nfs_serv.c revision 1.40 1 /* $NetBSD: nfs_serv.c,v 1.40 1998/02/05 08:00:21 mrg Exp $ */
2
3 /*
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Rick Macklem at The University of Guelph.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. 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 8.7 (Berkeley) 5/14/95
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/param.h>
62 #include <sys/systm.h>
63 #include <sys/proc.h>
64 #include <sys/file.h>
65 #include <sys/namei.h>
66 #include <sys/vnode.h>
67 #include <sys/mount.h>
68 #include <sys/socket.h>
69 #include <sys/socketvar.h>
70 #include <sys/mbuf.h>
71 #include <sys/dirent.h>
72 #include <sys/stat.h>
73 #include <sys/kernel.h>
74 #include <ufs/ufs/dir.h>
75
76 #include <vm/vm.h>
77
78 #if defined(UVM)
79 #include <uvm/uvm_extern.h>
80 #endif
81
82 #include <nfs/nfsproto.h>
83 #include <nfs/rpcv2.h>
84 #include <nfs/nfs.h>
85 #include <nfs/xdr_subs.h>
86 #include <nfs/nfsm_subs.h>
87 #include <nfs/nqnfs.h>
88 #include <nfs/nfs_var.h>
89
90 /* Global vars */
91 extern u_int32_t nfs_xdrneg1;
92 extern u_int32_t nfs_false, nfs_true;
93 extern enum vtype nv3tov_type[8];
94 extern struct nfsstats nfsstats;
95 extern nfstype nfsv2_type[9];
96 extern nfstype nfsv3_type[9];
97 extern struct nfs_public nfs_pub;
98 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
99
100 /*
101 * nfs v3 access service
102 */
103 int
104 nfsrv3_access(nfsd, slp, procp, mrq)
105 struct nfsrv_descript *nfsd;
106 struct nfssvc_sock *slp;
107 struct proc *procp;
108 struct mbuf **mrq;
109 {
110 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
111 struct mbuf *nam = nfsd->nd_nam;
112 caddr_t dpos = nfsd->nd_dpos;
113 struct ucred *cred = &nfsd->nd_cr;
114 struct vnode *vp;
115 nfsfh_t nfh;
116 fhandle_t *fhp;
117 register u_int32_t *tl;
118 register int32_t t1;
119 caddr_t bpos;
120 int error = 0, rdonly, cache = 0, getret;
121 char *cp2;
122 struct mbuf *mb, *mreq, *mb2;
123 struct vattr va;
124 u_long inmode, testmode, outmode;
125 u_quad_t frev;
126
127 fhp = &nfh.fh_generic;
128 nfsm_srvmtofh(fhp);
129 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
130 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
131 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
132 if (error) {
133 nfsm_reply(NFSX_UNSIGNED);
134 nfsm_srvpostop_attr(1, (struct vattr *)0);
135 return (0);
136 }
137 inmode = fxdr_unsigned(u_int32_t, *tl);
138 outmode = 0;
139 if ((inmode & NFSV3ACCESS_READ) &&
140 nfsrv_access(vp, VREAD, cred, rdonly, procp, 0) == 0)
141 outmode |= NFSV3ACCESS_READ;
142 if (vp->v_type != VDIR) {
143 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
144 if (testmode &&
145 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0)
146 outmode |= testmode;
147 if ((inmode & NFSV3ACCESS_EXECUTE) &&
148 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
149 outmode |= NFSV3ACCESS_EXECUTE;
150 } else {
151 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
152 NFSV3ACCESS_DELETE);
153 if (testmode &&
154 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0)
155 outmode |= testmode;
156 if ((inmode & NFSV3ACCESS_LOOKUP) &&
157 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
158 outmode |= NFSV3ACCESS_LOOKUP;
159 }
160 getret = VOP_GETATTR(vp, &va, cred, procp);
161 vput(vp);
162 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
163 nfsm_srvpostop_attr(getret, &va);
164 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
165 *tl = txdr_unsigned(outmode);
166 nfsm_srvdone;
167 }
168
169 /*
170 * nfs getattr service
171 */
172 int
173 nfsrv_getattr(nfsd, slp, procp, mrq)
174 struct nfsrv_descript *nfsd;
175 struct nfssvc_sock *slp;
176 struct proc *procp;
177 struct mbuf **mrq;
178 {
179 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
180 struct mbuf *nam = nfsd->nd_nam;
181 caddr_t dpos = nfsd->nd_dpos;
182 struct ucred *cred = &nfsd->nd_cr;
183 register struct nfs_fattr *fp;
184 struct vattr va;
185 struct vnode *vp;
186 nfsfh_t nfh;
187 fhandle_t *fhp;
188 register u_int32_t *tl;
189 register int32_t t1;
190 caddr_t bpos;
191 int error = 0, rdonly, cache;
192 char *cp2;
193 struct mbuf *mb, *mb2, *mreq;
194 u_quad_t frev;
195
196 fhp = &nfh.fh_generic;
197 nfsm_srvmtofh(fhp);
198 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
199 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
200 if (error) {
201 nfsm_reply(0);
202 return (0);
203 }
204 nqsrv_getl(vp, ND_READ);
205 error = VOP_GETATTR(vp, &va, cred, procp);
206 vput(vp);
207 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
208 if (error)
209 return (0);
210 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
211 nfsm_srvfillattr(&va, fp);
212 nfsm_srvdone;
213 }
214
215 /*
216 * nfs setattr service
217 */
218 int
219 nfsrv_setattr(nfsd, slp, procp, mrq)
220 struct nfsrv_descript *nfsd;
221 struct nfssvc_sock *slp;
222 struct proc *procp;
223 struct mbuf **mrq;
224 {
225 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
226 struct mbuf *nam = nfsd->nd_nam;
227 caddr_t dpos = nfsd->nd_dpos;
228 struct ucred *cred = &nfsd->nd_cr;
229 struct vattr va, preat;
230 register struct nfsv2_sattr *sp;
231 register struct nfs_fattr *fp;
232 struct vnode *vp;
233 nfsfh_t nfh;
234 fhandle_t *fhp;
235 register u_int32_t *tl;
236 register int32_t t1;
237 caddr_t bpos;
238 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1;
239 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
240 char *cp2;
241 struct mbuf *mb, *mb2, *mreq;
242 u_quad_t frev;
243 struct timespec guard;
244
245 fhp = &nfh.fh_generic;
246 nfsm_srvmtofh(fhp);
247 VATTR_NULL(&va);
248 if (v3) {
249 nfsm_srvsattr(&va);
250 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
251 gcheck = fxdr_unsigned(int, *tl);
252 if (gcheck) {
253 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
254 fxdr_nfsv3time(tl, &guard);
255 }
256 } else {
257 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
258 /*
259 * Nah nah nah nah na nah
260 * There is a bug in the Sun client that puts 0xffff in the mode
261 * field of sattr when it should put in 0xffffffff. The u_short
262 * doesn't sign extend.
263 * --> check the low order 2 bytes for 0xffff
264 */
265 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
266 va.va_mode = nfstov_mode(sp->sa_mode);
267 if (sp->sa_uid != nfs_xdrneg1)
268 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
269 if (sp->sa_gid != nfs_xdrneg1)
270 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
271 if (sp->sa_size != nfs_xdrneg1)
272 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
273 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
274 #ifdef notyet
275 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime);
276 #else
277 va.va_atime.tv_sec =
278 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
279 va.va_atime.tv_nsec = 0;
280 #endif
281 }
282 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
283 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime);
284
285 }
286
287 /*
288 * Now that we have all the fields, lets do it.
289 */
290 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
291 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
292 if (error) {
293 nfsm_reply(2 * NFSX_UNSIGNED);
294 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
295 return (0);
296 }
297 nqsrv_getl(vp, ND_WRITE);
298 if (v3) {
299 error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp);
300 if (!error && gcheck &&
301 (preat.va_ctime.tv_sec != guard.tv_sec ||
302 preat.va_ctime.tv_nsec != guard.tv_nsec))
303 error = NFSERR_NOT_SYNC;
304 if (error) {
305 vput(vp);
306 nfsm_reply(NFSX_WCCDATA(v3));
307 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
308 return (0);
309 }
310 }
311
312 /*
313 * If the size is being changed write acces is required, otherwise
314 * just check for a read only file system.
315 */
316 if (va.va_size == ((u_quad_t)((quad_t) -1))) {
317 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
318 error = EROFS;
319 goto out;
320 }
321 } else {
322 if (vp->v_type == VDIR) {
323 error = EISDIR;
324 goto out;
325 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
326 procp, 0)) != 0)
327 goto out;
328 }
329 error = VOP_SETATTR(vp, &va, cred, procp);
330 postat_ret = VOP_GETATTR(vp, &va, cred, procp);
331 if (!error)
332 error = postat_ret;
333 out:
334 vput(vp);
335 nfsm_reply(NFSX_WCCORFATTR(v3));
336 if (v3) {
337 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
338 return (0);
339 } else {
340 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
341 nfsm_srvfillattr(&va, fp);
342 }
343 nfsm_srvdone;
344 }
345
346 /*
347 * nfs lookup rpc
348 */
349 int
350 nfsrv_lookup(nfsd, slp, procp, mrq)
351 struct nfsrv_descript *nfsd;
352 struct nfssvc_sock *slp;
353 struct proc *procp;
354 struct mbuf **mrq;
355 {
356 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
357 struct mbuf *nam = nfsd->nd_nam;
358 caddr_t dpos = nfsd->nd_dpos;
359 struct ucred *cred = &nfsd->nd_cr;
360 register struct nfs_fattr *fp;
361 struct nameidata nd, ind, *ndp = &nd;
362 struct vnode *vp, *dirp;
363 nfsfh_t nfh;
364 fhandle_t *fhp;
365 register caddr_t cp;
366 register u_int32_t *tl;
367 register int32_t t1;
368 caddr_t bpos;
369 int error = 0, cache, len, dirattr_ret = 1;
370 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
371 char *cp2;
372 struct mbuf *mb, *mb2, *mreq;
373 struct vattr va, dirattr;
374 u_quad_t frev;
375
376 fhp = &nfh.fh_generic;
377 nfsm_srvmtofh(fhp);
378 nfsm_srvnamesiz(len);
379
380 pubflag = nfs_ispublicfh(fhp);
381
382 nd.ni_cnd.cn_cred = cred;
383 nd.ni_cnd.cn_nameiop = LOOKUP;
384 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
385 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
386 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
387
388 if (!error && pubflag) {
389 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
390 /*
391 * Setup call to lookup() to see if we can find
392 * the index file. Arguably, this doesn't belong
393 * in a kernel.. Ugh.
394 */
395 ind = nd;
396 VOP_UNLOCK(nd.ni_vp);
397 ind.ni_pathlen = strlen(nfs_pub.np_index);
398 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
399 nfs_pub.np_index;
400 ind.ni_startdir = nd.ni_vp;
401 VREF(ind.ni_startdir);
402 error = lookup(&ind);
403 if (!error) {
404 /*
405 * Found an index file. Get rid of
406 * the old references.
407 */
408 if (dirp)
409 vrele(dirp);
410 dirp = nd.ni_vp;
411 vrele(nd.ni_startdir);
412 ndp = &ind;
413 } else
414 error = 0;
415 }
416 /*
417 * If the public filehandle was used, check that this lookup
418 * didn't result in a filehandle outside the publicly exported
419 * filesystem.
