nfs_serv.c revision 1.51 1 /* $NetBSD: nfs_serv.c,v 1.51 1999/05/04 16:01:37 sommerfe 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 <vm/vm.h>
77
78 #include <uvm/uvm_extern.h>
79
80 #include <nfs/nfsproto.h>
81 #include <nfs/rpcv2.h>
82 #include <nfs/nfs.h>
83 #include <nfs/xdr_subs.h>
84 #include <nfs/nfsm_subs.h>
85 #include <nfs/nqnfs.h>
86 #include <nfs/nfs_var.h>
87
88 /* Global vars */
89 extern u_int32_t nfs_xdrneg1;
90 extern u_int32_t nfs_false, nfs_true;
91 extern enum vtype nv3tov_type[8];
92 extern struct nfsstats nfsstats;
93 extern nfstype nfsv2_type[9];
94 extern nfstype nfsv3_type[9];
95 extern struct nfs_public nfs_pub;
96 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
97
98 /*
99 * nfs v3 access service
100 */
101 int
102 nfsrv3_access(nfsd, slp, procp, mrq)
103 struct nfsrv_descript *nfsd;
104 struct nfssvc_sock *slp;
105 struct proc *procp;
106 struct mbuf **mrq;
107 {
108 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
109 struct mbuf *nam = nfsd->nd_nam;
110 caddr_t dpos = nfsd->nd_dpos;
111 struct ucred *cred = &nfsd->nd_cr;
112 struct vnode *vp;
113 nfsfh_t nfh;
114 fhandle_t *fhp;
115 register u_int32_t *tl;
116 register int32_t t1;
117 caddr_t bpos;
118 int error = 0, rdonly, cache = 0, getret;
119 char *cp2;
120 struct mbuf *mb, *mreq, *mb2;
121 struct vattr va;
122 u_long inmode, testmode, outmode;
123 u_quad_t frev;
124
125 fhp = &nfh.fh_generic;
126 nfsm_srvmtofh(fhp);
127 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
128 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
129 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
130 if (error) {
131 nfsm_reply(NFSX_UNSIGNED);
132 nfsm_srvpostop_attr(1, (struct vattr *)0);
133 return (0);
134 }
135 inmode = fxdr_unsigned(u_int32_t, *tl);
136 outmode = 0;
137 if ((inmode & NFSV3ACCESS_READ) &&
138 nfsrv_access(vp, VREAD, cred, rdonly, procp, 0) == 0)
139 outmode |= NFSV3ACCESS_READ;
140 if (vp->v_type != VDIR) {
141 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
142 if (testmode &&
143 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0)
144 outmode |= testmode;
145 if ((inmode & NFSV3ACCESS_EXECUTE) &&
146 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
147 outmode |= NFSV3ACCESS_EXECUTE;
148 } else {
149 testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
150 NFSV3ACCESS_DELETE);
151 if (testmode &&
152 nfsrv_access(vp, VWRITE, cred, rdonly, procp, 0) == 0)
153 outmode |= testmode;
154 if ((inmode & NFSV3ACCESS_LOOKUP) &&
155 nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
156 outmode |= NFSV3ACCESS_LOOKUP;
157 }
158 getret = VOP_GETATTR(vp, &va, cred, procp);
159 vput(vp);
160 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
161 nfsm_srvpostop_attr(getret, &va);
162 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
163 *tl = txdr_unsigned(outmode);
164 nfsm_srvdone;
165 }
166
167 /*
168 * nfs getattr service
169 */
170 int
171 nfsrv_getattr(nfsd, slp, procp, mrq)
172 struct nfsrv_descript *nfsd;
173 struct nfssvc_sock *slp;
174 struct proc *procp;
175 struct mbuf **mrq;
176 {
177 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
178 struct mbuf *nam = nfsd->nd_nam;
179 caddr_t dpos = nfsd->nd_dpos;
180 struct ucred *cred = &nfsd->nd_cr;
181 register struct nfs_fattr *fp;
182 struct vattr va;
183 struct vnode *vp;
184 nfsfh_t nfh;
185 fhandle_t *fhp;
186 register u_int32_t *tl;
187 register int32_t t1;
188 caddr_t bpos;
189 int error = 0, rdonly, cache;
190 char *cp2;
191 struct mbuf *mb, *mb2, *mreq;
192 u_quad_t frev;
193
194 fhp = &nfh.fh_generic;
195 nfsm_srvmtofh(fhp);
196 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
197 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
198 if (error) {
199 nfsm_reply(0);
200 return (0);
201 }
202 nqsrv_getl(vp, ND_READ);
203 error = VOP_GETATTR(vp, &va, cred, procp);
204 vput(vp);
205 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
206 if (error)
207 return (0);
208 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
209 nfsm_srvfillattr(&va, fp);
210 nfsm_srvdone;
211 }
212
213 /*
214 * nfs setattr service
215 */
216 int
217 nfsrv_setattr(nfsd, slp, procp, mrq)
218 struct nfsrv_descript *nfsd;
219 struct nfssvc_sock *slp;
220 struct proc *procp;
221 struct mbuf **mrq;
222 {
223 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
224 struct mbuf *nam = nfsd->nd_nam;
225 caddr_t dpos = nfsd->nd_dpos;
226 struct ucred *cred = &nfsd->nd_cr;
227 struct vattr va, preat;
228 register struct nfsv2_sattr *sp;
229 register struct nfs_fattr *fp;
230 struct vnode *vp;
231 nfsfh_t nfh;
232 fhandle_t *fhp;
233 register u_int32_t *tl;
234 register int32_t t1;
235 caddr_t bpos;
236 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1;
237 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
238 char *cp2;
239 struct mbuf *mb, *mb2, *mreq;
240 u_quad_t frev;
241 struct timespec guard;
242
243 fhp = &nfh.fh_generic;
244 nfsm_srvmtofh(fhp);
245 VATTR_NULL(&va);
246 if (v3) {
247 nfsm_srvsattr(&va);
248 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
249 gcheck = fxdr_unsigned(int, *tl);
250 if (gcheck) {
251 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
252 fxdr_nfsv3time(tl, &guard);
253 }
254 } else {
255 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
256 /*
257 * Nah nah nah nah na nah
258 * There is a bug in the Sun client that puts 0xffff in the mode
259 * field of sattr when it should put in 0xffffffff. The u_short
260 * doesn't sign extend.
261 * --> check the low order 2 bytes for 0xffff
262 */
263 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
264 va.va_mode = nfstov_mode(sp->sa_mode);
265 if (sp->sa_uid != nfs_xdrneg1)
266 va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
267 if (sp->sa_gid != nfs_xdrneg1)
268 va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
269 if (sp->sa_size != nfs_xdrneg1)
270 va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
271 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
272 #ifdef notyet
273 fxdr_nfsv2time(&sp->sa_atime, &va.va_atime);
274 #else
275 va.va_atime.tv_sec =
276 fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
277 va.va_atime.tv_nsec = 0;
278 #endif
279 }
280 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
281 fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime);
282
283 }
284
285 /*
286 * Now that we have all the fields, lets do it.
287 */
288 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
289 (nfsd->nd_flag & ND_KERBAUTH), FALSE);
290 if (error) {
291 nfsm_reply(2 * NFSX_UNSIGNED);
292 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
293 return (0);
294 }
295 nqsrv_getl(vp, ND_WRITE);
296 if (v3) {
297 error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp);
298 if (!error && gcheck &&
299 (preat.va_ctime.tv_sec != guard.tv_sec ||
300 preat.va_ctime.tv_nsec != guard.tv_nsec))
301 error = NFSERR_NOT_SYNC;
302 if (error) {
303 vput(vp);
304 nfsm_reply(NFSX_WCCDATA(v3));
305 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
306 return (0);
307 }
308 }
309
310 /*
311 * If the size is being changed write acces is required, otherwise
312 * just check for a read only file system.
313 */
314 if (va.va_size == ((u_quad_t)((quad_t) -1))) {
315 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
316 error = EROFS;
317 goto out;
318 }
319 } else {
320 if (vp->v_type == VDIR) {
321 error = EISDIR;
322 goto out;
323 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
324 procp, 0)) != 0)
325 goto out;
326 }
327 error = VOP_SETATTR(vp, &va, cred, procp);
328 postat_ret = VOP_GETATTR(vp, &va, cred, procp);
329 if (!error)
330 error = postat_ret;
331 out:
332 vput(vp);
333 nfsm_reply(NFSX_WCCORFATTR(v3));
334 if (v3) {
335 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
336 return (0);
337 } else {
338 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
339 nfsm_srvfillattr(&va, fp);
340 }
341 nfsm_srvdone;
342 }
343
344 /*
345 * nfs lookup rpc
346 */
347 int
348 nfsrv_lookup(nfsd, slp, procp, mrq)
349 struct nfsrv_descript *nfsd;
350 struct nfssvc_sock *slp;
351 struct proc *procp;
352 struct mbuf **mrq;
353 {
354 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
355 struct mbuf *nam = nfsd->nd_nam;
356 caddr_t dpos = nfsd->nd_dpos;
357 struct ucred *cred = &nfsd->nd_cr;
358 register struct nfs_fattr *fp;
359 struct nameidata nd, ind, *ndp = &nd;
360 struct vnode *vp, *dirp;
361 nfsfh_t nfh;
362 fhandle_t *fhp;
363 register caddr_t cp;
364 register u_int32_t *tl;
365 register int32_t t1;
366 caddr_t bpos;
367 int error = 0, cache, len, dirattr_ret = 1;
368 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
369 char *cp2;
370 struct mbuf *mb, *mb2, *mreq;
371 struct vattr va, dirattr;
372 u_quad_t frev;
373
374 fhp = &nfh.fh_generic;
375 nfsm_srvmtofh(fhp);
376 nfsm_srvnamesiz(len);
377
378 pubflag = nfs_ispublicfh(fhp);
379
380 nd.ni_cnd.cn_cred = cred;
381 nd.ni_cnd.cn_nameiop = LOOKUP;
382 nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
383 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
384 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
385
386 if (!error && pubflag) {
387 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
388 /*
389 * Setup call to lookup() to see if we can find
390 * the index file. Arguably, this doesn't belong
391 * in a kernel.. Ugh.
392 */
393 ind = nd;
394 VOP_UNLOCK(nd.ni_vp, 0);
395 ind.ni_pathlen = strlen(nfs_pub.np_index);
396 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
397 nfs_pub.np_index;
398 ind.ni_startdir = nd.ni_vp;
399 VREF(ind.ni_startdir);
400 error = lookup(&ind);
401 if (!error) {
402 /*
403 * Found an index file. Get rid of
404 * the old references.
405 */
406 if (dirp)
407 vrele(dirp);
408 dirp = nd.ni_vp;
409 vrele(nd.ni_startdir);
410 ndp = &ind;
411 } else
412 error = 0;
413 }
414 /*
415 * If the public filehandle was used, check that this lookup
416 * didn't result in a filehandle outside the publicly exported
417 * filesystem.
