nfs_serv.c revision 1.65 1 /* $NetBSD: nfs_serv.c,v 1.65 2002/09/27 19:30:46 bouyer 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 1.65 2002/09/27 19:30:46 bouyer 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 nfstype nfsv2_type[9];
95 extern 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, *mb2;
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, *mb2, *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, *mb2, *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, len, dirattr_ret = 1;
368 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
369 char *cp2;
370 struct mbuf *mb, *mb2, *mreq;
371 struct vattr va, dirattr;
372 u_quad_t frev;
373
374 fhp = &nfh.fh_generic;
375 nfsm_srvmtofh(fhp);
376 nfsm_srvnamesiz(len);
377
378 pubflag = nfs_ispublicfh(fhp);
379
380 nd.ni_cnd.cn_cred = cred;
381 nd.ni_cnd.cn_nameiop = LOOKUP;
382 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
383 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
384 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
385
386 if (!error && pubflag) {
387 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
388 /*
389 * Setup call to lookup() to see if we can find
390 * the index file. Arguably, this doesn't belong
391 * in a kernel.. Ugh.
392 */
393 ind = nd;
394 VOP_UNLOCK(nd.ni_vp, 0);
395 ind.ni_pathlen = strlen(nfs_pub.np_index);
396 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
397 nfs_pub.np_index;
398 ind.ni_startdir = nd.ni_vp;
399 VREF(ind.ni_startdir);
400 error = lookup(&ind);
401 if (!error) {
402 /*
403 * Found an index file. Get rid of
404 * the old references.
405 */
406 if (dirp)
407 vrele(dirp);
408 dirp = nd.ni_vp;
409 vrele(nd.ni_startdir);
410 ndp = &ind;
411 } else
412 error = 0;
413 }
414 /*
415 * If the public filehandle was used, check that this lookup
416 * didn't result in a filehandle outside the publicly exported
417 * filesystem.
418 */
419
420 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
421 vput(nd.ni_vp);
422 error = EPERM;
423 }
424 }
425
426 if (dirp) {
427 if (v3)
428 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred,
429 procp);
430 vrele(dirp);
431 }
432
433 if (error) {
434 nfsm_reply(NFSX_POSTOPATTR(v3));
435 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
436 return (0);
437 }
438
439 nqsrv_getl(ndp->ni_startdir, ND_READ);
440 vrele(ndp->ni_startdir);
441 PNBUF_PUT(nd.ni_cnd.cn_pnbuf);
442 vp = ndp->ni_vp;
443 memset((caddr_t)fhp, 0, sizeof(nfh));
444 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
445 error = VFS_VPTOFH(vp, &fhp->fh_fid);
446 if (!error)
447 error = VOP_GETATTR(vp, &va, cred, procp);
448 vput(vp);
449 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
450 if (error) {
451 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
452 return (0);
453 }
454 nfsm_srvfhtom(fhp, v3);
455 if (v3) {
456 nfsm_srvpostop_attr(0, &va);
457 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
458 } else {
459 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
460 nfsm_srvfillattr(&va, fp);
461 }
462 nfsm_srvdone;
463 }
464
465 /*
466 * nfs readlink service
467 */
468 int
469 nfsrv_readlink(nfsd, slp, procp, mrq)
470 struct nfsrv_descript *nfsd;
471 struct nfssvc_sock *slp;
472 struct proc *procp;
473 struct mbuf **mrq;
474 {
475 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
476 struct mbuf *nam = nfsd->nd_nam;
477 caddr_t dpos = nfsd->nd_dpos;
478 struct ucred *cred = &nfsd->nd_cr;
479 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
480 struct iovec *ivp = iv;
481 struct mbuf *mp;
482 u_int32_t *tl;
483 int32_t t1;
484 caddr_t bpos;
485 int error = 0, rdonly, cache, i, tlen, len, getret;
486 int v3 = (nfsd->nd_flag & ND_NFSV3);
487 char *cp2;
488 struct mbuf *mb, *mb2, *mp2 = NULL, *mp3 = NULL, *mreq;
489 struct vnode *vp;
490 struct vattr attr;
491 nfsfh_t nfh;
492 fhandle_t *fhp;
493 struct uio io, *uiop = &io;
494 u_quad_t frev;
495
496 fhp = &nfh.fh_generic;
497 nfsm_srvmtofh(fhp);
498 len = 0;
499 i = 0;
500 while (len < NFS_MAXPATHLEN) {
501 MGET(mp, M_WAIT, MT_DATA);
502 MCLGET(mp, M_WAIT);
503 mp->m_len = NFSMSIZ(mp);
504 if (len == 0)
505 mp3 = mp2 = mp;
506 else {
507 mp2->m_next = mp;
508 mp2 = mp;
509 }
510 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
511 mp->m_len = NFS_MAXPATHLEN-len;
512 len = NFS_MAXPATHLEN;
513 } else
514 len += mp->m_len;
515 ivp->iov_base = mtod(mp, caddr_t);
516 ivp->iov_len = mp->m_len;
517 i++;
518 ivp++;
519 }
520 uiop->uio_iov = iv;
521 uiop->uio_iovcnt = i;
522 uiop->uio_offset = 0;
523 uiop->uio_resid = len;
524 uiop->uio_rw = UIO_READ;
525 uiop->uio_segflg = UIO_SYSSPACE;
526 uiop->uio_procp = (struct proc *)0;
527 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
528 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
529 if (error) {
530 m_freem(mp3);
531 nfsm_reply(2 * NFSX_UNSIGNED);
532 nfsm_srvpostop_attr(1, (struct vattr *)0);
533 return (0);
534 }
535 if (vp->v_type != VLNK) {
536 if (v3)
537 error = EINVAL;
538 else
539 error = ENXIO;
540 goto out;
541 }
542 nqsrv_getl(vp, ND_READ);
543 error = VOP_READLINK(vp, uiop, cred);
544 out:
545 getret = VOP_GETATTR(vp, &attr, cred, procp);
546 vput(vp);
547 if (error)
548 m_freem(mp3);
549 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
550 if (v3) {
551 nfsm_srvpostop_attr(getret, &attr);
552 if (error)
553 return (0);
554 }
555 if (uiop->uio_resid > 0) {
556 len -= uiop->uio_resid;
557 tlen = nfsm_rndup(len);
558 nfsm_adj(mp3, NFS_MAXPATHLEN-tlen, tlen-len);
559 }
560 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
561 *tl = txdr_unsigned(len);
562 mb->m_next = mp3;
563 nfsm_srvdone;
564 }
565
566 /*
567 * nfs read service
568 */
569 int
570 nfsrv_read(nfsd, slp, procp, mrq)
571 struct nfsrv_descript *nfsd;
572 struct nfssvc_sock *slp;
573 struct proc *procp;
574 struct mbuf **mrq;
575 {
576 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
577 struct mbuf *nam = nfsd->nd_nam;
578 caddr_t dpos = nfsd->nd_dpos;
579 struct ucred *cred = &nfsd->nd_cr;
580 struct iovec *iv;
581 struct iovec *iv2;
582 struct mbuf *m;
583 struct nfs_fattr *fp;
584 u_int32_t *tl;
585 int32_t t1;
586 int i;
587 caddr_t bpos;
588 int error = 0, rdonly, cache, cnt, len, left, siz, tlen, getret;
589 int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen;
590 char *cp2;
591 struct mbuf *mb, *mb2, *mreq;
592 struct mbuf *m2;
593 struct vnode *vp;
594 nfsfh_t nfh;
595 fhandle_t *fhp;
596 struct uio io, *uiop = &io;
597 struct vattr va;
598 off_t off;
599 u_quad_t frev;
600
601 fhp = &nfh.fh_generic;
602 nfsm_srvmtofh(fhp);
603 if (v3) {
604 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
605 off = fxdr_hyper(tl);
606 } else {
607 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
608 off = (off_t)fxdr_unsigned(u_int32_t, *tl);
609 }
610 nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd));
611 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
612 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
613 if (error) {
614 nfsm_reply(2 * NFSX_UNSIGNED);
615 nfsm_srvpostop_attr(1, (struct vattr *)0);
616 return (0);
617 }
618 if (vp->v_type != VREG) {
619 if (v3)
620 error = EINVAL;
621 else
622 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
623 }
624 if (!error) {
625 nqsrv_getl(vp, ND_READ);
626 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0)
627 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1);
628 }
629 getret = VOP_GETATTR(vp, &va, cred, procp);
630 if (!error)
631 error = getret;
632 if (error) {
633 vput(vp);
634 nfsm_reply(NFSX_POSTOPATTR(v3));
635 nfsm_srvpostop_attr(getret, &va);
636 return (0);
637 }
638 if (off >= va.va_size)
639 cnt = 0;
640 else if ((off + reqlen) > va.va_size)
641 cnt = nfsm_rndup(va.va_size - off);
642 else
643 cnt = reqlen;
644 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
645 if (v3) {
646 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
647 *tl++ = nfs_true;
648 fp = (struct nfs_fattr *)tl;
649 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
650 } else {
651 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
652 fp = (struct nfs_fattr *)tl;
653 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
654 }
655 len = left = cnt;
656 if (cnt > 0) {
657 /*
658 * Generate the mbuf list with the uio_iov ref. to it.