420 */
421
422 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
423 vput(nd.ni_vp);
424 error = EPERM;
425 }
426 }
427
428 if (dirp) {
429 if (v3)
430 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred,
431 procp);
432 vrele(dirp);
433 }
434
435 if (error) {
436 nfsm_reply(NFSX_POSTOPATTR(v3));
437 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
438 return (0);
439 }
440
441 nqsrv_getl(ndp->ni_startdir, ND_READ);
442 vrele(ndp->ni_startdir);
443 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
444 vp = ndp->ni_vp;
445 bzero((caddr_t)fhp, sizeof(nfh));
446 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
447 error = VFS_VPTOFH(vp, &fhp->fh_fid);
448 if (!error)
449 error = VOP_GETATTR(vp, &va, cred, procp);
450 vput(vp);
451 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
452 if (error) {
453 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
454 return (0);
455 }
456 nfsm_srvfhtom(fhp, v3);
457 if (v3) {
458 nfsm_srvpostop_attr(0, &va);
459 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
460 } else {
461 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
462 nfsm_srvfillattr(&va, fp);
463 }
464 nfsm_srvdone;
465 }
466
467 /*
468 * nfs readlink service
469 */
470 int
471 nfsrv_readlink(nfsd, slp, procp, mrq)
472 struct nfsrv_descript *nfsd;
473 struct nfssvc_sock *slp;
474 struct proc *procp;
475 struct mbuf **mrq;
476 {
477 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
478 struct mbuf *nam = nfsd->nd_nam;
479 caddr_t dpos = nfsd->nd_dpos;
480 struct ucred *cred = &nfsd->nd_cr;
481 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
482 register struct iovec *ivp = iv;
483 register struct mbuf *mp;
484 register u_int32_t *tl;
485 register int32_t t1;
486 caddr_t bpos;
487 int error = 0, rdonly, cache, i, tlen, len, getret;
488 int v3 = (nfsd->nd_flag & ND_NFSV3);
489 char *cp2;
490 struct mbuf *mb, *mb2, *mp2 = NULL, *mp3 = NULL, *mreq;
491 struct vnode *vp;
492 struct vattr attr;
493 nfsfh_t nfh;
494 fhandle_t *fhp;
495 struct uio io, *uiop = &io;
496 u_quad_t frev;
497
498 fhp = &nfh.fh_generic;
499 nfsm_srvmtofh(fhp);
500 len = 0;
501 i = 0;
502 while (len < NFS_MAXPATHLEN) {
503 MGET(mp, M_WAIT, MT_DATA);
504 MCLGET(mp, M_WAIT);
505 mp->m_len = NFSMSIZ(mp);
506 if (len == 0)
507 mp3 = mp2 = mp;
508 else {
509 mp2->m_next = mp;
510 mp2 = mp;
511 }
512 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
513 mp->m_len = NFS_MAXPATHLEN-len;
514 len = NFS_MAXPATHLEN;
515 } else
516 len += mp->m_len;
517 ivp->iov_base = mtod(mp, caddr_t);
518 ivp->iov_len = mp->m_len;
519 i++;
520 ivp++;
521 }
522 uiop->uio_iov = iv;
523 uiop->uio_iovcnt = i;
524 uiop->uio_offset = 0;
525 uiop->uio_resid = len;
526 uiop->uio_rw = UIO_READ;
527 uiop->uio_segflg = UIO_SYSSPACE;
528 uiop->uio_procp = (struct proc *)0;
529 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
530 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
531 if (error) {
532 m_freem(mp3);
533 nfsm_reply(2 * NFSX_UNSIGNED);
534 nfsm_srvpostop_attr(1, (struct vattr *)0);
535 return (0);
536 }
537 if (vp->v_type != VLNK) {
538 if (v3)
539 error = EINVAL;
540 else
541 error = ENXIO;
542 goto out;
543 }
544 nqsrv_getl(vp, ND_READ);
545 error = VOP_READLINK(vp, uiop, cred);
546 out:
547 getret = VOP_GETATTR(vp, &attr, cred, procp);
548 vput(vp);
549 if (error)
550 m_freem(mp3);
551 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
552 if (v3) {
553 nfsm_srvpostop_attr(getret, &attr);
554 if (error)
555 return (0);
556 }
557 if (uiop->uio_resid > 0) {
558 len -= uiop->uio_resid;
559 tlen = nfsm_rndup(len);
560 nfsm_adj(mp3, NFS_MAXPATHLEN-tlen, tlen-len);
561 }
562 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
563 *tl = txdr_unsigned(len);
564 mb->m_next = mp3;
565 nfsm_srvdone;
566 }
567
568 /*
569 * nfs read service
570 */
571 int
572 nfsrv_read(nfsd, slp, procp, mrq)
573 struct nfsrv_descript *nfsd;
574 struct nfssvc_sock *slp;
575 struct proc *procp;
576 struct mbuf **mrq;
577 {
578 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
579 struct mbuf *nam = nfsd->nd_nam;
580 caddr_t dpos = nfsd->nd_dpos;
581 struct ucred *cred = &nfsd->nd_cr;
582 register struct iovec *iv;
583 struct iovec *iv2;
584 register struct mbuf *m;
585 register struct nfs_fattr *fp;
586 register u_int32_t *tl;
587 register int32_t t1;
588 register int i;
589 caddr_t bpos;
590 int error = 0, rdonly, cache, cnt, len, left, siz, tlen, getret;
591 int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen;
592 char *cp2;
593 struct mbuf *mb, *mb2, *mreq;
594 struct mbuf *m2;
595 struct vnode *vp;
596 nfsfh_t nfh;
597 fhandle_t *fhp;
598 struct uio io, *uiop = &io;
599 struct vattr va;
600 off_t off;
601 u_quad_t frev;
602
603 fhp = &nfh.fh_generic;
604 nfsm_srvmtofh(fhp);
605 if (v3) {
606 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
607 fxdr_hyper(tl, &off);
608 } else {
609 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
610 off = (off_t)fxdr_unsigned(u_int32_t, *tl);
611 }
612 nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd));
613 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
614 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
615 if (error) {
616 nfsm_reply(2 * NFSX_UNSIGNED);
617 nfsm_srvpostop_attr(1, (struct vattr *)0);
618 return (0);
619 }
620 if (vp->v_type != VREG) {
621 if (v3)
622 error = EINVAL;
623 else
624 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
625 }
626 if (!error) {
627 nqsrv_getl(vp, ND_READ);
628 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0)
629 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1);
630 }
631 getret = VOP_GETATTR(vp, &va, cred, procp);
632 if (!error)
633 error = getret;
634 if (error) {
635 vput(vp);
636 nfsm_reply(NFSX_POSTOPATTR(v3));
637 nfsm_srvpostop_attr(getret, &va);
638 return (0);
639 }
640 if (off >= va.va_size)
641 cnt = 0;
642 else if ((off + reqlen) > va.va_size)
643 cnt = nfsm_rndup(va.va_size - off);
644 else
645 cnt = reqlen;
646 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
647 if (v3) {
648 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
649 *tl++ = nfs_true;
650 fp = (struct nfs_fattr *)tl;
651 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
652 } else {
653 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
654 fp = (struct nfs_fattr *)tl;
655 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
656 }
657 len = left = cnt;
658 if (cnt > 0) {
659 /*
660 * Generate the mbuf list with the uio_iov ref. to it.
661 */
662 i = 0;
663 m = m2 = mb;
664 while (left > 0) {
665 siz = min(M_TRAILINGSPACE(m), left);
666 if (siz > 0) {
667 left -= siz;
668 i++;
669 }
670 if (left > 0) {
671 MGET(m, M_WAIT, MT_DATA);
672 MCLGET(m, M_WAIT);
673 m->m_len = 0;
674 m2->m_next = m;
675 m2 = m;
676 }
677 }
678 MALLOC(iv, struct iovec *, i * sizeof (struct iovec),
679 M_TEMP, M_WAITOK);
680 uiop->uio_iov = iv2 = iv;
681 m = mb;
682 left = cnt;
683 i = 0;
684 while (left > 0) {
685 if (m == NULL)
686 panic("nfsrv_read iov");
687 siz = min(M_TRAILINGSPACE(m), left);
688 if (siz > 0) {
689 iv->iov_base = mtod(m, caddr_t) + m->m_len;
690 iv->iov_len = siz;
691 m->m_len += siz;
692 left -= siz;
693 iv++;
694 i++;
695 }
696 m = m->m_next;
697 }
698 uiop->uio_iovcnt = i;
699 uiop->uio_offset = off;
700 uiop->uio_resid = cnt;
701 uiop->uio_rw = UIO_READ;
702 uiop->uio_segflg = UIO_SYSSPACE;
703 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
704 off = uiop->uio_offset;
705 FREE((caddr_t)iv2, M_TEMP);
706 if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){
707 if (!error)
708 error = getret;
709 m_freem(mreq);
710 vput(vp);
711 nfsm_reply(NFSX_POSTOPATTR(v3));
712 nfsm_srvpostop_attr(getret, &va);
713 return (0);
714 }
715 } else
716 uiop->uio_resid = 0;
717 vput(vp);
718 nfsm_srvfillattr(&va, fp);
719 len -= uiop->uio_resid;
720 tlen = nfsm_rndup(len);
721 if (cnt != tlen || tlen != len)
722 nfsm_adj(mb, cnt - tlen, tlen - len);
723 if (v3) {
724 *tl++ = txdr_unsigned(len);
725 if (len < reqlen)
726 *tl++ = nfs_true;
727 else
728 *tl++ = nfs_false;
729 }
730 *tl = txdr_unsigned(len);
731 nfsm_srvdone;
732 }
733
734 /*
735 * nfs write service
736 */
737 int
738 nfsrv_write(nfsd, slp, procp, mrq)
739 struct nfsrv_descript *nfsd;
740 struct nfssvc_sock *slp;
741 struct proc *procp;
742 struct mbuf **mrq;
743 {
744 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
745 struct mbuf *nam = nfsd->nd_nam;
746 caddr_t dpos = nfsd->nd_dpos;
747 struct ucred *cred = &nfsd->nd_cr;
748 register struct iovec *ivp;
749 register int i, cnt;
750 register struct mbuf *mp;
751 register struct nfs_fattr *fp;
752 struct iovec *iv;
753 struct vattr va, forat;
754 register u_int32_t *tl;
755 register int32_t t1;
756 caddr_t bpos;
757 int error = 0, rdonly, cache, len, forat_ret = 1;
758 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
759 int stable = NFSV3WRITE_FILESYNC;
760 int v3 = (nfsd->nd_flag & ND_NFSV3);
761 char *cp2;
762 struct mbuf *mb, *mb2, *mreq;
763 struct vnode *vp;
764 nfsfh_t nfh;
765 fhandle_t *fhp;
766 struct uio io, *uiop = &io;
767 off_t off;
768 u_quad_t frev;
769
770 if (mrep == NULL) {
771 *mrq = NULL;
772 return (0);
773 }
774 fhp = &nfh.fh_generic;
775 nfsm_srvmtofh(fhp);
776 if (v3) {
777 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
778 fxdr_hyper(tl, &off);
779 tl += 3;
780 stable = fxdr_unsigned(int, *tl++);
781 } else {
782 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
783 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
784 tl += 2;
785 }
786 retlen = len = fxdr_unsigned(int32_t, *tl);
787 cnt = i = 0;
788
789 /*
790 * For NFS Version 2, it is not obvious what a write of zero length
791 * should do, but I might as well be consistent with Version 3,
792 * which is to return ok so long as there are no permission problems.
793 */
794 if (len > 0) {
795 zeroing = 1;
796 mp = mrep;
797 while (mp) {
798 if (mp == md) {
799 zeroing = 0;
800 adjust = dpos - mtod(mp, caddr_t);
801 mp->m_len -= adjust;
802 if (mp->m_len > 0 && adjust > 0)
803 NFSMADV(mp, adjust);
804 }
805 if (zeroing)
806 mp->m_len = 0;
807 else if (mp->m_len > 0) {
808 i += mp->m_len;
809 if (i > len) {
810 mp->m_len -= (i - len);
811 zeroing = 1;
812 }
813 if (mp->m_len > 0)
814 cnt++;
815 }
816 mp = mp->m_next;
817 }
818 }
819 if (len > NFS_MAXDATA || len < 0 || i < len) {
820 error = EIO;
821 nfsm_reply(2 * NFSX_UNSIGNED);
822 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
823 return (0);
824 }
825 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
826 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
827 if (error) {
828 nfsm_reply(2 * NFSX_UNSIGNED);
829 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
830 return (0);
831 }
832 if (v3)
833 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
834 if (vp->v_type != VREG) {
835 if (v3)
836 error = EINVAL;
837 else
838 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
839 }
840 if (!error) {
841 nqsrv_getl(vp, ND_WRITE);
842 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
843 }
844 if (error) {
845 vput(vp);
846 nfsm_reply(NFSX_WCCDATA(v3));
847 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
848 return (0);
849 }
850
851 if (len > 0) {
852 MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP,
853 M_WAITOK);
854 uiop->uio_iov = iv = ivp;
855 uiop->uio_iovcnt = cnt;
856 mp = mrep;
857 while (mp) {
858 if (mp->m_len > 0) {
859 ivp->iov_base = mtod(mp, caddr_t);
860 ivp->iov_len = mp->m_len;
861 ivp++;
862 }
863 mp = mp->m_next;
864 }
865
866 /*
867 * XXX
868 * The IO_METASYNC flag indicates that all metadata (and not just
869 * enough to ensure data integrity) mus be written to stable storage
870 * synchronously.
871 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
872 */
873 if (stable == NFSV3WRITE_UNSTABLE)
874 ioflags = IO_NODELOCKED;
875 else if (stable == NFSV3WRITE_DATASYNC)
876 ioflags = (IO_SYNC | IO_NODELOCKED);
877 else
878 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
879 uiop->uio_resid = len;
880 uiop->uio_rw = UIO_WRITE;
881 uiop->uio_segflg = UIO_SYSSPACE;
882 uiop->uio_procp = (struct proc *)0;
883 uiop->uio_offset = off;
884 error = VOP_WRITE(vp, uiop, ioflags, cred);
885 nfsstats.srvvop_writes++;
886 FREE((caddr_t)iv, M_TEMP);
887 }
888 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
889 vput(vp);
890 if (!error)
891 error = aftat_ret;
892 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
893 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
894 if (v3) {
895 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
896 if (error)
897 return (0);
898 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
899 *tl++ = txdr_unsigned(retlen);
900 if (stable == NFSV3WRITE_UNSTABLE)
901 *tl++ = txdr_unsigned(stable);
902 else
903 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
904 /*
905 * Actually, there is no need to txdr these fields,
906 * but it may make the values more human readable,
907 * for debugging purposes.
908 */
909 *tl++ = txdr_unsigned(boottime.tv_sec);
910 *tl = txdr_unsigned(boottime.tv_usec);
911 } else {
912 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
913 nfsm_srvfillattr(&va, fp);
914 }
915 nfsm_srvdone;
916 }
917
918 /*
919 * NFS write service with write gathering support. Called when
920 * nfsrvw_procrastinate > 0.
921 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
922 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
923 * Jan. 1994.
