nfs_serv.c revision 1.55 1 /* $NetBSD: nfs_serv.c,v 1.55 2000/06/27 17:52:32 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 8.8 (Berkeley) 7/31/95
39 */
40
41 /*
42 * nfs version 2 and 3 server calls to vnode ops
43 * - these routines generally have 3 phases
44 * 1 - break down and validate rpc request in mbuf list
45 * 2 - do the vnode ops for the request
46 * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c)
47 * 3 - build the rpc reply in an mbuf list
48 * nb:
49 * - do not mix the phases, since the nfsm_?? macros can return failures
50 * on a bad rpc or similar and do not do any vrele() or vput()'s
51 *
52 * - the nfsm_reply() macro generates an nfs rpc reply with the nfs
53 * error number iff error != 0 whereas
54 * returning an error from the server function implies a fatal error
55 * such as a badly constructed rpc request that should be dropped without
56 * a reply.
57 * For Version 3, nfsm_reply() does not return for the error case, since
58 * most version 3 rpcs return more than the status for error cases.
59 */
60
61 #include <sys/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 <uvm/uvm_extern.h>
77
78 #include <nfs/nfsproto.h>
79 #include <nfs/rpcv2.h>
80 #include <nfs/nfs.h>
81 #include <nfs/xdr_subs.h>
82 #include <nfs/nfsm_subs.h>
83 #include <nfs/nqnfs.h>
84 #include <nfs/nfs_var.h>
85
86 /* Global vars */
87 extern u_int32_t nfs_xdrneg1;
88 extern u_int32_t nfs_false, nfs_true;
89 extern enum vtype nv3tov_type[8];
90 extern struct nfsstats nfsstats;
91 extern nfstype nfsv2_type[9];
92 extern nfstype nfsv3_type[9];
93 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
94
95 /*
96 * nfs v3 access service
97 */
98 int
99 nfsrv3_access(nfsd, slp, procp, mrq)
100 struct nfsrv_descript *nfsd;
101 struct nfssvc_sock *slp;
102 struct proc *procp;
103 struct mbuf **mrq;
104 {
105 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
106 struct mbuf *nam = nfsd->nd_nam;
107 caddr_t dpos = nfsd->nd_dpos;
108 struct ucred *cred = &nfsd->nd_cr;
109 struct vnode *vp;
110 nfsfh_t nfh;
111 fhandle_t *fhp;
112 u_int32_t *tl;
113 int32_t t1;
114 caddr_t bpos;
115 int error = 0, rdonly, cache = 0, getret;
116 char *cp2;
117 struct mbuf *mb, *mreq, *mb2;
118 struct vattr va;
119 u_long inmode, testmode, outmode;
120 u_quad_t frev;
121
122 fhp = &nfh.fh_generic;
123 nfsm_srvmtofh(fhp);
124 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
125 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
126 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
127 if (error) {
128 nfsm_reply(NFSX_UNSIGNED);
129 nfsm_srvpostop_attr(1, (struct vattr *)0);
130 return (0);
131 }
132 inmode = fxdr_unsigned(u_int32_t, *tl);
133 outmode = 0;
134 if ((inmode & NFSV3ACCESS_READ) &&
135 nfsrv_access(vp, VREAD, cred, rdonly, procp, 0) == 0)
136 outmode |= NFSV3ACCESS_READ;
137 if (vp->v_type != VDIR) {
138 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
139 if (testmode &&
140 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0)
141 outmode |= testmode;
142 if ((inmode & NFSV3ACCESS_EXECUTE) &&
143 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
144 outmode |= NFSV3ACCESS_EXECUTE;
145 } else {
146 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
147 NFSV3ACCESS_DELETE);
148 if (testmode &&
149 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0)
150 outmode |= testmode;
151 if ((inmode & NFSV3ACCESS_LOOKUP) &&
152 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
153 outmode |= NFSV3ACCESS_LOOKUP;
154 }
155 getret = VOP_GETATTR(vp, &va, cred, procp);
156 vput(vp);
157 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
158 nfsm_srvpostop_attr(getret, &va);
159 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
160 *tl = txdr_unsigned(outmode);
161 nfsm_srvdone;
162 }
163
164 /*
165 * nfs getattr service
166 */
167 int
168 nfsrv_getattr(nfsd, slp, procp, mrq)
169 struct nfsrv_descript *nfsd;
170 struct nfssvc_sock *slp;
171 struct proc *procp;
172 struct mbuf **mrq;
173 {
174 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
175 struct mbuf *nam = nfsd->nd_nam;
176 caddr_t dpos = nfsd->nd_dpos;
177 struct ucred *cred = &nfsd->nd_cr;
178 struct nfs_fattr *fp;
179 struct vattr va;
180 struct vnode *vp;
181 nfsfh_t nfh;
182 fhandle_t *fhp;
183 u_int32_t *tl;
184 int32_t t1;
185 caddr_t bpos;
186 int error = 0, rdonly, cache;
187 char *cp2;
188 struct mbuf *mb, *mb2, *mreq;
189 u_quad_t frev;
190
191 fhp = &nfh.fh_generic;
192 nfsm_srvmtofh(fhp);
193 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
194 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
195 if (error) {
196 nfsm_reply(0);
197 return (0);
198 }
199 nqsrv_getl(vp, ND_READ);
200 error = VOP_GETATTR(vp, &va, cred, procp);
201 vput(vp);
202 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
203 if (error)
204 return (0);
205 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
206 nfsm_srvfillattr(&va, fp);
207 nfsm_srvdone;
208 }
209
210 /*
211 * nfs setattr service
212 */
213 int
214 nfsrv_setattr(nfsd, slp, procp, mrq)
215 struct nfsrv_descript *nfsd;
216 struct nfssvc_sock *slp;
217 struct proc *procp;
218 struct mbuf **mrq;
219 {
220 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
221 struct mbuf *nam = nfsd->nd_nam;
222 caddr_t dpos = nfsd->nd_dpos;
223 struct ucred *cred = &nfsd->nd_cr;
224 struct vattr va, preat;
225 struct nfsv2_sattr *sp;
226 struct nfs_fattr *fp;
227 struct vnode *vp;
228 nfsfh_t nfh;
229 fhandle_t *fhp;
230 u_int32_t *tl;
231 int32_t t1;
232 caddr_t bpos;
233 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1;
234 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
235 char *cp2;
236 struct mbuf *mb, *mb2, *mreq;
237 u_quad_t frev;
238 struct timespec guard;
239
240 fhp = &nfh.fh_generic;
241 nfsm_srvmtofh(fhp);
242 VATTR_NULL(&va);
243 if (v3) {
244 nfsm_srvsattr(&va);
245 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
246 gcheck = fxdr_unsigned(int, *tl);
247 if (gcheck) {
248 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
249 fxdr_nfsv3time(tl, &guard);
250 }
251 } else {
252 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
253 /*
254 * Nah nah nah nah na nah
255 * There is a bug in the Sun client that puts 0xffff in the mode
256 * field of sattr when it should put in 0xffffffff. The u_short
257 * doesn't sign extend.
258 * --> check the low order 2 bytes for 0xffff
259 */
260 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
261 va.va_mode = nfstov_mode(sp->sa_mode);
262 if (sp->sa_uid != nfs_xdrneg1)
263 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
264 if (sp->sa_gid != nfs_xdrneg1)
265 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
266 if (sp->sa_size != nfs_xdrneg1)
267 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
268 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
269 #ifdef notyet
270 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime);
271 #else
272 va.va_atime.tv_sec =
273 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
274 va.va_atime.tv_nsec = 0;
275 #endif
276 }
277 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
278 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime);
279
280 }
281
282 /*
283 * Now that we have all the fields, lets do it.
284 */
285 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
286 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
287 if (error) {
288 nfsm_reply(2 * NFSX_UNSIGNED);
289 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
290 return (0);
291 }
292 nqsrv_getl(vp, ND_WRITE);
293 if (v3) {
294 error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp);
295 if (!error && gcheck &&
296 (preat.va_ctime.tv_sec != guard.tv_sec ||
297 preat.va_ctime.tv_nsec != guard.tv_nsec))
298 error = NFSERR_NOT_SYNC;
299 if (error) {
300 vput(vp);
301 nfsm_reply(NFSX_WCCDATA(v3));
302 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
303 return (0);
304 }
305 }
306
307 /*
308 * If the size is being changed write acces is required, otherwise
309 * just check for a read only file system.
310 */
311 if (va.va_size == ((u_quad_t)((quad_t) -1))) {
312 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
313 error = EROFS;
314 goto out;
315 }
316 } else {
317 if (vp->v_type == VDIR) {
318 error = EISDIR;
319 goto out;
320 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
321 procp, 0)) != 0)
322 goto out;
323 }
324 error = VOP_SETATTR(vp, &va, cred, procp);
325 postat_ret = VOP_GETATTR(vp, &va, cred, procp);
326 if (!error)
327 error = postat_ret;
328 out:
329 vput(vp);
330 nfsm_reply(NFSX_WCCORFATTR(v3));
331 if (v3) {
332 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
333 return (0);
334 } else {
335 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
336 nfsm_srvfillattr(&va, fp);
337 }
338 nfsm_srvdone;
339 }
340
341 /*
342 * nfs lookup rpc
343 */
344 int
345 nfsrv_lookup(nfsd, slp, procp, mrq)
346 struct nfsrv_descript *nfsd;
347 struct nfssvc_sock *slp;
348 struct proc *procp;
349 struct mbuf **mrq;
350 {
351 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
352 struct mbuf *nam = nfsd->nd_nam;
353 caddr_t dpos = nfsd->nd_dpos;
354 struct ucred *cred = &nfsd->nd_cr;
355 struct nfs_fattr *fp;
356 struct nameidata nd, ind, *ndp = &nd;
357 struct vnode *vp, *dirp;
358 nfsfh_t nfh;
359 fhandle_t *fhp;
360 caddr_t cp;
361 u_int32_t *tl;
362 int32_t t1;
363 caddr_t bpos;
364 int error = 0, cache, len, dirattr_ret = 1;
365 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
366 char *cp2;
367 struct mbuf *mb, *mb2, *mreq;
368 struct vattr va, dirattr;
369 u_quad_t frev;
370
371 fhp = &nfh.fh_generic;
372 nfsm_srvmtofh(fhp);
373 nfsm_srvnamesiz(len);
374
375 pubflag = nfs_ispublicfh(fhp);
376
377 nd.ni_cnd.cn_cred = cred;
378 nd.ni_cnd.cn_nameiop = LOOKUP;
379 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
380 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
381 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
382
383 if (!error && pubflag) {
384 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
385 /*
386 * Setup call to lookup() to see if we can find
387 * the index file. Arguably, this doesn't belong
388 * in a kernel.. Ugh.
389 */
390 ind = nd;
391 VOP_UNLOCK(nd.ni_vp, 0);
392 ind.ni_pathlen = strlen(nfs_pub.np_index);
393 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
394 nfs_pub.np_index;
395 ind.ni_startdir = nd.ni_vp;
396 VREF(ind.ni_startdir);
397 error = lookup(&ind);
398 if (!error) {
399 /*
400 * Found an index file. Get rid of
401 * the old references.
402 */
403 if (dirp)
404 vrele(dirp);
405 dirp = nd.ni_vp;
406 vrele(nd.ni_startdir);
407 ndp = &ind;
408 } else
409 error = 0;
410 }
411 /*
412 * If the public filehandle was used, check that this lookup
413 * didn't result in a filehandle outside the publicly exported
414 * filesystem.
