2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
7 * This code is derived from software contributed to Berkeley by
8 * Rick Macklem at The University of Guelph.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
40 * Rpc op calls, generally called from the vnode op calls or through the
41 * buffer cache, for NFS v2, 3 and 4.
42 * These do not normally make any changes to vnode arguments or use
43 * structures that might change between the VFS variants. The returned
44 * arguments are all at the end, after the NFSPROC_T *p one.
47 #include "opt_inet6.h"
49 #include <fs/nfs/nfsport.h>
50 #include <sys/sysctl.h>
51 #include <sys/taskqueue.h>
53 SYSCTL_DECL(_vfs_nfs);
55 static int nfsignore_eexist = 0;
56 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
57 &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
59 static int nfscl_dssameconn = 0;
60 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
61 &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
66 extern struct nfsstatsv1 nfsstatsv1;
67 extern int nfs_numnfscbd;
68 extern struct timeval nfsboottime;
69 extern u_int32_t newnfs_false, newnfs_true;
70 extern nfstype nfsv34_type[9];
71 extern int nfsrv_useacl;
72 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
73 extern int nfscl_debuglevel;
74 extern int nfs_pnfsiothreads;
76 int nfstest_outofseq = 0;
77 int nfscl_assumeposixlocks = 1;
78 int nfscl_enablecallb = 0;
79 short nfsv4_cbport = NFSV4_CBPORT;
80 int nfstest_openallsetattr = 0;
82 #define DIRHDSIZ offsetof(struct dirent, d_name)
85 * nfscl_getsameserver() can return one of three values:
86 * NFSDSP_USETHISSESSION - Use this session for the DS.
87 * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
89 * NFSDSP_NOTFOUND - No matching server was found.
92 NFSDSP_USETHISSESSION = 0,
93 NFSDSP_SEQTHISSESSION = 1,
98 * Do a write RPC on a DS data file, using this structure for the arguments,
99 * so that this function can be executed by a separate kernel process.
101 struct nfsclwritedsdorpc {
108 nfsv4stateid_t *stateidp;
121 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
122 struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
123 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
124 nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
125 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
126 struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
128 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
129 nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
130 struct nfsvattr *, struct nfsfh **, int *, int *, void *);
131 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
132 nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
133 NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
134 int *, void *, int *);
135 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
136 struct nfscllockowner *, u_int64_t, u_int64_t,
137 u_int32_t, struct ucred *, NFSPROC_T *, int);
138 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
139 struct acl *, nfsv4stateid_t *, void *);
140 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
141 uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
142 struct ucred *, NFSPROC_T *);
143 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
144 struct sockaddr_in6 *, sa_family_t, int, struct nfsclds **, NFSPROC_T *);
145 static void nfscl_initsessionslots(struct nfsclsession *);
146 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
147 nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
148 struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
150 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
151 nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
152 struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
153 struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
154 static struct mbuf *nfsm_copym(struct mbuf *, int, int);
155 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
156 struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
157 struct ucred *, NFSPROC_T *);
158 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
159 nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
160 struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
161 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
162 struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
163 struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
164 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
165 struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
166 struct ucred *, NFSPROC_T *);
167 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
168 struct nfsclds *, struct nfsclds **, uint32_t *);
169 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
170 struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
172 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
173 struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
174 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
175 uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
176 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
178 static int nfsrv_parselayoutget(struct nfsrv_descript *, nfsv4stateid_t *,
179 int *, struct nfsclflayouthead *);
180 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
181 int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
182 struct nfscldeleg **, struct ucred *, NFSPROC_T *);
183 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
184 nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
185 struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
186 struct nfsfh **, int *, int *, void *, int *);
187 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
188 int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
189 struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
190 struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
191 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
192 nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
193 struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
194 struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
195 int, int, int, int *, struct nfsclflayouthead *, int *);
196 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
197 uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
198 struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
199 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
200 int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
201 struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
203 int nfs_pnfsio(task_fn_t *, void *);
206 * nfs null call from vfs.
209 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
212 struct nfsrv_descript nfsd, *nd = &nfsd;
214 NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
215 error = nfscl_request(nd, vp, p, cred, NULL);
216 if (nd->nd_repstat && !error)
217 error = nd->nd_repstat;
218 mbuf_freem(nd->nd_mrep);
224 * For nfs version 3 and 4, use the access rpc to check accessibility. If file
225 * modes are changed on the server, accesses might still fail later.
228 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
229 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
232 u_int32_t mode, rmode;
235 mode = NFSACCESS_READ;
238 if (vnode_vtype(vp) == VDIR) {
240 mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
243 mode |= NFSACCESS_LOOKUP;
246 mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
248 mode |= NFSACCESS_EXECUTE;
252 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
254 error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
258 * The NFS V3 spec does not clarify whether or not
259 * the returned access bits can be a superset of
260 * the ones requested, so...
262 if (!error && (rmode & mode) != mode)
268 * The actual rpc, separated out for Darwin.