418 */
419
420 if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
421 vput(nd.ni_vp);
422 error = EPERM;
423 }
424 }
425
426 if (dirp) {
427 if (v3)
428 dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred,
429 procp);
430 vrele(dirp);
431 }
432
433 if (error) {
434 nfsm_reply(NFSX_POSTOPATTR(v3));
435 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
436 return (0);
437 }
438
439 nqsrv_getl(ndp->ni_startdir, ND_READ);
440 vrele(ndp->ni_startdir);
441 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
442 vp = ndp->ni_vp;
443 memset((caddr_t)fhp, 0, sizeof(nfh));
444 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
445 error = VFS_VPTOFH(vp, &fhp->fh_fid);
446 if (!error)
447 error = VOP_GETATTR(vp, &va, cred, procp);
448 vput(vp);
449 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
450 if (error) {
451 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
452 return (0);
453 }
454 nfsm_srvfhtom(fhp, v3);
455 if (v3) {
456 nfsm_srvpostop_attr(0, &va);
457 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
458 } else {
459 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
460 nfsm_srvfillattr(&va, fp);
461 }
462 nfsm_srvdone;
463 }
464
465 /*
466 * nfs readlink service
467 */
468 int
469 nfsrv_readlink(nfsd, slp, procp, mrq)
470 struct nfsrv_descript *nfsd;
471 struct nfssvc_sock *slp;
472 struct proc *procp;
473 struct mbuf **mrq;
474 {
475 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
476 struct mbuf *nam = nfsd->nd_nam;
477 caddr_t dpos = nfsd->nd_dpos;
478 struct ucred *cred = &nfsd->nd_cr;
479 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
480 register struct iovec *ivp = iv;
481 register struct mbuf *mp;
482 register u_int32_t *tl;
483 register int32_t t1;
484 caddr_t bpos;
485 int error = 0, rdonly, cache, i, tlen, len, getret;
486 int v3 = (nfsd->nd_flag & ND_NFSV3);
487 char *cp2;
488 struct mbuf *mb, *mb2, *mp2 = NULL, *mp3 = NULL, *mreq;
489 struct vnode *vp;
490 struct vattr attr;
491 nfsfh_t nfh;
492 fhandle_t *fhp;
493 struct uio io, *uiop = &io;
494 u_quad_t frev;
495
496 fhp = &nfh.fh_generic;
497 nfsm_srvmtofh(fhp);
498 len = 0;
499 i = 0;
500 while (len < NFS_MAXPATHLEN) {
501 MGET(mp, M_WAIT, MT_DATA);
502 MCLGET(mp, M_WAIT);
503 mp->m_len = NFSMSIZ(mp);
504 if (len == 0)
505 mp3 = mp2 = mp;
506 else {
507 mp2->m_next = mp;
508 mp2 = mp;
509 }
510 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
511 mp->m_len = NFS_MAXPATHLEN-len;
512 len = NFS_MAXPATHLEN;
513 } else
514 len += mp->m_len;
515 ivp->iov_base = mtod(mp, caddr_t);
516 ivp->iov_len = mp->m_len;
517 i++;
518 ivp++;
519 }
520 uiop->uio_iov = iv;
521 uiop->uio_iovcnt = i;
522 uiop->uio_offset = 0;
523 uiop->uio_resid = len;
524 uiop->uio_rw = UIO_READ;
525 uiop->uio_segflg = UIO_SYSSPACE;
526 uiop->uio_procp = (struct proc *)0;
527 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
528 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
529 if (error) {
530 m_freem(mp3);
531 nfsm_reply(2 * NFSX_UNSIGNED);
532 nfsm_srvpostop_attr(1, (struct vattr *)0);
533 return (0);
534 }
535 if (vp->v_type != VLNK) {
536 if (v3)
537 error = EINVAL;
538 else
539 error = ENXIO;
540 goto out;
541 }
542 nqsrv_getl(vp, ND_READ);
543 error = VOP_READLINK(vp, uiop, cred);
544 out:
545 getret = VOP_GETATTR(vp, &attr, cred, procp);
546 vput(vp);
547 if (error)
548 m_freem(mp3);
549 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
550 if (v3) {
551 nfsm_srvpostop_attr(getret, &attr);
552 if (error)
553 return (0);
554 }
555 if (uiop->uio_resid > 0) {
556 len -= uiop->uio_resid;
557 tlen = nfsm_rndup(len);
558 nfsm_adj(mp3, NFS_MAXPATHLEN-tlen, tlen-len);
559 }
560 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
561 *tl = txdr_unsigned(len);
562 mb->m_next = mp3;
563 nfsm_srvdone;
564 }
565
566 /*
567 * nfs read service
568 */
569 int
570 nfsrv_read(nfsd, slp, procp, mrq)
571 struct nfsrv_descript *nfsd;
572 struct nfssvc_sock *slp;
573 struct proc *procp;
574 struct mbuf **mrq;
575 {
576 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
577 struct mbuf *nam = nfsd->nd_nam;
578 caddr_t dpos = nfsd->nd_dpos;
579 struct ucred *cred = &nfsd->nd_cr;
580 register struct iovec *iv;
581 struct iovec *iv2;
582 register struct mbuf *m;
583 register struct nfs_fattr *fp;
584 register u_int32_t *tl;
585 register int32_t t1;
586 register int i;
587 caddr_t bpos;
588 int error = 0, rdonly, cache, cnt, len, left, siz, tlen, getret;
589 int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen;
590 char *cp2;
591 struct mbuf *mb, *mb2, *mreq;
592 struct mbuf *m2;
593 struct vnode *vp;
594 nfsfh_t nfh;
595 fhandle_t *fhp;
596 struct uio io, *uiop = &io;
597 struct vattr va;
598 off_t off;
599 u_quad_t frev;
600
601 fhp = &nfh.fh_generic;
602 nfsm_srvmtofh(fhp);
603 if (v3) {
604 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
605 off = fxdr_hyper(tl);
606 } else {
607 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
608 off = (off_t)fxdr_unsigned(u_int32_t, *tl);
609 }
610 nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd));
611 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
612 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
613 if (error) {
614 nfsm_reply(2 * NFSX_UNSIGNED);
615 nfsm_srvpostop_attr(1, (struct vattr *)0);
616 return (0);
617 }
618 if (vp->v_type != VREG) {
619 if (v3)
620 error = EINVAL;
621 else
622 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
623 }
624 if (!error) {
625 nqsrv_getl(vp, ND_READ);
626 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0)
627 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1);
628 }
629 getret = VOP_GETATTR(vp, &va, cred, procp);
630 if (!error)
631 error = getret;
632 if (error) {
633 vput(vp);
634 nfsm_reply(NFSX_POSTOPATTR(v3));
635 nfsm_srvpostop_attr(getret, &va);
636 return (0);
637 }
638 if (off >= va.va_size)
639 cnt = 0;
640 else if ((off + reqlen) > va.va_size)
641 cnt = nfsm_rndup(va.va_size - off);
642 else
643 cnt = reqlen;
644 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
645 if (v3) {
646 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
647 *tl++ = nfs_true;
648 fp = (struct nfs_fattr *)tl;
649 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
650 } else {
651 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
652 fp = (struct nfs_fattr *)tl;
653 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
654 }
655 len = left = cnt;
656 if (cnt > 0) {
657 /*
658 * Generate the mbuf list with the uio_iov ref. to it.
659 */
660 i = 0;
661 m = m2 = mb;
662 while (left > 0) {
663 siz = min(M_TRAILINGSPACE(m), left);
664 if (siz > 0) {
665 left -= siz;
666 i++;
667 }
668 if (left > 0) {
669 MGET(m, M_WAIT, MT_DATA);
670 MCLGET(m, M_WAIT);
671 m->m_len = 0;
672 m2->m_next = m;
673 m2 = m;
674 }
675 }
676 MALLOC(iv, struct iovec *, i * sizeof (struct iovec),
677 M_TEMP, M_WAITOK);
678 uiop->uio_iov = iv2 = iv;
679 m = mb;
680 left = cnt;
681 i = 0;
682 while (left > 0) {
683 if (m == NULL)
684 panic("nfsrv_read iov");
685 siz = min(M_TRAILINGSPACE(m), left);
686 if (siz > 0) {
687 iv->iov_base = mtod(m, caddr_t) + m->m_len;
688 iv->iov_len = siz;
689 m->m_len += siz;
690 left -= siz;
691 iv++;
692 i++;
693 }
694 m = m->m_next;
695 }
696 uiop->uio_iovcnt = i;
697 uiop->uio_offset = off;
698 uiop->uio_resid = cnt;
699 uiop->uio_rw = UIO_READ;
700 uiop->uio_segflg = UIO_SYSSPACE;
701 error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
702 off = uiop->uio_offset;
703 FREE((caddr_t)iv2, M_TEMP);
704 if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){
705 if (!error)
706 error = getret;
707 m_freem(mreq);
708 vput(vp);
709 nfsm_reply(NFSX_POSTOPATTR(v3));
710 nfsm_srvpostop_attr(getret, &va);
711 return (0);
712 }
713 } else
714 uiop->uio_resid = 0;
715 vput(vp);
716 nfsm_srvfillattr(&va, fp);
717 len -= uiop->uio_resid;
718 tlen = nfsm_rndup(len);
719 if (cnt != tlen || tlen != len)
720 nfsm_adj(mb, cnt - tlen, tlen - len);
721 if (v3) {
722 *tl++ = txdr_unsigned(len);
723 if (len < reqlen)
724 *tl++ = nfs_true;
725 else
726 *tl++ = nfs_false;
727 }
728 *tl = txdr_unsigned(len);
729 nfsm_srvdone;
730 }
731
732 /*
733 * nfs write service
734 */
735 int
736 nfsrv_write(nfsd, slp, procp, mrq)
737 struct nfsrv_descript *nfsd;
738 struct nfssvc_sock *slp;
739 struct proc *procp;
740 struct mbuf **mrq;
741 {
742 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
743 struct mbuf *nam = nfsd->nd_nam;
744 caddr_t dpos = nfsd->nd_dpos;
745 struct ucred *cred = &nfsd->nd_cr;
746 register struct iovec *ivp;
747 register int i, cnt;
748 register struct mbuf *mp;
749 register struct nfs_fattr *fp;
750 struct iovec *iv;
751 struct vattr va, forat;
752 register u_int32_t *tl;
753 register int32_t t1;
754 caddr_t bpos;
755 int error = 0, rdonly, cache, len, forat_ret = 1;
756 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
757 int stable = NFSV3WRITE_FILESYNC;
758 int v3 = (nfsd->nd_flag & ND_NFSV3);
759 char *cp2;
760 struct mbuf *mb, *mb2, *mreq;
761 struct vnode *vp;
762 nfsfh_t nfh;
763 fhandle_t *fhp;
764 struct uio io, *uiop = &io;
765 off_t off;
766 u_quad_t frev;
767
768 if (mrep == NULL) {
769 *mrq = NULL;
770 return (0);
771 }
772 fhp = &nfh.fh_generic;
773 nfsm_srvmtofh(fhp);
774 if (v3) {
775 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
776 off = fxdr_hyper(tl);
777 tl += 3;
778 stable = fxdr_unsigned(int, *tl++);
779 } else {
780 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
781 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
782 tl += 2;
783 }
784 retlen = len = fxdr_unsigned(int32_t, *tl);
785 cnt = i = 0;
786
787 /*
788 * For NFS Version 2, it is not obvious what a write of zero length
789 * should do, but I might as well be consistent with Version 3,
790 * which is to return ok so long as there are no permission problems.
791 */
792 if (len > 0) {
793 zeroing = 1;
794 mp = mrep;
795 while (mp) {
796 if (mp == md) {
797 zeroing = 0;
798 adjust = dpos - mtod(mp, caddr_t);
799 mp->m_len -= adjust;
800 if (mp->m_len > 0 && adjust > 0)
801 NFSMADV(mp, adjust);
802 }
803 if (zeroing)
804 mp->m_len = 0;
805 else if (mp->m_len > 0) {
806 i += mp->m_len;
807 if (i > len) {
808 mp->m_len -= (i - len);
809 zeroing = 1;
810 }
811 if (mp->m_len > 0)
812 cnt++;
813 }
814 mp = mp->m_next;
815 }
816 }
817 if (len > NFS_MAXDATA || len < 0 || i < len) {
818 error = EIO;
819 nfsm_reply(2 * NFSX_UNSIGNED);
820 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
821 return (0);
822 }
823 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
824 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
825 if (error) {
826 nfsm_reply(2 * NFSX_UNSIGNED);
827 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
828 return (0);
829 }
830 if (v3)
831 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
832 if (vp->v_type != VREG) {
833 if (v3)
834 error = EINVAL;
835 else
836 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
837 }
838 if (!error) {
839 nqsrv_getl(vp, ND_WRITE);
840 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
841 }
842 if (error) {
843 vput(vp);
844 nfsm_reply(NFSX_WCCDATA(v3));
845 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
846 return (0);
847 }
848
849 if (len > 0) {
850 MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP,
851 M_WAITOK);
852 uiop->uio_iov = iv = ivp;
853 uiop->uio_iovcnt = cnt;
854 mp = mrep;
855 while (mp) {
856 if (mp->m_len > 0) {
857 ivp->iov_base = mtod(mp, caddr_t);
858 ivp->iov_len = mp->m_len;
859 ivp++;
860 }
861 mp = mp->m_next;
862 }
863
864 /*
865 * XXX
866 * The IO_METASYNC flag indicates that all metadata (and not just
867 * enough to ensure data integrity) mus be written to stable storage
868 * synchronously.
869 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
870 */
871 if (stable == NFSV3WRITE_UNSTABLE)
872 ioflags = IO_NODELOCKED;
873 else if (stable == NFSV3WRITE_DATASYNC)
874 ioflags = (IO_SYNC | IO_NODELOCKED);
875 else
876 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
877 uiop->uio_resid = len;
878 uiop->uio_rw = UIO_WRITE;
879 uiop->uio_segflg = UIO_SYSSPACE;
880 uiop->uio_procp = (struct proc *)0;
881 uiop->uio_offset = off;
882 error = VOP_WRITE(vp, uiop, ioflags, cred);
883 nfsstats.srvvop_writes++;
884 FREE((caddr_t)iv, M_TEMP);
885 }
886 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
887 vput(vp);
888 if (!error)
889 error = aftat_ret;
890 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
891 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
892 if (v3) {
893 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
894 if (error)
895 return (0);
896 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
897 *tl++ = txdr_unsigned(retlen);
898 if (stable == NFSV3WRITE_UNSTABLE)
899 *tl++ = txdr_unsigned(stable);
900 else
901 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
902 /*
903 * Actually, there is no need to txdr these fields,
904 * but it may make the values more human readable,
905 * for debugging purposes.
906 */
907 *tl++ = txdr_unsigned(boottime.tv_sec);
908 *tl = txdr_unsigned(boottime.tv_usec);
909 } else {
910 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
911 nfsm_srvfillattr(&va, fp);
912 }
913 nfsm_srvdone;
914 }
915
916 /*
917 * NFS write service with write gathering support. Called when
918 * nfsrvw_procrastinate > 0.
919 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
920 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
921 * Jan. 1994.