924 */
925 int
926 nfsrv_writegather(ndp, slp, procp, mrq)
927 struct nfsrv_descript **ndp;
928 struct nfssvc_sock *slp;
929 struct proc *procp;
930 struct mbuf **mrq;
931 {
932 register struct iovec *ivp;
933 register struct mbuf *mp;
934 register struct nfsrv_descript *wp, *nfsd, *owp, *swp;
935 register struct nfs_fattr *fp;
936 register int i = 0;
937 struct iovec *iov;
938 struct nfsrvw_delayhash *wpp;
939 struct ucred *cred;
940 struct vattr va, forat;
941 register u_int32_t *tl;
942 register int32_t t1;
943 caddr_t bpos, dpos;
944 int error = 0, rdonly, cache, len = 0, forat_ret = 1;
945 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
946 char *cp2;
947 struct mbuf *mb, *mb2, *mreq, *mrep, *md;
948 struct vnode *vp;
949 struct uio io, *uiop = &io;
950 u_quad_t frev, cur_usec;
951
952 *mrq = NULL;
953 if (*ndp) {
954 nfsd = *ndp;
955 *ndp = NULL;
956 mrep = nfsd->nd_mrep;
957 md = nfsd->nd_md;
958 dpos = nfsd->nd_dpos;
959 cred = &nfsd->nd_cr;
960 v3 = (nfsd->nd_flag & ND_NFSV3);
961 LIST_INIT(&nfsd->nd_coalesce);
962 nfsd->nd_mreq = NULL;
963 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
964 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
965 nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
966
967 /*
968 * Now, get the write header..
969 */
970 nfsm_srvmtofh(&nfsd->nd_fh);
971 if (v3) {
972 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
973 fxdr_hyper(tl, &nfsd->nd_off);
974 tl += 3;
975 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
976 } else {
977 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
978 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
979 tl += 2;
980 }
981 len = fxdr_unsigned(int32_t, *tl);
982 nfsd->nd_len = len;
983 nfsd->nd_eoff = nfsd->nd_off + len;
984
985 /*
986 * Trim the header out of the mbuf list and trim off any trailing
987 * junk so that the mbuf list has only the write data.
988 */
989 zeroing = 1;
990 i = 0;
991 mp = mrep;
992 while (mp) {
993 if (mp == md) {
994 zeroing = 0;
995 adjust = dpos - mtod(mp, caddr_t);
996 mp->m_len -= adjust;
997 if (mp->m_len > 0 && adjust > 0)
998 NFSMADV(mp, adjust);
999 }
1000 if (zeroing)
1001 mp->m_len = 0;
1002 else {
1003 i += mp->m_len;
1004 if (i > len) {
1005 mp->m_len -= (i - len);
1006 zeroing = 1;
1007 }
1008 }
1009 mp = mp->m_next;
1010 }
1011 if (len > NFS_MAXDATA || len < 0 || i < len) {
1012 nfsmout:
1013 m_freem(mrep);
1014 error = EIO;
1015 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1016 if (v3)
1017 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1018 nfsd->nd_mreq = mreq;
1019 nfsd->nd_mrep = NULL;
1020 nfsd->nd_time = 0;
1021 }
1022
1023 /*
1024 * Add this entry to the hash and time queues.
1025 */
1026 s = splsoftclock();
1027 owp = NULL;
1028 wp = slp->ns_tq.lh_first;
1029 while (wp && wp->nd_time < nfsd->nd_time) {
1030 owp = wp;
1031 wp = wp->nd_tq.le_next;
1032 }
1033 if (owp) {
1034 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1035 } else {
1036 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1037 }
1038 if (nfsd->nd_mrep) {
1039 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1040 owp = NULL;
1041 wp = wpp->lh_first;
1042 while (wp &&
1043 bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) {
1044 owp = wp;
1045 wp = wp->nd_hash.le_next;
1046 }
1047 while (wp && wp->nd_off < nfsd->nd_off &&
1048 !bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) {
1049 owp = wp;
1050 wp = wp->nd_hash.le_next;
1051 }
1052 if (owp) {
1053 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1054
1055 /*
1056 * Search the hash list for overlapping entries and
1057 * coalesce.
1058 */
1059 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1060 wp = nfsd->nd_hash.le_next;
1061 if (NFSW_SAMECRED(owp, nfsd))
1062 nfsrvw_coalesce(owp, nfsd);
1063 }
1064 } else {
1065 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1066 }
1067 }
1068 splx(s);
1069 }
1070
1071 /*
1072 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1073 * and generate the associated reply mbuf list(s).
1074 */
1075 loop1:
1076 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1077 s = splsoftclock();
1078 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = owp) {
1079 owp = nfsd->nd_tq.le_next;
1080 if (nfsd->nd_time > cur_usec)
1081 break;
1082 if (nfsd->nd_mreq)
1083 continue;
1084 LIST_REMOVE(nfsd, nd_tq);
1085 LIST_REMOVE(nfsd, nd_hash);
1086 splx(s);
1087 mrep = nfsd->nd_mrep;
1088 nfsd->nd_mrep = NULL;
1089 cred = &nfsd->nd_cr;
1090 v3 = (nfsd->nd_flag & ND_NFSV3);
1091 forat_ret = aftat_ret = 1;
1092 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1093 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1094 FALSE);
1095 if (!error) {
1096 if (v3)
1097 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
1098 if (vp->v_type != VREG) {
1099 if (v3)
1100 error = EINVAL;
1101 else
1102 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1103 }
1104 } else
1105 vp = NULL;
1106 if (!error) {
1107 nqsrv_getl(vp, ND_WRITE);
1108 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
1109 }
1110
1111 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1112 ioflags = IO_NODELOCKED;
1113 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1114 ioflags = (IO_SYNC | IO_NODELOCKED);
1115 else
1116 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1117 uiop->uio_rw = UIO_WRITE;
1118 uiop->uio_segflg = UIO_SYSSPACE;
1119 uiop->uio_procp = (struct proc *)0;
1120 uiop->uio_offset = nfsd->nd_off;
1121 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1122 if (uiop->uio_resid > 0) {
1123 mp = mrep;
1124 i = 0;
1125 while (mp) {
1126 if (mp->m_len > 0)
1127 i++;
1128 mp = mp->m_next;
1129 }
1130 uiop->uio_iovcnt = i;
1131 MALLOC(iov, struct iovec *, i * sizeof (struct iovec),
1132 M_TEMP, M_WAITOK);
1133 uiop->uio_iov = ivp = iov;
1134 mp = mrep;
1135 while (mp) {
1136 if (mp->m_len > 0) {
1137 ivp->iov_base = mtod(mp, caddr_t);
1138 ivp->iov_len = mp->m_len;
1139 ivp++;
1140 }
1141 mp = mp->m_next;
1142 }
1143 if (!error) {
1144 error = VOP_WRITE(vp, uiop, ioflags, cred);
1145 nfsstats.srvvop_writes++;
1146 }
1147 FREE((caddr_t)iov, M_TEMP);
1148 }
1149 m_freem(mrep);
1150 if (vp) {
1151 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
1152 vput(vp);
1153 }
1154
1155 /*
1156 * Loop around generating replies for all write rpcs that have
1157 * now been completed.
1158 */
1159 swp = nfsd;
1160 do {
1161 if (error) {
1162 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1163 if (v3) {
1164 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1165 }
1166 } else {
1167 nfsm_writereply(NFSX_PREOPATTR(v3) +
1168 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1169 NFSX_WRITEVERF(v3), v3);
1170 if (v3) {
1171 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1172 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1173 *tl++ = txdr_unsigned(nfsd->nd_len);
1174 *tl++ = txdr_unsigned(swp->nd_stable);
1175 /*
1176 * Actually, there is no need to txdr these fields,
1177 * but it may make the values more human readable,
1178 * for debugging purposes.
1179 */
1180 *tl++ = txdr_unsigned(boottime.tv_sec);
1181 *tl = txdr_unsigned(boottime.tv_usec);
1182 } else {
1183 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1184 nfsm_srvfillattr(&va, fp);
1185 }
1186 }
1187 nfsd->nd_mreq = mreq;
1188 if (nfsd->nd_mrep)
1189 panic("nfsrv_write: nd_mrep not free");
1190
1191 /*
1192 * Done. Put it at the head of the timer queue so that
1193 * the final phase can return the reply.
1194 */
1195 s = splsoftclock();
1196 if (nfsd != swp) {
1197 nfsd->nd_time = 0;
1198 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1199 }
1200 nfsd = swp->nd_coalesce.lh_first;
1201 if (nfsd) {
1202 LIST_REMOVE(nfsd, nd_tq);
1203 }
1204 splx(s);
1205 } while (nfsd);
1206 s = splsoftclock();
1207 swp->nd_time = 0;
1208 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1209 splx(s);
1210 goto loop1;
1211 }
1212 splx(s);
1213
1214 /*
1215 * Search for a reply to return.
1216 */
1217 s = splsoftclock();
1218 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = nfsd->nd_tq.le_next)
1219 if (nfsd->nd_mreq) {
1220 LIST_REMOVE(nfsd, nd_tq);
1221 *mrq = nfsd->nd_mreq;
1222 *ndp = nfsd;
1223 break;
1224 }
1225 splx(s);
1226 return (0);
1227 }
1228
1229 /*
1230 * Coalesce the write request nfsd into owp. To do this we must:
1231 * - remove nfsd from the queues
1232 * - merge nfsd->nd_mrep into owp->nd_mrep
1233 * - update the nd_eoff and nd_stable for owp
1234 * - put nfsd on owp's nd_coalesce list
1235 * NB: Must be called at splsoftclock().
1236 */
1237 void
1238 nfsrvw_coalesce(owp, nfsd)
1239 register struct nfsrv_descript *owp;
1240 register struct nfsrv_descript *nfsd;
1241 {
1242 register int overlap;
1243 register struct mbuf *mp;
1244
1245 LIST_REMOVE(nfsd, nd_hash);
1246 LIST_REMOVE(nfsd, nd_tq);
1247 if (owp->nd_eoff < nfsd->nd_eoff) {
1248 overlap = owp->nd_eoff - nfsd->nd_off;
1249 if (overlap < 0)
1250 panic("nfsrv_coalesce: bad off");
1251 if (overlap > 0)
1252 m_adj(nfsd->nd_mrep, overlap);
1253 mp = owp->nd_mrep;
1254 while (mp->m_next)
1255 mp = mp->m_next;
1256 mp->m_next = nfsd->nd_mrep;
1257 owp->nd_eoff = nfsd->nd_eoff;
1258 } else
1259 m_freem(nfsd->nd_mrep);
1260 nfsd->nd_mrep = NULL;
1261 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1262 owp->nd_stable = NFSV3WRITE_FILESYNC;
1263 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1264 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1265 owp->nd_stable = NFSV3WRITE_DATASYNC;
1266 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1267 /*
1268 * nfsd might hold coalesce elements! Move them to owp.
1269 * Otherwise, requests may be lost and clients will be stuck.
1270 */
1271 if (nfsd->nd_coalesce.lh_first)
1272 {
1273 register struct nfsrv_descript *m;
1274
1275 while ((m = nfsd->nd_coalesce.lh_first))
1276 {
1277 LIST_REMOVE(m, nd_tq);
1278 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1279 }
1280 }
1281 }
1282
1283 /*
1284 * nfs create service
1285 * now does a truncate to 0 length via. setattr if it already exists
1286 */
1287 int
1288 nfsrv_create(nfsd, slp, procp, mrq)
1289 struct nfsrv_descript *nfsd;
1290 struct nfssvc_sock *slp;
1291 struct proc *procp;
1292 struct mbuf **mrq;
1293 {
1294 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1295 struct mbuf *nam = nfsd->nd_nam;
1296 caddr_t dpos = nfsd->nd_dpos;
1297 struct ucred *cred = &nfsd->nd_cr;
1298 register struct nfs_fattr *fp;
1299 struct vattr va, dirfor, diraft;
1300 register struct nfsv2_sattr *sp;
1301 register u_int32_t *tl;
1302 struct nameidata nd;
1303 register caddr_t cp;
1304 register int32_t t1;
1305 caddr_t bpos;
1306 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1307 int rdev = 0;
1308 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1309 char *cp2;
1310 struct mbuf *mb, *mb2, *mreq;
1311 struct vnode *vp = NULL, *dirp = NULL;
1312 nfsfh_t nfh;
1313 fhandle_t *fhp;
1314 u_quad_t frev, tempsize;
1315 u_char cverf[NFSX_V3CREATEVERF];
1316
1317 nd.ni_cnd.cn_nameiop = 0;
1318 fhp = &nfh.fh_generic;
1319 nfsm_srvmtofh(fhp);
1320 nfsm_srvnamesiz(len);
1321 nd.ni_cnd.cn_cred = cred;
1322 nd.ni_cnd.cn_nameiop = CREATE;
1323 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART;
1324 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1325 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1326 if (dirp) {
1327 if (v3)
1328 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1329 procp);
1330 else {
1331 vrele(dirp);
1332 dirp = (struct vnode *)0;
1333 }
1334 }
1335 if (error) {
1336 nfsm_reply(NFSX_WCCDATA(v3));
1337 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1338 if (dirp)
1339 vrele(dirp);
1340 return (0);
1341 }
1342 VATTR_NULL(&va);
1343 if (v3) {
1344 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1345 how = fxdr_unsigned(int, *tl);
1346 switch (how) {
1347 case NFSV3CREATE_GUARDED:
1348 if (nd.ni_vp) {
1349 error = EEXIST;
1350 break;
1351 }
1352 case NFSV3CREATE_UNCHECKED:
1353 nfsm_srvsattr(&va);
1354 break;
1355 case NFSV3CREATE_EXCLUSIVE:
1356 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
1357 bcopy(cp, cverf, NFSX_V3CREATEVERF);
1358 exclusive_flag = 1;
1359 if (nd.ni_vp == NULL)
1360 va.va_mode = 0;
1361 break;
1362 };
1363 va.va_type = VREG;
1364 } else {
1365 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1366 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1367 if (va.va_type == VNON)
1368 va.va_type = VREG;
1369 va.va_mode = nfstov_mode(sp->sa_mode);
1370 switch (va.va_type) {
1371 case VREG:
1372 tsize = fxdr_unsigned(int32_t, sp->sa_size);
1373 if (tsize != -1)
1374 va.va_size = (u_quad_t)tsize;
1375 break;
1376 case VCHR:
1377 case VBLK:
1378 case VFIFO:
1379 rdev = fxdr_unsigned(int32_t, sp->sa_size);
1380 break;
1381 default:
1382 break;
1383 };
1384 }
1385
1386 /*
1387 * Iff doesn't exist, create it
1388 * otherwise just truncate to 0 length
1389 * should I set the mode too ??
