Deleted Added
full compact
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_subs.c 8.8 (Berkeley) 5/22/95
37 */
38
39#include <sys/cdefs.h>
40__FBSDID("$FreeBSD: head/sys/nfsclient/nfs_subs.c 88091 2001-12-18 01:22:09Z iedowse $");
40__FBSDID("$FreeBSD: head/sys/nfsclient/nfs_subs.c 92783 2002-03-20 10:07:52Z jeff $");
41
42/*
43 * These functions support the macros and help fiddle mbuf chains for
44 * the nfs op functions. They do things like create the rpc header and
45 * copy data between mbuf chains and uio lists.
46 */
47
48#include <sys/param.h>
49#include <sys/systm.h>
50#include <sys/kernel.h>
51#include <sys/bio.h>
52#include <sys/buf.h>
53#include <sys/proc.h>
54#include <sys/mount.h>
55#include <sys/vnode.h>
56#include <sys/namei.h>
57#include <sys/mbuf.h>
58#include <sys/socket.h>
59#include <sys/stat.h>
60#include <sys/malloc.h>
61#include <sys/sysent.h>
62#include <sys/syscall.h>
63#include <sys/sysproto.h>
64
65#include <vm/vm.h>
66#include <vm/vm_object.h>
67#include <vm/vm_extern.h>
68#include <vm/vm_zone.h>
68#include <vm/uma.h>
69
70#include <nfs/rpcv2.h>
71#include <nfs/nfsproto.h>
72#include <nfsclient/nfs.h>
73#include <nfsclient/nfsnode.h>
74#include <nfs/xdr_subs.h>
75#include <nfsclient/nfsm_subs.h>
76#include <nfsclient/nfsmount.h>
77
78#include <netinet/in.h>
79
80/*
81 * Data items converted to xdr at startup, since they are constant
82 * This is kinda hokey, but may save a little time doing byte swaps
83 */
84u_int32_t nfs_xdrneg1;
85u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
86 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted;
87u_int32_t nfs_true, nfs_false;
88
89/* And other global data */
90static u_int32_t nfs_xid = 0;
91static enum vtype nv2tov_type[8]= {
92 VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON
93};
94
95int nfs_ticks;
96int nfs_pbuf_freecnt = -1; /* start out unlimited */
97
98struct nfs_reqq nfs_reqq;
99struct nfs_bufq nfs_bufq;
100
101static int nfs_prev_nfsclnt_sy_narg;
102static sy_call_t *nfs_prev_nfsclnt_sy_call;
103
104/*
105 * and the reverse mapping from generic to Version 2 procedure numbers
106 */
107int nfsv2_procid[NFS_NPROCS] = {
108 NFSV2PROC_NULL,
109 NFSV2PROC_GETATTR,
110 NFSV2PROC_SETATTR,
111 NFSV2PROC_LOOKUP,
112 NFSV2PROC_NOOP,
113 NFSV2PROC_READLINK,
114 NFSV2PROC_READ,
115 NFSV2PROC_WRITE,
116 NFSV2PROC_CREATE,
117 NFSV2PROC_MKDIR,
118 NFSV2PROC_SYMLINK,
119 NFSV2PROC_CREATE,
120 NFSV2PROC_REMOVE,
121 NFSV2PROC_RMDIR,
122 NFSV2PROC_RENAME,
123 NFSV2PROC_LINK,
124 NFSV2PROC_READDIR,
125 NFSV2PROC_NOOP,
126 NFSV2PROC_STATFS,
127 NFSV2PROC_NOOP,
128 NFSV2PROC_NOOP,
129 NFSV2PROC_NOOP,
130 NFSV2PROC_NOOP,
131};
132
133LIST_HEAD(nfsnodehashhead, nfsnode);
134
135/*
136 * Create the header for an rpc request packet
137 * The hsiz is the size of the rest of the nfs request header.
138 * (just used to decide if a cluster is a good idea)
139 */
140struct mbuf *
141nfsm_reqhead(struct vnode *vp, u_long procid, int hsiz)
142{
143 struct mbuf *mb;
144
145 MGET(mb, M_TRYWAIT, MT_DATA);
146 if (hsiz >= MINCLSIZE)
147 MCLGET(mb, M_TRYWAIT);
148 mb->m_len = 0;
149 return (mb);
150}
151
152/*
153 * Build the RPC header and fill in the authorization info.