922 */
923 int
924 nfsrv_writegather(ndp, slp, procp, mrq)
925 struct nfsrv_descript **ndp;
926 struct nfssvc_sock *slp;
927 struct proc *procp;
928 struct mbuf **mrq;
929 {
930 register struct iovec *ivp;
931 register struct mbuf *mp;
932 register struct nfsrv_descript *wp, *nfsd, *owp, *swp;
933 register struct nfs_fattr *fp;
934 register int i = 0;
935 struct iovec *iov;
936 struct nfsrvw_delayhash *wpp;
937 struct ucred *cred;
938 struct vattr va, forat;
939 register u_int32_t *tl;
940 register int32_t t1;
941 caddr_t bpos, dpos;
942 int error = 0, rdonly, cache, len = 0, forat_ret = 1;
943 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
944 char *cp2;
945 struct mbuf *mb, *mb2, *mreq, *mrep, *md;
946 struct vnode *vp;
947 struct uio io, *uiop = &io;
948 u_quad_t frev, cur_usec;
949
950 *mrq = NULL;
951 if (*ndp) {
952 nfsd = *ndp;
953 *ndp = NULL;
954 mrep = nfsd->nd_mrep;
955 md = nfsd->nd_md;
956 dpos = nfsd->nd_dpos;
957 cred = &nfsd->nd_cr;
958 v3 = (nfsd->nd_flag & ND_NFSV3);
959 LIST_INIT(&nfsd->nd_coalesce);
960 nfsd->nd_mreq = NULL;
961 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
962 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
963 nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
964
965 /*
966 * Now, get the write header..
967 */
968 nfsm_srvmtofh(&nfsd->nd_fh);
969 if (v3) {
970 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
971 nfsd->nd_off = fxdr_hyper(tl);
972 tl += 3;
973 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
974 } else {
975 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
976 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
977 tl += 2;
978 }
979 len = fxdr_unsigned(int32_t, *tl);
980 nfsd->nd_len = len;
981 nfsd->nd_eoff = nfsd->nd_off + len;
982
983 /*
984 * Trim the header out of the mbuf list and trim off any trailing
985 * junk so that the mbuf list has only the write data.
986 */
987 zeroing = 1;
988 i = 0;
989 mp = mrep;
990 while (mp) {
991 if (mp == md) {
992 zeroing = 0;
993 adjust = dpos - mtod(mp, caddr_t);
994 mp->m_len -= adjust;
995 if (mp->m_len > 0 && adjust > 0)
996 NFSMADV(mp, adjust);
997 }
998 if (zeroing)
999 mp->m_len = 0;
1000 else {
1001 i += mp->m_len;
1002 if (i > len) {
1003 mp->m_len -= (i - len);
1004 zeroing = 1;
1005 }
1006 }
1007 mp = mp->m_next;
1008 }
1009 if (len > NFS_MAXDATA || len < 0 || i < len) {
1010 nfsmout:
1011 m_freem(mrep);
1012 error = EIO;
1013 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1014 if (v3)
1015 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1016 nfsd->nd_mreq = mreq;
1017 nfsd->nd_mrep = NULL;
1018 nfsd->nd_time = 0;
1019 }
1020
1021 /*
1022 * Add this entry to the hash and time queues.
1023 */
1024 s = splsoftclock();
1025 owp = NULL;
1026 wp = slp->ns_tq.lh_first;
1027 while (wp && wp->nd_time < nfsd->nd_time) {
1028 owp = wp;
1029 wp = wp->nd_tq.le_next;
1030 }
1031 if (owp) {
1032 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1033 } else {
1034 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1035 }
1036 if (nfsd->nd_mrep) {
1037 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1038 owp = NULL;
1039 wp = wpp->lh_first;
1040 while (wp &&
1041 memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1042 owp = wp;
1043 wp = wp->nd_hash.le_next;
1044 }
1045 while (wp && wp->nd_off < nfsd->nd_off &&
1046 !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1047 owp = wp;
1048 wp = wp->nd_hash.le_next;
1049 }
1050 if (owp) {
1051 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1052
1053 /*
1054 * Search the hash list for overlapping entries and
1055 * coalesce.
1056 */
1057 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1058 wp = nfsd->nd_hash.le_next;
1059 if (NFSW_SAMECRED(owp, nfsd))
1060 nfsrvw_coalesce(owp, nfsd);
1061 }
1062 } else {
1063 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1064 }
1065 }
1066 splx(s);
1067 }
1068
1069 /*
1070 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1071 * and generate the associated reply mbuf list(s).
1072 */
1073 loop1:
1074 cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1075 s = splsoftclock();
1076 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = owp) {
1077 owp = nfsd->nd_tq.le_next;
1078 if (nfsd->nd_time > cur_usec)
1079 break;
1080 if (nfsd->nd_mreq)
1081 continue;
1082 LIST_REMOVE(nfsd, nd_tq);
1083 LIST_REMOVE(nfsd, nd_hash);
1084 splx(s);
1085 mrep = nfsd->nd_mrep;
1086 nfsd->nd_mrep = NULL;
1087 cred = &nfsd->nd_cr;
1088 v3 = (nfsd->nd_flag & ND_NFSV3);
1089 forat_ret = aftat_ret = 1;
1090 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1091 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1092 FALSE);
1093 if (!error) {
1094 if (v3)
1095 forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
1096 if (vp->v_type != VREG) {
1097 if (v3)
1098 error = EINVAL;
1099 else
1100 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1101 }
1102 } else
1103 vp = NULL;
1104 if (!error) {
1105 nqsrv_getl(vp, ND_WRITE);
1106 error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
1107 }
1108
1109 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1110 ioflags = IO_NODELOCKED;
1111 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1112 ioflags = (IO_SYNC | IO_NODELOCKED);
1113 else
1114 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1115 uiop->uio_rw = UIO_WRITE;
1116 uiop->uio_segflg = UIO_SYSSPACE;
1117 uiop->uio_procp = (struct proc *)0;
1118 uiop->uio_offset = nfsd->nd_off;
1119 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1120 if (uiop->uio_resid > 0) {
1121 mp = mrep;
1122 i = 0;
1123 while (mp) {
1124 if (mp->m_len > 0)
1125 i++;
1126 mp = mp->m_next;
1127 }
1128 uiop->uio_iovcnt = i;
1129 MALLOC(iov, struct iovec *, i * sizeof (struct iovec),
1130 M_TEMP, M_WAITOK);
1131 uiop->uio_iov = ivp = iov;
1132 mp = mrep;
1133 while (mp) {
1134 if (mp->m_len > 0) {
1135 ivp->iov_base = mtod(mp, caddr_t);
1136 ivp->iov_len = mp->m_len;
1137 ivp++;
1138 }
1139 mp = mp->m_next;
1140 }
1141 if (!error) {
1142 error = VOP_WRITE(vp, uiop, ioflags, cred);
1143 nfsstats.srvvop_writes++;
1144 }
1145 FREE((caddr_t)iov, M_TEMP);
1146 }
1147 m_freem(mrep);
1148 if (vp) {
1149 aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
1150 vput(vp);
1151 }
1152
1153 /*
1154 * Loop around generating replies for all write rpcs that have
1155 * now been completed.
1156 */
1157 swp = nfsd;
1158 do {
1159 if (error) {
1160 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1161 if (v3) {
1162 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1163 }
1164 } else {
1165 nfsm_writereply(NFSX_PREOPATTR(v3) +
1166 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1167 NFSX_WRITEVERF(v3), v3);
1168 if (v3) {
1169 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1170 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1171 *tl++ = txdr_unsigned(nfsd->nd_len);
1172 *tl++ = txdr_unsigned(swp->nd_stable);
1173 /*
1174 * Actually, there is no need to txdr these fields,
1175 * but it may make the values more human readable,
1176 * for debugging purposes.
1177 */
1178 *tl++ = txdr_unsigned(boottime.tv_sec);
1179 *tl = txdr_unsigned(boottime.tv_usec);
1180 } else {
1181 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1182 nfsm_srvfillattr(&va, fp);
1183 }
1184 }
1185 nfsd->nd_mreq = mreq;
1186 if (nfsd->nd_mrep)
1187 panic("nfsrv_write: nd_mrep not free");
1188
1189 /*
1190 * Done. Put it at the head of the timer queue so that
1191 * the final phase can return the reply.
1192 */
1193 s = splsoftclock();
1194 if (nfsd != swp) {
1195 nfsd->nd_time = 0;
1196 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1197 }
1198 nfsd = swp->nd_coalesce.lh_first;
1199 if (nfsd) {
1200 LIST_REMOVE(nfsd, nd_tq);
1201 }
1202 splx(s);
1203 } while (nfsd);
1204 s = splsoftclock();
1205 swp->nd_time = 0;
1206 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1207 splx(s);
1208 goto loop1;
1209 }
1210 splx(s);
1211
1212 /*
1213 * Search for a reply to return.
1214 */
1215 s = splsoftclock();
1216 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = nfsd->nd_tq.le_next)
1217 if (nfsd->nd_mreq) {
1218 LIST_REMOVE(nfsd, nd_tq);
1219 *mrq = nfsd->nd_mreq;
1220 *ndp = nfsd;
1221 break;
1222 }
1223 splx(s);
1224 return (0);
1225 }
1226
1227 /*
1228 * Coalesce the write request nfsd into owp. To do this we must:
1229 * - remove nfsd from the queues
1230 * - merge nfsd->nd_mrep into owp->nd_mrep
1231 * - update the nd_eoff and nd_stable for owp
1232 * - put nfsd on owp's nd_coalesce list
1233 * NB: Must be called at splsoftclock().
1234 */
1235 void
1236 nfsrvw_coalesce(owp, nfsd)
1237 register struct nfsrv_descript *owp;
1238 register struct nfsrv_descript *nfsd;
1239 {
1240 register int overlap;
1241 register struct mbuf *mp;
1242
1243 LIST_REMOVE(nfsd, nd_hash);
1244 LIST_REMOVE(nfsd, nd_tq);
1245 if (owp->nd_eoff < nfsd->nd_eoff) {
1246 overlap = owp->nd_eoff - nfsd->nd_off;
1247 if (overlap < 0)
1248 panic("nfsrv_coalesce: bad off");
1249 if (overlap > 0)
1250 m_adj(nfsd->nd_mrep, overlap);
1251 mp = owp->nd_mrep;
1252 while (mp->m_next)
1253 mp = mp->m_next;
1254 mp->m_next = nfsd->nd_mrep;
1255 owp->nd_eoff = nfsd->nd_eoff;
1256 } else
1257 m_freem(nfsd->nd_mrep);
1258 nfsd->nd_mrep = NULL;
1259 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1260 owp->nd_stable = NFSV3WRITE_FILESYNC;
1261 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1262 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1263 owp->nd_stable = NFSV3WRITE_DATASYNC;
1264 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1265 /*
1266 * nfsd might hold coalesce elements! Move them to owp.
1267 * Otherwise, requests may be lost and clients will be stuck.
1268 */
1269 if (nfsd->nd_coalesce.lh_first)
1270 {
1271 register struct nfsrv_descript *m;
1272
1273 while ((m = nfsd->nd_coalesce.lh_first))
1274 {
1275 LIST_REMOVE(m, nd_tq);
1276 LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1277 }
1278 }
1279 }
1280
1281 /*
1282 * nfs create service
1283 * now does a truncate to 0 length via. setattr if it already exists
1284 */
1285 int
1286 nfsrv_create(nfsd, slp, procp, mrq)
1287 struct nfsrv_descript *nfsd;
1288 struct nfssvc_sock *slp;
1289 struct proc *procp;
1290 struct mbuf **mrq;
1291 {
1292 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1293 struct mbuf *nam = nfsd->nd_nam;
1294 caddr_t dpos = nfsd->nd_dpos;
1295 struct ucred *cred = &nfsd->nd_cr;
1296 register struct nfs_fattr *fp;
1297 struct vattr va, dirfor, diraft;
1298 register struct nfsv2_sattr *sp;
1299 register u_int32_t *tl;
1300 struct nameidata nd;
1301 register caddr_t cp;
1302 register int32_t t1;
1303 caddr_t bpos;
1304 int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1305 int rdev = 0;
1306 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1307 char *cp2;
1308 struct mbuf *mb, *mb2, *mreq;
1309 struct vnode *vp = NULL, *dirp = NULL;
1310 nfsfh_t nfh;
1311 fhandle_t *fhp;
1312 u_quad_t frev, tempsize;
1313 u_char cverf[NFSX_V3CREATEVERF];
1314
1315 nd.ni_cnd.cn_nameiop = 0;
1316 fhp = &nfh.fh_generic;
1317 nfsm_srvmtofh(fhp);
1318 nfsm_srvnamesiz(len);
1319 nd.ni_cnd.cn_cred = cred;
1320 nd.ni_cnd.cn_nameiop = CREATE;
1321 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART;
1322 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1323 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1324 if (dirp) {
1325 if (v3)
1326 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1327 procp);
1328 else {
1329 vrele(dirp);
1330 dirp = (struct vnode *)0;
1331 }
1332 }
1333 if (error) {
1334 nfsm_reply(NFSX_WCCDATA(v3));
1335 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1336 if (dirp)
1337 vrele(dirp);
1338 return (0);
1339 }
1340 VATTR_NULL(&va);
1341 if (v3) {
1342 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1343 how = fxdr_unsigned(int, *tl);
1344 switch (how) {
1345 case NFSV3CREATE_GUARDED:
1346 if (nd.ni_vp) {
1347 error = EEXIST;
1348 break;
1349 }
1350 case NFSV3CREATE_UNCHECKED:
1351 nfsm_srvsattr(&va);
1352 break;
1353 case NFSV3CREATE_EXCLUSIVE:
1354 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
1355 memcpy(cverf, cp, NFSX_V3CREATEVERF);
1356 exclusive_flag = 1;
1357 if (nd.ni_vp == NULL)
1358 va.va_mode = 0;
1359 break;
1360 };
1361 va.va_type = VREG;
1362 } else {
1363 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1364 va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1365 if (va.va_type == VNON)
1366 va.va_type = VREG;
1367 va.va_mode = nfstov_mode(sp->sa_mode);
1368 switch (va.va_type) {
1369 case VREG:
1370 tsize = fxdr_unsigned(int32_t, sp->sa_size);
1371 if (tsize != -1)
1372 va.va_size = (u_quad_t)tsize;
1373 break;
1374 case VCHR:
1375 case VBLK:
1376 case VFIFO:
1377 rdev = fxdr_unsigned(int32_t, sp->sa_size);
1378 break;
1379 default:
1380 break;
1381 };
1382 }
1383
1384 /*
1385 * Iff doesn't exist, create it
1386 * otherwise just truncate to 0 length
1387 * should I set the mode too ??