1390 */
1391 if (nd.ni_vp == NULL) {
1392 if (va.va_type == VREG || va.va_type == VSOCK) {
1393 vrele(nd.ni_startdir);
1394 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1395 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1396 if (!error) {
1397 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1398 if (exclusive_flag) {
1399 exclusive_flag = 0;
1400 VATTR_NULL(&va);
1401 bcopy(cverf, (caddr_t)&va.va_atime,
1402 NFSX_V3CREATEVERF);
1403 error = VOP_SETATTR(nd.ni_vp, &va, cred,
1404 procp);
1405 }
1406 }
1407 } else if (va.va_type == VCHR || va.va_type == VBLK ||
1408 va.va_type == VFIFO) {
1409 if (va.va_type == VCHR && rdev == 0xffffffff)
1410 va.va_type = VFIFO;
1411 error = suser(cred, (u_short *)0);
1412 if (error) {
1413 vrele(nd.ni_startdir);
1414 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1415 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1416 vput(nd.ni_dvp);
1417 nfsm_reply(0);
1418 return (error);
1419 } else
1420 va.va_rdev = (dev_t)rdev;
1421 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1422 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1423 &va);
1424 if (error) {
1425 vrele(nd.ni_startdir);
1426 nfsm_reply(0);
1427 }
1428 nd.ni_cnd.cn_nameiop = LOOKUP;
1429 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART);
1430 nd.ni_cnd.cn_proc = procp;
1431 nd.ni_cnd.cn_cred = cred;
1432 if ((error = lookup(&nd)) != 0) {
1433 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1434 nfsm_reply(0);
1435 }
1436 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1437 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1438 vrele(nd.ni_dvp);
1439 vput(nd.ni_vp);
1440 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1441 error = EINVAL;
1442 nfsm_reply(0);
1443 }
1444 } else {
1445 vrele(nd.ni_startdir);
1446 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1447 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1448 vput(nd.ni_dvp);
1449 error = ENXIO;
1450 }
1451 vp = nd.ni_vp;
1452 } else {
1453 vrele(nd.ni_startdir);
1454 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1455 vp = nd.ni_vp;
1456 if (nd.ni_dvp == vp)
1457 vrele(nd.ni_dvp);
1458 else
1459 vput(nd.ni_dvp);
1460 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1461 if (va.va_size != -1) {
1462 error = nfsrv_access(vp, VWRITE, cred,
1463 (nd.ni_cnd.cn_flags & RDONLY), procp, 0);
1464 if (!error) {
1465 nqsrv_getl(vp, ND_WRITE);
1466 tempsize = va.va_size;
1467 VATTR_NULL(&va);
1468 va.va_size = tempsize;
1469 error = VOP_SETATTR(vp, &va, cred,
1470 procp);
1471 }
1472 if (error)
1473 vput(vp);
1474 }
1475 }
1476 if (!error) {
1477 bzero((caddr_t)fhp, sizeof(nfh));
1478 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1479 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1480 if (!error)
1481 error = VOP_GETATTR(vp, &va, cred, procp);
1482 vput(vp);
1483 }
1484 if (v3) {
1485 if (exclusive_flag && !error &&
1486 bcmp(cverf, (caddr_t)&va.va_atime, NFSX_V3CREATEVERF))
1487 error = EEXIST;
1488 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1489 vrele(dirp);
1490 }
1491 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1492 if (v3) {
1493 if (!error) {
1494 nfsm_srvpostop_fh(fhp);
1495 nfsm_srvpostop_attr(0, &va);
1496 }
1497 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1498 } else {
1499 nfsm_srvfhtom(fhp, v3);
1500 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1501 nfsm_srvfillattr(&va, fp);
1502 }
1503 return (0);
1504 nfsmout:
1505 if (dirp)
1506 vrele(dirp);
1507 if (nd.ni_cnd.cn_nameiop) {
1508 vrele(nd.ni_startdir);
1509 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1510 }
1511 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1512 if (nd.ni_dvp == nd.ni_vp)
1513 vrele(nd.ni_dvp);
1514 else
1515 vput(nd.ni_dvp);
1516 if (nd.ni_vp)
1517 vput(nd.ni_vp);
1518 return (error);
1519 }
1520
1521 /*
1522 * nfs v3 mknod service
1523 */
1524 int
1525 nfsrv_mknod(nfsd, slp, procp, mrq)
1526 struct nfsrv_descript *nfsd;
1527 struct nfssvc_sock *slp;
1528 struct proc *procp;
1529 struct mbuf **mrq;
1530 {
1531 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1532 struct mbuf *nam = nfsd->nd_nam;
1533 caddr_t dpos = nfsd->nd_dpos;
1534 struct ucred *cred = &nfsd->nd_cr;
1535 struct vattr va, dirfor, diraft;
1536 register u_int32_t *tl;
1537 struct nameidata nd;
1538 register int32_t t1;
1539 caddr_t bpos;
1540 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1541 u_int32_t major, minor;
1542 enum vtype vtyp;
1543 char *cp2;
1544 struct mbuf *mb, *mb2, *mreq;
1545 struct vnode *vp, *dirp = (struct vnode *)0;
1546 nfsfh_t nfh;
1547 fhandle_t *fhp;
1548 u_quad_t frev;
1549
1550 nd.ni_cnd.cn_nameiop = 0;
1551 fhp = &nfh.fh_generic;
1552 nfsm_srvmtofh(fhp);
1553 nfsm_srvnamesiz(len);
1554 nd.ni_cnd.cn_cred = cred;
1555 nd.ni_cnd.cn_nameiop = CREATE;
1556 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART;
1557 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1558 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1559 if (dirp)
1560 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp);
1561 if (error) {
1562 nfsm_reply(NFSX_WCCDATA(1));
1563 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1564 if (dirp)
1565 vrele(dirp);
1566 return (0);
1567 }
1568 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1569 vtyp = nfsv3tov_type(*tl);
1570 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1571 vrele(nd.ni_startdir);
1572 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1573 error = NFSERR_BADTYPE;
1574 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1575 vput(nd.ni_dvp);
1576 goto out;
1577 }
1578 VATTR_NULL(&va);
1579 nfsm_srvsattr(&va);
1580 if (vtyp == VCHR || vtyp == VBLK) {
1581 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1582 major = fxdr_unsigned(u_int32_t, *tl++);
1583 minor = fxdr_unsigned(u_int32_t, *tl);
1584 va.va_rdev = makedev(major, minor);
1585 }
1586
1587 /*
1588 * Iff doesn't exist, create it.
1589 */
1590 if (nd.ni_vp) {
1591 vrele(nd.ni_startdir);
1592 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1593 error = EEXIST;
1594 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1595 vput(nd.ni_dvp);
1596 goto out;
1597 }
1598 va.va_type = vtyp;
1599 if (vtyp == VSOCK) {
1600 vrele(nd.ni_startdir);
1601 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1602 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1603 if (!error)
1604 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1605 } else {
1606 error = suser(cred, (u_short *)0);
1607 if (error) {
1608 vrele(nd.ni_startdir);
1609 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1610 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1611 vput(nd.ni_dvp);
1612 goto out;
1613 }
1614 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1615 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1616 if (error) {
1617 vrele(nd.ni_startdir);
1618 goto out;
1619 }
1620 nd.ni_cnd.cn_nameiop = LOOKUP;
1621 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART);
1622 nd.ni_cnd.cn_proc = procp;
1623 nd.ni_cnd.cn_cred = procp->p_ucred;
1624 error = lookup(&nd);
1625 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1626 if (error)
1627 goto out;
1628 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1629 vrele(nd.ni_dvp);
1630 vput(nd.ni_vp);
1631 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1632 error = EINVAL;
1633 }
1634 }
1635 out:
1636 vp = nd.ni_vp;
1637 if (!error) {
1638 bzero((caddr_t)fhp, sizeof(nfh));
1639 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1640 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1641 if (!error)
1642 error = VOP_GETATTR(vp, &va, cred, procp);
1643 vput(vp);
1644 }
1645 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1646 vrele(dirp);
1647 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
1648 if (!error) {
1649 nfsm_srvpostop_fh(fhp);
1650 nfsm_srvpostop_attr(0, &va);
1651 }
1652 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1653 return (0);
1654 nfsmout:
1655 if (dirp)
1656 vrele(dirp);
1657 if (nd.ni_cnd.cn_nameiop) {
1658 vrele(nd.ni_startdir);
1659 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1660 }
1661 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1662 if (nd.ni_dvp == nd.ni_vp)
1663 vrele(nd.ni_dvp);
1664 else
1665 vput(nd.ni_dvp);
1666 if (nd.ni_vp)
1667 vput(nd.ni_vp);
1668 return (error);
1669 }
1670
1671 /*
1672 * nfs remove service
1673 */
1674 int
1675 nfsrv_remove(nfsd, slp, procp, mrq)
1676 struct nfsrv_descript *nfsd;
1677 struct nfssvc_sock *slp;
1678 struct proc *procp;
1679 struct mbuf **mrq;
1680 {
1681 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1682 struct mbuf *nam = nfsd->nd_nam;
1683 caddr_t dpos = nfsd->nd_dpos;
1684 struct ucred *cred = &nfsd->nd_cr;
1685 struct nameidata nd;
1686 register u_int32_t *tl;
1687 register int32_t t1;
1688 caddr_t bpos;
1689 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1690 int v3 = (nfsd->nd_flag & ND_NFSV3);
1691 char *cp2;
1692 struct mbuf *mb, *mreq;
1693 struct vnode *vp, *dirp;
1694 struct vattr dirfor, diraft;
1695 nfsfh_t nfh;
1696 fhandle_t *fhp;
1697 u_quad_t frev;
1698
1699 #ifndef nolint
1700 vp = (struct vnode *)0;
1701 #endif
1702 fhp = &nfh.fh_generic;
1703 nfsm_srvmtofh(fhp);
1704 nfsm_srvnamesiz(len);
1705 nd.ni_cnd.cn_cred = cred;
1706 nd.ni_cnd.cn_nameiop = DELETE;
1707 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1708 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1709 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1710 if (dirp) {
1711 if (v3)
1712 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1713 procp);
1714 else
1715 vrele(dirp);
1716 }
1717 if (!error) {
1718 vp = nd.ni_vp;
1719 if (vp->v_type == VDIR &&
1720 (error = suser(cred, (u_short *)0)) != 0)
1721 goto out;
1722 /*
1723 * The root of a mounted filesystem cannot be deleted.
1724 */
1725 if (vp->v_flag & VROOT) {
1726 error = EBUSY;
1727 goto out;
1728 }
1729 out:
1730 if (!error) {
1731 #if defined(UVM)
1732 (void)uvm_vnp_uncache(vp);
1733 #else
1734 (void)vnode_pager_uncache(vp);
1735 #endif
1736 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1737 nqsrv_getl(vp, ND_WRITE);
1738 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1739 } else {
1740 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1741 if (nd.ni_dvp == vp)
1742 vrele(nd.ni_dvp);
1743 else
1744 vput(nd.ni_dvp);
1745 vput(vp);
1746 }
1747 }
1748 if (dirp && v3) {
1749 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1750 vrele(dirp);
1751 }
1752 nfsm_reply(NFSX_WCCDATA(v3));
1753 if (v3) {
1754 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1755 return (0);
1756 }
1757 nfsm_srvdone;
1758 }
1759
1760 /*
1761 * nfs rename service
1762 */
1763 int
1764 nfsrv_rename(nfsd, slp, procp, mrq)
1765 struct nfsrv_descript *nfsd;
1766 struct nfssvc_sock *slp;
1767 struct proc *procp;
1768 struct mbuf **mrq;
1769 {
1770 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1771 struct mbuf *nam = nfsd->nd_nam;
1772 caddr_t dpos = nfsd->nd_dpos;
1773 struct ucred *cred = &nfsd->nd_cr;
1774 register u_int32_t *tl;
1775 register int32_t t1;
1776 caddr_t bpos;
1777 int error = 0, cache, len, len2, fdirfor_ret = 1, fdiraft_ret = 1;
1778 int tdirfor_ret = 1, tdiraft_ret = 1;
1779 int v3 = (nfsd->nd_flag & ND_NFSV3);
1780 char *cp2;
1781 struct mbuf *mb, *mreq;
1782 struct nameidata fromnd, tond;
1783 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
1784 struct vnode *tdirp = (struct vnode *)0;
1785 struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1786 nfsfh_t fnfh, tnfh;
1787 fhandle_t *ffhp, *tfhp;
1788 u_quad_t frev;
1789 uid_t saved_uid;
1790
1791 #ifndef nolint
1792 fvp = (struct vnode *)0;
1793 #endif
1794 ffhp = &fnfh.fh_generic;
1795 tfhp = &tnfh.fh_generic;
1796 fromnd.ni_cnd.cn_nameiop = 0;
1797 tond.ni_cnd.cn_nameiop = 0;
1798 nfsm_srvmtofh(ffhp);
1799 nfsm_srvnamesiz(len);
1800 /*
1801 * Remember our original uid so that we can reset cr_uid before
1802 * the second nfs_namei() call, in case it is remapped.