659 */
660 i = 0;
661 m = m2 = mb;
662 while (left > 0) {
663 siz = min(M_TRAILINGSPACE(m), left);
664 if (siz > 0) {
665 left -= siz;
666 i++;
667 }
668 if (left > 0) {
669 MGET(m, M_WAIT, MT_DATA);
670 MCLGET(m, M_WAIT);
671 m->m_len = 0;
672 m2->m_next = m;
673 m2 = m;
674 }
675 }
676 iv = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
677 uiop->uio_iov = iv2 = iv;
678 m = mb;
679 left = cnt;
680 i = 0;
681 while (left > 0) {
682 if (m == NULL)
683 panic("nfsrv_read iov");
684 siz = min(M_TRAILINGSPACE(m), left);
685 if (siz > 0) {
686 iv->iov_base = mtod(m, caddr_t) + m->m_len;
687 iv->iov_len = siz;
688 m->m_len += siz;
689 left -= siz;
690 iv++;
691 i++;
692 }
693 m = m->m_next;
694 }
695 uiop->uio_iovcnt = i;
696 uiop->uio_offset = off;
697 uiop->uio_resid = cnt;
698 uiop->uio_rw = UIO_READ;
699 uiop->uio_segflg = UIO_SYSSPACE;
700 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
701 off = uiop->uio_offset;
702 free((caddr_t)iv2, M_TEMP);
703 if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){
704 if (!error)
705 error = getret;
706 m_freem(mreq);
707 vput(vp);
708 nfsm_reply(NFSX_POSTOPATTR(v3));
709 nfsm_srvpostop_attr(getret, &va);
710 return (0);
711 }
712 } else
713 uiop->uio_resid = 0;
714 vput(vp);
715 nfsm_srvfillattr(&va, fp);
716 len -= uiop->uio_resid;
717 tlen = nfsm_rndup(len);
718 if (cnt != tlen || tlen != len)
719 nfsm_adj(mb, cnt - tlen, tlen - len);
720 if (v3) {
721 *tl++ = txdr_unsigned(len);
722 if (len < reqlen)
723 *tl++ = nfs_true;
724 else
725 *tl++ = nfs_false;
726 }
727 *tl = txdr_unsigned(len);
728 nfsm_srvdone;
729 }
730
731 /*
732 * nfs write service
733 */
734 int
735 nfsrv_write(nfsd, slp, procp, mrq)
736 struct nfsrv_descript *nfsd;
737 struct nfssvc_sock *slp;
738 struct proc *procp;
739 struct mbuf **mrq;
740 {
741 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
742 struct mbuf *nam = nfsd->nd_nam;
743 caddr_t dpos = nfsd->nd_dpos;
744 struct ucred *cred = &nfsd->nd_cr;
745 struct iovec *ivp;
746 int i, cnt;
747 struct mbuf *mp;
748 struct nfs_fattr *fp;
749 struct iovec *iv;
750 struct vattr va, forat;
751 u_int32_t *tl;
752 int32_t t1;
753 caddr_t bpos;
754 int error = 0, rdonly, cache, len, forat_ret = 1;
755 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
756 int stable = NFSV3WRITE_FILESYNC;
757 int v3 = (nfsd->nd_flag & ND_NFSV3);
758 char *cp2;
759 struct mbuf *mb, *mb2, *mreq;
760 struct vnode *vp;
761 nfsfh_t nfh;
762 fhandle_t *fhp;
763 struct uio io, *uiop = &io;
764 off_t off;
765 u_quad_t frev;
766
767 if (mrep == NULL) {
768 *mrq = NULL;
769 return (0);
770 }
771 fhp = &nfh.fh_generic;
772 nfsm_srvmtofh(fhp);
773 if (v3) {
774 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
775 off = fxdr_hyper(tl);
776 tl += 3;
777 stable = fxdr_unsigned(int, *tl++);
778 } else {
779 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
780 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
781 tl += 2;
782 }
783 retlen = len = fxdr_unsigned(int32_t, *tl);
784 cnt = i = 0;
785
786 /*
787 * For NFS Version 2, it is not obvious what a write of zero length
788 * should do, but I might as well be consistent with Version 3,
789 * which is to return ok so long as there are no permission problems.
790 */
791 if (len > 0) {
792 zeroing = 1;
793 mp = mrep;
794 while (mp) {
795 if (mp == md) {
796 zeroing = 0;
797 adjust = dpos - mtod(mp, caddr_t);
798 mp->m_len -= adjust;
799 if (mp->m_len > 0 && adjust > 0)
800 NFSMADV(mp, adjust);
801 }
802 if (zeroing)
803 mp->m_len = 0;
804 else if (mp->m_len > 0) {
805 i += mp->m_len;
806 if (i > len) {
807 mp->m_len -= (i - len);
808 zeroing = 1;
809 }
810 if (mp->m_len > 0)
811 cnt++;
812 }
813 mp = mp->m_next;
814 }
815 }
816 if (len > NFS_MAXDATA || len < 0 || i < len) {
817 error = EIO;
818 nfsm_reply(2 * NFSX_UNSIGNED);
819 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
820 return (0);
821 }
822 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
823 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
824 if (error) {
825 nfsm_reply(2 * NFSX_UNSIGNED);
826 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
827 return (0);
828 }
829 if (v3)
830 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
831 if (vp->v_type != VREG) {
832 if (v3)
833 error = EINVAL;
834 else
835 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
836 }
837 if (!error) {
838 nqsrv_getl(vp, ND_WRITE);
839 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
840 }
841 if (error) {
842 vput(vp);
843 nfsm_reply(NFSX_WCCDATA(v3));
844 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
845 return (0);
846 }
847
848 if (len > 0) {
849 ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK);
850 uiop->uio_iov = iv = ivp;
851 uiop->uio_iovcnt = cnt;
852 mp = mrep;
853 while (mp) {
854 if (mp->m_len > 0) {
855 ivp->iov_base = mtod(mp, caddr_t);
856 ivp->iov_len = mp->m_len;
857 ivp++;
858 }
859 mp = mp->m_next;
860 }
861
862 /*
863 * XXX
864 * The IO_METASYNC flag indicates that all metadata (and not just
865 * enough to ensure data integrity) mus be written to stable storage
866 * synchronously.
867 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
868 */
869 if (stable == NFSV3WRITE_UNSTABLE)
870 ioflags = IO_NODELOCKED;
871 else if (stable == NFSV3WRITE_DATASYNC)
872 ioflags = (IO_SYNC | IO_NODELOCKED);
873 else
874 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
875 uiop->uio_resid = len;
876 uiop->uio_rw = UIO_WRITE;
877 uiop->uio_segflg = UIO_SYSSPACE;
878 uiop->uio_procp = (struct proc *)0;
879 uiop->uio_offset = off;
880 error = VOP_WRITE(vp, uiop, ioflags, cred);
881 nfsstats.srvvop_writes++;
882 free((caddr_t)iv, M_TEMP);
883 }
884 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
885 vput(vp);
886 if (!error)
887 error = aftat_ret;
888 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
889 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
890 if (v3) {
891 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
892 if (error)
893 return (0);
894 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
895 *tl++ = txdr_unsigned(retlen);
896 if (stable == NFSV3WRITE_UNSTABLE)
897 *tl++ = txdr_unsigned(stable);
898 else
899 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
900 /*
901 * Actually, there is no need to txdr these fields,
902 * but it may make the values more human readable,
903 * for debugging purposes.
904 */
905 *tl++ = txdr_unsigned(boottime.tv_sec);
906 *tl = txdr_unsigned(boottime.tv_usec);
907 } else {
908 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
909 nfsm_srvfillattr(&va, fp);
910 }
911 nfsm_srvdone;
912 }
913
914 /*
915 * NFS write service with write gathering support. Called when
916 * nfsrvw_procrastinate > 0.
917 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
918 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
919 * Jan. 1994.
920 */
921 int
922 nfsrv_writegather(ndp, slp, procp, mrq)
923 struct nfsrv_descript **ndp;
924 struct nfssvc_sock *slp;
925 struct proc *procp;
926 struct mbuf **mrq;
927 {
928 struct iovec *ivp;
929 struct mbuf *mp;
930 struct nfsrv_descript *wp, *nfsd, *owp, *swp;
931 struct nfs_fattr *fp;
932 int i = 0;
933 struct iovec *iov;
934 struct nfsrvw_delayhash *wpp;
935 struct ucred *cred;
936 struct vattr va, forat;
937 u_int32_t *tl;
938 int32_t t1;
939 caddr_t bpos, dpos;
940 int error = 0, rdonly, cache, len = 0, forat_ret = 1;
941 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
942 char *cp2;
943 struct mbuf *mb, *mb2, *mreq, *mrep, *md;
944 struct vnode *vp;
945 struct uio io, *uiop = &io;
946 u_quad_t frev, cur_usec;
947
948 *mrq = NULL;
949 if (*ndp) {
950 nfsd = *ndp;
951 *ndp = NULL;
952 mrep = nfsd->nd_mrep;
953 md = nfsd->nd_md;
954 dpos = nfsd->nd_dpos;
955 cred = &nfsd->nd_cr;
956 v3 = (nfsd->nd_flag & ND_NFSV3);
957 LIST_INIT(&nfsd->nd_coalesce);
958 nfsd->nd_mreq = NULL;
959 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
960 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
961 nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
962
963 /*
964 * Now, get the write header..
965 */
966 nfsm_srvmtofh(&nfsd->nd_fh);
967 if (v3) {
968 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
969 nfsd->nd_off = fxdr_hyper(tl);
970 tl += 3;
971 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
972 } else {
973 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
974 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
975 tl += 2;
976 }
977 len = fxdr_unsigned(int32_t, *tl);
978 nfsd->nd_len = len;
979 nfsd->nd_eoff = nfsd->nd_off + len;
980
981 /*
982 * Trim the header out of the mbuf list and trim off any trailing
983 * junk so that the mbuf list has only the write data.
984 */
985 zeroing = 1;
986 i = 0;
987 mp = mrep;
988 while (mp) {
989 if (mp == md) {
990 zeroing = 0;
991 adjust = dpos - mtod(mp, caddr_t);
992 mp->m_len -= adjust;
993 if (mp->m_len > 0 && adjust > 0)
994 NFSMADV(mp, adjust);
995 }
996 if (zeroing)
997 mp->m_len = 0;
998 else {
999 i += mp->m_len;
1000 if (i > len) {
1001 mp->m_len -= (i - len);
1002 zeroing = 1;
1003 }
1004 }
1005 mp = mp->m_next;
1006 }
1007 if (len > NFS_MAXDATA || len < 0 || i < len) {
1008 nfsmout:
1009 m_freem(mrep);
1010 error = EIO;
1011 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1012 if (v3)
1013 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1014 nfsd->nd_mreq = mreq;
1015 nfsd->nd_mrep = NULL;
1016 nfsd->nd_time = 0;
1017 }
1018
1019 /*
1020 * Add this entry to the hash and time queues.
