Document addition of ICH6 UHCI.
[dragonfly.git] / sys / vfs / nfs / nfs_serv.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95
37 * $FreeBSD: src/sys/nfs/nfs_serv.c,v 1.93.2.6 2002/12/29 18:19:53 dillon Exp $
74f1caca 38 * $DragonFly: src/sys/vfs/nfs/nfs_serv.c,v 1.16 2004/06/02 14:43:04 eirikn Exp $
984263bc
MD
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 * Other notes:
61 * Warning: always pay careful attention to resource cleanup on return
62 * and note that nfsm_*() macros can terminate a procedure on certain
63 * errors.
64 *
65 * lookup() and namei()
66 * may return garbage in various structural fields/return elements
67 * if an error is returned, and may garbage up nd.ni_dvp even if no
68 * error is returned and you did not request LOCKPARENT or WANTPARENT.
69 *
70 * We use the ni_cnd.cn_flags 'HASBUF' flag to track whether the name
71 * buffer has been freed or not.
72 */
73
74#include <sys/param.h>
75#include <sys/systm.h>
76#include <sys/proc.h>
77#include <sys/namei.h>
78#include <sys/unistd.h>
79#include <sys/vnode.h>
80#include <sys/mount.h>
81#include <sys/socket.h>
82#include <sys/socketvar.h>
83#include <sys/malloc.h>
84#include <sys/mbuf.h>
85#include <sys/dirent.h>
86#include <sys/stat.h>
87#include <sys/kernel.h>
88#include <sys/sysctl.h>
89#include <sys/buf.h>
90
91#include <vm/vm.h>
92#include <vm/vm_extern.h>
93#include <vm/vm_zone.h>
94#include <vm/vm_object.h>
95
3020e3be
MD
96#include <sys/buf2.h>
97
1f2de5d4
MD
98#include "nfsproto.h"
99#include "rpcv2.h"
100#include "nfs.h"
101#include "xdr_subs.h"
102#include "nfsm_subs.h"
103#include "nqnfs.h"
984263bc
MD
104
105#ifdef NFSRV_DEBUG
106#define nfsdbprintf(info) printf info
107#else
108#define nfsdbprintf(info)
109#endif
110
111#define MAX_COMMIT_COUNT (1024 * 1024)
112
113#define NUM_HEURISTIC 64
114#define NHUSE_INIT 64
115#define NHUSE_INC 16
116#define NHUSE_MAX 2048
117
118static struct nfsheur {
119 struct vnode *nh_vp; /* vp to match (unreferenced pointer) */
120 off_t nh_nextr; /* next offset for sequential detection */
121 int nh_use; /* use count for selection */
122 int nh_seqcount; /* heuristic */
123} nfsheur[NUM_HEURISTIC];
124
125nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK,
126 NFFIFO, NFNON };
127#ifndef NFS_NOSERVER
128nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON,
129 NFCHR, NFNON };
130/* Global vars */
131extern u_int32_t nfs_xdrneg1;
132extern u_int32_t nfs_false, nfs_true;
133extern enum vtype nv3tov_type[8];
134extern struct nfsstats nfsstats;
135
136int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
137int nfsrvw_procrastinate_v3 = 0;
138
3a6f9faf 139static struct timespec nfsver;
984263bc
MD
140
141SYSCTL_DECL(_vfs_nfs);
142
143static int nfs_async;
144SYSCTL_INT(_vfs_nfs, OID_AUTO, async, CTLFLAG_RW, &nfs_async, 0, "");
145static int nfs_commit_blks;
146static int nfs_commit_miss;
147SYSCTL_INT(_vfs_nfs, OID_AUTO, commit_blks, CTLFLAG_RW, &nfs_commit_blks, 0, "");
148SYSCTL_INT(_vfs_nfs, OID_AUTO, commit_miss, CTLFLAG_RW, &nfs_commit_miss, 0, "");
149
a6ee311a
RG
150static int nfsrv_access (struct vnode *,int,struct ucred *,int,
151 struct thread *, int);
152static void nfsrvw_coalesce (struct nfsrv_descript *,
153 struct nfsrv_descript *);
984263bc
MD
154
155/*
156 * Clear nameidata fields that are tested in nsfmout cleanup code prior
157 * to using first nfsm macro (that might jump to the cleanup code).
158 */
159
160static __inline
161void
162ndclear(struct nameidata *nd)
163{
164 nd->ni_cnd.cn_flags = 0;
165 nd->ni_vp = NULL;
166 nd->ni_dvp = NULL;
167 nd->ni_startdir = NULL;
168}
169
170/*
171 * nfs v3 access service
172 */
173int
dadab5e9 174nfsrv3_access(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
e851b29e 175 struct thread *td, struct mbuf **mrq)
984263bc
MD
176{
177 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
178 struct sockaddr *nam = nfsd->nd_nam;
179 caddr_t dpos = nfsd->nd_dpos;
180 struct ucred *cred = &nfsd->nd_cr;
181 struct vnode *vp = NULL;
182 nfsfh_t nfh;
183 fhandle_t *fhp;
dadab5e9
MD
184 u_int32_t *tl;
185 int32_t t1;
984263bc
MD
186 caddr_t bpos;
187 int error = 0, rdonly, cache, getret;
188 char *cp2;
189 struct mbuf *mb, *mreq, *mb2;
190 struct vattr vattr, *vap = &vattr;
191 u_long testmode, nfsmode;
192 u_quad_t frev;
193
194 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
195#ifndef nolint
196 cache = 0;
197#endif
198 fhp = &nfh.fh_generic;
199 nfsm_srvmtofh(fhp);
200 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
201 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
202 (nfsd->nd_flag & ND_KERBAUTH), TRUE);
203 if (error) {
204 nfsm_reply(NFSX_UNSIGNED);
205 nfsm_srvpostop_attr(1, (struct vattr *)0);
206 error = 0;
207 goto nfsmout;
208 }
209 nfsmode = fxdr_unsigned(u_int32_t, *tl);
210 if ((nfsmode & NFSV3ACCESS_READ) &&
dadab5e9 211 nfsrv_access(vp, VREAD, cred, rdonly, td, 0))
984263bc
MD
212 nfsmode &= ~NFSV3ACCESS_READ;
213 if (vp->v_type == VDIR)
214 testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
215 NFSV3ACCESS_DELETE);
216 else
217 testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
218 if ((nfsmode & testmode) &&
dadab5e9 219 nfsrv_access(vp, VWRITE, cred, rdonly, td, 0))
984263bc
MD
220 nfsmode &= ~testmode;
221 if (vp->v_type == VDIR)
222 testmode = NFSV3ACCESS_LOOKUP;
223 else
224 testmode = NFSV3ACCESS_EXECUTE;
225 if ((nfsmode & testmode) &&
dadab5e9 226 nfsrv_access(vp, VEXEC, cred, rdonly, td, 0))
984263bc 227 nfsmode &= ~testmode;
3b568787 228 getret = VOP_GETATTR(vp, vap, td);
984263bc
MD
229 vput(vp);
230 vp = NULL;
231 nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
232 nfsm_srvpostop_attr(getret, vap);
233 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
234 *tl = txdr_unsigned(nfsmode);
235nfsmout:
236 if (vp)
237 vput(vp);
238 return(error);
239}
240
241/*
242 * nfs getattr service
243 */
244int
e851b29e
CP
245nfsrv_getattr(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
246 struct thread *td, struct mbuf **mrq)
984263bc
MD
247{
248 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
249 struct sockaddr *nam = nfsd->nd_nam;
250 caddr_t dpos = nfsd->nd_dpos;
251 struct ucred *cred = &nfsd->nd_cr;
dadab5e9 252 struct nfs_fattr *fp;
984263bc 253 struct vattr va;
dadab5e9 254 struct vattr *vap = &va;
984263bc
MD
255 struct vnode *vp = NULL;
256 nfsfh_t nfh;
257 fhandle_t *fhp;
dadab5e9
MD
258 u_int32_t *tl;
259 int32_t t1;
984263bc
MD
260 caddr_t bpos;
261 int error = 0, rdonly, cache;
262 char *cp2;
263 struct mbuf *mb, *mb2, *mreq;
264 u_quad_t frev;
265
266 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
267 fhp = &nfh.fh_generic;
268 nfsm_srvmtofh(fhp);
269 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
270 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
271 if (error) {
272 nfsm_reply(0);
273 error = 0;
274 goto nfsmout;
275 }
276 nqsrv_getl(vp, ND_READ);
3b568787 277 error = VOP_GETATTR(vp, vap, td);
984263bc
MD
278 vput(vp);
279 vp = NULL;
280 nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
281 if (error) {
282 error = 0;
283 goto nfsmout;
284 }
285 nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
286 nfsm_srvfillattr(vap, fp);
287 /* fall through */
288
289nfsmout:
290 if (vp)
291 vput(vp);
292 return(error);
293}
294
295/*
296 * nfs setattr service
297 */
298int
e851b29e
CP
299nfsrv_setattr(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
300 struct thread *td, struct mbuf **mrq)
984263bc
MD
301{
302 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
303 struct sockaddr *nam = nfsd->nd_nam;
304 caddr_t dpos = nfsd->nd_dpos;
305 struct ucred *cred = &nfsd->nd_cr;
306 struct vattr va, preat;
dadab5e9
MD
307 struct vattr *vap = &va;
308 struct nfsv2_sattr *sp;
309 struct nfs_fattr *fp;
984263bc
MD
310 struct vnode *vp = NULL;
311 nfsfh_t nfh;
312 fhandle_t *fhp;
dadab5e9
MD
313 u_int32_t *tl;
314 int32_t t1;
984263bc
MD
315 caddr_t bpos;
316 int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1;
317 int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
318 char *cp2;
319 struct mbuf *mb, *mb2, *mreq;
320 u_quad_t frev;
321 struct timespec guard;
322
323 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
324 fhp = &nfh.fh_generic;
325 nfsm_srvmtofh(fhp);
326 VATTR_NULL(vap);
327 if (v3) {
328 nfsm_srvsattr(vap);
329 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
330 gcheck = fxdr_unsigned(int, *tl);
331 if (gcheck) {
332 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
333 fxdr_nfsv3time(tl, &guard);
334 }
335 } else {
336 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
337 /*
338 * Nah nah nah nah na nah
339 * There is a bug in the Sun client that puts 0xffff in the mode
340 * field of sattr when it should put in 0xffffffff. The u_short
341 * doesn't sign extend.
342 * --> check the low order 2 bytes for 0xffff
343 */
344 if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
345 vap->va_mode = nfstov_mode(sp->sa_mode);
346 if (sp->sa_uid != nfs_xdrneg1)
347 vap->va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
348 if (sp->sa_gid != nfs_xdrneg1)
349 vap->va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
350 if (sp->sa_size != nfs_xdrneg1)
351 vap->va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
352 if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
353#ifdef notyet
354 fxdr_nfsv2time(&sp->sa_atime, &vap->va_atime);
355#else
356 vap->va_atime.tv_sec =
357 fxdr_unsigned(int32_t, sp->sa_atime.nfsv2_sec);
358 vap->va_atime.tv_nsec = 0;
359#endif
360 }
361 if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
362 fxdr_nfsv2time(&sp->sa_mtime, &vap->va_mtime);
363
364 }
365
366 /*
367 * Now that we have all the fields, lets do it.
368 */
369 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
370 (nfsd->nd_flag & ND_KERBAUTH), TRUE);
371 if (error) {
372 nfsm_reply(2 * NFSX_UNSIGNED);
373 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap);
374 error = 0;
375 goto nfsmout;
376 }
377
378 /*
379 * vp now an active resource, pay careful attention to cleanup
380 */
381
382 nqsrv_getl(vp, ND_WRITE);
383 if (v3) {
3b568787 384 error = preat_ret = VOP_GETATTR(vp, &preat, td);
984263bc
MD
385 if (!error && gcheck &&
386 (preat.va_ctime.tv_sec != guard.tv_sec ||
387 preat.va_ctime.tv_nsec != guard.tv_nsec))
388 error = NFSERR_NOT_SYNC;
389 if (error) {
390 vput(vp);
391 vp = NULL;
392 nfsm_reply(NFSX_WCCDATA(v3));
393 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap);
394 error = 0;
395 goto nfsmout;
396 }
397 }
398
399 /*
400 * If the size is being changed write acces is required, otherwise
401 * just check for a read only file system.
402 */
403 if (vap->va_size == ((u_quad_t)((quad_t) -1))) {
404 if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
405 error = EROFS;
406 goto out;
407 }
408 } else {
409 if (vp->v_type == VDIR) {
410 error = EISDIR;
411 goto out;
412 } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
dadab5e9 413 td, 0)) != 0){
984263bc 414 goto out;
dadab5e9 415 }
984263bc 416 }
dadab5e9 417 error = VOP_SETATTR(vp, vap, cred, td);
3b568787 418 postat_ret = VOP_GETATTR(vp, vap, td);
984263bc
MD
419 if (!error)
420 error = postat_ret;
421out:
422 vput(vp);
423 vp = NULL;
424 nfsm_reply(NFSX_WCCORFATTR(v3));
425 if (v3) {
426 nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap);
427 error = 0;
428 goto nfsmout;
429 } else {
430 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
431 nfsm_srvfillattr(vap, fp);
432 }
433 /* fall through */
434
435nfsmout:
436 if (vp)
437 vput(vp);
438 return(error);
439}
440
441/*
442 * nfs lookup rpc
443 */
444int
e851b29e
CP
445nfsrv_lookup(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
446 struct thread *td, struct mbuf **mrq)
984263bc
MD
447{
448 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
449 struct sockaddr *nam = nfsd->nd_nam;
450 caddr_t dpos = nfsd->nd_dpos;
451 struct ucred *cred = &nfsd->nd_cr;
dadab5e9 452 struct nfs_fattr *fp;
984263bc
MD
453 struct nameidata nd, ind, *ndp = &nd;
454 struct vnode *vp, *dirp = NULL;
455 nfsfh_t nfh;
456 fhandle_t *fhp;
dadab5e9
MD
457 caddr_t cp;
458 u_int32_t *tl;
459 int32_t t1;
984263bc
MD
460 caddr_t bpos;
461 int error = 0, cache, len, dirattr_ret = 1;
462 int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
463 char *cp2;
464 struct mbuf *mb, *mb2, *mreq;
465 struct vattr va, dirattr, *vap = &va;
466 u_quad_t frev;
467
468 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
469 ndclear(&nd);
470
471 fhp = &nfh.fh_generic;
472 nfsm_srvmtofh(fhp);
473 nfsm_srvnamesiz(len);
474
475 pubflag = nfs_ispublicfh(fhp);
476
477 nd.ni_cnd.cn_cred = cred;
2b69e610
MD
478 nd.ni_cnd.cn_nameiop = NAMEI_LOOKUP;
479 nd.ni_cnd.cn_flags = CNP_LOCKLEAF | CNP_SAVESTART;
984263bc 480 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
dadab5e9 481 &dirp, td, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
984263bc
MD
482
483 /*
484 * namei failure, only dirp to cleanup. Clear out garbarge from
485 * structure in case macros jump to nfsmout.
486 */
487
488 if (error) {
489 if (dirp) {
490 if (v3)
3b568787 491 dirattr_ret = VOP_GETATTR(dirp, &dirattr, td);
984263bc
MD
492 vrele(dirp);
493 dirp = NULL;
494 }
495 nfsm_reply(NFSX_POSTOPATTR(v3));
496 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
497 error = 0;
498 goto nfsmout;
499 }
500
501 /*
502 * Locate index file for public filehandle
503 *
504 * error is 0 on entry and 0 on exit from this block.
505 */
506
507 if (pubflag) {
508 if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
509 /*
510 * Setup call to lookup() to see if we can find
511 * the index file. Arguably, this doesn't belong
512 * in a kernel.. Ugh. If an error occurs, do not
513 * try to install an index file and then clear the
514 * error.
515 *
516 * When we replace nd with ind and redirect ndp,
517 * maintenance of ni_startdir and ni_vp shift to
518 * ind and we have to clean them up in the old nd.
519 * However, the cnd resource continues to be maintained
520 * via the original nd. Confused? You aren't alone!
521 */
522 ind = nd;
41a01a4d 523 VOP_UNLOCK(nd.ni_vp, NULL, 0, td);
984263bc
MD
524 ind.ni_pathlen = strlen(nfs_pub.np_index);
525 ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
526 nfs_pub.np_index;
527 ind.ni_startdir = nd.ni_vp;
597aea93 528 vref(ind.ni_startdir);
984263bc
MD
529
530 error = lookup(&ind);
531 ind.ni_dvp = NULL;
532
533 if (error == 0) {
534 /*
535 * Found an index file. Get rid of
536 * the old references. transfer nd.ni_vp'
537 */
538 if (dirp)
539 vrele(dirp);
540 dirp = nd.ni_vp;
541 nd.ni_vp = NULL;
542 vrele(nd.ni_startdir);
543 nd.ni_startdir = NULL;
544 ndp = &ind;
545 }
546 error = 0;
547 }
548 /*
549 * If the public filehandle was used, check that this lookup
550 * didn't result in a filehandle outside the publicly exported
551 * filesystem. We clear the poor vp here to avoid lockups due
552 * to NFS I/O.
