 struct mbuf **mrq;
107 {
108 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
109 struct mbuf *nam = nfsd->nd_nam;
110 caddr_t dpos = nfsd->nd_dpos;
111 struct ucred *cred = &nfsd->nd_cr;
112 struct vnode *vp;
113 nfsfh_t nfh;
114 fhandle_t *fhp;
115 u_int32_t *tl;
116 int32_t t1;
117 caddr_t bpos;
118 int error = 0, rdonly, cache = 0, getret;
119 char *cp2;
120 struct mbuf *mb, *mreq;
121 struct vattr va;
122 u_long inmode, testmode, outmode;
123 u_quad_t frev;
124
125 fhp = &nfh.fh_generic;
126 nfsm_srvmtofh(fhp);
127 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
128 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
129 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
130 if (error) {
131 nfsm_reply(NFSX_UNSIGNED);
132 nfsm_srvpostop_attr(1, (struct vattr *)0);
133 return (0);
134 }
135 inmode = fxdr_unsigned(u_int32_t, *tl);
136 outmode = 0;
137 if ((inmode & NFSV3ACCESS_READ) &&
138 nfsrv_access(vp, VREAD, cred, rdonly, procp, 0) == 0)
139 outmode |= NFSV3ACCESS_READ;
140 if (vp->v_type != VDIR) {
141 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
142 if (testmode &&
143 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0)
144 outmode |= testmode;
145 if ((inmode & NFSV3ACCESS_EXECUTE) &&
146 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
147 outmode |= NFSV3ACCESS_EXECUTE;
148 } else {
149 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
150 NFSV3ACCESS_DELETE);
151 if (testmode &&
152 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0)
153 outmode |= testmode;
154 if ((inmode & NFSV3ACCESS_LOOKUP) &&
155 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
156 outmode |= NFSV3ACCESS_LOOKUP;
157 }
158 getret = VOP_GETATTR(vp, &va, cred, procp);
159 vput(vp);
160 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
161 nfsm_srvpostop_attr(getret, &va);
162 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
163 *tl = txdr_unsigned(outmode);
164 nfsm_srvdone;
165 }
166
167 /*
168 * nfs getattr service
169 */
170 int
171 nfsrv_getattr(nfsd, slp, procp, mrq)
172 struct nfsrv_descript *nfsd;
173 struct nfssvc_sock *slp;
174 struct proc *procp;
175 struct mbuf **mrq;
176 {
177 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
178 struct mbuf *nam = nfsd->nd_nam;
179 caddr_t dpos = nfsd->nd_dpos;
180 struct ucred *cred = &nfsd->nd_cr;
181 struct nfs_fattr *fp;
182 struct vattr va;
183 struct vnode *vp;
184 nfsfh_t nfh;
185 fhandle_t *fhp;
186 u_int32_t *tl;
187 int32_t t1;
188 caddr_t bpos;
189 int error = 0, rdonly, cache;
190 char *cp2;
191 struct mbuf *mb, *mreq;
192 u_quad_t frev;
193
194 fhp = &nfh.fh_generic;
195 nfsm_srvmtofh(fhp);
196 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
197 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
198 if (error) {
199 nfsm_reply(0);
200 return (0);
201 }
202 nqsrv_getl(vp, ND_READ);
203 error = VOP_GETATTR(vp, &va, cred, procp);
204 vput(vp);
205 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
206 if (error)
207 return (0);
208 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
209 nfsm_srvfillattr(&va, fp);
210 nfsm_srvdone;
211 }
212
213 /*
214 * nfs setattr service
215 */
216 int
217 nfsrv_setattr(nfsd, slp, procp, mrq)
218 struct nfsrv_descript *nfsd;
219 struct nfssvc_sock *slp;
220 struct proc *procp;
221 struct mbuf **mrq;
222 {
223 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
224 struct mbuf *nam = nfsd->nd_nam;
225 caddr_t dpos = nfsd->nd_dpos;
226 struct ucred *cred = &nfsd->nd_cr;
227 struct vattr va, preat;
228 struct nfsv2_sattr *sp;
229 struct nfs_fattr *fp;
230 struct vnode *vp;
231 nfsfh_t nfh;
232 fhandle_t *fhp;
233 u_int32_t *tl;
234 int32_t t1;
235 caddr_t bpos;
236 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1;
237 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
238 char *cp2;
239 struct mbuf *mb, *mreq;
240 u_quad_t frev;
241 struct timespec guard;
242
243 fhp = &nfh.fh_generic;
244 nfsm_srvmtofh(fhp);
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 return (0);
294 }
295 nqsrv_getl(vp, ND_WRITE);
296 if (v3) {
297 error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp);
298 if (!error && gcheck &&
299 (preat.va_ctime.tv_sec != guard.tv_sec ||
300 preat.va_ctime.tv_nsec != guard.tv_nsec))
301 error = NFSERR_NOT_SYNC;
302 if (error) {
303 vput(vp);
304 nfsm_reply(NFSX_WCCDATA(v3));
305 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
306 return (0);
307 }
308 }
309
310 /*
311 * If the size is being changed write acces is required, otherwise
312 * just check for a read only file system.
313 */
314 if (va.va_size == ((u_quad_t)((quad_t) -1))) {
315 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
316 error = EROFS;
317 goto out;
318 }
319 } else {
320 if (vp->v_type == VDIR) {
321 error = EISDIR;
322 goto out;
323 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
324 procp, 0)) != 0)
325 goto out;
326 }
327 error = VOP_SETATTR(vp, &va, cred, procp);
328 postat_ret = VOP_GETATTR(vp, &va, cred, procp);
329 if (!error)
330 error = postat_ret;
331 out:
332 vput(vp);
333 nfsm_reply(NFSX_WCCORFATTR(v3));
334 if (v3) {
335 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
336 return (0);
337 } else {
338 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
339 nfsm_srvfillattr(&va, fp);
340 }
341 nfsm_srvdone;
342 }
343
344 /*
345 * nfs lookup rpc
346 */
347 int
348 nfsrv_lookup(nfsd, slp, procp, mrq)
349 struct nfsrv_descript *nfsd;
350 struct nfssvc_sock *slp;
351 struct proc *procp;
352 struct mbuf **mrq;
353 {
354 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
355 struct mbuf *nam = nfsd->nd_nam;
356 caddr_t dpos = nfsd->nd_dpos;
357 struct ucred *cred = &nfsd->nd_cr;
358 struct nfs_fattr *fp;
359 struct nameidata nd, ind, *ndp = &nd;
360 struct vnode *vp, *dirp;
361 nfsfh_t nfh;
362 fhandle_t *fhp;
363 caddr_t cp;
364 u_int32_t *tl;
365 int32_t t1;
366 caddr_t bpos;
367 int error = 0, cache, dirattr_ret = 1;
368 uint32_t len;
369 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
370 char *cp2;
371 struct mbuf *mb, *mreq;
372 struct vattr va, dirattr;
373 u_quad_t frev;
374
375 fhp = &nfh.fh_generic;
376 nfsm_srvmtofh(fhp);
377 nfsm_srvnamesiz(len);
378
379 pubflag = nfs_ispublicfh(fhp);
380
381 nd.ni_cnd.cn_cred = cred;
382 nd.ni_cnd.cn_nameiop = LOOKUP;
383 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
384 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
385 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
386
387 if (!error && pubflag) {
388 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
389 /*
390 * Setup call to lookup() to see if we can find
391 * the index file. Arguably, this doesn't belong
392 * in a kernel.. Ugh.
393 */
394 ind = nd;
395 VOP_UNLOCK(nd.ni_vp, 0);
396 ind.ni_pathlen = strlen(nfs_pub.np_index);
397 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
398 nfs_pub.np_index;
399 ind.ni_startdir = nd.ni_vp;
400 VREF(ind.ni_startdir);
401 error = lookup(&ind);
402 if (!error) {
403 /*
404 * Found an index file. Get rid of
405 * the old references.
406 */
407 if (dirp)
408 vrele(dirp);
409 dirp = nd.ni_vp;
410 vrele(nd.ni_startdir);
411 ndp = &ind;
412 } else
413 error = 0;
414 }
415 /*
416 * If the public filehandle was used, check that this lookup
417 * didn't result in a filehandle outside the publicly exported
418 * filesystem.
419 */
420
421 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
422 vput(nd.ni_vp);
423 error = EPERM;
424 }
425 }
426
427 if (dirp) {
428 if (v3)
429 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred,
430 procp);
431 vrele(dirp);
432 }
433
434 if (error) {
435 nfsm_reply(NFSX_POSTOPATTR(v3));
436 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
437 return (0);
438 }
439
440 nqsrv_getl(ndp->ni_startdir, ND_READ);
441 vrele(ndp->ni_startdir);
442 PNBUF_PUT(nd.ni_cnd.cn_pnbuf);
443 vp = ndp->ni_vp;
444 memset((caddr_t)fhp, 0, sizeof(nfh));
445 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
446 error = VFS_VPTOFH(vp, &fhp->fh_fid);
447 if (!error)
448 error = VOP_GETATTR(vp, &va, cred, procp);
449 vput(vp);
450 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
451 if (error) {
452 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
453 return (0);
454 }
455 nfsm_srvfhtom(fhp, v3);
456 if (v3) {
457 nfsm_srvpostop_attr(0, &va);
458 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
459 } else {
460 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
461 nfsm_srvfillattr(&va, fp);
462 }
463 nfsm_srvdone;
464 }
465
466 /*
467 * nfs readlink service
468 */
469 int
470 nfsrv_readlink(nfsd, slp, procp, mrq)
471 struct nfsrv_descript *nfsd;
472 struct nfssvc_sock *slp;
473 struct proc *procp;
474 struct mbuf **mrq;
475 {
476 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
477 struct mbuf *nam = nfsd->nd_nam;
478 caddr_t dpos = nfsd->nd_dpos;
479 struct ucred *cred = &nfsd->nd_cr;
480 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
481 struct iovec *ivp = iv;
482 struct mbuf *mp;
483 u_int32_t *tl;
484 int32_t t1;
485 caddr_t bpos;
486 int error = 0, rdonly, cache, i, tlen, len, getret;
487 int v3 = (nfsd->nd_flag & ND_NFSV3);
488 char *cp2;
489 struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq;
490 struct vnode *vp;
491 struct vattr attr;
492 nfsfh_t nfh;
493 fhandle_t *fhp;
494 struct uio io, *uiop = &io;
495 u_quad_t frev;
496
497 fhp = &nfh.fh_generic;
498 nfsm_srvmtofh(fhp);
499 len = 0;
500 i = 0;
501 while (len < NFS_MAXPATHLEN) {
502 mp = m_get(M_WAIT, MT_DATA);
503 MCLAIM(mp, &nfs_mowner);
504 m_clget(mp, M_WAIT);
505 mp->m_len = NFSMSIZ(mp);
506 if (len == 0)
507 mp3 = mp2 = mp;
508 else {
509 mp2->m_next = mp;
510 mp2 = mp;
511 }
512 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
513 mp->m_len = NFS_MAXPATHLEN-len;
514 len = NFS_MAXPATHLEN;
515 } else
516 len += mp->m_len;
517 ivp->iov_base = mtod(mp, caddr_t);
518 ivp->iov_len = mp->m_len;
519 i++;
520 ivp++;
521 }
522 uiop->uio_iov = iv;
523 uiop->uio_iovcnt = i;
524 uiop->uio_offset = 0;
525 uiop->uio_resid = len;
526 uiop->uio_rw = UIO_READ;
527 uiop->uio_segflg = UIO_SYSSPACE;
528 uiop->uio_procp = (struct proc *)0;
529 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
530 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
531 if (error) {
532 m_freem(mp3);
533 nfsm_reply(2 * NFSX_UNSIGNED);
534 nfsm_srvpostop_attr(1, (struct vattr *)0);
535 return (0);
536 }
537 if (vp->v_type != VLNK) {
538 if (v3)
539 error = EINVAL;
540 else
541 error = ENXIO;
542 goto out;
543 }
544 nqsrv_getl(vp, ND_READ);
545 error = VOP_READLINK(vp, uiop, cred);
546 out:
547 getret = VOP_GETATTR(vp, &attr, cred, procp);
548 vput(vp);
549 if (error)
550 m_freem(mp3);
551 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
552 if (v3) {
553 nfsm_srvpostop_attr(getret, &attr);
554 if (error)
555 return (0);
556 }
557 if (uiop->uio_resid > 0) {
558 len -= uiop->uio_resid;
559 tlen = nfsm_rndup(len);
560 nfsm_adj(mp3, NFS_MAXPATHLEN-tlen, tlen-len);
561 }
562 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
563 *tl = txdr_unsigned(len);
564 mb->m_next = mp3;
565 nfsm_srvdone;
566 }
567
568 /*
569 * nfs read service
570 */
571 int
572 nfsrv_read(nfsd, slp, procp, mrq)
573 struct nfsrv_descript *nfsd;
574 struct nfssvc_sock *slp;
575 struct proc *procp;
576 struct mbuf **mrq;
577 {
578 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
579 struct mbuf *nam = nfsd->nd_nam;
580 caddr_t dpos = nfsd->nd_dpos;
581 struct ucred *cred = &nfsd->nd_cr;
582 struct iovec *iv;
583 struct iovec *iv2;
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, cnt, len, left, siz, tlen, getret;
591 int v3 = (nfsd->nd_flag & ND_NFSV3);
592 uint32_t reqlen;
593 char *cp2;
594 struct mbuf *mb, *mreq;
595 struct mbuf *m2;
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_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd));
614 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
615 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
616 if (error) {
617 nfsm_reply(2 * NFSX_UNSIGNED);
618 nfsm_srvpostop_attr(1, (struct vattr *)0);
619 return (0);
620 }
621 if (vp->v_type != VREG) {
622 if (v3)
623 error = EINVAL;
624 else
625 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
626 }
627 if (!error) {
628 nqsrv_getl(vp, ND_READ);
629 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0)
630 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1);
631 }
632 getret = VOP_GETATTR(vp, &va, cred, procp);
633 if (!error)
634 error = getret;
635 if (error) {
636 vput(vp);
637 nfsm_reply(NFSX_POSTOPATTR(v3));
638 nfsm_srvpostop_attr(getret, &va);
639 return (0);
640 }
641 if (off >= va.va_size)
642 cnt = 0;
643 else if ((off + reqlen) > va.va_size)
644 cnt = nfsm_rndup(va.va_size - off);
645 else
646 cnt = reqlen;
647 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
648 if (v3) {
649 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
650 *tl++ = nfs_true;
651 fp = (struct nfs_fattr *)tl;
652 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
653 } else {
654 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
655 fp = (struct nfs_fattr *)tl;
656 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
657 }
658 len = left = cnt;
659 if (cnt > 0) {
660 /*
661 * Generate the mbuf list with the uio_iov ref. to it.