415 */
416
417 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
418 vput(nd.ni_vp);
419 error = EPERM;
420 }
421 }
422
423 if (dirp) {
424 if (v3)
425 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred,
426 procp);
427 vrele(dirp);
428 }
429
430 if (error) {
431 nfsm_reply(NFSX_POSTOPATTR(v3));
432 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
433 return (0);
434 }
435
436 nqsrv_getl(ndp->ni_startdir, ND_READ);
437 vrele(ndp->ni_startdir);
438 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
439 vp = ndp->ni_vp;
440 memset((caddr_t)fhp, 0, sizeof(nfh));
441 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
442 error = VFS_VPTOFH(vp, &fhp->fh_fid);
443 if (!error)
444 error = VOP_GETATTR(vp, &va, cred, procp);
445 vput(vp);
446 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
447 if (error) {
448 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
449 return (0);
450 }
451 nfsm_srvfhtom(fhp, v3);
452 if (v3) {
453 nfsm_srvpostop_attr(0, &va);
454 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
455 } else {
456 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
457 nfsm_srvfillattr(&va, fp);
458 }
459 nfsm_srvdone;
460 }
461
462 /*
463 * nfs readlink service
464 */
465 int
466 nfsrv_readlink(nfsd, slp, procp, mrq)
467 struct nfsrv_descript *nfsd;
468 struct nfssvc_sock *slp;
469 struct proc *procp;
470 struct mbuf **mrq;
471 {
472 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
473 struct mbuf *nam = nfsd->nd_nam;
474 caddr_t dpos = nfsd->nd_dpos;
475 struct ucred *cred = &nfsd->nd_cr;
476 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
477 struct iovec *ivp = iv;
478 struct mbuf *mp;
479 u_int32_t *tl;
480 int32_t t1;
481 caddr_t bpos;
482 int error = 0, rdonly, cache, i, tlen, len, getret;
483 int v3 = (nfsd->nd_flag & ND_NFSV3);
484 char *cp2;
485 struct mbuf *mb, *mb2, *mp2 = NULL, *mp3 = NULL, *mreq;
486 struct vnode *vp;
487 struct vattr attr;
488 nfsfh_t nfh;
489 fhandle_t *fhp;
490 struct uio io, *uiop = &io;
491 u_quad_t frev;
492
493 fhp = &nfh.fh_generic;
494 nfsm_srvmtofh(fhp);
495 len = 0;
496 i = 0;
497 while (len < NFS_MAXPATHLEN) {
498 MGET(mp, M_WAIT, MT_DATA);
499 MCLGET(mp, M_WAIT);
500 mp->m_len = NFSMSIZ(mp);
501 if (len == 0)
502 mp3 = mp2 = mp;
503 else {
504 mp2->m_next = mp;
505 mp2 = mp;
506 }
507 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
508 mp->m_len = NFS_MAXPATHLEN-len;
509 len = NFS_MAXPATHLEN;
510 } else
511 len += mp->m_len;
512 ivp->iov_base = mtod(mp, caddr_t);
513 ivp->iov_len = mp->m_len;
514 i++;
515 ivp++;
516 }
517 uiop->uio_iov = iv;
518 uiop->uio_iovcnt = i;
519 uiop->uio_offset = 0;
520 uiop->uio_resid = len;
521 uiop->uio_rw = UIO_READ;
522 uiop->uio_segflg = UIO_SYSSPACE;
523 uiop->uio_procp = (struct proc *)0;
524 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
525 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
526 if (error) {
527 m_freem(mp3);
528 nfsm_reply(2 * NFSX_UNSIGNED);
529 nfsm_srvpostop_attr(1, (struct vattr *)0);
530 return (0);
531 }
532 if (vp->v_type != VLNK) {
533 if (v3)
534 error = EINVAL;
535 else
536 error = ENXIO;
537 goto out;
538 }
539 nqsrv_getl(vp, ND_READ);
540 error = VOP_READLINK(vp, uiop, cred);
541 out:
542 getret = VOP_GETATTR(vp, &attr, cred, procp);
543 vput(vp);
544 if (error)
545 m_freem(mp3);
546 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
547 if (v3) {
548 nfsm_srvpostop_attr(getret, &attr);
549 if (error)
550 return (0);
551 }
552 if (uiop->uio_resid > 0) {
553 len -= uiop->uio_resid;
554 tlen = nfsm_rndup(len);
555 nfsm_adj(mp3, NFS_MAXPATHLEN-tlen, tlen-len);
556 }
557 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
558 *tl = txdr_unsigned(len);
559 mb->m_next = mp3;
560 nfsm_srvdone;
561 }
562
563 /*
564 * nfs read service
565 */
566 int
567 nfsrv_read(nfsd, slp, procp, mrq)
568 struct nfsrv_descript *nfsd;
569 struct nfssvc_sock *slp;
570 struct proc *procp;
571 struct mbuf **mrq;
572 {
573 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
574 struct mbuf *nam = nfsd->nd_nam;
575 caddr_t dpos = nfsd->nd_dpos;
576 struct ucred *cred = &nfsd->nd_cr;
577 struct iovec *iv;
578 struct iovec *iv2;
579 struct mbuf *m;
580 struct nfs_fattr *fp;
581 u_int32_t *tl;
582 int32_t t1;
583 int i;
584 caddr_t bpos;
585 int error = 0, rdonly, cache, cnt, len, left, siz, tlen, getret;
586 int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen;
587 char *cp2;
588 struct mbuf *mb, *mb2, *mreq;
589 struct mbuf *m2;
590 struct vnode *vp;
591 nfsfh_t nfh;
592 fhandle_t *fhp;
593 struct uio io, *uiop = &io;
594 struct vattr va;
595 off_t off;
596 u_quad_t frev;
597
598 fhp = &nfh.fh_generic;
599 nfsm_srvmtofh(fhp);
600 if (v3) {
601 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
602 off = fxdr_hyper(tl);
603 } else {
604 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
605 off = (off_t)fxdr_unsigned(u_int32_t, *tl);
606 }
607 nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd));
608 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
609 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
610 if (error) {
611 nfsm_reply(2 * NFSX_UNSIGNED);
612 nfsm_srvpostop_attr(1, (struct vattr *)0);
613 return (0);
614 }
615 if (vp->v_type != VREG) {
616 if (v3)
617 error = EINVAL;
618 else
619 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
620 }
621 if (!error) {
622 nqsrv_getl(vp, ND_READ);
623 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0)
624 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1);
625 }
626 getret = VOP_GETATTR(vp, &va, cred, procp);
627 if (!error)
628 error = getret;
629 if (error) {
630 vput(vp);
631 nfsm_reply(NFSX_POSTOPATTR(v3));
632 nfsm_srvpostop_attr(getret, &va);
633 return (0);
634 }
635 if (off >= va.va_size)
636 cnt = 0;
637 else if ((off + reqlen) > va.va_size)
638 cnt = nfsm_rndup(va.va_size - off);
639 else
640 cnt = reqlen;
641 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
642 if (v3) {
643 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
644 *tl++ = nfs_true;
645 fp = (struct nfs_fattr *)tl;
646 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
647 } else {
648 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
649 fp = (struct nfs_fattr *)tl;
650 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
651 }
652 len = left = cnt;
653 if (cnt > 0) {
654 /*
655 * Generate the mbuf list with the uio_iov ref. to it.
656 */
657 i = 0;
658 m = m2 = mb;
659 while (left > 0) {
660 siz = min(M_TRAILINGSPACE(m), left);
661 if (siz > 0) {
662 left -= siz;
663 i++;
664 }
665 if (left > 0) {
666 MGET(m, M_WAIT, MT_DATA);
667 MCLGET(m, M_WAIT);
668 m->m_len = 0;
669 m2->m_next = m;
670 m2 = m;
671 }
672 }
673 MALLOC(iv, struct iovec *, i * sizeof (struct iovec),
674 M_TEMP, M_WAITOK);
675 uiop->uio_iov = iv2 = iv;
676 m = mb;
677 left = cnt;
678 i = 0;
679 while (left > 0) {
680 if (m == NULL)
681 panic("nfsrv_read iov");
682 siz = min(M_TRAILINGSPACE(m), left);
683 if (siz > 0) {
684 iv->iov_base = mtod(m, caddr_t) + m->m_len;
685 iv->iov_len = siz;
686 m->m_len += siz;
687 left -= siz;
688 iv++;
689 i++;
690 }
691 m = m->m_next;
692 }
693 uiop->uio_iovcnt = i;
694 uiop->uio_offset = off;
695 uiop->uio_resid = cnt;
696 uiop->uio_rw = UIO_READ;
697 uiop->uio_segflg = UIO_SYSSPACE;
698 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
699 off = uiop->uio_offset;
700 FREE((caddr_t)iv2, M_TEMP);
701 if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){
702 if (!error)
703 error = getret;
704 m_freem(mreq);
705 vput(vp);
706 nfsm_reply(NFSX_POSTOPATTR(v3));
707 nfsm_srvpostop_attr(getret, &va);
708 return (0);
709 }
710 } else
711 uiop->uio_resid = 0;
712 vput(vp);
713 nfsm_srvfillattr(&va, fp);
714 len -= uiop->uio_resid;
715 tlen = nfsm_rndup(len);
716 if (cnt != tlen || tlen != len)
717 nfsm_adj(mb, cnt - tlen, tlen - len);
718 if (v3) {
719 *tl++ = txdr_unsigned(len);
720 if (len < reqlen)
721 *tl++ = nfs_true;
722 else
723 *tl++ = nfs_false;
724 }
725 *tl = txdr_unsigned(len);
726 nfsm_srvdone;
727 }
728
729 /*
730 * nfs write service
731 */
732 int
733 nfsrv_write(nfsd, slp, procp, mrq)
734 struct nfsrv_descript *nfsd;
735 struct nfssvc_sock *slp;
736 struct proc *procp;
737 struct mbuf **mrq;
738 {
739 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
740 struct mbuf *nam = nfsd->nd_nam;
741 caddr_t dpos = nfsd->nd_dpos;
742 struct ucred *cred = &nfsd->nd_cr;
743 struct iovec *ivp;
744 int i, cnt;
745 struct mbuf *mp;
746 struct nfs_fattr *fp;
747 struct iovec *iv;
748 struct vattr va, forat;
749 u_int32_t *tl;
750 int32_t t1;
751 caddr_t bpos;
752 int error = 0, rdonly, cache, len, forat_ret = 1;
753 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
754 int stable = NFSV3WRITE_FILESYNC;
755 int v3 = (nfsd->nd_flag & ND_NFSV3);
756 char *cp2;
757 struct mbuf *mb, *mb2, *mreq;
758 struct vnode *vp;
759 nfsfh_t nfh;
760 fhandle_t *fhp;
761 struct uio io, *uiop = &io;
762 off_t off;
763 u_quad_t frev;
764
765 if (mrep == NULL) {
766 *mrq = NULL;
767 return (0);
768 }
769 fhp = &nfh.fh_generic;
770 nfsm_srvmtofh(fhp);
771 if (v3) {
772 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
773 off = fxdr_hyper(tl);
774 tl += 3;
775 stable = fxdr_unsigned(int, *tl++);
776 } else {
777 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
778 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
779 tl += 2;
780 }
781 retlen = len = fxdr_unsigned(int32_t, *tl);
782 cnt = i = 0;
783
784 /*
785 * For NFS Version 2, it is not obvious what a write of zero length
786 * should do, but I might as well be consistent with Version 3,
787 * which is to return ok so long as there are no permission problems.
788 */
789 if (len > 0) {
790 zeroing = 1;
791 mp = mrep;
792 while (mp) {
793 if (mp == md) {
794 zeroing = 0;
795 adjust = dpos - mtod(mp, caddr_t);
796 mp->m_len -= adjust;
797 if (mp->m_len > 0 && adjust > 0)
798 NFSMADV(mp, adjust);
799 }
800 if (zeroing)
801 mp->m_len = 0;
802 else if (mp->m_len > 0) {
803 i += mp->m_len;
804 if (i > len) {
805 mp->m_len -= (i - len);
806 zeroing = 1;
807 }
808 if (mp->m_len > 0)
809 cnt++;
810 }
811 mp = mp->m_next;
812 }
813 }
814 if (len > NFS_MAXDATA || len < 0 || i < len) {
815 error = EIO;
816 nfsm_reply(2 * NFSX_UNSIGNED);
817 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
818 return (0);
819 }
820 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
821 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
822 if (error) {
823 nfsm_reply(2 * NFSX_UNSIGNED);
824 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
825 return (0);
826 }
827 if (v3)
828 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
829 if (vp->v_type != VREG) {
830 if (v3)
831 error = EINVAL;
832 else
833 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
834 }
835 if (!error) {
836 nqsrv_getl(vp, ND_WRITE);
837 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
838 }
839 if (error) {
840 vput(vp);
841 nfsm_reply(NFSX_WCCDATA(v3));
842 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
843 return (0);
844 }
845
846 if (len > 0) {
847 MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP,
848 M_WAITOK);
849 uiop->uio_iov = iv = ivp;
850 uiop->uio_iovcnt = cnt;
851 mp = mrep;
852 while (mp) {
853 if (mp->m_len > 0) {
854 ivp->iov_base = mtod(mp, caddr_t);
855 ivp->iov_len = mp->m_len;
856 ivp++;
857 }
858 mp = mp->m_next;
859 }
860
861 /*
862 * XXX
863 * The IO_METASYNC flag indicates that all metadata (and not just
864 * enough to ensure data integrity) mus be written to stable storage
865 * synchronously.