1388 */
1389 if (nd.ni_vp == NULL) {
1390 if (va.va_type == VREG || va.va_type == VSOCK) {
1391 vrele(nd.ni_startdir);
1392 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1393 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1394 if (!error) {
1395 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1396 if (exclusive_flag) {
1397 exclusive_flag = 0;
1398 VATTR_NULL(&va);
1399 memcpy((caddr_t)&va.va_atime, cverf,
1400 NFSX_V3CREATEVERF);
1401 error = VOP_SETATTR(nd.ni_vp, &va, cred,
1402 procp);
1403 }
1404 }
1405 } else if (va.va_type == VCHR || va.va_type == VBLK ||
1406 va.va_type == VFIFO) {
1407 if (va.va_type == VCHR && rdev == 0xffffffff)
1408 va.va_type = VFIFO;
1409 if (va.va_type != VFIFO &&
1410 (error = suser(cred, (u_short *)0))) {
1411 vrele(nd.ni_startdir);
1412 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1413 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1414 vput(nd.ni_dvp);
1415 nfsm_reply(0);
1416 return (error);
1417 } else
1418 va.va_rdev = (dev_t)rdev;
1419 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1420 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1421 &va);
1422 if (error) {
1423 vrele(nd.ni_startdir);
1424 nfsm_reply(0);
1425 }
1426 nd.ni_cnd.cn_nameiop = LOOKUP;
1427 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART);
1428 nd.ni_cnd.cn_proc = procp;
1429 nd.ni_cnd.cn_cred = cred;
1430 if ((error = lookup(&nd)) != 0) {
1431 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1432 nfsm_reply(0);
1433 }
1434 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1435 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1436 vrele(nd.ni_dvp);
1437 vput(nd.ni_vp);
1438 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1439 error = EINVAL;
1440 nfsm_reply(0);
1441 }
1442 } else {
1443 vrele(nd.ni_startdir);
1444 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1445 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1446 vput(nd.ni_dvp);
1447 error = ENXIO;
1448 }
1449 vp = nd.ni_vp;
1450 } else {
1451 vrele(nd.ni_startdir);
1452 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1453 vp = nd.ni_vp;
1454 if (nd.ni_dvp == vp)
1455 vrele(nd.ni_dvp);
1456 else
1457 vput(nd.ni_dvp);
1458 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1459 if (!error && va.va_size != -1) {
1460 error = nfsrv_access(vp, VWRITE, cred,
1461 (nd.ni_cnd.cn_flags & RDONLY), procp, 0);
1462 if (!error) {
1463 nqsrv_getl(vp, ND_WRITE);
1464 tempsize = va.va_size;
1465 VATTR_NULL(&va);
1466 va.va_size = tempsize;
1467 error = VOP_SETATTR(vp, &va, cred,
1468 procp);
1469 }
1470 }
1471 if (error)
1472 vput(vp);
1473 }
1474 if (!error) {
1475 memset((caddr_t)fhp, 0, sizeof(nfh));
1476 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1477 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1478 if (!error)
1479 error = VOP_GETATTR(vp, &va, cred, procp);
1480 vput(vp);
1481 }
1482 if (v3) {
1483 if (exclusive_flag && !error &&
1484 memcmp(cverf, (caddr_t)&va.va_atime, NFSX_V3CREATEVERF))
1485 error = EEXIST;
1486 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1487 vrele(dirp);
1488 }
1489 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1490 if (v3) {
1491 if (!error) {
1492 nfsm_srvpostop_fh(fhp);
1493 nfsm_srvpostop_attr(0, &va);
1494 }
1495 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1496 } else {
1497 nfsm_srvfhtom(fhp, v3);
1498 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1499 nfsm_srvfillattr(&va, fp);
1500 }
1501 return (0);
1502 nfsmout:
1503 if (dirp)
1504 vrele(dirp);
1505 if (nd.ni_cnd.cn_nameiop) {
1506 vrele(nd.ni_startdir);
1507 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1508 }
1509 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1510 if (nd.ni_dvp == nd.ni_vp)
1511 vrele(nd.ni_dvp);
1512 else
1513 vput(nd.ni_dvp);
1514 if (nd.ni_vp)
1515 vput(nd.ni_vp);
1516 return (error);
1517 }
1518
1519 /*
1520 * nfs v3 mknod service
1521 */
1522 int
1523 nfsrv_mknod(nfsd, slp, procp, mrq)
1524 struct nfsrv_descript *nfsd;
1525 struct nfssvc_sock *slp;
1526 struct proc *procp;
1527 struct mbuf **mrq;
1528 {
1529 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1530 struct mbuf *nam = nfsd->nd_nam;
1531 caddr_t dpos = nfsd->nd_dpos;
1532 struct ucred *cred = &nfsd->nd_cr;
1533 struct vattr va, dirfor, diraft;
1534 register u_int32_t *tl;
1535 struct nameidata nd;
1536 register int32_t t1;
1537 caddr_t bpos;
1538 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1539 u_int32_t major, minor;
1540 enum vtype vtyp;
1541 char *cp2;
1542 struct mbuf *mb, *mb2, *mreq;
1543 struct vnode *vp, *dirp = (struct vnode *)0;
1544 nfsfh_t nfh;
1545 fhandle_t *fhp;
1546 u_quad_t frev;
1547
1548 nd.ni_cnd.cn_nameiop = 0;
1549 fhp = &nfh.fh_generic;
1550 nfsm_srvmtofh(fhp);
1551 nfsm_srvnamesiz(len);
1552 nd.ni_cnd.cn_cred = cred;
1553 nd.ni_cnd.cn_nameiop = CREATE;
1554 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART;
1555 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1556 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1557 if (dirp)
1558 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp);
1559 if (error) {
1560 nfsm_reply(NFSX_WCCDATA(1));
1561 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1562 if (dirp)
1563 vrele(dirp);
1564 return (0);
1565 }
1566 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1567 vtyp = nfsv3tov_type(*tl);
1568 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1569 vrele(nd.ni_startdir);
1570 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1571 error = NFSERR_BADTYPE;
1572 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1573 if (nd.ni_dvp == nd.ni_vp)
1574 vrele(nd.ni_dvp);
1575 else
1576 vput(nd.ni_dvp);
1577 if (nd.ni_vp)
1578 vput(nd.ni_vp);
1579 goto out;
1580 }
1581 VATTR_NULL(&va);
1582 nfsm_srvsattr(&va);
1583 if (vtyp == VCHR || vtyp == VBLK) {
1584 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1585 major = fxdr_unsigned(u_int32_t, *tl++);
1586 minor = fxdr_unsigned(u_int32_t, *tl);
1587 va.va_rdev = makedev(major, minor);
1588 }
1589
1590 /*
1591 * Iff doesn't exist, create it.
1592 */
1593 if (nd.ni_vp) {
1594 vrele(nd.ni_startdir);
1595 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1596 error = EEXIST;
1597 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1598 if (nd.ni_dvp == nd.ni_vp)
1599 vrele(nd.ni_dvp);
1600 else
1601 vput(nd.ni_dvp);
1602 vput(nd.ni_vp);
1603 goto out;
1604 }
1605 va.va_type = vtyp;
1606 if (vtyp == VSOCK) {
1607 vrele(nd.ni_startdir);
1608 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1609 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1610 if (!error)
1611 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1612 } else {
1613 if (va.va_type != VFIFO &&
1614 (error = suser(cred, (u_short *)0))) {
1615 vrele(nd.ni_startdir);
1616 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1617 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1618 vput(nd.ni_dvp);
1619 goto out;
1620 }
1621 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1622 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1623 if (error) {
1624 vrele(nd.ni_startdir);
1625 goto out;
1626 }
1627 nd.ni_cnd.cn_nameiop = LOOKUP;
1628 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART);
1629 nd.ni_cnd.cn_proc = procp;
1630 nd.ni_cnd.cn_cred = procp->p_ucred;
1631 error = lookup(&nd);
1632 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
1633 if (error)
1634 goto out;
1635 if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1636 vrele(nd.ni_dvp);
1637 vput(nd.ni_vp);
1638 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1639 error = EINVAL;
1640 }
1641 }
1642 out:
1643 vp = nd.ni_vp;
1644 if (!error) {
1645 memset((caddr_t)fhp, 0, sizeof(nfh));
1646 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1647 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1648 if (!error)
1649 error = VOP_GETATTR(vp, &va, cred, procp);
1650 vput(vp);
1651 }
1652 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1653 vrele(dirp);
1654 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
1655 if (!error) {
1656 nfsm_srvpostop_fh(fhp);
1657 nfsm_srvpostop_attr(0, &va);
1658 }
1659 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1660 return (0);
1661 nfsmout:
1662 if (dirp)
1663 vrele(dirp);
1664 if (nd.ni_cnd.cn_nameiop) {
1665 vrele(nd.ni_startdir);
1666 free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI);
1667 }
1668 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1669 if (nd.ni_dvp == nd.ni_vp)
1670 vrele(nd.ni_dvp);
1671 else
1672 vput(nd.ni_dvp);
1673 if (nd.ni_vp)
1674 vput(nd.ni_vp);
1675 return (error);
1676 }
1677
1678 /*
1679 * nfs remove service
1680 */
1681 int
1682 nfsrv_remove(nfsd, slp, procp, mrq)
1683 struct nfsrv_descript *nfsd;
1684 struct nfssvc_sock *slp;
1685 struct proc *procp;
1686 struct mbuf **mrq;
1687 {
1688 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1689 struct mbuf *nam = nfsd->nd_nam;
1690 caddr_t dpos = nfsd->nd_dpos;
1691 struct ucred *cred = &nfsd->nd_cr;
1692 struct nameidata nd;
1693 register u_int32_t *tl;
1694 register int32_t t1;
1695 caddr_t bpos;
1696 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1697 int v3 = (nfsd->nd_flag & ND_NFSV3);
1698 char *cp2;
1699 struct mbuf *mb, *mreq;
1700 struct vnode *vp, *dirp;
1701 struct vattr dirfor, diraft;
1702 nfsfh_t nfh;
1703 fhandle_t *fhp;
1704 u_quad_t frev;
1705
1706 #ifndef nolint
1707 vp = (struct vnode *)0;
1708 #endif
1709 fhp = &nfh.fh_generic;
1710 nfsm_srvmtofh(fhp);
1711 nfsm_srvnamesiz(len);
1712 nd.ni_cnd.cn_cred = cred;
1713 nd.ni_cnd.cn_nameiop = DELETE;
1714 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1715 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1716 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1717 if (dirp) {
1718 if (v3)
1719 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1720 procp);
1721 else
1722 vrele(dirp);
1723 }
1724 if (!error) {
1725 vp = nd.ni_vp;
1726 if (vp->v_type == VDIR &&
1727 (error = suser(cred, (u_short *)0)) != 0)
1728 goto out;
1729 /*
1730 * The root of a mounted filesystem cannot be deleted.
1731 */
1732 if (vp->v_flag & VROOT) {
1733 error = EBUSY;
1734 goto out;
1735 }
1736 out:
1737 if (!error) {
1738 (void)uvm_vnp_uncache(vp);
1739 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1740 nqsrv_getl(vp, ND_WRITE);
1741 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1742 } else {
1743 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1744 if (nd.ni_dvp == vp)
1745 vrele(nd.ni_dvp);
1746 else
1747 vput(nd.ni_dvp);
1748 vput(vp);
1749 }
1750 }
1751 if (dirp && v3) {
1752 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1753 vrele(dirp);
1754 }
1755 nfsm_reply(NFSX_WCCDATA(v3));
1756 if (v3) {
1757 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1758 return (0);
1759 }
1760 nfsm_srvdone;
1761 }
1762
1763 /*
1764 * nfs rename service
1765 */
1766 int
1767 nfsrv_rename(nfsd, slp, procp, mrq)
1768 struct nfsrv_descript *nfsd;
1769 struct nfssvc_sock *slp;
1770 struct proc *procp;
1771 struct mbuf **mrq;
1772 {
1773 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1774 struct mbuf *nam = nfsd->nd_nam;
1775 caddr_t dpos = nfsd->nd_dpos;
1776 struct ucred *cred = &nfsd->nd_cr;
1777 register u_int32_t *tl;
1778 register int32_t t1;
1779 caddr_t bpos;
1780 int error = 0, cache, len, len2, fdirfor_ret = 1, fdiraft_ret = 1;
1781 int tdirfor_ret = 1, tdiraft_ret = 1;
1782 int v3 = (nfsd->nd_flag & ND_NFSV3);
1783 char *cp2;
1784 struct mbuf *mb, *mreq;
1785 struct nameidata fromnd, tond;
1786 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
1787 struct vnode *tdirp = (struct vnode *)0;
1788 struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1789 nfsfh_t fnfh, tnfh;
1790 fhandle_t *ffhp, *tfhp;
1791 u_quad_t frev;
1792 uid_t saved_uid;
1793
1794 #ifndef nolint
1795 fvp = (struct vnode *)0;
1796 #endif
1797 ffhp = &fnfh.fh_generic;
1798 tfhp = &tnfh.fh_generic;
1799 fromnd.ni_cnd.cn_nameiop = 0;
1800 tond.ni_cnd.cn_nameiop = 0;
1801 nfsm_srvmtofh(ffhp);
1802 nfsm_srvnamesiz(len);
1803 /*
1804 * Remember our original uid so that we can reset cr_uid before
1805 * the second nfs_namei() call, in case it is remapped.
