2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
32 * @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
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.
44 #include "opt_kdtrace.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
52 #include <sys/mount.h>
53 #include <sys/vnode.h>
54 #include <sys/namei.h>
56 #include <sys/socket.h>
58 #include <sys/malloc.h>
59 #include <sys/sysent.h>
60 #include <sys/syscall.h>
61 #include <sys/sysproto.h>
64 #include <vm/vm_object.h>
65 #include <vm/vm_extern.h>
68 #include <nfs/nfsproto.h>
69 #include <nfsclient/nfs.h>
70 #include <nfsclient/nfsnode.h>
71 #include <nfsclient/nfs_kdtrace.h>
72 #include <nfs/xdr_subs.h>
73 #include <nfsclient/nfsm_subs.h>
74 #include <nfsclient/nfsmount.h>
76 #include <netinet/in.h>
79 * Note that stdarg.h and the ANSI style va_start macro is used for both
80 * ANSI and traditional C compilers.
82 #include <machine/stdarg.h>
85 dtrace_nfsclient_attrcache_flush_probe_func_t
86 dtrace_nfsclient_attrcache_flush_done_probe;
87 uint32_t nfsclient_attrcache_flush_done_id;
89 dtrace_nfsclient_attrcache_get_hit_probe_func_t
90 dtrace_nfsclient_attrcache_get_hit_probe;
91 uint32_t nfsclient_attrcache_get_hit_id;
93 dtrace_nfsclient_attrcache_get_miss_probe_func_t
94 dtrace_nfsclient_attrcache_get_miss_probe;
95 uint32_t nfsclient_attrcache_get_miss_id;
97 dtrace_nfsclient_attrcache_load_probe_func_t
98 dtrace_nfsclient_attrcache_load_done_probe;
99 uint32_t nfsclient_attrcache_load_done_id;
100 #endif /* !KDTRACE_HOOKS */
103 * Data items converted to xdr at startup, since they are constant
104 * This is kinda hokey, but may save a little time doing byte swaps
106 u_int32_t nfs_xdrneg1;
107 u_int32_t nfs_true, nfs_false;
109 /* And other global data */
110 static u_int32_t nfs_xid = 0;
111 static enum vtype nv2tov_type[8]= {
112 VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON
116 int nfs_pbuf_freecnt = -1; /* start out unlimited */
118 struct nfs_bufq nfs_bufq;
119 static struct mtx nfs_xid_mtx;
122 * and the reverse mapping from generic to Version 2 procedure numbers
124 int nfsv2_procid[NFS_NPROCS] = {
150 LIST_HEAD(nfsnodehashhead, nfsnode);
157 mtx_lock(&nfs_xid_mtx);
159 /* Get a pretty random xid to start with */
163 * Skip zero xid if it should ever happen.
168 mtx_unlock(&nfs_xid_mtx);
173 * Create the header for an rpc request packet
174 * The hsiz is the size of the rest of the nfs request header.
175 * (just used to decide if a cluster is a good idea)
178 nfsm_reqhead(struct vnode *vp, u_long procid, int hsiz)
182 MGET(mb, M_WAIT, MT_DATA);
183 if (hsiz >= MINCLSIZE)
190 * copies a uio scatter/gather list to an mbuf chain.
191 * NOTE: can ony handle iovcnt == 1
194 nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
197 struct mbuf *mp, *mp2;
198 int xfer, left, mlen;
199 int uiosiz, clflg, rem;
202 KASSERT(uiop->uio_iovcnt == 1, ("nfsm_uiotombuf: iovcnt != 1"));
204 if (siz > MLEN) /* or should it >= MCLBYTES ?? */
208 rem = nfsm_rndup(siz)-siz;
211 left = uiop->uio_iov->iov_len;
212 uiocp = uiop->uio_iov->iov_base;
217 mlen = M_TRAILINGSPACE(mp);
219 MGET(mp, M_WAIT, MT_DATA);
225 mlen = M_TRAILINGSPACE(mp);
227 xfer = (left > mlen) ? mlen : left;
230 if (uiop->uio_iov->iov_op != NULL)
231 (*(uiop->uio_iov->iov_op))
232 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
235 if (uiop->uio_segflg == UIO_SYSSPACE)
236 bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
238 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
242 uiop->uio_offset += xfer;
243 uiop->uio_resid -= xfer;
245 uiop->uio_iov->iov_base =
246 (char *)uiop->uio_iov->iov_base + uiosiz;
247 uiop->uio_iov->iov_len -= uiosiz;
251 if (rem > M_TRAILINGSPACE(mp)) {
252 MGET(mp, M_WAIT, MT_DATA);
256 cp = mtod(mp, caddr_t)+mp->m_len;
257 for (left = 0; left < rem; left++)
262 *bpos = mtod(mp, caddr_t)+mp->m_len;
268 * Copy a string into mbufs for the hard cases...