1021 */
1022 s = splsoftclock();
1023 owp = NULL;
1024 wp = slp->ns_tq.lh_first;
1025 while (wp && wp->nd_time < nfsd->nd_time) {
1026 owp = wp;
1027 wp = wp->nd_tq.le_next;
1028 }
1029 if (owp) {
1030 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1031 } else {
1032 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1033 }
1034 if (nfsd->nd_mrep) {
1035 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1036 owp = NULL;
1037 wp = wpp->lh_first;
1038 while (wp &&
1039 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1040 owp = wp;
1041 wp = wp->nd_hash.le_next;
1042 }
1043 while (wp && wp->nd_off < nfsd->nd_off &&
1044 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1045 owp = wp;
1046 wp = wp->nd_hash.le_next;
1047 }
1048 if (owp) {
1049 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1050
1051 /*
1052 * Search the hash list for overlapping entries and
1053 * coalesce.
1054 */
1055 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1056 wp = nfsd->nd_hash.le_next;
1057 if (NFSW_SAMECRED(owp, nfsd))
1058 nfsrvw_coalesce(owp, nfsd);
1059 }
1060 } else {
1061 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1062 }
1063 }
1064 splx(s);
1065 }
1066
1067 /*
1068 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1069 * and generate the associated reply mbuf list(s).
1070 */
1071 loop1:
1072 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1073 s = splsoftclock();
1074 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = owp) {
1075 owp = nfsd->nd_tq.le_next;
1076 if (nfsd->nd_time > cur_usec)
1077 break;
1078 if (nfsd->nd_mreq)
1079 continue;
1080 LIST_REMOVE(nfsd, nd_tq);
1081 LIST_REMOVE(nfsd, nd_hash);
1082 splx(s);
1083 mrep = nfsd->nd_mrep;
1084 nfsd->nd_mrep = NULL;
1085 cred = &nfsd->nd_cr;
1086 v3 = (nfsd->nd_flag & ND_NFSV3);
1087 forat_ret = aftat_ret = 1;
1088 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1089 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1090 FALSE);
1091 if (!error) {
1092 if (v3)
1093 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
1094 if (vp->v_type != VREG) {
1095 if (v3)
1096 error = EINVAL;
1097 else
1098 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1099 }
1100 } else
1101 vp = NULL;
1102 if (!error) {
1103 nqsrv_getl(vp, ND_WRITE);
1104 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
1105 }
1106
1107 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1108 ioflags = IO_NODELOCKED;
1109 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1110 ioflags = (IO_SYNC | IO_NODELOCKED);
1111 else
1112 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1113 uiop->uio_rw = UIO_WRITE;
1114 uiop->uio_segflg = UIO_SYSSPACE;
1115 uiop->uio_procp = (struct proc *)0;
1116 uiop->uio_offset = nfsd->nd_off;
1117 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1118 if (uiop->uio_resid > 0) {
1119 mp = mrep;
1120 i = 0;
1121 while (mp) {
1122 if (mp->m_len > 0)
1123 i++;
1124 mp = mp->m_next;
1125 }
1126 uiop->uio_iovcnt = i;
1127 iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
1128 uiop->uio_iov = ivp = iov;
1129 mp = mrep;
1130 while (mp) {
1131 if (mp->m_len > 0) {
1132 ivp->iov_base = mtod(mp, caddr_t);
1133 ivp->iov_len = mp->m_len;
1134 ivp++;
1135 }
1136 mp = mp->m_next;
1137 }
1138 if (!error) {
1139 error = VOP_WRITE(vp, uiop, ioflags, cred);
1140 nfsstats.srvvop_writes++;
1141 }
1142 free((caddr_t)iov, M_TEMP);
1143 }
1144 m_freem(mrep);
1145 if (vp) {
1146 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
1147 vput(vp);
1148 }
1149
1150 /*
1151 * Loop around generating replies for all write rpcs that have
1152 * now been completed.
1153 */
1154 swp = nfsd;
1155 do {
1156 if (error) {
1157 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1158 if (v3) {
1159 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1160 }
1161 } else {
1162 nfsm_writereply(NFSX_PREOPATTR(v3) +
1163 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1164 NFSX_WRITEVERF(v3), v3);
1165 if (v3) {
1166 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1167 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1168 *tl++ = txdr_unsigned(nfsd->nd_len);
1169 *tl++ = txdr_unsigned(swp->nd_stable);
1170 /*
1171 * Actually, there is no need to txdr these fields,
1172 * but it may make the values more human readable,
1173 * for debugging purposes.
1174 */
1175 *tl++ = txdr_unsigned(boottime.tv_sec);
1176 *tl = txdr_unsigned(boottime.tv_usec);
1177 } else {
1178 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1179 nfsm_srvfillattr(&va, fp);
1180 }
1181 }
1182 nfsd->nd_mreq = mreq;
1183 if (nfsd->nd_mrep)
1184 panic("nfsrv_write: nd_mrep not free");
1185
1186 /*
1187 * Done. Put it at the head of the timer queue so that
1188 * the final phase can return the reply.
1189 */
1190 s = splsoftclock();
1191 if (nfsd != swp) {
1192 nfsd->nd_time = 0;
1193 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1194 }
1195 nfsd = swp->nd_coalesce.lh_first;
1196 if (nfsd) {
1197 LIST_REMOVE(nfsd, nd_tq);
1198 }
1199 splx(s);
1200 } while (nfsd);
1201 s = splsoftclock();
1202 swp->nd_time = 0;
1203 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1204 splx(s);
1205 goto loop1;
1206 }
1207 splx(s);
1208
1209 /*
1210 * Search for a reply to return.
1211 */
1212 s = splsoftclock();
1213 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = nfsd->nd_tq.le_next)
1214 if (nfsd->nd_mreq) {
1215 LIST_REMOVE(nfsd, nd_tq);
1216 *mrq = nfsd->nd_mreq;
1217 *ndp = nfsd;
1218 break;
1219 }
1220 splx(s);
1221 return (0);
1222 }
1223
1224 /*
1225 * Coalesce the write request nfsd into owp. To do this we must:
1226 * - remove nfsd from the queues
1227 * - merge nfsd->nd_mrep into owp->nd_mrep
1228 * - update the nd_eoff and nd_stable for owp
1229 * - put nfsd on owp's nd_coalesce list
1230 * NB: Must be called at splsoftclock().
1231 */
1232 void
1233 nfsrvw_coalesce(owp, nfsd)
1234 struct nfsrv_descript *owp;
1235 struct nfsrv_descript *nfsd;
1236 {
1237 int overlap;
1238 struct mbuf *mp;
1239
1240 LIST_REMOVE(nfsd, nd_hash);
1241 LIST_REMOVE(nfsd, nd_tq);
1242 if (owp->nd_eoff < nfsd->nd_eoff) {
1243 overlap = owp->nd_eoff - nfsd->nd_off;
1244 if (overlap < 0)
1245 panic("nfsrv_coalesce: bad off");
1246 if (overlap > 0)
1247 m_adj(nfsd->nd_mrep, overlap);
1248 mp = owp->nd_mrep;
1249 while (mp->m_next)
1250 mp = mp->m_next;
1251 mp->m_next = nfsd->nd_mrep;
1252 owp->nd_eoff = nfsd->nd_eoff;
1253 } else
1254 m_freem(nfsd->nd_mrep);
1255 nfsd->nd_mrep = NULL;
1256 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1257 owp->nd_stable = NFSV3WRITE_FILESYNC;
1258 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1259 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1260 owp->nd_stable = NFSV3WRITE_DATASYNC;
1261 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1262 /*
1263 * nfsd might hold coalesce elements! Move them to owp.
1264 * Otherwise, requests may be lost and clients will be stuck.
1265 */
1266 if (nfsd->nd_coalesce.lh_first)
1267 {
1268 struct nfsrv_descript *m;
1269
1270 while ((m = nfsd->nd_coalesce.lh_first))
1271 {
1272 LIST_REMOVE(m, nd_tq);
1273 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1274 }
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, *mb2, *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, *mb2, *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, len, len2, fdirfor_ret = 1, fdiraft_ret = 1;
1734 int tdirfor_ret = 1, tdiraft_ret = 1;
1735 int v3 = (nfsd->nd_flag & ND_NFSV3);
1736 char *cp2;
1737 struct mbuf *mb, *mreq;
1738 struct nameidata fromnd, tond;
1739 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
1740 struct vnode *tdirp = (struct vnode *)0;
1741 struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1742 nfsfh_t fnfh, tnfh;
1743 fhandle_t *ffhp, *tfhp;
1744 u_quad_t frev;
1745 uid_t saved_uid;
1746
1747 #ifndef nolint
1748 fvp = (struct vnode *)0;
1749 #endif
1750 ffhp = &fnfh.fh_generic;
1751 tfhp = &tnfh.fh_generic;
1752 fromnd.ni_cnd.cn_nameiop = 0;
1753 tond.ni_cnd.cn_nameiop = 0;
1754 nfsm_srvmtofh(ffhp);
1755 nfsm_srvnamesiz(len);
1756 /*
1757 * Remember our original uid so that we can reset cr_uid before
1758 * the second nfs_namei() call, in case it is remapped.