553 */
554
555 if (ndp->ni_vp->v_mount != nfs_pub.np_mount) {
556 vput(nd.ni_vp);
557 nd.ni_vp = NULL;
558 error = EPERM;
559 }
560 }
561
562 if (dirp) {
563 if (v3)
3b568787 564 dirattr_ret = VOP_GETATTR(dirp, &dirattr, td);
984263bc
MD
565 vrele(dirp);
566 dirp = NULL;
567 }
568
569 /*
570 * Resources at this point:
571 * ndp->ni_vp may not be NULL
572 *
573 */
574
575 if (error) {
576 nfsm_reply(NFSX_POSTOPATTR(v3));
577 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
578 error = 0;
579 goto nfsmout;
580 }
581
582 nqsrv_getl(ndp->ni_startdir, ND_READ);
583
584 /*
585 * Clear out some resources prior to potentially blocking. This
586 * is not as critical as ni_dvp resources in other routines, but
587 * it helps.
588 */
589 vrele(ndp->ni_startdir);
590 ndp->ni_startdir = NULL;
591 NDFREE(&nd, NDF_ONLY_PNBUF);
592
593 /*
594 * Get underlying attribute, then release remaining resources ( for
595 * the same potential blocking reason ) and reply.
596 */
597 vp = ndp->ni_vp;
598 bzero((caddr_t)fhp, sizeof(nfh));
599 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
600 error = VFS_VPTOFH(vp, &fhp->fh_fid);
601 if (!error)
3b568787 602 error = VOP_GETATTR(vp, vap, td);
984263bc
MD
603
604 vput(vp);
605 ndp->ni_vp = NULL;
606 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
607 if (error) {
608 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
609 error = 0;
610 goto nfsmout;
611 }
612 nfsm_srvfhtom(fhp, v3);
613 if (v3) {
614 nfsm_srvpostop_attr(0, vap);
615 nfsm_srvpostop_attr(dirattr_ret, &dirattr);
616 } else {
617 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
618 nfsm_srvfillattr(vap, fp);
619 }
620
621nfsmout:
622 if (dirp)
623 vrele(dirp);
624 NDFREE(&nd, NDF_ONLY_PNBUF);
625 if (ndp->ni_startdir)
626 vrele(ndp->ni_startdir);
627 if (ndp->ni_vp)
628 vput(ndp->ni_vp);
629 return (error);
630}
631
632/*
633 * nfs readlink service
634 */
635int
e851b29e
CP
636nfsrv_readlink(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
637 struct thread *td, struct mbuf **mrq)
984263bc
MD
638{
639 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
640 struct sockaddr *nam = nfsd->nd_nam;
641 caddr_t dpos = nfsd->nd_dpos;
642 struct ucred *cred = &nfsd->nd_cr;
643 struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
dadab5e9
MD
644 struct iovec *ivp = iv;
645 struct mbuf *mp;
646 u_int32_t *tl;
647 int32_t t1;
984263bc
MD
648 caddr_t bpos;
649 int error = 0, rdonly, cache, i, tlen, len, getret;
650 int v3 = (nfsd->nd_flag & ND_NFSV3);
651 char *cp2;
652 struct mbuf *mb, *mb2, *mp2, *mp3, *mreq;
653 struct vnode *vp = NULL;
654 struct vattr attr;
655 nfsfh_t nfh;
656 fhandle_t *fhp;
657 struct uio io, *uiop = &io;
658 u_quad_t frev;
659
660 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
661#ifndef nolint
662 mp2 = (struct mbuf *)0;
663#endif
664 mp3 = NULL;
665 fhp = &nfh.fh_generic;
666 nfsm_srvmtofh(fhp);
667 len = 0;
668 i = 0;
669 while (len < NFS_MAXPATHLEN) {
74f1caca
EN
670 MGET(mp, MB_WAIT, MT_DATA);
671 MCLGET(mp, MB_WAIT);
984263bc
MD
672 mp->m_len = NFSMSIZ(mp);
673 if (len == 0)
674 mp3 = mp2 = mp;
675 else {
676 mp2->m_next = mp;
677 mp2 = mp;
678 }
679 if ((len+mp->m_len) > NFS_MAXPATHLEN) {
680 mp->m_len = NFS_MAXPATHLEN-len;
681 len = NFS_MAXPATHLEN;
682 } else
683 len += mp->m_len;
684 ivp->iov_base = mtod(mp, caddr_t);
685 ivp->iov_len = mp->m_len;
686 i++;
687 ivp++;
688 }
689 uiop->uio_iov = iv;
690 uiop->uio_iovcnt = i;
691 uiop->uio_offset = 0;
692 uiop->uio_resid = len;
693 uiop->uio_rw = UIO_READ;
694 uiop->uio_segflg = UIO_SYSSPACE;
dadab5e9 695 uiop->uio_td = NULL;
984263bc
MD
696 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
697 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
698 if (error) {
699 nfsm_reply(2 * NFSX_UNSIGNED);
700 nfsm_srvpostop_attr(1, (struct vattr *)0);
701 error = 0;
702 goto nfsmout;
703 }
704 if (vp->v_type != VLNK) {
705 if (v3)
706 error = EINVAL;
707 else
708 error = ENXIO;
709 goto out;
710 }
711 nqsrv_getl(vp, ND_READ);
712 error = VOP_READLINK(vp, uiop, cred);
713out:
3b568787 714 getret = VOP_GETATTR(vp, &attr, td);
984263bc
MD
715 vput(vp);
716 vp = NULL;
717 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
718 if (v3) {
719 nfsm_srvpostop_attr(getret, &attr);
720 if (error) {
721 error = 0;
722 goto nfsmout;
723 }
724 }
725 if (uiop->uio_resid > 0) {
726 len -= uiop->uio_resid;
727 tlen = nfsm_rndup(len);
728 nfsm_adj(mp3, NFS_MAXPATHLEN-tlen, tlen-len);
729 }
730 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
731 *tl = txdr_unsigned(len);
732 mb->m_next = mp3;
733 mp3 = NULL;
734nfsmout:
735 if (mp3)
736 m_freem(mp3);
737 if (vp)
738 vput(vp);
739 return(error);
740}
741
742/*
743 * nfs read service
744 */
745int
e851b29e
CP
746nfsrv_read(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
747 struct thread *td, struct mbuf **mrq)
984263bc
MD
748{
749 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
750 struct sockaddr *nam = nfsd->nd_nam;
751 caddr_t dpos = nfsd->nd_dpos;
752 struct ucred *cred = &nfsd->nd_cr;
dadab5e9 753 struct iovec *iv;
984263bc 754 struct iovec *iv2;
dadab5e9
MD
755 struct mbuf *m;
756 struct nfs_fattr *fp;
757 u_int32_t *tl;
758 int32_t t1;
759 int i;
984263bc
MD
760 caddr_t bpos;
761 int error = 0, rdonly, cache, cnt, len, left, siz, tlen, getret;
762 int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen;
763 char *cp2;
764 struct mbuf *mb, *mb2, *mreq;
765 struct mbuf *m2;
766 struct vnode *vp = NULL;
767 nfsfh_t nfh;
768 fhandle_t *fhp;
769 struct uio io, *uiop = &io;
770 struct vattr va, *vap = &va;
771 struct nfsheur *nh;
772 off_t off;
773 int ioflag = 0;
774 u_quad_t frev;
775
776 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
777 fhp = &nfh.fh_generic;
778 nfsm_srvmtofh(fhp);
779 if (v3) {
780 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
781 off = fxdr_hyper(tl);
782 } else {
783 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
784 off = (off_t)fxdr_unsigned(u_int32_t, *tl);
785 }
786 nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd));
787
788 /*
789 * Reference vp. If an error occurs, vp will be invalid, but we
790 * have to NULL it just in case. The macros might goto nfsmout
791 * as well.
792 */
793
794 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
795 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
796 if (error) {
797 vp = NULL;
798 nfsm_reply(2 * NFSX_UNSIGNED);
799 nfsm_srvpostop_attr(1, (struct vattr *)0);
800 error = 0;
801 goto nfsmout;
802 }
803
804 if (vp->v_type != VREG) {
805 if (v3)
806 error = EINVAL;
807 else
808 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
809 }
810 if (!error) {
811 nqsrv_getl(vp, ND_READ);
dadab5e9
MD
812 if ((error = nfsrv_access(vp, VREAD, cred, rdonly, td, 1)) != 0)
813 error = nfsrv_access(vp, VEXEC, cred, rdonly, td, 1);
984263bc 814 }
3b568787 815 getret = VOP_GETATTR(vp, vap, td);
984263bc
MD
816 if (!error)
817 error = getret;
818 if (error) {
819 vput(vp);
820 vp = NULL;
821 nfsm_reply(NFSX_POSTOPATTR(v3));
822 nfsm_srvpostop_attr(getret, vap);
823 error = 0;
824 goto nfsmout;
825 }
826
827 /*
828 * Calculate byte count to read
829 */
830
831 if (off >= vap->va_size)
832 cnt = 0;
833 else if ((off + reqlen) > vap->va_size)
834 cnt = vap->va_size - off;
835 else
836 cnt = reqlen;
837
838 /*
839 * Calculate seqcount for heuristic
840 */
841
842 {
843 int hi;
844 int try = 4;
845
846 /*
847 * Locate best candidate
848 */
849
850 hi = ((int)(vm_offset_t)vp / sizeof(struct vnode)) & (NUM_HEURISTIC - 1);
851 nh = &nfsheur[hi];
852
853 while (try--) {
854 if (nfsheur[hi].nh_vp == vp) {
855 nh = &nfsheur[hi];
856 break;
857 }
858 if (nfsheur[hi].nh_use > 0)
859 --nfsheur[hi].nh_use;
860 hi = (hi + 1) & (NUM_HEURISTIC - 1);
861 if (nfsheur[hi].nh_use < nh->nh_use)
862 nh = &nfsheur[hi];
863 }
864
865 if (nh->nh_vp != vp) {
866 nh->nh_vp = vp;
867 nh->nh_nextr = off;
868 nh->nh_use = NHUSE_INIT;
869 if (off == 0)
870 nh->nh_seqcount = 4;
871 else
872 nh->nh_seqcount = 1;
873 }
874
875 /*
876 * Calculate heuristic
877 */
878
879 if ((off == 0 && nh->nh_seqcount > 0) || off == nh->nh_nextr) {
880 if (++nh->nh_seqcount > IO_SEQMAX)
881 nh->nh_seqcount = IO_SEQMAX;
882 } else if (nh->nh_seqcount > 1) {
883 nh->nh_seqcount = 1;
884 } else {
885 nh->nh_seqcount = 0;
886 }
887 nh->nh_use += NHUSE_INC;
888 if (nh->nh_use > NHUSE_MAX)
889 nh->nh_use = NHUSE_MAX;
890 ioflag |= nh->nh_seqcount << IO_SEQSHIFT;
891 }
892
893 nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
894 if (v3) {
895 nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
896 *tl++ = nfs_true;
897 fp = (struct nfs_fattr *)tl;
898 tl += (NFSX_V3FATTR / sizeof (u_int32_t));
899 } else {
900 nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
901 fp = (struct nfs_fattr *)tl;
902 tl += (NFSX_V2FATTR / sizeof (u_int32_t));
903 }
904 len = left = nfsm_rndup(cnt);
905 if (cnt > 0) {
906 /*
907 * Generate the mbuf list with the uio_iov ref. to it.
908 */
909 i = 0;
910 m = m2 = mb;
911 while (left > 0) {
912 siz = min(M_TRAILINGSPACE(m), left);
913 if (siz > 0) {
914 left -= siz;
915 i++;
916 }
917 if (left > 0) {
74f1caca
EN
918 MGET(m, MB_WAIT, MT_DATA);
919 MCLGET(m, MB_WAIT);
984263bc
MD
920 m->m_len = 0;
921 m2->m_next = m;
922 m2 = m;
923 }
924 }
925 MALLOC(iv, struct iovec *, i * sizeof (struct iovec),
926 M_TEMP, M_WAITOK);
927 uiop->uio_iov = iv2 = iv;
928 m = mb;
929 left = len;
930 i = 0;
931 while (left > 0) {
932 if (m == NULL)
933 panic("nfsrv_read iov");
934 siz = min(M_TRAILINGSPACE(m), left);
935 if (siz > 0) {
936 iv->iov_base = mtod(m, caddr_t) + m->m_len;
937 iv->iov_len = siz;
938 m->m_len += siz;
939 left -= siz;
940 iv++;
941 i++;
942 }
943 m = m->m_next;
944 }
945 uiop->uio_iovcnt = i;
946 uiop->uio_offset = off;
947 uiop->uio_resid = len;
948 uiop->uio_rw = UIO_READ;
949 uiop->uio_segflg = UIO_SYSSPACE;
950 error = VOP_READ(vp, uiop, IO_NODELOCKED | ioflag, cred);
951 off = uiop->uio_offset;
952 nh->nh_nextr = off;
953 FREE((caddr_t)iv2, M_TEMP);
3b568787 954 if (error || (getret = VOP_GETATTR(vp, vap, td))) {
984263bc
MD
955 if (!error)
956 error = getret;
957 m_freem(mreq);
958 vput(vp);
959 vp = NULL;
960 nfsm_reply(NFSX_POSTOPATTR(v3));
961 nfsm_srvpostop_attr(getret, vap);
962 error = 0;
963 goto nfsmout;
964 }
965 } else {
966 uiop->uio_resid = 0;
967 }
968 vput(vp);
969 vp = NULL;
970 nfsm_srvfillattr(vap, fp);
971 tlen = len - uiop->uio_resid;
972 cnt = cnt < tlen ? cnt : tlen;
973 tlen = nfsm_rndup(cnt);
974 if (len != tlen || tlen != cnt)
975 nfsm_adj(mb, len - tlen, tlen - cnt);
976 if (v3) {
977 *tl++ = txdr_unsigned(cnt);
978 if (len < reqlen)
979 *tl++ = nfs_true;
980 else
981 *tl++ = nfs_false;
982 }
983 *tl = txdr_unsigned(cnt);
984nfsmout:
985 if (vp)
986 vput(vp);
987 return(error);
988}
989
990/*
991 * nfs write service
992 */
993int
e851b29e
CP
994nfsrv_write(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
995 struct thread *td, struct mbuf **mrq)
984263bc
MD
996{
997 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
998 struct sockaddr *nam = nfsd->nd_nam;
999 caddr_t dpos = nfsd->nd_dpos;
1000 struct ucred *cred = &nfsd->nd_cr;
dadab5e9
MD
1001 struct iovec *ivp;
1002 int i, cnt;
1003 struct mbuf *mp;
1004 struct nfs_fattr *fp;
984263bc
MD
1005 struct iovec *iv;
1006 struct vattr va, forat;
dadab5e9
MD
1007 struct vattr *vap = &va;
1008 u_int32_t *tl;
1009 int32_t t1;
984263bc
MD
1010 caddr_t bpos;
1011 int error = 0, rdonly, cache, len, forat_ret = 1;
1012 int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
1013 int stable = NFSV3WRITE_FILESYNC;
1014 int v3 = (nfsd->nd_flag & ND_NFSV3);
1015 char *cp2;
1016 struct mbuf *mb, *mb2, *mreq;
1017 struct vnode *vp = NULL;
1018 nfsfh_t nfh;
1019 fhandle_t *fhp;
1020 struct uio io, *uiop = &io;
1021 off_t off;
1022 u_quad_t frev;
1023
1024 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
1025 if (mrep == NULL) {
1026 *mrq = NULL;
1027 error = 0;
1028 goto nfsmout;
1029 }
1030 fhp = &nfh.fh_generic;
1031 nfsm_srvmtofh(fhp);
1032 if (v3) {
1033 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1034 off = fxdr_hyper(tl);
1035 tl += 3;
1036 stable = fxdr_unsigned(int, *tl++);
1037 } else {
1038 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1039 off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
1040 tl += 2;
1041 if (nfs_async)
1042 stable = NFSV3WRITE_UNSTABLE;
1043 }
1044 retlen = len = fxdr_unsigned(int32_t, *tl);
1045 cnt = i = 0;
1046
1047 /*
1048 * For NFS Version 2, it is not obvious what a write of zero length
1049 * should do, but I might as well be consistent with Version 3,
1050 * which is to return ok so long as there are no permission problems.