271 nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
273 struct mbuf *m1 = NULL, *m2;
274 long left, xfer, len, tlen;
280 left = M_TRAILINGSPACE(m2);
282 tl = ((u_int32_t *)(*bpos));
283 *tl++ = txdr_unsigned(siz);
285 left -= NFSX_UNSIGNED;
286 m2->m_len += NFSX_UNSIGNED;
288 bcopy(cp, (caddr_t) tl, left);
295 /* Loop around adding mbufs */
297 MGET(m1, M_WAIT, MT_DATA);
300 m1->m_len = NFSMSIZ(m1);
303 tl = mtod(m1, u_int32_t *);
306 *tl++ = txdr_unsigned(siz);
307 m1->m_len -= NFSX_UNSIGNED;
308 tlen = NFSX_UNSIGNED;
311 if (siz < m1->m_len) {
312 len = nfsm_rndup(siz);
317 xfer = len = m1->m_len;
319 bcopy(cp, (caddr_t) tl, xfer);
320 m1->m_len = len+tlen;
325 *bpos = mtod(m1, caddr_t)+m1->m_len;
330 * Called once to initialize data structures...
333 nfs_init(struct vfsconf *vfsp)
337 nfsmount_zone = uma_zcreate("NFSMOUNT", sizeof(struct nfsmount),
338 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
339 nfs_true = txdr_unsigned(TRUE);
340 nfs_false = txdr_unsigned(FALSE);
341 nfs_xdrneg1 = txdr_unsigned(-1);
342 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
345 /* Ensure async daemons disabled */
346 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
347 nfs_iodwant[i] = NFSIOD_NOT_AVAILABLE;
348 nfs_iodmount[i] = NULL;
350 nfs_nhinit(); /* Init the nfsnode table */
353 * Initialize reply list and start timer
355 mtx_init(&nfs_iod_mtx, "NFS iod lock", NULL, MTX_DEF);
356 mtx_init(&nfs_xid_mtx, "NFS xid lock", NULL, MTX_DEF);
358 nfs_pbuf_freecnt = nswbuf / 2 + 1;
364 nfs_uninit(struct vfsconf *vfsp)
369 * Tell all nfsiod processes to exit. Clear nfs_iodmax, and wakeup
370 * any sleeping nfsiods so they check nfs_iodmax and exit.
372 mtx_lock(&nfs_iod_mtx);
374 for (i = 0; i < nfs_numasync; i++)
375 if (nfs_iodwant[i] == NFSIOD_AVAILABLE)
376 wakeup(&nfs_iodwant[i]);
377 /* The last nfsiod to exit will wake us up when nfs_numasync hits 0 */
379 msleep(&nfs_numasync, &nfs_iod_mtx, PWAIT, "ioddie", 0);
380 mtx_unlock(&nfs_iod_mtx);
382 uma_zdestroy(nfsmount_zone);
387 nfs_dircookie_lock(struct nfsnode *np)
389 mtx_lock(&np->n_mtx);
390 while (np->n_flag & NDIRCOOKIELK)
391 (void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
392 np->n_flag |= NDIRCOOKIELK;
393 mtx_unlock(&np->n_mtx);
397 nfs_dircookie_unlock(struct nfsnode *np)
399 mtx_lock(&np->n_mtx);
400 np->n_flag &= ~NDIRCOOKIELK;
402 mtx_unlock(&np->n_mtx);
406 nfs_upgrade_vnlock(struct vnode *vp)
410 ASSERT_VOP_LOCKED(vp, "nfs_upgrade_vnlock");
411 old_lock = VOP_ISLOCKED(vp);
412 if (old_lock != LK_EXCLUSIVE) {
413 KASSERT(old_lock == LK_SHARED,
414 ("nfs_upgrade_vnlock: wrong old_lock %d", old_lock));
415 /* Upgrade to exclusive lock, this might block */
416 vn_lock(vp, LK_UPGRADE | LK_RETRY);
422 nfs_downgrade_vnlock(struct vnode *vp, int old_lock)
424 if (old_lock != LK_EXCLUSIVE) {
425 KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock));
426 /* Downgrade from exclusive lock. */
427 vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
432 nfs_printf(const char *fmt, ...)