271 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
272 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
276 u_int32_t supported, rmode;
278 struct nfsrv_descript nfsd, *nd = &nfsd;
279 nfsattrbit_t attrbits;
283 NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
284 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
285 *tl = txdr_unsigned(mode);
286 if (nd->nd_flag & ND_NFSV4) {
288 * And do a Getattr op.
290 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
291 *tl = txdr_unsigned(NFSV4OP_GETATTR);
292 NFSGETATTR_ATTRBIT(&attrbits);
293 (void) nfsrv_putattrbit(nd, &attrbits);
295 error = nfscl_request(nd, vp, p, cred, stuff);
298 if (nd->nd_flag & ND_NFSV3) {
299 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
303 if (!nd->nd_repstat) {
304 if (nd->nd_flag & ND_NFSV4) {
305 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
306 supported = fxdr_unsigned(u_int32_t, *tl++);
308 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
310 rmode = fxdr_unsigned(u_int32_t, *tl);
311 if (nd->nd_flag & ND_NFSV4)
312 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
315 * It's not obvious what should be done about
316 * unsupported access modes. For now, be paranoid
317 * and clear the unsupported ones.
322 error = nd->nd_repstat;
324 mbuf_freem(nd->nd_mrep);
332 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
334 struct nfsclopen *op;
335 struct nfscldeleg *dp;
337 struct nfsnode *np = VTONFS(vp);
338 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
339 u_int32_t mode, clidrev;
340 int ret, newone, error, expireret = 0, retrycnt;
343 * For NFSv4, Open Ops are only done on Regular Files.
345 if (vnode_vtype(vp) != VREG)
349 mode |= NFSV4OPEN_ACCESSREAD;
351 mode |= NFSV4OPEN_ACCESSWRITE;
356 { char name[100]; int namel;
357 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
358 bcopy(NFS4NODENAME(np->n_v4), name, namel);
360 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
361 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
362 else printf(" fhl=0\n");
367 error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
368 cred, p, NULL, &op, &newone, &ret, 1);
372 if (nmp->nm_clp != NULL)
373 clidrev = nmp->nm_clp->nfsc_clientidrev;
376 if (ret == NFSCLOPEN_DOOPEN) {
377 if (np->n_v4 != NULL) {
379 * For the first attempt, try and get a layout, if
380 * pNFS is enabled for the mount.
382 if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
383 nfs_numnfscbd == 0 ||
384 (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
385 error = nfsrpc_openrpc(nmp, vp,
387 np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
388 np->n_fhp->nfh_len, mode, op,
389 NFS4NODENAME(np->n_v4),
390 np->n_v4->n4_namelen,
391 &dp, 0, 0x0, cred, p, 0, 0);
393 error = nfsrpc_getopenlayout(nmp, vp,
395 np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
396 np->n_fhp->nfh_len, mode, op,
397 NFS4NODENAME(np->n_v4),
398 np->n_v4->n4_namelen, &dp, cred, p);
401 OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
404 np->n_flag &= ~NDELEGMOD;
406 * Invalidate the attribute cache, so that
407 * attributes that pre-date the issue of a
408 * delegation are not cached, since the
409 * cached attributes will remain valid while
410 * the delegation is held.
412 NFSINVALATTRCACHE(np);
415 (void) nfscl_deleg(nmp->nm_mountp,
416 op->nfso_own->nfsow_clp,
417 nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
422 newnfs_copyincred(cred, &op->nfso_cred);
423 } else if (ret == NFSCLOPEN_SETCRED)
425 * This is a new local open on a delegation. It needs
426 * to have credentials so that an open can be done
427 * against the server during recovery.
429 newnfs_copyincred(cred, &op->nfso_cred);
432 * nfso_opencnt is the count of how many VOP_OPEN()s have
433 * been done on this Open successfully and a VOP_CLOSE()
434 * is expected for each of these.
435 * If error is non-zero, don't increment it, since the Open
436 * hasn't succeeded yet.