1759 */
1760 saved_uid = cred->cr_uid;
1761 fromnd.ni_cnd.cn_cred = cred;
1762 fromnd.ni_cnd.cn_nameiop = DELETE;
1763 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART;
1764 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
1765 &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1766 if (fdirp) {
1767 if (v3)
1768 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred,
1769 procp);
1770 else {
1771 vrele(fdirp);
1772 fdirp = (struct vnode *)0;
1773 }
1774 }
1775 if (error) {
1776 nfsm_reply(2 * NFSX_WCCDATA(v3));
1777 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1778 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1779 if (fdirp)
1780 vrele(fdirp);
1781 return (0);
1782 }
1783 fvp = fromnd.ni_vp;
1784 nfsm_srvmtofh(tfhp);
1785 nfsm_strsiz(len2, NFS_MAXNAMLEN);
1786 cred->cr_uid = saved_uid;
1787 tond.ni_cnd.cn_cred = cred;
1788 tond.ni_cnd.cn_nameiop = RENAME;
1789 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
1790 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
1791 &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1792 if (tdirp) {
1793 if (v3)
1794 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred,
1795 procp);
1796 else {
1797 vrele(tdirp);
1798 tdirp = (struct vnode *)0;
1799 }
1800 }
1801 if (error) {
1802 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1803 vrele(fromnd.ni_dvp);
1804 vrele(fvp);
1805 goto out1;
1806 }
1807 tdvp = tond.ni_dvp;
1808 tvp = tond.ni_vp;
1809 if (tvp != NULL) {
1810 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1811 if (v3)
1812 error = EEXIST;
1813 else
1814 error = EISDIR;
1815 goto out;
1816 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1817 if (v3)
1818 error = EEXIST;
1819 else
1820 error = ENOTDIR;
1821 goto out;
1822 }
1823 if (tvp->v_type == VDIR && tvp->v_mountedhere) {
1824 if (v3)
1825 error = EXDEV;
1826 else
1827 error = ENOTEMPTY;
1828 goto out;
1829 }
1830 }
1831 if (fvp->v_type == VDIR && fvp->v_mountedhere) {
1832 if (v3)
1833 error = EXDEV;
1834 else
1835 error = ENOTEMPTY;
1836 goto out;
1837 }
1838 if (fvp->v_mount != tdvp->v_mount) {
1839 if (v3)
1840 error = EXDEV;
1841 else
1842 error = ENOTEMPTY;
1843 goto out;
1844 }
1845 if (fvp == tdvp) {
1846 if (v3)
1847 error = EINVAL;
1848 else
1849 error = ENOTEMPTY;
1850 }
1851 /*
1852 * If source is the same as the destination (that is the
1853 * same vnode with the same name in the same directory),
1854 * then there is nothing to do.
1855 */
1856 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
1857 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1858 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
1859 fromnd.ni_cnd.cn_namelen))
1860 error = -1;
1861 out:
1862 if (!error) {
1863 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
1864 nqsrv_getl(tdvp, ND_WRITE);
1865 if (tvp) {
1866 nqsrv_getl(tvp, ND_WRITE);
1867 }
1868 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
1869 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
1870 } else {
1871 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
1872 if (tdvp == tvp)
1873 vrele(tdvp);
1874 else
1875 vput(tdvp);
1876 if (tvp)
1877 vput(tvp);
1878 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1879 vrele(fromnd.ni_dvp);
1880 vrele(fvp);
1881 if (error == -1)
1882 error = 0;
1883 }
1884 vrele(tond.ni_startdir);
1885 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
1886 out1:
1887 if (fdirp) {
1888 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp);
1889 vrele(fdirp);
1890 }
1891 if (tdirp) {
1892 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp);
1893 vrele(tdirp);
1894 }
1895 vrele(fromnd.ni_startdir);
1896 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
1897 nfsm_reply(2 * NFSX_WCCDATA(v3));
1898 if (v3) {
1899 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1900 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1901 }
1902 return (0);
1903
1904 nfsmout:
1905 if (fdirp)
1906 vrele(fdirp);
1907 if (tdirp)
1908 vrele(tdirp);
1909 if (tond.ni_cnd.cn_nameiop) {
1910 vrele(tond.ni_startdir);
1911 PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
1912 }
1913 if (fromnd.ni_cnd.cn_nameiop) {
1914 vrele(fromnd.ni_startdir);
1915 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
1916 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1917 vrele(fromnd.ni_dvp);
1918 vrele(fvp);
1919 }
1920 return (error);
1921 }
1922
1923 /*
1924 * nfs link service
1925 */
1926 int
1927 nfsrv_link(nfsd, slp, procp, mrq)
1928 struct nfsrv_descript *nfsd;
1929 struct nfssvc_sock *slp;
1930 struct proc *procp;
1931 struct mbuf **mrq;
1932 {
1933 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1934 struct mbuf *nam = nfsd->nd_nam;
1935 caddr_t dpos = nfsd->nd_dpos;
1936 struct ucred *cred = &nfsd->nd_cr;
1937 struct nameidata nd;
1938 u_int32_t *tl;
1939 int32_t t1;
1940 caddr_t bpos;
1941 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1;
1942 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
1943 char *cp2;
1944 struct mbuf *mb, *mreq;
1945 struct vnode *vp, *xp, *dirp = (struct vnode *)0;
1946 struct vattr dirfor, diraft, at;
1947 nfsfh_t nfh, dnfh;
1948 fhandle_t *fhp, *dfhp;
1949 u_quad_t frev;
1950
1951 fhp = &nfh.fh_generic;
1952 dfhp = &dnfh.fh_generic;
1953 nfsm_srvmtofh(fhp);
1954 nfsm_srvmtofh(dfhp);
1955 nfsm_srvnamesiz(len);
1956 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
1957 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1958 if (error) {
1959 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
1960 nfsm_srvpostop_attr(getret, &at);
1961 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1962 return (0);
1963 }
1964 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0)
1965 goto out1;
1966 nd.ni_cnd.cn_cred = cred;
1967 nd.ni_cnd.cn_nameiop = CREATE;
1968 nd.ni_cnd.cn_flags = LOCKPARENT;
1969 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
1970 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1971 if (dirp) {
1972 if (v3)
1973 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1974 procp);
1975 else {
1976 vrele(dirp);
1977 dirp = (struct vnode *)0;
1978 }
1979 }
1980 if (error)
1981 goto out1;
1982 xp = nd.ni_vp;
1983 if (xp != NULL) {
1984 error = EEXIST;
1985 goto out;
1986 }
1987 xp = nd.ni_dvp;
1988 if (vp->v_mount != xp->v_mount)
1989 error = EXDEV;
1990 out:
1991 if (!error) {
1992 nqsrv_getl(vp, ND_WRITE);
1993 nqsrv_getl(xp, ND_WRITE);
1994 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
1995 } else {
1996 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1997 if (nd.ni_dvp == nd.ni_vp)
1998 vrele(nd.ni_dvp);
1999 else
2000 vput(nd.ni_dvp);
2001 if (nd.ni_vp)
2002 vrele(nd.ni_vp);
2003 }
2004 out1:
2005 if (v3)
2006 getret = VOP_GETATTR(vp, &at, cred, procp);
2007 if (dirp) {
2008 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2009 vrele(dirp);
2010 }
2011 vrele(vp);
2012 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2013 if (v3) {
2014 nfsm_srvpostop_attr(getret, &at);
2015 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2016 return (0);
2017 }
2018 nfsm_srvdone;
2019 }
2020
2021 /*
2022 * nfs symbolic link service
2023 */
2024 int
2025 nfsrv_symlink(nfsd, slp, procp, mrq)
2026 struct nfsrv_descript *nfsd;
2027 struct nfssvc_sock *slp;
2028 struct proc *procp;
2029 struct mbuf **mrq;
2030 {
2031 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2032 struct mbuf *nam = nfsd->nd_nam;
2033 caddr_t dpos = nfsd->nd_dpos;
2034 struct ucred *cred = &nfsd->nd_cr;
2035 struct vattr va, dirfor, diraft;
2036 struct nameidata nd;
2037 u_int32_t *tl;
2038 int32_t t1;
2039 struct nfsv2_sattr *sp;
2040 char *bpos, *pathcp = NULL, *cp2;
2041 struct uio io;
2042 struct iovec iv;
2043 int error = 0, cache, len, len2, dirfor_ret = 1, diraft_ret = 1;
2044 int v3 = (nfsd->nd_flag & ND_NFSV3);
2045 struct mbuf *mb, *mreq, *mb2;
2046 struct vnode *dirp = (struct vnode *)0;
2047 nfsfh_t nfh;
2048 fhandle_t *fhp;
2049 u_quad_t frev;
2050
2051 nd.ni_cnd.cn_nameiop = 0;
2052 fhp = &nfh.fh_generic;
2053 nfsm_srvmtofh(fhp);
2054 nfsm_srvnamesiz(len);
2055 nd.ni_cnd.cn_cred = cred;
2056 nd.ni_cnd.cn_nameiop = CREATE;
2057 nd.ni_cnd.cn_flags = LOCKPARENT;
2058 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2059 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2060 if (dirp) {
2061 if (v3)
2062 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2063 procp);
2064 else {
2065 vrele(dirp);
2066 dirp = (struct vnode *)0;
2067 }
2068 }
2069 if (error)
2070 goto out;
2071 VATTR_NULL(&va);
2072 if (v3)
2073 nfsm_srvsattr(&va);
2074 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2075 pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK);
2076 iv.iov_base = pathcp;
2077 iv.iov_len = len2;
2078 io.uio_resid = len2;
2079 io.uio_offset = 0;
2080 io.uio_iov = &iv;
2081 io.uio_iovcnt = 1;
2082 io.uio_segflg = UIO_SYSSPACE;
2083 io.uio_rw = UIO_READ;
2084 io.uio_procp = (struct proc *)0;
2085 nfsm_mtouio(&io, len2);
2086 if (!v3) {
2087 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2088 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2089 }
2090 *(pathcp + len2) = '\0';
2091 if (nd.ni_vp) {
2092 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2093 if (nd.ni_dvp == nd.ni_vp)
2094 vrele(nd.ni_dvp);
2095 else
2096 vput(nd.ni_dvp);
2097 vrele(nd.ni_vp);
2098 error = EEXIST;
2099 goto out;
2100 }
2101 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2102 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2103 if (!error) {
2104 if (v3) {
2105 memset((caddr_t)fhp, 0, sizeof(nfh));
2106 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
2107 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2108 if (!error)
2109 error = VOP_GETATTR(nd.ni_vp, &va, cred,
2110 procp);
2111 vput(nd.ni_vp);
2112 } else {
2113 vput(nd.ni_vp);
2114 }
2115 }
2116 out:
2117 if (pathcp)
2118 free(pathcp, M_TEMP);
2119 if (dirp) {