662 */
663 i = 0;
664 m = m2 = mb;
665 while (left > 0) {
666 siz = min(M_TRAILINGSPACE(m), left);
667 if (siz > 0) {
668 left -= siz;
669 i++;
670 }
671 if (left > 0) {
672 m = m_get(M_WAIT, MT_DATA);
673 MCLAIM(m, &nfs_mowner);
674 m_clget(m, M_WAIT);
675 m->m_len = 0;
676 m2->m_next = m;
677 m2 = m;
678 }
679 }
680 iv = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
681 uiop->uio_iov = iv2 = iv;
682 m = mb;
683 left = cnt;
684 i = 0;
685 while (left > 0) {
686 if (m == NULL)
687 panic("nfsrv_read iov");
688 siz = min(M_TRAILINGSPACE(m), left);
689 if (siz > 0) {
690 iv->iov_base = mtod(m, caddr_t) + m->m_len;
691 iv->iov_len = siz;
692 m->m_len += siz;
693 left -= siz;
694 iv++;
695 i++;
696 }
697 m = m->m_next;
698 }
699 uiop->uio_iovcnt = i;
700 uiop->uio_offset = off;
701 uiop->uio_resid = cnt;
702 uiop->uio_rw = UIO_READ;
703 uiop->uio_segflg = UIO_SYSSPACE;
704 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
705 off = uiop->uio_offset;
706 free((caddr_t)iv2, M_TEMP);
707 if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){
708 if (!error)
709 error = getret;
710 m_freem(mreq);
711 vput(vp);
712 nfsm_reply(NFSX_POSTOPATTR(v3));
713 nfsm_srvpostop_attr(getret, &va);
714 return (0);
715 }
716 } else
717 uiop->uio_resid = 0;
718 vput(vp);
719 nfsm_srvfillattr(&va, fp);
720 len -= uiop->uio_resid;
721 tlen = nfsm_rndup(len);
722 if (cnt != tlen || tlen != len)
723 nfsm_adj(mb, cnt - tlen, tlen - len);
724 if (v3) {
725 *tl++ = txdr_unsigned(len);
726 if (len < reqlen)
727 *tl++ = nfs_true;
728 else
729 *tl++ = nfs_false;
730 }
731 *tl = txdr_unsigned(len);
732 nfsm_srvdone;
733 }
734
735 /*
736 * nfs write service
737 */
738 int
739 nfsrv_write(nfsd, slp, procp, mrq)
740 struct nfsrv_descript *nfsd;
741 struct nfssvc_sock *slp;
742 struct proc *procp;
743 struct mbuf **mrq;
744 {
745 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
746 struct mbuf *nam = nfsd->nd_nam;
747 caddr_t dpos = nfsd->nd_dpos;
748 struct ucred *cred = &nfsd->nd_cr;
749 struct iovec *ivp;
750 int i, cnt;
751 struct mbuf *mp;
752 struct nfs_fattr *fp;
753 struct iovec *iv;
754 struct vattr va, forat;
755 u_int32_t *tl;
756 int32_t t1;
757 caddr_t bpos;
758 int error = 0, rdonly, cache, len, forat_ret = 1;
759 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
760 int stable = NFSV3WRITE_FILESYNC;
761 int v3 = (nfsd->nd_flag & ND_NFSV3);
762 char *cp2;
763 struct mbuf *mb, *mreq;
764 struct vnode *vp;
765 nfsfh_t nfh;
766 fhandle_t *fhp;
767 struct uio io, *uiop = &io;
768 off_t off;
769 u_quad_t frev;
770
771 if (mrep == NULL) {
772 *mrq = NULL;
773 return (0);
774 }
775 fhp = &nfh.fh_generic;
776 nfsm_srvmtofh(fhp);
777 if (v3) {
778 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
779 off = fxdr_hyper(tl);
780 tl += 3;
781 stable = fxdr_unsigned(int, *tl++);
782 } else {
783 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
784 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
785 tl += 2;
786 }
787 retlen = len = fxdr_unsigned(int32_t, *tl);
788 cnt = i = 0;
789
790 /*
791 * For NFS Version 2, it is not obvious what a write of zero length
792 * should do, but I might as well be consistent with Version 3,
793 * which is to return ok so long as there are no permission problems.
794 */
795 if (len > 0) {
796 zeroing = 1;
797 mp = mrep;
798 while (mp) {
799 if (mp == md) {
800 zeroing = 0;
801 adjust = dpos - mtod(mp, caddr_t);
802 mp->m_len -= adjust;
803 if (mp->m_len > 0 && adjust > 0)
804 NFSMADV(mp, adjust);
805 }
806 if (zeroing)
807 mp->m_len = 0;
808 else if (mp->m_len > 0) {
809 i += mp->m_len;
810 if (i > len) {
811 mp->m_len -= (i - len);
812 zeroing = 1;
813 }
814 if (mp->m_len > 0)
815 cnt++;
816 }
817 mp = mp->m_next;
818 }
819 }
820 if (len > NFS_MAXDATA || len < 0 || i < len) {
821 error = EIO;
822 nfsm_reply(2 * NFSX_UNSIGNED);
823 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
824 return (0);
825 }
826 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
827 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
828 if (error) {
829 nfsm_reply(2 * NFSX_UNSIGNED);
830 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
831 return (0);
832 }
833 if (v3)
834 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
835 if (vp->v_type != VREG) {
836 if (v3)
837 error = EINVAL;
838 else
839 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
840 }
841 if (!error) {
842 nqsrv_getl(vp, ND_WRITE);
843 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
844 }
845 if (error) {
846 vput(vp);
847 nfsm_reply(NFSX_WCCDATA(v3));
848 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
849 return (0);
850 }
851
852 if (len > 0) {
853 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK);
854 uiop->uio_iov = iv = ivp;
855 uiop->uio_iovcnt = cnt;
856 mp = mrep;
857 while (mp) {
858 if (mp->m_len > 0) {
859 ivp->iov_base = mtod(mp, caddr_t);
860 ivp->iov_len = mp->m_len;
861 ivp++;
862 }
863 mp = mp->m_next;
864 }
865
866 /*
867 * XXX
868 * The IO_METASYNC flag indicates that all metadata (and not just
869 * enough to ensure data integrity) mus be written to stable storage
870 * synchronously.
871 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
872 */
873 if (stable == NFSV3WRITE_UNSTABLE)
874 ioflags = IO_NODELOCKED;
875 else if (stable == NFSV3WRITE_DATASYNC)
876 ioflags = (IO_SYNC | IO_NODELOCKED);
877 else
878 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
879 uiop->uio_resid = len;
880 uiop->uio_rw = UIO_WRITE;
881 uiop->uio_segflg = UIO_SYSSPACE;
882 uiop->uio_procp = (struct proc *)0;
883 uiop->uio_offset = off;
884 error = VOP_WRITE(vp, uiop, ioflags, cred);
885 nfsstats.srvvop_writes++;
886 free((caddr_t)iv, M_TEMP);
887 }
888 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
889 vput(vp);
890 if (!error)
891 error = aftat_ret;
892 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
893 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
894 if (v3) {
895 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
896 if (error)
897 return (0);
898 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
899 *tl++ = txdr_unsigned(retlen);
900 if (stable == NFSV3WRITE_UNSTABLE)
901 *tl++ = txdr_unsigned(stable);
902 else
903 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
904 /*
905 * Actually, there is no need to txdr these fields,
906 * but it may make the values more human readable,
907 * for debugging purposes.
908 */
909 *tl++ = txdr_unsigned(boottime.tv_sec);
910 *tl = txdr_unsigned(boottime.tv_usec);
911 } else {
912 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
913 nfsm_srvfillattr(&va, fp);
914 }
915 nfsm_srvdone;
916 }
917
918 /*
919 * NFS write service with write gathering support. Called when
920 * nfsrvw_procrastinate > 0.
921 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
922 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
923 * Jan. 1994.
924 */
925 int
926 nfsrv_writegather(ndp, slp, procp, mrq)
927 struct nfsrv_descript **ndp;
928 struct nfssvc_sock *slp;
929 struct proc *procp;
930 struct mbuf **mrq;
931 {
932 struct iovec *ivp;
933 struct mbuf *mp;
934 struct nfsrv_descript *wp, *nfsd, *owp, *swp;
935 struct nfs_fattr *fp;
936 int i = 0;
937 struct iovec *iov;
938 struct nfsrvw_delayhash *wpp;
939 struct ucred *cred;
940 struct vattr va, forat;
941 u_int32_t *tl;
942 int32_t t1;
943 caddr_t bpos, dpos;
944 int error = 0, rdonly, cache, len = 0, forat_ret = 1;
945 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
946 char *cp2;
947 struct mbuf *mb, *mreq, *mrep, *md;
948 struct vnode *vp;
949 struct uio io, *uiop = &io;
950 u_quad_t frev, cur_usec;
951
952 *mrq = NULL;
953 if (*ndp) {
954 nfsd = *ndp;
955 *ndp = NULL;
956 mrep = nfsd->nd_mrep;
957 md = nfsd->nd_md;
958 dpos = nfsd->nd_dpos;
959 cred = &nfsd->nd_cr;
960 v3 = (nfsd->nd_flag & ND_NFSV3);
961 LIST_INIT(&nfsd->nd_coalesce);
962 nfsd->nd_mreq = NULL;
963 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
964 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
965 nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
966
967 /*
968 * Now, get the write header..
969 */
970 nfsm_srvmtofh(&nfsd->nd_fh);
971 if (v3) {
972 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
973 nfsd->nd_off = fxdr_hyper(tl);
974 tl += 3;
975 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
976 } else {
977 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
978 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
979 tl += 2;
980 }
981 len = fxdr_unsigned(int32_t, *tl);
982 nfsd->nd_len = len;
983 nfsd->nd_eoff = nfsd->nd_off + len;
984
985 /*
986 * Trim the header out of the mbuf list and trim off any trailing
987 * junk so that the mbuf list has only the write data.