1803 */
1804 saved_uid = cred->cr_uid;
1805 fromnd.ni_cnd.cn_cred = cred;
1806 fromnd.ni_cnd.cn_nameiop = DELETE;
1807 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART;
1808 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
1809 &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1810 if (fdirp) {
1811 if (v3)
1812 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred,
1813 procp);
1814 else {
1815 vrele(fdirp);
1816 fdirp = (struct vnode *)0;
1817 }
1818 }
1819 if (error) {
1820 nfsm_reply(2 * NFSX_WCCDATA(v3));
1821 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1822 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1823 if (fdirp)
1824 vrele(fdirp);
1825 return (0);
1826 }
1827 fvp = fromnd.ni_vp;
1828 nfsm_srvmtofh(tfhp);
1829 nfsm_strsiz(len2, NFS_MAXNAMLEN);
1830 cred->cr_uid = saved_uid;
1831 tond.ni_cnd.cn_cred = cred;
1832 tond.ni_cnd.cn_nameiop = RENAME;
1833 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
1834 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
1835 &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1836 if (tdirp) {
1837 if (v3)
1838 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred,
1839 procp);
1840 else {
1841 vrele(tdirp);
1842 tdirp = (struct vnode *)0;
1843 }
1844 }
1845 if (error) {
1846 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1847 vrele(fromnd.ni_dvp);
1848 vrele(fvp);
1849 goto out1;
1850 }
1851 tdvp = tond.ni_dvp;
1852 tvp = tond.ni_vp;
1853 if (tvp != NULL) {
1854 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1855 if (v3)
1856 error = EEXIST;
1857 else
1858 error = EISDIR;
1859 goto out;
1860 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1861 if (v3)
1862 error = EEXIST;
1863 else
1864 error = ENOTDIR;
1865 goto out;
1866 }
1867 if (tvp->v_type == VDIR && tvp->v_mountedhere) {
1868 if (v3)
1869 error = EXDEV;
1870 else
1871 error = ENOTEMPTY;
1872 goto out;
1873 }
1874 }
1875 if (fvp->v_type == VDIR && fvp->v_mountedhere) {
1876 if (v3)
1877 error = EXDEV;
1878 else
1879 error = ENOTEMPTY;
1880 goto out;
1881 }
1882 if (fvp->v_mount != tdvp->v_mount) {
1883 if (v3)
1884 error = EXDEV;
1885 else
1886 error = ENOTEMPTY;
1887 goto out;
1888 }
1889 if (fvp == tdvp)
1890 if (v3)
1891 error = EINVAL;
1892 else
1893 error = ENOTEMPTY;
1894 /*
1895 * If source is the same as the destination (that is the
1896 * same vnode with the same name in the same directory),
1897 * then there is nothing to do.
1898 */
1899 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
1900 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1901 !bcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
1902 fromnd.ni_cnd.cn_namelen))
1903 error = -1;
1904 out:
1905 if (!error) {
1906 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
1907 nqsrv_getl(tdvp, ND_WRITE);
1908 if (tvp) {
1909 #if defined(UVM)
1910 (void)uvm_vnp_uncache(tvp);
1911 #else
1912 (void)vnode_pager_uncache(tvp);
1913 #endif
1914 nqsrv_getl(tvp, ND_WRITE);
1915 }
1916 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
1917 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
1918 } else {
1919 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
1920 if (tdvp == tvp)
1921 vrele(tdvp);
1922 else
1923 vput(tdvp);
1924 if (tvp)
1925 vput(tvp);
1926 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1927 vrele(fromnd.ni_dvp);
1928 vrele(fvp);
1929 if (error == -1)
1930 error = 0;
1931 }
1932 vrele(tond.ni_startdir);
1933 FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
1934 out1:
1935 if (fdirp) {
1936 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp);
1937 vrele(fdirp);
1938 }
1939 if (tdirp) {
1940 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp);
1941 vrele(tdirp);
1942 }
1943 vrele(fromnd.ni_startdir);
1944 FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
1945 nfsm_reply(2 * NFSX_WCCDATA(v3));
1946 if (v3) {
1947 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1948 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1949 }
1950 return (0);
1951
1952 nfsmout:
1953 if (fdirp)
1954 vrele(fdirp);
1955 if (tdirp)
1956 vrele(tdirp);
1957 if (tond.ni_cnd.cn_nameiop) {
1958 vrele(tond.ni_startdir);
1959 FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
1960 }
1961 if (fromnd.ni_cnd.cn_nameiop) {
1962 vrele(fromnd.ni_startdir);
1963 FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
1964 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1965 vrele(fromnd.ni_dvp);
1966 vrele(fvp);
1967 }
1968 return (error);
1969 }
1970
1971 /*
1972 * nfs link service
1973 */
1974 int
1975 nfsrv_link(nfsd, slp, procp, mrq)
1976 struct nfsrv_descript *nfsd;
1977 struct nfssvc_sock *slp;
1978 struct proc *procp;
1979 struct mbuf **mrq;
1980 {
1981 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1982 struct mbuf *nam = nfsd->nd_nam;
1983 caddr_t dpos = nfsd->nd_dpos;
1984 struct ucred *cred = &nfsd->nd_cr;
1985 struct nameidata nd;
1986 register u_int32_t *tl;
1987 register int32_t t1;
1988 caddr_t bpos;
1989 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1;
1990 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
1991 char *cp2;
1992 struct mbuf *mb, *mreq;
1993 struct vnode *vp, *xp, *dirp = (struct vnode *)0;
1994 struct vattr dirfor, diraft, at;
1995 nfsfh_t nfh, dnfh;
1996 fhandle_t *fhp, *dfhp;
1997 u_quad_t frev;
1998
1999 fhp = &nfh.fh_generic;
2000 dfhp = &dnfh.fh_generic;
2001 nfsm_srvmtofh(fhp);
2002 nfsm_srvmtofh(dfhp);
2003 nfsm_srvnamesiz(len);
2004 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
2005 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2006 if (error) {
2007 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2008 nfsm_srvpostop_attr(getret, &at);
2009 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2010 return (0);
2011 }
2012 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0)
2013 goto out1;
2014 nd.ni_cnd.cn_cred = cred;
2015 nd.ni_cnd.cn_nameiop = CREATE;
2016 nd.ni_cnd.cn_flags = LOCKPARENT;
2017 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
2018 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2019 if (dirp) {
2020 if (v3)
2021 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2022 procp);
2023 else {
2024 vrele(dirp);
2025 dirp = (struct vnode *)0;
2026 }
2027 }
2028 if (error)
2029 goto out1;
2030 xp = nd.ni_vp;
2031 if (xp != NULL) {
2032 error = EEXIST;
2033 goto out;
2034 }
2035 xp = nd.ni_dvp;
2036 if (vp->v_mount != xp->v_mount)
2037 error = EXDEV;
2038 out:
2039 if (!error) {
2040 nqsrv_getl(vp, ND_WRITE);
2041 nqsrv_getl(xp, ND_WRITE);
2042 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2043 } else {
2044 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2045 if (nd.ni_dvp == nd.ni_vp)
2046 vrele(nd.ni_dvp);
2047 else
2048 vput(nd.ni_dvp);
2049 if (nd.ni_vp)
2050 vrele(nd.ni_vp);
2051 }
2052 out1:
2053 if (v3)
2054 getret = VOP_GETATTR(vp, &at, cred, procp);
2055 if (dirp) {
2056 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2057 vrele(dirp);
2058 }
2059 vrele(vp);
2060 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2061 if (v3) {
2062 nfsm_srvpostop_attr(getret, &at);
2063 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2064 return (0);
2065 }
2066 nfsm_srvdone;
2067 }
2068
2069 /*
2070 * nfs symbolic link service
2071 */
2072 int
2073 nfsrv_symlink(nfsd, slp, procp, mrq)
2074 struct nfsrv_descript *nfsd;
2075 struct nfssvc_sock *slp;
2076 struct proc *procp;
2077 struct mbuf **mrq;
2078 {
2079 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2080 struct mbuf *nam = nfsd->nd_nam;
2081 caddr_t dpos = nfsd->nd_dpos;
2082 struct ucred *cred = &nfsd->nd_cr;
2083 struct vattr va, dirfor, diraft;
2084 struct nameidata nd;
2085 register u_int32_t *tl;
2086 register int32_t t1;
2087 struct nfsv2_sattr *sp;
2088 char *bpos, *pathcp = NULL, *cp2;
2089 struct uio io;
2090 struct iovec iv;
2091 int error = 0, cache, len, len2, dirfor_ret = 1, diraft_ret = 1;
2092 int v3 = (nfsd->nd_flag & ND_NFSV3);
2093 struct mbuf *mb, *mreq, *mb2;
2094 struct vnode *dirp = (struct vnode *)0;
2095 nfsfh_t nfh;
2096 fhandle_t *fhp;
2097 u_quad_t frev;
2098
2099 nd.ni_cnd.cn_nameiop = 0;
2100 fhp = &nfh.fh_generic;
2101 nfsm_srvmtofh(fhp);
2102 nfsm_srvnamesiz(len);
2103 nd.ni_cnd.cn_cred = cred;
2104 nd.ni_cnd.cn_nameiop = CREATE;
2105 nd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART;
2106 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2107 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2108 if (dirp) {
2109 if (v3)
2110 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2111 procp);
2112 else {
2113 vrele(dirp);
2114 dirp = (struct vnode *)0;
2115 }
2116 }
2117 if (error)
2118 goto out;
2119 VATTR_NULL(&va);
2120 if (v3)
2121 nfsm_srvsattr(&va);
2122 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2123 MALLOC(pathcp, caddr_t, len2 + 1, M_TEMP, M_WAITOK);
2124 iv.iov_base = pathcp;
2125 iv.iov_len = len2;
2126 io.uio_resid = len2;
2127 io.uio_offset = 0;
2128 io.uio_iov = &iv;
2129 io.uio_iovcnt = 1;
2130 io.uio_segflg = UIO_SYSSPACE;
2131 io.uio_rw = UIO_READ;
2132 io.uio_procp = (struct proc *)0;
2133 nfsm_mtouio(&io, len2);
2134 if (!v3) {
2135 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2136 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2137 }
2138 *(pathcp + len2) = '\0';
2139 if (nd.ni_vp) {
2140 vrele(nd.ni_startdir);
2141 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2142 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2143 if (nd.ni_dvp == nd.ni_vp)
2144 vrele(nd.ni_dvp);
2145 else
2146 vput(nd.ni_dvp);
2147 vrele(nd.ni_vp);
2148 error = EEXIST;
2149 goto out;
2150 }
2151 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2152 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2153 if (error)
2154 vrele(nd.ni_startdir);
2155 else {
2156 if (v3) {
2157 nd.ni_cnd.cn_nameiop = LOOKUP;
2158 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART | FOLLOW);
2159 nd.ni_cnd.cn_flags |= (NOFOLLOW | LOCKLEAF);
2160 nd.ni_cnd.cn_proc = procp;
2161 nd.ni_cnd.cn_cred = cred;
2162 error = lookup(&nd);
2163 if (!error) {
2164 bzero((caddr_t)fhp, sizeof(nfh));
2165 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
2166 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2167 if (!error)
2168 error = VOP_GETATTR(nd.ni_vp, &va, cred,
2169 procp);
2170 vput(nd.ni_vp);
2171 }
2172 } else
2173 vrele(nd.ni_startdir);
2174 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2175 }
2176 out:
2177 if (pathcp)
2178 FREE(pathcp, M_TEMP);
2179 if (dirp) {
2180 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2181 vrele(dirp);
2182 }
2183 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2184 if (v3) {
2185 if (!error) {
2186 nfsm_srvpostop_fh(fhp);
2187 nfsm_srvpostop_attr(0, &va);
2188 }
2189 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2190 }
2191 return (0);
2192 nfsmout:
2193 if (nd.ni_cnd.cn_nameiop) {
2194 vrele(nd.ni_startdir);
2195 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2196 }
2197 if (dirp)
2198 vrele(dirp);
2199 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2200 if (nd.ni_dvp == nd.ni_vp)
2201 vrele(nd.ni_dvp);
2202 else
2203 vput(nd.ni_dvp);
2204 if (nd.ni_vp)
2205 vrele(nd.ni_vp);
2206 if (pathcp)
2207 FREE(pathcp, M_TEMP);
2208 return (error);
2209 }
2210
2211 /*
2212 * nfs mkdir service
2213 */
2214 int
2215 nfsrv_mkdir(nfsd, slp, procp, mrq)
2216 struct nfsrv_descript *nfsd;
2217 struct nfssvc_sock *slp;
2218 struct proc *procp;
2219 struct mbuf **mrq;
2220 {
2221 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2222 struct mbuf *nam = nfsd->nd_nam;
2223 caddr_t dpos = nfsd->nd_dpos;
2224 struct ucred *cred = &nfsd->nd_cr;
2225 struct vattr va, dirfor, diraft;
2226 register struct nfs_fattr *fp;
2227 struct nameidata nd;
2228 register caddr_t cp;
2229 register u_int32_t *tl;
2230 register int32_t t1;
2231 caddr_t bpos;
2232 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2233 int v3 = (nfsd->nd_flag & ND_NFSV3);
2234 char *cp2;
2235 struct mbuf *mb, *mb2, *mreq;
2236 struct vnode *vp, *dirp = (struct vnode *)0;
2237 nfsfh_t nfh;
2238 fhandle_t *fhp;
2239 u_quad_t frev;
2240