154 * The authorization string argument is only used when the credentials
155 * come from outside of the kernel.
156 * Returns the head of the mbuf list.
157 */
158struct mbuf *
159nfsm_rpchead(struct ucred *cr, int nmflag, int procid, int auth_type,
160 int auth_len, struct mbuf *mrest, int mrest_len, struct mbuf **mbp,
161 u_int32_t *xidp)
162{
163 struct mbuf *mb;
164 u_int32_t *tl;
165 caddr_t bpos;
166 int i;
167 struct mbuf *mreq;
168 int grpsiz, authsiz;
169
170 authsiz = nfsm_rndup(auth_len);
171 MGETHDR(mb, M_TRYWAIT, MT_DATA);
172 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
173 MCLGET(mb, M_TRYWAIT);
174 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
175 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
176 } else {
177 MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
178 }
179 mb->m_len = 0;
180 mreq = mb;
181 bpos = mtod(mb, caddr_t);
182
183 /*
184 * First the RPC header.
185 */
186 tl = nfsm_build(u_int32_t *, 8 * NFSX_UNSIGNED);
187
188 /* Get a pretty random xid to start with */
189 if (!nfs_xid)
190 nfs_xid = random();
191 /*
192 * Skip zero xid if it should ever happen.
193 */
194 if (++nfs_xid == 0)
195 nfs_xid++;
196
197 *tl++ = *xidp = txdr_unsigned(nfs_xid);
198 *tl++ = rpc_call;
199 *tl++ = rpc_vers;
200 *tl++ = txdr_unsigned(NFS_PROG);
201 if (nmflag & NFSMNT_NFSV3) {
202 *tl++ = txdr_unsigned(NFS_VER3);
203 *tl++ = txdr_unsigned(procid);
204 } else {
205 *tl++ = txdr_unsigned(NFS_VER2);
206 *tl++ = txdr_unsigned(nfsv2_procid[procid]);
207 }
208
209 /*
210 * And then the authorization cred.
211 */
212 *tl++ = txdr_unsigned(auth_type);
213 *tl = txdr_unsigned(authsiz);
214 switch (auth_type) {
215 case RPCAUTH_UNIX:
216 tl = nfsm_build(u_int32_t *, auth_len);
217 *tl++ = 0; /* stamp ?? */
218 *tl++ = 0; /* NULL hostname */
219 *tl++ = txdr_unsigned(cr->cr_uid);
220 *tl++ = txdr_unsigned(cr->cr_groups[0]);
221 grpsiz = (auth_len >> 2) - 5;
222 *tl++ = txdr_unsigned(grpsiz);
223 for (i = 1; i <= grpsiz; i++)
224 *tl++ = txdr_unsigned(cr->cr_groups[i]);
225 break;
226 }
227
228 /*
229 * And the verifier...
230 */
231 tl = nfsm_build(u_int32_t *, 2 * NFSX_UNSIGNED);
232 *tl++ = txdr_unsigned(RPCAUTH_NULL);
233 *tl = 0;
234 mb->m_next = mrest;
235 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
236 mreq->m_pkthdr.rcvif = (struct ifnet *)0;
237 *mbp = mb;
238 return (mreq);
239}
240
241/*
242 * copies a uio scatter/gather list to an mbuf chain.