988 */
989 zeroing = 1;
990 i = 0;
991 mp = mrep;
992 while (mp) {
993 if (mp == md) {
994 zeroing = 0;
995 adjust = dpos - mtod(mp, caddr_t);
996 mp->m_len -= adjust;
997 if (mp->m_len > 0 && adjust > 0)
998 NFSMADV(mp, adjust);
999 }
1000 if (zeroing)
1001 mp->m_len = 0;
1002 else {
1003 i += mp->m_len;
1004 if (i > len) {
1005 mp->m_len -= (i - len);
1006 zeroing = 1;
1007 }
1008 }
1009 mp = mp->m_next;
1010 }
1011 if (len > NFS_MAXDATA || len < 0 || i < len) {
1012 nfsmout:
1013 m_freem(mrep);
1014 error = EIO;
1015 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1016 if (v3)
1017 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1018 nfsd->nd_mreq = mreq;
1019 nfsd->nd_mrep = NULL;
1020 nfsd->nd_time = 0;
1021 }
1022
1023 /*
1024 * Add this entry to the hash and time queues.
1025 */
1026 s = splsoftclock();
1027 owp = NULL;
1028 wp = LIST_FIRST(&slp->ns_tq);
1029 while (wp && wp->nd_time < nfsd->nd_time) {
1030 owp = wp;
1031 wp = LIST_NEXT(wp, nd_tq);
1032 }
1033 if (owp) {
1034 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1035 } else {
1036 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1037 }
1038 if (nfsd->nd_mrep) {
1039 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1040 owp = NULL;
1041 wp = LIST_FIRST(wpp);
1042 while (wp &&
1043 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1044 owp = wp;
1045 wp = LIST_NEXT(wp, nd_hash);
1046 }
1047 while (wp && wp->nd_off < nfsd->nd_off &&
1048 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1049 owp = wp;
1050 wp = LIST_NEXT(wp, nd_hash);
1051 }
1052 if (owp) {
1053 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1054
1055 /*
1056 * Search the hash list for overlapping entries and
1057 * coalesce.
1058 */
1059 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1060 wp = LIST_NEXT(nfsd, nd_hash);
1061 if (NFSW_SAMECRED(owp, nfsd))
1062 nfsrvw_coalesce(owp, nfsd);
1063 }
1064 } else {
1065 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1066 }
1067 }
1068 splx(s);
1069 }
1070
1071 /*
1072 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1073 * and generate the associated reply mbuf list(s).
1074 */
1075 loop1:
1076 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1077 s = splsoftclock();
1078 for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) {
1079 owp = LIST_NEXT(nfsd, nd_tq);
1080 if (nfsd->nd_time > cur_usec)
1081 break;
1082 if (nfsd->nd_mreq)
1083 continue;
1084 LIST_REMOVE(nfsd, nd_tq);
1085 LIST_REMOVE(nfsd, nd_hash);
1086 splx(s);
1087 mrep = nfsd->nd_mrep;
1088 nfsd->nd_mrep = NULL;
1089 cred = &nfsd->nd_cr;
1090 v3 = (nfsd->nd_flag & ND_NFSV3);
1091 forat_ret = aftat_ret = 1;
1092 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1093 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1094 FALSE);
1095 if (!error) {
1096 if (v3)
1097 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
1098 if (vp->v_type != VREG) {
1099 if (v3)
1100 error = EINVAL;
1101 else
1102 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1103 }
1104 } else
1105 vp = NULL;
1106 if (!error) {
1107 nqsrv_getl(vp, ND_WRITE);
1108 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
1109 }
1110
1111 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1112 ioflags = IO_NODELOCKED;
1113 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1114 ioflags = (IO_SYNC | IO_NODELOCKED);
1115 else
1116 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1117 uiop->uio_rw = UIO_WRITE;
1118 uiop->uio_segflg = UIO_SYSSPACE;
1119 uiop->uio_procp = (struct proc *)0;
1120 uiop->uio_offset = nfsd->nd_off;
1121 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1122 if (uiop->uio_resid > 0) {
1123 mp = mrep;
1124 i = 0;
1125 while (mp) {
1126 if (mp->m_len > 0)
1127 i++;
1128 mp = mp->m_next;
1129 }
1130 uiop->uio_iovcnt = i;
1131 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
1132 uiop->uio_iov = ivp = iov;
1133 mp = mrep;
1134 while (mp) {
1135 if (mp->m_len > 0) {
1136 ivp->iov_base = mtod(mp, caddr_t);
1137 ivp->iov_len = mp->m_len;
1138 ivp++;
1139 }
1140 mp = mp->m_next;
1141 }
1142 if (!error) {
1143 error = VOP_WRITE(vp, uiop, ioflags, cred);
1144 nfsstats.srvvop_writes++;
1145 }
1146 free((caddr_t)iov, M_TEMP);
1147 }
1148 m_freem(mrep);
1149 if (vp) {
1150 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
1151 vput(vp);
1152 }
1153
1154 /*
1155 * Loop around generating replies for all write rpcs that have
1156 * now been completed.
1157 */
1158 swp = nfsd;
1159 do {
1160 if (error) {
1161 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1162 if (v3) {
1163 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1164 }
1165 } else {
1166 nfsm_writereply(NFSX_PREOPATTR(v3) +
1167 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1168 NFSX_WRITEVERF(v3), v3);
1169 if (v3) {
1170 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1171 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1172 *tl++ = txdr_unsigned(nfsd->nd_len);
1173 *tl++ = txdr_unsigned(swp->nd_stable);
1174 /*
1175 * Actually, there is no need to txdr these fields,
1176 * but it may make the values more human readable,
1177 * for debugging purposes.
1178 */
1179 *tl++ = txdr_unsigned(boottime.tv_sec);
1180 *tl = txdr_unsigned(boottime.tv_usec);
1181 } else {
1182 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1183 nfsm_srvfillattr(&va, fp);
1184 }
1185 }
1186 nfsd->nd_mreq = mreq;
1187 if (nfsd->nd_mrep)
1188 panic("nfsrv_write: nd_mrep not free");
1189
1190 /*
1191 * Done. Put it at the head of the timer queue so that
1192 * the final phase can return the reply.
1193 */
1194 s = splsoftclock();
1195 if (nfsd != swp) {
1196 nfsd->nd_time = 0;
1197 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1198 }
1199 nfsd = LIST_FIRST(&swp->nd_coalesce);
1200 if (nfsd) {
1201 LIST_REMOVE(nfsd, nd_tq);
1202 }
1203 splx(s);
1204 } while (nfsd);
1205 s = splsoftclock();
1206 swp->nd_time = 0;
1207 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1208 splx(s);
1209 goto loop1;
1210 }
1211 splx(s);
1212
1213 /*
1214 * Search for a reply to return.
1215 */
1216 s = splsoftclock();
1217 LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) {
1218 if (nfsd->nd_mreq) {
1219 LIST_REMOVE(nfsd, nd_tq);
1220 *mrq = nfsd->nd_mreq;
1221 *ndp = nfsd;
1222 break;
1223 }
1224 }
1225 splx(s);
1226 return (0);
1227 }
1228
1229 /*
1230 * Coalesce the write request nfsd into owp. To do this we must:
1231 * - remove nfsd from the queues
1232 * - merge nfsd->nd_mrep into owp->nd_mrep
1233 * - update the nd_eoff and nd_stable for owp
1234 * - put nfsd on owp's nd_coalesce list
1235 * NB: Must be called at splsoftclock().
1236 */
1237 void
1238 nfsrvw_coalesce(owp, nfsd)
1239 struct nfsrv_descript *owp;
1240 struct nfsrv_descript *nfsd;
1241 {
1242 int overlap;
1243 struct mbuf *mp;
1244 struct nfsrv_descript *m;
1245
1246 LIST_REMOVE(nfsd, nd_hash);
1247 LIST_REMOVE(nfsd, nd_tq);
1248 if (owp->nd_eoff < nfsd->nd_eoff) {
1249 overlap = owp->nd_eoff - nfsd->nd_off;
1250 if (overlap < 0)
1251 panic("nfsrv_coalesce: bad off");
1252 if (overlap > 0)
1253 m_adj(nfsd->nd_mrep, overlap);
1254 mp = owp->nd_mrep;
1255 while (mp->m_next)
1256 mp = mp->m_next;
1257 mp->m_next = nfsd->nd_mrep;
1258 owp->nd_eoff = nfsd->nd_eoff;
1259 } else
1260 m_freem(nfsd->nd_mrep);
1261 nfsd->nd_mrep = NULL;
1262 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1263 owp->nd_stable = NFSV3WRITE_FILESYNC;
1264 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1265 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1266 owp->nd_stable = NFSV3WRITE_DATASYNC;
1267 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1268 /*
1269 * nfsd might hold coalesce elements! Move them to owp.
1270 * Otherwise, requests may be lost and clients will be stuck.
1271 */
1272 while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) {
1273 LIST_REMOVE(m, nd_tq);
1274 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1275 }
1276 }
1277
1278 /*
1279 * nfs create service
1280 * now does a truncate to 0 length via. setattr if it already exists
1281 */
1282 int
1283 nfsrv_create(nfsd, slp, procp, mrq)
1284 struct nfsrv_descript *nfsd;
1285 struct nfssvc_sock *slp;
1286 struct proc *procp;
1287 struct mbuf **mrq;
1288 {
1289 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1290 struct mbuf *nam = nfsd->nd_nam;
1291 caddr_t dpos = nfsd->nd_dpos;
1292 struct ucred *cred = &nfsd->nd_cr;
1293 struct nfs_fattr *fp;
1294 struct vattr va, dirfor, diraft;
1295 struct nfsv2_sattr *sp;
1296 u_int32_t *tl;
1297 struct nameidata nd;
1298 caddr_t cp;
1299 int32_t t1;
1300 caddr_t bpos;
1301 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1302 int rdev = 0;
1303 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1304 char *cp2;
1305 struct mbuf *mb, *mreq;
1306 struct vnode *vp = NULL, *dirp = NULL;
1307 nfsfh_t nfh;
1308 fhandle_t *fhp;
1309 u_quad_t frev, tempsize;
1310 u_char cverf[NFSX_V3CREATEVERF];
1311
1312 nd.ni_cnd.cn_nameiop = 0;
1313 fhp = &nfh.fh_generic;
1314 nfsm_srvmtofh(fhp);
1315 nfsm_srvnamesiz(len);
1316 nd.ni_cnd.cn_cred = cred;
1317 nd.ni_cnd.cn_nameiop = CREATE;
1318 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1319 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1320 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1321 if (dirp) {
1322 if (v3)
1323 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1324 procp);
1325 else {
1326 vrele(dirp);
1327 dirp = (struct vnode *)0;
1328 }
1329 }
1330 if (error) {
1331 nfsm_reply(NFSX_WCCDATA(v3));
1332 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1333 if (dirp)
1334 vrele(dirp);
1335 return (0);
1336 }
1337 VATTR_NULL(&va);
1338 if (v3) {
1339 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1340 how = fxdr_unsigned(int, *tl);
1341 switch (how) {
1342 case NFSV3CREATE_GUARDED:
1343 if (nd.ni_vp) {
1344 error = EEXIST;
1345 break;
1346 }
1347 case NFSV3CREATE_UNCHECKED:
1348 nfsm_srvsattr(&va);
1349 break;
1350 case NFSV3CREATE_EXCLUSIVE:
1351 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
1352 memcpy(cverf, cp, NFSX_V3CREATEVERF);
1353 exclusive_flag = 1;
1354 if (nd.ni_vp == NULL)
1355 va.va_mode = 0;
1356 break;
1357 };
1358 va.va_type = VREG;
1359 } else {
1360 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1361 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1362 if (va.va_type == VNON)
1363 va.va_type = VREG;
1364 va.va_mode = nfstov_mode(sp->sa_mode);
1365 switch (va.va_type) {
1366 case VREG:
1367 tsize = fxdr_unsigned(int32_t, sp->sa_size);
1368 if (tsize != -1)
1369 va.va_size = (u_quad_t)tsize;
1370 break;
1371 case VCHR:
1372 case VBLK:
1373 case VFIFO:
1374 rdev = fxdr_unsigned(int32_t, sp->sa_size);
1375 break;
1376 default:
1377 break;
1378 };
1379 }
1380
1381 /*
1382 * Iff doesn't exist, create it
1383 * otherwise just truncate to 0 length
1384 * should I set the mode too ??