2120 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2121 vrele(dirp);
2122 }
2123 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2124 if (v3) {
2125 if (!error) {
2126 nfsm_srvpostop_fh(fhp);
2127 nfsm_srvpostop_attr(0, &va);
2128 }
2129 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2130 }
2131 return (0);
2132 nfsmout:
2133 if (dirp)
2134 vrele(dirp);
2135 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2136 if (nd.ni_dvp == nd.ni_vp)
2137 vrele(nd.ni_dvp);
2138 else
2139 vput(nd.ni_dvp);
2140 if (nd.ni_vp)
2141 vrele(nd.ni_vp);
2142 if (pathcp)
2143 free(pathcp, M_TEMP);
2144 return (error);
2145 }
2146
2147 /*
2148 * nfs mkdir service
2149 */
2150 int
2151 nfsrv_mkdir(nfsd, slp, procp, mrq)
2152 struct nfsrv_descript *nfsd;
2153 struct nfssvc_sock *slp;
2154 struct proc *procp;
2155 struct mbuf **mrq;
2156 {
2157 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2158 struct mbuf *nam = nfsd->nd_nam;
2159 caddr_t dpos = nfsd->nd_dpos;
2160 struct ucred *cred = &nfsd->nd_cr;
2161 struct vattr va, dirfor, diraft;
2162 struct nfs_fattr *fp;
2163 struct nameidata nd;
2164 caddr_t cp;
2165 u_int32_t *tl;
2166 int32_t t1;
2167 caddr_t bpos;
2168 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2169 int v3 = (nfsd->nd_flag & ND_NFSV3);
2170 char *cp2;
2171 struct mbuf *mb, *mb2, *mreq;
2172 struct vnode *vp, *dirp = (struct vnode *)0;
2173 nfsfh_t nfh;
2174 fhandle_t *fhp;
2175 u_quad_t frev;
2176
2177 fhp = &nfh.fh_generic;
2178 nfsm_srvmtofh(fhp);
2179 nfsm_srvnamesiz(len);
2180 nd.ni_cnd.cn_cred = cred;
2181 nd.ni_cnd.cn_nameiop = CREATE;
2182 nd.ni_cnd.cn_flags = LOCKPARENT;
2183 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2184 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2185 if (dirp) {
2186 if (v3)
2187 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2188 procp);
2189 else {
2190 vrele(dirp);
2191 dirp = (struct vnode *)0;
2192 }
2193 }
2194 if (error) {
2195 nfsm_reply(NFSX_WCCDATA(v3));
2196 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2197 if (dirp)
2198 vrele(dirp);
2199 return (0);
2200 }
2201 VATTR_NULL(&va);
2202 if (v3) {
2203 nfsm_srvsattr(&va);
2204 } else {
2205 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2206 va.va_mode = nfstov_mode(*tl++);
2207 }
2208 va.va_type = VDIR;
2209 vp = nd.ni_vp;
2210 if (vp != NULL) {
2211 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2212 if (nd.ni_dvp == vp)
2213 vrele(nd.ni_dvp);
2214 else
2215 vput(nd.ni_dvp);
2216 vrele(vp);
2217 error = EEXIST;
2218 goto out;
2219 }
2220 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2221 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2222 if (!error) {
2223 vp = nd.ni_vp;
2224 memset((caddr_t)fhp, 0, sizeof(nfh));
2225 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
2226 error = VFS_VPTOFH(vp, &fhp->fh_fid);
2227 if (!error)
2228 error = VOP_GETATTR(vp, &va, cred, procp);
2229 vput(vp);
2230 }
2231 out:
2232 if (dirp) {
2233 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2234 vrele(dirp);
2235 }
2236 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2237 if (v3) {
2238 if (!error) {
2239 nfsm_srvpostop_fh(fhp);
2240 nfsm_srvpostop_attr(0, &va);
2241 }
2242 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2243 } else {
2244 nfsm_srvfhtom(fhp, v3);
2245 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2246 nfsm_srvfillattr(&va, fp);
2247 }
2248 return (0);
2249 nfsmout:
2250 if (dirp)
2251 vrele(dirp);
2252 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2253 if (nd.ni_dvp == nd.ni_vp)
2254 vrele(nd.ni_dvp);
2255 else
2256 vput(nd.ni_dvp);
2257 if (nd.ni_vp)
2258 vrele(nd.ni_vp);
2259 return (error);
2260 }
2261
2262 /*
2263 * nfs rmdir service
2264 */
2265 int
2266 nfsrv_rmdir(nfsd, slp, procp, mrq)
2267 struct nfsrv_descript *nfsd;
2268 struct nfssvc_sock *slp;
2269 struct proc *procp;
2270 struct mbuf **mrq;
2271 {
2272 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2273 struct mbuf *nam = nfsd->nd_nam;
2274 caddr_t dpos = nfsd->nd_dpos;
2275 struct ucred *cred = &nfsd->nd_cr;
2276 u_int32_t *tl;
2277 int32_t t1;
2278 caddr_t bpos;
2279 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2280 int v3 = (nfsd->nd_flag & ND_NFSV3);
2281 char *cp2;
2282 struct mbuf *mb, *mreq;
2283 struct vnode *vp, *dirp = (struct vnode *)0;
2284 struct vattr dirfor, diraft;
2285 nfsfh_t nfh;
2286 fhandle_t *fhp;
2287 struct nameidata nd;
2288 u_quad_t frev;
2289
2290 fhp = &nfh.fh_generic;
2291 nfsm_srvmtofh(fhp);
2292 nfsm_srvnamesiz(len);
2293 nd.ni_cnd.cn_cred = cred;
2294 nd.ni_cnd.cn_nameiop = DELETE;
2295 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2296 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2297 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2298 if (dirp) {
2299 if (v3)
2300 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2301 procp);
2302 else {
2303 vrele(dirp);
2304 dirp = (struct vnode *)0;
2305 }
2306 }
2307 if (error) {
2308 nfsm_reply(NFSX_WCCDATA(v3));
2309 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2310 if (dirp)
2311 vrele(dirp);
2312 return (0);
2313 }
2314 vp = nd.ni_vp;
2315 if (vp->v_type != VDIR) {
2316 error = ENOTDIR;
2317 goto out;
2318 }
2319 /*
2320 * No rmdir "." please.
2321 */
2322 if (nd.ni_dvp == vp) {
2323 error = EINVAL;
2324 goto out;
2325 }
2326 /*
2327 * The root of a mounted filesystem cannot be deleted.
2328 */
2329 if (vp->v_flag & VROOT)
2330 error = EBUSY;
2331 out:
2332 if (!error) {
2333 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2334 nqsrv_getl(vp, ND_WRITE);
2335 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2336 } else {
2337 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2338 if (nd.ni_dvp == nd.ni_vp)
2339 vrele(nd.ni_dvp);
2340 else
2341 vput(nd.ni_dvp);
2342 vput(vp);
2343 }
2344 if (dirp) {
2345 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2346 vrele(dirp);
2347 }
2348 nfsm_reply(NFSX_WCCDATA(v3));
2349 if (v3) {
2350 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2351 return (0);
2352 }
2353 nfsm_srvdone;
2354 }
2355
2356 /*
2357 * nfs readdir service
2358 * - mallocs what it thinks is enough to read
2359 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR
2360 * - calls VOP_READDIR()
2361 * - loops around building the reply
2362 * if the output generated exceeds count break out of loop
2363 * The nfsm_clget macro is used here so that the reply will be packed
2364 * tightly in mbuf clusters.
2365 * - it only knows that it has encountered eof when the VOP_READDIR()
2366 * reads nothing
2367 * - as such one readdir rpc will return eof false although you are there
2368 * and then the next will return eof
2369 * - it trims out records with d_fileno == 0
2370 * this doesn't matter for Unix clients, but they might confuse clients
2371 * for other os'.
2372 * - it trims out records with d_type == DT_WHT
2373 * these cannot be seen through NFS (unless we extend the protocol)
2374 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2375 * than requested, but this may not apply to all filesystems. For
2376 * example, client NFS does not { although it is never remote mounted
2377 * anyhow }
2378 * The alternate call nfsrv_readdirplus() does lookups as well.
2379 * PS: The NFS protocol spec. does not clarify what the "count" byte
2380 * argument is a count of.. just name strings and file id's or the
2381 * entire reply rpc or ...
2382 * I tried just file name and id sizes and it confused the Sun client,
2383 * so I am using the full rpc size now. The "paranoia.." comment refers
2384 * to including the status longwords that are not a part of the dir.
2385 * "entry" structures, but are in the rpc.
2386 */
2387 struct flrep {
2388 nfsuint64 fl_off;
2389 u_int32_t fl_postopok;
2390 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2391 u_int32_t fl_fhok;
2392 u_int32_t fl_fhsize;
2393 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2394 };
2395
2396 int
2397 nfsrv_readdir(nfsd, slp, procp, mrq)
2398 struct nfsrv_descript *nfsd;
2399 struct nfssvc_sock *slp;
2400 struct proc *procp;
2401 struct mbuf **mrq;
2402 {
2403 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2404 struct mbuf *nam = nfsd->nd_nam;
2405 caddr_t dpos = nfsd->nd_dpos;
2406 struct ucred *cred = &nfsd->nd_cr;
2407 char *bp, *be;
2408 struct mbuf *mp;
2409 struct dirent *dp;
2410 caddr_t cp;
2411 u_int32_t *tl;
2412 int32_t t1;
2413 caddr_t bpos;
2414 struct mbuf *mb, *mb2, *mreq, *mp2;
2415 char *cpos, *cend, *cp2, *rbuf;
2416 struct vnode *vp;
2417 struct vattr at;
2418 nfsfh_t nfh;
2419 fhandle_t *fhp;
2420 struct uio io;
2421 struct iovec iv;
2422 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2423 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies;
2424 int v3 = (nfsd->nd_flag & ND_NFSV3);
2425 u_quad_t frev, off, toff, verf;
2426 off_t *cookies = NULL, *cookiep;
2427 nfsuint64 jar;
2428
2429 fhp = &nfh.fh_generic;
2430 nfsm_srvmtofh(fhp);
2431 if (v3) {
2432 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2433 toff = fxdr_hyper(tl);
2434 tl += 2;
2435 verf = fxdr_hyper(tl);
2436 tl += 2;
2437 } else {
2438 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2439 toff = fxdr_unsigned(u_quad_t, *tl++);
2440 }
2441 off = toff;
2442 cnt = fxdr_unsigned(int, *tl);
2443 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2444 xfer = NFS_SRVMAXDATA(nfsd);
2445 if (siz > xfer)
2446 siz = xfer;
2447 fullsiz = siz;
2448 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2449 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2450 if (!error && vp->v_type != VDIR) {
2451 error = ENOTDIR;
2452 vput(vp);
2453 }
2454 if (error) {
2455 nfsm_reply(NFSX_UNSIGNED);
2456 nfsm_srvpostop_attr(getret, &at);
2457 return (0);
2458 }
2459 nqsrv_getl(vp, ND_READ);
2460 if (v3) {
2461 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2462 #ifdef NFS3_STRICTVERF
2463 /*
2464 * XXX This check is too strict for Solaris 2.5 clients.