444 * Attribute cache routines.
445 * nfs_loadattrcache() - loads or updates the cache contents from attributes
446 * that are on the mbuf list
447 * nfs_getattrcache() - returns valid attributes if found in cache, returns
452 * Load the attribute cache (that lives in the nfsnode entry) with
453 * the values on the mbuf list and
455 * copy the attributes to *vaper
458 nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
459 struct vattr *vaper, int dontshrink)
461 struct vnode *vp = *vpp;
463 struct nfs_fattr *fp;
464 struct nfsnode *np = NULL;
471 struct timespec mtime, mtime_save;
472 int v3 = NFS_ISV3(vp);
476 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
477 cp2 = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, M_WAIT);
482 fp = (struct nfs_fattr *)cp2;
484 vtyp = nfsv3tov_type(fp->fa_type);
485 vmode = fxdr_unsigned(u_short, fp->fa_mode);
486 rdev = makedev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
487 fxdr_unsigned(int, fp->fa3_rdev.specdata2));
488 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
490 vtyp = nfsv2tov_type(fp->fa_type);
491 vmode = fxdr_unsigned(u_short, fp->fa_mode);
495 * The duplicate information returned in fa_type and fa_mode
496 * is an ambiguity in the NFS version 2 protocol.
498 * VREG should be taken literally as a regular file. If a
499 * server intents to return some type information differently
500 * in the upper bits of the mode field (e.g. for sockets, or
501 * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
502 * leave the examination of the mode bits even in the VREG
503 * case to avoid breakage for bogus servers, but we make sure
504 * that there are actually type bits set in the upper part of
505 * fa_mode (and failing that, trust the va_type field).
507 * NFSv3 cleared the issue, and requires fa_mode to not
508 * contain any type information (while also introduing sockets
509 * and FIFOs for fa_type).
511 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
512 vtyp = IFTOVT(vmode);
513 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
514 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
517 * Really ugly NFSv2 kludge.
519 if (vtyp == VCHR && rdev == 0xffffffff)
524 * If v_type == VNON it is a new node, so fill in the v_type,
525 * n_mtime fields. Check to see if it represents a special
526 * device, and if so, check for a possible alias. Once the
527 * correct vnode has been obtained, fill in the rest of the
531 mtx_lock(&np->n_mtx);
532 if (vp->v_type != vtyp) {
534 if (vp->v_type == VFIFO)
535 vp->v_op = &nfs_fifoops;
540 vap->va_mode = (vmode & 07777);
542 mtime_save = vap->va_mtime;
543 vap->va_mtime = mtime;
544 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
546 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
547 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
548 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
549 vap->va_size = fxdr_hyper(&fp->fa3_size);
550 vap->va_blocksize = NFS_FABLKSIZE;
551 vap->va_bytes = fxdr_hyper(&fp->fa3_used);
552 vap->va_fileid = fxdr_unsigned(int32_t,
553 fp->fa3_fileid.nfsuquad[1]);
554 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
555 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
559 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
560 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
561 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
562 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
563 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
564 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
566 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
567 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
569 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
570 fp->fa2_ctime.nfsv2_sec);
571 vap->va_ctime.tv_nsec = 0;
572 vap->va_gen = fxdr_unsigned(u_int32_t, fp->fa2_ctime.nfsv2_usec);
575 np->n_attrstamp = time_second;
576 if (vap->va_size != np->n_size) {
577 if (vap->va_type == VREG) {
578 if (dontshrink && vap->va_size < np->n_size) {
580 * We've been told not to shrink the file;
581 * zero np->n_attrstamp to indicate that
582 * the attributes are stale.