243 * NOTE: can ony handle iovcnt == 1
244 */
245int
246nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
247{
248 char *uiocp;
249 struct mbuf *mp, *mp2;
250 int xfer, left, mlen;
251 int uiosiz, clflg, rem;
252 char *cp;
253
254#ifdef DIAGNOSTIC
255 if (uiop->uio_iovcnt != 1)
256 panic("nfsm_uiotombuf: iovcnt != 1");
257#endif
258
259 if (siz > MLEN) /* or should it >= MCLBYTES ?? */
260 clflg = 1;
261 else
262 clflg = 0;
263 rem = nfsm_rndup(siz)-siz;
264 mp = mp2 = *mq;
265 while (siz > 0) {
266 left = uiop->uio_iov->iov_len;
267 uiocp = uiop->uio_iov->iov_base;
268 if (left > siz)
269 left = siz;
270 uiosiz = left;
271 while (left > 0) {
272 mlen = M_TRAILINGSPACE(mp);
273 if (mlen == 0) {
274 MGET(mp, M_TRYWAIT, MT_DATA);
275 if (clflg)
276 MCLGET(mp, M_TRYWAIT);
277 mp->m_len = 0;
278 mp2->m_next = mp;
279 mp2 = mp;
280 mlen = M_TRAILINGSPACE(mp);
281 }
282 xfer = (left > mlen) ? mlen : left;
283#ifdef notdef
284 /* Not Yet.. */
285 if (uiop->uio_iov->iov_op != NULL)
286 (*(uiop->uio_iov->iov_op))
287 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
288 else
289#endif
290 if (uiop->uio_segflg == UIO_SYSSPACE)
291 bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
292 else
293 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
294 mp->m_len += xfer;
295 left -= xfer;
296 uiocp += xfer;
297 uiop->uio_offset += xfer;
298 uiop->uio_resid -= xfer;
299 }
300 uiop->uio_iov->iov_base += uiosiz;
301 uiop->uio_iov->iov_len -= uiosiz;
302 siz -= uiosiz;
303 }
304 if (rem > 0) {
305 if (rem > M_TRAILINGSPACE(mp)) {
306 MGET(mp, M_TRYWAIT, MT_DATA);
307 mp->m_len = 0;
308 mp2->m_next = mp;
309 }
310 cp = mtod(mp, caddr_t)+mp->m_len;
311 for (left = 0; left < rem; left++)
312 *cp++ = '\0';
313 mp->m_len += rem;
314 *bpos = cp;
315 } else
316 *bpos = mtod(mp, caddr_t)+mp->m_len;
317 *mq = mp;
318 return (0);
319}
320
321/*
322 * Copy a string into mbufs for the hard cases...
323 */
324int
325nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
326{
327 struct mbuf *m1 = NULL, *m2;
328 long left, xfer, len, tlen;
329 u_int32_t *tl;
330 int putsize;
331
332 putsize = 1;
333 m2 = *mb;
334 left = M_TRAILINGSPACE(m2);
335 if (left > 0) {
336 tl = ((u_int32_t *)(*bpos));
337 *tl++ = txdr_unsigned(siz);
338 putsize = 0;
339 left -= NFSX_UNSIGNED;
340 m2->m_len += NFSX_UNSIGNED;
341 if (left > 0) {
342 bcopy(cp, (caddr_t) tl, left);
343 siz -= left;
344 cp += left;
345 m2->m_len += left;
346 left = 0;
347 }
348 }
349 /* Loop around adding mbufs */
350 while (siz > 0) {
351 MGET(m1, M_TRYWAIT, MT_DATA);
352 if (siz > MLEN)
353 MCLGET(m1, M_TRYWAIT);
354 m1->m_len = NFSMSIZ(m1);
355 m2->m_next = m1;
356 m2 = m1;
357 tl = mtod(m1, u_int32_t *);
358 tlen = 0;
359 if (putsize) {
360 *tl++ = txdr_unsigned(siz);
361 m1->m_len -= NFSX_UNSIGNED;
362 tlen = NFSX_UNSIGNED;
363 putsize = 0;
364 }
365 if (siz < m1->m_len) {
366 len = nfsm_rndup(siz);
367 xfer = siz;
368 if (xfer < len)
369 *(tl+(xfer>>2)) = 0;
370 } else {
371 xfer = len = m1->m_len;
372 }
373 bcopy(cp, (caddr_t) tl, xfer);
374 m1->m_len = len+tlen;
375 siz -= xfer;
376 cp += xfer;
377 }
378 *mb = m1;
379 *bpos = mtod(m1, caddr_t)+m1->m_len;
380 return (0);
381}
382
383/*
384 * Called once to initialize data structures...