440 nfscl_openrelease(nmp, op, error, newone);
441 if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
442 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
443 error == NFSERR_BADSESSION) {
444 (void) nfs_catnap(PZERO, error, "nfs_open");
445 } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
447 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
450 } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
451 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
452 error == NFSERR_BADSESSION ||
453 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
454 expireret == 0 && clidrev != 0 && retrycnt < 4));
455 if (error && retrycnt >= 4)
461 * the actual open rpc
464 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
465 u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
466 u_int8_t *name, int namelen, struct nfscldeleg **dpp,
467 int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
468 int syscred, int recursed)
471 struct nfsrv_descript nfsd, *nd = &nfsd;
472 struct nfscldeleg *dp, *ndp = NULL;
473 struct nfsvattr nfsva;
474 u_int32_t rflags, deleg;
475 nfsattrbit_t attrbits;
476 int error, ret, acesize, limitby;
477 struct nfsclsession *tsep;
481 nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
482 NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
483 *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
484 *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
485 *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
486 tsep = nfsmnt_mdssession(nmp);
487 *tl++ = tsep->nfsess_clientid.lval[0];
488 *tl = tsep->nfsess_clientid.lval[1];
489 (void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
490 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
491 *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
493 *tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
494 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
495 *tl = txdr_unsigned(delegtype);
498 *tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
499 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
500 if (NFSHASNFSV4N(nmp))
503 *tl++ = dp->nfsdl_stateid.seqid;
504 *tl++ = dp->nfsdl_stateid.other[0];
505 *tl++ = dp->nfsdl_stateid.other[1];
506 *tl = dp->nfsdl_stateid.other[2];
508 *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
510 (void) nfsm_strtom(nd, name, namelen);
512 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
513 *tl = txdr_unsigned(NFSV4OP_GETATTR);
514 NFSZERO_ATTRBIT(&attrbits);
515 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
516 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
517 (void) nfsrv_putattrbit(nd, &attrbits);
519 nd->nd_flag |= ND_USEGSSNAME;
520 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
521 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
524 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
525 if (!nd->nd_repstat) {
526 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
528 op->nfso_stateid.seqid = *tl++;
529 op->nfso_stateid.other[0] = *tl++;
530 op->nfso_stateid.other[1] = *tl++;
531 op->nfso_stateid.other[2] = *tl;
532 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
533 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
536 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
537 deleg = fxdr_unsigned(u_int32_t, *tl);
538 if (deleg == NFSV4OPEN_DELEGATEREAD ||
539 deleg == NFSV4OPEN_DELEGATEWRITE) {
540 if (!(op->nfso_own->nfsow_clp->nfsc_flags &
541 NFSCLFLAGS_FIRSTDELEG))
542 op->nfso_own->nfsow_clp->nfsc_flags |=
543 (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
545 sizeof (struct nfscldeleg) + newfhlen,
546 M_NFSCLDELEG, M_WAITOK);
547 LIST_INIT(&ndp->nfsdl_owner);
548 LIST_INIT(&ndp->nfsdl_lock);
549 ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
550 ndp->nfsdl_fhlen = newfhlen;
551 NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
552 newnfs_copyincred(cred, &ndp->nfsdl_cred);
553 nfscl_lockinit(&ndp->nfsdl_rwlock);
554 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
556 ndp->nfsdl_stateid.seqid = *tl++;
557 ndp->nfsdl_stateid.other[0] = *tl++;
558 ndp->nfsdl_stateid.other[1] = *tl++;
559 ndp->nfsdl_stateid.other[2] = *tl++;
560 ret = fxdr_unsigned(int, *tl);
561 if (deleg == NFSV4OPEN_DELEGATEWRITE) {
562 ndp->nfsdl_flags = NFSCLDL_WRITE;
564 * Indicates how much the file can grow.
566 NFSM_DISSECT(tl, u_int32_t *,
568 limitby = fxdr_unsigned(int, *tl++);
570 case NFSV4OPEN_LIMITSIZE:
571 ndp->nfsdl_sizelimit = fxdr_hyper(tl);
573 case NFSV4OPEN_LIMITBLOCKS:
574 ndp->nfsdl_sizelimit =
575 fxdr_unsigned(u_int64_t, *tl++);
576 ndp->nfsdl_sizelimit *=
577 fxdr_unsigned(u_int64_t, *tl);
580 error = NFSERR_BADXDR;
584 ndp->nfsdl_flags = NFSCLDL_READ;
587 ndp->nfsdl_flags |= NFSCLDL_RECALL;
588 error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
592 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
593 error = NFSERR_BADXDR;
596 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
597 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
598 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
599 NULL, NULL, NULL, p, cred);
603 ndp->nfsdl_change = nfsva.na_filerev;
604 ndp->nfsdl_modtime = nfsva.na_mtime;
605 ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
607 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
609 ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
611 if (ret == NFSERR_DELAY)
612 (void) nfs_catnap(PZERO, ret, "nfs_open");
613 } while (ret == NFSERR_DELAY);
616 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
617 nfscl_assumeposixlocks)
618 op->nfso_posixlock = 1;
620 op->nfso_posixlock = 0;
623 * If the server is handing out delegations, but we didn't
624 * get one because an OpenConfirm was required, try the
625 * Open again, to get a delegation. This is a harmless no-op,
626 * from a server's point of view.