2465 */
2466 if (!error && toff && verf != at.va_filerev)
2467 error = NFSERR_BAD_COOKIE;
2468 #endif
2469 }
2470 if (!error)
2471 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2472 if (error) {
2473 vput(vp);
2474 nfsm_reply(NFSX_POSTOPATTR(v3));
2475 nfsm_srvpostop_attr(getret, &at);
2476 return (0);
2477 }
2478 VOP_UNLOCK(vp, 0);
2479 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2480 again:
2481 iv.iov_base = rbuf;
2482 iv.iov_len = fullsiz;
2483 io.uio_iov = &iv;
2484 io.uio_iovcnt = 1;
2485 io.uio_offset = (off_t)off;
2486 io.uio_resid = fullsiz;
2487 io.uio_segflg = UIO_SYSSPACE;
2488 io.uio_rw = UIO_READ;
2489 io.uio_procp = (struct proc *)0;
2490 eofflag = 0;
2491 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2492
2493 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2494
2495 off = (off_t)io.uio_offset;
2496 if (!cookies && !error)
2497 error = NFSERR_PERM;
2498 if (v3) {
2499 getret = VOP_GETATTR(vp, &at, cred, procp);
2500 if (!error)
2501 error = getret;
2502 }
2503
2504 VOP_UNLOCK(vp, 0);
2505 if (error) {
2506 vrele(vp);
2507 free((caddr_t)rbuf, M_TEMP);
2508 if (cookies)
2509 free((caddr_t)cookies, M_TEMP);
2510 nfsm_reply(NFSX_POSTOPATTR(v3));
2511 nfsm_srvpostop_attr(getret, &at);
2512 return (0);
2513 }
2514 if (io.uio_resid) {
2515 siz -= io.uio_resid;
2516
2517 /*
2518 * If nothing read, return eof
2519 * rpc reply
2520 */
2521 if (siz == 0) {
2522 vrele(vp);
2523 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2524 2 * NFSX_UNSIGNED);
2525 if (v3) {
2526 nfsm_srvpostop_attr(getret, &at);
2527 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2528 txdr_hyper(at.va_filerev, tl);
2529 tl += 2;
2530 } else
2531 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2532 *tl++ = nfs_false;
2533 *tl = nfs_true;
2534 free((caddr_t)rbuf, M_TEMP);
2535 free((caddr_t)cookies, M_TEMP);
2536 return (0);
2537 }
2538 }
2539
2540 /*
2541 * Check for degenerate cases of nothing useful read.
2542 * If so go try again
2543 */
2544 cpos = rbuf;
2545 cend = rbuf + siz;
2546 dp = (struct dirent *)cpos;
2547 cookiep = cookies;
2548
2549 while (cpos < cend && ncookies > 0 &&
2550 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2551 cpos += dp->d_reclen;
2552 dp = (struct dirent *)cpos;
2553 cookiep++;
2554 ncookies--;
2555 }
2556 if (cpos >= cend || ncookies == 0) {
2557 toff = off;
2558 siz = fullsiz;
2559 goto again;
2560 }
2561
2562 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
2563 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2564 if (v3) {
2565 nfsm_srvpostop_attr(getret, &at);
2566 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2567 txdr_hyper(at.va_filerev, tl);
2568 }
2569 mp = mp2 = mb;
2570 bp = bpos;
2571 be = bp + M_TRAILINGSPACE(mp);
2572
2573 /* Loop through the records and build reply */
2574 while (cpos < cend && ncookies > 0) {
2575 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2576 nlen = dp->d_namlen;
2577 rem = nfsm_rndup(nlen)-nlen;
2578 len += (4 * NFSX_UNSIGNED + nlen + rem);
2579 if (v3)
2580 len += 2 * NFSX_UNSIGNED;
2581 if (len > cnt) {
2582 eofflag = 0;
2583 break;
2584 }
2585 /*
2586 * Build the directory record xdr from
2587 * the dirent entry.
2588 */
2589 nfsm_clget;
2590 *tl = nfs_true;
2591 bp += NFSX_UNSIGNED;
2592 if (v3) {
2593 nfsm_clget;
2594 *tl = 0;
2595 bp += NFSX_UNSIGNED;
2596 }
2597 nfsm_clget;
2598 *tl = txdr_unsigned(dp->d_fileno);
2599 bp += NFSX_UNSIGNED;
2600 nfsm_clget;
2601 *tl = txdr_unsigned(nlen);
2602 bp += NFSX_UNSIGNED;
2603
2604 /* And loop around copying the name */
2605 xfer = nlen;
2606 cp = dp->d_name;
2607 while (xfer > 0) {
2608 nfsm_clget;
2609 if ((bp+xfer) > be)
2610 tsiz = be-bp;
2611 else
2612 tsiz = xfer;
2613 memcpy(bp, cp, tsiz);
2614 bp += tsiz;
2615 xfer -= tsiz;
2616 if (xfer > 0)
2617 cp += tsiz;
2618 }
2619 /* And null pad to an int32_t boundary */
2620 for (i = 0; i < rem; i++)
2621 *bp++ = '\0';
2622 nfsm_clget;
2623
2624 /* Finish off the record */
2625 txdr_hyper(*cookiep, &jar);
2626 if (v3) {
2627 *tl = jar.nfsuquad[0];
2628 bp += NFSX_UNSIGNED;
2629 nfsm_clget;
2630 }
2631 *tl = jar.nfsuquad[1];
2632 bp += NFSX_UNSIGNED;
2633 }
2634 cpos += dp->d_reclen;
2635 dp = (struct dirent *)cpos;
2636 cookiep++;
2637 ncookies--;
2638 }
2639 vrele(vp);
2640 nfsm_clget;
2641 *tl = nfs_false;
2642 bp += NFSX_UNSIGNED;
2643 nfsm_clget;
2644 if (eofflag)
2645 *tl = nfs_true;
2646 else
2647 *tl = nfs_false;
2648 bp += NFSX_UNSIGNED;
2649 if (mp != mb) {
2650 if (bp < be)
2651 mp->m_len = bp - mtod(mp, caddr_t);
2652 } else
2653 mp->m_len += bp - bpos;
2654 free((caddr_t)rbuf, M_TEMP);
2655 free((caddr_t)cookies, M_TEMP);
2656 nfsm_srvdone;
2657 }
2658
2659 int
2660 nfsrv_readdirplus(nfsd, slp, procp, mrq)
2661 struct nfsrv_descript *nfsd;
2662 struct nfssvc_sock *slp;
2663 struct proc *procp;
2664 struct mbuf **mrq;
2665 {
2666 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2667 struct mbuf *nam = nfsd->nd_nam;
2668 caddr_t dpos = nfsd->nd_dpos;
2669 struct ucred *cred = &nfsd->nd_cr;
2670 char *bp, *be;
2671 struct mbuf *mp;
2672 struct dirent *dp;
2673 caddr_t cp;
2674 u_int32_t *tl;
2675 int32_t t1;
2676 caddr_t bpos;
2677 struct mbuf *mb, *mb2, *mreq, *mp2;
2678 char *cpos, *cend, *cp2, *rbuf;
2679 struct vnode *vp, *nvp;
2680 struct flrep fl;
2681 nfsfh_t nfh;
2682 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
2683 struct uio io;
2684 struct iovec iv;
2685 struct vattr va, at, *vap = &va;
2686 struct nfs_fattr *fp;
2687 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2688 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies;
2689 u_quad_t frev, off, toff, verf;
2690 off_t *cookies = NULL, *cookiep;
2691
2692 fhp = &nfh.fh_generic;
2693 nfsm_srvmtofh(fhp);
2694 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2695 toff = fxdr_hyper(tl);
2696 tl += 2;
2697 verf = fxdr_hyper(tl);
2698 tl += 2;
2699 siz = fxdr_unsigned(int, *tl++);
2700 cnt = fxdr_unsigned(int, *tl);
2701 off = toff;
2702 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2703 xfer = NFS_SRVMAXDATA(nfsd);
2704 if (siz > xfer)
2705 siz = xfer;
2706 fullsiz = siz;
2707 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2708 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2709 if (!error && vp->v_type != VDIR) {
2710 error = ENOTDIR;
2711 vput(vp);
2712 }
2713 if (error) {
2714 nfsm_reply(NFSX_UNSIGNED);
2715 nfsm_srvpostop_attr(getret, &at);
2716 return (0);
2717 }
2718 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2719 #ifdef NFS3_STRICTVERF
2720 /*
2721 * XXX This check is too strict for Solaris 2.5 clients.
2722 */
2723 if (!error && toff && verf != at.va_filerev)
2724 error = NFSERR_BAD_COOKIE;
2725 #endif
2726 if (!error) {
2727 nqsrv_getl(vp, ND_READ);
2728 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2729 }
2730 if (error) {
2731 vput(vp);
2732 nfsm_reply(NFSX_V3POSTOPATTR);
2733 nfsm_srvpostop_attr(getret, &at);
2734 return (0);
2735 }
2736 VOP_UNLOCK(vp, 0);
2737
2738 rbuf = malloc(siz, M_TEMP, M_WAITOK);
2739 again:
2740 iv.iov_base = rbuf;
2741 iv.iov_len = fullsiz;
2742 io.uio_iov = &iv;
2743 io.uio_iovcnt = 1;
2744 io.uio_offset = (off_t)off;
2745 io.uio_resid = fullsiz;
2746 io.uio_segflg = UIO_SYSSPACE;
2747 io.uio_rw = UIO_READ;
2748 io.uio_procp = (struct proc *)0;
2749 eofflag = 0;
2750
2751 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2752
2753 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2754
2755 off = (u_quad_t)io.uio_offset;
2756 getret = VOP_GETATTR(vp, &at, cred, procp);
2757
2758 VOP_UNLOCK(vp, 0);
2759
2760 /*
2761 * If the VGET operation doesn't work for this filesystem,
2762 * we can't support readdirplus. Returning NOTSUPP should
2763 * make clients fall back to plain readdir.