385 */
386int
387nfs_init(struct vfsconf *vfsp)
388{
389 int i;
390
391 nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
391 nfsmount_zone = uma_zcreate("NFSMOUNT", sizeof(struct nfsmount),
392 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
393 rpc_vers = txdr_unsigned(RPC_VER2);
394 rpc_call = txdr_unsigned(RPC_CALL);
395 rpc_reply = txdr_unsigned(RPC_REPLY);
396 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
397 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
398 rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
399 rpc_autherr = txdr_unsigned(RPC_AUTHERR);
400 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
401 nfs_true = txdr_unsigned(TRUE);
402 nfs_false = txdr_unsigned(FALSE);
403 nfs_xdrneg1 = txdr_unsigned(-1);
404 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
405 if (nfs_ticks < 1)
406 nfs_ticks = 1;
407 /* Ensure async daemons disabled */
408 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
409 nfs_iodwant[i] = (struct proc *)0;
410 nfs_iodmount[i] = (struct nfsmount *)0;
411 }
412 nfs_nhinit(); /* Init the nfsnode table */
413
414 /*
415 * Initialize reply list and start timer
416 */
417 TAILQ_INIT(&nfs_reqq);
418
419 nfs_timer(0);
420
421 nfs_prev_nfsclnt_sy_narg = sysent[SYS_nfsclnt].sy_narg;
422 sysent[SYS_nfsclnt].sy_narg = 2;
423 nfs_prev_nfsclnt_sy_call = sysent[SYS_nfsclnt].sy_call;
424 sysent[SYS_nfsclnt].sy_call = (sy_call_t *)nfsclnt;
425
426 nfs_pbuf_freecnt = nswbuf / 2 + 1;
427
428 return (0);
429}
430
431int
432nfs_uninit(struct vfsconf *vfsp)
433{
434
435 untimeout(nfs_timer, (void *)NULL, nfs_timer_handle);
436 sysent[SYS_nfsclnt].sy_narg = nfs_prev_nfsclnt_sy_narg;
437 sysent[SYS_nfsclnt].sy_call = nfs_prev_nfsclnt_sy_call;
438 return (0);
439}
440
441/*
442 * Attribute cache routines.
443 * nfs_loadattrcache() - loads or updates the cache contents from attributes
444 * that are on the mbuf list
445 * nfs_getattrcache() - returns valid attributes if found in cache, returns
446 * error otherwise
447 */
448
449/*
450 * Load the attribute cache (that lives in the nfsnode entry) with
451 * the values on the mbuf list and
452 * Iff vap not NULL
453 * copy the attributes to *vaper
454 */
455int
456nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
457 struct vattr *vaper, int dontshrink)
458{
459 struct vnode *vp = *vpp;
460 struct vattr *vap;
461 struct nfs_fattr *fp;
462 struct nfsnode *np;
463 int32_t t1;
464 caddr_t cp2;
465 int rdev;
466 struct mbuf *md;
467 enum vtype vtyp;
468 u_short vmode;
469 struct timespec mtime;
470 int v3 = NFS_ISV3(vp);
471
472 md = *mdp;
473 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
474 cp2 = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1);
475 if (cp2 == NULL)
476 return EBADRPC;
477 fp = (struct nfs_fattr *)cp2;
478 if (v3) {
479 vtyp = nfsv3tov_type(fp->fa_type);
480 vmode = fxdr_unsigned(u_short, fp->fa_mode);
481 rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
482 fxdr_unsigned(int, fp->fa3_rdev.specdata2));
483 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
484 } else {
485 vtyp = nfsv2tov_type(fp->fa_type);
486 vmode = fxdr_unsigned(u_short, fp->fa_mode);
487 /*
488 * XXX
489 *
490 * The duplicate information returned in fa_type and fa_mode
491 * is an ambiguity in the NFS version 2 protocol.
492 *
493 * VREG should be taken literally as a regular file. If a
494 * server intents to return some type information differently
495 * in the upper bits of the mode field (e.g. for sockets, or
496 * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
497 * leave the examination of the mode bits even in the VREG
498 * case to avoid breakage for bogus servers, but we make sure
499 * that there are actually type bits set in the upper part of
500 * fa_mode (and failing that, trust the va_type field).
501 *
502 * NFSv3 cleared the issue, and requires fa_mode to not
503 * contain any type information (while also introduing sockets
504 * and FIFOs for fa_type).
505 */
506 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
507 vtyp = IFTOVT(vmode);
508 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
509 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
510
511 /*
512 * Really ugly NFSv2 kludge.
513 */
514 if (vtyp == VCHR && rdev == 0xffffffff)
515 vtyp = VFIFO;
516 }
517
518 /*
519 * If v_type == VNON it is a new node, so fill in the v_type,
520 * n_mtime fields. Check to see if it represents a special
521 * device, and if so, check for a possible alias. Once the
522 * correct vnode has been obtained, fill in the rest of the
523 * information.
