[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ]

FreeBSD/Linux Kernel Cross Reference
sys/nfsclient/nfs_subs.c

Version: -  FREEBSD  -  FREEBSD7  -  FREEBSD70  -  FREEBSD6  -  FREEBSD63  -  FREEBSD62  -  FREEBSD61  -  FREEBSD60  -  FREEBSD5  -  FREEBSD55  -  FREEBSD54  -  FREEBSD53  -  FREEBSD52  -  FREEBSD51  -  FREEBSD50  -  FREEBSD4  -  FREEBSD3  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  OPENSOLARIS  -  minix-3-1-1  -  TRUSTEDBSD-SEBSD  -  FREEBSD-LIBC  -  FREEBSD7-LIBC  -  FREEBSD6-LIBC  -  GLIBC27 
SearchContext: -  none  -  excerpts  -  bigexcerpts 

  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  * 4. Neither the name of the University nor the names of its contributors
 17  *    may be used to endorse or promote products derived from this software
 18  *    without specific prior written permission.
 19  *
 20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 30  * SUCH DAMAGE.
 31  *
 32  *      @(#)nfs_subs.c  8.8 (Berkeley) 5/22/95
 33  */
 34 
 35 #include <sys/cdefs.h>
 36 __FBSDID("$FreeBSD: src/sys/nfsclient/nfs_subs.c,v 1.158 2008/11/03 10:38:00 dfr Exp $");
 37 
 38 /*
 39  * These functions support the macros and help fiddle mbuf chains for
 40  * the nfs op functions. They do things like create the rpc header and
 41  * copy data between mbuf chains and uio lists.
 42  */
 43 
 44 #include <sys/param.h>
 45 #include <sys/systm.h>
 46 #include <sys/kernel.h>
 47 #include <sys/bio.h>
 48 #include <sys/buf.h>
 49 #include <sys/proc.h>
 50 #include <sys/mount.h>
 51 #include <sys/vnode.h>
 52 #include <sys/namei.h>
 53 #include <sys/mbuf.h>
 54 #include <sys/socket.h>
 55 #include <sys/stat.h>
 56 #include <sys/malloc.h>
 57 #include <sys/sysent.h>
 58 #include <sys/syscall.h>
 59 #include <sys/sysproto.h>
 60 
 61 #include <vm/vm.h>
 62 #include <vm/vm_object.h>
 63 #include <vm/vm_extern.h>
 64 #include <vm/uma.h>
 65 
 66 #include <rpc/rpcclnt.h>
 67 
 68 #include <nfs/rpcv2.h>
 69 #include <nfs/nfsproto.h>
 70 #include <nfsclient/nfs.h>
 71 #include <nfsclient/nfsnode.h>
 72 #include <nfs/xdr_subs.h>
 73 #include <nfsclient/nfsm_subs.h>
 74 #include <nfsclient/nfsmount.h>
 75 
 76 #include <netinet/in.h>
 77 
 78 /*
 79  * Note that stdarg.h and the ANSI style va_start macro is used for both
 80  * ANSI and traditional C compilers.