1051 */
1052 if (len > 0) {
1053 zeroing = 1;
1054 mp = mrep;
1055 while (mp) {
1056 if (mp == md) {
1057 zeroing = 0;
1058 adjust = dpos - mtod(mp, caddr_t);
1059 mp->m_len -= adjust;
1060 if (mp->m_len > 0 && adjust > 0)
1061 NFSMADV(mp, adjust);
1062 }
1063 if (zeroing)
1064 mp->m_len = 0;
1065 else if (mp->m_len > 0) {
1066 i += mp->m_len;
1067 if (i > len) {
1068 mp->m_len -= (i - len);
1069 zeroing = 1;
1070 }
1071 if (mp->m_len > 0)
1072 cnt++;
1073 }
1074 mp = mp->m_next;
1075 }
1076 }
1077 if (len > NFS_MAXDATA || len < 0 || i < len) {
1078 error = EIO;
1079 nfsm_reply(2 * NFSX_UNSIGNED);
1080 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap);
1081 error = 0;
1082 goto nfsmout;
1083 }
1084 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
1085 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
1086 if (error) {
1087 vp = NULL;
1088 nfsm_reply(2 * NFSX_UNSIGNED);
1089 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap);
1090 error = 0;
1091 goto nfsmout;
1092 }
1093 if (v3)
3b568787 1094 forat_ret = VOP_GETATTR(vp, &forat, td);
984263bc
MD
1095 if (vp->v_type != VREG) {
1096 if (v3)
1097 error = EINVAL;
1098 else
1099 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1100 }
1101 if (!error) {
1102 nqsrv_getl(vp, ND_WRITE);
dadab5e9 1103 error = nfsrv_access(vp, VWRITE, cred, rdonly, td, 1);
984263bc
MD
1104 }
1105 if (error) {
1106 vput(vp);
1107 vp = NULL;
1108 nfsm_reply(NFSX_WCCDATA(v3));
1109 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap);
1110 error = 0;
1111 goto nfsmout;
1112 }
1113
1114 if (len > 0) {
1115 MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP,
1116 M_WAITOK);
1117 uiop->uio_iov = iv = ivp;
1118 uiop->uio_iovcnt = cnt;
1119 mp = mrep;
1120 while (mp) {
1121 if (mp->m_len > 0) {
1122 ivp->iov_base = mtod(mp, caddr_t);
1123 ivp->iov_len = mp->m_len;
1124 ivp++;
1125 }
1126 mp = mp->m_next;
1127 }
1128
1129 /*
1130 * XXX
1131 * The IO_METASYNC flag indicates that all metadata (and not just
1132 * enough to ensure data integrity) mus be written to stable storage
1133 * synchronously.
1134 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
1135 */
1136 if (stable == NFSV3WRITE_UNSTABLE)
1137 ioflags = IO_NODELOCKED;
1138 else if (stable == NFSV3WRITE_DATASYNC)
1139 ioflags = (IO_SYNC | IO_NODELOCKED);
1140 else
1141 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1142 uiop->uio_resid = len;
1143 uiop->uio_rw = UIO_WRITE;
1144 uiop->uio_segflg = UIO_SYSSPACE;
dadab5e9 1145 uiop->uio_td = NULL;
984263bc
MD
1146 uiop->uio_offset = off;
1147 error = VOP_WRITE(vp, uiop, ioflags, cred);
1148 nfsstats.srvvop_writes++;
1149 FREE((caddr_t)iv, M_TEMP);
1150 }
3b568787 1151 aftat_ret = VOP_GETATTR(vp, vap, td);
984263bc
MD
1152 vput(vp);
1153 vp = NULL;
1154 if (!error)
1155 error = aftat_ret;
1156 nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
1157 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
1158 if (v3) {
1159 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap);
1160 if (error) {
1161 error = 0;
1162 goto nfsmout;
1163 }
1164 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1165 *tl++ = txdr_unsigned(retlen);
1166 /*
1167 * If nfs_async is set, then pretend the write was FILESYNC.
1168 */
1169 if (stable == NFSV3WRITE_UNSTABLE && !nfs_async)
1170 *tl++ = txdr_unsigned(stable);
1171 else
1172 *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
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 if (nfsver.tv_sec == 0)
1179 nfsver = boottime;
1180 *tl++ = txdr_unsigned(nfsver.tv_sec);
3a6f9faf 1181 *tl = txdr_unsigned(nfsver.tv_nsec / 1000);
984263bc
MD
1182 } else {
1183 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1184 nfsm_srvfillattr(vap, fp);
1185 }
1186nfsmout:
1187 if (vp)
1188 vput(vp);
1189 return(error);
1190}
1191
1192/*
1193 * NFS write service with write gathering support. Called when
1194 * nfsrvw_procrastinate > 0.
1195 * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
1196 * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
1197 * Jan. 1994.
1198 */
1199int
e851b29e
CP
1200nfsrv_writegather(struct nfsrv_descript **ndp, struct nfssvc_sock *slp,
1201 struct thread *td, struct mbuf **mrq)
984263bc 1202{
dadab5e9
MD
1203 struct iovec *ivp;
1204 struct mbuf *mp;
1205 struct nfsrv_descript *wp, *nfsd, *owp, *swp;
1206 struct nfs_fattr *fp;
1207 int i;
984263bc
MD
1208 struct iovec *iov;
1209 struct nfsrvw_delayhash *wpp;
1210 struct ucred *cred;
1211 struct vattr va, forat;
dadab5e9
MD
1212 u_int32_t *tl;
1213 int32_t t1;
984263bc
MD
1214 caddr_t bpos, dpos;
1215 int error = 0, rdonly, cache, len, forat_ret = 1;
1216 int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
1217 char *cp2;
1218 struct mbuf *mb, *mb2, *mreq, *mrep, *md;
1219 struct vnode *vp = NULL;
1220 struct uio io, *uiop = &io;
1221 u_quad_t frev, cur_usec;
1222
1223 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
1224#ifndef nolint
1225 i = 0;
1226 len = 0;
1227#endif
1228 *mrq = NULL;
1229 if (*ndp) {
1230 nfsd = *ndp;
1231 *ndp = NULL;
1232 mrep = nfsd->nd_mrep;
1233 md = nfsd->nd_md;
1234 dpos = nfsd->nd_dpos;
1235 cred = &nfsd->nd_cr;
1236 v3 = (nfsd->nd_flag & ND_NFSV3);
1237 LIST_INIT(&nfsd->nd_coalesce);
1238 nfsd->nd_mreq = NULL;
1239 nfsd->nd_stable = NFSV3WRITE_FILESYNC;
1240 cur_usec = nfs_curusec();
1241 nfsd->nd_time = cur_usec +
1242 (v3 ? nfsrvw_procrastinate_v3 : nfsrvw_procrastinate);
1243
1244 /*
1245 * Now, get the write header..
1246 */
1247 nfsm_srvmtofh(&nfsd->nd_fh);
1248 if (v3) {
1249 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1250 nfsd->nd_off = fxdr_hyper(tl);
1251 tl += 3;
1252 nfsd->nd_stable = fxdr_unsigned(int, *tl++);
1253 } else {
1254 nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1255 nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
1256 tl += 2;
1257 if (nfs_async)
1258 nfsd->nd_stable = NFSV3WRITE_UNSTABLE;
1259 }
1260 len = fxdr_unsigned(int32_t, *tl);
1261 nfsd->nd_len = len;
1262 nfsd->nd_eoff = nfsd->nd_off + len;
1263
1264 /*
1265 * Trim the header out of the mbuf list and trim off any trailing
1266 * junk so that the mbuf list has only the write data.
1267 */
1268 zeroing = 1;
1269 i = 0;
1270 mp = mrep;
1271 while (mp) {
1272 if (mp == md) {
1273 zeroing = 0;
1274 adjust = dpos - mtod(mp, caddr_t);
1275 mp->m_len -= adjust;
1276 if (mp->m_len > 0 && adjust > 0)
1277 NFSMADV(mp, adjust);
1278 }
1279 if (zeroing)
1280 mp->m_len = 0;
1281 else {
1282 i += mp->m_len;
1283 if (i > len) {
1284 mp->m_len -= (i - len);
1285 zeroing = 1;
1286 }
1287 }
1288 mp = mp->m_next;
1289 }
1290 if (len > NFS_MAXDATA || len < 0 || i < len) {
1291nfsmout:
1292 m_freem(mrep);
1293 error = EIO;
1294 nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1295 if (v3)
1296 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1297 nfsd->nd_mreq = mreq;
1298 nfsd->nd_mrep = NULL;
1299 nfsd->nd_time = 0;
1300 }
1301
1302 /*
1303 * Add this entry to the hash and time queues.
1304 */
1305 s = splsoftclock();
1306 owp = NULL;
1307 wp = slp->ns_tq.lh_first;
1308 while (wp && wp->nd_time < nfsd->nd_time) {
1309 owp = wp;
1310 wp = wp->nd_tq.le_next;
1311 }
1312 NFS_DPF(WG, ("Q%03x", nfsd->nd_retxid & 0xfff));
1313 if (owp) {
1314 LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1315 } else {
1316 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1317 }
1318 if (nfsd->nd_mrep) {
1319 wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1320 owp = NULL;
1321 wp = wpp->lh_first;
1322 while (wp &&
1323 bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) {
1324 owp = wp;
1325 wp = wp->nd_hash.le_next;
1326 }
1327 while (wp && wp->nd_off < nfsd->nd_off &&
1328 !bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) {
1329 owp = wp;
1330 wp = wp->nd_hash.le_next;
1331 }
1332 if (owp) {
1333 LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1334
1335 /*
1336 * Search the hash list for overlapping entries and
1337 * coalesce.
1338 */
1339 for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1340 wp = nfsd->nd_hash.le_next;
1341 if (NFSW_SAMECRED(owp, nfsd))
1342 nfsrvw_coalesce(owp, nfsd);
1343 }
1344 } else {
1345 LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1346 }
1347 }
1348 splx(s);
1349 }
1350
1351 /*
1352 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1353 * and generate the associated reply mbuf list(s).
1354 */
1355loop1:
1356 cur_usec = nfs_curusec();
1357 s = splsoftclock();
1358 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = owp) {
1359 owp = nfsd->nd_tq.le_next;
1360 if (nfsd->nd_time > cur_usec)
1361 break;
1362 if (nfsd->nd_mreq)
1363 continue;
1364 NFS_DPF(WG, ("P%03x", nfsd->nd_retxid & 0xfff));
1365 LIST_REMOVE(nfsd, nd_tq);
1366 LIST_REMOVE(nfsd, nd_hash);
1367 splx(s);
1368 mrep = nfsd->nd_mrep;
1369 nfsd->nd_mrep = NULL;
1370 cred = &nfsd->nd_cr;
1371 v3 = (nfsd->nd_flag & ND_NFSV3);
1372 forat_ret = aftat_ret = 1;
1373 error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1374 nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
1375 if (!error) {
1376 if (v3)
3b568787 1377 forat_ret = VOP_GETATTR(vp, &forat, td);
984263bc
MD
1378 if (vp->v_type != VREG) {
1379 if (v3)
1380 error = EINVAL;
1381 else
1382 error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1383 }
1384 } else {
1385 vp = NULL;
1386 }
1387 if (!error) {
1388 nqsrv_getl(vp, ND_WRITE);
dadab5e9 1389 error = nfsrv_access(vp, VWRITE, cred, rdonly, td, 1);
984263bc
MD
1390 }
1391
1392 if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1393 ioflags = IO_NODELOCKED;
1394 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1395 ioflags = (IO_SYNC | IO_NODELOCKED);
1396 else
1397 ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1398 uiop->uio_rw = UIO_WRITE;
1399 uiop->uio_segflg = UIO_SYSSPACE;
dadab5e9 1400 uiop->uio_td = NULL;
984263bc
MD
1401 uiop->uio_offset = nfsd->nd_off;
1402 uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1403 if (uiop->uio_resid > 0) {
1404 mp = mrep;
1405 i = 0;
1406 while (mp) {
1407 if (mp->m_len > 0)
1408 i++;
1409 mp = mp->m_next;
1410 }
1411 uiop->uio_iovcnt = i;
1412 MALLOC(iov, struct iovec *, i * sizeof (struct iovec),
1413 M_TEMP, M_WAITOK);
1414 uiop->uio_iov = ivp = iov;
1415 mp = mrep;
1416 while (mp) {
1417 if (mp->m_len > 0) {
1418 ivp->iov_base = mtod(mp, caddr_t);
1419 ivp->iov_len = mp->m_len;
1420 ivp++;
1421 }
1422 mp = mp->m_next;
1423 }
1424 if (!error) {
1425 error = VOP_WRITE(vp, uiop, ioflags, cred);
1426 nfsstats.srvvop_writes++;
1427 }
1428 FREE((caddr_t)iov, M_TEMP);
1429 }
1430 m_freem(mrep);
1431 if (vp) {
3b568787 1432 aftat_ret = VOP_GETATTR(vp, &va, td);
984263bc
MD
1433 vput(vp);
1434 vp = NULL;
1435 }
1436
1437 /*
1438 * Loop around generating replies for all write rpcs that have
1439 * now been completed.
1440 */
1441 swp = nfsd;
1442 do {
1443 NFS_DPF(WG, ("R%03x", nfsd->nd_retxid & 0xfff));
1444 if (error) {
1445 nfsm_writereply(NFSX_WCCDATA(v3), v3);
1446 if (v3) {
1447 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1448 }
1449 } else {
1450 nfsm_writereply(NFSX_PREOPATTR(v3) +
1451 NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1452 NFSX_WRITEVERF(v3), v3);
1453 if (v3) {
1454 nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1455 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1456 *tl++ = txdr_unsigned(nfsd->nd_len);
1457 *tl++ = txdr_unsigned(swp->nd_stable);
1458 /*
1459 * Actually, there is no need to txdr these fields,
1460 * but it may make the values more human readable,
1461 * for debugging purposes.
1462 */
1463 if (nfsver.tv_sec == 0)
1464 nfsver = boottime;
1465 *tl++ = txdr_unsigned(nfsver.tv_sec);
3a6f9faf 1466 *tl = txdr_unsigned(nfsver.tv_nsec / 1000);
984263bc
MD
1467 } else {
1468 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1469 nfsm_srvfillattr(&va, fp);
1470 }
1471 }
1472 nfsd->nd_mreq = mreq;
1473 if (nfsd->nd_mrep)
1474 panic("nfsrv_write: nd_mrep not free");
1475
1476 /*
1477 * Done. Put it at the head of the timer queue so that
1478 * the final phase can return the reply.
1479 */
1480 s = splsoftclock();
1481 if (nfsd != swp) {
1482 nfsd->nd_time = 0;
1483 LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1484 }
1485 nfsd = swp->nd_coalesce.lh_first;
1486 if (nfsd) {
1487 LIST_REMOVE(nfsd, nd_tq);
1488 }
1489 splx(s);
1490 } while (nfsd);
1491 s = splsoftclock();
1492 swp->nd_time = 0;
1493 LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1494 splx(s);
1495 goto loop1;
1496 }
1497 splx(s);
1498
1499 /*
1500 * Search for a reply to return.
1501 */
1502 s = splsoftclock();
1503 for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = nfsd->nd_tq.le_next)
1504 if (nfsd->nd_mreq) {
1505 NFS_DPF(WG, ("X%03x", nfsd->nd_retxid & 0xfff));
1506 LIST_REMOVE(nfsd, nd_tq);
1507 *mrq = nfsd->nd_mreq;
1508 *ndp = nfsd;
1509 break;
1510 }
1511 splx(s);
1512 return (0);
1513}
1514
1515/*
1516 * Coalesce the write request nfsd into owp. To do this we must:
1517 * - remove nfsd from the queues
1518 * - merge nfsd->nd_mrep into owp->nd_mrep
1519 * - update the nd_eoff and nd_stable for owp
1520 * - put nfsd on owp's nd_coalesce list
1521 * NB: Must be called at splsoftclock().
1522 */
1523static void
dadab5e9 1524nfsrvw_coalesce(struct nfsrv_descript *owp, struct nfsrv_descript *nfsd)
984263bc 1525{
dadab5e9
MD
1526 int overlap;
1527 struct mbuf *mp;
984263bc
MD
1528 struct nfsrv_descript *p;
1529
1530 NFS_DPF(WG, ("C%03x-%03x",
1531 nfsd->nd_retxid & 0xfff, owp->nd_retxid & 0xfff));
1532 LIST_REMOVE(nfsd, nd_hash);
1533 LIST_REMOVE(nfsd, nd_tq);
1534 if (owp->nd_eoff < nfsd->nd_eoff) {
1535 overlap = owp->nd_eoff - nfsd->nd_off;
1536 if (overlap < 0)
1537 panic("nfsrv_coalesce: bad off");
1538 if (overlap > 0)
1539 m_adj(nfsd->nd_mrep, overlap);
1540 mp = owp->nd_mrep;
1541 while (mp->m_next)
1542 mp = mp->m_next;
1543 mp->m_next = nfsd->nd_mrep;
1544 owp->nd_eoff = nfsd->nd_eoff;
1545 } else
1546 m_freem(nfsd->nd_mrep);
1547 nfsd->nd_mrep = NULL;
1548 if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1549 owp->nd_stable = NFSV3WRITE_FILESYNC;
1550 else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1551 owp->nd_stable == NFSV3WRITE_UNSTABLE)
1552 owp->nd_stable = NFSV3WRITE_DATASYNC;
1553 LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1554
1555 /*
1556 * If nfsd had anything else coalesced into it, transfer them
1557 * to owp, otherwise their replies will never get sent.