1806 */
1807 saved_uid = cred->cr_uid;
1808 fromnd.ni_cnd.cn_cred = cred;
1809 fromnd.ni_cnd.cn_nameiop = DELETE;
1810 fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART;
1811 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
1812 &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1813 if (fdirp) {
1814 if (v3)
1815 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred,
1816 procp);
1817 else {
1818 vrele(fdirp);
1819 fdirp = (struct vnode *)0;
1820 }
1821 }
1822 if (error) {
1823 nfsm_reply(2 * NFSX_WCCDATA(v3));
1824 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1825 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1826 if (fdirp)
1827 vrele(fdirp);
1828 return (0);
1829 }
1830 fvp = fromnd.ni_vp;
1831 nfsm_srvmtofh(tfhp);
1832 nfsm_strsiz(len2, NFS_MAXNAMLEN);
1833 cred->cr_uid = saved_uid;
1834 tond.ni_cnd.cn_cred = cred;
1835 tond.ni_cnd.cn_nameiop = RENAME;
1836 tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
1837 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
1838 &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1839 if (tdirp) {
1840 if (v3)
1841 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred,
1842 procp);
1843 else {
1844 vrele(tdirp);
1845 tdirp = (struct vnode *)0;
1846 }
1847 }
1848 if (error) {
1849 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1850 vrele(fromnd.ni_dvp);
1851 vrele(fvp);
1852 goto out1;
1853 }
1854 tdvp = tond.ni_dvp;
1855 tvp = tond.ni_vp;
1856 if (tvp != NULL) {
1857 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1858 if (v3)
1859 error = EEXIST;
1860 else
1861 error = EISDIR;
1862 goto out;
1863 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1864 if (v3)
1865 error = EEXIST;
1866 else
1867 error = ENOTDIR;
1868 goto out;
1869 }
1870 if (tvp->v_type == VDIR && tvp->v_mountedhere) {
1871 if (v3)
1872 error = EXDEV;
1873 else
1874 error = ENOTEMPTY;
1875 goto out;
1876 }
1877 }
1878 if (fvp->v_type == VDIR && fvp->v_mountedhere) {
1879 if (v3)
1880 error = EXDEV;
1881 else
1882 error = ENOTEMPTY;
1883 goto out;
1884 }
1885 if (fvp->v_mount != tdvp->v_mount) {
1886 if (v3)
1887 error = EXDEV;
1888 else
1889 error = ENOTEMPTY;
1890 goto out;
1891 }
1892 if (fvp == tdvp) {
1893 if (v3)
1894 error = EINVAL;
1895 else
1896 error = ENOTEMPTY;
1897 }
1898 /*
1899 * If source is the same as the destination (that is the
1900 * same vnode with the same name in the same directory),
1901 * then there is nothing to do.
1902 */
1903 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
1904 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1905 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
1906 fromnd.ni_cnd.cn_namelen))
1907 error = -1;
1908 out:
1909 if (!error) {
1910 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
1911 nqsrv_getl(tdvp, ND_WRITE);
1912 if (tvp) {
1913 (void)uvm_vnp_uncache(tvp);
1914 nqsrv_getl(tvp, ND_WRITE);
1915 }
1916 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
1917 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
1918 } else {
1919 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
1920 if (tdvp == tvp)
1921 vrele(tdvp);
1922 else
1923 vput(tdvp);
1924 if (tvp)
1925 vput(tvp);
1926 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1927 vrele(fromnd.ni_dvp);
1928 vrele(fvp);
1929 if (error == -1)
1930 error = 0;
1931 }
1932 vrele(tond.ni_startdir);
1933 FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
1934 out1:
1935 if (fdirp) {
1936 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp);
1937 vrele(fdirp);
1938 }
1939 if (tdirp) {
1940 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp);
1941 vrele(tdirp);
1942 }
1943 vrele(fromnd.ni_startdir);
1944 FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
1945 nfsm_reply(2 * NFSX_WCCDATA(v3));
1946 if (v3) {
1947 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1948 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1949 }
1950 return (0);
1951
1952 nfsmout:
1953 if (fdirp)
1954 vrele(fdirp);
1955 if (tdirp)
1956 vrele(tdirp);
1957 if (tond.ni_cnd.cn_nameiop) {
1958 vrele(tond.ni_startdir);
1959 FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI);
1960 }
1961 if (fromnd.ni_cnd.cn_nameiop) {
1962 vrele(fromnd.ni_startdir);
1963 FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI);
1964 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1965 vrele(fromnd.ni_dvp);
1966 vrele(fvp);
1967 }
1968 return (error);
1969 }
1970
1971 /*
1972 * nfs link service
1973 */
1974 int
1975 nfsrv_link(nfsd, slp, procp, mrq)
1976 struct nfsrv_descript *nfsd;
1977 struct nfssvc_sock *slp;
1978 struct proc *procp;
1979 struct mbuf **mrq;
1980 {
1981 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1982 struct mbuf *nam = nfsd->nd_nam;
1983 caddr_t dpos = nfsd->nd_dpos;
1984 struct ucred *cred = &nfsd->nd_cr;
1985 struct nameidata nd;
1986 register u_int32_t *tl;
1987 register int32_t t1;
1988 caddr_t bpos;
1989 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1;
1990 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
1991 char *cp2;
1992 struct mbuf *mb, *mreq;
1993 struct vnode *vp, *xp, *dirp = (struct vnode *)0;
1994 struct vattr dirfor, diraft, at;
1995 nfsfh_t nfh, dnfh;
1996 fhandle_t *fhp, *dfhp;
1997 u_quad_t frev;
1998
1999 fhp = &nfh.fh_generic;
2000 dfhp = &dnfh.fh_generic;
2001 nfsm_srvmtofh(fhp);
2002 nfsm_srvmtofh(dfhp);
2003 nfsm_srvnamesiz(len);
2004 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
2005 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2006 if (error) {
2007 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2008 nfsm_srvpostop_attr(getret, &at);
2009 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2010 return (0);
2011 }
2012 if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0)
2013 goto out1;
2014 nd.ni_cnd.cn_cred = cred;
2015 nd.ni_cnd.cn_nameiop = CREATE;
2016 nd.ni_cnd.cn_flags = LOCKPARENT;
2017 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
2018 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2019 if (dirp) {
2020 if (v3)
2021 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2022 procp);
2023 else {
2024 vrele(dirp);
2025 dirp = (struct vnode *)0;
2026 }
2027 }
2028 if (error)
2029 goto out1;
2030 xp = nd.ni_vp;
2031 if (xp != NULL) {
2032 error = EEXIST;
2033 goto out;
2034 }
2035 xp = nd.ni_dvp;
2036 if (vp->v_mount != xp->v_mount)
2037 error = EXDEV;
2038 out:
2039 if (!error) {
2040 nqsrv_getl(vp, ND_WRITE);
2041 nqsrv_getl(xp, ND_WRITE);
2042 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2043 } else {
2044 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2045 if (nd.ni_dvp == nd.ni_vp)
2046 vrele(nd.ni_dvp);
2047 else
2048 vput(nd.ni_dvp);
2049 if (nd.ni_vp)
2050 vrele(nd.ni_vp);
2051 }
2052 out1:
2053 if (v3)
2054 getret = VOP_GETATTR(vp, &at, cred, procp);
2055 if (dirp) {
2056 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2057 vrele(dirp);
2058 }
2059 vrele(vp);
2060 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2061 if (v3) {
2062 nfsm_srvpostop_attr(getret, &at);
2063 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2064 return (0);
2065 }
2066 nfsm_srvdone;
2067 }
2068
2069 /*
2070 * nfs symbolic link service
2071 */
2072 int
2073 nfsrv_symlink(nfsd, slp, procp, mrq)
2074 struct nfsrv_descript *nfsd;
2075 struct nfssvc_sock *slp;
2076 struct proc *procp;
2077 struct mbuf **mrq;
2078 {
2079 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2080 struct mbuf *nam = nfsd->nd_nam;
2081 caddr_t dpos = nfsd->nd_dpos;
2082 struct ucred *cred = &nfsd->nd_cr;
2083 struct vattr va, dirfor, diraft;
2084 struct nameidata nd;
2085 register u_int32_t *tl;
2086 register int32_t t1;
2087 struct nfsv2_sattr *sp;
2088 char *bpos, *pathcp = NULL, *cp2;
2089 struct uio io;
2090 struct iovec iv;
2091 int error = 0, cache, len, len2, dirfor_ret = 1, diraft_ret = 1;
2092 int v3 = (nfsd->nd_flag & ND_NFSV3);
2093 struct mbuf *mb, *mreq, *mb2;
2094 struct vnode *dirp = (struct vnode *)0;
2095 nfsfh_t nfh;
2096 fhandle_t *fhp;
2097 u_quad_t frev;
2098
2099 nd.ni_cnd.cn_nameiop = 0;
2100 fhp = &nfh.fh_generic;
2101 nfsm_srvmtofh(fhp);
2102 nfsm_srvnamesiz(len);
2103 nd.ni_cnd.cn_cred = cred;
2104 nd.ni_cnd.cn_nameiop = CREATE;
2105 nd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART;
2106 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2107 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2108 if (dirp) {
2109 if (v3)
2110 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2111 procp);
2112 else {
2113 vrele(dirp);
2114 dirp = (struct vnode *)0;
2115 }
2116 }
2117 if (error)
2118 goto out;
2119 VATTR_NULL(&va);
2120 if (v3)
2121 nfsm_srvsattr(&va);
2122 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2123 MALLOC(pathcp, caddr_t, len2 + 1, M_TEMP, M_WAITOK);
2124 iv.iov_base = pathcp;
2125 iv.iov_len = len2;
2126 io.uio_resid = len2;
2127 io.uio_offset = 0;
2128 io.uio_iov = &iv;
2129 io.uio_iovcnt = 1;
2130 io.uio_segflg = UIO_SYSSPACE;
2131 io.uio_rw = UIO_READ;
2132 io.uio_procp = (struct proc *)0;
2133 nfsm_mtouio(&io, len2);
2134 if (!v3) {
2135 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2136 va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2137 }
2138 *(pathcp + len2) = '\0';
2139 if (nd.ni_vp) {
2140 vrele(nd.ni_startdir);
2141 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2142 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2143 if (nd.ni_dvp == nd.ni_vp)
2144 vrele(nd.ni_dvp);
2145 else
2146 vput(nd.ni_dvp);
2147 vrele(nd.ni_vp);
2148 error = EEXIST;
2149 goto out;
2150 }
2151 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2152 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2153 if (error)
2154 vrele(nd.ni_startdir);
2155 else {
2156 if (v3) {
2157 nd.ni_cnd.cn_nameiop = LOOKUP;
2158 nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART | FOLLOW);
2159 nd.ni_cnd.cn_flags |= (NOFOLLOW | LOCKLEAF);
2160 nd.ni_cnd.cn_proc = procp;
2161 nd.ni_cnd.cn_cred = cred;
2162 error = lookup(&nd);
2163 if (!error) {
2164 memset((caddr_t)fhp, 0, sizeof(nfh));
2165 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
2166 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2167 if (!error)
2168 error = VOP_GETATTR(nd.ni_vp, &va, cred,
2169 procp);
2170 vput(nd.ni_vp);
2171 }
2172 } else
2173 vrele(nd.ni_startdir);
2174 FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2175 }
2176 out:
2177 if (pathcp)
2178 FREE(pathcp, M_TEMP);
2179 if (dirp) {
2180 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2181 vrele(dirp);
2182 }
2183 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2184 if (v3) {
2185 if (!error) {
2186 nfsm_srvpostop_fh(fhp);
2187 nfsm_srvpostop_attr(0, &va);
2188 }
2189 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2190 }
2191 return (0);
2192 nfsmout:
2193 if (nd.ni_cnd.cn_nameiop) {
2194 vrele(nd.ni_startdir);
2195 free(nd.ni_cnd.cn_pnbuf, M_NAMEI);
2196 }
2197 if (dirp)
2198 vrele(dirp);
2199 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2200 if (nd.ni_dvp == nd.ni_vp)
2201 vrele(nd.ni_dvp);
2202 else
2203 vput(nd.ni_dvp);
2204 if (nd.ni_vp)
2205 vrele(nd.ni_vp);
2206 if (pathcp)
2207 FREE(pathcp, M_TEMP);
2208 return (error);
2209 }
2210
2211 /*
2212 * nfs mkdir service
2213 */
2214 int
2215 nfsrv_mkdir(nfsd, slp, procp, mrq)
