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 $");
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>
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);
392 rpc_vers = txdr_unsigned(RPC_VER2);
393 rpc_call = txdr_unsigned(RPC_CALL);
394 rpc_reply = txdr_unsigned(RPC_REPLY);
395 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
396 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
397 rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
398 rpc_autherr = txdr_unsigned(RPC_AUTHERR);
399 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
400 nfs_true = txdr_unsigned(TRUE);
401 nfs_false = txdr_unsigned(FALSE);
402 nfs_xdrneg1 = txdr_unsigned(-1);
403 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
404 if (nfs_ticks < 1)
405 nfs_ticks = 1;
406 /* Ensure async daemons disabled */
407 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
408 nfs_iodwant[i] = (struct proc *)0;
409 nfs_iodmount[i] = (struct nfsmount *)0;
410 }
411 nfs_nhinit(); /* Init the nfsnode table */
412
413 /*
414 * Initialize reply list and start timer
415 */
416 TAILQ_INIT(&nfs_reqq);
417
418 nfs_timer(0);
419
420 nfs_prev_nfsclnt_sy_narg = sysent[SYS_nfsclnt].sy_narg;
421 sysent[SYS_nfsclnt].sy_narg = 2;
422 nfs_prev_nfsclnt_sy_call = sysent[SYS_nfsclnt].sy_call;
423 sysent[SYS_nfsclnt].sy_call = (sy_call_t *)nfsclnt;
424
425 nfs_pbuf_freecnt = nswbuf / 2 + 1;
426
427 return (0);
428}
429
430int
431nfs_uninit(struct vfsconf *vfsp)
432{
433
434 untimeout(nfs_timer, (void *)NULL, nfs_timer_handle);
435 sysent[SYS_nfsclnt].sy_narg = nfs_prev_nfsclnt_sy_narg;
436 sysent[SYS_nfsclnt].sy_call = nfs_prev_nfsclnt_sy_call;
437 return (0);
438}
439
440/*
441 * Attribute cache routines.
442 * nfs_loadattrcache() - loads or updates the cache contents from attributes
443 * that are on the mbuf list
444 * nfs_getattrcache() - returns valid attributes if found in cache, returns
445 * error otherwise
446 */
447
448/*
449 * Load the attribute cache (that lives in the nfsnode entry) with
450 * the values on the mbuf list and
451 * Iff vap not NULL
452 * copy the attributes to *vaper
453 */
454int
455nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
456 struct vattr *vaper, int dontshrink)
457{
458 struct vnode *vp = *vpp;
459 struct vattr *vap;
460 struct nfs_fattr *fp;
461 struct nfsnode *np;
462 int32_t t1;
463 caddr_t cp2;
464 int rdev;
465 struct mbuf *md;
466 enum vtype vtyp;
467 u_short vmode;
468 struct timespec mtime;
469 int v3 = NFS_ISV3(vp);
470
471 md = *mdp;
472 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
473 cp2 = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1);
474 if (cp2 == NULL)
475 return EBADRPC;
476 fp = (struct nfs_fattr *)cp2;
477 if (v3) {
478 vtyp = nfsv3tov_type(fp->fa_type);
479 vmode = fxdr_unsigned(u_short, fp->fa_mode);
480 rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
481 fxdr_unsigned(int, fp->fa3_rdev.specdata2));
482 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
483 } else {
484 vtyp = nfsv2tov_type(fp->fa_type);
485 vmode = fxdr_unsigned(u_short, fp->fa_mode);
486 /*
487 * XXX
488 *
489 * The duplicate information returned in fa_type and fa_mode
490 * is an ambiguity in the NFS version 2 protocol.
491 *
492 * VREG should be taken literally as a regular file. If a
493 * server intents to return some type information differently
494 * in the upper bits of the mode field (e.g. for sockets, or
495 * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
496 * leave the examination of the mode bits even in the VREG
497 * case to avoid breakage for bogus servers, but we make sure
498 * that there are actually type bits set in the upper part of
499 * fa_mode (and failing that, trust the va_type field).
500 *
501 * NFSv3 cleared the issue, and requires fa_mode to not
502 * contain any type information (while also introduing sockets
503 * and FIFOs for fa_type).
504 */
505 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
506 vtyp = IFTOVT(vmode);
507 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
508 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
509
510 /*
511 * Really ugly NFSv2 kludge.