 81  */
 82 #include <machine/stdarg.h>
 83 
 84 /*
 85  * Data items converted to xdr at startup, since they are constant
 86  * This is kinda hokey, but may save a little time doing byte swaps
 87  */
 88 u_int32_t       nfs_xdrneg1;
 89 u_int32_t       rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
 90                     rpc_mismatch, rpc_auth_unix, rpc_msgaccepted;
 91 u_int32_t       nfs_true, nfs_false;
 92 
 93 /* And other global data */
 94 static u_int32_t nfs_xid = 0;
 95 static enum vtype nv2tov_type[8]= {
 96         VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON,  VNON
 97 };
 98 
 99 int             nfs_ticks;
100 int             nfs_pbuf_freecnt = -1;  /* start out unlimited */
101 
102 #ifdef NFS_LEGACYRPC
103 struct nfs_reqq nfs_reqq;
104 struct mtx nfs_reqq_mtx;
105 #endif
106 struct nfs_bufq nfs_bufq;
107 static struct mtx nfs_xid_mtx;
108 
109 /*
110  * and the reverse mapping from generic to Version 2 procedure numbers
111  */
112 int nfsv2_procid[NFS_NPROCS] = {
113         NFSV2PROC_NULL,
114         NFSV2PROC_GETATTR,
115         NFSV2PROC_SETATTR,
116         NFSV2PROC_LOOKUP,
117         NFSV2PROC_NOOP,
118         NFSV2PROC_READLINK,
119         NFSV2PROC_READ,
120         NFSV2PROC_WRITE,
121         NFSV2PROC_CREATE,
122         NFSV2PROC_MKDIR,
123         NFSV2PROC_SYMLINK,
124         NFSV2PROC_CREATE,
125         NFSV2PROC_REMOVE,
126         NFSV2PROC_RMDIR,
127         NFSV2PROC_RENAME,
128         NFSV2PROC_LINK,
129         NFSV2PROC_READDIR,
130         NFSV2PROC_NOOP,
131         NFSV2PROC_STATFS,
132         NFSV2PROC_NOOP,
133         NFSV2PROC_NOOP,
134         NFSV2PROC_NOOP,
135         NFSV2PROC_NOOP,
136 };
137 
138 LIST_HEAD(nfsnodehashhead, nfsnode);
139 
140 u_int32_t
141 nfs_xid_gen(void)
142 {
143         uint32_t xid;
144 
145         mtx_lock(&nfs_xid_mtx);
146 
147         /* Get a pretty random xid to start with */
148         if (!nfs_xid)
149                 nfs_xid = random();
150         /*
151          * Skip zero xid if it should ever happen.
152          */
153         if (++nfs_xid == 0)
154                 nfs_xid++;
155         xid = nfs_xid;
156         mtx_unlock(&nfs_xid_mtx);
157         return xid;
158 }
159 
160 /*
161  * Create the header for an rpc request packet
162  * The hsiz is the size of the rest of the nfs request header.
163  * (just used to decide if a cluster is a good idea)
164  */
165 struct mbuf *
166 nfsm_reqhead(struct vnode *vp, u_long procid, int hsiz)
167 {
168         struct mbuf *mb;
169 
170         MGET(mb, M_WAIT, MT_DATA);
171         if (hsiz >= MINCLSIZE)
172                 MCLGET(mb, M_WAIT);
173         mb->m_len = 0;
174         return (mb);
175 }
176 
177 /*
178  * Build the RPC header and fill in the authorization info.
179  * The authorization string argument is only used when the credentials
180  * come from outside of the kernel.
181  * Returns the head of the mbuf list.
182  */
183 struct mbuf *
184 nfsm_rpchead(struct ucred *cr, int nmflag, int procid, int auth_type,
185     int auth_len, struct mbuf *mrest, int mrest_len, struct mbuf **mbp,
186     u_int32_t **xidpp)
187 {
188         struct mbuf *mb;
189         u_int32_t *tl;
190         caddr_t bpos;
191         int i;
192         struct mbuf *mreq;
193         int grpsiz, authsiz;
194 
195         authsiz = nfsm_rndup(auth_len);
196         MGETHDR(mb, M_WAIT, MT_DATA);
197         if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
198                 MCLGET(mb, M_WAIT);
199         } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
200                 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
201         } else {
202                 MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
203         }
204         mb->m_len = 0;
205         mreq = mb;
206         bpos = mtod(mb, caddr_t);
207 
208         /*
209          * First the RPC header.
210          */
211         tl = nfsm_build(u_int32_t *, 8 * NFSX_UNSIGNED);
212 
213         *xidpp = tl;
214         *tl++ = txdr_unsigned(nfs_xid_gen());
215         *tl++ = rpc_call;
216         *tl++ = rpc_vers;
217         *tl++ = txdr_unsigned(NFS_PROG);
218         if (nmflag & NFSMNT_NFSV3) {
219                 *tl++ = txdr_unsigned(NFS_VER3);
220                 *tl++ = txdr_unsigned(procid);
221         } else {
222                 *tl++ = txdr_unsigned(NFS_VER2);
223                 *tl++ = txdr_unsigned(nfsv2_procid[procid]);
224         }
225 
226         /*
227          * And then the authorization cred.