628 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
629 (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
630 && !error && dp == NULL && ndp == NULL && !recursed) {
632 ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
633 newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
634 cred, p, syscred, 1);
635 if (ret == NFSERR_DELAY)
636 (void) nfs_catnap(PZERO, ret, "nfs_open2");
637 } while (ret == NFSERR_DELAY);
640 free(ndp, M_NFSCLDELEG);
643 if (ret == NFSERR_STALECLIENTID ||
644 ret == NFSERR_STALEDONTRECOVER ||
645 ret == NFSERR_BADSESSION)
650 if (nd->nd_repstat != 0 && error == 0)
651 error = nd->nd_repstat;
652 if (error == NFSERR_STALECLIENTID)
653 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
657 else if (ndp != NULL)
658 free(ndp, M_NFSCLDELEG);
659 mbuf_freem(nd->nd_mrep);
667 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
668 struct ucred *cred, NFSPROC_T *p)
671 struct nfsrv_descript nfsd, *nd = &nfsd;
674 NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
675 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
676 if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
679 *tl++ = op->nfso_stateid.seqid;
680 *tl++ = op->nfso_stateid.other[0];
681 *tl++ = op->nfso_stateid.other[1];
682 *tl++ = op->nfso_stateid.other[2];
683 *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
684 *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
685 *tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
686 error = nfscl_request(nd, vp, p, cred, NULL);
689 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
690 if (!nd->nd_repstat) {
691 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
692 op->nfso_stateid.seqid = *tl++;
693 op->nfso_stateid.other[0] = *tl++;
694 op->nfso_stateid.other[1] = *tl++;
695 op->nfso_stateid.other[2] = *tl;
697 if (nd->nd_repstat && error == 0)
698 error = nd->nd_repstat;
699 if (error == NFSERR_STALESTATEID)
700 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
702 mbuf_freem(nd->nd_mrep);
707 * V4 Close operation.
710 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
712 struct nfsclclient *clp;
715 if (vnode_vtype(vp) != VREG)
718 error = nfscl_doclose(vp, &clp, p);
720 error = nfscl_getclose(vp, &clp);
724 nfscl_clientrelease(clp);
732 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
734 struct nfsrv_descript nfsd, *nd = &nfsd;
735 struct nfscllockowner *lp, *nlp;
736 struct nfscllock *lop, *nlop;
738 u_int64_t off = 0, len = 0;
739 u_int32_t type = NFSV4LOCKT_READ;
740 int error, do_unlock, trycnt;
742 tcred = newnfs_getcred();
743 newnfs_copycred(&op->nfso_cred, tcred);
745 * (Theoretically this could be done in the same
746 * compound as the close, but having multiple
747 * sequenced Ops in the same compound might be
748 * too scary for some servers.)
750 if (op->nfso_posixlock) {
753 type = NFSV4LOCKT_READ;
757 * Since this function is only called from VOP_INACTIVE(), no
758 * other thread will be manipulating this Open. As such, the
759 * lock lists are not being changed by other threads, so it should
760 * be safe to do this without locking.
762 LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
764 LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
765 if (op->nfso_posixlock == 0) {
766 off = lop->nfslo_first;
767 len = lop->nfslo_end - lop->nfslo_first;
768 if (lop->nfslo_type == F_WRLCK)
769 type = NFSV4LOCKT_WRITE;
771 type = NFSV4LOCKT_READ;
776 error = nfsrpc_locku(nd, nmp, lp, off,
777 len, type, tcred, p, 0);
778 if ((nd->nd_repstat == NFSERR_GRACE ||
779 nd->nd_repstat == NFSERR_DELAY) &&
781 (void) nfs_catnap(PZERO,
784 } while ((nd->nd_repstat == NFSERR_GRACE ||
785 nd->nd_repstat == NFSERR_DELAY) &&
786 error == 0 && trycnt++ < 5);
787 if (op->nfso_posixlock)
790 nfscl_freelock(lop, 0);
793 * Do a ReleaseLockOwner.
794 * The lock owner name nfsl_owner may be used by other opens for
795 * other files but the lock_owner4 name that nfsrpc_rellockown()
796 * puts on the wire has the file handle for this file appended
797 * to it, so it can be done now.
799 (void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
800 lp->nfsl_open->nfso_fhlen, tcred, p);
804 * There could be other Opens for different files on the same
805 * OpenOwner, so locking is required.
808 nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
811 error = nfscl_tryclose(op, tcred, nmp, p);
812 if (error == NFSERR_GRACE)
813 (void) nfs_catnap(PZERO, error, "nfs_close");
814 } while (error == NFSERR_GRACE);
816 nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
818 LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
819 nfscl_freelockowner(lp, 0);
820 nfscl_freeopen(op, 0);
826 * The actual Close RPC.
829 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
830 struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
836 nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
837 op->nfso_fhlen, NULL, NULL, 0, 0);
838 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
839 *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
840 if (NFSHASNFSV4N(nmp))
843 *tl++ = op->nfso_stateid.seqid;
844 *tl++ = op->nfso_stateid.other[0];
845 *tl++ = op->nfso_stateid.other[1];
846 *tl = op->nfso_stateid.other[2];
848 nd->nd_flag |= ND_USEGSSNAME;
849 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
850 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
853 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
854 if (nd->nd_repstat == 0)
855 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
856 error = nd->nd_repstat;
857 if (error == NFSERR_STALESTATEID)
858 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
860 mbuf_freem(nd->nd_mrep);
865 * V4 Open Confirm RPC.