512 */
513 if (vtyp == VCHR && rdev == 0xffffffff)
514 vtyp = VFIFO;
515 }
516
517 /*
518 * If v_type == VNON it is a new node, so fill in the v_type,
519 * n_mtime fields. Check to see if it represents a special
520 * device, and if so, check for a possible alias. Once the
521 * correct vnode has been obtained, fill in the rest of the
522 * information.
523 */
524 np = VTONFS(vp);
525 if (vp->v_type != vtyp) {
526 vp->v_type = vtyp;
527 if (vp->v_type == VFIFO) {
528 vp->v_op = fifo_nfsv2nodeop_p;
529 }
530 if (vp->v_type == VCHR || vp->v_type == VBLK) {
531 vp->v_op = spec_nfsv2nodeop_p;
532 vp = addaliasu(vp, rdev);
533 np->n_vnode = vp;
534 }
535 np->n_mtime = mtime.tv_sec;
536 }
537 vap = &np->n_vattr;
538 vap->va_type = vtyp;
539 vap->va_mode = (vmode & 07777);
540 vap->va_rdev = rdev;
541 vap->va_mtime = mtime;
542 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
543 if (v3) {
544 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
545 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
546 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
547 vap->va_size = fxdr_hyper(&fp->fa3_size);
548 vap->va_blocksize = NFS_FABLKSIZE;
549 vap->va_bytes = fxdr_hyper(&fp->fa3_used);
550 vap->va_fileid = fxdr_unsigned(int32_t,
551 fp->fa3_fileid.nfsuquad[1]);
552 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
553 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
554 vap->va_flags = 0;
555 vap->va_filerev = 0;
556 } else {
557 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
558 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
559 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
560 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
561 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
562 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
563 * NFS_FABLKSIZE;
564 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
565 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
566 vap->va_flags = 0;
567 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
568 fp->fa2_ctime.nfsv2_sec);
569 vap->va_ctime.tv_nsec = 0;
570 vap->va_gen = fxdr_unsigned(u_int32_t, fp->fa2_ctime.nfsv2_usec);
571 vap->va_filerev = 0;
572 }
573 np->n_attrstamp = time_second;
574 if (vap->va_size != np->n_size) {
575 if (vap->va_type == VREG) {
576 if (dontshrink && vap->va_size < np->n_size) {
577 /*
578 * We've been told not to shrink the file;
579 * zero np->n_attrstamp to indicate that
580 * the attributes are stale.
581 */
582 vap->va_size = np->n_size;
583 np->n_attrstamp = 0;
584 } else if (np->n_flag & NMODIFIED) {
585 if (vap->va_size < np->n_size)
586 vap->va_size = np->n_size;
587 else
588 np->n_size = vap->va_size;
589 } else {
590 np->n_size = vap->va_size;
591 }
592 vnode_pager_setsize(vp, np->n_size);
593 } else {
594 np->n_size = vap->va_size;
595 }
596 }
597 if (vaper != NULL) {
598 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
599 if (np->n_flag & NCHG) {
600 if (np->n_flag & NACC)
601 vaper->va_atime = np->n_atim;
602 if (np->n_flag & NUPD)
603 vaper->va_mtime = np->n_mtim;
604 }
605 }
606 return (0);
607}
608
609#ifdef NFS_ACDEBUG
610#include <sys/sysctl.h>
611SYSCTL_DECL(_vfs_nfs);
612static int nfs_acdebug;
613SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
614#endif
615
616/*
617 * Check the time stamp
618 * If the cache is valid, copy contents to *vap and return 0
619 * otherwise return an error
620 */
621int
622nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
623{
624 struct nfsnode *np;
625 struct vattr *vap;
626 struct nfsmount *nmp;
627 int timeo;
628
629 np = VTONFS(vp);
630 vap = &np->n_vattr;
631 nmp = VFSTONFS(vp->v_mount);
632 /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
633 timeo = (time_second - np->n_mtime) / 10;
634
635#ifdef NFS_ACDEBUG
636 if (nfs_acdebug>1)
637 printf("nfs_getattrcache: initial timeo = %d\n", timeo);