584 vap->va_size = np->n_size;
586 KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
587 } else if (np->n_flag & NMODIFIED) {
589 * We've modified the file: Use the larger
590 * of our size, and the server's size.
592 if (vap->va_size < np->n_size) {
593 vap->va_size = np->n_size;
595 np->n_size = vap->va_size;
596 np->n_flag |= NSIZECHANGED;
599 np->n_size = vap->va_size;
600 np->n_flag |= NSIZECHANGED;
602 vnode_pager_setsize(vp, np->n_size);
604 np->n_size = vap->va_size;
608 * The following checks are added to prevent a race between (say)
609 * a READDIR+ and a WRITE.
610 * READDIR+, WRITE requests sent out.
611 * READDIR+ resp, WRITE resp received on client.
612 * However, the WRITE resp was handled before the READDIR+ resp
613 * causing the post op attrs from the write to be loaded first
614 * and the attrs from the READDIR+ to be loaded later. If this
615 * happens, we have stale attrs loaded into the attrcache.
616 * We detect this by for the mtime moving back. We invalidate the
617 * attrcache when this happens.
619 if (timespeccmp(&mtime_save, &vap->va_mtime, >)) {
620 /* Size changed or mtime went backwards */
622 KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
625 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
626 if (np->n_flag & NCHG) {
627 if (np->n_flag & NACC)
628 vaper->va_atime = np->n_atim;
629 if (np->n_flag & NUPD)
630 vaper->va_mtime = np->n_mtim;
635 if (np->n_attrstamp != 0)
636 KDTRACE_NFS_ATTRCACHE_LOAD_DONE(vp, &np->n_vattr, 0);
638 mtx_unlock(&np->n_mtx);
642 KDTRACE_NFS_ATTRCACHE_LOAD_DONE(vp, NULL, error);
648 #include <sys/sysctl.h>
649 SYSCTL_DECL(_vfs_nfs);
650 static int nfs_acdebug;
651 SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0,
652 "Toggle acdebug (access cache debug) flag");
656 * Check the time stamp
657 * If the cache is valid, copy contents to *vap and return 0
658 * otherwise return an error
661 nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
665 struct nfsmount *nmp;
670 nmp = VFSTONFS(vp->v_mount);
672 mtx_lock(&Giant); /* nfs_printf() */
674 mtx_lock(&np->n_mtx);
675 /* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
676 timeo = (time_second - np->n_mtime.tv_sec) / 10;
680 nfs_printf("nfs_getattrcache: initial timeo = %d\n", timeo);
683 if (vap->va_type == VDIR) {
684 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
685 timeo = nmp->nm_acdirmin;
686 else if (timeo > nmp->nm_acdirmax)
687 timeo = nmp->nm_acdirmax;
689 if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
690 timeo = nmp->nm_acregmin;
691 else if (timeo > nmp->nm_acregmax)
692 timeo = nmp->nm_acregmax;
697 nfs_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
698 nmp->nm_acregmin, nmp->nm_acregmax,
699 nmp->nm_acdirmin, nmp->nm_acdirmax);
702 nfs_printf("nfs_getattrcache: age = %d; final timeo = %d\n",
703 (time_second - np->n_attrstamp), timeo);
706 if ((time_second - np->n_attrstamp) >= timeo) {
707 nfsstats.attrcache_misses++;
708 mtx_unlock(&np->n_mtx);
709 KDTRACE_NFS_ATTRCACHE_GET_MISS(vp);
712 nfsstats.attrcache_hits++;
713 if (vap->va_size != np->n_size) {
714 if (vap->va_type == VREG) {
715 if (np->n_flag & NMODIFIED) {
716 if (vap->va_size < np->n_size)
717 vap->va_size = np->n_size;
719 np->n_size = vap->va_size;
721 np->n_size = vap->va_size;
723 vnode_pager_setsize(vp, np->n_size);
725 np->n_size = vap->va_size;
728 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
729 if (np->n_flag & NCHG) {
730 if (np->n_flag & NACC)
731 vaper->va_atime = np->n_atim;
732 if (np->n_flag & NUPD)
733 vaper->va_mtime = np->n_mtim;
735 mtx_unlock(&np->n_mtx);
737 mtx_unlock(&Giant); /* nfs_printf() */
739 KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap);
744 * Purge all cached information about an NFS vnode including name
745 * cache entries, the attribute cache, and the access cache. This is
746 * called when an NFS request for a node fails with a stale
750 nfs_purgecache(struct vnode *vp)
757 mtx_lock(&np->n_mtx);
759 KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
760 for (i = 0; i < NFS_ACCESSCACHESIZE; i++)
761 np->n_accesscache[i].stamp = 0;
762 KDTRACE_NFS_ACCESSCACHE_FLUSH_DONE(vp);
763 mtx_unlock(&np->n_mtx);
766 static nfsuint64 nfs_nullcookie = { { 0, 0 } };
768 * This function finds the directory cookie that corresponds to the
769 * logical byte offset given.