868 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
869 struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
872 struct nfsrv_descript nfsd, *nd = &nfsd;
873 struct nfsmount *nmp;
876 nmp = VFSTONFS(vnode_mount(vp));
877 if (NFSHASNFSV4N(nmp))
878 return (0); /* No confirmation for NFSv4.1. */
879 nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
881 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
882 *tl++ = op->nfso_stateid.seqid;
883 *tl++ = op->nfso_stateid.other[0];
884 *tl++ = op->nfso_stateid.other[1];
885 *tl++ = op->nfso_stateid.other[2];
886 *tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
887 error = nfscl_request(nd, vp, p, cred, NULL);
890 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
891 if (!nd->nd_repstat) {
892 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
893 op->nfso_stateid.seqid = *tl++;
894 op->nfso_stateid.other[0] = *tl++;
895 op->nfso_stateid.other[1] = *tl++;
896 op->nfso_stateid.other[2] = *tl;
898 error = nd->nd_repstat;
899 if (error == NFSERR_STALESTATEID)
900 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
902 mbuf_freem(nd->nd_mrep);
907 * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
908 * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
911 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
912 struct ucred *cred, NFSPROC_T *p)
915 struct nfsrv_descript nfsd;
916 struct nfsrv_descript *nd = &nfsd;
917 nfsattrbit_t attrbits;
918 u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
920 int error, isinet6 = 0, callblen;
923 static u_int32_t rev = 0;
926 struct nfsclsession *tsep;
928 if (nfsboottime.tv_sec == 0)
929 NFSSETBOOTTIME(nfsboottime);
930 clp->nfsc_rev = rev++;
931 if (NFSHASNFSV4N(nmp)) {
933 * Either there was no previous session or the
934 * previous session has failed, so...
935 * do an ExchangeID followed by the CreateSession.
937 error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq,
938 NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp, cred, p);
939 NFSCL_DEBUG(1, "aft exch=%d\n", error);
941 error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
943 dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
947 * The old sessions cannot be safely free'd
948 * here, since they may still be used by
952 if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
953 tsep = NFSMNT_MDSSESSION(nmp);
954 TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
957 * Wake up RPCs waiting for a slot on the
958 * old session. These will then fail with
959 * NFSERR_BADSESSION and be retried with the
960 * new session by nfsv4_setsequence().
961 * Also wakeup() processes waiting for the
965 wakeup(&tsep->nfsess_slots);
966 wakeup(&nmp->nm_sess);
969 nfscl_freenfsclds(dsp);
970 NFSCL_DEBUG(1, "aft createsess=%d\n", error);
971 if (error == 0 && reclaim == 0) {
972 error = nfsrpc_reclaimcomplete(nmp, cred, p);
973 NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
974 if (error == NFSERR_COMPLETEALREADY ||
975 error == NFSERR_NOTSUPP)
976 /* Ignore this error. */
983 * Allocate a single session structure for NFSv4.0, because some of
984 * the fields are used by NFSv4.0 although it doesn't do a session.
986 dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
987 mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
988 mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
990 TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
991 tsep = NFSMNT_MDSSESSION(nmp);
994 nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
995 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
996 *tl++ = txdr_unsigned(nfsboottime.tv_sec);
997 *tl = txdr_unsigned(clp->nfsc_rev);
998 (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1001 * set up the callback address
1003 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1004 *tl = txdr_unsigned(NFS_CALLBCKPROG);
1005 callblen = strlen(nfsv4_callbackaddr);
1007 cp = nfscl_getmyip(nmp, &a6, &isinet6);
1008 if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1009 (callblen > 0 || cp != NULL)) {
1010 port = htons(nfsv4_cbport);
1011 cp2 = (u_int8_t *)&port;
1013 if ((callblen > 0 &&
1014 strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1015 char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1017 (void) nfsm_strtom(nd, "tcp6", 4);
1018 if (callblen == 0) {
1019 ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1022 ip6add = nfsv4_callbackaddr;
1024 snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1025 ip6add, cp2[0], cp2[1]);
1029 (void) nfsm_strtom(nd, "tcp", 3);
1031 snprintf(addr, INET6_ADDRSTRLEN + 9,
1032 "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1033 cp[2], cp[3], cp2[0], cp2[1]);
1035 snprintf(addr, INET6_ADDRSTRLEN + 9,
1036 "%s.%d.%d", nfsv4_callbackaddr,
1039 (void) nfsm_strtom(nd, addr, strlen(addr));
1041 (void) nfsm_strtom(nd, "tcp", 3);
1042 (void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1044 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1045 *tl = txdr_unsigned(clp->nfsc_cbident);
1046 nd->nd_flag |= ND_USEGSSNAME;
1047 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1048 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1051 if (nd->nd_repstat == 0) {
1052 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1053 tsep->nfsess_clientid.lval[0] = *tl++;
1054 tsep->nfsess_clientid.lval[1] = *tl++;
1055 confirm.lval[0] = *tl++;
1056 confirm.lval[1] = *tl;
1057 mbuf_freem(nd->nd_mrep);
1063 nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1065 NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1066 *tl++ = tsep->nfsess_clientid.lval[0];
1067 *tl++ = tsep->nfsess_clientid.lval[1];
1068 *tl++ = confirm.lval[0];
1069 *tl = confirm.lval[1];
1070 nd->nd_flag |= ND_USEGSSNAME;
1071 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1072 cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1075 mbuf_freem(nd->nd_mrep);
1077 if (nd->nd_repstat == 0) {
1078 nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
1079 nmp->nm_fhsize, NULL, NULL, 0, 0);
1080 NFSZERO_ATTRBIT(&attrbits);
1081 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1082 (void) nfsrv_putattrbit(nd, &attrbits);
1083 nd->nd_flag |= ND_USEGSSNAME;