638#endif
639
640 if (vap->va_type == VDIR) {
641 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
642 timeo = nmp->nm_acdirmin;
643 else if (timeo > nmp->nm_acdirmax)
644 timeo = nmp->nm_acdirmax;
645 } else {
646 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
647 timeo = nmp->nm_acregmin;
648 else if (timeo > nmp->nm_acregmax)
649 timeo = nmp->nm_acregmax;
650 }
651
652#ifdef NFS_ACDEBUG
653 if (nfs_acdebug > 2)
654 printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
655 nmp->nm_acregmin, nmp->nm_acregmax,
656 nmp->nm_acdirmin, nmp->nm_acdirmax);
657
658 if (nfs_acdebug)
659 printf("nfs_getattrcache: age = %d; final timeo = %d\n",
660 (time_second - np->n_attrstamp), timeo);
661#endif
662
663 if ((time_second - np->n_attrstamp) >= timeo) {
664 nfsstats.attrcache_misses++;
665 return (ENOENT);
666 }
667 nfsstats.attrcache_hits++;
668 if (vap->va_size != np->n_size) {
669 if (vap->va_type == VREG) {
670 if (np->n_flag & NMODIFIED) {
671 if (vap->va_size < np->n_size)
672 vap->va_size = np->n_size;
673 else
674 np->n_size = vap->va_size;
675 } else {
676 np->n_size = vap->va_size;
677 }
678 vnode_pager_setsize(vp, np->n_size);
679 } else {
680 np->n_size = vap->va_size;
681 }
682 }
683 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
684 if (np->n_flag & NCHG) {
685 if (np->n_flag & NACC)
686 vaper->va_atime = np->n_atim;
687 if (np->n_flag & NUPD)
688 vaper->va_mtime = np->n_mtim;
689 }
690 return (0);
691}
692
693static nfsuint64 nfs_nullcookie = { { 0, 0 } };
694/*
695 * This function finds the directory cookie that corresponds to the
696 * logical byte offset given.
697 */
698nfsuint64 *
699nfs_getcookie(struct nfsnode *np, off_t off, int add)
700{
701 struct nfsdmap *dp, *dp2;
702 int pos;
703
704 pos = (uoff_t)off / NFS_DIRBLKSIZ;
705 if (pos == 0 || off < 0) {
706#ifdef DIAGNOSTIC
707 if (add)
708 panic("nfs getcookie add at <= 0");
709#endif
710 return (&nfs_nullcookie);
711 }
712 pos--;
713 dp = LIST_FIRST(&np->n_cookies);
714 if (!dp) {
715 if (add) {
716 MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
717 M_NFSDIROFF, M_WAITOK);
718 dp->ndm_eocookie = 0;
719 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
720 } else
721 return ((nfsuint64 *)0);
722 }
723 while (pos >= NFSNUMCOOKIES) {
724 pos -= NFSNUMCOOKIES;
725 if (LIST_NEXT(dp, ndm_list)) {
726 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
727 pos >= dp->ndm_eocookie)
728 return ((nfsuint64 *)0);
729 dp = LIST_NEXT(dp, ndm_list);
730 } else if (add) {
731 MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
732 M_NFSDIROFF, M_WAITOK);
733 dp2->ndm_eocookie = 0;
734 LIST_INSERT_AFTER(dp, dp2, ndm_list);
735 dp = dp2;
736 } else
737 return ((nfsuint64 *)0);
738 }
739 if (pos >= dp->ndm_eocookie) {
740 if (add)
741 dp->ndm_eocookie = pos + 1;
742 else
743 return ((nfsuint64 *)0);
744 }
745 return (&dp->ndm_cookies[pos]);
746}
747
748/*
749 * Invalidate cached directory information, except for the actual directory
750 * blocks (which are invalidated separately).
751 * Done mainly to avoid the use of stale offset cookies.
752 */
753void
754nfs_invaldir(struct vnode *vp)
755{
756 struct nfsnode *np = VTONFS(vp);
757
758#ifdef DIAGNOSTIC
759 if (vp->v_type != VDIR)
760 panic("nfs: invaldir not dir");
761#endif
762 np->n_direofoffset = 0;
763 np->n_cookieverf.nfsuquad[0] = 0;
764 np->n_cookieverf.nfsuquad[1] = 0;
765 if (LIST_FIRST(&np->n_cookies))
766 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
767}
768
769/*
770 * The write verifier has changed (probably due to a server reboot), so all
771 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
772 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
773 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the
774 * mount point.
775 *
776 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
777 * writes are not clusterable.