524 */
525 np = VTONFS(vp);
526 if (vp->v_type != vtyp) {
527 vp->v_type = vtyp;
528 if (vp->v_type == VFIFO) {
529 vp->v_op = fifo_nfsv2nodeop_p;
530 }
531 if (vp->v_type == VCHR || vp->v_type == VBLK) {
532 vp->v_op = spec_nfsv2nodeop_p;
533 vp = addaliasu(vp, rdev);
534 np->n_vnode = vp;
535 }
536 np->n_mtime = mtime.tv_sec;
537 }
538 vap = &np->n_vattr;
539 vap->va_type = vtyp;
540 vap->va_mode = (vmode & 07777);
541 vap->va_rdev = rdev;
542 vap->va_mtime = mtime;
543 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
544 if (v3) {
545 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
546 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
547 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
548 vap->va_size = fxdr_hyper(&fp->fa3_size);
549 vap->va_blocksize = NFS_FABLKSIZE;
550 vap->va_bytes = fxdr_hyper(&fp->fa3_used);
551 vap->va_fileid = fxdr_unsigned(int32_t,
552 fp->fa3_fileid.nfsuquad[1]);
553 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
554 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
555 vap->va_flags = 0;
556 vap->va_filerev = 0;
557 } else {
558 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
559 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
560 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
561 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
562 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
563 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
564 * NFS_FABLKSIZE;
565 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
566 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
567 vap->va_flags = 0;
568 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
569 fp->fa2_ctime.nfsv2_sec);
570 vap->va_ctime.tv_nsec = 0;
571 vap->va_gen = fxdr_unsigned(u_int32_t, fp->fa2_ctime.nfsv2_usec);
572 vap->va_filerev = 0;
573 }
574 np->n_attrstamp = time_second;
575 if (vap->va_size != np->n_size) {
576 if (vap->va_type == VREG) {
577 if (dontshrink && vap->va_size < np->n_size) {
578 /*
579 * We've been told not to shrink the file;
580 * zero np->n_attrstamp to indicate that
581 * the attributes are stale.
582 */
583 vap->va_size = np->n_size;
584 np->n_attrstamp = 0;
585 } else if (np->n_flag & NMODIFIED) {
586 if (vap->va_size < np->n_size)
587 vap->va_size = np->n_size;
588 else
589 np->n_size = vap->va_size;
590 } else {
591 np->n_size = vap->va_size;
592 }
593 vnode_pager_setsize(vp, np->n_size);
594 } else {
595 np->n_size = vap->va_size;
596 }
597 }
598 if (vaper != NULL) {
599 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
600 if (np->n_flag & NCHG) {
601 if (np->n_flag & NACC)
602 vaper->va_atime = np->n_atim;
603 if (np->n_flag & NUPD)
604 vaper->va_mtime = np->n_mtim;
605 }
606 }
607 return (0);
608}
609
610#ifdef NFS_ACDEBUG
611#include <sys/sysctl.h>
612SYSCTL_DECL(_vfs_nfs);
613static int nfs_acdebug;
614SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
615#endif
616
617/*
618 * Check the time stamp
619 * If the cache is valid, copy contents to *vap and return 0
620 * otherwise return an error
621 */
622int
623nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
624{
625 struct nfsnode *np;
626 struct vattr *vap;
627 struct nfsmount *nmp;
628 int timeo;
629
630 np = VTONFS(vp);
631 vap = &np->n_vattr;
632 nmp = VFSTONFS(vp->v_mount);
633 /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
634 timeo = (time_second - np->n_mtime) / 10;
635
636#ifdef NFS_ACDEBUG
637 if (nfs_acdebug>1)
638 printf("nfs_getattrcache: initial timeo = %d\n", timeo);
639#endif
640
641 if (vap->va_type == VDIR) {
642 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
643 timeo = nmp->nm_acdirmin;
644 else if (timeo > nmp->nm_acdirmax)