1084 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1085 cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1088 if (nd->nd_repstat == 0) {
1089 error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
1090 NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
1093 clp->nfsc_renew = NFSCL_RENEW(lease);
1094 clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
1095 clp->nfsc_clientidrev++;
1096 if (clp->nfsc_clientidrev == 0)
1097 clp->nfsc_clientidrev++;
1101 error = nd->nd_repstat;
1103 mbuf_freem(nd->nd_mrep);
1111 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1112 struct nfsvattr *nap, void *stuff)
1114 struct nfsrv_descript nfsd, *nd = &nfsd;
1116 nfsattrbit_t attrbits;
1118 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1119 if (nd->nd_flag & ND_NFSV4) {
1120 NFSGETATTR_ATTRBIT(&attrbits);
1121 (void) nfsrv_putattrbit(nd, &attrbits);
1123 error = nfscl_request(nd, vp, p, cred, stuff);
1126 if (!nd->nd_repstat)
1127 error = nfsm_loadattr(nd, nap);
1129 error = nd->nd_repstat;
1130 mbuf_freem(nd->nd_mrep);
1135 * nfs getattr call with non-vnode arguemnts.
1138 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1139 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1142 struct nfsrv_descript nfsd, *nd = &nfsd;
1143 int error, vers = NFS_VER2;
1144 nfsattrbit_t attrbits;
1146 nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1147 if (nd->nd_flag & ND_NFSV4) {
1149 NFSGETATTR_ATTRBIT(&attrbits);
1150 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1151 (void) nfsrv_putattrbit(nd, &attrbits);
1152 } else if (nd->nd_flag & ND_NFSV3) {
1156 nd->nd_flag |= ND_USEGSSNAME;
1157 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1158 NFS_PROG, vers, NULL, 1, xidp, NULL);
1161 if (nd->nd_repstat == 0) {
1162 if ((nd->nd_flag & ND_NFSV4) != 0)
1163 error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1164 NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1167 error = nfsm_loadattr(nd, nap);
1169 error = nd->nd_repstat;
1170 mbuf_freem(nd->nd_mrep);
1175 * Do an nfs setattr operation.
1178 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1179 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1182 int error, expireret = 0, openerr, retrycnt;
1183 u_int32_t clidrev = 0, mode;
1184 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1186 nfsv4stateid_t stateid;
1189 if (nmp->nm_clp != NULL)
1190 clidrev = nmp->nm_clp->nfsc_clientidrev;
1191 if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1192 mode = NFSV4OPEN_ACCESSWRITE;
1194 mode = NFSV4OPEN_ACCESSREAD;
1199 if (NFSHASNFSV4(nmp)) {
1200 nfhp = VTONFS(vp)->n_fhp;
1201 error = nfscl_getstateid(vp, nfhp->nfh_fh,
1202 nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1203 if (error && vnode_vtype(vp) == VREG &&
1204 (mode == NFSV4OPEN_ACCESSWRITE ||
1205 nfstest_openallsetattr)) {
1207 * No Open stateid, so try and open the file
1210 if (mode == NFSV4OPEN_ACCESSWRITE)
1211 openerr = nfsrpc_open(vp, FWRITE, cred,
1214 openerr = nfsrpc_open(vp, FREAD, cred,
1217 (void) nfscl_getstateid(vp,
1218 nfhp->nfh_fh, nfhp->nfh_len,
1219 mode, 0, cred, p, &stateid, &lckp);
1223 error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1224 rnap, attrflagp, stuff);
1226 error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1228 if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1230 nmp->nm_state |= NFSSTA_OPENMODE;
1233 if (error == NFSERR_STALESTATEID)
1234 nfscl_initiate_recovery(nmp->nm_clp);
1236 nfscl_lockderef(lckp);
1238 (void) nfsrpc_close(vp, 0, p);
1239 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1240 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1241 error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1242 (void) nfs_catnap(PZERO, error, "nfs_setattr");
1243 } else if ((error == NFSERR_EXPIRED ||
1244 error == NFSERR_BADSTATEID) && clidrev != 0) {
1245 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1248 } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1249 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1250 error == NFSERR_BADSESSION ||
1251 (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1252 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1253 expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1254 (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1256 if (error && retrycnt >= 4)
1262 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1263 nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1264 struct nfsvattr *rnap, int *attrflagp, void *stuff)
1267 struct nfsrv_descript nfsd, *nd = &nfsd;
1269 nfsattrbit_t attrbits;
1272 NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1273 if (nd->nd_flag & ND_NFSV4)
1274 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1275 vap->va_type = vnode_vtype(vp);
1276 nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1277 if (nd->nd_flag & ND_NFSV3) {
1278 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1280 } else if (nd->nd_flag & ND_NFSV4) {
1281 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1282 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1283 NFSGETATTR_ATTRBIT(&attrbits);
1284 (void) nfsrv_putattrbit(nd, &attrbits);
1286 error = nfscl_request(nd, vp, p, cred, stuff);
1289 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1290 error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1291 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1292 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1293 if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1294 error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1295 mbuf_freem(nd->nd_mrep);
1296 if (nd->nd_repstat && !error)
1297 error = nd->nd_repstat;
1305 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1306 NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1307 struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1310 struct nfsrv_descript nfsd, *nd = &nfsd;
1311 struct nfsmount *nmp;
1314 nfsattrbit_t attrbits;
1315 int error = 0, lookupp = 0;
1319 if (vnode_vtype(dvp) != VDIR)
1321 nmp = VFSTONFS(vnode_mount(dvp));
1322 if (len > NFS_MAXNAMLEN)
1323 return (ENAMETOOLONG);
1324 if (NFSHASNFSV4(nmp) && len == 1 &&
1327 * Just return the current dir's fh.