1558 */
1559 for (p = nfsd->nd_coalesce.lh_first; p;
1560 p = nfsd->nd_coalesce.lh_first) {
1561 LIST_REMOVE(p, nd_tq);
1562 LIST_INSERT_HEAD(&owp->nd_coalesce, p, nd_tq);
1563 }
1564}
1565
1566/*
1567 * nfs create service
1568 * now does a truncate to 0 length via. setattr if it already exists
1569 */
1570int
e851b29e
CP
1571nfsrv_create(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
1572 struct thread *td, struct mbuf **mrq)
984263bc
MD
1573{
1574 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1575 struct sockaddr *nam = nfsd->nd_nam;
1576 caddr_t dpos = nfsd->nd_dpos;
1577 struct ucred *cred = &nfsd->nd_cr;
dadab5e9 1578 struct nfs_fattr *fp;
984263bc 1579 struct vattr va, dirfor, diraft;
dadab5e9
MD
1580 struct vattr *vap = &va;
1581 struct nfsv2_sattr *sp;
1582 u_int32_t *tl;
984263bc 1583 struct nameidata nd;
dadab5e9 1584 int32_t t1;
984263bc
MD
1585 caddr_t bpos;
1586 int error = 0, rdev, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1587 int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1588 caddr_t cp;
1589 char *cp2;
1590 struct mbuf *mb, *mb2, *mreq;
1591 struct vnode *dirp = (struct vnode *)0;
1592 nfsfh_t nfh;
1593 fhandle_t *fhp;
1594 u_quad_t frev, tempsize;
1595 u_char cverf[NFSX_V3CREATEVERF];
1596
1597 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
1598#ifndef nolint
1599 rdev = 0;
1600#endif
1601 ndclear(&nd);
1602
1603 fhp = &nfh.fh_generic;
1604 nfsm_srvmtofh(fhp);
1605 nfsm_srvnamesiz(len);
1606
1607 nd.ni_cnd.cn_cred = cred;
2b69e610
MD
1608 nd.ni_cnd.cn_nameiop = NAMEI_CREATE;
1609 nd.ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_LOCKLEAF | CNP_SAVESTART;
984263bc
MD
1610
1611 /*
1612 * Call namei and do initial cleanup to get a few things
1613 * out of the way. If we get an initial error we cleanup
1614 * and return here to avoid special-casing the invalid nd
1615 * structure through the rest of the case. dirp may be
1616 * set even if an error occurs, but the nd structure will not
1617 * be valid at all if an error occurs so we have to invalidate it
1618 * prior to calling nfsm_reply ( which might goto nfsmout ).
1619 */
1620 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
dadab5e9 1621 &dirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc
MD
1622 if (dirp) {
1623 if (v3) {
3b568787 1624 dirfor_ret = VOP_GETATTR(dirp, &dirfor, td);
984263bc
MD
1625 } else {
1626 vrele(dirp);
1627 dirp = NULL;
1628 }
1629 }
1630 if (error) {
1631 nfsm_reply(NFSX_WCCDATA(v3));
1632 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1633 error = 0;
1634 goto nfsmout;
1635 }
1636
1637 /*
1638 * No error. Continue. State:
1639 *
1640 * startdir is valid ( we release this immediately )
1641 * dirp may be valid
1642 * nd.ni_vp may be valid
1643 * nd.ni_dvp is valid
1644 *
1645 * The error state is set through the code and we may also do some
1646 * opportunistic releasing of vnodes to avoid holding locks through
1647 * NFS I/O. The cleanup at the end is a catch-all
1648 */
1649
1650 VATTR_NULL(vap);
1651 if (v3) {
1652 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1653 how = fxdr_unsigned(int, *tl);
1654 switch (how) {
1655 case NFSV3CREATE_GUARDED:
1656 if (nd.ni_vp) {
1657 error = EEXIST;
1658 break;
1659 }
1660 /* fall through */
1661 case NFSV3CREATE_UNCHECKED:
1662 nfsm_srvsattr(vap);
1663 break;
1664 case NFSV3CREATE_EXCLUSIVE:
1665 nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
1666 bcopy(cp, cverf, NFSX_V3CREATEVERF);
1667 exclusive_flag = 1;
1668 break;
1669 };
1670 vap->va_type = VREG;
1671 } else {
1672 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1673 vap->va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1674 if (vap->va_type == VNON)
1675 vap->va_type = VREG;
1676 vap->va_mode = nfstov_mode(sp->sa_mode);
1677 switch (vap->va_type) {
1678 case VREG:
1679 tsize = fxdr_unsigned(int32_t, sp->sa_size);
1680 if (tsize != -1)
1681 vap->va_size = (u_quad_t)tsize;
1682 break;
1683 case VCHR:
1684 case VBLK:
1685 case VFIFO:
1686 rdev = fxdr_unsigned(long, sp->sa_size);
1687 break;
1688 default:
1689 break;
1690 };
1691 }
1692
1693 /*
1694 * Iff doesn't exist, create it
1695 * otherwise just truncate to 0 length
1696 * should I set the mode too ?
1697 *
1698 * The only possible error we can have at this point is EEXIST.
1699 * nd.ni_vp will also be non-NULL in that case.
1700 */
1701 if (nd.ni_vp == NULL) {
1702 if (vap->va_mode == (mode_t)VNOVAL)
1703 vap->va_mode = 0;
1704 if (vap->va_type == VREG || vap->va_type == VSOCK) {
1705 nqsrv_getl(nd.ni_dvp, ND_WRITE);
bc0c094e
MD
1706 error = VOP_CREATE(nd.ni_dvp, NCPNULL, &nd.ni_vp,
1707 &nd.ni_cnd, vap);
984263bc
MD
1708 if (error)
1709 NDFREE(&nd, NDF_ONLY_PNBUF);
1710 else {
1711 nfsrv_object_create(nd.ni_vp);
1712 if (exclusive_flag) {
1713 exclusive_flag = 0;
1714 VATTR_NULL(vap);
1715 bcopy(cverf, (caddr_t)&vap->va_atime,
1716 NFSX_V3CREATEVERF);
1717 error = VOP_SETATTR(nd.ni_vp, vap, cred,
dadab5e9 1718 td);
984263bc
MD
1719 }
1720 }
1721 } else if (
1722 vap->va_type == VCHR ||
1723 vap->va_type == VBLK ||
1724 vap->va_type == VFIFO
1725 ) {
1726 /*
1727 * Handle SysV FIFO node special cases. All other
1728 * devices require super user to access.
1729 */
1730 if (vap->va_type == VCHR && rdev == 0xffffffff)
1731 vap->va_type = VFIFO;
1732 if (vap->va_type != VFIFO &&
dadab5e9 1733 (error = suser_cred(cred, 0))) {
984263bc
MD
1734 goto nfsmreply0;
1735 }
1736 vap->va_rdev = rdev;
1737 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1738
bc0c094e
MD
1739 error = VOP_MKNOD(nd.ni_dvp, NCPNULL, &nd.ni_vp,
1740 &nd.ni_cnd, vap);
984263bc
MD
1741 if (error) {
1742 NDFREE(&nd, NDF_ONLY_PNBUF);
1743 goto nfsmreply0;
1744 }
1745 vput(nd.ni_vp);
1746 nd.ni_vp = NULL;
1747
1748 /*
1749 * release dvp prior to lookup
1750 */
1751 vput(nd.ni_dvp);
1752 nd.ni_dvp = NULL;
1753
1754 /*
1755 * Setup for lookup.
1756 *
1757 * Even though LOCKPARENT was cleared, ni_dvp may
1758 * be garbage.
1759 */
2b69e610
MD
1760 nd.ni_cnd.cn_nameiop = NAMEI_LOOKUP;
1761 nd.ni_cnd.cn_flags &= ~(CNP_LOCKPARENT);
dadab5e9 1762 nd.ni_cnd.cn_td = td;
984263bc
MD
1763 nd.ni_cnd.cn_cred = cred;
1764
1765 error = lookup(&nd);
1766 nd.ni_dvp = NULL;
1767
1768 if (error != 0) {
1769 nfsm_reply(0);
1770 /* fall through on certain errors */
1771 }
1772 nfsrv_object_create(nd.ni_vp);
2b69e610 1773 if (nd.ni_cnd.cn_flags & CNP_ISSYMLINK) {
984263bc
MD
1774 error = EINVAL;
1775 goto nfsmreply0;
1776 }
1777 } else {
1778 error = ENXIO;
1779 }
1780 } else {
1781 if (vap->va_size != -1) {
1782 error = nfsrv_access(nd.ni_vp, VWRITE, cred,
2b69e610 1783 (nd.ni_cnd.cn_flags & CNP_RDONLY), td, 0);
984263bc
MD
1784 if (!error) {
1785 nqsrv_getl(nd.ni_vp, ND_WRITE);
1786 tempsize = vap->va_size;
1787 VATTR_NULL(vap);
1788 vap->va_size = tempsize;
dadab5e9 1789 error = VOP_SETATTR(nd.ni_vp, vap, cred, td);
984263bc
MD
1790 }
1791 }
1792 }
1793
1794 if (!error) {
1795 bzero((caddr_t)fhp, sizeof(nfh));
1796 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
1797 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
1798 if (!error)
3b568787 1799 error = VOP_GETATTR(nd.ni_vp, vap, td);
984263bc
MD
1800 }
1801 if (v3) {
1802 if (exclusive_flag && !error &&
1803 bcmp(cverf, (caddr_t)&vap->va_atime, NFSX_V3CREATEVERF))
1804 error = EEXIST;
3b568787 1805 diraft_ret = VOP_GETATTR(dirp, &diraft, td);
984263bc
MD
1806 vrele(dirp);
1807 dirp = NULL;
1808 }
1809 nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1810 if (v3) {
1811 if (!error) {
1812 nfsm_srvpostop_fh(fhp);
1813 nfsm_srvpostop_attr(0, vap);
1814 }
1815 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1816 error = 0;
1817 } else {
1818 nfsm_srvfhtom(fhp, v3);
1819 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1820 nfsm_srvfillattr(vap, fp);
1821 }
1822 goto nfsmout;
1823
1824nfsmreply0:
1825 nfsm_reply(0);
1826 error = 0;
1827 /* fall through */
1828
1829nfsmout:
1830 if (nd.ni_startdir) {
1831 vrele(nd.ni_startdir);
1832 nd.ni_startdir = NULL;
1833 }
1834 if (dirp)
1835 vrele(dirp);
1836 NDFREE(&nd, NDF_ONLY_PNBUF);
1837 if (nd.ni_dvp) {
1838 if (nd.ni_dvp == nd.ni_vp)
1839 vrele(nd.ni_dvp);
1840 else
1841 vput(nd.ni_dvp);
1842 }
1843 if (nd.ni_vp)
1844 vput(nd.ni_vp);
1845 return (error);
1846}
1847
1848/*
1849 * nfs v3 mknod service
1850 */
1851int
e851b29e
CP
1852nfsrv_mknod(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
1853 struct thread *td, struct mbuf **mrq)
984263bc
MD
1854{
1855 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1856 struct sockaddr *nam = nfsd->nd_nam;
1857 caddr_t dpos = nfsd->nd_dpos;
1858 struct ucred *cred = &nfsd->nd_cr;
1859 struct vattr va, dirfor, diraft;
dadab5e9
MD
1860 struct vattr *vap = &va;
1861 u_int32_t *tl;
984263bc 1862 struct nameidata nd;
dadab5e9 1863 int32_t t1;
984263bc
MD
1864 caddr_t bpos;
1865 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1866 u_int32_t major, minor;
1867 enum vtype vtyp;
1868 char *cp2;
1869 struct mbuf *mb, *mb2, *mreq;
1870 struct vnode *vp, *dirp = (struct vnode *)0;
1871 nfsfh_t nfh;
1872 fhandle_t *fhp;
1873 u_quad_t frev;
1874
1875 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
1876 ndclear(&nd);
1877
1878 fhp = &nfh.fh_generic;
1879 nfsm_srvmtofh(fhp);
1880 nfsm_srvnamesiz(len);
1881
1882 nd.ni_cnd.cn_cred = cred;
2b69e610
MD
1883 nd.ni_cnd.cn_nameiop = NAMEI_CREATE;
1884 nd.ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_LOCKLEAF | CNP_SAVESTART;
984263bc
MD
1885
1886 /*
1887 * Handle nfs_namei() call. If an error occurs, the nd structure
1888 * is not valid. However, nfsm_*() routines may still jump to
1889 * nfsmout.
1890 */
1891
1892 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
dadab5e9 1893 &dirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc 1894 if (dirp)
3b568787 1895 dirfor_ret = VOP_GETATTR(dirp, &dirfor, td);
984263bc
MD
1896 if (error) {
1897 nfsm_reply(NFSX_WCCDATA(1));
1898 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1899 error = 0;
1900 goto nfsmout;
1901 }
1902 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1903 vtyp = nfsv3tov_type(*tl);
1904 if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1905 error = NFSERR_BADTYPE;
1906 goto out;
1907 }
1908 VATTR_NULL(vap);
1909 nfsm_srvsattr(vap);
1910 if (vtyp == VCHR || vtyp == VBLK) {
1911 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1912 major = fxdr_unsigned(u_int32_t, *tl++);
1913 minor = fxdr_unsigned(u_int32_t, *tl);
1914 vap->va_rdev = makeudev(major, minor);
1915 }
1916
1917 /*
1918 * Iff doesn't exist, create it.
1919 */
1920 if (nd.ni_vp) {
1921 error = EEXIST;
1922 goto out;
1923 }
1924 vap->va_type = vtyp;
1925 if (vap->va_mode == (mode_t)VNOVAL)
1926 vap->va_mode = 0;
1927 if (vtyp == VSOCK) {
1928 vrele(nd.ni_startdir);
1929 nd.ni_startdir = NULL;
1930 nqsrv_getl(nd.ni_dvp, ND_WRITE);
bc0c094e
MD
1931 error = VOP_CREATE(nd.ni_dvp, NCPNULL, &nd.ni_vp,
1932 &nd.ni_cnd, vap);
984263bc
MD
1933 if (error)
1934 NDFREE(&nd, NDF_ONLY_PNBUF);
1935 } else {
dadab5e9 1936 if (vtyp != VFIFO && (error = suser_cred(cred, 0)))
984263bc
MD
1937 goto out;
1938 nqsrv_getl(nd.ni_dvp, ND_WRITE);
1939
bc0c094e
MD
1940 error = VOP_MKNOD(nd.ni_dvp, NCPNULL, &nd.ni_vp,
1941 &nd.ni_cnd, vap);
984263bc
MD
1942 if (error) {
1943 NDFREE(&nd, NDF_ONLY_PNBUF);
1944 goto out;
1945 }
1946 vput(nd.ni_vp);
1947 nd.ni_vp = NULL;
1948
1949 /*
1950 * Release dvp prior to lookup
1951 */
1952 vput(nd.ni_dvp);
1953 nd.ni_dvp = NULL;
1954
dadab5e9 1955 KKASSERT(td->td_proc);
2b69e610
MD
1956 nd.ni_cnd.cn_nameiop = NAMEI_LOOKUP;
1957 nd.ni_cnd.cn_flags &= ~(CNP_LOCKPARENT);
dadab5e9
MD
1958 nd.ni_cnd.cn_td = td;
1959 nd.ni_cnd.cn_cred = td->td_proc->p_ucred;
984263bc
MD
1960
1961 error = lookup(&nd);
1962 nd.ni_dvp = NULL;
1963
1964 if (error)
1965 goto out;
2b69e610 1966 if (nd.ni_cnd.cn_flags & CNP_ISSYMLINK)
984263bc
MD
1967 error = EINVAL;
1968 }
1969
1970 /*
1971 * send response, cleanup, return.