228          */
229         *tl++ = txdr_unsigned(auth_type);
230         *tl = txdr_unsigned(authsiz);
231         switch (auth_type) {
232         case RPCAUTH_UNIX:
233                 tl = nfsm_build(u_int32_t *, auth_len);
234                 *tl++ = 0;              /* stamp ?? */
235                 *tl++ = 0;              /* NULL hostname */
236                 *tl++ = txdr_unsigned(cr->cr_uid);
237                 *tl++ = txdr_unsigned(cr->cr_groups[0]);
238                 grpsiz = (auth_len >> 2) - 5;
239                 *tl++ = txdr_unsigned(grpsiz);
240                 for (i = 1; i <= grpsiz; i++)
241                         *tl++ = txdr_unsigned(cr->cr_groups[i]);
242                 break;
243         }
244 
245         /*
246          * And the verifier...
247          */
248         tl = nfsm_build(u_int32_t *, 2 * NFSX_UNSIGNED);
249         *tl++ = txdr_unsigned(RPCAUTH_NULL);
250         *tl = 0;
251         mb->m_next = mrest;
252         mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
253         mreq->m_pkthdr.rcvif = NULL;
254         *mbp = mb;
255         return (mreq);
256 }
257 
258 /*
259  * copies a uio scatter/gather list to an mbuf chain.
260  * NOTE: can ony handle iovcnt == 1
261  */
262 int
263 nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
264 {
265         char *uiocp;
266         struct mbuf *mp, *mp2;
267         int xfer, left, mlen;
268         int uiosiz, clflg, rem;
269         char *cp;
270 
271 #ifdef DIAGNOSTIC
272         if (uiop->uio_iovcnt != 1)
273                 panic("nfsm_uiotombuf: iovcnt != 1");
274 #endif
275 
276         if (siz > MLEN)         /* or should it >= MCLBYTES ?? */
277                 clflg = 1;
278         else
279                 clflg = 0;
280         rem = nfsm_rndup(siz)-siz;
281         mp = mp2 = *mq;
282         while (siz > 0) {
283                 left = uiop->uio_iov->iov_len;
284                 uiocp = uiop->uio_iov->iov_base;
285                 if (left > siz)
286                         left = siz;
287                 uiosiz = left;
288                 while (left > 0) {
289                         mlen = M_TRAILINGSPACE(mp);
290                         if (mlen == 0) {
291                                 MGET(mp, M_WAIT, MT_DATA);
292                                 if (clflg)
293                                         MCLGET(mp, M_WAIT);
294                                 mp->m_len = 0;
295                                 mp2->m_next = mp;
296                                 mp2 = mp;
297                                 mlen = M_TRAILINGSPACE(mp);
298                         }
299                         xfer = (left > mlen) ? mlen : left;
300 #ifdef notdef
301                         /* Not Yet.. */
302                         if (uiop->uio_iov->iov_op != NULL)
303                                 (*(uiop->uio_iov->iov_op))
304                                 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
305                         else
306 #endif
307                         if (uiop->uio_segflg == UIO_SYSSPACE)
308                                 bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
309                         else
310                                 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
311                         mp->m_len += xfer;
312                         left -= xfer;
313                         uiocp += xfer;
314                         uiop->uio_offset += xfer;
315                         uiop->uio_resid -= xfer;
316                 }
317                 uiop->uio_iov->iov_base =
318                     (char *)uiop->uio_iov->iov_base + uiosiz;
319                 uiop->uio_iov->iov_len -= uiosiz;
320                 siz -= uiosiz;
321         }
322         if (rem > 0) {
323                 if (rem > M_TRAILINGSPACE(mp)) {
324                         MGET(mp, M_WAIT, MT_DATA);
325                         mp->m_len = 0;
326                         mp2->m_next = mp;
327                 }
328                 cp = mtod(mp, caddr_t)+mp->m_len;
329                 for (left = 0; left < rem; left++)
330                         *cp++ = '\0';
331                 mp->m_len += rem;
332                 *bpos = cp;
333         } else
334                 *bpos = mtod(mp, caddr_t)+mp->m_len;
335         *mq = mp;
336         return (0);
337 }
338 
339 /*
340  * Copy a string into mbufs for the hard cases...