772 nfs_getcookie(struct nfsnode *np, off_t off, int add)
774 struct nfsdmap *dp, *dp2;
776 nfsuint64 *retval = NULL;
778 pos = (uoff_t)off / NFS_DIRBLKSIZ;
779 if (pos == 0 || off < 0) {
780 KASSERT(!add, ("nfs getcookie add at <= 0"));
781 return (&nfs_nullcookie);
784 dp = LIST_FIRST(&np->n_cookies);
787 dp = malloc(sizeof (struct nfsdmap),
788 M_NFSDIROFF, M_WAITOK);
789 dp->ndm_eocookie = 0;
790 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
794 while (pos >= NFSNUMCOOKIES) {
795 pos -= NFSNUMCOOKIES;
796 if (LIST_NEXT(dp, ndm_list)) {
797 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
798 pos >= dp->ndm_eocookie)
800 dp = LIST_NEXT(dp, ndm_list);
802 dp2 = malloc(sizeof (struct nfsdmap),
803 M_NFSDIROFF, M_WAITOK);
804 dp2->ndm_eocookie = 0;
805 LIST_INSERT_AFTER(dp, dp2, ndm_list);
810 if (pos >= dp->ndm_eocookie) {
812 dp->ndm_eocookie = pos + 1;
816 retval = &dp->ndm_cookies[pos];
822 * Invalidate cached directory information, except for the actual directory
823 * blocks (which are invalidated separately).
824 * Done mainly to avoid the use of stale offset cookies.
827 nfs_invaldir(struct vnode *vp)
829 struct nfsnode *np = VTONFS(vp);
831 KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir"));
832 nfs_dircookie_lock(np);
833 np->n_direofoffset = 0;
834 np->n_cookieverf.nfsuquad[0] = 0;
835 np->n_cookieverf.nfsuquad[1] = 0;
836 if (LIST_FIRST(&np->n_cookies))
837 LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
838 nfs_dircookie_unlock(np);
842 * The write verifier has changed (probably due to a server reboot), so all
843 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
844 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
845 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the
848 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
849 * writes are not clusterable.
852 nfs_clearcommit(struct mount *mp)
854 struct vnode *vp, *nvp;
855 struct buf *bp, *nbp;
859 MNT_VNODE_FOREACH(vp, mp, nvp) {
862 if (vp->v_iflag & VI_DOOMED) {
870 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
871 if (!BUF_ISLOCKED(bp) &&
872 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
873 == (B_DELWRI | B_NEEDCOMMIT))
874 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
884 * Helper functions for former macros. Some of these should be
885 * moved to their callers.