866 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
867 */
868 if (stable == NFSV3WRITE_UNSTABLE)
869 ioflags = IO_NODELOCKED;
870 else if (stable == NFSV3WRITE_DATASYNC)
871 ioflags = (IO_SYNC | IO_NODELOCKED);
872 else
873 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
874 uiop->uio_resid = len;
875 uiop->uio_rw = UIO_WRITE;
876 uiop->uio_segflg = UIO_SYSSPACE;
877 uiop->uio_procp = (struct proc *)0;
878 uiop->uio_offset = off;
879 error = VOP_WRITE(vp, uiop, ioflags, cred);
880 nfsstats.srvvop_writes++;
881 FREE((caddr_t)iv, M_TEMP);
882 }
883 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
884 vput(vp);
885 if (!error)
886 error = aftat_ret;
887 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
888 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
889 if (v3) {
890 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
891 if (error)
892 return (0);
893 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
894 *tl++ = txdr_unsigned(retlen);
895 if (stable == NFSV3WRITE_UNSTABLE)
896 *tl++ = txdr_unsigned(stable);
897 else
898 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
899 /*
900 * Actually, there is no need to txdr these fields,
901 * but it may make the values more human readable,
902 * for debugging purposes.
903 */
904 *tl++ = txdr_unsigned(boottime.tv_sec);
905 *tl = txdr_unsigned(boottime.tv_usec);
906 } else {
907 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
908 nfsm_srvfillattr(&va, fp);
909 }
910 nfsm_srvdone;
911 }
912
913 /*
914 * NFS write service with write gathering support. Called when
915 * nfsrvw_procrastinate > 0.
916 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
917 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
918 * Jan. 1994.
919 */
920 int
921 nfsrv_writegather(ndp, slp, procp, mrq)
922 struct nfsrv_descript **ndp;
923 struct nfssvc_sock *slp;
924 struct proc *procp;
925 struct mbuf **mrq;
926 {
927 struct iovec *ivp;
928 struct mbuf *mp;
929 struct nfsrv_descript *wp, *nfsd, *owp, *swp;
930 struct nfs_fattr *fp;
931 int i = 0;
932 struct iovec *iov;
933 struct nfsrvw_delayhash *wpp;
934 struct ucred *cred;
935 struct vattr va, forat;
936 u_int32_t *tl;
937 int32_t t1;
938 caddr_t bpos, dpos;
939 int error = 0, rdonly, cache, len = 0, forat_ret = 1;
940 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
941 char *cp2;
942 struct mbuf *mb, *mb2, *mreq, *mrep, *md;
943 struct vnode *vp;
944 struct uio io, *uiop = &io;
945 u_quad_t frev, cur_usec;
946
947 *mrq = NULL;
948 if (*ndp) {
949 nfsd = *ndp;
950 *ndp = NULL;
951 mrep = nfsd->nd_mrep;
952 md = nfsd->nd_md;
953 dpos = nfsd->nd_dpos;
954 cred = &nfsd->nd_cr;
955 v3 = (nfsd->nd_flag & ND_NFSV3);
956 LIST_INIT(&nfsd->nd_coalesce);
957 nfsd->nd_mreq = NULL;
958 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
959 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
960 nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
961
962 /*
963 * Now, get the write header..
964 */
965 nfsm_srvmtofh(&nfsd->nd_fh);
966 if (v3) {
967 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
968 nfsd->nd_off = fxdr_hyper(tl);
969 tl += 3;
970 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
971 } else {
972 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
973 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
974 tl += 2;
975 }
976 len = fxdr_unsigned(int32_t, *tl);
977 nfsd->nd_len = len;
978 nfsd->nd_eoff = nfsd->nd_off + len;
979
980 /*
981 * Trim the header out of the mbuf list and trim off any trailing
982 * junk so that the mbuf list has only the write data.
983 */
984 zeroing = 1;
985 i = 0;
986 mp = mrep;
987 while (mp) {
988 if (mp == md) {
989 zeroing = 0;
990 adjust = dpos - mtod(mp, caddr_t);
991 mp->m_len -= adjust;
992 if (mp->m_len > 0 && adjust > 0)
993 NFSMADV(mp, adjust);
994 }
995 if (zeroing)
996 mp->m_len = 0;
997 else {
998 i += mp->m_len;
999 if (i > len) {
1000 mp->m_len -= (i - len);
1001 zeroing = 1;
1002 }
1003 }
1004 mp = mp->m_next;
1005 }
1006 if (len > NFS_MAXDATA || len < 0 || i < len) {
1007 nfsmout:
1008 m_freem(mrep);
1009 error = EIO;
1010 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1011 if (v3)
1012 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1013 nfsd->nd_mreq = mreq;
1014 nfsd->nd_mrep = NULL;
1015 nfsd->nd_time = 0;
1016 }
1017
1018 /*
1019 * Add this entry to the hash and time queues.
1020 */
1021 s = splsoftclock();
1022 owp = NULL;
1023 wp = slp->ns_tq.lh_first;
1024 while (wp && wp->nd_time < nfsd->nd_time) {
1025 owp = wp;
1026 wp = wp->nd_tq.le_next;
1027 }
1028 if (owp) {
1029 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1030 } else {
1031 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1032 }
1033 if (nfsd->nd_mrep) {
1034 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1035 owp = NULL;
1036 wp = wpp->lh_first;
1037 while (wp &&
1038 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1039 owp = wp;
1040 wp = wp->nd_hash.le_next;
1041 }
1042 while (wp && wp->nd_off < nfsd->nd_off &&
1043 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1044 owp = wp;
1045 wp = wp->nd_hash.le_next;
1046 }
1047 if (owp) {
1048 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1049
1050 /*
1051 * Search the hash list for overlapping entries and
1052 * coalesce.
1053 */
1054 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1055 wp = nfsd->nd_hash.le_next;
1056 if (NFSW_SAMECRED(owp, nfsd))
1057 nfsrvw_coalesce(owp, nfsd);
1058 }
1059 } else {
1060 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1061 }
1062 }
1063 splx(s);
1064 }
1065
1066 /*
1067 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1068 * and generate the associated reply mbuf list(s).
1069 */
1070 loop1:
1071 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1072 s = splsoftclock();
1073 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = owp) {
1074 owp = nfsd->nd_tq.le_next;
1075 if (nfsd->nd_time > cur_usec)
1076 break;
1077 if (nfsd->nd_mreq)
1078 continue;
1079 LIST_REMOVE(nfsd, nd_tq);
1080 LIST_REMOVE(nfsd, nd_hash);
1081 splx(s);
1082 mrep = nfsd->nd_mrep;
1083 nfsd->nd_mrep = NULL;
1084 cred = &nfsd->nd_cr;
1085 v3 = (nfsd->nd_flag & ND_NFSV3);
1086 forat_ret = aftat_ret = 1;
1087 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1088 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1089 FALSE);
1090 if (!error) {
1091 if (v3)
1092 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
1093 if (vp->v_type != VREG) {
1094 if (v3)
1095 error = EINVAL;
1096 else
1097 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1098 }
1099 } else
1100 vp = NULL;
1101 if (!error) {
1102 nqsrv_getl(vp, ND_WRITE);
1103 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
1104 }
1105
1106 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1107 ioflags = IO_NODELOCKED;
1108 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1109 ioflags = (IO_SYNC | IO_NODELOCKED);
1110 else
1111 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1112 uiop->uio_rw = UIO_WRITE;
1113 uiop->uio_segflg = UIO_SYSSPACE;
1114 uiop->uio_procp = (struct proc *)0;
1115 uiop->uio_offset = nfsd->nd_off;
1116 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1117 if (uiop->uio_resid > 0) {
1118 mp = mrep;
1119 i = 0;
1120 while (mp) {
1121 if (mp->m_len > 0)
1122 i++;
1123 mp = mp->m_next;
1124 }
1125 uiop->uio_iovcnt = i;
1126 MALLOC(iov, struct iovec *, i * sizeof (struct iovec),
1127 M_TEMP, M_WAITOK);
1128 uiop->uio_iov = ivp = iov;
1129 mp = mrep;
1130 while (mp) {
1131 if (mp->m_len > 0) {
1132 ivp->iov_base = mtod(mp, caddr_t);
1133 ivp->iov_len = mp->m_len;
1134 ivp++;
1135 }
1136 mp = mp->m_next;
1137 }
1138 if (!error) {
1139 error = VOP_WRITE(vp, uiop, ioflags, cred);
1140 nfsstats.srvvop_writes++;
1141 }
1142 FREE((caddr_t)iov, M_TEMP);
1143 }
1144 m_freem(mrep);
1145 if (vp) {
1146 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
1147 vput(vp);
1148 }
1149
1150 /*
1151 * Loop around generating replies for all write rpcs that have
1152 * now been completed.
1153 */
1154 swp = nfsd;
1155 do {
1156 if (error) {
1157 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1158 if (v3) {
1159 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1160 }
1161 } else {
1162 nfsm_writereply(NFSX_PREOPATTR(v3) +
1163 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1164 NFSX_WRITEVERF(v3), v3);
1165 if (v3) {
1166 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1167 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1168 *tl++ = txdr_unsigned(nfsd->nd_len);
1169 *tl++ = txdr_unsigned(swp->nd_stable);
1170 /*
1171 * Actually, there is no need to txdr these fields,
1172 * but it may make the values more human readable,
1173 * for debugging purposes.
1174 */
1175 *tl++ = txdr_unsigned(boottime.tv_sec);
1176 *tl = txdr_unsigned(boottime.tv_usec);
1177 } else {
1178 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1179 nfsm_srvfillattr(&va, fp);
1180 }
1181 }
1182 nfsd->nd_mreq = mreq;
1183 if (nfsd->nd_mrep)
1184 panic("nfsrv_write: nd_mrep not free");
1185
1186 /*
1187 * Done. Put it at the head of the timer queue so that
1188 * the final phase can return the reply.
1189 */
1190 s = splsoftclock();
1191 if (nfsd != swp) {
1192 nfsd->nd_time = 0;
1193 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1194 }
1195 nfsd = swp->nd_coalesce.lh_first;
1196 if (nfsd) {
1197 LIST_REMOVE(nfsd, nd_tq);
1198 }
1199 splx(s);
1200 } while (nfsd);
1201 s = splsoftclock();
1202 swp->nd_time = 0;
1203 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1204 splx(s);
1205 goto loop1;
1206 }
1207 splx(s);
1208
1209 /*
1210 * Search for a reply to return.
1211 */
1212 s = splsoftclock();
1213 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = nfsd->nd_tq.le_next)
1214 if (nfsd->nd_mreq) {
1215 LIST_REMOVE(nfsd, nd_tq);
1216 *mrq = nfsd->nd_mreq;
1217 *ndp = nfsd;
1218 break;
1219 }
1220 splx(s);
1221 return (0);
1222 }
1223
1224 /*
1225 * Coalesce the write request nfsd into owp. To do this we must:
1226 * - remove nfsd from the queues
1227 * - merge nfsd->nd_mrep into owp->nd_mrep
1228 * - update the nd_eoff and nd_stable for owp
1229 * - put nfsd on owp's nd_coalesce list
1230 * NB: Must be called at splsoftclock().
1231 */
1232 void
1233 nfsrvw_coalesce(owp, nfsd)
1234 struct nfsrv_descript *owp;
1235 struct nfsrv_descript *nfsd;
1236 {
1237 int overlap;
1238 struct mbuf *mp;
1239
1240 LIST_REMOVE(nfsd, nd_hash);
1241 LIST_REMOVE(nfsd, nd_tq);
1242 if (owp->nd_eoff < nfsd->nd_eoff) {
1243 overlap = owp->nd_eoff - nfsd->nd_off;
1244 if (overlap < 0)
1245 panic("nfsrv_coalesce: bad off");
1246 if (overlap > 0)
1247 m_adj(nfsd->nd_mrep, overlap);
1248 mp = owp->nd_mrep;
1249 while (mp->m_next)
1250 mp = mp->m_next;
1251 mp->m_next = nfsd->nd_mrep;
1252 owp->nd_eoff = nfsd->nd_eoff;
1253 } else
1254 m_freem(nfsd->nd_mrep);
1255 nfsd->nd_mrep = NULL;
1256 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1257 owp->nd_stable = NFSV3WRITE_FILESYNC;
1258 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1259 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1260 owp->nd_stable = NFSV3WRITE_DATASYNC;
1261 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1262 /*
1263 * nfsd might hold coalesce elements! Move them to owp.
1264 * Otherwise, requests may be lost and clients will be stuck.