645 timeo = nmp->nm_acdirmax;
646 } else {
647 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
648 timeo = nmp->nm_acregmin;
649 else if (timeo > nmp->nm_acregmax)
650 timeo = nmp->nm_acregmax;
651 }
652
653#ifdef NFS_ACDEBUG
654 if (nfs_acdebug > 2)
655 printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
656 nmp->nm_acregmin, nmp->nm_acregmax,
657 nmp->nm_acdirmin, nmp->nm_acdirmax);
658
659 if (nfs_acdebug)
660 printf("nfs_getattrcache: age = %d; final timeo = %d\n",
661 (time_second - np->n_attrstamp), timeo);
662#endif
663
664 if ((time_second - np->n_attrstamp) >= timeo) {
665 nfsstats.attrcache_misses++;
666 return (ENOENT);
667 }
668 nfsstats.attrcache_hits++;
669 if (vap->va_size != np->n_size) {
670 if (vap->va_type == VREG) {
671 if (np->n_flag & NMODIFIED) {
672 if (vap->va_size < np->n_size)
673 vap->va_size = np->n_size;
674 else
675 np->n_size = vap->va_size;
676 } else {
677 np->n_size = vap->va_size;
678 }
679 vnode_pager_setsize(vp, np->n_size);
680 } else {
681 np->n_size = vap->va_size;
682 }
683 }
684 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
685 if (np->n_flag & NCHG) {
686 if (np->n_flag & NACC)
687 vaper->va_atime = np->n_atim;
688 if (np->n_flag & NUPD)
689 vaper->va_mtime = np->n_mtim;
690 }
691 return (0);
692}
693
694static nfsuint64 nfs_nullcookie = { { 0, 0 } };
695/*
696 * This function finds the directory cookie that corresponds to the
697 * logical byte offset given.
698 */
699nfsuint64 *
700nfs_getcookie(struct nfsnode *np, off_t off, int add)
701{
702 struct nfsdmap *dp, *dp2;
703 int pos;
704
705 pos = (uoff_t)off / NFS_DIRBLKSIZ;
706 if (pos == 0 || off < 0) {
707#ifdef DIAGNOSTIC
708 if (add)
709 panic("nfs getcookie add at <= 0");
710#endif
711 return (&nfs_nullcookie);
712 }
713 pos--;
714 dp = LIST_FIRST(&np->n_cookies);
715 if (!dp) {
716 if (add) {
717 MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
718 M_NFSDIROFF, M_WAITOK);
719 dp->ndm_eocookie = 0;
720 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
721 } else
722 return ((nfsuint64 *)0);
723 }
724 while (pos >= NFSNUMCOOKIES) {
725 pos -= NFSNUMCOOKIES;
726 if (LIST_NEXT(dp, ndm_list)) {
727 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
728 pos >= dp->ndm_eocookie)
729 return ((nfsuint64 *)0);
730 dp = LIST_NEXT(dp, ndm_list);
731 } else if (add) {
732 MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
733 M_NFSDIROFF, M_WAITOK);
734 dp2->ndm_eocookie = 0;
735 LIST_INSERT_AFTER(dp, dp2, ndm_list);
736 dp = dp2;
737 } else
738 return ((nfsuint64 *)0);
739 }
740 if (pos >= dp->ndm_eocookie) {
741 if (add)
742 dp->ndm_eocookie = pos + 1;
743 else
744 return ((nfsuint64 *)0);
745 }
746 return (&dp->ndm_cookies[pos]);
747}
748
749/*
750 * Invalidate cached directory information, except for the actual directory
751 * blocks (which are invalidated separately).
752 * Done mainly to avoid the use of stale offset cookies.
753 */
754void
755nfs_invaldir(struct vnode *vp)
756{
757 struct nfsnode *np = VTONFS(vp);
758
759#ifdef DIAGNOSTIC
760 if (vp->v_type != VDIR)
761 panic("nfs: invaldir not dir");
762#endif
763 np->n_direofoffset = 0;
764 np->n_cookieverf.nfsuquad[0] = 0;
765 np->n_cookieverf.nfsuquad[1] = 0;
766 if (LIST_FIRST(&np->n_cookies))
767 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
768}
769
770/*
771 * The write verifier has changed (probably due to a server reboot), so all
772 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
773 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
774 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the
775 * mount point.
776 *
777 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
778 * writes are not clusterable.