1972 */
1973out:
1974 if (nd.ni_startdir) {
1975 vrele(nd.ni_startdir);
1976 nd.ni_startdir = NULL;
1977 }
1978 NDFREE(&nd, NDF_ONLY_PNBUF);
1979 if (nd.ni_dvp) {
1980 if (nd.ni_dvp == nd.ni_vp)
1981 vrele(nd.ni_dvp);
1982 else
1983 vput(nd.ni_dvp);
1984 nd.ni_dvp = NULL;
1985 }
1986 vp = nd.ni_vp;
1987 if (!error) {
1988 bzero((caddr_t)fhp, sizeof(nfh));
1989 fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
1990 error = VFS_VPTOFH(vp, &fhp->fh_fid);
1991 if (!error)
3b568787 1992 error = VOP_GETATTR(vp, vap, td);
984263bc
MD
1993 }
1994 if (vp) {
1995 vput(vp);
1996 vp = NULL;
1997 nd.ni_vp = NULL;
1998 }
3b568787 1999 diraft_ret = VOP_GETATTR(dirp, &diraft, td);
984263bc
MD
2000 if (dirp) {
2001 vrele(dirp);
2002 dirp = NULL;
2003 }
2004 nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
2005 if (!error) {
2006 nfsm_srvpostop_fh(fhp);
2007 nfsm_srvpostop_attr(0, vap);
2008 }
2009 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2010 return (0);
2011nfsmout:
2012 if (dirp)
2013 vrele(dirp);
2014 if (nd.ni_startdir)
2015 vrele(nd.ni_startdir);
2016 NDFREE(&nd, NDF_ONLY_PNBUF);
2017 if (nd.ni_dvp) {
2018 if (nd.ni_dvp == nd.ni_vp)
2019 vrele(nd.ni_dvp);
2020 else
2021 vput(nd.ni_dvp);
2022 }
2023 if (nd.ni_vp)
2024 vput(nd.ni_vp);
2025 return (error);
2026}
2027
2028/*
2029 * nfs remove service
2030 */
2031int
e851b29e
CP
2032nfsrv_remove(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
2033 struct thread *td, struct mbuf **mrq)
984263bc
MD
2034{
2035 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2036 struct sockaddr *nam = nfsd->nd_nam;
2037 caddr_t dpos = nfsd->nd_dpos;
2038 struct ucred *cred = &nfsd->nd_cr;
2039 struct nameidata nd;
dadab5e9
MD
2040 u_int32_t *tl;
2041 int32_t t1;
984263bc
MD
2042 caddr_t bpos;
2043 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2044 int v3 = (nfsd->nd_flag & ND_NFSV3);
2045 char *cp2;
2046 struct mbuf *mb, *mreq;
2047 struct vnode *dirp;
2048 struct vattr dirfor, diraft;
2049 nfsfh_t nfh;
2050 fhandle_t *fhp;
2051 u_quad_t frev;
2052
2053 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
2054 ndclear(&nd);
2055
2056 fhp = &nfh.fh_generic;
2057 nfsm_srvmtofh(fhp);
2058 nfsm_srvnamesiz(len);
2059
2060 nd.ni_cnd.cn_cred = cred;
2b69e610
MD
2061 nd.ni_cnd.cn_nameiop = NAMEI_DELETE;
2062 nd.ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_LOCKLEAF;
984263bc 2063 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
dadab5e9 2064 &dirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc
MD
2065 if (dirp) {
2066 if (v3) {
3b568787 2067 dirfor_ret = VOP_GETATTR(dirp, &dirfor, td);
984263bc
MD
2068 } else {
2069 vrele(dirp);
2070 dirp = NULL;
2071 }
2072 }
2073 if (error == 0) {
2074 if (nd.ni_vp->v_type == VDIR) {
2075 error = EPERM; /* POSIX */
2076 goto out;
2077 }
2078 /*
2079 * The root of a mounted filesystem cannot be deleted.
2080 */
2081 if (nd.ni_vp->v_flag & VROOT) {
2082 error = EBUSY;
2083 goto out;
2084 }
2085out:
2086 if (!error) {
2087 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2088 nqsrv_getl(nd.ni_vp, ND_WRITE);
bc0c094e 2089 error = VOP_REMOVE(nd.ni_dvp, NCPNULL, nd.ni_vp, &nd.ni_cnd);
984263bc
MD
2090 NDFREE(&nd, NDF_ONLY_PNBUF);
2091 }
2092 }
2093 if (dirp && v3) {
3b568787 2094 diraft_ret = VOP_GETATTR(dirp, &diraft, td);
984263bc
MD
2095 vrele(dirp);
2096 dirp = NULL;
2097 }
2098 nfsm_reply(NFSX_WCCDATA(v3));
2099 if (v3) {
2100 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2101 error = 0;
2102 }
2103nfsmout:
2104 NDFREE(&nd, NDF_ONLY_PNBUF);
2105 if (nd.ni_dvp) {
2106 if (nd.ni_dvp == nd.ni_vp)
2107 vrele(nd.ni_dvp);
2108 else
2109 vput(nd.ni_dvp);
2110 }
2111 if (nd.ni_vp)
2112 vput(nd.ni_vp);
2113 return(error);
2114}
2115
2116/*
2117 * nfs rename service
2118 */
2119int
e851b29e
CP
2120nfsrv_rename(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
2121 struct thread *td, struct mbuf **mrq)
984263bc
MD
2122{
2123 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2124 struct sockaddr *nam = nfsd->nd_nam;
2125 caddr_t dpos = nfsd->nd_dpos;
2126 struct ucred *cred = &nfsd->nd_cr;
dadab5e9
MD
2127 u_int32_t *tl;
2128 int32_t t1;
984263bc
MD
2129 caddr_t bpos;
2130 int error = 0, cache, len, len2, fdirfor_ret = 1, fdiraft_ret = 1;
2131 int tdirfor_ret = 1, tdiraft_ret = 1;
2132 int v3 = (nfsd->nd_flag & ND_NFSV3);
2133 char *cp2;
2134 struct mbuf *mb, *mreq;
2135 struct nameidata fromnd, tond;
2136 struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
2137 struct vnode *tdirp = (struct vnode *)0;
2138 struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
2139 nfsfh_t fnfh, tnfh;
2140 fhandle_t *ffhp, *tfhp;
2141 u_quad_t frev;
2142 uid_t saved_uid;
2143
2144 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
2145#ifndef nolint
2146 fvp = (struct vnode *)0;
2147#endif
2148 ffhp = &fnfh.fh_generic;
2149 tfhp = &tnfh.fh_generic;
2150
2151 /*
2152 * Clear fields incase goto nfsmout occurs from macro.
2153 */
2154
2155 ndclear(&fromnd);
2156 ndclear(&tond);
2157
2158 nfsm_srvmtofh(ffhp);
2159 nfsm_srvnamesiz(len);
2160 /*
2161 * Remember our original uid so that we can reset cr_uid before
2162 * the second nfs_namei() call, in case it is remapped.
2163 */
2164 saved_uid = cred->cr_uid;
2165 fromnd.ni_cnd.cn_cred = cred;
2b69e610
MD
2166 fromnd.ni_cnd.cn_nameiop = NAMEI_DELETE;
2167 fromnd.ni_cnd.cn_flags = CNP_WANTPARENT | CNP_SAVESTART;
984263bc 2168 error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
dadab5e9 2169 &dpos, &fdirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc
MD
2170 if (fdirp) {
2171 if (v3) {
3b568787 2172 fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, td);
984263bc
MD
2173 } else {
2174 vrele(fdirp);
2175 fdirp = NULL;
2176 }
2177 }
2178 if (error) {
2179 nfsm_reply(2 * NFSX_WCCDATA(v3));
2180 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
2181 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
2182 error = 0;
2183 goto nfsmout;
2184 }
2185 fvp = fromnd.ni_vp;
2186 nfsm_srvmtofh(tfhp);
2187 nfsm_strsiz(len2, NFS_MAXNAMLEN);
2188 cred->cr_uid = saved_uid;
2189 tond.ni_cnd.cn_cred = cred;
2b69e610
MD
2190 tond.ni_cnd.cn_nameiop = NAMEI_RENAME;
2191 tond.ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_LOCKLEAF |
2192 CNP_NOCACHE | CNP_SAVESTART;
984263bc 2193 error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
dadab5e9 2194 &dpos, &tdirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc
MD
2195 if (tdirp) {
2196 if (v3) {
3b568787 2197 tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, td);
984263bc
MD
2198 } else {
2199 vrele(tdirp);
2200 tdirp = NULL;
2201 }
2202 }
2203 if (error)
2204 goto out1;
2205
2206 tdvp = tond.ni_dvp;
2207 tvp = tond.ni_vp;
2208 if (tvp != NULL) {
2209 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
2210 if (v3)
2211 error = EEXIST;
2212 else
2213 error = EISDIR;
2214 goto out;
2215 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
2216 if (v3)
2217 error = EEXIST;
2218 else
2219 error = ENOTDIR;
2220 goto out;
2221 }
2222 if (tvp->v_type == VDIR && tvp->v_mountedhere) {
2223 if (v3)
2224 error = EXDEV;
2225 else
2226 error = ENOTEMPTY;
2227 goto out;
2228 }
2229 }
2230 if (fvp->v_type == VDIR && fvp->v_mountedhere) {
2231 if (v3)
2232 error = EXDEV;
2233 else
2234 error = ENOTEMPTY;
2235 goto out;
2236 }
2237 if (fvp->v_mount != tdvp->v_mount) {
2238 if (v3)
2239 error = EXDEV;
2240 else
2241 error = ENOTEMPTY;
2242 goto out;
2243 }
2244 if (fvp == tdvp) {
2245 if (v3)
2246 error = EINVAL;
2247 else
2248 error = ENOTEMPTY;
2249 }
2250 /*
2251 * If source is the same as the destination (that is the
2252 * same vnode with the same name in the same directory),
2253 * then there is nothing to do.
2254 */
2255 if (fvp == tvp && fromnd.ni_dvp == tdvp &&
2256 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2257 !bcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
2258 fromnd.ni_cnd.cn_namelen))
2259 error = -1;
2260out:
2261 if (!error) {
2262 /*
2263 * The VOP_RENAME function releases all vnode references &
2264 * locks prior to returning so we need to clear the pointers
2265 * to bypass cleanup code later on.
2266 */
2267 nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
2268 nqsrv_getl(tdvp, ND_WRITE);
2269 if (tvp) {
2270 nqsrv_getl(tvp, ND_WRITE);
2271 }
bc0c094e
MD
2272 error = VOP_RENAME(fromnd.ni_dvp, NCPNULL, fromnd.ni_vp, &fromnd.ni_cnd,
2273 tond.ni_dvp, NCPNULL, tond.ni_vp, &tond.ni_cnd);
984263bc
MD
2274 fromnd.ni_dvp = NULL;
2275 fromnd.ni_vp = NULL;
2276 tond.ni_dvp = NULL;
2277 tond.ni_vp = NULL;
2278 if (error) {
2b69e610
MD
2279 fromnd.ni_cnd.cn_flags &= ~CNP_HASBUF;
2280 tond.ni_cnd.cn_flags &= ~CNP_HASBUF;
984263bc
MD
2281 }
2282 } else {
2283 if (error == -1)
2284 error = 0;
2285 }
2286 /* fall through */
2287
2288out1:
2289 if (fdirp)
3b568787 2290 fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, td);
984263bc 2291 if (tdirp)
3b568787 2292 tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, td);
984263bc
MD
2293 nfsm_reply(2 * NFSX_WCCDATA(v3));
2294 if (v3) {
2295 nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
2296 nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
2297 }
2298 error = 0;
2299 /* fall through */
2300
2301nfsmout:
2302 /*
2303 * Clear out tond related fields
2304 */
2305 if (tdirp)
2306 vrele(tdirp);
2307 if (tond.ni_startdir)
2308 vrele(tond.ni_startdir);
2309 NDFREE(&tond, NDF_ONLY_PNBUF);
2310 if (tond.ni_dvp) {
2311 if (tond.ni_dvp == tond.ni_vp)
2312 vrele(tond.ni_dvp);
2313 else
2314 vput(tond.ni_dvp);
2315 }
2316 if (tond.ni_vp)
2317 vput(tond.ni_vp);
2318
2319 /*
2320 * Clear out fromnd related fields
2321 */
2322 if (fdirp)
2323 vrele(fdirp);
2324 if (fromnd.ni_startdir)
2325 vrele(fromnd.ni_startdir);
2326 NDFREE(&fromnd, NDF_ONLY_PNBUF);
2327 if (fromnd.ni_dvp)
2328 vrele(fromnd.ni_dvp);
2329 if (fromnd.ni_vp)
2330 vrele(fromnd.ni_vp);
2331
2332 return (error);
2333}
2334
2335/*
2336 * nfs link service
2337 */
2338int
e851b29e
CP
2339nfsrv_link(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
2340 struct thread *td, struct mbuf **mrq)
984263bc
MD
2341{
2342 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2343 struct sockaddr *nam = nfsd->nd_nam;
2344 caddr_t dpos = nfsd->nd_dpos;
2345 struct ucred *cred = &nfsd->nd_cr;
2346 struct nameidata nd;
dadab5e9
MD
2347 u_int32_t *tl;
2348 int32_t t1;
984263bc
MD
2349 caddr_t bpos;
2350 int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1;
2351 int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
2352 char *cp2;
2353 struct mbuf *mb, *mreq;
2354 struct vnode *vp = NULL, *xp, *dirp = (struct vnode *)0;
2355 struct vattr dirfor, diraft, at;
2356 nfsfh_t nfh, dnfh;
2357 fhandle_t *fhp, *dfhp;
2358 u_quad_t frev;
2359
2360 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
2361 ndclear(&nd);
2362
2363 fhp = &nfh.fh_generic;
2364 dfhp = &dnfh.fh_generic;
2365 nfsm_srvmtofh(fhp);
2366 nfsm_srvmtofh(dfhp);
2367 nfsm_srvnamesiz(len);
2368
2369 error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
2370 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
2371 if (error) {
2372 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2373 nfsm_srvpostop_attr(getret, &at);
2374 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2375 vp = NULL;
2376 error = 0;
2377 goto nfsmout;
2378 }
2379 if (vp->v_type == VDIR) {
2380 error = EPERM; /* POSIX */
2381 goto out1;
2382 }
2383 nd.ni_cnd.cn_cred = cred;
2b69e610
MD
2384 nd.ni_cnd.cn_nameiop = NAMEI_CREATE;
2385 nd.ni_cnd.cn_flags = CNP_LOCKPARENT;
984263bc 2386 error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
dadab5e9 2387 &dirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc
MD
2388 if (dirp) {
2389 if (v3) {
3b568787 2390 dirfor_ret = VOP_GETATTR(dirp, &dirfor, td);
984263bc
MD
2391 } else {
2392 vrele(dirp);
2393 dirp = NULL;
2394 }
2395 }
2396 if (error)
2397 goto out1;
2398
2399 xp = nd.ni_vp;
2400 if (xp != NULL) {
2401 error = EEXIST;
2402 goto out;
2403 }
2404 xp = nd.ni_dvp;
2405 if (vp->v_mount != xp->v_mount)
2406 error = EXDEV;
2407out:
2408 if (!error) {
2409 nqsrv_getl(vp, ND_WRITE);
2410 nqsrv_getl(xp, ND_WRITE);
bc0c094e 2411 error = VOP_LINK(nd.ni_dvp, NCPNULL, vp, &nd.ni_cnd);
984263bc
MD
2412 NDFREE(&nd, NDF_ONLY_PNBUF);
2413 }
2414 /* fall through */
2415
2416out1:
2417 if (v3)
3b568787 2418 getret = VOP_GETATTR(vp, &at, td);
984263bc 2419 if (dirp)
3b568787 2420 diraft_ret = VOP_GETATTR(dirp, &diraft, td);
984263bc
MD
2421 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2422 if (v3) {
2423 nfsm_srvpostop_attr(getret, &at);
2424 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2425 error = 0;
2426 }
2427 /* fall through */
2428
2429nfsmout:
2430 NDFREE(&nd, NDF_ONLY_PNBUF);
2431 if (dirp)
2432 vrele(dirp);
2433 if (vp)
2434 vrele(vp);
2435 if (nd.ni_dvp) {
2436 if (nd.ni_dvp == nd.ni_vp)
2437 vrele(nd.ni_dvp);
2438 else
2439 vput(nd.ni_dvp);
2440 }
2441 if (nd.ni_vp)
2442 vrele(nd.ni_vp);
2443 return(error);
2444}
2445
2446/*
2447 * nfs symbolic link service
2448 */
2449int
e851b29e
CP
2450nfsrv_symlink(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
2451 struct thread *td, struct mbuf **mrq)
984263bc
MD
2452{
2453 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2454 struct sockaddr *nam = nfsd->nd_nam;
2455 caddr_t dpos = nfsd->nd_dpos;
2456 struct ucred *cred = &nfsd->nd_cr;
2457 struct vattr va, dirfor, diraft;
2458 struct nameidata nd;
dadab5e9
MD
2459 struct vattr *vap = &va;
2460 u_int32_t *tl;
2461 int32_t t1;
984263bc
MD
2462 struct nfsv2_sattr *sp;
2463 char *bpos, *pathcp = (char *)0, *cp2;
2464 struct uio io;
2465 struct iovec iv;
2466 int error = 0, cache, len, len2, dirfor_ret = 1, diraft_ret = 1;
2467 int v3 = (nfsd->nd_flag & ND_NFSV3);
2468 struct mbuf *mb, *mreq, *mb2;
2469 struct vnode *dirp = (struct vnode *)0;
2470 nfsfh_t nfh;
2471 fhandle_t *fhp;
2472 u_quad_t frev;
2473
2474 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
2475 ndclear(&nd);
2476
2477 fhp = &nfh.fh_generic;
2478 nfsm_srvmtofh(fhp);
2479 nfsm_srvnamesiz(len);
2480 nd.ni_cnd.cn_cred = cred;
2b69e610
MD
2481 nd.ni_cnd.cn_nameiop = NAMEI_CREATE;
2482 nd.ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_SAVESTART;
984263bc 2483 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
dadab5e9 2484 &dirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc
MD
2485 if (dirp) {
2486 if (v3) {
3b568787 2487 dirfor_ret = VOP_GETATTR(dirp, &dirfor, td);
984263bc
MD
2488 } else {
2489 vrele(dirp);
2490 dirp = NULL;
2491 }
2492 }
2493 if (error)
2494 goto out;
2495
2496 VATTR_NULL(vap);
2497 if (v3)
2498 nfsm_srvsattr(vap);
2499 nfsm_strsiz(len2, NFS_MAXPATHLEN);
2500 MALLOC(pathcp, caddr_t, len2 + 1, M_TEMP, M_WAITOK);
2501 iv.iov_base = pathcp;
2502 iv.iov_len = len2;
2503 io.uio_resid = len2;
2504 io.uio_offset = 0;
2505 io.uio_iov = &iv;
2506 io.uio_iovcnt = 1;
2507 io.uio_segflg = UIO_SYSSPACE;
2508 io.uio_rw = UIO_READ;
dadab5e9 2509 io.uio_td = NULL;
984263bc
MD
2510 nfsm_mtouio(&io, len2);
2511 if (!v3) {
2512 nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2513 vap->va_mode = nfstov_mode(sp->sa_mode);
2514 }
2515 *(pathcp + len2) = '\0';
2516 if (nd.ni_vp) {
2517 error = EEXIST;
2518 goto out;
2519 }
2520
2521 /*
2522 * issue symlink op. SAVESTART is set so the underlying path component
2523 * is only freed by the VOP if an error occurs.