2216 struct nfsrv_descript *nfsd;
2217 struct nfssvc_sock *slp;
2218 struct proc *procp;
2219 struct mbuf **mrq;
2220 {
2221 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2222 struct mbuf *nam = nfsd->nd_nam;
2223 caddr_t dpos = nfsd->nd_dpos;
2224 struct ucred *cred = &nfsd->nd_cr;
2225 struct vattr va, dirfor, diraft;
2226 register struct nfs_fattr *fp;
2227 struct nameidata nd;
2228 register caddr_t cp;
2229 register u_int32_t *tl;
2230 register int32_t t1;
2231 caddr_t bpos;
2232 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2233 int v3 = (nfsd->nd_flag & ND_NFSV3);
2234 char *cp2;
2235 struct mbuf *mb, *mb2, *mreq;
2236 struct vnode *vp, *dirp = (struct vnode *)0;
2237 nfsfh_t nfh;
2238 fhandle_t *fhp;
2239 u_quad_t frev;
2240
2241 fhp = &nfh.fh_generic;
2242 nfsm_srvmtofh(fhp);
2243 nfsm_srvnamesiz(len);
2244 nd.ni_cnd.cn_cred = cred;
2245 nd.ni_cnd.cn_nameiop = CREATE;
2246 nd.ni_cnd.cn_flags = LOCKPARENT;
2247 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2248 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2249 if (dirp) {
2250 if (v3)
2251 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2252 procp);
2253 else {
2254 vrele(dirp);
2255 dirp = (struct vnode *)0;
2256 }
2257 }
2258 if (error) {
2259 nfsm_reply(NFSX_WCCDATA(v3));
2260 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2261 if (dirp)
2262 vrele(dirp);
2263 return (0);
2264 }
2265 VATTR_NULL(&va);
2266 if (v3) {
2267 nfsm_srvsattr(&va);
2268 } else {
2269 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2270 va.va_mode = nfstov_mode(*tl++);
2271 }
2272 va.va_type = VDIR;
2273 vp = nd.ni_vp;
2274 if (vp != NULL) {
2275 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2276 if (nd.ni_dvp == vp)
2277 vrele(nd.ni_dvp);
2278 else
2279 vput(nd.ni_dvp);
2280 vrele(vp);
2281 error = EEXIST;
2282 goto out;
2283 }
2284 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2285 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2286 if (!error) {
2287 vp = nd.ni_vp;
2288 memset((caddr_t)fhp, 0, sizeof(nfh));
2289 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
2290 error = VFS_VPTOFH(vp, &fhp->fh_fid);
2291 if (!error)
2292 error = VOP_GETATTR(vp, &va, cred, procp);
2293 vput(vp);
2294 }
2295 out:
2296 if (dirp) {
2297 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2298 vrele(dirp);
2299 }
2300 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2301 if (v3) {
2302 if (!error) {
2303 nfsm_srvpostop_fh(fhp);
2304 nfsm_srvpostop_attr(0, &va);
2305 }
2306 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2307 } else {
2308 nfsm_srvfhtom(fhp, v3);
2309 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2310 nfsm_srvfillattr(&va, fp);
2311 }
2312 return (0);
2313 nfsmout:
2314 if (dirp)
2315 vrele(dirp);
2316 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2317 if (nd.ni_dvp == nd.ni_vp)
2318 vrele(nd.ni_dvp);
2319 else
2320 vput(nd.ni_dvp);
2321 if (nd.ni_vp)
2322 vrele(nd.ni_vp);
2323 return (error);
2324 }
2325
2326 /*
2327 * nfs rmdir service
2328 */
2329 int
2330 nfsrv_rmdir(nfsd, slp, procp, mrq)
2331 struct nfsrv_descript *nfsd;
2332 struct nfssvc_sock *slp;
2333 struct proc *procp;
2334 struct mbuf **mrq;
2335 {
2336 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2337 struct mbuf *nam = nfsd->nd_nam;
2338 caddr_t dpos = nfsd->nd_dpos;
2339 struct ucred *cred = &nfsd->nd_cr;
2340 register u_int32_t *tl;
2341 register int32_t t1;
2342 caddr_t bpos;
2343 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2344 int v3 = (nfsd->nd_flag & ND_NFSV3);
2345 char *cp2;
2346 struct mbuf *mb, *mreq;
2347 struct vnode *vp, *dirp = (struct vnode *)0;
2348 struct vattr dirfor, diraft;
2349 nfsfh_t nfh;
2350 fhandle_t *fhp;
2351 struct nameidata nd;
2352 u_quad_t frev;
2353
2354 fhp = &nfh.fh_generic;
2355 nfsm_srvmtofh(fhp);
2356 nfsm_srvnamesiz(len);
2357 nd.ni_cnd.cn_cred = cred;
2358 nd.ni_cnd.cn_nameiop = DELETE;
2359 nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2360 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2361 &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2362 if (dirp) {
2363 if (v3)
2364 dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2365 procp);
2366 else {
2367 vrele(dirp);
2368 dirp = (struct vnode *)0;
2369 }
2370 }
2371 if (error) {
2372 nfsm_reply(NFSX_WCCDATA(v3));
2373 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2374 if (dirp)
2375 vrele(dirp);
2376 return (0);
2377 }
2378 vp = nd.ni_vp;
2379 if (vp->v_type != VDIR) {
2380 error = ENOTDIR;
2381 goto out;
2382 }
2383 /*
2384 * No rmdir "." please.
2385 */
2386 if (nd.ni_dvp == vp) {
2387 error = EINVAL;
2388 goto out;
2389 }
2390 /*
2391 * The root of a mounted filesystem cannot be deleted.
2392 */
2393 if (vp->v_flag & VROOT)
2394 error = EBUSY;
2395 out:
2396 if (!error) {
2397 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2398 nqsrv_getl(vp, ND_WRITE);
2399 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2400 } else {
2401 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2402 if (nd.ni_dvp == nd.ni_vp)
2403 vrele(nd.ni_dvp);
2404 else
2405 vput(nd.ni_dvp);
2406 vput(vp);
2407 }
2408 if (dirp) {
2409 diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2410 vrele(dirp);
2411 }
2412 nfsm_reply(NFSX_WCCDATA(v3));
2413 if (v3) {
2414 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2415 return (0);
2416 }
2417 nfsm_srvdone;
2418 }
2419
2420 /*
2421 * nfs readdir service
2422 * - mallocs what it thinks is enough to read
2423 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR
2424 * - calls VOP_READDIR()
2425 * - loops around building the reply
2426 * if the output generated exceeds count break out of loop
2427 * The nfsm_clget macro is used here so that the reply will be packed
2428 * tightly in mbuf clusters.
2429 * - it only knows that it has encountered eof when the VOP_READDIR()
2430 * reads nothing
2431 * - as such one readdir rpc will return eof false although you are there
2432 * and then the next will return eof
2433 * - it trims out records with d_fileno == 0
2434 * this doesn't matter for Unix clients, but they might confuse clients
2435 * for other os'.
2436 * - it trims out records with d_type == DT_WHT
2437 * these cannot be seen through NFS (unless we extend the protocol)
2438 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2439 * than requested, but this may not apply to all filesystems. For
2440 * example, client NFS does not { although it is never remote mounted
2441 * anyhow }
2442 * The alternate call nfsrv_readdirplus() does lookups as well.
2443 * PS: The NFS protocol spec. does not clarify what the "count" byte
2444 * argument is a count of.. just name strings and file id's or the
2445 * entire reply rpc or ...
2446 * I tried just file name and id sizes and it confused the Sun client,
2447 * so I am using the full rpc size now. The "paranoia.." comment refers
2448 * to including the status longwords that are not a part of the dir.
2449 * "entry" structures, but are in the rpc.
2450 */
2451 struct flrep {
2452 nfsuint64 fl_off;
2453 u_int32_t fl_postopok;
2454 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2455 u_int32_t fl_fhok;
2456 u_int32_t fl_fhsize;
2457 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2458 };
2459
2460 int
2461 nfsrv_readdir(nfsd, slp, procp, mrq)
2462 struct nfsrv_descript *nfsd;
2463 struct nfssvc_sock *slp;
2464 struct proc *procp;
2465 struct mbuf **mrq;
2466 {
2467 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2468 struct mbuf *nam = nfsd->nd_nam;
2469 caddr_t dpos = nfsd->nd_dpos;
2470 struct ucred *cred = &nfsd->nd_cr;
2471 register char *bp, *be;
2472 register struct mbuf *mp;
2473 register struct dirent *dp;
2474 register caddr_t cp;
2475 register u_int32_t *tl;
2476 register int32_t t1;
2477 caddr_t bpos;
2478 struct mbuf *mb, *mb2, *mreq, *mp2;
2479 char *cpos, *cend, *cp2, *rbuf;
2480 struct vnode *vp;
2481 struct vattr at;
2482 nfsfh_t nfh;
2483 fhandle_t *fhp;
2484 struct uio io;
2485 struct iovec iv;
2486 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2487 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies;
2488 int v3 = (nfsd->nd_flag & ND_NFSV3);
2489 u_quad_t frev, off, toff, verf;
2490 off_t *cookies = NULL, *cookiep;
2491 nfsuint64 jar;
2492
2493 fhp = &nfh.fh_generic;
2494 nfsm_srvmtofh(fhp);
2495 if (v3) {
2496 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2497 toff = fxdr_hyper(tl);
2498 tl += 2;
2499 verf = fxdr_hyper(tl);
2500 tl += 2;
2501 } else {
2502 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2503 toff = fxdr_unsigned(u_quad_t, *tl++);
2504 }
2505 off = toff;
2506 cnt = fxdr_unsigned(int, *tl);
2507 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2508 xfer = NFS_SRVMAXDATA(nfsd);
2509 if (siz > xfer)
2510 siz = xfer;
2511 fullsiz = siz;
2512 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2513 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2514 if (!error && vp->v_type != VDIR) {
2515 error = ENOTDIR;
2516 vput(vp);
2517 }
2518 if (error) {
2519 nfsm_reply(NFSX_UNSIGNED);
2520 nfsm_srvpostop_attr(getret, &at);
2521 return (0);
2522 }
2523 nqsrv_getl(vp, ND_READ);
2524 if (v3) {
2525 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2526 #ifdef NFS3_STRICTVERF
2527 /*
2528 * XXX This check is too strict for Solaris 2.5 clients.
2529 */
2530 if (!error && toff && verf != at.va_filerev)
2531 error = NFSERR_BAD_COOKIE;
2532 #endif
2533 }
2534 if (!error)
2535 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2536 if (error) {
2537 vput(vp);
2538 nfsm_reply(NFSX_POSTOPATTR(v3));
2539 nfsm_srvpostop_attr(getret, &at);
2540 return (0);
2541 }
2542 VOP_UNLOCK(vp, 0);
2543 MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
2544 again:
2545 iv.iov_base = rbuf;
2546 iv.iov_len = fullsiz;
2547 io.uio_iov = &iv;
2548 io.uio_iovcnt = 1;
2549 io.uio_offset = (off_t)off;
2550 io.uio_resid = fullsiz;
2551 io.uio_segflg = UIO_SYSSPACE;
2552 io.uio_rw = UIO_READ;
2553 io.uio_procp = (struct proc *)0;
2554 eofflag = 0;
2555 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2556
2557 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2558
2559 off = (off_t)io.uio_offset;
2560 if (!cookies && !error)
2561 error = NFSERR_PERM;
2562 if (v3) {
2563 getret = VOP_GETATTR(vp, &at, cred, procp);
2564 if (!error)
2565 error = getret;
2566 }
2567
2568 VOP_UNLOCK(vp, 0);
2569 if (error) {
2570 vrele(vp);
2571 FREE((caddr_t)rbuf, M_TEMP);
2572 if (cookies)
2573 free((caddr_t)cookies, M_TEMP);
2574 nfsm_reply(NFSX_POSTOPATTR(v3));
2575 nfsm_srvpostop_attr(getret, &at);
2576 return (0);
2577 }
2578 if (io.uio_resid) {
2579 siz -= io.uio_resid;
2580
2581 /*
2582 * If nothing read, return eof
2583 * rpc reply
2584 */
2585 if (siz == 0) {
2586 vrele(vp);
2587 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2588 2 * NFSX_UNSIGNED);
2589 if (v3) {
2590 nfsm_srvpostop_attr(getret, &at);
2591 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2592 txdr_hyper(at.va_filerev, tl);
2593 tl += 2;
2594 } else
2595 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2596 *tl++ = nfs_false;
2597 *tl = nfs_true;
2598 FREE((caddr_t)rbuf, M_TEMP);
2599 FREE((caddr_t)cookies, M_TEMP);
2600 return (0);
2601 }
2602 }
2603
2604 /*
2605 * Check for degenerate cases of nothing useful read.