1385 */
1386 if (nd.ni_vp == NULL) {
1387 if (va.va_type == VREG || va.va_type == VSOCK) {
1388 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1389 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1390 if (!error) {
1391 if (exclusive_flag) {
1392 exclusive_flag = 0;
1393 VATTR_NULL(&va);
1394 memcpy((caddr_t)&va.va_atime, cverf,
1395 NFSX_V3CREATEVERF);
1396 error = VOP_SETATTR(nd.ni_vp, &va, cred,
1397 procp);
1398 }
1399 }
1400 } else if (va.va_type == VCHR || va.va_type == VBLK ||
1401 va.va_type == VFIFO) {
1402 if (va.va_type == VCHR && rdev == 0xffffffff)
1403 va.va_type = VFIFO;
1404 if (va.va_type != VFIFO &&
1405 (error = suser(cred, (u_short *)0))) {
1406 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1407 vput(nd.ni_dvp);
1408 nfsm_reply(0);
1409 return (error);
1410 } else
1411 va.va_rdev = (dev_t)rdev;
1412 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1413 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1414 &va);
1415 if (error) {
1416 nfsm_reply(0);
1417 }
1418 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1419 vrele(nd.ni_dvp);
1420 vput(nd.ni_vp);
1421 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1422 error = EINVAL;
1423 nfsm_reply(0);
1424 }
1425 } else {
1426 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1427 vput(nd.ni_dvp);
1428 error = ENXIO;
1429 }
1430 vp = nd.ni_vp;
1431 } else {
1432 vp = nd.ni_vp;
1433 if (nd.ni_dvp == vp)
1434 vrele(nd.ni_dvp);
1435 else
1436 vput(nd.ni_dvp);
1437 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1438 if (!error && va.va_size != -1) {
1439 error = nfsrv_access(vp, VWRITE, cred,
1440 (nd.ni_cnd.cn_flags & RDONLY), procp, 0);
1441 if (!error) {
1442 nqsrv_getl(vp, ND_WRITE);
1443 tempsize = va.va_size;
1444 VATTR_NULL(&va);
1445 va.va_size = tempsize;
1446 error = VOP_SETATTR(vp, &va, cred,
1447 procp);
1448 }
1449 }
1450 if (error)
1451 vput(vp);
1452 }
1453 if (!error) {
1454 memset((caddr_t)fhp, 0, sizeof(nfh));
1455 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1456 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1457 if (!error)
1458 error = VOP_GETATTR(vp, &va, cred, procp);
1459 vput(vp);
1460 }
1461 if (v3) {
1462 if (exclusive_flag && !error &&
1463 memcmp(cverf, (caddr_t)&va.va_atime, NFSX_V3CREATEVERF))
1464 error = EEXIST;
1465 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1466 vrele(dirp);
1467 }
1468 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1469 if (v3) {
1470 if (!error) {
1471 nfsm_srvpostop_fh(fhp);
1472 nfsm_srvpostop_attr(0, &va);
1473 }
1474 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1475 } else {
1476 nfsm_srvfhtom(fhp, v3);
1477 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1478 nfsm_srvfillattr(&va, fp);
1479 }
1480 return (0);
1481 nfsmout:
1482 if (dirp)
1483 vrele(dirp);
1484 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1485 if (nd.ni_dvp == nd.ni_vp)
1486 vrele(nd.ni_dvp);
1487 else
1488 vput(nd.ni_dvp);
1489 if (nd.ni_vp)
1490 vput(nd.ni_vp);
1491 return (error);
1492 }
1493
1494 /*
1495 * nfs v3 mknod service
1496 */
1497 int
1498 nfsrv_mknod(nfsd, slp, procp, mrq)
1499 struct nfsrv_descript *nfsd;
1500 struct nfssvc_sock *slp;
1501 struct proc *procp;
1502 struct mbuf **mrq;
1503 {
1504 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1505 struct mbuf *nam = nfsd->nd_nam;
1506 caddr_t dpos = nfsd->nd_dpos;
1507 struct ucred *cred = &nfsd->nd_cr;
1508 struct vattr va, dirfor, diraft;
1509 u_int32_t *tl;
1510 struct nameidata nd;
1511 int32_t t1;
1512 caddr_t bpos;
1513 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1514 u_int32_t major, minor;
1515 enum vtype vtyp;
1516 char *cp2;
1517 struct mbuf *mb, *mreq;
1518 struct vnode *vp, *dirp = (struct vnode *)0;
1519 nfsfh_t nfh;
1520 fhandle_t *fhp;
1521 u_quad_t frev;
1522
1523 nd.ni_cnd.cn_nameiop = 0;
1524 fhp = &nfh.fh_generic;
1525 nfsm_srvmtofh(fhp);
1526 nfsm_srvnamesiz(len);
1527 nd.ni_cnd.cn_cred = cred;
1528 nd.ni_cnd.cn_nameiop = CREATE;
1529 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1530 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1531 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1532 if (dirp)
1533 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp);
1534 if (error) {
1535 nfsm_reply(NFSX_WCCDATA(1));
1536 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1537 if (dirp)
1538 vrele(dirp);
1539 return (0);
1540 }
1541 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1542 vtyp = nfsv3tov_type(*tl);
1543 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1544 error = NFSERR_BADTYPE;
1545 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1546 if (nd.ni_dvp == nd.ni_vp)
1547 vrele(nd.ni_dvp);
1548 else
1549 vput(nd.ni_dvp);
1550 if (nd.ni_vp)
1551 vput(nd.ni_vp);
1552 goto out;
1553 }
1554 VATTR_NULL(&va);
1555 nfsm_srvsattr(&va);
1556 if (vtyp == VCHR || vtyp == VBLK) {
1557 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1558 major = fxdr_unsigned(u_int32_t, *tl++);
1559 minor = fxdr_unsigned(u_int32_t, *tl);
1560 va.va_rdev = makedev(major, minor);
1561 }
1562
1563 /*
1564 * Iff doesn't exist, create it.
1565 */
1566 if (nd.ni_vp) {
1567 error = EEXIST;
1568 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1569 if (nd.ni_dvp == nd.ni_vp)
1570 vrele(nd.ni_dvp);
1571 else
1572 vput(nd.ni_dvp);
1573 vput(nd.ni_vp);
1574 goto out;
1575 }
1576 va.va_type = vtyp;
1577 if (vtyp == VSOCK) {
1578 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1579 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1580 } else {
1581 if (va.va_type != VFIFO &&
1582 (error = suser(cred, (u_short *)0))) {
1583 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1584 vput(nd.ni_dvp);
1585 goto out;
1586 }
1587 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1588 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1589 if (error) {
1590 goto out;
1591 }
1592 if (error)
1593 goto out;
1594 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1595 vput(nd.ni_vp);
1596 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1597 error = EINVAL;
1598 }
1599 }
1600 out:
1601 vp = nd.ni_vp;
1602 if (!error) {
1603 memset((caddr_t)fhp, 0, sizeof(nfh));
1604 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1605 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1606 if (!error)
1607 error = VOP_GETATTR(vp, &va, cred, procp);
1608 vput(vp);
1609 }
1610 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1611 vrele(dirp);
1612 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
1613 if (!error) {
1614 nfsm_srvpostop_fh(fhp);
1615 nfsm_srvpostop_attr(0, &va);
1616 }
1617 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1618 return (0);
1619 nfsmout:
1620 if (dirp)
1621 vrele(dirp);
1622 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1623 if (nd.ni_dvp == nd.ni_vp)
1624 vrele(nd.ni_dvp);
1625 else
1626 vput(nd.ni_dvp);
1627 if (nd.ni_vp)
1628 vput(nd.ni_vp);
1629 return (error);
1630 }
1631
1632 /*
1633 * nfs remove service
1634 */
1635 int
1636 nfsrv_remove(nfsd, slp, procp, mrq)
1637 struct nfsrv_descript *nfsd;
1638 struct nfssvc_sock *slp;
1639 struct proc *procp;
1640 struct mbuf **mrq;
1641 {
1642 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1643 struct mbuf *nam = nfsd->nd_nam;
1644 caddr_t dpos = nfsd->nd_dpos;
1645 struct ucred *cred = &nfsd->nd_cr;
1646 struct nameidata nd;
1647 u_int32_t *tl;
1648 int32_t t1;
1649 caddr_t bpos;
1650 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1651 int v3 = (nfsd->nd_flag & ND_NFSV3);
1652 char *cp2;
1653 struct mbuf *mb, *mreq;
1654 struct vnode *vp, *dirp;
1655 struct vattr dirfor, diraft;
1656 nfsfh_t nfh;
1657 fhandle_t *fhp;
1658 u_quad_t frev;
1659
1660 #ifndef nolint
1661 vp = (struct vnode *)0;
1662 #endif
1663 fhp = &nfh.fh_generic;
1664 nfsm_srvmtofh(fhp);
1665 nfsm_srvnamesiz(len);
1666 nd.ni_cnd.cn_cred = cred;
1667 nd.ni_cnd.cn_nameiop = DELETE;
1668 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1669 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1670 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1671 if (dirp) {
1672 if (v3)
1673 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1674 procp);
1675 else
1676 vrele(dirp);
1677 }
1678 if (!error) {
1679 vp = nd.ni_vp;
1680 if (vp->v_type == VDIR &&
1681 (error = suser(cred, (u_short *)0)) != 0)
1682 goto out;
1683 /*
1684 * The root of a mounted filesystem cannot be deleted.
1685 */
1686 if (vp->v_flag & VROOT) {
1687 error = EBUSY;
1688 goto out;
1689 }
1690 out:
1691 if (!error) {
1692 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1693 nqsrv_getl(vp, ND_WRITE);
1694 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1695 } else {
1696 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1697 if (nd.ni_dvp == vp)
1698 vrele(nd.ni_dvp);
1699 else
1700 vput(nd.ni_dvp);
1701 vput(vp);
1702 }
1703 }
1704 if (dirp && v3) {
1705 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1706 vrele(dirp);
1707 }
1708 nfsm_reply(NFSX_WCCDATA(v3));
1709 if (v3) {
1710 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1711 return (0);
1712 }
1713 nfsm_srvdone;
1714 }
1715
1716 /*
1717 * nfs rename service
1718 */
1719 int
1720 nfsrv_rename(nfsd, slp, procp, mrq)
1721 struct nfsrv_descript *nfsd;
1722 struct nfssvc_sock *slp;
1723 struct proc *procp;
1724 struct mbuf **mrq;
1725 {
1726 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1727 struct mbuf *nam = nfsd->nd_nam;
1728 caddr_t dpos = nfsd->nd_dpos;
1729 struct ucred *cred = &nfsd->nd_cr;
1730 u_int32_t *tl;
1731 int32_t t1;
1732 caddr_t bpos;
1733 int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1;
1734 uint32_t len, len2;
1735 int tdirfor_ret = 1, tdiraft_ret = 1;
1736 int v3 = (nfsd->nd_flag & ND_NFSV3);
1737 char *cp2;
1738 struct mbuf *mb, *mreq;
1739 struct nameidata fromnd, tond;
1740 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
1741 struct vnode *tdirp = (struct vnode *)0;
1742 struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1743 nfsfh_t fnfh, tnfh;
1744 fhandle_t *ffhp, *tfhp;
1745 u_quad_t frev;
1746 uid_t saved_uid;
1747
1748 #ifndef nolint
1749 fvp = (struct vnode *)0;
1750 #endif
1751 ffhp = &fnfh.fh_generic;
1752 tfhp = &tnfh.fh_generic;
1753 fromnd.ni_cnd.cn_nameiop = 0;
1754 tond.ni_cnd.cn_nameiop = 0;
1755 nfsm_srvmtofh(ffhp);
1756 nfsm_srvnamesiz(len);
1757 /*
1758 * Remember our original uid so that we can reset cr_uid before
1759 * the second nfs_namei() call, in case it is remapped.