778 */
779void
780nfs_clearcommit(struct mount *mp)
781{
782 struct vnode *vp, *nvp;
783 struct buf *bp, *nbp;
784 int s;
785
786 GIANT_REQUIRED;
787
788 s = splbio();
789 mtx_lock(&mntvnode_mtx);
790loop:
791 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp; vp = nvp) {
792 if (vp->v_mount != mp) /* Paranoia */
793 goto loop;
794 nvp = TAILQ_NEXT(vp, v_nmntvnodes);
795 for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
796 nbp = TAILQ_NEXT(bp, b_vnbufs);
797 if (BUF_REFCNT(bp) == 0 &&
798 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
799 == (B_DELWRI | B_NEEDCOMMIT))
800 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
801 }
802 }
803 mtx_unlock(&mntvnode_mtx);
804 splx(s);
805}
806
807/*
808 * Helper functions for former macros. Some of these should be
809 * moved to their callers.
810 */
811
812int
813nfsm_mtofh_xx(struct vnode *d, struct vnode **v, int v3, int *f,
814 struct mbuf **md, caddr_t *dpos)
815{
816 struct nfsnode *ttnp;
817 struct vnode *ttvp;
818 nfsfh_t *ttfhp;
819 u_int32_t *tl;
820 int ttfhsize;
821 int t1;
822
823 if (v3) {
824 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
825 if (tl == NULL)
826 return EBADRPC;
827 *f = fxdr_unsigned(int, *tl);
828 } else
829 *f = 1;
830 if (*f) {
831 t1 = nfsm_getfh_xx(&ttfhp, &ttfhsize, (v3), md, dpos);
832 if (t1 != 0)
833 return t1;
834 t1 = nfs_nget(d->v_mount, ttfhp, ttfhsize, &ttnp);
835 if (t1 != 0)
836 return t1;
837 *v = NFSTOV(ttnp);
838 }
839 if (v3) {
840 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
841 if (tl == NULL)
842 return EBADRPC;
843 if (*f)
844 *f = fxdr_unsigned(int, *tl);
845 else if (fxdr_unsigned(int, *tl))
846 nfsm_adv_xx(NFSX_V3FATTR, md, dpos);
847 }
848 if (*f) {
849 ttvp = *v;
850 t1 = nfs_loadattrcache(&ttvp, md, dpos, (struct vattr *)0, 0);
851 if (t1)
852 return t1;
853 *v = ttvp;
854 }
855 return 0;
856}
857
858int
859nfsm_getfh_xx(nfsfh_t **f, int *s, int v3, struct mbuf **md, caddr_t *dpos)
860{
861 u_int32_t *tl;
862
863 if (v3) {
864 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
865 if (tl == NULL)
866 return EBADRPC;
867 *s = fxdr_unsigned(int, *tl);
868 if (*s <= 0 || *s > NFSX_V3FHMAX)
869 return EBADRPC;
870 } else
871 *s = NFSX_V2FH;
872 *f = nfsm_dissect_xx(nfsm_rndup(*s), md, dpos);
873 if (*f == NULL)
874 return EBADRPC;
875 else
876 return 0;
877}
878
879
880int
881nfsm_loadattr_xx(struct vnode **v, struct vattr *va, struct mbuf **md,
882 caddr_t *dpos)
883{
884 int t1;
885
886 struct vnode *ttvp = *v;
887 t1 = nfs_loadattrcache(&ttvp, md, dpos, va, 0);
888 if (t1 != 0)
889 return t1;
890 *v = ttvp;
891 return 0;
892}
893
894int
895nfsm_postop_attr_xx(struct vnode **v, int *f, struct mbuf **md,
896 caddr_t *dpos)
897{
898 u_int32_t *tl;
899 int t1;
900
901 struct vnode *ttvp = *v;
902 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
903 if (tl == NULL)
904 return EBADRPC;
905 *f = fxdr_unsigned(int, *tl);
906 if (*f != 0) {
907 t1 = nfs_loadattrcache(&ttvp, md, dpos, (struct vattr *)0, 1);
908 if (t1 != 0) {
909 *f = 0;
910 return t1;
911 }
912 *v = ttvp;
913 }
914 return 0;
915}
916
917int
918nfsm_wcc_data_xx(struct vnode **v, int *f, struct mbuf **md, caddr_t *dpos)
919{
920 u_int32_t *tl;
921 int ttattrf, ttretf = 0;
922 int t1;
923
924 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
925 if (tl == NULL)
926 return EBADRPC;
927 if (*tl == nfs_true) {
928 tl = nfsm_dissect_xx(6 * NFSX_UNSIGNED, md, dpos);
929 if (tl == NULL)
930 return EBADRPC;
931 if (*f)
932 ttretf = (VTONFS(*v)->n_mtime ==
933 fxdr_unsigned(u_int32_t, *(tl + 2)));
934 }