779 */
780void
781nfs_clearcommit(struct mount *mp)
782{
783 struct vnode *vp, *nvp;
784 struct buf *bp, *nbp;
785 int s;
786
787 GIANT_REQUIRED;
788
789 s = splbio();
790 mtx_lock(&mntvnode_mtx);
791loop:
792 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp; vp = nvp) {
793 if (vp->v_mount != mp) /* Paranoia */
794 goto loop;
795 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
796 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
797 nbp = TAILQ_NEXT(bp, b_vnbufs);
798 if (BUF_REFCNT(bp) == 0 &&
799 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
800 == (B_DELWRI | B_NEEDCOMMIT))
801 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
802 }
803 }
804 mtx_unlock(&mntvnode_mtx);
805 splx(s);
806}
807
808/*
809 * Helper functions for former macros. Some of these should be
810 * moved to their callers.
811 */
812
813int
814nfsm_mtofh_xx(struct vnode *d, struct vnode **v, int v3, int *f,
815 struct mbuf **md, caddr_t *dpos)
816{
817 struct nfsnode *ttnp;
818 struct vnode *ttvp;
819 nfsfh_t *ttfhp;
820 u_int32_t *tl;
821 int ttfhsize;
822 int t1;
823
824 if (v3) {
825 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
826 if (tl == NULL)
827 return EBADRPC;
828 *f = fxdr_unsigned(int, *tl);
829 } else
830 *f = 1;
831 if (*f) {
832 t1 = nfsm_getfh_xx(&ttfhp, &ttfhsize, (v3), md, dpos);
833 if (t1 != 0)
834 return t1;
835 t1 = nfs_nget(d->v_mount, ttfhp, ttfhsize, &ttnp);
836 if (t1 != 0)
837 return t1;
838 *v = NFSTOV(ttnp);
839 }
840 if (v3) {
841 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
842 if (tl == NULL)
843 return EBADRPC;
844 if (*f)
845 *f = fxdr_unsigned(int, *tl);
846 else if (fxdr_unsigned(int, *tl))
847 nfsm_adv_xx(NFSX_V3FATTR, md, dpos);
848 }
849 if (*f) {
850 ttvp = *v;
851 t1 = nfs_loadattrcache(&ttvp, md, dpos, (struct vattr *)0, 0);
852 if (t1)
853 return t1;
854 *v = ttvp;
855 }
856 return 0;
857}
858
859int
860nfsm_getfh_xx(nfsfh_t **f, int *s, int v3, struct mbuf **md, caddr_t *dpos)
861{
862 u_int32_t *tl;
863
864 if (v3) {
865 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
866 if (tl == NULL)
867 return EBADRPC;
868 *s = fxdr_unsigned(int, *tl);
869 if (*s <= 0 || *s > NFSX_V3FHMAX)
870 return EBADRPC;
871 } else
872 *s = NFSX_V2FH;
873 *f = nfsm_dissect_xx(nfsm_rndup(*s), md, dpos);
874 if (*f == NULL)
875 return EBADRPC;
876 else
877 return 0;
878}
879
880
881int
882nfsm_loadattr_xx(struct vnode **v, struct vattr *va, struct mbuf **md,
883 caddr_t *dpos)
884{
885 int t1;
886
887 struct vnode *ttvp = *v;
888 t1 = nfs_loadattrcache(&ttvp, md, dpos, va, 0);
889 if (t1 != 0)
890 return t1;
891 *v = ttvp;
892 return 0;
893}
894
895int
896nfsm_postop_attr_xx(struct vnode **v, int *f, struct mbuf **md,
897 caddr_t *dpos)
898{
899 u_int32_t *tl;
900 int t1;
901
902 struct vnode *ttvp = *v;
903 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
904 if (tl == NULL)
905 return EBADRPC;
906 *f = fxdr_unsigned(int, *tl);
907 if (*f != 0) {
908 t1 = nfs_loadattrcache(&ttvp, md, dpos, (struct vattr *)0, 1);
909 if (t1 != 0) {
910 *f = 0;
911 return t1;
912 }
913 *v = ttvp;
914 }
915 return 0;
916}
917
918int
919nfsm_wcc_data_xx(struct vnode **v, int *f, struct mbuf **md, caddr_t *dpos)
920{
921 u_int32_t *tl;
922 int ttattrf, ttretf = 0;
923 int t1;
924
925 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
926 if (tl == NULL)
927 return EBADRPC;
928 if (*tl == nfs_true) {
929 tl = nfsm_dissect_xx(6 * NFSX_UNSIGNED, md, dpos);
930 if (tl == NULL)
931 return EBADRPC;
932 if (*f)
933 ttretf = (VTONFS(*v)->n_mtime ==
934 fxdr_unsigned(u_int32_t, *(tl + 2)));
935 }