2524 */
2525 if (vap->va_mode == (mode_t)VNOVAL)
2526 vap->va_mode = 0;
2527 nqsrv_getl(nd.ni_dvp, ND_WRITE);
bc0c094e 2528 error = VOP_SYMLINK(nd.ni_dvp, NCPNULL, &nd.ni_vp, &nd.ni_cnd, vap, pathcp);
984263bc
MD
2529 if (error)
2530 NDFREE(&nd, NDF_ONLY_PNBUF);
2531 else
2532 vput(nd.ni_vp);
2533 nd.ni_vp = NULL;
2534 /*
2535 * releases directory prior to potential lookup op.
2536 */
2537 vput(nd.ni_dvp);
2538 nd.ni_dvp = NULL;
2539
2540 if (error == 0) {
2541 if (v3) {
2542 /*
2543 * Issue lookup. Leave SAVESTART set so we can easily free
2544 * the name buffer later on.
2545 *
2546 * since LOCKPARENT is not set, ni_dvp will be garbage on
2547 * return whether an error occurs or not.
2548 */
2b69e610
MD
2549 nd.ni_cnd.cn_nameiop = NAMEI_LOOKUP;
2550 nd.ni_cnd.cn_flags &= ~(CNP_LOCKPARENT | CNP_FOLLOW);
6fc0e645 2551 nd.ni_cnd.cn_flags |= CNP_LOCKLEAF;
dadab5e9 2552 nd.ni_cnd.cn_td = td;
984263bc
MD
2553 nd.ni_cnd.cn_cred = cred;
2554
2555 error = lookup(&nd);
2556 nd.ni_dvp = NULL;
2557
2558 if (error == 0) {
2559 bzero((caddr_t)fhp, sizeof(nfh));
2560 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
2561 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2562 if (!error)
3b568787 2563 error = VOP_GETATTR(nd.ni_vp, vap, td);
984263bc
MD
2564 vput(nd.ni_vp);
2565 nd.ni_vp = NULL;
2566 }
2567 }
2568 }
2569out:
2570 /*
2571 * These releases aren't strictly required, does even doing them
2572 * make any sense? XXX can nfsm_reply() block?
2573 */
2574 if (pathcp) {
2575 FREE(pathcp, M_TEMP);
2576 pathcp = NULL;
2577 }
2578 if (dirp) {
3b568787 2579 diraft_ret = VOP_GETATTR(dirp, &diraft, td);
984263bc
MD
2580 vrele(dirp);
2581 dirp = NULL;
2582 }
2583 if (nd.ni_startdir) {
2584 vrele(nd.ni_startdir);
2585 nd.ni_startdir = NULL;
2586 }
2587 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2588 if (v3) {
2589 if (!error) {
2590 nfsm_srvpostop_fh(fhp);
2591 nfsm_srvpostop_attr(0, vap);
2592 }
2593 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2594 }
2595 error = 0;
2596 /* fall through */
2597
2598nfsmout:
2599 NDFREE(&nd, NDF_ONLY_PNBUF);
2600 if (nd.ni_dvp) {
2601 if (nd.ni_dvp == nd.ni_vp)
2602 vrele(nd.ni_dvp);
2603 else
2604 vput(nd.ni_dvp);
2605 }
2606 if (nd.ni_vp)
2607 vrele(nd.ni_vp);
2608 if (nd.ni_startdir)
2609 vrele(nd.ni_startdir);
2610 if (dirp)
2611 vrele(dirp);
2612 if (pathcp)
2613 FREE(pathcp, M_TEMP);
2614
2615 return (error);
2616}
2617
2618/*
2619 * nfs mkdir service
2620 */
2621int
e851b29e
CP
2622nfsrv_mkdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
2623 struct thread *td, struct mbuf **mrq)
984263bc
MD
2624{
2625 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2626 struct sockaddr *nam = nfsd->nd_nam;
2627 caddr_t dpos = nfsd->nd_dpos;
2628 struct ucred *cred = &nfsd->nd_cr;
2629 struct vattr va, dirfor, diraft;
dadab5e9
MD
2630 struct vattr *vap = &va;
2631 struct nfs_fattr *fp;
984263bc 2632 struct nameidata nd;
dadab5e9
MD
2633 caddr_t cp;
2634 u_int32_t *tl;
2635 int32_t t1;
984263bc
MD
2636 caddr_t bpos;
2637 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2638 int v3 = (nfsd->nd_flag & ND_NFSV3);
2639 char *cp2;
2640 struct mbuf *mb, *mb2, *mreq;
2641 struct vnode *dirp = NULL;
2642 int vpexcl = 0;
2643 nfsfh_t nfh;
2644 fhandle_t *fhp;
2645 u_quad_t frev;
2646
2647 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
2648 ndclear(&nd);
2649
2650 fhp = &nfh.fh_generic;
2651 nfsm_srvmtofh(fhp);
2652 nfsm_srvnamesiz(len);
2653 nd.ni_cnd.cn_cred = cred;
2b69e610
MD
2654 nd.ni_cnd.cn_nameiop = NAMEI_CREATE;
2655 nd.ni_cnd.cn_flags = CNP_LOCKPARENT;
984263bc
MD
2656
2657 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
dadab5e9 2658 &dirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc
MD
2659 if (dirp) {
2660 if (v3) {
3b568787 2661 dirfor_ret = VOP_GETATTR(dirp, &dirfor, td);
984263bc
MD
2662 } else {
2663 vrele(dirp);
2664 dirp = NULL;
2665 }
2666 }
2667 if (error) {
2668 nfsm_reply(NFSX_WCCDATA(v3));
2669 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2670 error = 0;
2671 goto nfsmout;
2672 }
2673 VATTR_NULL(vap);
2674 if (v3) {
2675 nfsm_srvsattr(vap);
2676 } else {
2677 nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2678 vap->va_mode = nfstov_mode(*tl++);
2679 }
2680
2681 /*
2682 * At this point nd.ni_dvp is referenced and exclusively locked and
2683 * nd.ni_vp, if it exists, is referenced but not locked.
2684 */
2685
2686 vap->va_type = VDIR;
2687 if (nd.ni_vp != NULL) {
2688 NDFREE(&nd, NDF_ONLY_PNBUF);
2689 error = EEXIST;
2690 goto out;
2691 }
2692
2693 /*
2694 * Issue mkdir op. Since SAVESTART is not set, the pathname
2695 * component is freed by the VOP call. This will fill-in
2696 * nd.ni_vp, reference, and exclusively lock it.
2697 */
2698 if (vap->va_mode == (mode_t)VNOVAL)
2699 vap->va_mode = 0;
2700 nqsrv_getl(nd.ni_dvp, ND_WRITE);
bc0c094e 2701 error = VOP_MKDIR(nd.ni_dvp, NCPNULL, &nd.ni_vp, &nd.ni_cnd, vap);
984263bc
MD
2702 NDFREE(&nd, NDF_ONLY_PNBUF);
2703 vpexcl = 1;
2704
2705 vput(nd.ni_dvp);
2706 nd.ni_dvp = NULL;
2707
2708 if (!error) {
2709 bzero((caddr_t)fhp, sizeof(nfh));
2710 fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid;
2711 error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2712 if (!error)
3b568787 2713 error = VOP_GETATTR(nd.ni_vp, vap, td);
984263bc
MD
2714 }
2715out:
2716 if (dirp)
3b568787 2717 diraft_ret = VOP_GETATTR(dirp, &diraft, td);
984263bc
MD
2718 nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2719 if (v3) {
2720 if (!error) {
2721 nfsm_srvpostop_fh(fhp);
2722 nfsm_srvpostop_attr(0, vap);
2723 }
2724 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2725 } else {
2726 nfsm_srvfhtom(fhp, v3);
2727 nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2728 nfsm_srvfillattr(vap, fp);
2729 }
2730 error = 0;
2731 /* fall through */
2732
2733nfsmout:
2734 if (dirp)
2735 vrele(dirp);
2736 if (nd.ni_dvp) {
2737 NDFREE(&nd, NDF_ONLY_PNBUF);
2738 if (nd.ni_dvp == nd.ni_vp && vpexcl)
2739 vrele(nd.ni_dvp);
2740 else
2741 vput(nd.ni_dvp);
2742 }
2743 if (nd.ni_vp) {
2744 if (vpexcl)
2745 vput(nd.ni_vp);
2746 else
2747 vrele(nd.ni_vp);
2748 }
2749 return (error);
2750}
2751
2752/*
2753 * nfs rmdir service
2754 */
2755int
e851b29e
CP
2756nfsrv_rmdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
2757 struct thread *td, struct mbuf **mrq)
984263bc
MD
2758{
2759 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2760 struct sockaddr *nam = nfsd->nd_nam;
2761 caddr_t dpos = nfsd->nd_dpos;
2762 struct ucred *cred = &nfsd->nd_cr;
dadab5e9
MD
2763 u_int32_t *tl;
2764 int32_t t1;
984263bc
MD
2765 caddr_t bpos;
2766 int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2767 int v3 = (nfsd->nd_flag & ND_NFSV3);
2768 char *cp2;
2769 struct mbuf *mb, *mreq;
2770 struct vnode *vp, *dirp = (struct vnode *)0;
2771 struct vattr dirfor, diraft;
2772 nfsfh_t nfh;
2773 fhandle_t *fhp;
2774 struct nameidata nd;
2775 u_quad_t frev;
2776
2777 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
2778 ndclear(&nd);
2779
2780 fhp = &nfh.fh_generic;
2781 nfsm_srvmtofh(fhp);
2782 nfsm_srvnamesiz(len);
2783 nd.ni_cnd.cn_cred = cred;
2b69e610
MD
2784 nd.ni_cnd.cn_nameiop = NAMEI_DELETE;
2785 nd.ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_LOCKLEAF;
984263bc 2786 error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
dadab5e9 2787 &dirp, td, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
984263bc
MD
2788 if (dirp) {
2789 if (v3) {
3b568787 2790 dirfor_ret = VOP_GETATTR(dirp, &dirfor, td);
984263bc
MD
2791 } else {
2792 vrele(dirp);
2793 dirp = NULL;
2794 }
2795 }
2796 if (error) {
2797 nfsm_reply(NFSX_WCCDATA(v3));
2798 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2799 error = 0;
2800 goto nfsmout;
2801 }
2802 vp = nd.ni_vp;
2803 if (vp->v_type != VDIR) {
2804 error = ENOTDIR;
2805 goto out;
2806 }
2807 /*
2808 * No rmdir "." please.
2809 */
2810 if (nd.ni_dvp == vp) {
2811 error = EINVAL;
2812 goto out;
2813 }
2814 /*
2815 * The root of a mounted filesystem cannot be deleted.
2816 */
2817 if (vp->v_flag & VROOT)
2818 error = EBUSY;
2819out:
2820 /*
2821 * Issue or abort op. Since SAVESTART is not set, path name
2822 * component is freed by the VOP after either.
2823 */
2824 if (!error) {
2825 nqsrv_getl(nd.ni_dvp, ND_WRITE);
2826 nqsrv_getl(vp, ND_WRITE);
bc0c094e 2827 error = VOP_RMDIR(nd.ni_dvp, NCPNULL, nd.ni_vp, &nd.ni_cnd);
984263bc
MD
2828 }
2829 NDFREE(&nd, NDF_ONLY_PNBUF);
2830
2831 if (dirp)
3b568787 2832 diraft_ret = VOP_GETATTR(dirp, &diraft, td);
984263bc
MD
2833 nfsm_reply(NFSX_WCCDATA(v3));
2834 if (v3) {
2835 nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2836 error = 0;
2837 }
2838 /* fall through */
2839
2840nfsmout:
2841 NDFREE(&nd, NDF_ONLY_PNBUF);
2842 if (dirp)
2843 vrele(dirp);
2844 if (nd.ni_dvp) {
2845 if (nd.ni_dvp == nd.ni_vp)
2846 vrele(nd.ni_dvp);
2847 else
2848 vput(nd.ni_dvp);
2849 }
2850 if (nd.ni_vp)
2851 vput(nd.ni_vp);
2852
2853 return(error);
2854}
2855
2856/*
2857 * nfs readdir service
2858 * - mallocs what it thinks is enough to read
2859 * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR
2860 * - calls VOP_READDIR()
2861 * - loops around building the reply
2862 * if the output generated exceeds count break out of loop
2863 * The nfsm_clget macro is used here so that the reply will be packed
2864 * tightly in mbuf clusters.
2865 * - it only knows that it has encountered eof when the VOP_READDIR()
2866 * reads nothing
2867 * - as such one readdir rpc will return eof false although you are there
2868 * and then the next will return eof
2869 * - it trims out records with d_fileno == 0
2870 * this doesn't matter for Unix clients, but they might confuse clients
2871 * for other os'.
2872 * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2873 * than requested, but this may not apply to all filesystems. For
2874 * example, client NFS does not { although it is never remote mounted
2875 * anyhow }
2876 * The alternate call nfsrv_readdirplus() does lookups as well.
2877 * PS: The NFS protocol spec. does not clarify what the "count" byte
2878 * argument is a count of.. just name strings and file id's or the
2879 * entire reply rpc or ...
2880 * I tried just file name and id sizes and it confused the Sun client,
2881 * so I am using the full rpc size now. The "paranoia.." comment refers
2882 * to including the status longwords that are not a part of the dir.
2883 * "entry" structures, but are in the rpc.