935 t1 = nfsm_postop_attr_xx(v, &ttattrf, md, dpos);
936 if (t1)
937 return t1;
938 if (*f)
939 *f = ttretf;
940 else
941 *f = ttattrf;
942 return 0;
943}
944
945int
946nfsm_strtom_xx(const char *a, int s, int m, struct mbuf **mb, caddr_t *bpos)
947{
948 u_int32_t *tl;
949 int t1;
950
951 if (s > m)
952 return ENAMETOOLONG;
953 t1 = nfsm_rndup(s) + NFSX_UNSIGNED;
954 if (t1 <= M_TRAILINGSPACE(*mb)) {
955 tl = nfsm_build_xx(t1, mb, bpos);
956 *tl++ = txdr_unsigned(s);
957 *(tl + ((t1 >> 2) - 2)) = 0;
958 bcopy(a, tl, s);
959 } else {
960 t1 = nfsm_strtmbuf(mb, bpos, a, s);
961 if (t1 != 0)
962 return t1;
963 }
964 return 0;
965}
966
967int
968nfsm_fhtom_xx(struct vnode *v, int v3, struct mbuf **mb, caddr_t *bpos)
969{
970 u_int32_t *tl;
971 int t1;
972 caddr_t cp;
973
974 if (v3) {
975 t1 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED;
976 if (t1 < M_TRAILINGSPACE(*mb)) {
977 tl = nfsm_build_xx(t1, mb, bpos);
978 *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize);
979 *(tl + ((t1 >> 2) - 2)) = 0;
980 bcopy(VTONFS(v)->n_fhp, tl, VTONFS(v)->n_fhsize);
981 } else {
982 t1 = nfsm_strtmbuf(mb, bpos,
983 (const char *)VTONFS(v)->n_fhp,
984 VTONFS(v)->n_fhsize);
985 if (t1 != 0)
986 return t1;
987 }
988 } else {
989 cp = nfsm_build_xx(NFSX_V2FH, mb, bpos);
990 bcopy(VTONFS(v)->n_fhp, cp, NFSX_V2FH);
991 }
992 return 0;
993}
994
995void
996nfsm_v3attrbuild_xx(struct vattr *va, int full, struct mbuf **mb,
997 caddr_t *bpos)
998{
999 u_int32_t *tl;
1000
1001 if (va->va_mode != (mode_t)VNOVAL) {
1002 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1003 *tl++ = nfs_true;
1004 *tl = txdr_unsigned(va->va_mode);
1005 } else {
1006 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1007 *tl = nfs_false;
1008 }
1009 if (full && va->va_uid != (uid_t)VNOVAL) {
1010 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1011 *tl++ = nfs_true;
1012 *tl = txdr_unsigned(va->va_uid);
1013 } else {
1014 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1015 *tl = nfs_false;
1016 }
1017 if (full && va->va_gid != (gid_t)VNOVAL) {
1018 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1019 *tl++ = nfs_true;
1020 *tl = txdr_unsigned(va->va_gid);
1021 } else {
1022 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1023 *tl = nfs_false;
1024 }
1025 if (full && va->va_size != VNOVAL) {
1026 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1027 *tl++ = nfs_true;
1028 txdr_hyper(va->va_size, tl);
1029 } else {
1030 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1031 *tl = nfs_false;
1032 }
1033 if (va->va_atime.tv_sec != VNOVAL) {
1034 if (va->va_atime.tv_sec != time_second) {
1035 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1036 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
1037 txdr_nfsv3time(&va->va_atime, tl);
1038 } else {
1039 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1040 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
1041 }
1042 } else {
1043 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1044 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
1045 }
1046 if (va->va_mtime.tv_sec != VNOVAL) {
1047 if (va->va_mtime.tv_sec != time_second) {
1048 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1049 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
1050 txdr_nfsv3time(&va->va_mtime, tl);
1051 } else {
1052 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1053 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
1054 }
1055 } else {
1056 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1057 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
1058 }
1059}