341  */
342 int
343 nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
344 {
345         struct mbuf *m1 = NULL, *m2;
346         long left, xfer, len, tlen;
347         u_int32_t *tl;
348         int putsize;
349 
350         putsize = 1;
351         m2 = *mb;
352         left = M_TRAILINGSPACE(m2);
353         if (left > 0) {
354                 tl = ((u_int32_t *)(*bpos));
355                 *tl++ = txdr_unsigned(siz);
356                 putsize = 0;
357                 left -= NFSX_UNSIGNED;
358                 m2->m_len += NFSX_UNSIGNED;
359                 if (left > 0) {
360                         bcopy(cp, (caddr_t) tl, left);
361                         siz -= left;
362                         cp += left;
363                         m2->m_len += left;
364                         left = 0;
365                 }
366         }
367         /* Loop around adding mbufs */
368         while (siz > 0) {
369                 MGET(m1, M_WAIT, MT_DATA);
370                 if (siz > MLEN)
371                         MCLGET(m1, M_WAIT);
372                 m1->m_len = NFSMSIZ(m1);
373                 m2->m_next = m1;
374                 m2 = m1;
375                 tl = mtod(m1, u_int32_t *);
376                 tlen = 0;
377                 if (putsize) {
378                         *tl++ = txdr_unsigned(siz);
379                         m1->m_len -= NFSX_UNSIGNED;
380                         tlen = NFSX_UNSIGNED;
381                         putsize = 0;
382                 }
383                 if (siz < m1->m_len) {
384                         len = nfsm_rndup(siz);
385                         xfer = siz;
386                         if (xfer < len)
387                                 *(tl+(xfer>>2)) = 0;
388                 } else {
389                         xfer = len = m1->m_len;
390                 }
391                 bcopy(cp, (caddr_t) tl, xfer);
392                 m1->m_len = len+tlen;
393                 siz -= xfer;
394                 cp += xfer;
395         }
396         *mb = m1;
397         *bpos = mtod(m1, caddr_t)+m1->m_len;
398         return (0);
399 }
400 
401 /*
402  * Called once to initialize data structures...
403  */
404 int
405 nfs_init(struct vfsconf *vfsp)
406 {
407         int i;
408 
409         nfsmount_zone = uma_zcreate("NFSMOUNT", sizeof(struct nfsmount),
410             NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
411         rpc_vers = txdr_unsigned(RPC_VER2);
412         rpc_call = txdr_unsigned(RPC_CALL);
413         rpc_reply = txdr_unsigned(RPC_REPLY);
414         rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
415         rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
416         rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
417         rpc_autherr = txdr_unsigned(RPC_AUTHERR);
418         rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
419         nfs_true = txdr_unsigned(TRUE);
420         nfs_false = txdr_unsigned(FALSE);
421         nfs_xdrneg1 = txdr_unsigned(-1);
422         nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
423         if (nfs_ticks < 1)
424                 nfs_ticks = 1;
425         /* Ensure async daemons disabled */
426         for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
427                 nfs_iodwant[i] = NULL;
428                 nfs_iodmount[i] = NULL;
429         }
430         nfs_nhinit();                   /* Init the nfsnode table */
431 
432         /*
433          * Initialize reply list and start timer
434          */
435 #ifdef NFS_LEGACYRPC
436         TAILQ_INIT(&nfs_reqq);
437         mtx_init(&nfs_reqq_mtx, "NFS reqq lock", NULL, MTX_DEF);
438         callout_init(&nfs_callout, CALLOUT_MPSAFE);
439 #endif
440         mtx_init(&nfs_iod_mtx, "NFS iod lock", NULL, MTX_DEF);
441         mtx_init(&nfs_xid_mtx, "NFS xid lock", NULL, MTX_DEF);
442 
443         nfs_pbuf_freecnt = nswbuf / 2 + 1;
444 
445         return (0);
446 }
447 
448 int
449 nfs_uninit(struct vfsconf *vfsp)
450 {
451         int i;
452 
453 #ifdef NFS_LEGACYRPC
454         callout_stop(&nfs_callout);
455 
456         KASSERT(TAILQ_EMPTY(&nfs_reqq),
457             ("nfs_uninit: request queue not empty"));
458 #endif
459 
460         /*
461          * Tell all nfsiod processes to exit. Clear nfs_iodmax, and wakeup
462          * any sleeping nfsiods so they check nfs_iodmax and exit.