1760 */
1761 saved_uid = cred->cr_uid;
1762 fromnd.ni_cnd.cn_cred = cred;
1763 fromnd.ni_cnd.cn_nameiop = DELETE;
1764 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART;
1765 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
1766 &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1767 if (fdirp) {
1768 if (v3)
1769 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred,
1770 procp);
1771 else {
1772 vrele(fdirp);
1773 fdirp = (struct vnode *)0;
1774 }
1775 }
1776 if (error) {
1777 nfsm_reply(2 * NFSX_WCCDATA(v3));
1778 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1779 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1780 if (fdirp)
1781 vrele(fdirp);
1782 return (0);
1783 }
1784 fvp = fromnd.ni_vp;
1785 nfsm_srvmtofh(tfhp);
1786 if (v3) {
1787 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
1788 len2 = fxdr_unsigned(uint32_t, *tl);
1789 /* len2 will be checked by nfs_namei */
1790 }
1791 else {
1792 /* NFSv2 */
1793 nfsm_strsiz(len2, NFS_MAXNAMLEN);
1794 }
1795 cred->cr_uid = saved_uid;
1796 tond.ni_cnd.cn_cred = cred;
1797 tond.ni_cnd.cn_nameiop = RENAME;
1798 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
1799 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
1800 &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1801 if (tdirp) {
1802 if (v3)
1803 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred,
1804 procp);
1805 else {
1806 vrele(tdirp);
1807 tdirp = (struct vnode *)0;
1808 }
1809 }
1810 if (error) {
1811 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1812 vrele(fromnd.ni_dvp);
1813 vrele(fvp);
1814 goto out1;
1815 }
1816 tdvp = tond.ni_dvp;
1817 tvp = tond.ni_vp;
1818 if (tvp != NULL) {
1819 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1820 if (v3)
1821 error = EEXIST;
1822 else
1823 error = EISDIR;
1824 goto out;
1825 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1826 if (v3)
1827 error = EEXIST;
1828 else
1829 error = ENOTDIR;
1830 goto out;
1831 }
1832 if (tvp->v_type == VDIR && tvp->v_mountedhere) {
1833 if (v3)
1834 error = EXDEV;
1835 else
1836 error = ENOTEMPTY;
1837 goto out;
1838 }
1839 }
1840 if (fvp->v_type == VDIR && fvp->v_mountedhere) {
1841 if (v3)
1842 error = EXDEV;
1843 else
1844 error = ENOTEMPTY;
1845 goto out;
1846 }
1847 if (fvp->v_mount != tdvp->v_mount) {
1848 if (v3)
1849 error = EXDEV;
1850 else
1851 error = ENOTEMPTY;
1852 goto out;
1853 }
1854 if (fvp == tdvp) {
1855 if (v3)
1856 error = EINVAL;
1857 else
1858 error = ENOTEMPTY;
1859 }
1860 /*
1861 * If source is the same as the destination (that is the
1862 * same vnode with the same name in the same directory),
1863 * then there is nothing to do.
1864 */
1865 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
1866 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1867 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
1868 fromnd.ni_cnd.cn_namelen))
1869 error = -1;
1870 out:
1871 if (!error) {
1872 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
1873 nqsrv_getl(tdvp, ND_WRITE);
1874 if (tvp) {
1875 nqsrv_getl(tvp, ND_WRITE);
1876 }
1877 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
1878 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
1879 } else {
1880 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
1881 if (tdvp == tvp)
1882 vrele(tdvp);
1883 else
1884 vput(tdvp);
1885 if (tvp)
1886 vput(tvp);
1887 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1888 vrele(fromnd.ni_dvp);
1889 vrele(fvp);
1890 if (error == -1)
1891 error = 0;
1892 }
1893 vrele(tond.ni_startdir);
1894 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
1895 out1:
1896 if (fdirp) {
1897 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp);
1898 vrele(fdirp);
1899 }
1900 if (tdirp) {
1901 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp);
1902 vrele(tdirp);
1903 }
1904 vrele(fromnd.ni_startdir);
1905 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
1906 nfsm_reply(2 * NFSX_WCCDATA(v3));
1907 if (v3) {
1908 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1909 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1910 }
1911 return (0);
1912
1913 nfsmout:
1914 if (fdirp)
1915 vrele(fdirp);
1916 if (tdirp)
1917 vrele(tdirp);
1918 if (tond.ni_cnd.cn_nameiop) {
1919 vrele(tond.ni_startdir);
1920 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
1921 }
1922 if (fromnd.ni_cnd.cn_nameiop) {
1923 vrele(fromnd.ni_startdir);
1924 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
1925 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1926 vrele(fromnd.ni_dvp);
1927 vrele(fvp);
1928 }
1929 return (error);
1930 }
1931
1932 /*
1933 * nfs link service
1934 */
1935 int
1936 nfsrv_link(nfsd, slp, procp, mrq)
1937 struct nfsrv_descript *nfsd;
1938 struct nfssvc_sock *slp;
1939 struct proc *procp;
1940 struct mbuf **mrq;
1941 {
1942 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1943 struct mbuf *nam = nfsd->nd_nam;
1944 caddr_t dpos = nfsd->nd_dpos;
1945 struct ucred *cred = &nfsd->nd_cr;
1946 struct nameidata nd;
1947 u_int32_t *tl;
1948 int32_t t1;
1949 caddr_t bpos;
1950 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1;
1951 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
1952 char *cp2;
1953 struct mbuf *mb, *mreq;
1954 struct vnode *vp, *xp, *dirp = (struct vnode *)0;
1955 struct vattr dirfor, diraft, at;
1956 nfsfh_t nfh, dnfh;
1957 fhandle_t *fhp, *dfhp;
1958 u_quad_t frev;
1959
1960 fhp = &nfh.fh_generic;
1961 dfhp = &dnfh.fh_generic;
1962 nfsm_srvmtofh(fhp);
1963 nfsm_srvmtofh(dfhp);
1964 nfsm_srvnamesiz(len);
1965 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
1966 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1967 if (error) {
1968 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
1969 nfsm_srvpostop_attr(getret, &at);
1970 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1971 return (0);
1972 }
1973 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0)
1974 goto out1;
1975 nd.ni_cnd.cn_cred = cred;
1976 nd.ni_cnd.cn_nameiop = CREATE;
1977 nd.ni_cnd.cn_flags = LOCKPARENT;
1978 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
1979 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1980 if (dirp) {
1981 if (v3)
1982 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1983 procp);
1984 else {
1985 vrele(dirp);
1986 dirp = (struct vnode *)0;
1987 }
1988 }
1989 if (error)
1990 goto out1;
1991 xp = nd.ni_vp;
1992 if (xp != NULL) {
1993 error = EEXIST;
1994 goto out;
1995 }
1996 xp = nd.ni_dvp;
1997 if (vp->v_mount != xp->v_mount)
1998 error = EXDEV;
1999 out:
2000 if (!error) {
2001 nqsrv_getl(vp, ND_WRITE);
2002 nqsrv_getl(xp, ND_WRITE);
2003 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2004 } else {
2005 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2006 if (nd.ni_dvp == nd.ni_vp)
2007 vrele(nd.ni_dvp);
2008 else
2009 vput(nd.ni_dvp);
2010 if (nd.ni_vp)
2011 vrele(nd.ni_vp);
2012 }
2013 out1:
2014 if (v3)
2015 getret = VOP_GETATTR(vp, &at, cred, procp);
2016 if (dirp) {
2017 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2018 vrele(dirp);
2019 }
2020 vrele(vp);
2021 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2022 if (v3) {
2023 nfsm_srvpostop_attr(getret, &at);
2024 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2025 return (0);
2026 }
2027 nfsm_srvdone;
2028 }
2029
2030 /*
2031 * nfs symbolic link service
2032 */
2033 int
2034 nfsrv_symlink(nfsd, slp, procp, mrq)
2035 struct nfsrv_descript *nfsd;
2036 struct nfssvc_sock *slp;
2037 struct proc *procp;
2038 struct mbuf **mrq;
2039 {
2040 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2041 struct mbuf *nam = nfsd->nd_nam;
2042 caddr_t dpos = nfsd->nd_dpos;
2043 struct ucred *cred = &nfsd->nd_cr;
2044 struct vattr va, dirfor, diraft;
2045 struct nameidata nd;
2046 u_int32_t *tl;
2047 int32_t t1;
2048 struct nfsv2_sattr *sp;
2049 char *bpos, *pathcp = NULL, *cp2;
2050 struct uio io;
2051 struct iovec iv;
2052 int error = 0, cache, dirfor_ret = 1, diraft_ret = 1;
2053 uint32_t len, len2;
2054 int v3 = (nfsd->nd_flag & ND_NFSV3);
2055 struct mbuf *mb, *mreq;
2056 struct vnode *dirp = (struct vnode *)0;
2057 nfsfh_t nfh;
2058 fhandle_t *fhp;
2059 u_quad_t frev;
2060
2061 nd.ni_cnd.cn_nameiop = 0;
2062 fhp = &nfh.fh_generic;
2063 nfsm_srvmtofh(fhp);
2064 nfsm_srvnamesiz(len);
2065 nd.ni_cnd.cn_cred = cred;
2066 nd.ni_cnd.cn_nameiop = CREATE;
2067 nd.ni_cnd.cn_flags = LOCKPARENT;
2068 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2069 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2070 if (dirp) {
2071 if (v3)
2072 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2073 procp);
2074 else {
2075 vrele(dirp);
2076 dirp = (struct vnode *)0;
2077 }
2078 }
2079 if (error)
2080 goto out;
2081 VATTR_NULL(&va);
2082 if (v3) {
2083 nfsm_srvsattr(&va);
2084 nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
2085 len2 = fxdr_unsigned(uint32_t, *tl);
2086 if (len2 > PATH_MAX) {
2087 /* XXX should check _PC_NO_TRUNC */
2088 error = ENAMETOOLONG;
2089 goto abortop;
2090 }
2091 }
2092 else {
2093 /* NFSv2 */
2094 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2095 }
2096 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK);
2097 iv.iov_base = pathcp;
2098 iv.iov_len = len2;
2099 io.uio_resid = len2;
2100 io.uio_offset = 0;
2101 io.uio_iov = &iv;
2102 io.uio_iovcnt = 1;
2103 io.uio_segflg = UIO_SYSSPACE;
2104 io.uio_rw = UIO_READ;
2105 io.uio_procp = (struct proc *)0;
2106 nfsm_mtouio(&io, len2);
2107 if (!v3) {
2108 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2109 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2110 }
2111 *(pathcp + len2) = '\0';
2112 if (nd.ni_vp) {
2113 error = EEXIST;
2114 abortop:
2115 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2116 if (nd.ni_dvp == nd.ni_vp)
2117 vrele(nd.ni_dvp);
2118 else
2119 vput(nd.ni_dvp);
2120 if (nd.ni_vp)
2121 vrele(nd.ni_vp);
2122 goto out;
2123 }
2124 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2125 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2126 if (!error) {
2127 if (v3) {
2128 memset((caddr_t)fhp, 0, sizeof(nfh));
2129 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
2130 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2131 if (!error)
2132 error = VOP_GETATTR(nd.ni_vp, &va, cred,
2133 procp);
2134 vput(nd.ni_vp);
2135 } else {
2136 vput(nd.ni_vp);
2137 }
2138 }
2139 out:
2140 if (pathcp)
2141 free(pathcp, M_TEMP);
2142 if (dirp) {
2143 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2144 vrele(dirp);
2145 }
2146 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2147 if (v3) {
2148 if (!error) {
2149 nfsm_srvpostop_fh(fhp);
2150 nfsm_srvpostop_attr(0, &va);
2151 }
2152 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2153 }
2154 return (0);
2155 nfsmout:
2156 if (dirp)
2157 vrele(dirp);
2158 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2159 if (nd.ni_dvp == nd.ni_vp)
2160 vrele(nd.ni_dvp);
2161 else
2162 vput(nd.ni_dvp);
2163 if (nd.ni_vp)
2164 vrele(nd.ni_vp);
2165 if (pathcp)
2166 free(pathcp, M_TEMP);
2167 return (error);
2168 }
2169
2170 /*
2171 * nfs mkdir service
2172 */
2173 int
2174 nfsrv_mkdir(nfsd, slp, procp, mrq)
2175 struct nfsrv_descript *nfsd;
2176 struct nfssvc_sock *slp;
2177 struct proc *procp;
2178 struct mbuf **mrq;
2179 {
2180 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2181 struct mbuf *nam = nfsd->nd_nam;
2182 caddr_t dpos = nfsd->nd_dpos;
2183 struct ucred *cred = &nfsd->nd_cr;
2184 struct vattr va, dirfor, diraft;
2185 struct nfs_fattr *fp;
2186 struct nameidata nd;
2187 caddr_t cp;
2188 u_int32_t *tl;
2189 int32_t t1;
2190 caddr_t bpos;
2191 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2192 int v3 = (nfsd->nd_flag & ND_NFSV3);
2193 char *cp2;
2194 struct mbuf *mb, *mreq;
2195 struct vnode *vp, *dirp = (struct vnode *)0;
2196 nfsfh_t nfh;
2197 fhandle_t *fhp;
2198 u_quad_t frev;
2199
2200 fhp = &nfh.fh_generic;
2201 nfsm_srvmtofh(fhp);
2202 nfsm_srvnamesiz(len);
2203 nd.ni_cnd.cn_cred = cred;
2204 nd.ni_cnd.cn_nameiop = CREATE;
2205 nd.ni_cnd.cn_flags = LOCKPARENT;
2206 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2207 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2208 if (dirp) {
2209 if (v3)
2210 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2211 procp);
2212 else {
2213 vrele(dirp);
2214 dirp = (struct vnode *)0;
2215 }
2216 }
2217 if (error) {
2218 nfsm_reply(NFSX_WCCDATA(v3));
2219 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2220 if (dirp)
2221 vrele(dirp);
2222 return (0);
2223 }
2224 VATTR_NULL(&va);
2225 if (v3) {
2226 nfsm_srvsattr(&va);
2227 } else {
2228 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2229 va.va_mode = nfstov_mode(*tl++);
2230 }
2231 va.va_type = VDIR;
2232 vp = nd.ni_vp;
2233 if (vp != NULL) {
2234 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2235 if (nd.ni_dvp == vp)
2236 vrele(nd.ni_dvp);
2237 else
2238 vput(nd.ni_dvp);
2239 vrele(vp);
2240 error = EEXIST;
2241 goto out;
2242 }
2243 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2244 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2245 if (!error) {
2246 vp = nd.ni_vp;
2247 memset((caddr_t)fhp, 0, sizeof(nfh));
2248 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
2249 error = VFS_VPTOFH(vp, &fhp->fh_fid);
2250 if (!error)
2251 error = VOP_GETATTR(vp, &va, cred, procp);
2252 vput(vp);
2253 }
2254 out:
2255 if (dirp) {
2256 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2257 vrele(dirp);
2258 }
2259 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2260 if (v3) {
2261 if (!error) {
2262 nfsm_srvpostop_fh(fhp);
2263 nfsm_srvpostop_attr(0, &va);
2264 }
2265 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2266 } else {
2267 nfsm_srvfhtom(fhp, v3);
2268 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2269 nfsm_srvfillattr(&va, fp);
2270 }
2271 return (0);
2272 nfsmout:
2273 if (dirp)
2274 vrele(dirp);
2275 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2276 if (nd.ni_dvp == nd.ni_vp)
2277 vrele(nd.ni_dvp);
2278 else
2279 vput(nd.ni_dvp);
2280 if (nd.ni_vp)
2281 vrele(nd.ni_vp);
2282 return (error);
2283 }
2284
2285 /*
2286 * nfs rmdir service
2287 */
2288 int
2289 nfsrv_rmdir(nfsd, slp, procp, mrq)
2290 struct nfsrv_descript *nfsd;
2291 struct nfssvc_sock *slp;
2292 struct proc *procp;
2293 struct mbuf **mrq;
2294 {
2295 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2296 struct mbuf *nam = nfsd->nd_nam;
2297 caddr_t dpos = nfsd->nd_dpos;
2298 struct ucred *cred = &nfsd->nd_cr;
2299 u_int32_t *tl;
2300 int32_t t1;
2301 caddr_t bpos;
2302 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2303 int v3 = (nfsd->nd_flag & ND_NFSV3);
2304 char *cp2;
2305 struct mbuf *mb, *mreq;
2306 struct vnode *vp, *dirp = (struct vnode *)0;
2307 struct vattr dirfor, diraft;
2308 nfsfh_t nfh;
2309 fhandle_t *fhp;
2310 struct nameidata nd;
2311 u_quad_t frev;
2312
2313 fhp = &nfh.fh_generic;
2314 nfsm_srvmtofh(fhp);
2315 nfsm_srvnamesiz(len);
2316 nd.ni_cnd.cn_cred = cred;
2317 nd.ni_cnd.cn_nameiop = DELETE;
2318 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2319 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2320 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2321 if (dirp) {
2322 if (v3)
2323 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2324 procp);
2325 else {
2326 vrele(dirp);
2327 dirp = (struct vnode *)0;
2328 }
2329 }
2330 if (error) {
2331 nfsm_reply(NFSX_WCCDATA(v3));
2332 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2333 if (dirp)
2334 vrele(dirp);
2335 return (0);
2336 }
2337 vp = nd.ni_vp;
2338 if (vp->v_type != VDIR) {
2339 error = ENOTDIR;
2340 goto out;
2341 }
2342 /*
2343 * No rmdir "." please.