1265 */
1266 if (nfsd->nd_coalesce.lh_first)
1267 {
1268 struct nfsrv_descript *m;
1269
1270 while ((m = nfsd->nd_coalesce.lh_first))
1271 {
1272 LIST_REMOVE(m, nd_tq);
1273 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1274 }
1275 }
1276 }
1277
1278 /*
1279 * nfs create service
1280 * now does a truncate to 0 length via. setattr if it already exists
1281 */
1282 int
1283 nfsrv_create(nfsd, slp, procp, mrq)
1284 struct nfsrv_descript *nfsd;
1285 struct nfssvc_sock *slp;
1286 struct proc *procp;
1287 struct mbuf **mrq;
1288 {
1289 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1290 struct mbuf *nam = nfsd->nd_nam;
1291 caddr_t dpos = nfsd->nd_dpos;
1292 struct ucred *cred = &nfsd->nd_cr;
1293 struct nfs_fattr *fp;
1294 struct vattr va, dirfor, diraft;
1295 struct nfsv2_sattr *sp;
1296 u_int32_t *tl;
1297 struct nameidata nd;
1298 caddr_t cp;
1299 int32_t t1;
1300 caddr_t bpos;
1301 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1302 int rdev = 0;
1303 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1304 char *cp2;
1305 struct mbuf *mb, *mb2, *mreq;
1306 struct vnode *vp = NULL, *dirp = NULL;
1307 nfsfh_t nfh;
1308 fhandle_t *fhp;
1309 u_quad_t frev, tempsize;
1310 u_char cverf[NFSX_V3CREATEVERF];
1311
1312 nd.ni_cnd.cn_nameiop = 0;
1313 fhp = &nfh.fh_generic;
1314 nfsm_srvmtofh(fhp);
1315 nfsm_srvnamesiz(len);
1316 nd.ni_cnd.cn_cred = cred;
1317 nd.ni_cnd.cn_nameiop = CREATE;
1318 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART;
1319 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1320 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1321 if (dirp) {
1322 if (v3)
1323 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1324 procp);
1325 else {
1326 vrele(dirp);
1327 dirp = (struct vnode *)0;
1328 }
1329 }
1330 if (error) {
1331 nfsm_reply(NFSX_WCCDATA(v3));
1332 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1333 if (dirp)
1334 vrele(dirp);
1335 return (0);
1336 }
1337 VATTR_NULL(&va);
1338 if (v3) {
1339 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1340 how = fxdr_unsigned(int, *tl);
1341 switch (how) {
1342 case NFSV3CREATE_GUARDED:
1343 if (nd.ni_vp) {
1344 error = EEXIST;
1345 break;
1346 }
1347 case NFSV3CREATE_UNCHECKED:
1348 nfsm_srvsattr(&va);
1349 break;
1350 case NFSV3CREATE_EXCLUSIVE:
1351 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
1352 memcpy(cverf, cp, NFSX_V3CREATEVERF);
1353 exclusive_flag = 1;
1354 if (nd.ni_vp == NULL)
1355 va.va_mode = 0;
1356 break;
1357 };
1358 va.va_type = VREG;
1359 } else {
1360 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1361 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1362 if (va.va_type == VNON)
1363 va.va_type = VREG;
1364 va.va_mode = nfstov_mode(sp->sa_mode);
1365 switch (va.va_type) {
1366 case VREG:
1367 tsize = fxdr_unsigned(int32_t, sp->sa_size);
1368 if (tsize != -1)
1369 va.va_size = (u_quad_t)tsize;
1370 break;
1371 case VCHR:
1372 case VBLK:
1373 case VFIFO:
1374 rdev = fxdr_unsigned(int32_t, sp->sa_size);
1375 break;
1376 default:
1377 break;
1378 };
1379 }
1380
1381 /*
1382 * Iff doesn't exist, create it
1383 * otherwise just truncate to 0 length
1384 * should I set the mode too ??
1385 */
1386 if (nd.ni_vp == NULL) {
1387 if (va.va_type == VREG || va.va_type == VSOCK) {
1388 vrele(nd.ni_startdir);
1389 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1390 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1391 if (!error) {
1392 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1393 if (exclusive_flag) {
1394 exclusive_flag = 0;
1395 VATTR_NULL(&va);
1396 memcpy((caddr_t)&va.va_atime, cverf,
1397 NFSX_V3CREATEVERF);
1398 error = VOP_SETATTR(nd.ni_vp, &va, cred,
1399 procp);
1400 }
1401 }
1402 } else if (va.va_type == VCHR || va.va_type == VBLK ||
1403 va.va_type == VFIFO) {
1404 if (va.va_type == VCHR && rdev == 0xffffffff)
1405 va.va_type = VFIFO;
1406 if (va.va_type != VFIFO &&
1407 (error = suser(cred, (u_short *)0))) {
1408 vrele(nd.ni_startdir);
1409 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1410 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1411 vput(nd.ni_dvp);
1412 nfsm_reply(0);
1413 return (error);
1414 } else
1415 va.va_rdev = (dev_t)rdev;
1416 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1417 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1418 &va);
1419 if (error) {
1420 vrele(nd.ni_startdir);
1421 nfsm_reply(0);
1422 }
1423 nd.ni_cnd.cn_nameiop = LOOKUP;
1424 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART);
1425 nd.ni_cnd.cn_proc = procp;
1426 nd.ni_cnd.cn_cred = cred;
1427 if ((error = lookup(&nd)) != 0) {
1428 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1429 nfsm_reply(0);
1430 }
1431 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1432 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1433 vrele(nd.ni_dvp);
1434 vput(nd.ni_vp);
1435 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1436 error = EINVAL;
1437 nfsm_reply(0);
1438 }
1439 } else {
1440 vrele(nd.ni_startdir);
1441 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1442 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1443 vput(nd.ni_dvp);
1444 error = ENXIO;
1445 }
1446 vp = nd.ni_vp;
1447 } else {
1448 vrele(nd.ni_startdir);
1449 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1450 vp = nd.ni_vp;
1451 if (nd.ni_dvp == vp)
1452 vrele(nd.ni_dvp);
1453 else
1454 vput(nd.ni_dvp);
1455 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1456 if (!error && va.va_size != -1) {
1457 error = nfsrv_access(vp, VWRITE, cred,
1458 (nd.ni_cnd.cn_flags & RDONLY), procp, 0);
1459 if (!error) {
1460 nqsrv_getl(vp, ND_WRITE);
1461 tempsize = va.va_size;
1462 VATTR_NULL(&va);
1463 va.va_size = tempsize;
1464 error = VOP_SETATTR(vp, &va, cred,
1465 procp);
1466 }
1467 }
1468 if (error)
1469 vput(vp);
1470 }
1471 if (!error) {
1472 memset((caddr_t)fhp, 0, sizeof(nfh));
1473 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1474 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1475 if (!error)
1476 error = VOP_GETATTR(vp, &va, cred, procp);
1477 vput(vp);
1478 }
1479 if (v3) {
1480 if (exclusive_flag && !error &&
1481 memcmp(cverf, (caddr_t)&va.va_atime, NFSX_V3CREATEVERF))
1482 error = EEXIST;
1483 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1484 vrele(dirp);
1485 }
1486 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1487 if (v3) {
1488 if (!error) {
1489 nfsm_srvpostop_fh(fhp);
1490 nfsm_srvpostop_attr(0, &va);
1491 }
1492 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1493 } else {
1494 nfsm_srvfhtom(fhp, v3);
1495 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1496 nfsm_srvfillattr(&va, fp);
1497 }
1498 return (0);
1499 nfsmout:
1500 if (dirp)
1501 vrele(dirp);
1502 if (nd.ni_cnd.cn_nameiop) {
1503 vrele(nd.ni_startdir);
1504 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1505 }
1506 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1507 if (nd.ni_dvp == nd.ni_vp)
1508 vrele(nd.ni_dvp);
1509 else
1510 vput(nd.ni_dvp);
1511 if (nd.ni_vp)
1512 vput(nd.ni_vp);
1513 return (error);
1514 }
1515
1516 /*
1517 * nfs v3 mknod service
1518 */
1519 int
1520 nfsrv_mknod(nfsd, slp, procp, mrq)
1521 struct nfsrv_descript *nfsd;
1522 struct nfssvc_sock *slp;
1523 struct proc *procp;
1524 struct mbuf **mrq;
1525 {
1526 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1527 struct mbuf *nam = nfsd->nd_nam;
1528 caddr_t dpos = nfsd->nd_dpos;
1529 struct ucred *cred = &nfsd->nd_cr;
1530 struct vattr va, dirfor, diraft;
1531 u_int32_t *tl;
1532 struct nameidata nd;
1533 int32_t t1;
1534 caddr_t bpos;
1535 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1536 u_int32_t major, minor;
1537 enum vtype vtyp;
1538 char *cp2;
1539 struct mbuf *mb, *mb2, *mreq;
1540 struct vnode *vp, *dirp = (struct vnode *)0;
1541 nfsfh_t nfh;
1542 fhandle_t *fhp;
1543 u_quad_t frev;
1544
1545 nd.ni_cnd.cn_nameiop = 0;
1546 fhp = &nfh.fh_generic;
1547 nfsm_srvmtofh(fhp);
1548 nfsm_srvnamesiz(len);
1549 nd.ni_cnd.cn_cred = cred;
1550 nd.ni_cnd.cn_nameiop = CREATE;
1551 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART;
1552 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1553 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1554 if (dirp)
1555 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp);
1556 if (error) {
1557 nfsm_reply(NFSX_WCCDATA(1));
1558 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1559 if (dirp)
1560 vrele(dirp);
1561 return (0);
1562 }
1563 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1564 vtyp = nfsv3tov_type(*tl);
1565 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1566 vrele(nd.ni_startdir);
1567 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1568 error = NFSERR_BADTYPE;
1569 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1570 if (nd.ni_dvp == nd.ni_vp)
1571 vrele(nd.ni_dvp);
1572 else
1573 vput(nd.ni_dvp);
1574 if (nd.ni_vp)
1575 vput(nd.ni_vp);
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 if (nd.ni_dvp == nd.ni_vp)
1596 vrele(nd.ni_dvp);
1597 else
1598 vput(nd.ni_dvp);
1599 vput(nd.ni_vp);
1600 goto out;
1601 }
1602 va.va_type = vtyp;
1603 if (vtyp == VSOCK) {
1604 vrele(nd.ni_startdir);
1605 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1606 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1607 if (!error)
1608 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1609 } else {
1610 if (va.va_type != VFIFO &&
1611 (error = suser(cred, (u_short *)0))) {
1612 vrele(nd.ni_startdir);
1613 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1614 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1615 vput(nd.ni_dvp);
1616 goto out;
1617 }
1618 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1619 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1620 if (error) {
1621 vrele(nd.ni_startdir);
1622 goto out;
1623 }
1624 nd.ni_cnd.cn_nameiop = LOOKUP;
1625 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART);
1626 nd.ni_cnd.cn_proc = procp;
1627 nd.ni_cnd.cn_cred = procp->p_ucred;
1628 error = lookup(&nd);
1629 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1630 if (error)
1631 goto out;
1632 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1633 vrele(nd.ni_dvp);
1634 vput(nd.ni_vp);
1635 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1636 error = EINVAL;
1637 }
1638 }
1639 out:
1640 vp = nd.ni_vp;
1641 if (!error) {
1642 memset((caddr_t)fhp, 0, sizeof(nfh));
1643 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1644 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1645 if (!error)
1646 error = VOP_GETATTR(vp, &va, cred, procp);
1647 vput(vp);
1648 }
1649 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1650 vrele(dirp);
1651 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
1652 if (!error) {
1653 nfsm_srvpostop_fh(fhp);
1654 nfsm_srvpostop_attr(0, &va);
1655 }
1656 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1657 return (0);
1658 nfsmout:
1659 if (dirp)
1660 vrele(dirp);
1661 if (nd.ni_cnd.cn_nameiop) {
1662 vrele(nd.ni_startdir);
1663 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1664 }
1665 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1666 if (nd.ni_dvp == nd.ni_vp)
1667 vrele(nd.ni_dvp);
1668 else
1669 vput(nd.ni_dvp);
1670 if (nd.ni_vp)
1671 vput(nd.ni_vp);
1672 return (error);
1673 }
1674
1675 /*
1676 * nfs remove service
1677 */
1678 int
1679 nfsrv_remove(nfsd, slp, procp, mrq)
1680 struct nfsrv_descript *nfsd;
1681 struct nfssvc_sock *slp;
1682 struct proc *procp;
1683 struct mbuf **mrq;
1684 {
1685 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1686 struct mbuf *nam = nfsd->nd_nam;
1687 caddr_t dpos = nfsd->nd_dpos;
1688 struct ucred *cred = &nfsd->nd_cr;
1689 struct nameidata nd;
1690 u_int32_t *tl;
1691 int32_t t1;
1692 caddr_t bpos;
1693 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1694 int v3 = (nfsd->nd_flag & ND_NFSV3);
1695 char *cp2;
1696 struct mbuf *mb, *mreq;
1697 struct vnode *vp, *dirp;
1698 struct vattr dirfor, diraft;
1699 nfsfh_t nfh;
1700 fhandle_t *fhp;
1701 u_quad_t frev;
1702
1703 #ifndef nolint
1704 vp = (struct vnode *)0;
1705 #endif
1706 fhp = &nfh.fh_generic;
1707 nfsm_srvmtofh(fhp);
1708 nfsm_srvnamesiz(len);
1709 nd.ni_cnd.cn_cred = cred;
1710 nd.ni_cnd.cn_nameiop = DELETE;
1711 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1712 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1713 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1714 if (dirp) {
1715 if (v3)
1716 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1717 procp);
1718 else
1719 vrele(dirp);
1720 }
1721 if (!error) {
1722 vp = nd.ni_vp;
1723 if (vp->v_type == VDIR &&
1724 (error = suser(cred, (u_short *)0)) != 0)
1725 goto out;
1726 /*
1727 * The root of a mounted filesystem cannot be deleted.