2884 */
2885struct flrep {
2886 nfsuint64 fl_off;
2887 u_int32_t fl_postopok;
2888 u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2889 u_int32_t fl_fhok;
2890 u_int32_t fl_fhsize;
2891 u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2892};
2893
2894int
e851b29e
CP
2895nfsrv_readdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
2896 struct thread *td, struct mbuf **mrq)
984263bc
MD
2897{
2898 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2899 struct sockaddr *nam = nfsd->nd_nam;
2900 caddr_t dpos = nfsd->nd_dpos;
2901 struct ucred *cred = &nfsd->nd_cr;
dadab5e9
MD
2902 char *bp, *be;
2903 struct mbuf *mp;
2904 struct dirent *dp;
2905 caddr_t cp;
2906 u_int32_t *tl;
2907 int32_t t1;
984263bc
MD
2908 caddr_t bpos;
2909 struct mbuf *mb, *mb2, *mreq, *mp2;
2910 char *cpos, *cend, *cp2, *rbuf;
2911 struct vnode *vp = NULL;
2912 struct vattr at;
2913 nfsfh_t nfh;
2914 fhandle_t *fhp;
2915 struct uio io;
2916 struct iovec iv;
2917 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2918 int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies;
2919 int v3 = (nfsd->nd_flag & ND_NFSV3);
2920 u_quad_t frev, off, toff, verf;
2921 u_long *cookies = NULL, *cookiep; /* needs to be int64_t or off_t */
2922
2923 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
2924 fhp = &nfh.fh_generic;
2925 nfsm_srvmtofh(fhp);
2926 if (v3) {
2927 nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2928 toff = fxdr_hyper(tl);
2929 tl += 2;
2930 verf = fxdr_hyper(tl);
2931 tl += 2;
2932 } else {
2933 nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2934 toff = fxdr_unsigned(u_quad_t, *tl++);
2935 verf = 0; /* shut up gcc */
2936 }
2937 off = toff;
2938 cnt = fxdr_unsigned(int, *tl);
2939 siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2940 xfer = NFS_SRVMAXDATA(nfsd);
2941 if (cnt > xfer)
2942 cnt = xfer;
2943 if (siz > xfer)
2944 siz = xfer;
2945 fullsiz = siz;
2946 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2947 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
2948 if (!error && vp->v_type != VDIR) {
2949 error = ENOTDIR;
2950 vput(vp);
2951 vp = NULL;
2952 }
2953 if (error) {
2954 nfsm_reply(NFSX_UNSIGNED);
2955 nfsm_srvpostop_attr(getret, &at);
2956 error = 0;
2957 goto nfsmout;
2958 }
2959
2960 /*
2961 * Obtain lock on vnode for this section of the code
2962 */
2963
2964 nqsrv_getl(vp, ND_READ);
2965 if (v3) {
3b568787 2966 error = getret = VOP_GETATTR(vp, &at, td);
984263bc
MD
2967#if 0
2968 /*
2969 * XXX This check may be too strict for Solaris 2.5 clients.
2970 */
2971 if (!error && toff && verf && verf != at.va_filerev)
2972 error = NFSERR_BAD_COOKIE;
2973#endif
2974 }
2975 if (!error)
dadab5e9 2976 error = nfsrv_access(vp, VEXEC, cred, rdonly, td, 0);
984263bc
MD
2977 if (error) {
2978 vput(vp);
2979 vp = NULL;
2980 nfsm_reply(NFSX_POSTOPATTR(v3));
2981 nfsm_srvpostop_attr(getret, &at);
2982 error = 0;
2983 goto nfsmout;
2984 }
41a01a4d 2985 VOP_UNLOCK(vp, NULL, 0, td);
984263bc
MD
2986
2987 /*
2988 * end section. Allocate rbuf and continue
2989 */
2990 MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
2991again:
2992 iv.iov_base = rbuf;
2993 iv.iov_len = fullsiz;
2994 io.uio_iov = &iv;
2995 io.uio_iovcnt = 1;
2996 io.uio_offset = (off_t)off;
2997 io.uio_resid = fullsiz;
2998 io.uio_segflg = UIO_SYSSPACE;
2999 io.uio_rw = UIO_READ;
dadab5e9 3000 io.uio_td = NULL;
984263bc 3001 eofflag = 0;
41a01a4d 3002 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
3003 if (cookies) {
3004 free((caddr_t)cookies, M_TEMP);
3005 cookies = NULL;
3006 }
3007 error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies);
3008 off = (off_t)io.uio_offset;
3009 if (!cookies && !error)
3010 error = NFSERR_PERM;
3011 if (v3) {
3b568787 3012 getret = VOP_GETATTR(vp, &at, td);
984263bc
MD
3013 if (!error)
3014 error = getret;
3015 }
41a01a4d 3016 VOP_UNLOCK(vp, NULL, 0, td);
984263bc
MD
3017 if (error) {
3018 vrele(vp);
3019 vp = NULL;
3020 free((caddr_t)rbuf, M_TEMP);
3021 if (cookies)
3022 free((caddr_t)cookies, M_TEMP);
3023 nfsm_reply(NFSX_POSTOPATTR(v3));
3024 nfsm_srvpostop_attr(getret, &at);
3025 error = 0;
3026 goto nfsmout;
3027 }
3028 if (io.uio_resid) {
3029 siz -= io.uio_resid;
3030
3031 /*
3032 * If nothing read, return eof
3033 * rpc reply
3034 */
3035 if (siz == 0) {
3036 vrele(vp);
3037 vp = NULL;
3038 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
3039 2 * NFSX_UNSIGNED);
3040 if (v3) {
3041 nfsm_srvpostop_attr(getret, &at);
3042 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3043 txdr_hyper(at.va_filerev, tl);
3044 tl += 2;
3045 } else
3046 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3047 *tl++ = nfs_false;
3048 *tl = nfs_true;
3049 FREE((caddr_t)rbuf, M_TEMP);
3050 FREE((caddr_t)cookies, M_TEMP);
3051 error = 0;
3052 goto nfsmout;
3053 }
3054 }
3055
3056 /*
3057 * Check for degenerate cases of nothing useful read.
3058 * If so go try again
3059 */
3060 cpos = rbuf;
3061 cend = rbuf + siz;
3062 dp = (struct dirent *)cpos;
3063 cookiep = cookies;
3064 /*
3065 * For some reason FreeBSD's ufs_readdir() chooses to back the
3066 * directory offset up to a block boundary, so it is necessary to
3067 * skip over the records that preceed the requested offset. This
3068 * requires the assumption that file offset cookies monotonically
3069 * increase.
3070 */
3071 while (cpos < cend && ncookies > 0 &&
3072 (dp->d_fileno == 0 || dp->d_type == DT_WHT ||
3073 ((u_quad_t)(*cookiep)) <= toff)) {
3074 cpos += dp->d_reclen;
3075 dp = (struct dirent *)cpos;
3076 cookiep++;
3077 ncookies--;
3078 }
3079 if (cpos >= cend || ncookies == 0) {
3080 toff = off;
3081 siz = fullsiz;
3082 goto again;
3083 }
3084
3085 len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */
3086 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
3087 if (v3) {
3088 nfsm_srvpostop_attr(getret, &at);
3089 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3090 txdr_hyper(at.va_filerev, tl);
3091 }
3092 mp = mp2 = mb;
3093 bp = bpos;
3094 be = bp + M_TRAILINGSPACE(mp);
3095
3096 /* Loop through the records and build reply */
3097 while (cpos < cend && ncookies > 0) {
3098 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
3099 nlen = dp->d_namlen;
3100 rem = nfsm_rndup(nlen) - nlen;
3101 len += (4 * NFSX_UNSIGNED + nlen + rem);
3102 if (v3)
3103 len += 2 * NFSX_UNSIGNED;
3104 if (len > cnt) {
3105 eofflag = 0;
3106 break;
3107 }
3108 /*
3109 * Build the directory record xdr from
3110 * the dirent entry.
3111 */
3112 nfsm_clget;
3113 *tl = nfs_true;
3114 bp += NFSX_UNSIGNED;
3115 if (v3) {
3116 nfsm_clget;
3117 *tl = 0;
3118 bp += NFSX_UNSIGNED;
3119 }
3120 nfsm_clget;
3121 *tl = txdr_unsigned(dp->d_fileno);
3122 bp += NFSX_UNSIGNED;
3123 nfsm_clget;
3124 *tl = txdr_unsigned(nlen);
3125 bp += NFSX_UNSIGNED;
3126
3127 /* And loop around copying the name */
3128 xfer = nlen;
3129 cp = dp->d_name;
3130 while (xfer > 0) {
3131 nfsm_clget;
3132 if ((bp+xfer) > be)
3133 tsiz = be-bp;
3134 else
3135 tsiz = xfer;
3136 bcopy(cp, bp, tsiz);
3137 bp += tsiz;
3138 xfer -= tsiz;
3139 if (xfer > 0)
3140 cp += tsiz;
3141 }
3142 /* And null pad to a int32_t boundary */
3143 for (i = 0; i < rem; i++)
3144 *bp++ = '\0';
3145 nfsm_clget;
3146
3147 /* Finish off the record */
3148 if (v3) {
3149 *tl = 0;
3150 bp += NFSX_UNSIGNED;
3151 nfsm_clget;
3152 }
3153 *tl = txdr_unsigned(*cookiep);
3154 bp += NFSX_UNSIGNED;
3155 }
3156 cpos += dp->d_reclen;
3157 dp = (struct dirent *)cpos;
3158 cookiep++;
3159 ncookies--;
3160 }
3161 vrele(vp);
3162 vp = NULL;
3163 nfsm_clget;
3164 *tl = nfs_false;
3165 bp += NFSX_UNSIGNED;
3166 nfsm_clget;
3167 if (eofflag)
3168 *tl = nfs_true;
3169 else
3170 *tl = nfs_false;
3171 bp += NFSX_UNSIGNED;
3172 if (mp != mb) {
3173 if (bp < be)
3174 mp->m_len = bp - mtod(mp, caddr_t);
3175 } else
3176 mp->m_len += bp - bpos;
3177 FREE((caddr_t)rbuf, M_TEMP);
3178 FREE((caddr_t)cookies, M_TEMP);
3179
3180nfsmout:
3181 if (vp)
3182 vrele(vp);
3183 return(error);
3184}
3185
3186int
e851b29e
CP
3187nfsrv_readdirplus(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3188 struct thread *td, struct mbuf **mrq)
984263bc
MD
3189{
3190 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3191 struct sockaddr *nam = nfsd->nd_nam;
3192 caddr_t dpos = nfsd->nd_dpos;
3193 struct ucred *cred = &nfsd->nd_cr;
dadab5e9
MD
3194 char *bp, *be;
3195 struct mbuf *mp;
3196 struct dirent *dp;
3197 caddr_t cp;
3198 u_int32_t *tl;
3199 int32_t t1;
984263bc
MD
3200 caddr_t bpos;
3201 struct mbuf *mb, *mb2, *mreq, *mp2;
3202 char *cpos, *cend, *cp2, *rbuf;
3203 struct vnode *vp = NULL, *nvp;
3204 struct flrep fl;
3205 nfsfh_t nfh;
3206 fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
3207 struct uio io;
3208 struct iovec iv;
3209 struct vattr va, at, *vap = &va;
3210 struct nfs_fattr *fp;
3211 int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
3212 int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies;
3213 u_quad_t frev, off, toff, verf;
3214 u_long *cookies = NULL, *cookiep; /* needs to be int64_t or off_t */
3215
3216 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
3217 fhp = &nfh.fh_generic;
3218 nfsm_srvmtofh(fhp);
3219 nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3220 toff = fxdr_hyper(tl);
3221 tl += 2;
3222 verf = fxdr_hyper(tl);
3223 tl += 2;
3224 siz = fxdr_unsigned(int, *tl++);
3225 cnt = fxdr_unsigned(int, *tl);
3226 off = toff;
3227 siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
3228 xfer = NFS_SRVMAXDATA(nfsd);
3229 if (cnt > xfer)
3230 cnt = xfer;
3231 if (siz > xfer)
3232 siz = xfer;
3233 fullsiz = siz;
3234 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3235 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
3236 if (!error && vp->v_type != VDIR) {
3237 error = ENOTDIR;
3238 vput(vp);
3239 vp = NULL;
3240 }
3241 if (error) {
3242 nfsm_reply(NFSX_UNSIGNED);
3243 nfsm_srvpostop_attr(getret, &at);
3244 error = 0;
3245 goto nfsmout;
3246 }
3b568787 3247 error = getret = VOP_GETATTR(vp, &at, td);
984263bc
MD
3248#if 0
3249 /*
3250 * XXX This check may be too strict for Solaris 2.5 clients.
3251 */
3252 if (!error && toff && verf && verf != at.va_filerev)
3253 error = NFSERR_BAD_COOKIE;
3254#endif
3255 if (!error) {
3256 nqsrv_getl(vp, ND_READ);
dadab5e9 3257 error = nfsrv_access(vp, VEXEC, cred, rdonly, td, 0);
984263bc
MD
3258 }
3259 if (error) {
3260 vput(vp);
3261 vp = NULL;
3262 nfsm_reply(NFSX_V3POSTOPATTR);
3263 nfsm_srvpostop_attr(getret, &at);
3264 error = 0;
3265 goto nfsmout;
3266 }
41a01a4d 3267 VOP_UNLOCK(vp, NULL, 0, td);
984263bc
MD
3268 MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK);
3269again:
3270 iv.iov_base = rbuf;
3271 iv.iov_len = fullsiz;
3272 io.uio_iov = &iv;
3273 io.uio_iovcnt = 1;
3274 io.uio_offset = (off_t)off;
3275 io.uio_resid = fullsiz;
3276 io.uio_segflg = UIO_SYSSPACE;
3277 io.uio_rw = UIO_READ;
dadab5e9 3278 io.uio_td = NULL;
984263bc 3279 eofflag = 0;
41a01a4d 3280 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td);
984263bc
MD
3281 if (cookies) {
3282 free((caddr_t)cookies, M_TEMP);
3283 cookies = NULL;
3284 }
3285 error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies);
3286 off = (u_quad_t)io.uio_offset;
3b568787 3287 getret = VOP_GETATTR(vp, &at, td);
41a01a4d 3288 VOP_UNLOCK(vp, NULL, 0, td);
984263bc
MD
3289 if (!cookies && !error)
3290 error = NFSERR_PERM;
3291 if (!error)
3292 error = getret;
3293 if (error) {
3294 vrele(vp);
3295 vp = NULL;
3296 if (cookies)
3297 free((caddr_t)cookies, M_TEMP);
3298 free((caddr_t)rbuf, M_TEMP);
3299 nfsm_reply(NFSX_V3POSTOPATTR);
3300 nfsm_srvpostop_attr(getret, &at);
3301 error = 0;
3302 goto nfsmout;
3303 }
3304 if (io.uio_resid) {
3305 siz -= io.uio_resid;
3306
3307 /*
3308 * If nothing read, return eof
3309 * rpc reply
3310 */
3311 if (siz == 0) {
3312 vrele(vp);
3313 vp = NULL;
3314 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
3315 2 * NFSX_UNSIGNED);
3316 nfsm_srvpostop_attr(getret, &at);
3317 nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3318 txdr_hyper(at.va_filerev, tl);
3319 tl += 2;
3320 *tl++ = nfs_false;
3321 *tl = nfs_true;
3322 FREE((caddr_t)cookies, M_TEMP);
3323 FREE((caddr_t)rbuf, M_TEMP);
3324 error = 0;
3325 goto nfsmout;
3326 }
3327 }
3328
3329 /*
3330 * Check for degenerate cases of nothing useful read.
3331 * If so go try again
3332 */
3333 cpos = rbuf;
3334 cend = rbuf + siz;
3335 dp = (struct dirent *)cpos;
3336 cookiep = cookies;
3337 /*
3338 * For some reason FreeBSD's ufs_readdir() chooses to back the
3339 * directory offset up to a block boundary, so it is necessary to
3340 * skip over the records that preceed the requested offset. This
3341 * requires the assumption that file offset cookies monotonically
3342 * increase.
3343 */
3344 while (cpos < cend && ncookies > 0 &&
3345 (dp->d_fileno == 0 || dp->d_type == DT_WHT ||
3346 ((u_quad_t)(*cookiep)) <= toff)) {
3347 cpos += dp->d_reclen;
3348 dp = (struct dirent *)cpos;
3349 cookiep++;
3350 ncookies--;
3351 }
3352 if (cpos >= cend || ncookies == 0) {
3353 toff = off;
3354 siz = fullsiz;
3355 goto again;
3356 }
3357
3358 /*
3359 * Probe one of the directory entries to see if the filesystem
3360 * supports VGET.
3361 */
3362 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp) == EOPNOTSUPP) {
3363 error = NFSERR_NOTSUPP;
3364 vrele(vp);
3365 vp = NULL;
3366 free((caddr_t)cookies, M_TEMP);
3367 free((caddr_t)rbuf, M_TEMP);
3368 nfsm_reply(NFSX_V3POSTOPATTR);
3369 nfsm_srvpostop_attr(getret, &at);
3370 error = 0;
3371 goto nfsmout;
3372 }
3373 vput(nvp);
3374 nvp = NULL;
3375
3376 dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
3377 nfsm_reply(cnt);
3378 nfsm_srvpostop_attr(getret, &at);
3379 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3380 txdr_hyper(at.va_filerev, tl);
3381 mp = mp2 = mb;
3382 bp = bpos;
3383 be = bp + M_TRAILINGSPACE(mp);
3384
3385 /* Loop through the records and build reply */
3386 while (cpos < cend && ncookies > 0) {
3387 if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
3388 nlen = dp->d_namlen;
3389 rem = nfsm_rndup(nlen)-nlen;
3390
3391 /*
3392 * For readdir_and_lookup get the vnode using
3393 * the file number.
3394 */
3395 if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
3396 goto invalid;
3397 bzero((caddr_t)nfhp, NFSX_V3FH);
3398 nfhp->fh_fsid =
3399 nvp->v_mount->mnt_stat.f_fsid;
3400 if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
3401 vput(nvp);
3402 nvp = NULL;
3403 goto invalid;
3404 }
3b568787 3405 if (VOP_GETATTR(nvp, vap, td)) {
984263bc
MD
3406 vput(nvp);
3407 nvp = NULL;
3408 goto invalid;
3409 }
3410 vput(nvp);
3411 nvp = NULL;
3412
3413 /*
3414 * If either the dircount or maxcount will be
3415 * exceeded, get out now. Both of these lengths
3416 * are calculated conservatively, including all
3417 * XDR overheads.