2344 */
2345 if (nd.ni_dvp == vp) {
2346 error = EINVAL;
2347 goto out;
2348 }
2349 /*
2350 * The root of a mounted filesystem cannot be deleted.
2351 */
2352 if (vp->v_flag & VROOT)
2353 error = EBUSY;
2354 out:
2355 if (!error) {
2356 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2357 nqsrv_getl(vp, ND_WRITE);
2358 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2359 } else {
2360 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2361 if (nd.ni_dvp == nd.ni_vp)
2362 vrele(nd.ni_dvp);
2363 else
2364 vput(nd.ni_dvp);
2365 vput(vp);
2366 }
2367 if (dirp) {
2368 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2369 vrele(dirp);
2370 }
2371 nfsm_reply(NFSX_WCCDATA(v3));
2372 if (v3) {
2373 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2374 return (0);
2375 }
2376 nfsm_srvdone;
2377 }
2378
2379 /*
2380 * nfs readdir service
2381 * - mallocs what it thinks is enough to read
2382 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR
2383 * - calls VOP_READDIR()
2384 * - loops around building the reply
2385 * if the output generated exceeds count break out of loop
2386 * The nfsm_clget macro is used here so that the reply will be packed
2387 * tightly in mbuf clusters.
2388 * - it only knows that it has encountered eof when the VOP_READDIR()
2389 * reads nothing
2390 * - as such one readdir rpc will return eof false although you are there
2391 * and then the next will return eof
2392 * - it trims out records with d_fileno == 0
2393 * this doesn't matter for Unix clients, but they might confuse clients
2394 * for other os'.
2395 * - it trims out records with d_type == DT_WHT
2396 * these cannot be seen through NFS (unless we extend the protocol)
2397 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2398 * than requested, but this may not apply to all filesystems. For
2399 * example, client NFS does not { although it is never remote mounted
2400 * anyhow }
2401 * The alternate call nfsrv_readdirplus() does lookups as well.
2402 * PS: The NFS protocol spec. does not clarify what the "count" byte
2403 * argument is a count of.. just name strings and file id's or the
2404 * entire reply rpc or ...
2405 * I tried just file name and id sizes and it confused the Sun client,
2406 * so I am using the full rpc size now. The "paranoia.." comment refers
2407 * to including the status longwords that are not a part of the dir.
2408 * "entry" structures, but are in the rpc.
2409 */
2410 struct flrep {
2411 nfsuint64 fl_off;
2412 u_int32_t fl_postopok;
2413 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2414 u_int32_t fl_fhok;
2415 u_int32_t fl_fhsize;
2416 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2417 };
2418
2419 int
2420 nfsrv_readdir(nfsd, slp, procp, mrq)
2421 struct nfsrv_descript *nfsd;
2422 struct nfssvc_sock *slp;
2423 struct proc *procp;
2424 struct mbuf **mrq;
2425 {
2426 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2427 struct mbuf *nam = nfsd->nd_nam;
2428 caddr_t dpos = nfsd->nd_dpos;
2429 struct ucred *cred = &nfsd->nd_cr;
2430 char *bp, *be;
2431 struct mbuf *mp;
2432 struct dirent *dp;
2433 caddr_t cp;
2434 u_int32_t *tl;
2435 int32_t t1;
2436 caddr_t bpos;
2437 struct mbuf *mb, *mreq, *mp2;
2438 char *cpos, *cend, *cp2, *rbuf;
2439 struct vnode *vp;
2440 struct vattr at;
2441 nfsfh_t nfh;
2442 fhandle_t *fhp;
2443 struct uio io;
2444 struct iovec iv;
2445 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2446 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies;
2447 int v3 = (nfsd->nd_flag & ND_NFSV3);
2448 u_quad_t frev, off, toff, verf;
2449 off_t *cookies = NULL, *cookiep;
2450 nfsuint64 jar;
2451
2452 fhp = &nfh.fh_generic;
2453 nfsm_srvmtofh(fhp);
2454 if (v3) {
2455 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2456 toff = fxdr_hyper(tl);
2457 tl += 2;
2458 verf = fxdr_hyper(tl);
2459 tl += 2;
2460 } else {
2461 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2462 toff = fxdr_unsigned(u_quad_t, *tl++);
2463 }
2464 off = toff;
2465 cnt = fxdr_unsigned(int, *tl);
2466 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2467 xfer = NFS_SRVMAXDATA(nfsd);
2468 if (siz > xfer)
2469 siz = xfer;
2470 fullsiz = siz;
2471 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2472 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2473 if (!error && vp->v_type != VDIR) {
2474 error = ENOTDIR;
2475 vput(vp);
2476 }
2477 if (error) {
2478 nfsm_reply(NFSX_UNSIGNED);
2479 nfsm_srvpostop_attr(getret, &at);
2480 return (0);
2481 }
2482 nqsrv_getl(vp, ND_READ);
2483 if (v3) {
2484 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2485 #ifdef NFS3_STRICTVERF
2486 /*
2487 * XXX This check is too strict for Solaris 2.5 clients.
2488 */
2489 if (!error && toff && verf != at.va_filerev)
2490 error = NFSERR_BAD_COOKIE;
2491 #endif
2492 }
2493 if (!error)
2494 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2495 if (error) {
2496 vput(vp);
2497 nfsm_reply(NFSX_POSTOPATTR(v3));
2498 nfsm_srvpostop_attr(getret, &at);
2499 return (0);
2500 }
2501 VOP_UNLOCK(vp, 0);
2502 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2503 again:
2504 iv.iov_base = rbuf;
2505 iv.iov_len = fullsiz;
2506 io.uio_iov = &iv;
2507 io.uio_iovcnt = 1;
2508 io.uio_offset = (off_t)off;
2509 io.uio_resid = fullsiz;
2510 io.uio_segflg = UIO_SYSSPACE;
2511 io.uio_rw = UIO_READ;
2512 io.uio_procp = (struct proc *)0;
2513 eofflag = 0;
2514 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2515
2516 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2517
2518 off = (off_t)io.uio_offset;
2519 if (!cookies && !error)
2520 error = NFSERR_PERM;
2521 if (v3) {
2522 getret = VOP_GETATTR(vp, &at, cred, procp);
2523 if (!error)
2524 error = getret;
2525 }
2526
2527 VOP_UNLOCK(vp, 0);
2528 if (error) {
2529 vrele(vp);
2530 free((caddr_t)rbuf, M_TEMP);
2531 if (cookies)
2532 free((caddr_t)cookies, M_TEMP);
2533 nfsm_reply(NFSX_POSTOPATTR(v3));
2534 nfsm_srvpostop_attr(getret, &at);
2535 return (0);
2536 }
2537 if (io.uio_resid) {
2538 siz -= io.uio_resid;
2539
2540 /*
2541 * If nothing read, return eof
2542 * rpc reply
2543 */
2544 if (siz == 0) {
2545 vrele(vp);
2546 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2547 2 * NFSX_UNSIGNED);
2548 if (v3) {
2549 nfsm_srvpostop_attr(getret, &at);
2550 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2551 txdr_hyper(at.va_filerev, tl);
2552 tl += 2;
2553 } else
2554 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2555 *tl++ = nfs_false;
2556 *tl = nfs_true;
2557 free((caddr_t)rbuf, M_TEMP);
2558 free((caddr_t)cookies, M_TEMP);
2559 return (0);
2560 }
2561 }
2562
2563 /*
2564 * Check for degenerate cases of nothing useful read.
2565 * If so go try again
2566 */
2567 cpos = rbuf;
2568 cend = rbuf + siz;
2569 dp = (struct dirent *)cpos;
2570 cookiep = cookies;
2571
2572 while (cpos < cend && ncookies > 0 &&
2573 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2574 cpos += dp->d_reclen;
2575 dp = (struct dirent *)cpos;
2576 cookiep++;
2577 ncookies--;
2578 }
2579 if (cpos >= cend || ncookies == 0) {
2580 toff = off;
2581 siz = fullsiz;
2582 goto again;
2583 }
2584
2585 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
2586 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2587 if (v3) {
2588 nfsm_srvpostop_attr(getret, &at);
2589 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2590 txdr_hyper(at.va_filerev, tl);
2591 }
2592 mp = mp2 = mb;
2593 bp = bpos;
2594 be = bp + M_TRAILINGSPACE(mp);
2595
2596 /* Loop through the records and build reply */
2597 while (cpos < cend && ncookies > 0) {
2598 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2599 nlen = dp->d_namlen;
2600 rem = nfsm_rndup(nlen)-nlen;
2601 len += (4 * NFSX_UNSIGNED + nlen + rem);
2602 if (v3)
2603 len += 2 * NFSX_UNSIGNED;
2604 if (len > cnt) {
2605 eofflag = 0;
2606 break;
2607 }
2608 /*
2609 * Build the directory record xdr from
2610 * the dirent entry.