1728 */
1729 if (vp->v_flag & VROOT) {
1730 error = EBUSY;
1731 goto out;
1732 }
1733 out:
1734 if (!error) {
1735 (void)uvm_vnp_uncache(vp);
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 u_int32_t *tl;
1775 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 /*
1896 * If source is the same as the destination (that is the
1897 * same vnode with the same name in the same directory),
1898 * then there is nothing to do.
1899 */
1900 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
1901 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1902 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
1903 fromnd.ni_cnd.cn_namelen))
1904 error = -1;
1905 out:
1906 if (!error) {
1907 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
1908 nqsrv_getl(tdvp, ND_WRITE);
1909 if (tvp) {
1910 (void)uvm_vnp_uncache(tvp);
1911 nqsrv_getl(tvp, ND_WRITE);
1912 }
1913 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
1914 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
1915 } else {
1916 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
1917 if (tdvp == tvp)
1918 vrele(tdvp);
1919 else
1920 vput(tdvp);
1921 if (tvp)
1922 vput(tvp);
1923 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1924 vrele(fromnd.ni_dvp);
1925 vrele(fvp);
1926 if (error == -1)
1927 error = 0;
1928 }
1929 vrele(tond.ni_startdir);
1930 FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
1931 out1:
1932 if (fdirp) {
1933 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp);
1934 vrele(fdirp);
1935 }
1936 if (tdirp) {
1937 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp);
1938 vrele(tdirp);
1939 }
1940 vrele(fromnd.ni_startdir);
1941 FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
1942 nfsm_reply(2 * NFSX_WCCDATA(v3));
1943 if (v3) {
1944 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1945 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1946 }
1947 return (0);
1948
1949 nfsmout:
1950 if (fdirp)
1951 vrele(fdirp);
1952 if (tdirp)
1953 vrele(tdirp);
1954 if (tond.ni_cnd.cn_nameiop) {
1955 vrele(tond.ni_startdir);
1956 FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
1957 }
1958 if (fromnd.ni_cnd.cn_nameiop) {
1959 vrele(fromnd.ni_startdir);
1960 FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
1961 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1962 vrele(fromnd.ni_dvp);
1963 vrele(fvp);
1964 }
1965 return (error);
1966 }
1967
1968 /*
1969 * nfs link service
1970 */
1971 int
1972 nfsrv_link(nfsd, slp, procp, mrq)
1973 struct nfsrv_descript *nfsd;
1974 struct nfssvc_sock *slp;
1975 struct proc *procp;
1976 struct mbuf **mrq;
1977 {
1978 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1979 struct mbuf *nam = nfsd->nd_nam;
1980 caddr_t dpos = nfsd->nd_dpos;
1981 struct ucred *cred = &nfsd->nd_cr;
1982 struct nameidata nd;
1983 u_int32_t *tl;
1984 int32_t t1;
1985 caddr_t bpos;
1986 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1;
1987 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
1988 char *cp2;
1989 struct mbuf *mb, *mreq;
1990 struct vnode *vp, *xp, *dirp = (struct vnode *)0;
1991 struct vattr dirfor, diraft, at;
1992 nfsfh_t nfh, dnfh;
1993 fhandle_t *fhp, *dfhp;
1994 u_quad_t frev;
1995
1996 fhp = &nfh.fh_generic;
1997 dfhp = &dnfh.fh_generic;
1998 nfsm_srvmtofh(fhp);
1999 nfsm_srvmtofh(dfhp);
2000 nfsm_srvnamesiz(len);
2001 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
2002 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2003 if (error) {
2004 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2005 nfsm_srvpostop_attr(getret, &at);
2006 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2007 return (0);
2008 }
2009 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0)
2010 goto out1;
2011 nd.ni_cnd.cn_cred = cred;
2012 nd.ni_cnd.cn_nameiop = CREATE;
2013 nd.ni_cnd.cn_flags = LOCKPARENT;
2014 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
2015 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2016 if (dirp) {
2017 if (v3)
2018 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2019 procp);
2020 else {
2021 vrele(dirp);
2022 dirp = (struct vnode *)0;
2023 }
2024 }
2025 if (error)
2026 goto out1;
2027 xp = nd.ni_vp;
2028 if (xp != NULL) {
2029 error = EEXIST;
2030 goto out;
2031 }
2032 xp = nd.ni_dvp;
2033 if (vp->v_mount != xp->v_mount)
2034 error = EXDEV;
2035 out:
2036 if (!error) {
2037 nqsrv_getl(vp, ND_WRITE);
2038 nqsrv_getl(xp, ND_WRITE);
2039 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2040 } else {
2041 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2042 if (nd.ni_dvp == nd.ni_vp)
2043 vrele(nd.ni_dvp);
2044 else
2045 vput(nd.ni_dvp);
2046 if (nd.ni_vp)
2047 vrele(nd.ni_vp);
2048 }
2049 out1:
2050 if (v3)
2051 getret = VOP_GETATTR(vp, &at, cred, procp);
2052 if (dirp) {
2053 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2054 vrele(dirp);
2055 }
2056 vrele(vp);
2057 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2058 if (v3) {
2059 nfsm_srvpostop_attr(getret, &at);
2060 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2061 return (0);
2062 }
2063 nfsm_srvdone;
2064 }
2065
2066 /*
2067 * nfs symbolic link service
2068 */
2069 int
2070 nfsrv_symlink(nfsd, slp, procp, mrq)
2071 struct nfsrv_descript *nfsd;
2072 struct nfssvc_sock *slp;
2073 struct proc *procp;
2074 struct mbuf **mrq;
2075 {
2076 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2077 struct mbuf *nam = nfsd->nd_nam;
2078 caddr_t dpos = nfsd->nd_dpos;
2079 struct ucred *cred = &nfsd->nd_cr;
2080 struct vattr va, dirfor, diraft;
2081 struct nameidata nd;
2082 u_int32_t *tl;
2083 int32_t t1;
2084 struct nfsv2_sattr *sp;
2085 char *bpos, *pathcp = NULL, *cp2;
2086 struct uio io;
2087 struct iovec iv;
2088 int error = 0, cache, len, len2, dirfor_ret = 1, diraft_ret = 1;
2089 int v3 = (nfsd->nd_flag & ND_NFSV3);
2090 struct mbuf *mb, *mreq, *mb2;
2091 struct vnode *dirp = (struct vnode *)0;
2092 nfsfh_t nfh;
2093 fhandle_t *fhp;
2094 u_quad_t frev;
2095
2096 nd.ni_cnd.cn_nameiop = 0;
2097 fhp = &nfh.fh_generic;
2098 nfsm_srvmtofh(fhp);
2099 nfsm_srvnamesiz(len);
2100 nd.ni_cnd.cn_cred = cred;
2101 nd.ni_cnd.cn_nameiop = CREATE;
2102 nd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART;
2103 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2104 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2105 if (dirp) {
2106 if (v3)
2107 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2108 procp);
2109 else {
2110 vrele(dirp);
2111 dirp = (struct vnode *)0;
2112 }
2113 }
2114 if (error)
2115 goto out;
2116 VATTR_NULL(&va);
2117 if (v3)
2118 nfsm_srvsattr(&va);
2119 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2120 MALLOC(pathcp, caddr_t, len2 + 1, M_TEMP, M_WAITOK);
2121 iv.iov_base = pathcp;
2122 iv.iov_len = len2;
2123 io.uio_resid = len2;
2124 io.uio_offset = 0;
2125 io.uio_iov = &iv;
2126 io.uio_iovcnt = 1;
2127 io.uio_segflg = UIO_SYSSPACE;
2128 io.uio_rw = UIO_READ;
2129 io.uio_procp = (struct proc *)0;
2130 nfsm_mtouio(&io, len2);
2131 if (!v3) {
2132 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2133 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2134 }
2135 *(pathcp + len2) = '\0';
2136 if (nd.ni_vp) {
2137 vrele(nd.ni_startdir);
2138 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2139 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2140 if (nd.ni_dvp == nd.ni_vp)
2141 vrele(nd.ni_dvp);
2142 else
2143 vput(nd.ni_dvp);
2144 vrele(nd.ni_vp);
2145 error = EEXIST;
2146 goto out;
2147 }
2148 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2149 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2150 if (error)
2151 vrele(nd.ni_startdir);
2152 else {
2153 if (v3) {
2154 nd.ni_cnd.cn_nameiop = LOOKUP;
2155 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART | FOLLOW);
2156 nd.ni_cnd.cn_flags |= (NOFOLLOW | LOCKLEAF);
2157 nd.ni_cnd.cn_proc = procp;
2158 nd.ni_cnd.cn_cred = cred;
2159 error = lookup(&nd);
2160 if (!error) {
2161 memset((caddr_t)fhp, 0, sizeof(nfh));
2162 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
2163 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2164 if (!error)
2165 error = VOP_GETATTR(nd.ni_vp, &va, cred,
2166 procp);
2167 vput(nd.ni_vp);
2168 }
2169 } else
2170 vrele(nd.ni_startdir);
2171 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2172 }
2173 out:
2174 if (pathcp)
2175 FREE(pathcp, M_TEMP);
2176 if (dirp) {
2177 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2178 vrele(dirp);
2179 }
2180 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2181 if (v3) {
2182 if (!error) {
2183 nfsm_srvpostop_fh(fhp);
2184 nfsm_srvpostop_attr(0, &va);
2185 }
2186 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2187 }
2188 return (0);
2189 nfsmout:
2190 if (nd.ni_cnd.cn_nameiop) {
2191 vrele(nd.ni_startdir);
2192 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2193 }
2194 if (dirp)
2195 vrele(dirp);
2196 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2197 if (nd.ni_dvp == nd.ni_vp)
2198 vrele(nd.ni_dvp);
2199 else
2200 vput(nd.ni_dvp);
2201 if (nd.ni_vp)
2202 vrele(nd.ni_vp);
2203 if (pathcp)
2204 FREE(pathcp, M_TEMP);
2205 return (error);
2206 }
2207
2208 /*
2209 * nfs mkdir service
2210 */
2211 int
2212 nfsrv_mkdir(nfsd, slp, procp, mrq)
2213 struct nfsrv_descript *nfsd;
2214 struct nfssvc_sock *slp;
2215 struct proc *procp;
2216 struct mbuf **mrq;
2217 {
2218 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2219 struct mbuf *nam = nfsd->nd_nam;
2220 caddr_t dpos = nfsd->nd_dpos;
2221 struct ucred *cred = &nfsd->nd_cr;
2222 struct vattr va, dirfor, diraft;
2223 struct nfs_fattr *fp;
2224 struct nameidata nd;
2225 caddr_t cp;
2226 u_int32_t *tl;
2227 int32_t t1;
2228 caddr_t bpos;
2229 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2230 int v3 = (nfsd->nd_flag & ND_NFSV3);
2231 char *cp2;
2232 struct mbuf *mb, *mb2, *mreq;
2233 struct vnode *vp, *dirp = (struct vnode *)0;
2234 nfsfh_t nfh;
2235 fhandle_t *fhp;
2236 u_quad_t frev;
2237
2238 fhp = &nfh.fh_generic;
2239 nfsm_srvmtofh(fhp);
2240 nfsm_srvnamesiz(len);
2241 nd.ni_cnd.cn_cred = cred;
2242 nd.ni_cnd.cn_nameiop = CREATE;
2243 nd.ni_cnd.cn_flags = LOCKPARENT;
2244 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2245 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2246 if (dirp) {
2247 if (v3)
2248 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2249 procp);
2250 else {
2251 vrele(dirp);
2252 dirp = (struct vnode *)0;
2253 }
2254 }
2255 if (error) {
2256 nfsm_reply(NFSX_WCCDATA(v3));
2257 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2258 if (dirp)
2259 vrele(dirp);
2260 return (0);
2261 }
2262 VATTR_NULL(&va);
2263 if (v3) {
2264 nfsm_srvsattr(&va);
2265 } else {
2266 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2267 va.va_mode = nfstov_mode(*tl++);
2268 }
2269 va.va_type = VDIR;
2270 vp = nd.ni_vp;
2271 if (vp != NULL) {
2272 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2273 if (nd.ni_dvp == vp)
2274 vrele(nd.ni_dvp);
2275 else
2276 vput(nd.ni_dvp);
2277 vrele(vp);
2278 error = EEXIST;
2279 goto out;
2280 }
2281 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2282 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2283 if (!error) {
2284 vp = nd.ni_vp;
2285 memset((caddr_t)fhp, 0, sizeof(nfh));
2286 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
2287 error = VFS_VPTOFH(vp, &fhp->fh_fid);
2288 if (!error)
2289 error = VOP_GETATTR(vp, &va, cred, procp);
2290 vput(vp);
2291 }
2292 out:
2293 if (dirp) {
2294 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2295 vrele(dirp);
2296 }
2297 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2298 if (v3) {
2299 if (!error) {
2300 nfsm_srvpostop_fh(fhp);
2301 nfsm_srvpostop_attr(0, &va);
2302 }
2303 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2304 } else {
2305 nfsm_srvfhtom(fhp, v3);
2306 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2307 nfsm_srvfillattr(&va, fp);
2308 }
2309 return (0);
2310 nfsmout:
2311 if (dirp)
2312 vrele(dirp);
2313 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2314 if (nd.ni_dvp == nd.ni_vp)
2315 vrele(nd.ni_dvp);
2316 else
2317 vput(nd.ni_dvp);
2318 if (nd.ni_vp)
2319 vrele(nd.ni_vp);
2320 return (error);
2321 }
2322
2323 /*
2324 * nfs rmdir service
2325 */
2326 int
2327 nfsrv_rmdir(nfsd, slp, procp, mrq)
2328 struct nfsrv_descript *nfsd;
2329 struct nfssvc_sock *slp;
2330 struct proc *procp;
2331 struct mbuf **mrq;
2332 {
2333 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2334 struct mbuf *nam = nfsd->nd_nam;
2335 caddr_t dpos = nfsd->nd_dpos;
2336 struct ucred *cred = &nfsd->nd_cr;
2337 u_int32_t *tl;
2338 int32_t t1;
2339 caddr_t bpos;
2340 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2341 int v3 = (nfsd->nd_flag & ND_NFSV3);
2342 char *cp2;
2343 struct mbuf *mb, *mreq;
2344 struct vnode *vp, *dirp = (struct vnode *)0;
2345 struct vattr dirfor, diraft;
2346 nfsfh_t nfh;
2347 fhandle_t *fhp;
2348 struct nameidata nd;
2349 u_quad_t frev;
2350
2351 fhp = &nfh.fh_generic;
2352 nfsm_srvmtofh(fhp);
2353 nfsm_srvnamesiz(len);
2354 nd.ni_cnd.cn_cred = cred;
2355 nd.ni_cnd.cn_nameiop = DELETE;
2356 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2357 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2358 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2359 if (dirp) {
2360 if (v3)
2361 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2362 procp);
2363 else {
2364 vrele(dirp);
2365 dirp = (struct vnode *)0;
2366 }
2367 }
2368 if (error) {
2369 nfsm_reply(NFSX_WCCDATA(v3));
2370 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2371 if (dirp)
2372 vrele(dirp);
2373 return (0);
2374 }
2375 vp = nd.ni_vp;
2376 if (vp->v_type != VDIR) {
2377 error = ENOTDIR;
2378 goto out;
2379 }
2380 /*
2381 * No rmdir "." please.