2241 fhp = &nfh.fh_generic;
2242 nfsm_srvmtofh(fhp);
2243 nfsm_srvnamesiz(len);
2244 nd.ni_cnd.cn_cred = cred;
2245 nd.ni_cnd.cn_nameiop = CREATE;
2246 nd.ni_cnd.cn_flags = LOCKPARENT;
2247 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2248 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2249 if (dirp) {
2250 if (v3)
2251 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2252 procp);
2253 else {
2254 vrele(dirp);
2255 dirp = (struct vnode *)0;
2256 }
2257 }
2258 if (error) {
2259 nfsm_reply(NFSX_WCCDATA(v3));
2260 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2261 if (dirp)
2262 vrele(dirp);
2263 return (0);
2264 }
2265 VATTR_NULL(&va);
2266 if (v3) {
2267 nfsm_srvsattr(&va);
2268 } else {
2269 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2270 va.va_mode = nfstov_mode(*tl++);
2271 }
2272 va.va_type = VDIR;
2273 vp = nd.ni_vp;
2274 if (vp != NULL) {
2275 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2276 if (nd.ni_dvp == vp)
2277 vrele(nd.ni_dvp);
2278 else
2279 vput(nd.ni_dvp);
2280 vrele(vp);
2281 error = EEXIST;
2282 goto out;
2283 }
2284 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2285 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2286 if (!error) {
2287 vp = nd.ni_vp;
2288 bzero((caddr_t)fhp, sizeof(nfh));
2289 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
2290 error = VFS_VPTOFH(vp, &fhp->fh_fid);
2291 if (!error)
2292 error = VOP_GETATTR(vp, &va, cred, procp);
2293 vput(vp);
2294 }
2295 out:
2296 if (dirp) {
2297 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2298 vrele(dirp);
2299 }
2300 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2301 if (v3) {
2302 if (!error) {
2303 nfsm_srvpostop_fh(fhp);
2304 nfsm_srvpostop_attr(0, &va);
2305 }
2306 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2307 } else {
2308 nfsm_srvfhtom(fhp, v3);
2309 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2310 nfsm_srvfillattr(&va, fp);
2311 }
2312 return (0);
2313 nfsmout:
2314 if (dirp)
2315 vrele(dirp);
2316 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2317 if (nd.ni_dvp == nd.ni_vp)
2318 vrele(nd.ni_dvp);
2319 else
2320 vput(nd.ni_dvp);
2321 if (nd.ni_vp)
2322 vrele(nd.ni_vp);
2323 return (error);
2324 }
2325
2326 /*
2327 * nfs rmdir service
2328 */
2329 int
2330 nfsrv_rmdir(nfsd, slp, procp, mrq)
2331 struct nfsrv_descript *nfsd;
2332 struct nfssvc_sock *slp;
2333 struct proc *procp;
2334 struct mbuf **mrq;
2335 {
2336 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2337 struct mbuf *nam = nfsd->nd_nam;
2338 caddr_t dpos = nfsd->nd_dpos;
2339 struct ucred *cred = &nfsd->nd_cr;
2340 register u_int32_t *tl;
2341 register int32_t t1;
2342 caddr_t bpos;
2343 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2344 int v3 = (nfsd->nd_flag & ND_NFSV3);
2345 char *cp2;
2346 struct mbuf *mb, *mreq;
2347 struct vnode *vp, *dirp = (struct vnode *)0;
2348 struct vattr dirfor, diraft;
2349 nfsfh_t nfh;
2350 fhandle_t *fhp;
2351 struct nameidata nd;
2352 u_quad_t frev;
2353
2354 fhp = &nfh.fh_generic;
2355 nfsm_srvmtofh(fhp);
2356 nfsm_srvnamesiz(len);
2357 nd.ni_cnd.cn_cred = cred;
2358 nd.ni_cnd.cn_nameiop = DELETE;
2359 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2360 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2361 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2362 if (dirp) {
2363 if (v3)
2364 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2365 procp);
2366 else {
2367 vrele(dirp);
2368 dirp = (struct vnode *)0;
2369 }
2370 }
2371 if (error) {
2372 nfsm_reply(NFSX_WCCDATA(v3));
2373 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2374 if (dirp)
2375 vrele(dirp);
2376 return (0);
2377 }
2378 vp = nd.ni_vp;
2379 if (vp->v_type != VDIR) {
2380 error = ENOTDIR;
2381 goto out;
2382 }
2383 /*
2384 * No rmdir "." please.
2385 */
2386 if (nd.ni_dvp == vp) {
2387 error = EINVAL;
2388 goto out;
2389 }
2390 /*
2391 * The root of a mounted filesystem cannot be deleted.
2392 */
2393 if (vp->v_flag & VROOT)
2394 error = EBUSY;
2395 out:
2396 if (!error) {
2397 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2398 nqsrv_getl(vp, ND_WRITE);
2399 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2400 } else {
2401 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2402 if (nd.ni_dvp == nd.ni_vp)
2403 vrele(nd.ni_dvp);
2404 else
2405 vput(nd.ni_dvp);
2406 vput(vp);
2407 }
2408 if (dirp) {
2409 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2410 vrele(dirp);
2411 }
2412 nfsm_reply(NFSX_WCCDATA(v3));
2413 if (v3) {
2414 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2415 return (0);
2416 }
2417 nfsm_srvdone;
2418 }
2419
2420 /*
2421 * nfs readdir service
2422 * - mallocs what it thinks is enough to read
2423 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR
2424 * - calls VOP_READDIR()
2425 * - loops around building the reply
2426 * if the output generated exceeds count break out of loop
2427 * The nfsm_clget macro is used here so that the reply will be packed
2428 * tightly in mbuf clusters.
2429 * - it only knows that it has encountered eof when the VOP_READDIR()
2430 * reads nothing
2431 * - as such one readdir rpc will return eof false although you are there
2432 * and then the next will return eof
2433 * - it trims out records with d_fileno == 0
2434 * this doesn't matter for Unix clients, but they might confuse clients
2435 * for other os'.
2436 * - it trims out records with d_type == DT_WHT
2437 * these cannot be seen through NFS (unless we extend the protocol)
2438 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2439 * than requested, but this may not apply to all filesystems. For
2440 * example, client NFS does not { although it is never remote mounted
2441 * anyhow }
2442 * The alternate call nfsrv_readdirplus() does lookups as well.
2443 * PS: The NFS protocol spec. does not clarify what the "count" byte
2444 * argument is a count of.. just name strings and file id's or the
2445 * entire reply rpc or ...
2446 * I tried just file name and id sizes and it confused the Sun client,
2447 * so I am using the full rpc size now. The "paranoia.." comment refers
2448 * to including the status longwords that are not a part of the dir.
2449 * "entry" structures, but are in the rpc.
2450 */
2451 struct flrep {
2452 nfsuint64 fl_off;
2453 u_int32_t fl_postopok;
2454 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2455 u_int32_t fl_fhok;
2456 u_int32_t fl_fhsize;
2457 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2458 };
2459
2460 int
2461 nfsrv_readdir(nfsd, slp, procp, mrq)
2462 struct nfsrv_descript *nfsd;
2463 struct nfssvc_sock *slp;
2464 struct proc *procp;
2465 struct mbuf **mrq;
2466 {
2467 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2468 struct mbuf *nam = nfsd->nd_nam;
2469 caddr_t dpos = nfsd->nd_dpos;
2470 struct ucred *cred = &nfsd->nd_cr;
2471 register char *bp, *be;
2472 register struct mbuf *mp;
2473 register struct dirent *dp;
2474 register caddr_t cp;
2475 register u_int32_t *tl;
2476 register int32_t t1;
2477 caddr_t bpos;
2478 struct mbuf *mb, *mb2, *mreq, *mp2;
2479 char *cpos, *cend, *cp2, *rbuf;
2480 struct vnode *vp;
2481 struct vattr at;
2482 nfsfh_t nfh;
2483 fhandle_t *fhp;
2484 struct uio io;
2485 struct iovec iv;
2486 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2487 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies;
2488 int v3 = (nfsd->nd_flag & ND_NFSV3);
2489 u_quad_t frev, off, toff, verf;
2490 off_t *cookies = NULL, *cookiep;
2491 nfsuint64 jar;
2492
2493 fhp = &nfh.fh_generic;
2494 nfsm_srvmtofh(fhp);
2495 if (v3) {
2496 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2497 fxdr_hyper(tl, &toff);
2498 tl += 2;
2499 fxdr_hyper(tl, &verf);
2500 tl += 2;
2501 } else {
2502 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2503 toff = fxdr_unsigned(u_quad_t, *tl++);
2504 }
2505 off = toff;
2506 cnt = fxdr_unsigned(int, *tl);
2507 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2508 xfer = NFS_SRVMAXDATA(nfsd);
2509 if (siz > xfer)
2510 siz = xfer;
2511 fullsiz = siz;
2512 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2513 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2514 if (!error && vp->v_type != VDIR) {
2515 error = ENOTDIR;
2516 vput(vp);
2517 }
2518 if (error) {
2519 nfsm_reply(NFSX_UNSIGNED);
2520 nfsm_srvpostop_attr(getret, &at);
2521 return (0);
2522 }
2523 nqsrv_getl(vp, ND_READ);
2524 if (v3) {
2525 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2526 #ifdef NFS3_STRICTVERF
2527 /*
2528 * XXX This check is too strict for Solaris 2.5 clients.
2529 */
2530 if (!error && toff && verf != at.va_filerev)
2531 error = NFSERR_BAD_COOKIE;
2532 #endif
2533 }
2534 if (!error)
2535 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2536 if (error) {
2537 vput(vp);
2538 nfsm_reply(NFSX_POSTOPATTR(v3));
2539 nfsm_srvpostop_attr(getret, &at);
2540 return (0);
2541 }
2542 #ifdef Lite2_integrated
2543 VOP_UNLOCK(vp, 0, procp);
2544 #else
2545 VOP_UNLOCK(vp);
2546 #endif
2547 MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
2548 ncookies = siz / (5 * NFSX_UNSIGNED); /*7 for V3, but it's an est. so*/
2549 MALLOC(cookies, off_t *, ncookies * sizeof (off_t), M_TEMP,
2550 M_WAITOK);
2551 again:
2552 iv.iov_base = rbuf;
2553 iv.iov_len = fullsiz;
2554 io.uio_iov = &iv;
2555 io.uio_iovcnt = 1;
2556 io.uio_offset = (off_t)off;
2557 io.uio_resid = fullsiz;
2558 io.uio_segflg = UIO_SYSSPACE;
2559 io.uio_rw = UIO_READ;
2560 io.uio_procp = (struct proc *)0;
2561 eofflag = 0;
2562 #ifdef Lite2_integrated
2563 VOP_LOCK(vp, 0, procp);
2564 #else
2565 VOP_LOCK(vp);
2566 #endif
2567
2568 error = VOP_READDIR(vp, &io, cred, &eofflag, cookies, ncookies);
2569
2570 off = (off_t)io.uio_offset;
2571 if (!cookies && !error)
2572 error = NFSERR_PERM;
2573 if (v3) {
2574 getret = VOP_GETATTR(vp, &at, cred, procp);
2575 if (!error)
2576 error = getret;
2577 }
2578
2579 #ifdef Lite2_integrated
2580 VOP_UNLOCK(vp, 0, procp);
2581 #else
2582 VOP_UNLOCK(vp);
2583 #endif
2584 if (error) {
2585 vrele(vp);
2586 free((caddr_t)rbuf, M_TEMP);
2587 if (cookies)
2588 free((caddr_t)cookies, M_TEMP);
2589 nfsm_reply(NFSX_POSTOPATTR(v3));
2590 nfsm_srvpostop_attr(getret, &at);
2591 return (0);
2592 }
2593 if (io.uio_resid) {
2594 siz -= io.uio_resid;
2595
2596 /*
2597 * If nothing read, return eof
2598 * rpc reply
2599 */
2600 if (siz == 0) {
2601 vrele(vp);
2602 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2603 2 * NFSX_UNSIGNED);
2604 if (v3) {
2605 nfsm_srvpostop_attr(getret, &at);
2606 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2607 txdr_hyper(&at.va_filerev, tl);
2608 tl += 2;
2609 } else
2610 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2611 *tl++ = nfs_false;
2612 *tl = nfs_true;
2613 FREE((caddr_t)rbuf, M_TEMP);
2614 FREE((caddr_t)cookies, M_TEMP);
2615 return (0);
2616 }
2617 }
2618
2619 /*
2620 * Check for degenerate cases of nothing useful read.
2621 * If so go try again
2622 */
2623 cpos = rbuf;
2624 cend = rbuf + siz;
2625 dp = (struct dirent *)cpos;
2626 cookiep = cookies;
2627
2628 while (cpos < cend && ncookies > 0 &&
2629 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2630 cpos += dp->d_reclen;
2631 dp = (struct dirent *)cpos;
2632 cookiep++;
2633 ncookies--;
2634 }
2635 if (cpos >= cend || ncookies == 0) {
2636 toff = off;
2637 siz = fullsiz;
2638 goto again;
2639 }
2640
2641 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
2642 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2643 if (v3) {
2644 nfsm_srvpostop_attr(getret, &at);
2645 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2646 txdr_hyper(&at.va_filerev, tl);
2647 }
2648 mp = mp2 = mb;
2649 bp = bpos;
2650 be = bp + M_TRAILINGSPACE(mp);
2651
2652 /* Loop through the records and build reply */
2653 while (cpos < cend && ncookies > 0) {
2654 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2655 nlen = dp->d_namlen;
2656 rem = nfsm_rndup(nlen)-nlen;
2657 len += (4 * NFSX_UNSIGNED + nlen + rem);
2658 if (v3)
2659 len += 2 * NFSX_UNSIGNED;
2660 if (len > cnt) {
2661 eofflag = 0;
2662 break;
2663 }
2664 /*
2665 * Build the directory record xdr from
2666 * the dirent entry.