2611 */
2612 nfsm_clget;
2613 *tl = nfs_true;
2614 bp += NFSX_UNSIGNED;
2615 if (v3) {
2616 nfsm_clget;
2617 *tl = 0;
2618 bp += NFSX_UNSIGNED;
2619 }
2620 nfsm_clget;
2621 *tl = txdr_unsigned(dp->d_fileno);
2622 bp += NFSX_UNSIGNED;
2623 nfsm_clget;
2624 *tl = txdr_unsigned(nlen);
2625 bp += NFSX_UNSIGNED;
2626
2627 /* And loop around copying the name */
2628 xfer = nlen;
2629 cp = dp->d_name;
2630 while (xfer > 0) {
2631 nfsm_clget;
2632 if ((bp+xfer) > be)
2633 tsiz = be-bp;
2634 else
2635 tsiz = xfer;
2636 memcpy(bp, cp, tsiz);
2637 bp += tsiz;
2638 xfer -= tsiz;
2639 if (xfer > 0)
2640 cp += tsiz;
2641 }
2642 /* And null pad to an int32_t boundary */
2643 for (i = 0; i < rem; i++)
2644 *bp++ = '\0';
2645 nfsm_clget;
2646
2647 /* Finish off the record */
2648 txdr_hyper(*cookiep, &jar);
2649 if (v3) {
2650 *tl = jar.nfsuquad[0];
2651 bp += NFSX_UNSIGNED;
2652 nfsm_clget;
2653 }
2654 *tl = jar.nfsuquad[1];
2655 bp += NFSX_UNSIGNED;
2656 }
2657 cpos += dp->d_reclen;
2658 dp = (struct dirent *)cpos;
2659 cookiep++;
2660 ncookies--;
2661 }
2662 vrele(vp);
2663 nfsm_clget;
2664 *tl = nfs_false;
2665 bp += NFSX_UNSIGNED;
2666 nfsm_clget;
2667 if (eofflag)
2668 *tl = nfs_true;
2669 else
2670 *tl = nfs_false;
2671 bp += NFSX_UNSIGNED;
2672 if (mp != mb) {
2673 if (bp < be)
2674 mp->m_len = bp - mtod(mp, caddr_t);
2675 } else
2676 mp->m_len += bp - bpos;
2677 free((caddr_t)rbuf, M_TEMP);
2678 free((caddr_t)cookies, M_TEMP);
2679 nfsm_srvdone;
2680 }
2681
2682 int
2683 nfsrv_readdirplus(nfsd, slp, procp, mrq)
2684 struct nfsrv_descript *nfsd;
2685 struct nfssvc_sock *slp;
2686 struct proc *procp;
2687 struct mbuf **mrq;
2688 {
2689 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2690 struct mbuf *nam = nfsd->nd_nam;
2691 caddr_t dpos = nfsd->nd_dpos;
2692 struct ucred *cred = &nfsd->nd_cr;
2693 char *bp, *be;
2694 struct mbuf *mp;
2695 struct dirent *dp;
2696 caddr_t cp;
2697 u_int32_t *tl;
2698 int32_t t1;
2699 caddr_t bpos;
2700 struct mbuf *mb, *mreq, *mp2;
2701 char *cpos, *cend, *cp2, *rbuf;
2702 struct vnode *vp, *nvp;
2703 struct flrep fl;
2704 nfsfh_t nfh;
2705 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
2706 struct uio io;
2707 struct iovec iv;
2708 struct vattr va, at, *vap = &va;
2709 struct nfs_fattr *fp;
2710 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2711 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies;
2712 u_quad_t frev, off, toff, verf;
2713 off_t *cookies = NULL, *cookiep;
2714
2715 fhp = &nfh.fh_generic;
2716 nfsm_srvmtofh(fhp);
2717 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2718 toff = fxdr_hyper(tl);
2719 tl += 2;
2720 verf = fxdr_hyper(tl);
2721 tl += 2;
2722 siz = fxdr_unsigned(int, *tl++);
2723 cnt = fxdr_unsigned(int, *tl);
2724 off = toff;
2725 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2726 xfer = NFS_SRVMAXDATA(nfsd);
2727 if (siz > xfer)
2728 siz = xfer;
2729 fullsiz = siz;
2730 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2731 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2732 if (!error && vp->v_type != VDIR) {
2733 error = ENOTDIR;
2734 vput(vp);
2735 }
2736 if (error) {
2737 nfsm_reply(NFSX_UNSIGNED);
2738 nfsm_srvpostop_attr(getret, &at);
2739 return (0);
2740 }
2741 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2742 #ifdef NFS3_STRICTVERF
2743 /*
2744 * XXX This check is too strict for Solaris 2.5 clients.
2745 */
2746 if (!error && toff && verf != at.va_filerev)
2747 error = NFSERR_BAD_COOKIE;
2748 #endif
2749 if (!error) {
2750 nqsrv_getl(vp, ND_READ);
2751 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2752 }
2753 if (error) {
2754 vput(vp);
2755 nfsm_reply(NFSX_V3POSTOPATTR);
2756 nfsm_srvpostop_attr(getret, &at);
2757 return (0);
2758 }
2759 VOP_UNLOCK(vp, 0);
2760
2761 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2762 again:
2763 iv.iov_base = rbuf;
2764 iv.iov_len = fullsiz;
2765 io.uio_iov = &iv;
2766 io.uio_iovcnt = 1;
2767 io.uio_offset = (off_t)off;
2768 io.uio_resid = fullsiz;
2769 io.uio_segflg = UIO_SYSSPACE;
2770 io.uio_rw = UIO_READ;
2771 io.uio_procp = (struct proc *)0;
2772 eofflag = 0;
2773
2774 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2775
2776 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2777
2778 off = (u_quad_t)io.uio_offset;
2779 getret = VOP_GETATTR(vp, &at, cred, procp);
2780
2781 VOP_UNLOCK(vp, 0);
2782
2783 /*
2784 * If the VGET operation doesn't work for this filesystem,
2785 * we can't support readdirplus. Returning NOTSUPP should
2786 * make clients fall back to plain readdir.
2787 * There's no need to check for VPTOFH as well, we wouldn't
2788 * even be here otherwise.
2789 */
2790 if (!getret) {
2791 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
2792 getret = (getret == EOPNOTSUPP) ?
2793 NFSERR_NOTSUPP : NFSERR_IO;
2794 else
2795 vput(nvp);
2796 }
2797
2798 if (!cookies && !error)
2799 error = NFSERR_PERM;
2800 if (!error)
2801 error = getret;
2802 if (error) {
2803 vrele(vp);
2804 if (cookies)
2805 free((caddr_t)cookies, M_TEMP);
2806 free((caddr_t)rbuf, M_TEMP);
2807 nfsm_reply(NFSX_V3POSTOPATTR);
2808 nfsm_srvpostop_attr(getret, &at);
2809 return (0);
2810 }
2811 if (io.uio_resid) {
2812 siz -= io.uio_resid;
2813
2814 /*
2815 * If nothing read, return eof
2816 * rpc reply
2817 */
2818 if (siz == 0) {
2819 vrele(vp);
2820 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
2821 2 * NFSX_UNSIGNED);
2822 nfsm_srvpostop_attr(getret, &at);
2823 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2824 txdr_hyper(at.va_filerev, tl);
2825 tl += 2;
2826 *tl++ = nfs_false;
2827 *tl = nfs_true;
2828 free((caddr_t)cookies, M_TEMP);
2829 free((caddr_t)rbuf, M_TEMP);
2830 return (0);
2831 }
2832 }
2833
2834 /*
2835 * Check for degenerate cases of nothing useful read.
2836 * If so go try again
2837 */
2838 cpos = rbuf;
2839 cend = rbuf + siz;
2840 dp = (struct dirent *)cpos;
2841 cookiep = cookies;
2842
2843 while (cpos < cend && ncookies > 0 &&
2844 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2845 cpos += dp->d_reclen;
2846 dp = (struct dirent *)cpos;
2847 cookiep++;
2848 ncookies--;
2849 }
2850 if (cpos >= cend || ncookies == 0) {
2851 toff = off;
2852 siz = fullsiz;
2853 goto again;
2854 }
2855
2856 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
2857 nfsm_reply(cnt);
2858 nfsm_srvpostop_attr(getret, &at);
2859 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2860 txdr_hyper(at.va_filerev, tl);
2861 mp = mp2 = mb;
2862 bp = bpos;
2863 be = bp + M_TRAILINGSPACE(mp);
2864
2865 /* Loop through the records and build reply */
2866 while (cpos < cend && ncookies > 0) {
2867 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2868 nlen = dp->d_namlen;
2869 rem = nfsm_rndup(nlen)-nlen;
2870
2871 /*
2872 * For readdir_and_lookup get the vnode using
2873 * the file number.
2874 */
2875 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
2876 goto invalid;
2877 memset((caddr_t)nfhp, 0, NFSX_V3FH);
2878 nfhp->fh_fsid =
2879 nvp->v_mount->mnt_stat.f_fsid;
2880 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
2881 vput(nvp);
2882 goto invalid;
2883 }
2884 if (VOP_GETATTR(nvp, vap, cred, procp)) {
2885 vput(nvp);
2886 goto invalid;
2887 }
2888 vput(nvp);
2889
2890 /*
2891 * If either the dircount or maxcount will be
2892 * exceeded, get out now. Both of these lengths
2893 * are calculated conservatively, including all
2894 * XDR overheads.
2895 */
2896 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
2897 NFSX_V3POSTOPATTR);
2898 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
2899 if (len > cnt || dirlen > fullsiz) {
2900 eofflag = 0;
2901 break;
2902 }
2903
2904 /*
2905 * Build the directory record xdr from
2906 * the dirent entry.
2907 */
2908 fp = (struct nfs_fattr *)&fl.fl_fattr;
2909 nfsm_srvfillattr(vap, fp);
2910 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
2911 fl.fl_fhok = nfs_true;
2912 fl.fl_postopok = nfs_true;
2913 txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
2914
2915 nfsm_clget;
2916 *tl = nfs_true;
2917 bp += NFSX_UNSIGNED;
2918 nfsm_clget;
2919 *tl = 0;
2920 bp += NFSX_UNSIGNED;
2921 nfsm_clget;
2922 *tl = txdr_unsigned(dp->d_fileno);
2923 bp += NFSX_UNSIGNED;
2924 nfsm_clget;
2925 *tl = txdr_unsigned(nlen);
2926 bp += NFSX_UNSIGNED;
2927
2928 /* And loop around copying the name */
2929 xfer = nlen;
2930 cp = dp->d_name;
2931 while (xfer > 0) {
2932 nfsm_clget;
2933 if ((bp + xfer) > be)
2934 tsiz = be - bp;
2935 else
2936 tsiz = xfer;
2937 memcpy(bp, cp, tsiz);
2938 bp += tsiz;
2939 xfer -= tsiz;
2940 if (xfer > 0)
2941 cp += tsiz;
2942 }
2943 /* And null pad to an int32_t boundary */
2944 for (i = 0; i < rem; i++)
2945 *bp++ = '\0';
2946
2947 /*
2948 * Now copy the flrep structure out.