2606 * If so go try again
2607 */
2608 cpos = rbuf;
2609 cend = rbuf + siz;
2610 dp = (struct dirent *)cpos;
2611 cookiep = cookies;
2612
2613 while (cpos < cend && ncookies > 0 &&
2614 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2615 cpos += dp->d_reclen;
2616 dp = (struct dirent *)cpos;
2617 cookiep++;
2618 ncookies--;
2619 }
2620 if (cpos >= cend || ncookies == 0) {
2621 toff = off;
2622 siz = fullsiz;
2623 goto again;
2624 }
2625
2626 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
2627 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2628 if (v3) {
2629 nfsm_srvpostop_attr(getret, &at);
2630 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2631 txdr_hyper(at.va_filerev, tl);
2632 }
2633 mp = mp2 = mb;
2634 bp = bpos;
2635 be = bp + M_TRAILINGSPACE(mp);
2636
2637 /* Loop through the records and build reply */
2638 while (cpos < cend && ncookies > 0) {
2639 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2640 nlen = dp->d_namlen;
2641 rem = nfsm_rndup(nlen)-nlen;
2642 len += (4 * NFSX_UNSIGNED + nlen + rem);
2643 if (v3)
2644 len += 2 * NFSX_UNSIGNED;
2645 if (len > cnt) {
2646 eofflag = 0;
2647 break;
2648 }
2649 /*
2650 * Build the directory record xdr from
2651 * the dirent entry.
2652 */
2653 nfsm_clget;
2654 *tl = nfs_true;
2655 bp += NFSX_UNSIGNED;
2656 if (v3) {
2657 nfsm_clget;
2658 *tl = 0;
2659 bp += NFSX_UNSIGNED;
2660 }
2661 nfsm_clget;
2662 *tl = txdr_unsigned(dp->d_fileno);
2663 bp += NFSX_UNSIGNED;
2664 nfsm_clget;
2665 *tl = txdr_unsigned(nlen);
2666 bp += NFSX_UNSIGNED;
2667
2668 /* And loop around copying the name */
2669 xfer = nlen;
2670 cp = dp->d_name;
2671 while (xfer > 0) {
2672 nfsm_clget;
2673 if ((bp+xfer) > be)
2674 tsiz = be-bp;
2675 else
2676 tsiz = xfer;
2677 memcpy(bp, cp, tsiz);
2678 bp += tsiz;
2679 xfer -= tsiz;
2680 if (xfer > 0)
2681 cp += tsiz;
2682 }
2683 /* And null pad to an int32_t boundary */
2684 for (i = 0; i < rem; i++)
2685 *bp++ = '\0';
2686 nfsm_clget;
2687
2688 /* Finish off the record */
2689 txdr_hyper(*cookiep, &jar);
2690 if (v3) {
2691 *tl = jar.nfsuquad[0];
2692 bp += NFSX_UNSIGNED;
2693 nfsm_clget;
2694 }
2695 *tl = jar.nfsuquad[1];
2696 bp += NFSX_UNSIGNED;
2697 }
2698 cpos += dp->d_reclen;
2699 dp = (struct dirent *)cpos;
2700 cookiep++;
2701 ncookies--;
2702 }
2703 vrele(vp);
2704 nfsm_clget;
2705 *tl = nfs_false;
2706 bp += NFSX_UNSIGNED;
2707 nfsm_clget;
2708 if (eofflag)
2709 *tl = nfs_true;
2710 else
2711 *tl = nfs_false;
2712 bp += NFSX_UNSIGNED;
2713 if (mp != mb) {
2714 if (bp < be)
2715 mp->m_len = bp - mtod(mp, caddr_t);
2716 } else
2717 mp->m_len += bp - bpos;
2718 FREE((caddr_t)rbuf, M_TEMP);
2719 FREE((caddr_t)cookies, M_TEMP);
2720 nfsm_srvdone;
2721 }
2722
2723 int
2724 nfsrv_readdirplus(nfsd, slp, procp, mrq)
2725 struct nfsrv_descript *nfsd;
2726 struct nfssvc_sock *slp;
2727 struct proc *procp;
2728 struct mbuf **mrq;
2729 {
2730 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2731 struct mbuf *nam = nfsd->nd_nam;
2732 caddr_t dpos = nfsd->nd_dpos;
2733 struct ucred *cred = &nfsd->nd_cr;
2734 register char *bp, *be;
2735 register struct mbuf *mp;
2736 register struct dirent *dp;
2737 register caddr_t cp;
2738 register u_int32_t *tl;
2739 register int32_t t1;
2740 caddr_t bpos;
2741 struct mbuf *mb, *mb2, *mreq, *mp2;
2742 char *cpos, *cend, *cp2, *rbuf;
2743 struct vnode *vp, *nvp;
2744 struct flrep fl;
2745 nfsfh_t nfh;
2746 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
2747 struct uio io;
2748 struct iovec iv;
2749 struct vattr va, at, *vap = &va;
2750 struct nfs_fattr *fp;
2751 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2752 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies;
2753 u_quad_t frev, off, toff, verf;
2754 off_t *cookies = NULL, *cookiep;
2755
2756 fhp = &nfh.fh_generic;
2757 nfsm_srvmtofh(fhp);
2758 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2759 toff = fxdr_hyper(tl);
2760 tl += 2;
2761 verf = fxdr_hyper(tl);
2762 tl += 2;
2763 siz = fxdr_unsigned(int, *tl++);
2764 cnt = fxdr_unsigned(int, *tl);
2765 off = toff;
2766 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2767 xfer = NFS_SRVMAXDATA(nfsd);
2768 if (siz > xfer)
2769 siz = xfer;
2770 fullsiz = siz;
2771 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2772 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2773 if (!error && vp->v_type != VDIR) {
2774 error = ENOTDIR;
2775 vput(vp);
2776 }
2777 if (error) {
2778 nfsm_reply(NFSX_UNSIGNED);
2779 nfsm_srvpostop_attr(getret, &at);
2780 return (0);
2781 }
2782 error = getret = VOP_GETATTR(vp, &at, cred, procp);
2783 #ifdef NFS3_STRICTVERF
2784 /*
2785 * XXX This check is too strict for Solaris 2.5 clients.
2786 */
2787 if (!error && toff && verf != at.va_filerev)
2788 error = NFSERR_BAD_COOKIE;
2789 #endif
2790 if (!error) {
2791 nqsrv_getl(vp, ND_READ);
2792 error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2793 }
2794 if (error) {
2795 vput(vp);
2796 nfsm_reply(NFSX_V3POSTOPATTR);
2797 nfsm_srvpostop_attr(getret, &at);
2798 return (0);
2799 }
2800 VOP_UNLOCK(vp, 0);
2801
2802 MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
2803 again:
2804 iv.iov_base = rbuf;
2805 iv.iov_len = fullsiz;
2806 io.uio_iov = &iv;
2807 io.uio_iovcnt = 1;
2808 io.uio_offset = (off_t)off;
2809 io.uio_resid = fullsiz;
2810 io.uio_segflg = UIO_SYSSPACE;
2811 io.uio_rw = UIO_READ;
2812 io.uio_procp = (struct proc *)0;
2813 eofflag = 0;
2814
2815 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2816
2817 error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2818
2819 off = (u_quad_t)io.uio_offset;
2820 getret = VOP_GETATTR(vp, &at, cred, procp);
2821
2822 VOP_UNLOCK(vp, 0);
2823
2824 /*
2825 * If the VGET operation doesn't work for this filesystem,
2826 * we can't support readdirplus. Returning NOTSUPP should
2827 * make clients fall back to plain readdir.
2828 * There's no need to check for VPTOFH as well, we wouldn't
2829 * even be here otherwise.
2830 */
2831 if (!getret) {
2832 if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
2833 getret = (getret == EOPNOTSUPP) ?
2834 NFSERR_NOTSUPP : NFSERR_IO;
2835 else
2836 vput(nvp);
2837 }
2838
2839 if (!cookies && !error)
2840 error = NFSERR_PERM;
2841 if (!error)
2842 error = getret;
2843 if (error) {
2844 vrele(vp);
2845 if (cookies)
2846 FREE((caddr_t)cookies, M_TEMP);
2847 FREE((caddr_t)rbuf, M_TEMP);
2848 nfsm_reply(NFSX_V3POSTOPATTR);
2849 nfsm_srvpostop_attr(getret, &at);
2850 return (0);
2851 }
2852 if (io.uio_resid) {
2853 siz -= io.uio_resid;
2854
2855 /*
2856 * If nothing read, return eof
2857 * rpc reply
2858 */
2859 if (siz == 0) {
2860 vrele(vp);
2861 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
2862 2 * NFSX_UNSIGNED);
2863 nfsm_srvpostop_attr(getret, &at);
2864 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2865 txdr_hyper(at.va_filerev, tl);
2866 tl += 2;
2867 *tl++ = nfs_false;
2868 *tl = nfs_true;
2869 FREE((caddr_t)cookies, M_TEMP);
2870 FREE((caddr_t)rbuf, M_TEMP);
2871 return (0);
2872 }
2873 }
2874
2875 /*
2876 * Check for degenerate cases of nothing useful read.
2877 * If so go try again
2878 */
2879 cpos = rbuf;
2880 cend = rbuf + siz;
2881 dp = (struct dirent *)cpos;
2882 cookiep = cookies;
2883
2884 while (cpos < cend && ncookies > 0 &&
2885 (dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2886 cpos += dp->d_reclen;
2887 dp = (struct dirent *)cpos;
2888 cookiep++;
2889 ncookies--;
2890 }
2891 if (cpos >= cend || ncookies == 0) {
2892 toff = off;
2893 siz = fullsiz;
2894 goto again;
2895 }
2896
2897 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
2898 nfsm_reply(cnt);
2899 nfsm_srvpostop_attr(getret, &at);
2900 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2901 txdr_hyper(at.va_filerev, tl);
2902 mp = mp2 = mb;
2903 bp = bpos;
2904 be = bp + M_TRAILINGSPACE(mp);
2905
2906 /* Loop through the records and build reply */
2907 while (cpos < cend && ncookies > 0) {
2908 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2909 nlen = dp->d_namlen;
2910 rem = nfsm_rndup(nlen)-nlen;
2911
2912 /*
2913 * For readdir_and_lookup get the vnode using
2914 * the file number.
2915 */
2916 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
2917 goto invalid;
2918 memset((caddr_t)nfhp, 0, NFSX_V3FH);
2919 nfhp->fh_fsid =
2920 nvp->v_mount->mnt_stat.f_fsid;
2921 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
2922 vput(nvp);
2923 goto invalid;
2924 }
2925 if (VOP_GETATTR(nvp, vap, cred, procp)) {
2926 vput(nvp);
2927 goto invalid;
2928 }
2929 vput(nvp);
2930
2931 /*
2932 * If either the dircount or maxcount will be
2933 * exceeded, get out now. Both of these lengths
2934 * are calculated conservatively, including all
2935 * XDR overheads.
2936 */
2937 len += (7 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
2938 NFSX_V3POSTOPATTR);
2939 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
2940 if (len > cnt || dirlen > fullsiz) {
2941 eofflag = 0;
2942 break;
2943 }
2944
2945 /*
2946 * Build the directory record xdr from
2947 * the dirent entry.
2948 */
2949 fp = (struct nfs_fattr *)&fl.fl_fattr;
2950 nfsm_srvfillattr(vap, fp);
2951 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
2952 fl.fl_fhok = nfs_true;
2953 fl.fl_postopok = nfs_true;
2954 txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
2955
2956 nfsm_clget;
2957 *tl = nfs_true;
2958 bp += NFSX_UNSIGNED;
2959 nfsm_clget;
2960 *tl = 0;
2961 bp += NFSX_UNSIGNED;
2962 nfsm_clget;
2963 *tl = txdr_unsigned(dp->d_fileno);
2964 bp += NFSX_UNSIGNED;
2965 nfsm_clget;
2966 *tl = txdr_unsigned(nlen);
2967 bp += NFSX_UNSIGNED;
2968
2969 /* And loop around copying the name */
2970 xfer = nlen;
2971 cp = dp->d_name;
2972 while (xfer > 0) {
2973 nfsm_clget;
2974 if ((bp + xfer) > be)
2975 tsiz = be - bp;
2976 else
2977 tsiz = xfer;
2978 memcpy(bp, cp, tsiz);
2979 bp += tsiz;
2980 xfer -= tsiz;
2981 if (xfer > 0)
2982 cp += tsiz;
2983 }
2984 /* And null pad to an int32_t boundary */
2985 for (i = 0; i < rem; i++)
2986 *bp++ = '\0';
2987
2988 /*
2989 * Now copy the flrep structure out.