2382 */
2383 if (nd.ni_dvp == vp) {
2384 error = EINVAL;
2385 goto out;
2386 }
2387 /*
2388 * The root of a mounted filesystem cannot be deleted.
2389 */
2390 if (vp->v_flag & VROOT)
2391 error = EBUSY;
2392 out:
2393 if (!error) {
2394 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2395 nqsrv_getl(vp, ND_WRITE);
2396 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2397 } else {
2398 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2399 if (nd.ni_dvp == nd.ni_vp)
2400 vrele(nd.ni_dvp);
2401 else
2402 vput(nd.ni_dvp);
2403 vput(vp);
2404 }
2405 if (dirp) {
2406 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2407 vrele(dirp);
2408 }
2409 nfsm_reply(NFSX_WCCDATA(v3));
2410 if (v3) {
2411 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2412 return (0);
2413 }
2414 nfsm_srvdone;
2415 }
2416
2417 /*
2418 * nfs readdir service
2419 * - mallocs what it thinks is enough to read
2420 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR
2421 * - calls VOP_READDIR()
2422 * - loops around building the reply
2423 * if the output generated exceeds count break out of loop
2424 * The nfsm_clget macro is used here so that the reply will be packed
2425 * tightly in mbuf clusters.
2426 * - it only knows that it has encountered eof when the VOP_READDIR()
2427 * reads nothing
2428 * - as such one readdir rpc will return eof false although you are there
2429 * and then the next will return eof
2430 * - it trims out records with d_fileno == 0
2431 * this doesn't matter for Unix clients, but they might confuse clients
2432 * for other os'.
2433 * - it trims out records with d_type == DT_WHT
2434 * these cannot be seen through NFS (unless we extend the protocol)
2435 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2436 * than requested, but this may not apply to all filesystems. For
2437 * example, client NFS does not { although it is never remote mounted
2438 * anyhow }
2439 * The alternate call nfsrv_readdirplus() does lookups as well.
2440 * PS: The NFS protocol spec. does not clarify what the "count" byte
2441 * argument is a count of.. just name strings and file id's or the
2442 * entire reply rpc or ...
2443 * I tried just file name and id sizes and it confused the Sun client,
2444 * so I am using the full rpc size now. The "paranoia.." comment refers
2445 * to including the status longwords that are not a part of the dir.
2446 * "entry" structures, but are in the rpc.
2447 */
2448 struct flrep {
2449 nfsuint64 fl_off;
2450 u_int32_t fl_postopok;
2451 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2452 u_int32_t fl_fhok;
2453 u_int32_t fl_fhsize;
2454 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2455 };
2456
2457 int
2458 nfsrv_readdir(nfsd, slp, procp, mrq)
2459 struct nfsrv_descript *nfsd;
2460 struct nfssvc_sock *slp;
2461 struct proc *procp;
2462 struct mbuf **mrq;
2463 {
2464 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2465 struct mbuf *nam = nfsd->nd_nam;
2466 caddr_t dpos = nfsd->nd_dpos;
2467 struct ucred *cred = &nfsd->nd_cr;
2468 char *bp, *be;
2469 struct mbuf *mp;
2470 struct dirent *dp;
2471 caddr_t cp;
2472 u_int32_t *tl;
2473 int32_t t1;
2474 caddr_t bpos;
2475 struct mbuf *mb, *mb2, *mreq, *mp2;
2476 char *cpos, *cend, *cp2, *rbuf;
2477 struct vnode *vp;
2478 struct vattr at;
2479 nfsfh_t nfh;
2480 fhandle_t *fhp;
2481 struct uio io;
2482 struct iovec iv;
2483 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2484 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies;
2485 int v3 = (nfsd->nd_flag & ND_NFSV3);
2486 u_quad_t frev, off, toff, verf;
2487 off_t *cookies = NULL, *cookiep;
2488 nfsuint64 jar;
2489
2490 fhp = &nfh.fh_generic;
2491 nfsm_srvmtofh(fhp);
2492 if (v3) {
2493 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2494 toff = fxdr_hyper(tl);
2495 tl += 2;
2496 verf = fxdr_hyper(tl);
2497 tl += 2;
2498 } else {
2499 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2500 toff = fxdr_unsigned(u_quad_t, *tl++);
2501 }
2502 off = toff;
2503 cnt = fxdr_unsigned(int, *tl);
2504 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2505 xfer = NFS_SRVMAXDATA(nfsd);
2506 if (siz > xfer)
2507 siz = xfer;
2508 fullsiz = siz;
2509 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2510 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2511 if (!error && vp->v_type != VDIR) {
2512 error = ENOTDIR;
2513 vput(vp);
2514 }
2515 if (error) {
2516 nfsm_reply(NFSX_UNSIGNED);
2517 nfsm_srvpostop_attr(getret, &at);
2518 return (0);
2519 }
2520 nqsrv_getl(vp, ND_READ);
2521 if (v3) {
2522 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2523 #ifdef NFS3_STRICTVERF
2524 /*
2525 * XXX This check is too strict for Solaris 2.5 clients.
2526 */
2527 if (!error && toff && verf != at.va_filerev)
2528 error = NFSERR_BAD_COOKIE;
2529 #endif
2530 }
2531 if (!error)
2532 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2533 if (error) {
2534 vput(vp);
2535 nfsm_reply(NFSX_POSTOPATTR(v3));
2536 nfsm_srvpostop_attr(getret, &at);
2537 return (0);
2538 }
2539 VOP_UNLOCK(vp, 0);
2540 MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
2541 again:
2542 iv.iov_base = rbuf;
2543 iv.iov_len = fullsiz;
2544 io.uio_iov = &iv;
2545 io.uio_iovcnt = 1;
2546 io.uio_offset = (off_t)off;
2547 io.uio_resid = fullsiz;
2548 io.uio_segflg = UIO_SYSSPACE;
2549 io.uio_rw = UIO_READ;
2550 io.uio_procp = (struct proc *)0;
2551 eofflag = 0;
2552 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2553
2554 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2555
2556 off = (off_t)io.uio_offset;
2557 if (!cookies && !error)
2558 error = NFSERR_PERM;
2559 if (v3) {
2560 getret = VOP_GETATTR(vp, &at, cred, procp);
2561 if (!error)
2562 error = getret;
2563 }
2564
2565 VOP_UNLOCK(vp, 0);
2566 if (error) {
2567 vrele(vp);
2568 FREE((caddr_t)rbuf, M_TEMP);
2569 if (cookies)
2570 free((caddr_t)cookies, M_TEMP);
2571 nfsm_reply(NFSX_POSTOPATTR(v3));
2572 nfsm_srvpostop_attr(getret, &at);
2573 return (0);
2574 }
2575 if (io.uio_resid) {
2576 siz -= io.uio_resid;
2577
2578 /*
2579 * If nothing read, return eof
2580 * rpc reply
2581 */
2582 if (siz == 0) {
2583 vrele(vp);
2584 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2585 2 * NFSX_UNSIGNED);
2586 if (v3) {
2587 nfsm_srvpostop_attr(getret, &at);
2588 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2589 txdr_hyper(at.va_filerev, tl);
2590 tl += 2;
2591 } else
2592 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2593 *tl++ = nfs_false;
2594 *tl = nfs_true;
2595 FREE((caddr_t)rbuf, M_TEMP);
2596 FREE((caddr_t)cookies, M_TEMP);
2597 return (0);
2598 }
2599 }
2600
2601 /*
2602 * Check for degenerate cases of nothing useful read.
2603 * If so go try again
2604 */
2605 cpos = rbuf;
2606 cend = rbuf + siz;
2607 dp = (struct dirent *)cpos;
2608 cookiep = cookies;
2609
2610 while (cpos < cend && ncookies > 0 &&
2611 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2612 cpos += dp->d_reclen;
2613 dp = (struct dirent *)cpos;
2614 cookiep++;
2615 ncookies--;
2616 }
2617 if (cpos >= cend || ncookies == 0) {
2618 toff = off;
2619 siz = fullsiz;
2620 goto again;
2621 }
2622
2623 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
2624 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2625 if (v3) {
2626 nfsm_srvpostop_attr(getret, &at);
2627 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2628 txdr_hyper(at.va_filerev, tl);
2629 }
2630 mp = mp2 = mb;
2631 bp = bpos;
2632 be = bp + M_TRAILINGSPACE(mp);
2633
2634 /* Loop through the records and build reply */
2635 while (cpos < cend && ncookies > 0) {
2636 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2637 nlen = dp->d_namlen;
2638 rem = nfsm_rndup(nlen)-nlen;
2639 len += (4 * NFSX_UNSIGNED + nlen + rem);
2640 if (v3)
2641 len += 2 * NFSX_UNSIGNED;
2642 if (len > cnt) {
2643 eofflag = 0;
2644 break;
2645 }
2646 /*
2647 * Build the directory record xdr from
2648 * the dirent entry.