2949 */
2950 xfer = sizeof (struct flrep);
2951 cp = (caddr_t)&fl;
2952 while (xfer > 0) {
2953 nfsm_clget;
2954 if ((bp + xfer) > be)
2955 tsiz = be - bp;
2956 else
2957 tsiz = xfer;
2958 memcpy(bp, cp, tsiz);
2959 bp += tsiz;
2960 xfer -= tsiz;
2961 if (xfer > 0)
2962 cp += tsiz;
2963 }
2964 }
2965 invalid:
2966 cpos += dp->d_reclen;
2967 dp = (struct dirent *)cpos;
2968 cookiep++;
2969 ncookies--;
2970 }
2971 vrele(vp);
2972 nfsm_clget;
2973 *tl = nfs_false;
2974 bp += NFSX_UNSIGNED;
2975 nfsm_clget;
2976 if (eofflag)
2977 *tl = nfs_true;
2978 else
2979 *tl = nfs_false;
2980 bp += NFSX_UNSIGNED;
2981 if (mp != mb) {
2982 if (bp < be)
2983 mp->m_len = bp - mtod(mp, caddr_t);
2984 } else
2985 mp->m_len += bp - bpos;
2986 free((caddr_t)cookies, M_TEMP);
2987 free((caddr_t)rbuf, M_TEMP);
2988 nfsm_srvdone;
2989 }
2990
2991 /*
2992 * nfs commit service
2993 */
2994 int
2995 nfsrv_commit(nfsd, slp, procp, mrq)
2996 struct nfsrv_descript *nfsd;
2997 struct nfssvc_sock *slp;
2998 struct proc *procp;
2999 struct mbuf **mrq;
3000 {
3001 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3002 struct mbuf *nam = nfsd->nd_nam;
3003 caddr_t dpos = nfsd->nd_dpos;
3004 struct ucred *cred = &nfsd->nd_cr;
3005 struct vattr bfor, aft;
3006 struct vnode *vp;
3007 nfsfh_t nfh;
3008 fhandle_t *fhp;
3009 u_int32_t *tl;
3010 int32_t t1;
3011 caddr_t bpos;
3012 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache;
3013 char *cp2;
3014 struct mbuf *mb, *mreq;
3015 u_quad_t frev, off, end;
3016
3017 #ifndef nolint
3018 cache = 0;
3019 #endif
3020 fhp = &nfh.fh_generic;
3021 nfsm_srvmtofh(fhp);
3022 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3023
3024 off = fxdr_hyper(tl);
3025 tl += 2;
3026 cnt = fxdr_unsigned(int, *tl);
3027 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3028 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3029 if (error) {
3030 nfsm_reply(2 * NFSX_UNSIGNED);
3031 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3032 return (0);
3033 }
3034 for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
3035 end = (cnt > 0) ? off + cnt : vp->v_size;
3036 if (end < off || end > vp->v_size)
3037 end = vp->v_size;
3038 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, procp);
3039 aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
3040 vput(vp);
3041 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3042 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3043 if (!error) {
3044 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3045 *tl++ = txdr_unsigned(boottime.tv_sec);
3046 *tl = txdr_unsigned(boottime.tv_usec);
3047 } else
3048 return (0);
3049 nfsm_srvdone;
3050 }
3051
3052 /*
3053 * nfs statfs service
3054 */
3055 int
3056 nfsrv_statfs(nfsd, slp, procp, mrq)
3057 struct nfsrv_descript *nfsd;
3058 struct nfssvc_sock *slp;
3059 struct proc *procp;
3060 struct mbuf **mrq;
3061 {
3062 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3063 struct mbuf *nam = nfsd->nd_nam;
3064 caddr_t dpos = nfsd->nd_dpos;
3065 struct ucred *cred = &nfsd->nd_cr;
3066 struct statfs *sf;
3067 struct nfs_statfs *sfp;
3068 u_int32_t *tl;
3069 int32_t t1;
3070 caddr_t bpos;
3071 int error = 0, rdonly, cache, getret = 1;
3072 int v3 = (nfsd->nd_flag & ND_NFSV3);
3073 char *cp2;
3074 struct mbuf *mb, *mreq;
3075 struct vnode *vp;
3076 struct vattr at;
3077 nfsfh_t nfh;
3078 fhandle_t *fhp;
3079 struct statfs statfs;
3080 u_quad_t frev, tval;
3081
3082 #ifndef nolint
3083 cache = 0;
3084 #endif
3085 fhp = &nfh.fh_generic;
3086 nfsm_srvmtofh(fhp);
3087 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3088 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3089 if (error) {
3090 nfsm_reply(NFSX_UNSIGNED);
3091 nfsm_srvpostop_attr(getret, &at);
3092 return (0);
3093 }
3094 sf = &statfs;
3095 error = VFS_STATFS(vp->v_mount, sf, procp);
3096 getret = VOP_GETATTR(vp, &at, cred, procp);
3097 vput(vp);
3098 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3099 if (v3)
3100 nfsm_srvpostop_attr(getret, &at);
3101 if (error)
3102 return (0);
3103 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3104 if (v3) {
3105 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_bsize);
3106 txdr_hyper(tval, &sfp->sf_tbytes);
3107 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_bsize);
3108 txdr_hyper(tval, &sfp->sf_fbytes);
3109 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_bsize);
3110 txdr_hyper(tval, &sfp->sf_abytes);
3111 tval = (u_quad_t)sf->f_files;
3112 txdr_hyper(tval, &sfp->sf_tfiles);
3113 tval = (u_quad_t)sf->f_ffree;
3114 txdr_hyper(tval, &sfp->sf_ffiles);
3115 txdr_hyper(tval, &sfp->sf_afiles);
3116 sfp->sf_invarsec = 0;
3117 } else {
3118 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3119 sfp->sf_bsize = txdr_unsigned(sf->f_bsize);
3120 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3121 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3122 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3123 }
3124 nfsm_srvdone;
3125 }
3126
3127 /*
3128 * nfs fsinfo service
3129 */
3130 int
3131 nfsrv_fsinfo(nfsd, slp, procp, mrq)
3132 struct nfsrv_descript *nfsd;
3133 struct nfssvc_sock *slp;
3134 struct proc *procp;
3135 struct mbuf **mrq;
3136 {
3137 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3138 struct mbuf *nam = nfsd->nd_nam;
3139 caddr_t dpos = nfsd->nd_dpos;
3140 struct ucred *cred = &nfsd->nd_cr;
3141 u_int32_t *tl;
3142 struct nfsv3_fsinfo *sip;
3143 int32_t t1;
3144 caddr_t bpos;
3145 int error = 0, rdonly, cache, getret = 1;
3146 uint32_t maxdata;
3147 char *cp2;
3148 struct mbuf *mb, *mreq;
3149 struct vnode *vp;
3150 struct vattr at;
3151 nfsfh_t nfh;
3152 fhandle_t *fhp;
3153 u_quad_t frev, maxfsize;
3154 struct statfs sb;
3155
3156 #ifndef nolint
3157 cache = 0;
3158 #endif
3159 fhp = &nfh.fh_generic;
3160 nfsm_srvmtofh(fhp);
3161 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3162 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3163 if (error) {
3164 nfsm_reply(NFSX_UNSIGNED);
3165 nfsm_srvpostop_attr(getret, &at);
3166 return (0);
3167 }
3168
3169 /* XXX Try to make a guess on the max file size. */
3170 VFS_STATFS(vp->v_mount, &sb, (struct proc *)0);
3171 maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1;
3172
3173 getret = VOP_GETATTR(vp, &at, cred, procp);
3174 vput(vp);
3175 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3176 nfsm_srvpostop_attr(getret, &at);
3177 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3178
3179 /*
3180 * XXX
3181 * There should be file system VFS OP(s) to get this information.
3182 * For now, assume ufs.
3183 */
3184 if (slp->ns_so->so_type == SOCK_DGRAM)
3185 maxdata = NFS_MAXDGRAMDATA;
3186 else
3187 maxdata = NFS_MAXDATA;
3188 sip->fs_rtmax = txdr_unsigned(maxdata);
3189 sip->fs_rtpref = txdr_unsigned(maxdata);
3190 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3191 sip->fs_wtmax = txdr_unsigned(maxdata);
3192 sip->fs_wtpref = txdr_unsigned(maxdata);
3193 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3194 sip->fs_dtpref = txdr_unsigned(maxdata);
3195 txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3196 sip->fs_timedelta.nfsv3_sec = 0;
3197 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3198 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3199 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3200 NFSV3FSINFO_CANSETTIME);
3201 nfsm_srvdone;
3202 }
3203
3204 /*
3205 * nfs pathconf service
3206 */
3207 int
3208 nfsrv_pathconf(nfsd, slp, procp, mrq)
3209 struct nfsrv_descript *nfsd;
3210 struct nfssvc_sock *slp;
3211 struct proc *procp;
3212 struct mbuf **mrq;
3213 {
3214 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3215 struct mbuf *nam = nfsd->nd_nam;
3216 caddr_t dpos = nfsd->nd_dpos;
3217 struct ucred *cred = &nfsd->nd_cr;
3218 u_int32_t *tl;
3219 struct nfsv3_pathconf *pc;
3220 int32_t t1;
3221 caddr_t bpos;
3222 int error = 0, rdonly, cache, getret = 1;
3223 register_t linkmax, namemax, chownres, notrunc;
3224 char *cp2;
3225 struct mbuf *mb, *mreq;
3226 struct vnode *vp;
3227 struct vattr at;
3228 nfsfh_t nfh;
3229 fhandle_t *fhp;
3230 u_quad_t frev;
3231
3232 #ifndef nolint
3233 cache = 0;
3234 #endif
3235 fhp = &nfh.fh_generic;
3236 nfsm_srvmtofh(fhp);
3237 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3238 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3239 if (error) {
3240 nfsm_reply(NFSX_UNSIGNED);
3241 nfsm_srvpostop_attr(getret, &at);
3242 return (0);
3243 }
3244 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3245 if (!error)
3246 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3247 if (!error)
3248 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3249 if (!error)
3250 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc);
3251 getret = VOP_GETATTR(vp, &at, cred, procp);
3252 vput(vp);
3253 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3254 nfsm_srvpostop_attr(getret, &at);
3255 if (error)
3256 return (0);
3257 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3258
3259 pc->pc_linkmax = txdr_unsigned(linkmax);
3260 pc->pc_namemax = txdr_unsigned(namemax);
3261 pc->pc_notrunc = txdr_unsigned(notrunc);
3262 pc->pc_chownrestricted = txdr_unsigned(chownres);
3263
3264 /*
3265 * These should probably be supported by VOP_PATHCONF(), but
3266 * until msdosfs is exportable (why would you want to?), the
3267 * Unix defaults should be ok.
3268 */
3269 pc->pc_caseinsensitive = nfs_false;
3270 pc->pc_casepreserving = nfs_true;
3271 nfsm_srvdone;
3272 }
3273
3274 /*
3275 * Null operation, used by clients to ping server
3276 */
3277 /* ARGSUSED */
3278 int
3279 nfsrv_null(nfsd, slp, procp, mrq)
3280 struct nfsrv_descript *nfsd;
3281 struct nfssvc_sock *slp;
3282 struct proc *procp;
3283 struct mbuf **mrq;
3284 {
3285 struct mbuf *mrep = nfsd->nd_mrep;
3286 caddr_t bpos;
3287 int error = NFSERR_RETVOID, cache = 0;
3288 struct mbuf *mb, *mreq;
3289 u_quad_t frev;
3290
3291 nfsm_reply(0);
3292 return (0);
3293 }
3294
3295 /*
3296 * No operation, used for obsolete procedures
3297 */
3298 /* ARGSUSED */
3299 int
3300 nfsrv_noop(nfsd, slp, procp, mrq)
3301 struct nfsrv_descript *nfsd;
3302 struct nfssvc_sock *slp;
3303 struct proc *procp;
3304 struct mbuf **mrq;
3305 {
3306 struct mbuf *mrep = nfsd->nd_mrep;
3307 caddr_t bpos;
3308 int error, cache = 0;
3309 struct mbuf *mb, *mreq;
3310 u_quad_t frev;
3311
3312 if (nfsd->nd_repstat)
3313 error = nfsd->nd_repstat;
3314 else
3315 error = EPROCUNAVAIL;
3316 nfsm_reply(0);
3317 return (0);
3318 }
3319
3320 /*
3321 * Perform access checking for vnodes obtained from file handles that would
3322 * refer to files already opened by a Unix client. You cannot just use
3323 * vn_writechk() and VOP_ACCESS() for two reasons.
3324 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3325 * 2 - The owner is to be given access irrespective of mode bits for some
3326 * operations, so that processes that chmod after opening a file don't
3327 * break. I don't like this because it opens a security hole, but since
3328 * the nfs server opens a security hole the size of a barn door anyhow,
3329 * what the heck.
3330 *
3331 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3332 * will return EPERM instead of EACCESS. EPERM is always an error.
3333 */
3334 int
3335 nfsrv_access(vp, flags, cred, rdonly, p, override)
3336 struct vnode *vp;
3337 int flags;
3338 struct ucred *cred;
3339 int rdonly;
3340 struct proc *p;
3341 {
3342 struct vattr vattr;
3343 int error;
3344 if (flags & VWRITE) {
3345 /* Just vn_writechk() changed to check rdonly */
3346 /*
3347 * Disallow write attempts on read-only file systems;
3348 * unless the file is a socket or a block or character
3349 * device resident on the file system.
3350 */
3351 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3352 switch (vp->v_type) {
3353 case VREG:
3354 case VDIR:
3355 case VLNK:
3356 return (EROFS);
3357 default:
3358 break;
3359 }
3360 }
3361
3362 /*
3363 * If the vnode is in use as a process's text,
3364 * we can't allow writing.
3365 */
3366 if (vp->v_flag & VTEXT)
3367 return (ETXTBSY);
3368 }
3369 error = VOP_GETATTR(vp, &vattr, cred, p);
3370 if (error)
3371 return (error);
3372 error = VOP_ACCESS(vp, flags, cred, p);
3373 /*
3374 * Allow certain operations for the owner (reads and writes
3375 * on files that are already open).
3376 */
3377 if (override && error == EACCES && cred->cr_uid == vattr.va_uid)
3378 error = 0;
3379 return error;
3380 }
3381