1330 nfhp = malloc(sizeof (struct nfsfh) +
1331 np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1332 nfhp->nfh_len = np->n_fhp->nfh_len;
1333 NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1337 if (NFSHASNFSV4(nmp) && len == 2 &&
1338 name[0] == '.' && name[1] == '.') {
1340 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1342 NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1343 (void) nfsm_strtom(nd, name, len);
1345 if (nd->nd_flag & ND_NFSV4) {
1346 NFSGETATTR_ATTRBIT(&attrbits);
1347 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1348 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1349 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1350 (void) nfsrv_putattrbit(nd, &attrbits);
1352 error = nfscl_request(nd, dvp, p, cred, stuff);
1355 if (nd->nd_repstat) {
1357 * When an NFSv4 Lookupp returns ENOENT, it means that
1358 * the lookup is at the root of an fs, so return this dir.
1360 if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1362 nfhp = malloc(sizeof (struct nfsfh) +
1363 np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1364 nfhp->nfh_len = np->n_fhp->nfh_len;
1365 NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1367 mbuf_freem(nd->nd_mrep);
1370 if (nd->nd_flag & ND_NFSV3)
1371 error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1372 else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1374 /* Load the directory attributes. */
1375 error = nfsm_loadattr(nd, dnap);
1381 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1382 /* Load the directory attributes. */
1383 error = nfsm_loadattr(nd, dnap);
1387 /* Skip over the Lookup and GetFH operation status values. */
1388 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1390 error = nfsm_getfh(nd, nfhpp);
1394 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1395 if ((nd->nd_flag & ND_NFSV3) && !error)
1396 error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1398 mbuf_freem(nd->nd_mrep);
1399 if (!error && nd->nd_repstat)
1400 error = nd->nd_repstat;
1405 * Do a readlink rpc.
1408 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1409 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1412 struct nfsrv_descript nfsd, *nd = &nfsd;
1413 struct nfsnode *np = VTONFS(vp);
1414 nfsattrbit_t attrbits;
1415 int error, len, cangetattr = 1;
1418 NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1419 if (nd->nd_flag & ND_NFSV4) {
1421 * And do a Getattr op.
1423 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1424 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1425 NFSGETATTR_ATTRBIT(&attrbits);
1426 (void) nfsrv_putattrbit(nd, &attrbits);
1428 error = nfscl_request(nd, vp, p, cred, stuff);
1431 if (nd->nd_flag & ND_NFSV3)
1432 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1433 if (!nd->nd_repstat && !error) {
1434 NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1436 * This seems weird to me, but must have been added to
1437 * FreeBSD for some reason. The only thing I can think of
1438 * is that there was/is some server that replies with
1439 * more link data than it should?
1441 if (len == NFS_MAXPATHLEN) {
1443 if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1449 error = nfsm_mbufuio(nd, uiop, len);
1450 if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1451 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1453 if (nd->nd_repstat && !error)
1454 error = nd->nd_repstat;
1456 mbuf_freem(nd->nd_mrep);
1464 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1465 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1467 int error, expireret = 0, retrycnt;
1468 u_int32_t clidrev = 0;
1469 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1470 struct nfsnode *np = VTONFS(vp);
1471 struct ucred *newcred;
1472 struct nfsfh *nfhp = NULL;
1473 nfsv4stateid_t stateid;
1476 if (nmp->nm_clp != NULL)
1477 clidrev = nmp->nm_clp->nfsc_clientidrev;
1479 if (NFSHASNFSV4(nmp)) {
1481 newcred = NFSNEWCRED(cred);
1486 if (NFSHASNFSV4(nmp))
1487 (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1488 NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1490 error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1492 if (error == NFSERR_OPENMODE) {
1494 nmp->nm_state |= NFSSTA_OPENMODE;
1497 if (error == NFSERR_STALESTATEID)
1498 nfscl_initiate_recovery(nmp->nm_clp);
1500 nfscl_lockderef(lckp);
1501 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1502 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1503 error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1504 (void) nfs_catnap(PZERO, error, "nfs_read");
1505 } else if ((error == NFSERR_EXPIRED ||
1506 error == NFSERR_BADSTATEID) && clidrev != 0) {
1507 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1510 } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1511 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1512 error == NFSERR_BADSESSION ||
1513 (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1514 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1515 expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1516 (error == NFSERR_OPENMODE && retrycnt < 4));
1517 if (error && retrycnt >= 4)
1519 if (NFSHASNFSV4(nmp))
1520 NFSFREECRED(newcred);
1525 * The actual read RPC.