3418 */
3419 len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
3420 NFSX_V3POSTOPATTR);
3421 dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
3422 if (len > cnt || dirlen > fullsiz) {
3423 eofflag = 0;
3424 break;
3425 }
3426
3427 /*
3428 * Build the directory record xdr from
3429 * the dirent entry.
3430 */
3431 fp = (struct nfs_fattr *)&fl.fl_fattr;
3432 nfsm_srvfillattr(vap, fp);
3433 fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
3434 fl.fl_fhok = nfs_true;
3435 fl.fl_postopok = nfs_true;
3436 fl.fl_off.nfsuquad[0] = 0;
3437 fl.fl_off.nfsuquad[1] = txdr_unsigned(*cookiep);
3438
3439 nfsm_clget;
3440 *tl = nfs_true;
3441 bp += NFSX_UNSIGNED;
3442 nfsm_clget;
3443 *tl = 0;
3444 bp += NFSX_UNSIGNED;
3445 nfsm_clget;
3446 *tl = txdr_unsigned(dp->d_fileno);
3447 bp += NFSX_UNSIGNED;
3448 nfsm_clget;
3449 *tl = txdr_unsigned(nlen);
3450 bp += NFSX_UNSIGNED;
3451
3452 /* And loop around copying the name */
3453 xfer = nlen;
3454 cp = dp->d_name;
3455 while (xfer > 0) {
3456 nfsm_clget;
3457 if ((bp + xfer) > be)
3458 tsiz = be - bp;
3459 else
3460 tsiz = xfer;
3461 bcopy(cp, bp, tsiz);
3462 bp += tsiz;
3463 xfer -= tsiz;
3464 if (xfer > 0)
3465 cp += tsiz;
3466 }
3467 /* And null pad to a int32_t boundary */
3468 for (i = 0; i < rem; i++)
3469 *bp++ = '\0';
3470
3471 /*
3472 * Now copy the flrep structure out.
3473 */
3474 xfer = sizeof (struct flrep);
3475 cp = (caddr_t)&fl;
3476 while (xfer > 0) {
3477 nfsm_clget;
3478 if ((bp + xfer) > be)
3479 tsiz = be - bp;
3480 else
3481 tsiz = xfer;
3482 bcopy(cp, bp, tsiz);
3483 bp += tsiz;
3484 xfer -= tsiz;
3485 if (xfer > 0)
3486 cp += tsiz;
3487 }
3488 }
3489invalid:
3490 cpos += dp->d_reclen;
3491 dp = (struct dirent *)cpos;
3492 cookiep++;
3493 ncookies--;
3494 }
3495 vrele(vp);
3496 vp = NULL;
3497 nfsm_clget;
3498 *tl = nfs_false;
3499 bp += NFSX_UNSIGNED;
3500 nfsm_clget;
3501 if (eofflag)
3502 *tl = nfs_true;
3503 else
3504 *tl = nfs_false;
3505 bp += NFSX_UNSIGNED;
3506 if (mp != mb) {
3507 if (bp < be)
3508 mp->m_len = bp - mtod(mp, caddr_t);
3509 } else
3510 mp->m_len += bp - bpos;
3511 FREE((caddr_t)cookies, M_TEMP);
3512 FREE((caddr_t)rbuf, M_TEMP);
3513nfsmout:
3514 if (vp)
3515 vrele(vp);
3516 return(error);
3517}
3518
3519/*
3520 * nfs commit service
3521 */
3522int
e851b29e
CP
3523nfsrv_commit(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3524 struct thread *td, struct mbuf **mrq)
984263bc
MD
3525{
3526 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3527 struct sockaddr *nam = nfsd->nd_nam;
3528 caddr_t dpos = nfsd->nd_dpos;
3529 struct ucred *cred = &nfsd->nd_cr;
3530 struct vattr bfor, aft;
3531 struct vnode *vp = NULL;
3532 nfsfh_t nfh;
3533 fhandle_t *fhp;
dadab5e9
MD
3534 u_int32_t *tl;
3535 int32_t t1;
984263bc
MD
3536 caddr_t bpos;
3537 int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache;
3538 char *cp2;
3539 struct mbuf *mb, *mb2, *mreq;
3540 u_quad_t frev, off;
3541
3542 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
3543#ifndef nolint
3544 cache = 0;
3545#endif
3546 fhp = &nfh.fh_generic;
3547 nfsm_srvmtofh(fhp);
3548 nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3549
3550 /*
3551 * XXX At this time VOP_FSYNC() does not accept offset and byte
3552 * count parameters, so these arguments are useless (someday maybe).
3553 */
3554 off = fxdr_hyper(tl);
3555 tl += 2;
3556 cnt = fxdr_unsigned(int, *tl);
3557 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3558 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
3559 if (error) {
3560 nfsm_reply(2 * NFSX_UNSIGNED);
3561 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3562 error = 0;
3563 goto nfsmout;
3564 }
3b568787 3565 for_ret = VOP_GETATTR(vp, &bfor, td);
984263bc
MD
3566
3567 if (cnt > MAX_COMMIT_COUNT) {
3568 /*
3569 * Give up and do the whole thing
3570 */
3571 if (vp->v_object &&
3572 (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) {
3573 vm_object_page_clean(vp->v_object, 0, 0, OBJPC_SYNC);
3574 }
3b568787 3575 error = VOP_FSYNC(vp, MNT_WAIT, td);
984263bc
MD
3576 } else {
3577 /*
3578 * Locate and synchronously write any buffers that fall
3579 * into the requested range. Note: we are assuming that
3580 * f_iosize is a power of 2.
3581 */
3582 int iosize = vp->v_mount->mnt_stat.f_iosize;
3583 int iomask = iosize - 1;
3584 int s;
3585 daddr_t lblkno;
3586
3587 /*
3588 * Align to iosize boundry, super-align to page boundry.
3589 */
3590 if (off & iomask) {
3591 cnt += off & iomask;
3592 off &= ~(u_quad_t)iomask;
3593 }
3594 if (off & PAGE_MASK) {
3595 cnt += off & PAGE_MASK;
3596 off &= ~(u_quad_t)PAGE_MASK;
3597 }
3598 lblkno = off / iosize;
3599
3600 if (vp->v_object &&
3601 (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) {
3602 vm_object_page_clean(vp->v_object, off / PAGE_SIZE, (cnt + PAGE_MASK) / PAGE_SIZE, OBJPC_SYNC);
3603 }
3604
3605 s = splbio();
3606 while (cnt > 0) {
3607 struct buf *bp;
3608
3609 /*
3610 * If we have a buffer and it is marked B_DELWRI we
3611 * have to lock and write it. Otherwise the prior
3612 * write is assumed to have already been committed.
3613 */
3614 if ((bp = gbincore(vp, lblkno)) != NULL && (bp->b_flags & B_DELWRI)) {
3615 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
3616 BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL);
3617 continue; /* retry */
3618 }
3619 bremfree(bp);
3620 bp->b_flags &= ~B_ASYNC;
3621 VOP_BWRITE(bp->b_vp, bp);
3622 ++nfs_commit_miss;
3623 }
3624 ++nfs_commit_blks;
3625 if (cnt < iosize)
3626 break;
3627 cnt -= iosize;
3628 ++lblkno;
3629 }
3630 splx(s);
3631 }
3632
3b568787 3633 aft_ret = VOP_GETATTR(vp, &aft, td);
984263bc
MD
3634 vput(vp);
3635 vp = NULL;
3636 nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3637 nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3638 if (!error) {
3639 nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3640 if (nfsver.tv_sec == 0)
3641 nfsver = boottime;
3642 *tl++ = txdr_unsigned(nfsver.tv_sec);
3a6f9faf 3643 *tl = txdr_unsigned(nfsver.tv_nsec / 1000);
984263bc
MD
3644 } else {
3645 error = 0;
3646 }
3647nfsmout:
3648 if (vp)
3649 vput(vp);
3650 return(error);
3651}
3652
3653/*
3654 * nfs statfs service
3655 */
3656int
e851b29e
CP
3657nfsrv_statfs(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3658 struct thread *td, struct mbuf **mrq)
984263bc
MD
3659{
3660 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3661 struct sockaddr *nam = nfsd->nd_nam;
3662 caddr_t dpos = nfsd->nd_dpos;
3663 struct ucred *cred = &nfsd->nd_cr;
dadab5e9
MD
3664 struct statfs *sf;
3665 struct nfs_statfs *sfp;
3666 u_int32_t *tl;
3667 int32_t t1;
984263bc
MD
3668 caddr_t bpos;
3669 int error = 0, rdonly, cache, getret = 1;
3670 int v3 = (nfsd->nd_flag & ND_NFSV3);
3671 char *cp2;
3672 struct mbuf *mb, *mb2, *mreq;
3673 struct vnode *vp = NULL;
3674 struct vattr at;
3675 nfsfh_t nfh;
3676 fhandle_t *fhp;
3677 struct statfs statfs;
3678 u_quad_t frev, tval;
3679
3680 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
3681#ifndef nolint
3682 cache = 0;
3683#endif
3684 fhp = &nfh.fh_generic;
3685 nfsm_srvmtofh(fhp);
3686 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3687 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
3688 if (error) {
3689 nfsm_reply(NFSX_UNSIGNED);
3690 nfsm_srvpostop_attr(getret, &at);
3691 error = 0;
3692 goto nfsmout;
3693 }
3694 sf = &statfs;
dadab5e9 3695 error = VFS_STATFS(vp->v_mount, sf, td);
3b568787 3696 getret = VOP_GETATTR(vp, &at, td);
984263bc
MD
3697 vput(vp);
3698 vp = NULL;
3699 nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3700 if (v3)
3701 nfsm_srvpostop_attr(getret, &at);
3702 if (error) {
3703 error = 0;
3704 goto nfsmout;
3705 }
3706 nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3707 if (v3) {
3708 tval = (u_quad_t)sf->f_blocks;
3709 tval *= (u_quad_t)sf->f_bsize;
3710 txdr_hyper(tval, &sfp->sf_tbytes);
3711 tval = (u_quad_t)sf->f_bfree;
3712 tval *= (u_quad_t)sf->f_bsize;
3713 txdr_hyper(tval, &sfp->sf_fbytes);
3714 tval = (u_quad_t)sf->f_bavail;
3715 tval *= (u_quad_t)sf->f_bsize;
3716 txdr_hyper(tval, &sfp->sf_abytes);
3717 sfp->sf_tfiles.nfsuquad[0] = 0;
3718 sfp->sf_tfiles.nfsuquad[1] = txdr_unsigned(sf->f_files);
3719 sfp->sf_ffiles.nfsuquad[0] = 0;
3720 sfp->sf_ffiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree);
3721 sfp->sf_afiles.nfsuquad[0] = 0;
3722 sfp->sf_afiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree);
3723 sfp->sf_invarsec = 0;
3724 } else {
3725 sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3726 sfp->sf_bsize = txdr_unsigned(sf->f_bsize);
3727 sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3728 sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3729 sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3730 }
3731nfsmout:
3732 if (vp)
3733 vput(vp);
3734 return(error);
3735}
3736
3737/*
3738 * nfs fsinfo service
3739 */
3740int
e851b29e
CP
3741nfsrv_fsinfo(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3742 struct thread *td, struct mbuf **mrq)
984263bc
MD
3743{
3744 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3745 struct sockaddr *nam = nfsd->nd_nam;
3746 caddr_t dpos = nfsd->nd_dpos;
3747 struct ucred *cred = &nfsd->nd_cr;
dadab5e9
MD
3748 u_int32_t *tl;
3749 struct nfsv3_fsinfo *sip;
3750 int32_t t1;
984263bc
MD
3751 caddr_t bpos;
3752 int error = 0, rdonly, cache, getret = 1, pref;
3753 char *cp2;
3754 struct mbuf *mb, *mb2, *mreq;
3755 struct vnode *vp = NULL;
3756 struct vattr at;
3757 nfsfh_t nfh;
3758 fhandle_t *fhp;
3759 u_quad_t frev, maxfsize;
3760 struct statfs sb;
3761
3762 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
3763#ifndef nolint
3764 cache = 0;
3765#endif
3766 fhp = &nfh.fh_generic;
3767 nfsm_srvmtofh(fhp);
3768 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3769 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
3770 if (error) {
3771 nfsm_reply(NFSX_UNSIGNED);
3772 nfsm_srvpostop_attr(getret, &at);
3773 error = 0;
3774 goto nfsmout;
3775 }
3776
3777 /* XXX Try to make a guess on the max file size. */
dadab5e9 3778 VFS_STATFS(vp->v_mount, &sb, td);
984263bc
MD
3779 maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1;
3780
3b568787 3781 getret = VOP_GETATTR(vp, &at, td);
984263bc
MD
3782 vput(vp);
3783 vp = NULL;
3784 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3785 nfsm_srvpostop_attr(getret, &at);
3786 nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3787
3788 /*
3789 * XXX
3790 * There should be file system VFS OP(s) to get this information.
3791 * For now, assume ufs.
3792 */
3793 if (slp->ns_so->so_type == SOCK_DGRAM)
3794 pref = NFS_MAXDGRAMDATA;
3795 else
3796 pref = NFS_MAXDATA;
3797 sip->fs_rtmax = txdr_unsigned(NFS_MAXDATA);
3798 sip->fs_rtpref = txdr_unsigned(pref);
3799 sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3800 sip->fs_wtmax = txdr_unsigned(NFS_MAXDATA);
3801 sip->fs_wtpref = txdr_unsigned(pref);
3802 sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3803 sip->fs_dtpref = txdr_unsigned(pref);
3804 txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3805 sip->fs_timedelta.nfsv3_sec = 0;
3806 sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3807 sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3808 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3809 NFSV3FSINFO_CANSETTIME);
3810nfsmout:
3811 if (vp)
3812 vput(vp);
3813 return(error);
3814}
3815
3816/*
3817 * nfs pathconf service
3818 */
3819int
e851b29e
CP
3820nfsrv_pathconf(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
3821 struct thread *td, struct mbuf **mrq)
984263bc
MD
3822{
3823 struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3824 struct sockaddr *nam = nfsd->nd_nam;
3825 caddr_t dpos = nfsd->nd_dpos;
3826 struct ucred *cred = &nfsd->nd_cr;
dadab5e9
MD
3827 u_int32_t *tl;
3828 struct nfsv3_pathconf *pc;
3829 int32_t t1;
984263bc
MD
3830 caddr_t bpos;
3831 int error = 0, rdonly, cache, getret = 1;
3832 register_t linkmax, namemax, chownres, notrunc;
3833 char *cp2;
3834 struct mbuf *mb, *mb2, *mreq;
3835 struct vnode *vp = NULL;
3836 struct vattr at;
3837 nfsfh_t nfh;
3838 fhandle_t *fhp;
3839 u_quad_t frev;
3840
3841 nfsdbprintf(("%s %d\n", __FILE__, __LINE__));
3842#ifndef nolint
3843 cache = 0;
3844#endif
3845 fhp = &nfh.fh_generic;
3846 nfsm_srvmtofh(fhp);
3847 error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3848 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), TRUE);
3849 if (error) {
3850 nfsm_reply(NFSX_UNSIGNED);
3851 nfsm_srvpostop_attr(getret, &at);
3852 error = 0;
3853 goto nfsmout;
3854 }
3855 error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3856 if (!error)
3857 error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3858 if (!error)
3859 error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3860 if (!error)
3861 error = VOP_PATHCONF(vp, _PC_NO_TRUNC, &notrunc);
3b568787 3862 getret = VOP_GETATTR(vp, &at, td);
984263bc
MD
3863 vput(vp);
3864 vp = NULL;
3865 nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3866 nfsm_srvpostop_attr(getret, &at);
3867 if (error) {
3868 error = 0;
3869 goto nfsmout;
3870 }
3871 nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3872
3873 pc->pc_linkmax = txdr_unsigned(linkmax);
3874 pc->pc_namemax = txdr_unsigned(namemax);
3875 pc->pc_notrunc = txdr_unsigned(notrunc);
3876 pc->pc_chownrestricted = txdr_unsigned(chownres);
3877
3878 /*
3879 * These should probably be supported by VOP_PATHCONF(), but
3880 * until msdosfs is exportable (why would you want to?), the
3881 * Unix defaults should be ok.
3882 */
3883 pc->pc_caseinsensitive = nfs_false;
3884 pc->pc_casepreserving = nfs_true;
3885nfsmout:
3886 if (vp)
3887 vput(vp);
3888 return(error);
3889}
3890
3891/*
3892 * Null operation, used by clients to ping server
3893