2764 * There's no need to check for VPTOFH as well, we wouldn't
2765 * even be here otherwise.
2766 */
2767 if (!getret) {
2768 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
2769 getret = (getret == EOPNOTSUPP) ?
2770 NFSERR_NOTSUPP : NFSERR_IO;
2771 else
2772 vput(nvp);
2773 }
2774
2775 if (!cookies && !error)
2776 error = NFSERR_PERM;
2777 if (!error)
2778 error = getret;
2779 if (error) {
2780 vrele(vp);
2781 if (cookies)
2782 free((caddr_t)cookies, M_TEMP);
2783 free((caddr_t)rbuf, M_TEMP);
2784 nfsm_reply(NFSX_V3POSTOPATTR);
2785 nfsm_srvpostop_attr(getret, &at);
2786 return (0);
2787 }
2788 if (io.uio_resid) {
2789 siz -= io.uio_resid;
2790
2791 /*
2792 * If nothing read, return eof
2793 * rpc reply
2794 */
2795 if (siz == 0) {
2796 vrele(vp);
2797 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
2798 2 * NFSX_UNSIGNED);
2799 nfsm_srvpostop_attr(getret, &at);
2800 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2801 txdr_hyper(at.va_filerev, tl);
2802 tl += 2;
2803 *tl++ = nfs_false;
2804 *tl = nfs_true;
2805 free((caddr_t)cookies, M_TEMP);
2806 free((caddr_t)rbuf, M_TEMP);
2807 return (0);
2808 }
2809 }
2810
2811 /*
2812 * Check for degenerate cases of nothing useful read.
2813 * If so go try again
2814 */
2815 cpos = rbuf;
2816 cend = rbuf + siz;
2817 dp = (struct dirent *)cpos;
2818 cookiep = cookies;
2819
2820 while (cpos < cend && ncookies > 0 &&
2821 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2822 cpos += dp->d_reclen;
2823 dp = (struct dirent *)cpos;
2824 cookiep++;
2825 ncookies--;
2826 }
2827 if (cpos >= cend || ncookies == 0) {
2828 toff = off;
2829 siz = fullsiz;
2830 goto again;
2831 }
2832
2833 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
2834 nfsm_reply(cnt);
2835 nfsm_srvpostop_attr(getret, &at);
2836 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2837 txdr_hyper(at.va_filerev, tl);
2838 mp = mp2 = mb;
2839 bp = bpos;
2840 be = bp + M_TRAILINGSPACE(mp);
2841
2842 /* Loop through the records and build reply */
2843 while (cpos < cend && ncookies > 0) {
2844 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2845 nlen = dp->d_namlen;
2846 rem = nfsm_rndup(nlen)-nlen;
2847
2848 /*
2849 * For readdir_and_lookup get the vnode using
2850 * the file number.
2851 */
2852 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
2853 goto invalid;
2854 memset((caddr_t)nfhp, 0, NFSX_V3FH);
2855 nfhp->fh_fsid =
2856 nvp->v_mount->mnt_stat.f_fsid;
2857 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
2858 vput(nvp);
2859 goto invalid;
2860 }
2861 if (VOP_GETATTR(nvp, vap, cred, procp)) {
2862 vput(nvp);
2863 goto invalid;
2864 }
2865 vput(nvp);
2866
2867 /*
2868 * If either the dircount or maxcount will be
2869 * exceeded, get out now. Both of these lengths
2870 * are calculated conservatively, including all
2871 * XDR overheads.
2872 */
2873 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
2874 NFSX_V3POSTOPATTR);
2875 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
2876 if (len > cnt || dirlen > fullsiz) {
2877 eofflag = 0;
2878 break;
2879 }
2880
2881 /*
2882 * Build the directory record xdr from
2883 * the dirent entry.
2884 */
2885 fp = (struct nfs_fattr *)&fl.fl_fattr;
2886 nfsm_srvfillattr(vap, fp);
2887 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
2888 fl.fl_fhok = nfs_true;
2889 fl.fl_postopok = nfs_true;
2890 txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
2891
2892 nfsm_clget;
2893 *tl = nfs_true;
2894 bp += NFSX_UNSIGNED;
2895 nfsm_clget;
2896 *tl = 0;
2897 bp += NFSX_UNSIGNED;
2898 nfsm_clget;
2899 *tl = txdr_unsigned(dp->d_fileno);
2900 bp += NFSX_UNSIGNED;
2901 nfsm_clget;
2902 *tl = txdr_unsigned(nlen);
2903 bp += NFSX_UNSIGNED;
2904
2905 /* And loop around copying the name */
2906 xfer = nlen;
2907 cp = dp->d_name;
2908 while (xfer > 0) {
2909 nfsm_clget;
2910 if ((bp + xfer) > be)
2911 tsiz = be - bp;
2912 else
2913 tsiz = xfer;
2914 memcpy(bp, cp, tsiz);
2915 bp += tsiz;
2916 xfer -= tsiz;
2917 if (xfer > 0)
2918 cp += tsiz;
2919 }
2920 /* And null pad to an int32_t boundary */
2921 for (i = 0; i < rem; i++)
2922 *bp++ = '\0';
2923
2924 /*
2925 * Now copy the flrep structure out.
2926 */
2927 xfer = sizeof (struct flrep);
2928 cp = (caddr_t)&fl;
2929 while (xfer > 0) {
2930 nfsm_clget;
2931 if ((bp + xfer) > be)
2932 tsiz = be - bp;
2933 else
2934 tsiz = xfer;
2935 memcpy(bp, cp, tsiz);
2936 bp += tsiz;
2937 xfer -= tsiz;
2938 if (xfer > 0)
2939 cp += tsiz;
2940 }
2941 }
2942 invalid:
2943 cpos += dp->d_reclen;
2944 dp = (struct dirent *)cpos;
2945 cookiep++;
2946 ncookies--;
2947 }
2948 vrele(vp);
2949 nfsm_clget;
2950 *tl = nfs_false;
2951 bp += NFSX_UNSIGNED;
2952 nfsm_clget;
2953 if (eofflag)
2954 *tl = nfs_true;
2955 else
2956 *tl = nfs_false;
2957 bp += NFSX_UNSIGNED;
2958 if (mp != mb) {
2959 if (bp < be)
2960 mp->m_len = bp - mtod(mp, caddr_t);
2961 } else
2962 mp->m_len += bp - bpos;
2963 free((caddr_t)cookies, M_TEMP);
2964 free((caddr_t)rbuf, M_TEMP);
2965 nfsm_srvdone;
2966 }
2967
2968 /*
2969 * nfs commit service
2970 */
2971 int
2972 nfsrv_commit(nfsd, slp, procp, mrq)
2973 struct nfsrv_descript *nfsd;
2974 struct nfssvc_sock *slp;
2975 struct proc *procp;
2976 struct mbuf **mrq;
2977 {
2978 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2979 struct mbuf *nam = nfsd->nd_nam;
2980 caddr_t dpos = nfsd->nd_dpos;
2981 struct ucred *cred = &nfsd->nd_cr;
2982 struct vattr bfor, aft;
2983 struct vnode *vp;
2984 nfsfh_t nfh;
2985 fhandle_t *fhp;
2986 u_int32_t *tl;
2987 int32_t t1;
2988 caddr_t bpos;
2989 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache;
2990 char *cp2;
2991 struct mbuf *mb, *mb2, *mreq;
2992 u_quad_t frev, off, end;
2993
2994 #ifndef nolint
2995 cache = 0;
2996 #endif
2997 fhp = &nfh.fh_generic;
2998 nfsm_srvmtofh(fhp);
2999 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3000
3001 off = fxdr_hyper(tl);
3002 tl += 2;
3003 cnt = fxdr_unsigned(int, *tl);
3004 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3005 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3006 if (error) {
3007 nfsm_reply(2 * NFSX_UNSIGNED);
3008 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3009 return (0);
3010 }
3011 for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
3012 end = (cnt > 0) ? off + cnt : vp->v_size;
3013 if (end < off || end > vp->v_size)
3014 end = vp->v_size;
3015 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, procp);
3016 aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
3017 vput(vp);
3018 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3019 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3020 if (!error) {
3021 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3022 *tl++ = txdr_unsigned(boottime.tv_sec);
3023 *tl = txdr_unsigned(boottime.tv_usec);
3024 } else
3025 return (0);
3026 nfsm_srvdone;
3027 }
3028
3029 /*
3030 * nfs statfs service
3031 */
3032 int
3033 nfsrv_statfs(nfsd, slp, procp, mrq)
3034 struct nfsrv_descript *nfsd;
3035 struct nfssvc_sock *slp;
3036 struct proc *procp;
3037 struct mbuf **mrq;
3038 {
3039 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3040 struct mbuf *nam = nfsd->nd_nam;
3041 caddr_t dpos = nfsd->nd_dpos;
3042 struct ucred *cred = &nfsd->nd_cr;
3043 struct statfs *sf;
3044 struct nfs_statfs *sfp;
3045 u_int32_t *tl;
3046 int32_t t1;
3047 caddr_t bpos;
3048 int error = 0, rdonly, cache, getret = 1;
3049 int v3 = (nfsd->nd_flag & ND_NFSV3);
3050 char *cp2;
3051 struct mbuf *mb, *mb2, *mreq;
3052 struct vnode *vp;
3053 struct vattr at;
3054 nfsfh_t nfh;
3055 fhandle_t *fhp;
3056 struct statfs statfs;
3057 u_quad_t frev, tval;
3058
3059 #ifndef nolint
3060 cache = 0;
3061 #endif
3062 fhp = &nfh.fh_generic;
3063 nfsm_srvmtofh(fhp);
3064 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3065 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3066 if (error) {
3067 nfsm_reply(NFSX_UNSIGNED);
3068 nfsm_srvpostop_attr(getret, &at);
3069 return (0);
3070 }
3071 sf = &statfs;
3072 error = VFS_STATFS(vp->v_mount, sf, procp);
3073 getret = VOP_GETATTR(vp, &at, cred, procp);
3074 vput(vp);
3075 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3076 if (v3)
3077 nfsm_srvpostop_attr(getret, &at);
3078 if (error)
3079 return (0);
3080 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3081 if (v3) {