463          */
464         mtx_lock(&nfs_iod_mtx);
465         nfs_iodmax = 0;
466         for (i = 0; i < nfs_numasync; i++)
467                 if (nfs_iodwant[i])
468                         wakeup(&nfs_iodwant[i]);
469         /* The last nfsiod to exit will wake us up when nfs_numasync hits 0 */
470         while (nfs_numasync)
471                 msleep(&nfs_numasync, &nfs_iod_mtx, PWAIT, "ioddie", 0);
472         mtx_unlock(&nfs_iod_mtx);
473         nfs_nhuninit();
474         uma_zdestroy(nfsmount_zone);
475         return (0);
476 }
477 
478 void 
479 nfs_dircookie_lock(struct nfsnode *np)
480 {
481         mtx_lock(&np->n_mtx);
482         while (np->n_flag & NDIRCOOKIELK)
483                 (void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
484         np->n_flag |= NDIRCOOKIELK;
485         mtx_unlock(&np->n_mtx);
486 }
487 
488 void 
489 nfs_dircookie_unlock(struct nfsnode *np)
490 {
491         mtx_lock(&np->n_mtx);
492         np->n_flag &= ~NDIRCOOKIELK;
493         wakeup(&np->n_flag);
494         mtx_unlock(&np->n_mtx);
495 }
496 
497 int
498 nfs_upgrade_vnlock(struct vnode *vp)
499 {
500         int old_lock;
501         
502         if ((old_lock = VOP_ISLOCKED(vp)) != LK_EXCLUSIVE) {
503                 if (old_lock == LK_SHARED) {
504                         /* Upgrade to exclusive lock, this might block */
505                         vn_lock(vp, LK_UPGRADE | LK_RETRY);
506                 } else {
507                         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
508                 }
509         }
510         return old_lock;
511 }
512 
513 void
514 nfs_downgrade_vnlock(struct vnode *vp, int old_lock)
515 {
516         if (old_lock != LK_EXCLUSIVE) {
517                 if (old_lock == LK_SHARED) {
518                         /* Downgrade from exclusive lock, this might block */
519                         vn_lock(vp, LK_DOWNGRADE);
520                 } else {
521                         VOP_UNLOCK(vp, 0);
522                 }
523         }
524 }
525 
526 void
527 nfs_printf(const char *fmt, ...)
528 {
529         va_list ap;
530 
531         mtx_lock(&Giant);
532         va_start(ap, fmt);
533         printf(fmt, ap);
534         va_end(ap);
535         mtx_unlock(&Giant);
536 }
537 
538 /*
539  * Attribute cache routines.
540  * nfs_loadattrcache() - loads or updates the cache contents from attributes
541  *      that are on the mbuf list
542  * nfs_getattrcache() - returns valid attributes if found in cache, returns
543  *      error otherwise
544  */
545 
546 /*
547  * Load the attribute cache (that lives in the nfsnode entry) with
548  * the values on the mbuf list and
549  * Iff vap not NULL
550  *    copy the attributes to *vaper
551  */
552 int
553 nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
554                   struct vattr *vaper, int dontshrink)
555 {
556         struct vnode *vp = *vpp;
557         struct vattr *vap;
558         struct nfs_fattr *fp;
559         struct nfsnode *np;
560         int32_t t1;
561         caddr_t cp2;
562         int rdev;
563         struct mbuf *md;
564         enum vtype vtyp;
565         u_short vmode;
566         struct timespec mtime, mtime_save;
567         int v3 = NFS_ISV3(vp);
568         struct thread *td = curthread;
569 
570         md = *mdp;
571         t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
572         cp2 = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, M_WAIT);
573         if (cp2 == NULL)
574                 return EBADRPC;
575         fp = (struct nfs_fattr *)cp2;
576         if (v3) {
577                 vtyp = nfsv3tov_type(fp->fa_type);
578                 vmode = fxdr_unsigned(u_short, fp->fa_mode);
579                 rdev = makedev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
580                         fxdr_unsigned(int, fp->fa3_rdev.specdata2));
581                 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
582         } else {
583                 vtyp = nfsv2tov_type(fp->fa_type);
584                 vmode = fxdr_unsigned(u_short, fp->fa_mode);
585                 /*
586                  * XXX
587                  *
588                  * The duplicate information returned in fa_type and fa_mode
589                  * is an ambiguity in the NFS version 2 protocol.