2667 */
2668 nfsm_clget;
2669 *tl = nfs_true;
2670 bp += NFSX_UNSIGNED;
2671 if (v3) {
2672 nfsm_clget;
2673 *tl = 0;
2674 bp += NFSX_UNSIGNED;
2675 }
2676 nfsm_clget;
2677 *tl = txdr_unsigned(dp->d_fileno);
2678 bp += NFSX_UNSIGNED;
2679 nfsm_clget;
2680 *tl = txdr_unsigned(nlen);
2681 bp += NFSX_UNSIGNED;
2682
2683 /* And loop around copying the name */
2684 xfer = nlen;
2685 cp = dp->d_name;
2686 while (xfer > 0) {
2687 nfsm_clget;
2688 if ((bp+xfer) > be)
2689 tsiz = be-bp;
2690 else
2691 tsiz = xfer;
2692 bcopy(cp, bp, tsiz);
2693 bp += tsiz;
2694 xfer -= tsiz;
2695 if (xfer > 0)
2696 cp += tsiz;
2697 }
2698 /* And null pad to an int32_t boundary */
2699 for (i = 0; i < rem; i++)
2700 *bp++ = '\0';
2701 nfsm_clget;
2702
2703 /* Finish off the record */
2704 txdr_hyper(cookiep, &jar);
2705 if (v3) {
2706 *tl = jar.nfsuquad[0];
2707 bp += NFSX_UNSIGNED;
2708 nfsm_clget;
2709 }
2710 *tl = jar.nfsuquad[1];
2711 bp += NFSX_UNSIGNED;
2712 }
2713 cpos += dp->d_reclen;
2714 dp = (struct dirent *)cpos;
2715 cookiep++;
2716 ncookies--;
2717 }
2718 vrele(vp);
2719 nfsm_clget;
2720 *tl = nfs_false;
2721 bp += NFSX_UNSIGNED;
2722 nfsm_clget;
2723 if (eofflag)
2724 *tl = nfs_true;
2725 else
2726 *tl = nfs_false;
2727 bp += NFSX_UNSIGNED;
2728 if (mp != mb) {
2729 if (bp < be)
2730 mp->m_len = bp - mtod(mp, caddr_t);
2731 } else
2732 mp->m_len += bp - bpos;
2733 FREE((caddr_t)rbuf, M_TEMP);
2734 FREE((caddr_t)cookies, M_TEMP);
2735 nfsm_srvdone;
2736 }
2737
2738 int
2739 nfsrv_readdirplus(nfsd, slp, procp, mrq)
2740 struct nfsrv_descript *nfsd;
2741 struct nfssvc_sock *slp;
2742 struct proc *procp;
2743 struct mbuf **mrq;
2744 {
2745 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2746 struct mbuf *nam = nfsd->nd_nam;
2747 caddr_t dpos = nfsd->nd_dpos;
2748 struct ucred *cred = &nfsd->nd_cr;
2749 register char *bp, *be;
2750 register struct mbuf *mp;
2751 register struct dirent *dp;
2752 register caddr_t cp;
2753 register u_int32_t *tl;
2754 register int32_t t1;
2755 caddr_t bpos;
2756 struct mbuf *mb, *mb2, *mreq, *mp2;
2757 char *cpos, *cend, *cp2, *rbuf;
2758 struct vnode *vp, *nvp;
2759 struct flrep fl;
2760 nfsfh_t nfh;
2761 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
2762 struct uio io;
2763 struct iovec iv;
2764 struct vattr va, at, *vap = &va;
2765 struct nfs_fattr *fp;
2766 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2767 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies;
2768 u_quad_t frev, off, toff, verf;
2769 off_t *cookies = NULL, *cookiep;
2770
2771 fhp = &nfh.fh_generic;
2772 nfsm_srvmtofh(fhp);
2773 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2774 fxdr_hyper(tl, &toff);
2775 tl += 2;
2776 fxdr_hyper(tl, &verf);
2777 tl += 2;
2778 siz = fxdr_unsigned(int, *tl++);
2779 cnt = fxdr_unsigned(int, *tl);
2780 off = toff;
2781 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2782 xfer = NFS_SRVMAXDATA(nfsd);
2783 if (siz > xfer)
2784 siz = xfer;
2785 fullsiz = siz;
2786 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2787 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2788 if (!error && vp->v_type != VDIR) {
2789 error = ENOTDIR;
2790 vput(vp);
2791 }
2792 if (error) {
2793 nfsm_reply(NFSX_UNSIGNED);
2794 nfsm_srvpostop_attr(getret, &at);
2795 return (0);
2796 }
2797 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2798 #ifdef NFS3_STRICTVERF
2799 /*
2800 * XXX This check is too strict for Solaris 2.5 clients.
2801 */
2802 if (!error && toff && verf != at.va_filerev)
2803 error = NFSERR_BAD_COOKIE;
2804 #endif
2805 if (!error) {
2806 nqsrv_getl(vp, ND_READ);
2807 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2808 }
2809 if (error) {
2810 vput(vp);
2811 nfsm_reply(NFSX_V3POSTOPATTR);
2812 nfsm_srvpostop_attr(getret, &at);
2813 return (0);
2814 }
2815 #ifdef Lite2_integrated
2816 VOP_UNLOCK(vp, 0, procp);
2817 #else
2818 VOP_UNLOCK(vp);
2819 #endif
2820
2821 MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
2822 ncookies = siz / (7 * NFSX_UNSIGNED);
2823 MALLOC(cookies, off_t *, ncookies * sizeof (off_t), M_TEMP,
2824 M_WAITOK);
2825 again:
2826 iv.iov_base = rbuf;
2827 iv.iov_len = fullsiz;
2828 io.uio_iov = &iv;
2829 io.uio_iovcnt = 1;
2830 io.uio_offset = (off_t)off;
2831 io.uio_resid = fullsiz;
2832 io.uio_segflg = UIO_SYSSPACE;
2833 io.uio_rw = UIO_READ;
2834 io.uio_procp = (struct proc *)0;
2835 eofflag = 0;
2836
2837 #ifdef Lite2_integrated
2838 VOP_LOCK(vp, 0, procp);
2839 #else
2840 VOP_LOCK(vp);
2841 #endif
2842 error = VOP_READDIR(vp, &io, cred, &eofflag, cookies, ncookies);
2843
2844 off = (u_quad_t)io.uio_offset;
2845 getret = VOP_GETATTR(vp, &at, cred, procp);
2846
2847 #ifdef Lite2_integrated
2848 VOP_UNLOCK(vp, 0, procp);
2849 #else
2850 VOP_UNLOCK(vp);
2851 #endif
2852
2853 /*
2854 * If the VGET operation doesn't work for this filesystem,
2855 * we can't support readdirplus. Returning NOTSUPP should
2856 * make clients fall back to plain readdir.
2857 * There's no need to check for VPTOFH as well, we wouldn't
2858 * even be here otherwise.
2859 */
2860 if (!getret) {
2861 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
2862 getret = (getret == EOPNOTSUPP) ?
2863 NFSERR_NOTSUPP : NFSERR_IO;
2864 else
2865 vput(nvp);
2866 }
2867
2868 if (!cookies && !error)
2869 error = NFSERR_PERM;
2870 if (!error)
2871 error = getret;
2872 if (error) {
2873 vrele(vp);
2874 if (cookies)
2875 free((caddr_t)cookies, M_TEMP);
2876 free((caddr_t)rbuf, M_TEMP);
2877 nfsm_reply(NFSX_V3POSTOPATTR);
2878 nfsm_srvpostop_attr(getret, &at);
2879 return (0);
2880 }
2881 if (io.uio_resid) {
2882 siz -= io.uio_resid;
2883
2884 /*
2885 * If nothing read, return eof
2886 * rpc reply
2887 */
2888 if (siz == 0) {
2889 vrele(vp);
2890 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
2891 2 * NFSX_UNSIGNED);
2892 nfsm_srvpostop_attr(getret, &at);
2893 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2894 txdr_hyper(&at.va_filerev, tl);
2895 tl += 2;
2896 *tl++ = nfs_false;
2897 *tl = nfs_true;
2898 FREE((caddr_t)cookies, M_TEMP);
2899 FREE((caddr_t)rbuf, M_TEMP);
2900 return (0);
2901 }
2902 }
2903
2904 /*
2905 * Check for degenerate cases of nothing useful read.
2906 * If so go try again
2907 */
2908 cpos = rbuf;
2909 cend = rbuf + siz;
2910 dp = (struct dirent *)cpos;
2911 cookiep = cookies;
2912
2913 while (cpos < cend && ncookies > 0 &&
2914 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2915 cpos += dp->d_reclen;
2916 dp = (struct dirent *)cpos;
2917 cookiep++;
2918 ncookies--;
2919 }
2920 if (cpos >= cend || ncookies == 0) {
2921 toff = off;
2922 siz = fullsiz;
2923 goto again;
2924 }
2925
2926 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
2927 nfsm_reply(cnt);
2928 nfsm_srvpostop_attr(getret, &at);
2929 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2930 txdr_hyper(&at.va_filerev, tl);
2931 mp = mp2 = mb;
2932 bp = bpos;
2933 be = bp + M_TRAILINGSPACE(mp);
2934
2935 /* Loop through the records and build reply */
2936 while (cpos < cend && ncookies > 0) {
2937 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2938 nlen = dp->d_namlen;
2939 rem = nfsm_rndup(nlen)-nlen;
2940
2941 /*
2942 * For readdir_and_lookup get the vnode using
2943 * the file number.
2944 */
2945 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
2946 goto invalid;
2947 bzero((caddr_t)nfhp, NFSX_V3FH);
2948 nfhp->fh_fsid =
2949 nvp->v_mount->mnt_stat.f_fsid;
2950 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
2951 vput(nvp);
2952 goto invalid;
2953 }
2954 if (VOP_GETATTR(nvp, vap, cred, procp)) {
2955 vput(nvp);
2956 goto invalid;
2957 }
2958 vput(nvp);
2959
2960 /*
2961 * If either the dircount or maxcount will be
2962 * exceeded, get out now. Both of these lengths
2963 * are calculated conservatively, including all
2964 * XDR overheads.
2965 */
2966 len += (7 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
2967 NFSX_V3POSTOPATTR);
2968 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
2969 if (len > cnt || dirlen > fullsiz) {
2970 eofflag = 0;
2971 break;
2972 }
2973
2974 /*
2975 * Build the directory record xdr from
2976 * the dirent entry.
2977 */
2978 fp = (struct nfs_fattr *)&fl.fl_fattr;
2979 nfsm_srvfillattr(vap, fp);
2980 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
2981 fl.fl_fhok = nfs_true;
2982 fl.fl_postopok = nfs_true;
2983 txdr_hyper(cookiep, fl.fl_off.nfsuquad);
2984
2985 nfsm_clget;
2986 *tl = nfs_true;
2987 bp += NFSX_UNSIGNED;
2988 nfsm_clget;
2989 *tl = 0;
2990 bp += NFSX_UNSIGNED;
2991 nfsm_clget;
2992 *tl = txdr_unsigned(dp->d_fileno);
2993 bp += NFSX_UNSIGNED;
2994 nfsm_clget;
2995 *tl = txdr_unsigned(nlen);
2996 bp += NFSX_UNSIGNED;
2997
2998 /* And loop around copying the name */
2999 xfer = nlen;
3000 cp = dp->d_name;
3001 while (xfer > 0) {
3002 nfsm_clget;
3003 if ((bp + xfer) > be)
3004 tsiz = be - bp;
3005 else
3006 tsiz = xfer;
3007 bcopy(cp, bp, tsiz);
3008 bp += tsiz;
3009 xfer -= tsiz;
3010 if (xfer > 0)
3011 cp += tsiz;
3012 }
3013 /* And null pad to an int32_t boundary */
3014 for (i = 0; i < rem; i++)
3015 *bp++ = '\0';
3016
3017 /*
3018 * Now copy the flrep structure out.
3019 */
3020 xfer = sizeof (struct flrep);
3021 cp = (caddr_t)&fl;
3022 while (xfer > 0) {
3023 nfsm_clget;
3024 if ((bp + xfer) > be)
3025 tsiz = be - bp;
3026 else
3027 tsiz = xfer;
3028 bcopy(cp, bp, tsiz);
3029 bp += tsiz;
3030 xfer -= tsiz;
3031 if (xfer > 0)
3032 cp += tsiz;
3033 }
3034 }
3035 invalid:
3036 cpos += dp->d_reclen;
3037 dp = (struct dirent *)cpos;
3038 cookiep++;
3039 ncookies--;
3040 }
3041 vrele(vp);
3042 nfsm_clget;
3043 *tl = nfs_false;
3044 bp += NFSX_UNSIGNED;
3045 nfsm_clget;
3046 if (eofflag)
3047 *tl = nfs_true;
3048 else
3049 *tl = nfs_false;
3050 bp += NFSX_UNSIGNED;
3051 if (mp != mb) {
3052 if (bp < be)
3053 mp->m_len = bp - mtod(mp, caddr_t);
3054 } else
3055 mp->m_len += bp - bpos;
3056 FREE((caddr_t)cookies, M_TEMP);
3057 FREE((caddr_t)rbuf, M_TEMP);
3058 nfsm_srvdone;
3059 }
3060
3061 /*
3062 * nfs commit service
3063 */
3064 int
3065 nfsrv_commit(nfsd, slp, procp, mrq)
3066 struct nfsrv_descript *nfsd;
3067 struct nfssvc_sock *slp;
3068 struct proc *procp;
3069 struct mbuf **mrq;
3070 {
3071 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3072 struct mbuf *nam = nfsd->nd_nam;
3073 caddr_t dpos = nfsd->nd_dpos;
3074 struct ucred *cred = &nfsd->nd_cr;
3075 struct vattr bfor, aft;
3076 struct vnode *vp;
3077 nfsfh_t nfh;
3078 fhandle_t *fhp;
3079 register u_int32_t *tl;
3080 register int32_t t1;
3081 caddr_t bpos;
3082 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache;
3083 char *cp2;
3084 struct mbuf *mb, *mb2, *mreq;
3085 u_quad_t frev, off;
3086
3087 #ifndef nolint
3088 cache = 0;
3089 #endif
3090 fhp = &nfh.fh_generic;
3091 nfsm_srvmtofh(fhp);
3092 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3093
3094 /*
3095 * XXX At this time VOP_FSYNC() does not accept offset and byte
3096 * count parameters, so these arguments are useless (someday maybe).