2990 */
2991 xfer = sizeof (struct flrep);
2992 cp = (caddr_t)&fl;
2993 while (xfer > 0) {
2994 nfsm_clget;
2995 if ((bp + xfer) > be)
2996 tsiz = be - bp;
2997 else
2998 tsiz = xfer;
2999 memcpy(bp, cp, tsiz);
3000 bp += tsiz;
3001 xfer -= tsiz;
3002 if (xfer > 0)
3003 cp += tsiz;
3004 }
3005 }
3006 invalid:
3007 cpos += dp->d_reclen;
3008 dp = (struct dirent *)cpos;
3009 cookiep++;
3010 ncookies--;
3011 }
3012 vrele(vp);
3013 nfsm_clget;
3014 *tl = nfs_false;
3015 bp += NFSX_UNSIGNED;
3016 nfsm_clget;
3017 if (eofflag)
3018 *tl = nfs_true;
3019 else
3020 *tl = nfs_false;
3021 bp += NFSX_UNSIGNED;
3022 if (mp != mb) {
3023 if (bp < be)
3024 mp->m_len = bp - mtod(mp, caddr_t);
3025 } else
3026 mp->m_len += bp - bpos;
3027 FREE((caddr_t)cookies, M_TEMP);
3028 FREE((caddr_t)rbuf, M_TEMP);
3029 nfsm_srvdone;
3030 }
3031
3032 /*
3033 * nfs commit service
3034 */
3035 int
3036 nfsrv_commit(nfsd, slp, procp, mrq)
3037 struct nfsrv_descript *nfsd;
3038 struct nfssvc_sock *slp;
3039 struct proc *procp;
3040 struct mbuf **mrq;
3041 {
3042 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3043 struct mbuf *nam = nfsd->nd_nam;
3044 caddr_t dpos = nfsd->nd_dpos;
3045 struct ucred *cred = &nfsd->nd_cr;
3046 struct vattr bfor, aft;
3047 struct vnode *vp;
3048 nfsfh_t nfh;
3049 fhandle_t *fhp;
3050 register u_int32_t *tl;
3051 register int32_t t1;
3052 caddr_t bpos;
3053 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache;
3054 char *cp2;
3055 struct mbuf *mb, *mb2, *mreq;
3056 u_quad_t frev, off;
3057
3058 #ifndef nolint
3059 cache = 0;
3060 #endif
3061 fhp = &nfh.fh_generic;
3062 nfsm_srvmtofh(fhp);
3063 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3064
3065 /*
3066 * XXX At this time VOP_FSYNC() does not accept offset and byte
3067 * count parameters, so these arguments are useless (someday maybe).
3068 */
3069 off = fxdr_hyper(tl);
3070 tl += 2;
3071 cnt = fxdr_unsigned(int, *tl);
3072 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3073 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3074 if (error) {
3075 nfsm_reply(2 * NFSX_UNSIGNED);
3076 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3077 return (0);
3078 }
3079 for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
3080 error = VOP_FSYNC(vp, cred, FSYNC_WAIT, procp);
3081 aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
3082 vput(vp);
3083 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3084 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3085 if (!error) {
3086 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3087 *tl++ = txdr_unsigned(boottime.tv_sec);
3088 *tl = txdr_unsigned(boottime.tv_usec);
3089 } else
3090 return (0);
3091 nfsm_srvdone;
3092 }
3093
3094 /*
3095 * nfs statfs service
3096 */
3097 int
3098 nfsrv_statfs(nfsd, slp, procp, mrq)
3099 struct nfsrv_descript *nfsd;
3100 struct nfssvc_sock *slp;
3101 struct proc *procp;
3102 struct mbuf **mrq;
3103 {
3104 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3105 struct mbuf *nam = nfsd->nd_nam;
3106 caddr_t dpos = nfsd->nd_dpos;
3107 struct ucred *cred = &nfsd->nd_cr;
3108 register struct statfs *sf;
3109 register struct nfs_statfs *sfp;
3110 register u_int32_t *tl;
3111 register int32_t t1;
3112 caddr_t bpos;
3113 int error = 0, rdonly, cache, getret = 1;
3114 int v3 = (nfsd->nd_flag & ND_NFSV3);
3115 char *cp2;
3116 struct mbuf *mb, *mb2, *mreq;
3117 struct vnode *vp;
3118 struct vattr at;
3119 nfsfh_t nfh;
3120 fhandle_t *fhp;
3121 struct statfs statfs;
3122 u_quad_t frev, tval;
3123
3124 #ifndef nolint
3125 cache = 0;
3126 #endif
3127 fhp = &nfh.fh_generic;
3128 nfsm_srvmtofh(fhp);
3129 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3130 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3131 if (error) {
3132 nfsm_reply(NFSX_UNSIGNED);
3133 nfsm_srvpostop_attr(getret, &at);
3134 return (0);
3135 }
3136 sf = &statfs;
3137 error = VFS_STATFS(vp->v_mount, sf, procp);
3138 getret = VOP_GETATTR(vp, &at, cred, procp);
3139 vput(vp);
3140 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3141 if (v3)
3142 nfsm_srvpostop_attr(getret, &at);
3143 if (error)
3144 return (0);
3145 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3146 if (v3) {
3147 tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_bsize);
3148 txdr_hyper(tval, &sfp->sf_tbytes);
3149 tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_bsize);
3150 txdr_hyper(tval, &sfp->sf_fbytes);
3151 tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_bsize);
3152 txdr_hyper(tval, &sfp->sf_abytes);
3153 tval = (u_quad_t)sf->f_files;
3154 txdr_hyper(tval, &sfp->sf_tfiles);
3155 tval = (u_quad_t)sf->f_ffree;
3156 txdr_hyper(tval, &sfp->sf_ffiles);
3157 txdr_hyper(tval, &sfp->sf_afiles);
3158 sfp->sf_invarsec = 0;
3159 } else {
3160 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3161 sfp->sf_bsize = txdr_unsigned(sf->f_bsize);
3162 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3163 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3164 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3165 }
3166 nfsm_srvdone;
3167 }
3168
3169 /*
3170 * nfs fsinfo service
3171 */
3172 int
3173 nfsrv_fsinfo(nfsd, slp, procp, mrq)
3174 struct nfsrv_descript *nfsd;
3175 struct nfssvc_sock *slp;
3176 struct proc *procp;
3177 struct mbuf **mrq;
3178 {
3179 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3180 struct mbuf *nam = nfsd->nd_nam;
3181 caddr_t dpos = nfsd->nd_dpos;
3182 struct ucred *cred = &nfsd->nd_cr;
3183 register u_int32_t *tl;
3184 register struct nfsv3_fsinfo *sip;
3185 register int32_t t1;
3186 caddr_t bpos;
3187 int error = 0, rdonly, cache, getret = 1, pref;
3188 char *cp2;
3189 struct mbuf *mb, *mb2, *mreq;
3190 struct vnode *vp;
3191 struct vattr at;
3192 nfsfh_t nfh;
3193 fhandle_t *fhp;
3194 u_quad_t frev, maxfsize;
3195 struct statfs sb;
3196
3197 #ifndef nolint
3198 cache = 0;
3199 #endif
3200 fhp = &nfh.fh_generic;
3201 nfsm_srvmtofh(fhp);
3202 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3203 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3204 if (error) {
3205 nfsm_reply(NFSX_UNSIGNED);
3206 nfsm_srvpostop_attr(getret, &at);
3207 return (0);
3208 }
3209
3210 /* XXX Try to make a guess on the max file size. */
3211 VFS_STATFS(vp->v_mount, &sb, (struct proc *)0);
3212 maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1;
3213
3214 getret = VOP_GETATTR(vp, &at, cred, procp);
3215 vput(vp);
3216 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3217 nfsm_srvpostop_attr(getret, &at);
3218 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3219
3220 /*
3221 * XXX
3222 * There should be file system VFS OP(s) to get this information.
3223 * For now, assume ufs.
3224 */
3225 if (slp->ns_so->so_type == SOCK_DGRAM)
3226 pref = NFS_MAXDGRAMDATA;
3227 else
3228 pref = NFS_MAXDATA;
3229 sip->fs_rtmax = txdr_unsigned(NFS_MAXDATA);
3230 sip->fs_rtpref = txdr_unsigned(pref);
3231 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3232 sip->fs_wtmax = txdr_unsigned(NFS_MAXDATA);
3233 sip->fs_wtpref = txdr_unsigned(pref);
3234 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3235 sip->fs_dtpref = txdr_unsigned(pref);
3236 txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3237 sip->fs_timedelta.nfsv3_sec = 0;
3238 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3239 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3240 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3241 NFSV3FSINFO_CANSETTIME);
3242 nfsm_srvdone;
3243 }
3244
3245 /*
3246 * nfs pathconf service
3247 */
3248 int
3249 nfsrv_pathconf(nfsd, slp, procp, mrq)
3250 struct nfsrv_descript *nfsd;
3251 struct nfssvc_sock *slp;
3252 struct proc *procp;
3253 struct mbuf **mrq;
3254 {
3255 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3256 struct mbuf *nam = nfsd->nd_nam;
3257 caddr_t dpos = nfsd->nd_dpos;
3258 struct ucred *cred = &nfsd->nd_cr;
3259 register u_int32_t *tl;
3260 register struct nfsv3_pathconf *pc;
3261 register int32_t t1;
3262 caddr_t bpos;
3263 int error = 0, rdonly, cache, getret = 1;
3264 register_t linkmax, namemax, chownres, notrunc;
3265 char *cp2;
3266 struct mbuf *mb, *mb2, *mreq;
3267 struct vnode *vp;
3268 struct vattr at;
3269 nfsfh_t nfh;
3270 fhandle_t *fhp;
3271 u_quad_t frev;
3272
3273 #ifndef nolint
3274 cache = 0;
3275 #endif
3276 fhp = &nfh.fh_generic;
3277 nfsm_srvmtofh(fhp);
3278 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3279 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3280 if (error) {
3281 nfsm_reply(NFSX_UNSIGNED);
3282 nfsm_srvpostop_attr(getret, &at);
3283 return (0);
3284 }
3285 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3286 if (!error)
3287 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3288 if (!error)
3289 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3290 if (!error)
3291 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc);
3292 getret = VOP_GETATTR(vp, &at, cred, procp);
3293 vput(vp);
3294 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3295 nfsm_srvpostop_attr(getret, &at);
3296 if (error)
3297 return (0);
3298 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3299
3300 pc->pc_linkmax = txdr_unsigned(linkmax);
3301 pc->pc_namemax = txdr_unsigned(namemax);
3302 pc->pc_notrunc = txdr_unsigned(notrunc);
3303 pc->pc_chownrestricted = txdr_unsigned(chownres);
3304
3305 /*
3306 * These should probably be supported by VOP_PATHCONF(), but
3307 * until msdosfs is exportable (why would you want to?), the
3308 * Unix defaults should be ok.
3309 */
3310 pc->pc_caseinsensitive = nfs_false;
3311 pc->pc_casepreserving = nfs_true;
3312 nfsm_srvdone;
3313 }
3314
3315 /*
3316 * Null operation, used by clients to ping server
3317 */
3318 /* ARGSUSED */
3319 int
3320 nfsrv_null(nfsd, slp, procp, mrq)
3321 struct nfsrv_descript *nfsd;
3322 struct nfssvc_sock *slp;
3323 struct proc *procp;
3324 struct mbuf **mrq;
3325 {
3326 struct mbuf *mrep = nfsd->nd_mrep;
3327 caddr_t bpos;
3328 int error = NFSERR_RETVOID, cache = 0;
3329 struct mbuf *mb, *mreq;
3330 u_quad_t frev;
3331
3332 nfsm_reply(0);
3333 return (0);
3334 }
3335
3336 /*
3337 * No operation, used for obsolete procedures
3338 */
3339 /* ARGSUSED */
3340 int
3341 nfsrv_noop(nfsd, slp, procp, mrq)
3342 struct nfsrv_descript *nfsd;
3343 struct nfssvc_sock *slp;
3344 struct proc *procp;
3345 struct mbuf **mrq;
3346 {
3347 struct mbuf *mrep = nfsd->nd_mrep;
3348 caddr_t bpos;
3349 int error, cache = 0;
3350 struct mbuf *mb, *mreq;
3351 u_quad_t frev;
3352
3353 if (nfsd->nd_repstat)
3354 error = nfsd->nd_repstat;
3355 else
3356 error = EPROCUNAVAIL;
3357 nfsm_reply(0);
3358 return (0);
3359 }
3360
3361 /*
3362 * Perform access checking for vnodes obtained from file handles that would
3363 * refer to files already opened by a Unix client. You cannot just use
3364 * vn_writechk() and VOP_ACCESS() for two reasons.
3365 * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3366 * 2 - The owner is to be given access irrespective of mode bits for some
3367 * operations, so that processes that chmod after opening a file don't
3368 * break. I don't like this because it opens a security hole, but since
3369 * the nfs server opens a security hole the size of a barn door anyhow,
3370 * what the heck.
3371 *
3372 * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3373 * will return EPERM instead of EACCESS. EPERM is always an error.
3374 */
3375 int
3376 nfsrv_access(vp, flags, cred, rdonly, p, override)
3377 register struct vnode *vp;
3378 int flags;
3379 register struct ucred *cred;
3380 int rdonly;
3381 struct proc *p;
3382 {
3383 struct vattr vattr;
3384 int error;
3385 if (flags & VWRITE) {
3386 /* Just vn_writechk() changed to check rdonly */
3387 /*
3388 * Disallow write attempts on read-only file systems;
3389 * unless the file is a socket or a block or character
3390 * device resident on the file system.
3391 */
3392 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3393 switch (vp->v_type) {
3394 case VREG:
3395 case VDIR:
3396 case VLNK:
3397 return (EROFS);
3398 default:
3399 break;
3400 }
3401 }
3402 /*
3403 * If there's shared text associated with
3404 * the inode, try to free it up once. If
3405 * we fail, we can't allow writing.
3406 */
3407 if ((vp->v_flag & VTEXT) && !uvm_vnp_uncache(vp))
3408 return (ETXTBSY);
3409 }
3410 error = VOP_GETATTR(vp, &vattr, cred, p);
3411 if (error)
3412 return (error);
3413 error = VOP_ACCESS(vp, flags, cred, p);
3414 /*
3415 * Allow certain operations for the owner (reads and writes
3416 * on files that are already open).
3417 */
3418 if (override && error == EACCES && cred->cr_uid == vattr.va_uid)
3419 error = 0;
3420 return error;
3421 }
3422