2649 */
2650 nfsm_clget;
2651 *tl = nfs_true;
2652 bp += NFSX_UNSIGNED;
2653 if (v3) {
2654 nfsm_clget;
2655 *tl = 0;
2656 bp += NFSX_UNSIGNED;
2657 }
2658 nfsm_clget;
2659 *tl = txdr_unsigned(dp->d_fileno);
2660 bp += NFSX_UNSIGNED;
2661 nfsm_clget;
2662 *tl = txdr_unsigned(nlen);
2663 bp += NFSX_UNSIGNED;
2664
2665 /* And loop around copying the name */
2666 xfer = nlen;
2667 cp = dp->d_name;
2668 while (xfer > 0) {
2669 nfsm_clget;
2670 if ((bp+xfer) > be)
2671 tsiz = be-bp;
2672 else
2673 tsiz = xfer;
2674 memcpy(bp, cp, tsiz);
2675 bp += tsiz;
2676 xfer -= tsiz;
2677 if (xfer > 0)
2678 cp += tsiz;
2679 }
2680 /* And null pad to an int32_t boundary */
2681 for (i = 0; i < rem; i++)
2682 *bp++ = '\0';
2683 nfsm_clget;
2684
2685 /* Finish off the record */
2686 txdr_hyper(*cookiep, &jar);
2687 if (v3) {
2688 *tl = jar.nfsuquad[0];
2689 bp += NFSX_UNSIGNED;
2690 nfsm_clget;
2691 }
2692 *tl = jar.nfsuquad[1];
2693 bp += NFSX_UNSIGNED;
2694 }
2695 cpos += dp->d_reclen;
2696 dp = (struct dirent *)cpos;
2697 cookiep++;
2698 ncookies--;
2699 }
2700 vrele(vp);
2701 nfsm_clget;
2702 *tl = nfs_false;
2703 bp += NFSX_UNSIGNED;
2704 nfsm_clget;
2705 if (eofflag)
2706 *tl = nfs_true;
2707 else
2708 *tl = nfs_false;
2709 bp += NFSX_UNSIGNED;
2710 if (mp != mb) {
2711 if (bp < be)
2712 mp->m_len = bp - mtod(mp, caddr_t);
2713 } else
2714 mp->m_len += bp - bpos;
2715 FREE((caddr_t)rbuf, M_TEMP);
2716 FREE((caddr_t)cookies, M_TEMP);
2717 nfsm_srvdone;
2718 }
2719
2720 int
2721 nfsrv_readdirplus(nfsd, slp, procp, mrq)
2722 struct nfsrv_descript *nfsd;
2723 struct nfssvc_sock *slp;
2724 struct proc *procp;
2725 struct mbuf **mrq;
2726 {
2727 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2728 struct mbuf *nam = nfsd->nd_nam;
2729 caddr_t dpos = nfsd->nd_dpos;
2730 struct ucred *cred = &nfsd->nd_cr;
2731 char *bp, *be;
2732 struct mbuf *mp;
2733 struct dirent *dp;
2734 caddr_t cp;
2735 u_int32_t *tl;
2736 int32_t t1;
2737 caddr_t bpos;
2738 struct mbuf *mb, *mb2, *mreq, *mp2;
2739 char *cpos, *cend, *cp2, *rbuf;
2740 struct vnode *vp, *nvp;
2741 struct flrep fl;
2742 nfsfh_t nfh;
2743 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
2744 struct uio io;
2745 struct iovec iv;
2746 struct vattr va, at, *vap = &va;
2747 struct nfs_fattr *fp;
2748 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2749 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies;
2750 u_quad_t frev, off, toff, verf;
2751 off_t *cookies = NULL, *cookiep;
2752
2753 fhp = &nfh.fh_generic;
2754 nfsm_srvmtofh(fhp);
2755 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2756 toff = fxdr_hyper(tl);
2757 tl += 2;
2758 verf = fxdr_hyper(tl);
2759 tl += 2;
2760 siz = fxdr_unsigned(int, *tl++);
2761 cnt = fxdr_unsigned(int, *tl);
2762 off = toff;
2763 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2764 xfer = NFS_SRVMAXDATA(nfsd);
2765 if (siz > xfer)
2766 siz = xfer;
2767 fullsiz = siz;
2768 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2769 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2770 if (!error && vp->v_type != VDIR) {
2771 error = ENOTDIR;
2772 vput(vp);
2773 }
2774 if (error) {
2775 nfsm_reply(NFSX_UNSIGNED);
2776 nfsm_srvpostop_attr(getret, &at);
2777 return (0);
2778 }
2779 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2780 #ifdef NFS3_STRICTVERF
2781 /*
2782 * XXX This check is too strict for Solaris 2.5 clients.
2783 */
2784 if (!error && toff && verf != at.va_filerev)
2785 error = NFSERR_BAD_COOKIE;
2786 #endif
2787 if (!error) {
2788 nqsrv_getl(vp, ND_READ);
2789 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2790 }
2791 if (error) {
2792 vput(vp);
2793 nfsm_reply(NFSX_V3POSTOPATTR);
2794 nfsm_srvpostop_attr(getret, &at);
2795 return (0);
2796 }
2797 VOP_UNLOCK(vp, 0);
2798
2799 MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
2800 again:
2801 iv.iov_base = rbuf;
2802 iv.iov_len = fullsiz;
2803 io.uio_iov = &iv;
2804 io.uio_iovcnt = 1;
2805 io.uio_offset = (off_t)off;
2806 io.uio_resid = fullsiz;
2807 io.uio_segflg = UIO_SYSSPACE;
2808 io.uio_rw = UIO_READ;
2809 io.uio_procp = (struct proc *)0;
2810 eofflag = 0;
2811
2812 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2813
2814 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2815
2816 off = (u_quad_t)io.uio_offset;
2817 getret = VOP_GETATTR(vp, &at, cred, procp);
2818
2819 VOP_UNLOCK(vp, 0);
2820
2821 /*
2822 * If the VGET operation doesn't work for this filesystem,
2823 * we can't support readdirplus. Returning NOTSUPP should
2824 * make clients fall back to plain readdir.
2825 * There's no need to check for VPTOFH as well, we wouldn't
2826 * even be here otherwise.
2827 */
2828 if (!getret) {
2829 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
2830 getret = (getret == EOPNOTSUPP) ?
2831 NFSERR_NOTSUPP : NFSERR_IO;
2832 else
2833 vput(nvp);
2834 }
2835
2836 if (!cookies && !error)
2837 error = NFSERR_PERM;
2838 if (!error)
2839 error = getret;
2840 if (error) {
2841 vrele(vp);
2842 if (cookies)
2843 FREE((caddr_t)cookies, M_TEMP);
2844 FREE((caddr_t)rbuf, M_TEMP);
2845 nfsm_reply(NFSX_V3POSTOPATTR);
2846 nfsm_srvpostop_attr(getret, &at);
2847 return (0);
2848 }
2849 if (io.uio_resid) {
2850 siz -= io.uio_resid;
2851
2852 /*
2853 * If nothing read, return eof
2854 * rpc reply
2855 */
2856 if (siz == 0) {
2857 vrele(vp);
2858 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
2859 2 * NFSX_UNSIGNED);
2860 nfsm_srvpostop_attr(getret, &at);
2861 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2862 txdr_hyper(at.va_filerev, tl);
2863 tl += 2;
2864 *tl++ = nfs_false;
2865 *tl = nfs_true;
2866 FREE((caddr_t)cookies, M_TEMP);
2867 FREE((caddr_t)rbuf, M_TEMP);
2868 return (0);
2869 }
2870 }
2871
2872 /*
2873 * Check for degenerate cases of nothing useful read.
2874 * If so go try again
2875 */
2876 cpos = rbuf;
2877 cend = rbuf + siz;
2878 dp = (struct dirent *)cpos;
2879 cookiep = cookies;
2880
2881 while (cpos < cend && ncookies > 0 &&
2882 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2883 cpos += dp->d_reclen;
2884 dp = (struct dirent *)cpos;
2885 cookiep++;
2886 ncookies--;
2887 }
2888 if (cpos >= cend || ncookies == 0) {
2889 toff = off;
2890 siz = fullsiz;
2891 goto again;
2892 }
2893
2894 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
2895 nfsm_reply(cnt);
2896 nfsm_srvpostop_attr(getret, &at);
2897 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2898 txdr_hyper(at.va_filerev, tl);
2899 mp = mp2 = mb;
2900 bp = bpos;
2901 be = bp + M_TRAILINGSPACE(mp);
2902
2903 /* Loop through the records and build reply */
2904 while (cpos < cend && ncookies > 0) {
2905 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2906 nlen = dp->d_namlen;
2907 rem = nfsm_rndup(nlen)-nlen;
2908
2909 /*
2910 * For readdir_and_lookup get the vnode using
2911 * the file number.
2912 */
2913 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
2914 goto invalid;
2915 memset((caddr_t)nfhp, 0, NFSX_V3FH);
2916 nfhp->fh_fsid =
2917 nvp->v_mount->mnt_stat.f_fsid;
2918 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
2919 vput(nvp);
2920 goto invalid;
2921 }
2922 if (VOP_GETATTR(nvp, vap, cred, procp)) {
2923 vput(nvp);
2924 goto invalid;
2925 }
2926 vput(nvp);
2927
2928 /*
2929 * If either the dircount or maxcount will be
2930 * exceeded, get out now. Both of these lengths
2931 * are calculated conservatively, including all
2932 * XDR overheads.
2933 */
2934 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
2935 NFSX_V3POSTOPATTR);
2936 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
2937 if (len > cnt || dirlen > fullsiz) {
2938 eofflag = 0;
2939 break;
2940 }
2941
2942 /*
2943 * Build the directory record xdr from
2944 * the dirent entry.
2945 */
2946 fp = (struct nfs_fattr *)&fl.fl_fattr;
2947 nfsm_srvfillattr(vap, fp);
2948 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
2949 fl.fl_fhok = nfs_true;
2950 fl.fl_postopok = nfs_true;
2951 txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
2952
2953 nfsm_clget;
2954 *tl = nfs_true;
2955 bp += NFSX_UNSIGNED;
2956 nfsm_clget;
2957 *tl = 0;
2958 bp += NFSX_UNSIGNED;
2959 nfsm_clget;
2960 *tl = txdr_unsigned(dp->d_fileno);
2961 bp += NFSX_UNSIGNED;
2962 nfsm_clget;
2963 *tl = txdr_unsigned(nlen);
2964 bp += NFSX_UNSIGNED;
2965
2966 /* And loop around copying the name */
2967 xfer = nlen;
2968 cp = dp->d_name;
2969 while (xfer > 0) {
2970 nfsm_clget;
2971 if ((bp + xfer) > be)
2972 tsiz = be - bp;
2973 else
2974 tsiz = xfer;
2975 memcpy(bp, cp, tsiz);
2976 bp += tsiz;
2977 xfer -= tsiz;
2978 if (xfer > 0)
2979 cp += tsiz;
2980 }
2981 /* And null pad to an int32_t boundary */
2982 for (i = 0; i < rem; i++)
2983 *bp++ = '\0';
2984
2985 /*
2986 * Now copy the flrep structure out.
2987 */
2988 xfer = sizeof (struct flrep);
2989 cp = (caddr_t)&fl;
2990 while (xfer > 0) {
2991 nfsm_clget;
2992 if ((bp + xfer) > be)
2993 tsiz = be - bp;
2994 else
2995 tsiz = xfer;
2996 memcpy(bp, cp, tsiz);
2997 bp += tsiz;
2998 xfer -= tsiz;
2999 if (xfer > 0)
3000 cp += tsiz;
3001 }
3002 }
3003 invalid:
3004 cpos += dp->d_reclen;
3005 dp = (struct dirent *)cpos;
3006 cookiep++;
3007 ncookies--;
3008 }
3009 vrele(vp);
3010 nfsm_clget;
3011 *tl = nfs_false;
3012 bp += NFSX_UNSIGNED;
3013 nfsm_clget;
3014 if (eofflag)
3015 *tl = nfs_true;
3016 else
3017 *tl = nfs_false;
3018 bp += NFSX_UNSIGNED;
3019 if (mp != mb) {
3020 if (bp < be)
3021 mp->m_len = bp - mtod(mp, caddr_t);
3022 } else
3023 mp->m_len += bp - bpos;
3024 FREE((caddr_t)cookies, M_TEMP);
3025 FREE((caddr_t)rbuf, M_TEMP);
3026 nfsm_srvdone;
3027 }
3028
3029 /*
3030 * nfs commit service
3031 */
3032 int
3033 nfsrv_commit(nfsd, slp, procp, mrq)
3034 struct nfsrv_descript *nfsd;
3035 struct nfssvc_sock *slp;
3036 struct proc *procp;
3037 struct mbuf **mrq;
3038 {
3039 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3040 struct mbuf *nam = nfsd->nd_nam;
3041 caddr_t dpos = nfsd->nd_dpos;
3042 struct ucred *cred = &nfsd->nd_cr;
3043 struct vattr bfor, aft;
3044 struct vnode *vp;
3045 nfsfh_t nfh;
3046 fhandle_t *fhp;
3047 u_int32_t *tl;
3048 int32_t t1;
3049 caddr_t bpos;
3050 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache;
3051 char *cp2;
3052 struct mbuf *mb, *mb2, *mreq;
3053 u_quad_t frev, off;
3054
3055 #ifndef nolint
3056 cache = 0;
3057 #endif
3058 fhp = &nfh.fh_generic;
3059 nfsm_srvmtofh(fhp);
3060 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3061
3062 /*
3063 * XXX At this time VOP_FSYNC() does not accept offset and byte
3064 * count parameters, so these arguments are useless (someday maybe).