3097 */
3098 fxdr_hyper(tl, &off);
3099 tl += 2;
3100 cnt = fxdr_unsigned(int, *tl);
3101 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3102 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3103 if (error) {
3104 nfsm_reply(2 * NFSX_UNSIGNED);
3105 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3106 return (0);
3107 }
3108 for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
3109 error = VOP_FSYNC(vp, cred, MNT_WAIT, procp);
3110 aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
3111 vput(vp);
3112 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3113 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3114 if (!error) {
3115 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3116 *tl++ = txdr_unsigned(boottime.tv_sec);
3117 *tl = txdr_unsigned(boottime.tv_usec);
3118 } else
3119 return (0);
3120 nfsm_srvdone;
3121 }
3122
3123 /*
3124 * nfs statfs service
3125 */
3126 int
3127 nfsrv_statfs(nfsd, slp, procp, mrq)
3128 struct nfsrv_descript *nfsd;
3129 struct nfssvc_sock *slp;
3130 struct proc *procp;
3131 struct mbuf **mrq;
3132 {
3133 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3134 struct mbuf *nam = nfsd->nd_nam;
3135 caddr_t dpos = nfsd->nd_dpos;
3136 struct ucred *cred = &nfsd->nd_cr;
3137 register struct statfs *sf;
3138 register struct nfs_statfs *sfp;
3139 register u_int32_t *tl;
3140 register int32_t t1;
3141 caddr_t bpos;
3142 int error = 0, rdonly, cache, getret = 1;
3143 int v3 = (nfsd->nd_flag & ND_NFSV3);
3144 char *cp2;
3145 struct mbuf *mb, *mb2, *mreq;
3146 struct vnode *vp;
3147 struct vattr at;
3148 nfsfh_t nfh;
3149 fhandle_t *fhp;
3150 struct statfs statfs;
3151 u_quad_t frev, tval;
3152
3153 #ifndef nolint
3154 cache = 0;
3155 #endif
3156 fhp = &nfh.fh_generic;
3157 nfsm_srvmtofh(fhp);
3158 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3159 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3160 if (error) {
3161 nfsm_reply(NFSX_UNSIGNED);
3162 nfsm_srvpostop_attr(getret, &at);
3163 return (0);
3164 }
3165 sf = &statfs;
3166 error = VFS_STATFS(vp->v_mount, sf, procp);
3167 getret = VOP_GETATTR(vp, &at, cred, procp);
3168 vput(vp);
3169 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3170 if (v3)
3171 nfsm_srvpostop_attr(getret, &at);
3172 if (error)
3173 return (0);
3174 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3175 if (v3) {
3176 tval = (u_quad_t)sf->f_blocks;
3177 tval *= (u_quad_t)sf->f_bsize;
3178 txdr_hyper(&tval, &sfp->sf_tbytes);
3179 tval = (u_quad_t)sf->f_bfree;
3180 tval *= (u_quad_t)sf->f_bsize;
3181 txdr_hyper(&tval, &sfp->sf_fbytes);
3182 tval = (u_quad_t)sf->f_bavail;
3183 tval *= (u_quad_t)sf->f_bsize;
3184 txdr_hyper(&tval, &sfp->sf_abytes);
3185 sfp->sf_tfiles.nfsuquad[0] = 0;
3186 sfp->sf_tfiles.nfsuquad[1] = txdr_unsigned(sf->f_files);
3187 sfp->sf_ffiles.nfsuquad[0] = 0;
3188 sfp->sf_ffiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree);
3189 sfp->sf_afiles.nfsuquad[0] = 0;
3190 sfp->sf_afiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree);
3191 sfp->sf_invarsec = 0;
3192 } else {
3193 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3194 sfp->sf_bsize = txdr_unsigned(sf->f_bsize);
3195 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3196 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3197 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3198 }
3199 nfsm_srvdone;
3200 }
3201
3202 /*
3203 * nfs fsinfo service
3204 */
3205 int
3206 nfsrv_fsinfo(nfsd, slp, procp, mrq)
3207 struct nfsrv_descript *nfsd;
3208 struct nfssvc_sock *slp;
3209 struct proc *procp;
3210 struct mbuf **mrq;
3211 {
3212 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3213 struct mbuf *nam = nfsd->nd_nam;
3214 caddr_t dpos = nfsd->nd_dpos;
3215 struct ucred *cred = &nfsd->nd_cr;
3216 register u_int32_t *tl;
3217 register struct nfsv3_fsinfo *sip;
3218 register int32_t t1;
3219 caddr_t bpos;
3220 int error = 0, rdonly, cache, getret = 1, pref;
3221 char *cp2;
3222 struct mbuf *mb, *mb2, *mreq;
3223 struct vnode *vp;
3224 struct vattr at;
3225 nfsfh_t nfh;
3226 fhandle_t *fhp;
3227 u_quad_t frev, maxfsize;
3228 struct statfs sb;
3229
3230 #ifndef nolint
3231 cache = 0;
3232 #endif
3233 fhp = &nfh.fh_generic;
3234 nfsm_srvmtofh(fhp);
3235 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3236 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3237 if (error) {
3238 nfsm_reply(NFSX_UNSIGNED);
3239 nfsm_srvpostop_attr(getret, &at);
3240 return (0);
3241 }
3242
3243 /* XXX Try to make a guess on the max file size. */
3244 VFS_STATFS(vp->v_mount, &sb, (struct proc *)0);
3245 maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1;
3246
3247 getret = VOP_GETATTR(vp, &at, cred, procp);
3248 vput(vp);
3249 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3250 nfsm_srvpostop_attr(getret, &at);
3251 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3252
3253 /*
3254 * XXX
3255 * There should be file system VFS OP(s) to get this information.
3256 * For now, assume ufs.
3257 */
3258 if (slp->ns_so->so_type == SOCK_DGRAM)
3259 pref = NFS_MAXDGRAMDATA;
3260 else
3261 pref = NFS_MAXDATA;
3262 sip->fs_rtmax = txdr_unsigned(NFS_MAXDATA);
3263 sip->fs_rtpref = txdr_unsigned(pref);
3264 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3265 sip->fs_wtmax = txdr_unsigned(NFS_MAXDATA);
3266 sip->fs_wtpref = txdr_unsigned(pref);
3267 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3268 sip->fs_dtpref = txdr_unsigned(pref);
3269 txdr_hyper(&maxfsize, &sip->fs_maxfilesize);
3270 sip->fs_timedelta.nfsv3_sec = 0;
3271 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3272 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3273 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3274 NFSV3FSINFO_CANSETTIME);
3275 nfsm_srvdone;
3276 }
3277
3278 /*
3279 * nfs pathconf service
3280 */
3281 int
3282 nfsrv_pathconf(nfsd, slp, procp, mrq)
3283 struct nfsrv_descript *nfsd;
3284 struct nfssvc_sock *slp;
3285 struct proc *procp;
3286 struct mbuf **mrq;
3287 {
3288 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3289 struct mbuf *nam = nfsd->nd_nam;
3290 caddr_t dpos = nfsd->nd_dpos;
3291 struct ucred *cred = &nfsd->nd_cr;
3292 register u_int32_t *tl;
3293 register struct nfsv3_pathconf *pc;
3294 register int32_t t1;
3295 caddr_t bpos;
3296 int error = 0, rdonly, cache, getret = 1;
3297 register_t linkmax, namemax, chownres, notrunc;
3298 char *cp2;
3299 struct mbuf *mb, *mb2, *mreq;
3300 struct vnode *vp;
3301 struct vattr at;
3302 nfsfh_t nfh;
3303 fhandle_t *fhp;
3304 u_quad_t frev;
3305
3306 #ifndef nolint
3307 cache = 0;
3308 #endif
3309 fhp = &nfh.fh_generic;
3310 nfsm_srvmtofh(fhp);
3311 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3312 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3313 if (error) {
3314 nfsm_reply(NFSX_UNSIGNED);
3315 nfsm_srvpostop_attr(getret, &at);
3316 return (0);
3317 }
3318 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3319 if (!error)
3320 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3321 if (!error)
3322 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3323 if (!error)
3324 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc);
3325 getret = VOP_GETATTR(vp, &at, cred, procp);
3326 vput(vp);
3327 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3328 nfsm_srvpostop_attr(getret, &at);
3329 if (error)
3330 return (0);
3331 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3332
3333 pc->pc_linkmax = txdr_unsigned(linkmax);
3334 pc->pc_namemax = txdr_unsigned(namemax);
3335 pc->pc_notrunc = txdr_unsigned(notrunc);
3336 pc->pc_chownrestricted = txdr_unsigned(chownres);
3337
3338 /*
3339 * These should probably be supported by VOP_PATHCONF(), but
3340 * until msdosfs is exportable (why would you want to?), the
3341 * Unix defaults should be ok.
3342 */
3343 pc->pc_caseinsensitive = nfs_false;
3344 pc->pc_casepreserving = nfs_true;
3345 nfsm_srvdone;
3346 }
3347
3348 /*
3349 * Null operation, used by clients to ping server
3350 */
3351 /* ARGSUSED */
3352 int
3353 nfsrv_null(nfsd, slp, procp, mrq)
3354 struct nfsrv_descript *nfsd;
3355 struct nfssvc_sock *slp;
3356 struct proc *procp;
3357 struct mbuf **mrq;
3358 {
3359 struct mbuf *mrep = nfsd->nd_mrep;
3360 caddr_t bpos;
3361 int error = NFSERR_RETVOID, cache = 0;
3362 struct mbuf *mb, *mreq;
3363 u_quad_t frev;
3364
3365 nfsm_reply(0);
3366 return (0);
3367 }
3368
3369 /*
3370 * No operation, used for obsolete procedures
3371 */
3372 /* ARGSUSED */
3373 int
3374 nfsrv_noop(nfsd, slp, procp, mrq)
3375 struct nfsrv_descript *nfsd;
3376 struct nfssvc_sock *slp;
3377 struct proc *procp;
3378 struct mbuf **mrq;
3379 {
3380 struct mbuf *mrep = nfsd->nd_mrep;
3381 caddr_t bpos;
3382 int error, cache = 0;
3383 struct mbuf *mb, *mreq;
3384 u_quad_t frev;
3385
3386 if (nfsd->nd_repstat)
3387 error = nfsd->nd_repstat;
3388 else
3389 error = EPROCUNAVAIL;
3390 nfsm_reply(0);
3391 return (0);
3392 }
3393
3394 /*
3395 * Perform access checking for vnodes obtained from file handles that would
3396 * refer to files already opened by a Unix client. You cannot just use
3397 * vn_writechk() and VOP_ACCESS() for two reasons.
3398 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3399 * 2 - The owner is to be given access irrespective of mode bits for some
3400 * operations, so that processes that chmod after opening a file don't
3401 * break. I don't like this because it opens a security hole, but since
3402 * the nfs server opens a security hole the size of a barn door anyhow,
3403 * what the heck.
3404 *
3405 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3406 * will return EPERM instead of EACCESS. EPERM is always an error.
3407 */
3408 int
3409 nfsrv_access(vp, flags, cred, rdonly, p, override)
3410 register struct vnode *vp;
3411 int flags;
3412 register struct ucred *cred;
3413 int rdonly;
3414 struct proc *p;
3415 {
3416 struct vattr vattr;
3417 int error;
3418 if (flags & VWRITE) {
3419 /* Just vn_writechk() changed to check rdonly */
3420 /*
3421 * Disallow write attempts on read-only file systems;
3422 * unless the file is a socket or a block or character
3423 * device resident on the file system.
3424 */
3425 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3426 switch (vp->v_type) {
3427 case VREG:
3428 case VDIR:
3429 case VLNK:
3430 return (EROFS);
3431 default:
3432 break;
3433 }
3434 }
3435 /*
3436 * If there's shared text associated with
3437 * the inode, try to free it up once. If
3438 * we fail, we can't allow writing.
3439 */
3440 #if defined(UVM)
3441 if ((vp->v_flag & VTEXT) && !uvm_vnp_uncache(vp))
3442 return (ETXTBSY);
3443 #else
3444 if ((vp->v_flag & VTEXT) && !vnode_pager_uncache(vp))
3445 return (ETXTBSY);
3446 #endif
3447 }
3448 error = VOP_GETATTR(vp, &vattr, cred, p);
3449 if (error)
3450 return (error);
3451 error = VOP_ACCESS(vp, flags, cred, p);
3452 /*
3453 * Allow certain operations for the owner (reads and writes
3454 * on files that are already open).
3455 */
3456 if (override && error == EACCES && cred->cr_uid == vattr.va_uid)
3457 error = 0;
3458 return error;
3459 }
3460