936 t1 = nfsm_postop_attr_xx(v, &ttattrf, md, dpos);
937 if (t1)
938 return t1;
939 if (*f)
940 *f = ttretf;
941 else
942 *f = ttattrf;
943 return 0;
944}
945
946int
947nfsm_strtom_xx(const char *a, int s, int m, struct mbuf **mb, caddr_t *bpos)
948{
949 u_int32_t *tl;
950 int t1;
951
952 if (s > m)
953 return ENAMETOOLONG;
954 t1 = nfsm_rndup(s) + NFSX_UNSIGNED;
955 if (t1 <= M_TRAILINGSPACE(*mb)) {
956 tl = nfsm_build_xx(t1, mb, bpos);
957 *tl++ = txdr_unsigned(s);
958 *(tl + ((t1 >> 2) - 2)) = 0;
959 bcopy(a, tl, s);
960 } else {
961 t1 = nfsm_strtmbuf(mb, bpos, a, s);
962 if (t1 != 0)
963 return t1;
964 }
965 return 0;
966}
967
968int
969nfsm_fhtom_xx(struct vnode *v, int v3, struct mbuf **mb, caddr_t *bpos)
970{
971 u_int32_t *tl;
972 int t1;
973 caddr_t cp;
974
975 if (v3) {
976 t1 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED;
977 if (t1 < M_TRAILINGSPACE(*mb)) {
978 tl = nfsm_build_xx(t1, mb, bpos);
979 *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize);
980 *(tl + ((t1 >> 2) - 2)) = 0;
981 bcopy(VTONFS(v)->n_fhp, tl, VTONFS(v)->n_fhsize);
982 } else {
983 t1 = nfsm_strtmbuf(mb, bpos,
984 (const char *)VTONFS(v)->n_fhp,
985 VTONFS(v)->n_fhsize);
986 if (t1 != 0)
987 return t1;
988 }
989 } else {
990 cp = nfsm_build_xx(NFSX_V2FH, mb, bpos);
991 bcopy(VTONFS(v)->n_fhp, cp, NFSX_V2FH);
992 }
993 return 0;
994}
995
996void
997nfsm_v3attrbuild_xx(struct vattr *va, int full, struct mbuf **mb,
998 caddr_t *bpos)
999{
1000 u_int32_t *tl;
1001
1002 if (va->va_mode != (mode_t)VNOVAL) {
1003 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1004 *tl++ = nfs_true;
1005 *tl = txdr_unsigned(va->va_mode);
1006 } else {
1007 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1008 *tl = nfs_false;
1009 }
1010 if (full && va->va_uid != (uid_t)VNOVAL) {
1011 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1012 *tl++ = nfs_true;
1013 *tl = txdr_unsigned(va->va_uid);
1014 } else {
1015 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1016 *tl = nfs_false;
1017 }
1018 if (full && va->va_gid != (gid_t)VNOVAL) {
1019 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1020 *tl++ = nfs_true;
1021 *tl = txdr_unsigned(va->va_gid);
1022 } else {
1023 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1024 *tl = nfs_false;
1025 }
1026 if (full && va->va_size != VNOVAL) {
1027 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1028 *tl++ = nfs_true;
1029 txdr_hyper(va->va_size, tl);
1030 } else {
1031 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1032 *tl = nfs_false;
1033 }
1034 if (va->va_atime.tv_sec != VNOVAL) {
1035 if (va->va_atime.tv_sec != time_second) {
1036 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1037 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
1038 txdr_nfsv3time(&va->va_atime, tl);
1039 } else {
1040 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1041 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
1042 }
1043 } else {
1044 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1045 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
1046 }
1047 if (va->va_mtime.tv_sec != VNOVAL) {
1048 if (va->va_mtime.tv_sec != time_second) {
1049 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1050 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
1051 txdr_nfsv3time(&va->va_mtime, tl);
1052 } else {
1053 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1054 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
1055 }
1056 } else {
1057 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1058 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
1059 }
1060}