889 nfsm_mtofh_xx(struct vnode *d, struct vnode **v, int v3, int *f,
890 struct mbuf **md, caddr_t *dpos)
892 struct nfsnode *ttnp;
900 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
903 *f = fxdr_unsigned(int, *tl);
907 t1 = nfsm_getfh_xx(&ttfhp, &ttfhsize, (v3), md, dpos);
910 t1 = nfs_nget(d->v_mount, ttfhp, ttfhsize, &ttnp, LK_EXCLUSIVE);
916 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
920 *f = fxdr_unsigned(int, *tl);
921 else if (fxdr_unsigned(int, *tl))
922 nfsm_adv_xx(NFSX_V3FATTR, md, dpos);
926 t1 = nfs_loadattrcache(&ttvp, md, dpos, NULL, 0);
935 nfsm_getfh_xx(nfsfh_t **f, int *s, int v3, struct mbuf **md, caddr_t *dpos)
940 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
943 *s = fxdr_unsigned(int, *tl);
944 if (*s <= 0 || *s > NFSX_V3FHMAX)
948 *f = nfsm_dissect_xx(nfsm_rndup(*s), md, dpos);
957 nfsm_loadattr_xx(struct vnode **v, struct vattr *va, struct mbuf **md,
962 struct vnode *ttvp = *v;
963 t1 = nfs_loadattrcache(&ttvp, md, dpos, va, 0);
971 nfsm_postop_attr_xx(struct vnode **v, int *f, struct mbuf **md,
977 struct vnode *ttvp = *v;
978 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
981 *f = fxdr_unsigned(int, *tl);
983 t1 = nfs_loadattrcache(&ttvp, md, dpos, NULL, 1);
994 nfsm_wcc_data_xx(struct vnode **v, int *f, struct mbuf **md, caddr_t *dpos)
997 int ttattrf, ttretf = 0;
1000 tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
1003 if (*tl == nfs_true) {
1004 tl = nfsm_dissect_xx(6 * NFSX_UNSIGNED, md, dpos);
1007 mtx_lock(&(VTONFS(*v))->n_mtx);
1009 ttretf = (VTONFS(*v)->n_mtime.tv_sec == fxdr_unsigned(u_int32_t, *(tl + 2)) &&
1010 VTONFS(*v)->n_mtime.tv_nsec == fxdr_unsigned(u_int32_t, *(tl + 3)));
1011 mtx_unlock(&(VTONFS(*v))->n_mtx);
1013 t1 = nfsm_postop_attr_xx(v, &ttattrf, md, dpos);
1024 nfsm_strtom_xx(const char *a, int s, int m, struct mbuf **mb, caddr_t *bpos)
1030 return ENAMETOOLONG;
1031 t1 = nfsm_rndup(s) + NFSX_UNSIGNED;
1032 if (t1 <= M_TRAILINGSPACE(*mb)) {
1033 tl = nfsm_build_xx(t1, mb, bpos);
1034 *tl++ = txdr_unsigned(s);
1035 *(tl + ((t1 >> 2) - 2)) = 0;
1038 t1 = nfsm_strtmbuf(mb, bpos, a, s);
1046 nfsm_fhtom_xx(struct vnode *v, int v3, struct mbuf **mb, caddr_t *bpos)
1053 t1 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED;
1054 if (t1 < M_TRAILINGSPACE(*mb)) {
1055 tl = nfsm_build_xx(t1, mb, bpos);
1056 *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize);
1057 *(tl + ((t1 >> 2) - 2)) = 0;
1058 bcopy(VTONFS(v)->n_fhp, tl, VTONFS(v)->n_fhsize);
1060 t1 = nfsm_strtmbuf(mb, bpos,
1061 (const char *)VTONFS(v)->n_fhp,
1062 VTONFS(v)->n_fhsize);
1067 cp = nfsm_build_xx(NFSX_V2FH, mb, bpos);
1068 bcopy(VTONFS(v)->n_fhp, cp, NFSX_V2FH);
1074 nfsm_v3attrbuild_xx(struct vattr *va, int full, struct mbuf **mb,
1079 if (va->va_mode != (mode_t)VNOVAL) {
1080 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1082 *tl = txdr_unsigned(va->va_mode);
1084 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1087 if (full && va->va_uid != (uid_t)VNOVAL) {
1088 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1090 *tl = txdr_unsigned(va->va_uid);
1092 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1095 if (full && va->va_gid != (gid_t)VNOVAL) {
1096 tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
1098 *tl = txdr_unsigned(va->va_gid);
1100 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1103 if (full && va->va_size != VNOVAL) {
1104 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1106 txdr_hyper(va->va_size, tl);
1108 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1111 if (va->va_atime.tv_sec != VNOVAL) {
1112 if (va->va_atime.tv_sec != time_second) {
1113 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1114 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
1115 txdr_nfsv3time(&va->va_atime, tl);
1117 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1118 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
1121 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1122 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
1124 if (va->va_mtime.tv_sec != VNOVAL) {
1125 if (va->va_mtime.tv_sec != time_second) {
1126 tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
1127 *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
1128 txdr_nfsv3time(&va->va_mtime, tl);
1130 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1131 *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
1134 tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
1135 *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);