590                  *
591                  * VREG should be taken literally as a regular file.  If a
592                  * server intents to return some type information differently
593                  * in the upper bits of the mode field (e.g. for sockets, or
594                  * FIFOs), NFSv2 mandates fa_type to be VNON.  Anyway, we
595                  * leave the examination of the mode bits even in the VREG
596                  * case to avoid breakage for bogus servers, but we make sure
597                  * that there are actually type bits set in the upper part of
598                  * fa_mode (and failing that, trust the va_type field).
599                  *
600                  * NFSv3 cleared the issue, and requires fa_mode to not
601                  * contain any type information (while also introduing sockets
602                  * and FIFOs for fa_type).
603                  */
604                 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
605                         vtyp = IFTOVT(vmode);
606                 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
607                 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
608 
609                 /*
610                  * Really ugly NFSv2 kludge.
611                  */
612                 if (vtyp == VCHR && rdev == 0xffffffff)
613                         vtyp = VFIFO;
614         }
615 
616         /*
617          * If v_type == VNON it is a new node, so fill in the v_type,
618          * n_mtime fields. Check to see if it represents a special
619          * device, and if so, check for a possible alias. Once the
620          * correct vnode has been obtained, fill in the rest of the
621          * information.
622          */
623         np = VTONFS(vp);
624         mtx_lock(&np->n_mtx);
625         if (vp->v_type != vtyp) {
626                 vp->v_type = vtyp;
627                 if (vp->v_type == VFIFO)
628                         vp->v_op = &nfs_fifoops;
629                 np->n_mtime = mtime;
630         }
631         vap = &np->n_vattr;
632         vap->va_type = vtyp;
633         vap->va_mode = (vmode & 07777);
634         vap->va_rdev = rdev;
635         mtime_save = vap->va_mtime;
636         vap->va_mtime = mtime;
637         vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
638         if (v3) {
639                 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
640                 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
641                 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
642                 vap->va_size = fxdr_hyper(&fp->fa3_size);
643                 vap->va_blocksize = NFS_FABLKSIZE;
644                 vap->va_bytes = fxdr_hyper(&fp->fa3_used);
645                 vap->va_fileid = fxdr_unsigned(int32_t,
646                     fp->fa3_fileid.nfsuquad[1]);
647                 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
648                 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
649                 vap->va_flags = 0;
650                 vap->va_filerev = 0;
651         } else {
652                 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
653                 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
654                 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
655                 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
656                 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
657                 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
658                     * NFS_FABLKSIZE;
659                 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
660                 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
661                 vap->va_flags = 0;
662                 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
663                     fp->fa2_ctime.nfsv2_sec);
664                 vap->va_ctime.tv_nsec = 0;
665                 vap->va_gen = fxdr_unsigned(u_int32_t, fp->fa2_ctime.nfsv2_usec);
666                 vap->va_filerev = 0;
667         }
668         np->n_attrstamp = time_second;
669         /* Timestamp the NFS otw getattr fetch */
670         if (td->td_proc) {
671                 np->n_ac_ts_tid = td->td_tid;
672                 np->n_ac_ts_pid = td->td_proc->p_pid;
673                 np->n_ac_ts_syscalls = td->td_syscalls;
674         } else
675                 bzero(&np->n_ac_ts, sizeof(struct nfs_attrcache_timestamp));
676         
677         if (vap->va_size != np->n_size) {
678                 if (vap->va_type == VREG) {
679                         if (dontshrink && vap->va_size < np->n_size) {
680                                 /*
681                                  * We've been told not to shrink the file;
682                                  * zero np->n_attrstamp to indicate that
683                                  * the attributes are stale.
684                                  */
685                                 vap->va_size = np->n_size;
686                                 np->n_attrstamp = 0;
687                         } else if (np->n_flag & NMODIFIED) {
688                                 /*
689                                  * We've modified the file: Use the larger
690                                  * of our size, and the server's size.
691                                  */
692                                 if (vap->va_size < np->n_size) {
693                                         vap->va_size = np->n_size;
694                                 } else {
695                                         np->n_size = vap->va_size;
696                                         np->n_flag |= NSIZECHANGED;
697                                 }
698                         } else {
699                                 np->n_size = vap->va_size;
700                                 np->n_flag |= NSIZECHANGED;
701                         }
702                         vnode_pager_setsize(vp, np->n_size);
703                 } else {
704                         np->n_size = vap->va_size;
705                 }
706         }
707         /*
708          * The following checks are added to prevent a race between (say)
709          * a READDIR+ and a WRITE. 