3082 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_bsize);
3083 txdr_hyper(tval, &sfp->sf_tbytes);
3084 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_bsize);
3085 txdr_hyper(tval, &sfp->sf_fbytes);
3086 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_bsize);
3087 txdr_hyper(tval, &sfp->sf_abytes);
3088 tval = (u_quad_t)sf->f_files;
3089 txdr_hyper(tval, &sfp->sf_tfiles);
3090 tval = (u_quad_t)sf->f_ffree;
3091 txdr_hyper(tval, &sfp->sf_ffiles);
3092 txdr_hyper(tval, &sfp->sf_afiles);
3093 sfp->sf_invarsec = 0;
3094 } else {
3095 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3096 sfp->sf_bsize = txdr_unsigned(sf->f_bsize);
3097 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3098 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3099 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3100 }
3101 nfsm_srvdone;
3102 }
3103
3104 /*
3105 * nfs fsinfo service
3106 */
3107 int
3108 nfsrv_fsinfo(nfsd, slp, procp, mrq)
3109 struct nfsrv_descript *nfsd;
3110 struct nfssvc_sock *slp;
3111 struct proc *procp;
3112 struct mbuf **mrq;
3113 {
3114 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3115 struct mbuf *nam = nfsd->nd_nam;
3116 caddr_t dpos = nfsd->nd_dpos;
3117 struct ucred *cred = &nfsd->nd_cr;
3118 u_int32_t *tl;
3119 struct nfsv3_fsinfo *sip;
3120 int32_t t1;
3121 caddr_t bpos;
3122 int error = 0, rdonly, cache, getret = 1, pref;
3123 char *cp2;
3124 struct mbuf *mb, *mb2, *mreq;
3125 struct vnode *vp;
3126 struct vattr at;
3127 nfsfh_t nfh;
3128 fhandle_t *fhp;
3129 u_quad_t frev, maxfsize;
3130 struct statfs sb;
3131
3132 #ifndef nolint
3133 cache = 0;
3134 #endif
3135 fhp = &nfh.fh_generic;
3136 nfsm_srvmtofh(fhp);
3137 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3138 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3139 if (error) {
3140 nfsm_reply(NFSX_UNSIGNED);
3141 nfsm_srvpostop_attr(getret, &at);
3142 return (0);
3143 }
3144
3145 /* XXX Try to make a guess on the max file size. */
3146 VFS_STATFS(vp->v_mount, &sb, (struct proc *)0);
3147 maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1;
3148
3149 getret = VOP_GETATTR(vp, &at, cred, procp);
3150 vput(vp);
3151 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3152 nfsm_srvpostop_attr(getret, &at);
3153 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3154
3155 /*
3156 * XXX
3157 * There should be file system VFS OP(s) to get this information.
3158 * For now, assume ufs.
3159 */
3160 if (slp->ns_so->so_type == SOCK_DGRAM)
3161 pref = NFS_MAXDGRAMDATA;
3162 else
3163 pref = NFS_MAXDATA;
3164 sip->fs_rtmax = txdr_unsigned(NFS_MAXDATA);
3165 sip->fs_rtpref = txdr_unsigned(pref);
3166 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3167 sip->fs_wtmax = txdr_unsigned(NFS_MAXDATA);
3168 sip->fs_wtpref = txdr_unsigned(pref);
3169 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3170 sip->fs_dtpref = txdr_unsigned(pref);
3171 txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3172 sip->fs_timedelta.nfsv3_sec = 0;
3173 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3174 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3175 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3176 NFSV3FSINFO_CANSETTIME);
3177 nfsm_srvdone;
3178 }
3179
3180 /*
3181 * nfs pathconf service
3182 */
3183 int
3184 nfsrv_pathconf(nfsd, slp, procp, mrq)
3185 struct nfsrv_descript *nfsd;
3186 struct nfssvc_sock *slp;
3187 struct proc *procp;
3188 struct mbuf **mrq;
3189 {
3190 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3191 struct mbuf *nam = nfsd->nd_nam;
3192 caddr_t dpos = nfsd->nd_dpos;
3193 struct ucred *cred = &nfsd->nd_cr;
3194 u_int32_t *tl;
3195 struct nfsv3_pathconf *pc;
3196 int32_t t1;
3197 caddr_t bpos;
3198 int error = 0, rdonly, cache, getret = 1;
3199 register_t linkmax, namemax, chownres, notrunc;
3200 char *cp2;
3201 struct mbuf *mb, *mb2, *mreq;
3202 struct vnode *vp;
3203 struct vattr at;
3204 nfsfh_t nfh;
3205 fhandle_t *fhp;
3206 u_quad_t frev;
3207
3208 #ifndef nolint
3209 cache = 0;
3210 #endif
3211 fhp = &nfh.fh_generic;
3212 nfsm_srvmtofh(fhp);
3213 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3214 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3215 if (error) {
3216 nfsm_reply(NFSX_UNSIGNED);
3217 nfsm_srvpostop_attr(getret, &at);
3218 return (0);
3219 }
3220 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3221 if (!error)
3222 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3223 if (!error)
3224 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3225 if (!error)
3226 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc);
3227 getret = VOP_GETATTR(vp, &at, cred, procp);
3228 vput(vp);
3229 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3230 nfsm_srvpostop_attr(getret, &at);
3231 if (error)
3232 return (0);
3233 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3234
3235 pc->pc_linkmax = txdr_unsigned(linkmax);
3236 pc->pc_namemax = txdr_unsigned(namemax);
3237 pc->pc_notrunc = txdr_unsigned(notrunc);
3238 pc->pc_chownrestricted = txdr_unsigned(chownres);
3239
3240 /*
3241 * These should probably be supported by VOP_PATHCONF(), but
3242 * until msdosfs is exportable (why would you want to?), the
3243 * Unix defaults should be ok.
3244 */
3245 pc->pc_caseinsensitive = nfs_false;
3246 pc->pc_casepreserving = nfs_true;
3247 nfsm_srvdone;
3248 }
3249
3250 /*
3251 * Null operation, used by clients to ping server
3252 */
3253 /* ARGSUSED */
3254 int
3255 nfsrv_null(nfsd, slp, procp, mrq)
3256 struct nfsrv_descript *nfsd;
3257 struct nfssvc_sock *slp;
3258 struct proc *procp;
3259 struct mbuf **mrq;
3260 {
3261 struct mbuf *mrep = nfsd->nd_mrep;
3262 caddr_t bpos;
3263 int error = NFSERR_RETVOID, cache = 0;
3264 struct mbuf *mb, *mreq;
3265 u_quad_t frev;
3266
3267 nfsm_reply(0);
3268 return (0);
3269 }
3270
3271 /*
3272 * No operation, used for obsolete procedures
3273 */
3274 /* ARGSUSED */
3275 int
3276 nfsrv_noop(nfsd, slp, procp, mrq)
3277 struct nfsrv_descript *nfsd;
3278 struct nfssvc_sock *slp;
3279 struct proc *procp;
3280 struct mbuf **mrq;
3281 {
3282 struct mbuf *mrep = nfsd->nd_mrep;
3283 caddr_t bpos;
3284 int error, cache = 0;
3285 struct mbuf *mb, *mreq;
3286 u_quad_t frev;
3287
3288 if (nfsd->nd_repstat)
3289 error = nfsd->nd_repstat;
3290 else
3291 error = EPROCUNAVAIL;
3292 nfsm_reply(0);
3293 return (0);
3294 }
3295
3296 /*
3297 * Perform access checking for vnodes obtained from file handles that would
3298 * refer to files already opened by a Unix client. You cannot just use
3299 * vn_writechk() and VOP_ACCESS() for two reasons.
3300 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3301 * 2 - The owner is to be given access irrespective of mode bits for some
3302 * operations, so that processes that chmod after opening a file don't
3303 * break. I don't like this because it opens a security hole, but since
3304 * the nfs server opens a security hole the size of a barn door anyhow,
3305 * what the heck.
3306 *
3307 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3308 * will return EPERM instead of EACCESS. EPERM is always an error.
3309 */
3310 int
3311 nfsrv_access(vp, flags, cred, rdonly, p, override)
3312 struct vnode *vp;
3313 int flags;
3314 struct ucred *cred;
3315 int rdonly;
3316 struct proc *p;
3317 {
3318 struct vattr vattr;
3319 int error;
3320 if (flags & VWRITE) {
3321 /* Just vn_writechk() changed to check rdonly */
3322 /*
3323 * Disallow write attempts on read-only file systems;
3324 * unless the file is a socket or a block or character
3325 * device resident on the file system.
3326 */
3327 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3328 switch (vp->v_type) {
3329 case VREG:
3330 case VDIR:
3331 case VLNK:
3332 return (EROFS);
3333 default:
3334 break;
3335 }
3336 }
3337
3338 /*
3339 * If the vnode is in use as a process's text,
3340 * we can't allow writing.
3341 */
3342 if (vp->v_flag & VTEXT)
3343 return (ETXTBSY);
3344 }
3345 error = VOP_GETATTR(vp, &vattr, cred, p);
3346 if (error)
3347 return (error);
3348 error = VOP_ACCESS(vp, flags, cred, p);
3349 /*
3350 * Allow certain operations for the owner (reads and writes
3351 * on files that are already open).
3352 */
3353 if (override && error == EACCES && cred->cr_uid == vattr.va_uid)
3354 error = 0;
3355 return error;
3356 }
3357