3065 */
3066 off = fxdr_hyper(tl);
3067 tl += 2;
3068 cnt = fxdr_unsigned(int, *tl);
3069 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3070 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3071 if (error) {
3072 nfsm_reply(2 * NFSX_UNSIGNED);
3073 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3074 return (0);
3075 }
3076 for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
3077 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, procp);
3078 aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
3079 vput(vp);
3080 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3081 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3082 if (!error) {
3083 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3084 *tl++ = txdr_unsigned(boottime.tv_sec);
3085 *tl = txdr_unsigned(boottime.tv_usec);
3086 } else
3087 return (0);
3088 nfsm_srvdone;
3089 }
3090
3091 /*
3092 * nfs statfs service
3093 */
3094 int
3095 nfsrv_statfs(nfsd, slp, procp, mrq)
3096 struct nfsrv_descript *nfsd;
3097 struct nfssvc_sock *slp;
3098 struct proc *procp;
3099 struct mbuf **mrq;
3100 {
3101 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3102 struct mbuf *nam = nfsd->nd_nam;
3103 caddr_t dpos = nfsd->nd_dpos;
3104 struct ucred *cred = &nfsd->nd_cr;
3105 struct statfs *sf;
3106 struct nfs_statfs *sfp;
3107 u_int32_t *tl;
3108 int32_t t1;
3109 caddr_t bpos;
3110 int error = 0, rdonly, cache, getret = 1;
3111 int v3 = (nfsd->nd_flag & ND_NFSV3);
3112 char *cp2;
3113 struct mbuf *mb, *mb2, *mreq;
3114 struct vnode *vp;
3115 struct vattr at;
3116 nfsfh_t nfh;
3117 fhandle_t *fhp;
3118 struct statfs statfs;
3119 u_quad_t frev, tval;
3120
3121 #ifndef nolint
3122 cache = 0;
3123 #endif
3124 fhp = &nfh.fh_generic;
3125 nfsm_srvmtofh(fhp);
3126 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3127 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3128 if (error) {
3129 nfsm_reply(NFSX_UNSIGNED);
3130 nfsm_srvpostop_attr(getret, &at);
3131 return (0);
3132 }
3133 sf = &statfs;
3134 error = VFS_STATFS(vp->v_mount, sf, procp);
3135 getret = VOP_GETATTR(vp, &at, cred, procp);
3136 vput(vp);
3137 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3138 if (v3)
3139 nfsm_srvpostop_attr(getret, &at);
3140 if (error)
3141 return (0);
3142 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3143 if (v3) {
3144 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_bsize);
3145 txdr_hyper(tval, &sfp->sf_tbytes);
3146 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_bsize);
3147 txdr_hyper(tval, &sfp->sf_fbytes);
3148 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_bsize);
3149 txdr_hyper(tval, &sfp->sf_abytes);
3150 tval = (u_quad_t)sf->f_files;
3151 txdr_hyper(tval, &sfp->sf_tfiles);
3152 tval = (u_quad_t)sf->f_ffree;
3153 txdr_hyper(tval, &sfp->sf_ffiles);
3154 txdr_hyper(tval, &sfp->sf_afiles);
3155 sfp->sf_invarsec = 0;
3156 } else {
3157 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3158 sfp->sf_bsize = txdr_unsigned(sf->f_bsize);
3159 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3160 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3161 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3162 }
3163 nfsm_srvdone;
3164 }
3165
3166 /*
3167 * nfs fsinfo service
3168 */
3169 int
3170 nfsrv_fsinfo(nfsd, slp, procp, mrq)
3171 struct nfsrv_descript *nfsd;
3172 struct nfssvc_sock *slp;
3173 struct proc *procp;
3174 struct mbuf **mrq;
3175 {
3176 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3177 struct mbuf *nam = nfsd->nd_nam;
3178 caddr_t dpos = nfsd->nd_dpos;
3179 struct ucred *cred = &nfsd->nd_cr;
3180 u_int32_t *tl;
3181 struct nfsv3_fsinfo *sip;
3182 int32_t t1;
3183 caddr_t bpos;
3184 int error = 0, rdonly, cache, getret = 1, pref;
3185 char *cp2;
3186 struct mbuf *mb, *mb2, *mreq;
3187 struct vnode *vp;
3188 struct vattr at;
3189 nfsfh_t nfh;
3190 fhandle_t *fhp;
3191 u_quad_t frev, maxfsize;
3192 struct statfs sb;
3193
3194 #ifndef nolint
3195 cache = 0;
3196 #endif
3197 fhp = &nfh.fh_generic;
3198 nfsm_srvmtofh(fhp);
3199 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3200 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3201 if (error) {
3202 nfsm_reply(NFSX_UNSIGNED);
3203 nfsm_srvpostop_attr(getret, &at);
3204 return (0);
3205 }
3206
3207 /* XXX Try to make a guess on the max file size. */
3208 VFS_STATFS(vp->v_mount, &sb, (struct proc *)0);
3209 maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1;
3210
3211 getret = VOP_GETATTR(vp, &at, cred, procp);
3212 vput(vp);
3213 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3214 nfsm_srvpostop_attr(getret, &at);
3215 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3216
3217 /*
3218 * XXX
3219 * There should be file system VFS OP(s) to get this information.
3220 * For now, assume ufs.
3221 */
3222 if (slp->ns_so->so_type == SOCK_DGRAM)
3223 pref = NFS_MAXDGRAMDATA;
3224 else
3225 pref = NFS_MAXDATA;
3226 sip->fs_rtmax = txdr_unsigned(NFS_MAXDATA);
3227 sip->fs_rtpref = txdr_unsigned(pref);
3228 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3229 sip->fs_wtmax = txdr_unsigned(NFS_MAXDATA);
3230 sip->fs_wtpref = txdr_unsigned(pref);
3231 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3232 sip->fs_dtpref = txdr_unsigned(pref);
3233 txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3234 sip->fs_timedelta.nfsv3_sec = 0;
3235 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3236 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3237 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3238 NFSV3FSINFO_CANSETTIME);
3239 nfsm_srvdone;
3240 }
3241
3242 /*
3243 * nfs pathconf service
3244 */
3245 int
3246 nfsrv_pathconf(nfsd, slp, procp, mrq)
3247 struct nfsrv_descript *nfsd;
3248 struct nfssvc_sock *slp;
3249 struct proc *procp;
3250 struct mbuf **mrq;
3251 {
3252 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3253 struct mbuf *nam = nfsd->nd_nam;
3254 caddr_t dpos = nfsd->nd_dpos;
3255 struct ucred *cred = &nfsd->nd_cr;
3256 u_int32_t *tl;
3257 struct nfsv3_pathconf *pc;
3258 int32_t t1;
3259 caddr_t bpos;
3260 int error = 0, rdonly, cache, getret = 1;
3261 register_t linkmax, namemax, chownres, notrunc;
3262 char *cp2;
3263 struct mbuf *mb, *mb2, *mreq;
3264 struct vnode *vp;
3265 struct vattr at;
3266 nfsfh_t nfh;
3267 fhandle_t *fhp;
3268 u_quad_t frev;
3269
3270 #ifndef nolint
3271 cache = 0;
3272 #endif
3273 fhp = &nfh.fh_generic;
3274 nfsm_srvmtofh(fhp);
3275 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3276 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3277 if (error) {
3278 nfsm_reply(NFSX_UNSIGNED);
3279 nfsm_srvpostop_attr(getret, &at);
3280 return (0);
3281 }
3282 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3283 if (!error)
3284 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3285 if (!error)
3286 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3287 if (!error)
3288 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc);
3289 getret = VOP_GETATTR(vp, &at, cred, procp);
3290 vput(vp);
3291 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3292 nfsm_srvpostop_attr(getret, &at);
3293 if (error)
3294 return (0);
3295 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3296
3297 pc->pc_linkmax = txdr_unsigned(linkmax);
3298 pc->pc_namemax = txdr_unsigned(namemax);
3299 pc->pc_notrunc = txdr_unsigned(notrunc);
3300 pc->pc_chownrestricted = txdr_unsigned(chownres);
3301
3302 /*
3303 * These should probably be supported by VOP_PATHCONF(), but
3304 * until msdosfs is exportable (why would you want to?), the
3305 * Unix defaults should be ok.
3306 */
3307 pc->pc_caseinsensitive = nfs_false;
3308 pc->pc_casepreserving = nfs_true;
3309 nfsm_srvdone;
3310 }
3311
3312 /*
3313 * Null operation, used by clients to ping server
3314 */
3315 /* ARGSUSED */
3316 int
3317 nfsrv_null(nfsd, slp, procp, mrq)
3318 struct nfsrv_descript *nfsd;
3319 struct nfssvc_sock *slp;
3320 struct proc *procp;
3321 struct mbuf **mrq;
3322 {
3323 struct mbuf *mrep = nfsd->nd_mrep;
3324 caddr_t bpos;
3325 int error = NFSERR_RETVOID, cache = 0;
3326 struct mbuf *mb, *mreq;
3327 u_quad_t frev;
3328
3329 nfsm_reply(0);
3330 return (0);
3331 }
3332
3333 /*
3334 * No operation, used for obsolete procedures
3335 */
3336 /* ARGSUSED */
3337 int
3338 nfsrv_noop(nfsd, slp, procp, mrq)
3339 struct nfsrv_descript *nfsd;
3340 struct nfssvc_sock *slp;
3341 struct proc *procp;
3342 struct mbuf **mrq;
3343 {
3344 struct mbuf *mrep = nfsd->nd_mrep;
3345 caddr_t bpos;
3346 int error, cache = 0;
3347 struct mbuf *mb, *mreq;
3348 u_quad_t frev;
3349
3350 if (nfsd->nd_repstat)
3351 error = nfsd->nd_repstat;
3352 else
3353 error = EPROCUNAVAIL;
3354 nfsm_reply(0);
3355 return (0);
3356 }
3357
3358 /*
3359 * Perform access checking for vnodes obtained from file handles that would
3360 * refer to files already opened by a Unix client. You cannot just use
3361 * vn_writechk() and VOP_ACCESS() for two reasons.
3362 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3363 * 2 - The owner is to be given access irrespective of mode bits for some
3364 * operations, so that processes that chmod after opening a file don't
3365 * break. I don't like this because it opens a security hole, but since
3366 * the nfs server opens a security hole the size of a barn door anyhow,
3367 * what the heck.
3368 *
3369 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3370 * will return EPERM instead of EACCESS. EPERM is always an error.
3371 */
3372 int
3373 nfsrv_access(vp, flags, cred, rdonly, p, override)
3374 struct vnode *vp;
3375 int flags;
3376 struct ucred *cred;
3377 int rdonly;
3378 struct proc *p;
3379 {
3380 struct vattr vattr;
3381 int error;
3382 if (flags & VWRITE) {
3383 /* Just vn_writechk() changed to check rdonly */
3384 /*
3385 * Disallow write attempts on read-only file systems;
3386 * unless the file is a socket or a block or character
3387 * device resident on the file system.
3388 */
3389 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3390 switch (vp->v_type) {
3391 case VREG:
3392 case VDIR:
3393 case VLNK:
3394 return (EROFS);
3395 default:
3396 break;
3397 }
3398 }
3399 /*
3400 * If there's shared text associated with
3401 * the inode, try to free it up once. If
3402 * we fail, we can't allow writing.
3403 */
3404 if ((vp->v_flag & VTEXT) && !uvm_vnp_uncache(vp))
3405 return (ETXTBSY);
3406 }
3407 error = VOP_GETATTR(vp, &vattr, cred, p);
3408 if (error)
3409 return (error);
3410 error = VOP_ACCESS(vp, flags, cred, p);
3411 /*
3412 * Allow certain operations for the owner (reads and writes
3413 * on files that are already open).
3414 */
3415 if (override && error == EACCES && cred->cr_uid == vattr.va_uid)
3416 error = 0;
3417 return error;
3418 }
3419