710          * READDIR+, WRITE requests sent out.
711          * READDIR+ resp, WRITE resp received on client.
712          * However, the WRITE resp was handled before the READDIR+ resp
713          * causing the post op attrs from the write to be loaded first
714          * and the attrs from the READDIR+ to be loaded later. If this 
715          * happens, we have stale attrs loaded into the attrcache.
716          * We detect this by for the mtime moving back. We invalidate the 
717          * attrcache when this happens.
718          */
719         if (timespeccmp(&mtime_save, &vap->va_mtime, >))
720                 /* Size changed or mtime went backwards */
721                 np->n_attrstamp = 0;
722         if (vaper != NULL) {
723                 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
724                 if (np->n_flag & NCHG) {
725                         if (np->n_flag & NACC)
726                                 vaper->va_atime = np->n_atim;
727                         if (np->n_flag & NUPD)
728                                 vaper->va_mtime = np->n_mtim;
729                 }
730         }
731         mtx_unlock(&np->n_mtx);
732         return (0);
733 }
734 
735 #ifdef NFS_ACDEBUG
736 #include <sys/sysctl.h>
737 SYSCTL_DECL(_vfs_nfs);
738 static int nfs_acdebug;
739 SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0,
740     "Toggle acdebug (access cache debug) flag");
741 #endif
742 
743 /*
744  * Check the time stamp
745  * If the cache is valid, copy contents to *vap and return 0
746  * otherwise return an error
747  */
748 int
749 nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
750 {
751         struct nfsnode *np;
752         struct vattr *vap;
753         struct nfsmount *nmp;
754         int timeo;
755         
756         np = VTONFS(vp);
757         vap = &np->n_vattr;
758         nmp = VFSTONFS(vp->v_mount);
759 #ifdef NFS_ACDEBUG
760         mtx_lock(&Giant);       /* nfs_printf() */
761 #endif
762         mtx_lock(&np->n_mtx);
763         /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
764         timeo = (time_second - np->n_mtime.tv_sec) / 10;
765 
766 #ifdef NFS_ACDEBUG
767         if (nfs_acdebug>1)
768                 nfs_printf("nfs_getattrcache: initial timeo = %d\n", timeo);
769 #endif
770 
771         if (vap->va_type == VDIR) {
772                 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
773                         timeo = nmp->nm_acdirmin;
774                 else if (timeo > nmp->nm_acdirmax)
775                         timeo = nmp->nm_acdirmax;
776         } else {
777                 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
778                         timeo = nmp->nm_acregmin;
779                 else if (timeo > nmp->nm_acregmax)
780                         timeo = nmp->nm_acregmax;
781         }
782 
783 #ifdef NFS_ACDEBUG
784         if (nfs_acdebug > 2)
785                 nfs_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
786                            nmp->nm_acregmin, nmp->nm_acregmax,
787                            nmp->nm_acdirmin, nmp->nm_acdirmax);
788 
789         if (nfs_acdebug)
790                 nfs_printf("nfs_getattrcache: age = %d; final timeo = %d\n",
791                            (time_second - np->n_attrstamp), timeo);
792 #endif
793 
794         if ((time_second - np->n_attrstamp) >= timeo) {
795                 nfsstats.attrcache_misses++;
796                 mtx_unlock(&np->n_mtx);
797                 return( ENOENT);
798         }
799         nfsstats.attrcache_hits++;
800         if (vap->va_size != np->n_size) {
801                 if (vap->va_type == VREG) {
802                         if (np->n_flag & NMODIFIED) {
803                                 if (vap->va_size < np->n_size)
804                                         vap->va_size = np->n_size;
805                                 else
806                                         np->n_size = vap->va_size;
807                         } else {
808                                 np->n_size = vap->va_size;
809                         }
810                         vnode_pager_setsize(vp, np->n_size);
811                 } else {
812                         np->n_size = vap->va_size;
813                 }
814         }
815         bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
816         if (np->n_flag & NCHG) {
817                 if (np->n_flag & NACC)
818                         vaper->va_atime = np->n_atim;
819                 if (np->n_flag & NUPD)
820                         vaper->va_mtime = np->n_mtim;
821         }
822         mtx_unlock(&np->n_mtx);
823 #ifdef NFS_ACDEBUG
824         mtx_unlock(&Giant);     /* nfs_printf() */
825 #endif
826         return (0);
827 }
828 
829 static nfsuint64 nfs_nullcookie = { { 0, 0 } };
830 /*
831  * This function finds the directory cookie that corresponds to the
832  * logical byte offset given.
833  */
834 nfsuint64 *
835 nfs_getcookie(struct nfsnode *np, off_t off, int add)
836 {
837         struct nfsdmap *dp, *dp2;
838         int pos;
839         nfsuint64 *retval = NULL;
840         
841         pos = (uoff_t)off / NFS_DIRBLKSIZ;
842         if (pos == 0 || off < 0) {
843 #ifdef DIAGNOSTIC
844                 if (add)
845                         panic("nfs getcookie add at <= 0");
846 #endif
847                 return (&nfs_nullcookie);
848         }
849         pos--;
850         dp = LIST_FIRST(&np->n_cookies);
851         if (!dp) {
852                 if (add) {
853                         dp = malloc(sizeof (struct nfsdmap),
854                                 M_NFSDIROFF, M_WAITOK);
855                         dp->ndm_eocookie = 0;
856                         LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
857                 } else
858                         goto out;
859         }
860         while (pos >= NFSNUMCOOKIES) {
861                 pos -= NFSNUMCOOKIES;
862                 if (LIST_NEXT(dp, ndm_list)) {
863                         if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
864                             pos >= dp->ndm_eocookie)
865                                 goto out;
866                         dp = LIST_NEXT(dp, ndm_list);
867                 } else if (add) {
868                         dp2 = malloc(sizeof (struct nfsdmap),
869                                 M_NFSDIROFF, M_WAITOK);
870                         dp2->ndm_eocookie = 0;
871                         LIST_INSERT_AFTER(dp, dp2, ndm_list);
872                         dp = dp2;
873                 } else
874                         goto out;
875         }
876         if (pos >= dp->ndm_eocookie) {
877                 if (add)
878                         dp->ndm_eocookie = pos + 1;
879                 else
880                         goto out;
881         }
882         retval = &dp->ndm_cookies[pos];
883 out:
884         return (retval);
885 }
886 
887 /*
888  * Invalidate cached directory information, except for the actual directory
889  * blocks (which are invalidated separately).
890  * Done mainly to avoid the use of stale offset cookies.
891  */
892 void
893 nfs_invaldir(struct vnode *vp)
894 {
895         struct nfsnode *np = VTONFS(vp);
896 
897 #ifdef DIAGNOSTIC
898         if (vp->v_type != VDIR)
899                 panic("nfs: invaldir not dir");
900 #endif
901         nfs_dircookie_lock(np);
902         np->n_direofoffset = 0;
903         np->n_cookieverf.nfsuquad[0] = 0;
904         np->n_cookieverf.nfsuquad[1] = 0;
905         if (LIST_FIRST(&np->n_cookies))
906                 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
907         nfs_dircookie_unlock(np);
908 }
909 
910 /*
911  * The write verifier has changed (probably due to a server reboot), so all
912  * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
913  * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
914  * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
915  * mount point.
916  *
917  * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
918  * writes are not clusterable.
919  */
920 void
921 nfs_clearcommit(struct mount *mp)
922 {
923         struct vnode *vp, *nvp;
924         struct buf *bp, *nbp;
925         struct bufobj *bo;
926 
927         MNT_ILOCK(mp);
928         MNT_VNODE_FOREACH(vp, mp, nvp) {
929                 bo = &vp->v_bufobj;
930                 VI_LOCK(vp);
931                 if (vp->v_iflag & VI_DOOMED) {
932                         VI_UNLOCK(vp);
933                         continue;
934                 }
935                 vholdl(vp);
936                 VI_UNLOCK(vp);
937                 MNT_IUNLOCK(mp);
938                 BO_LOCK(bo);
939                 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
940                         if (!BUF_ISLOCKED(bp) &&
941                             (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
942                                 == (B_DELWRI | B_NEEDCOMMIT))
943                                 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
944                 }
945                 BO_UNLOCK(bo);
946                 vdrop(vp);
947                 MNT_ILOCK(mp);
948         }
949         MNT_IUNLOCK(mp);
950 }
951 
952 /*
953  * Helper functions for former macros.  Some of these should be
954  * moved to their callers.
955  */