1528 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1529 nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1530 int *attrflagp, void *stuff)
1533 int error = 0, len, retlen, tsiz, eof = 0;
1534 struct nfsrv_descript nfsd;
1535 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1536 struct nfsrv_descript *nd = &nfsd;
1541 tsiz = uio_uio_resid(uiop);
1542 tmp_off = uiop->uio_offset + tsiz;
1544 if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1548 rsize = nmp->nm_rsize;
1553 len = (tsiz > rsize) ? rsize : tsiz;
1554 NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1555 if (nd->nd_flag & ND_NFSV4)
1556 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1557 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1558 if (nd->nd_flag & ND_NFSV2) {
1559 *tl++ = txdr_unsigned(uiop->uio_offset);
1560 *tl++ = txdr_unsigned(len);
1563 txdr_hyper(uiop->uio_offset, tl);
1564 *(tl + 2) = txdr_unsigned(len);
1567 * Since I can't do a Getattr for NFSv4 for Write, there
1568 * doesn't seem any point in doing one here, either.
1569 * (See the comment in nfsrpc_writerpc() for more info.)
1571 error = nfscl_request(nd, vp, p, cred, stuff);
1574 if (nd->nd_flag & ND_NFSV3) {
1575 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1576 } else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1577 error = nfsm_loadattr(nd, nap);
1581 if (nd->nd_repstat || error) {
1583 error = nd->nd_repstat;
1586 if (nd->nd_flag & ND_NFSV3) {
1587 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1588 eof = fxdr_unsigned(int, *(tl + 1));
1589 } else if (nd->nd_flag & ND_NFSV4) {
1590 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1591 eof = fxdr_unsigned(int, *tl);
1593 NFSM_STRSIZ(retlen, len);
1594 error = nfsm_mbufuio(nd, uiop, retlen);
1597 mbuf_freem(nd->nd_mrep);
1600 if (!(nd->nd_flag & ND_NFSV2)) {
1601 if (eof || retlen == 0)
1603 } else if (retlen < len)
1608 if (nd->nd_mrep != NULL)
1609 mbuf_freem(nd->nd_mrep);
1614 * nfs write operation
1615 * When called_from_strategy != 0, it should return EIO for an error that
1616 * indicates recovery is in progress, so that the buffer will be left
1617 * dirty and be written back to the server later. If it loops around,
1618 * the recovery thread could get stuck waiting for the buffer and recovery
1619 * will then deadlock.
1622 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1623 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1624 void *stuff, int called_from_strategy)
1626 int error, expireret = 0, retrycnt, nostateid;
1627 u_int32_t clidrev = 0;
1628 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1629 struct nfsnode *np = VTONFS(vp);
1630 struct ucred *newcred;
1631 struct nfsfh *nfhp = NULL;
1632 nfsv4stateid_t stateid;
1636 if (nmp->nm_clp != NULL)
1637 clidrev = nmp->nm_clp->nfsc_clientidrev;
1639 if (NFSHASNFSV4(nmp)) {
1640 newcred = NFSNEWCRED(cred);
1647 if (NFSHASNFSV4(nmp)) {
1648 (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1649 NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1651 if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1652 stateid.other[2] == 0) {
1654 NFSCL_DEBUG(1, "stateid0 in write\n");
1659 * If there is no stateid for NFSv4, it means this is an
1660 * extraneous write after close. Basically a poorly
1661 * implemented buffer cache. Just don't do the write.
1666 error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1667 newcred, &stateid, p, nap, attrflagp, stuff);
1668 if (error == NFSERR_STALESTATEID)
1669 nfscl_initiate_recovery(nmp->nm_clp);
1671 nfscl_lockderef(lckp);
1672 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1673 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1674 error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1675 (void) nfs_catnap(PZERO, error, "nfs_write");
1676 } else if ((error == NFSERR_EXPIRED ||
1677 error == NFSERR_BADSTATEID) && clidrev != 0) {
1678 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1681 } while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1682 ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1683 error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1684 (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1685 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1686 expireret == 0 && clidrev != 0 && retrycnt < 4));
1687 if (error != 0 && (retrycnt >= 4 ||
1688 ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1689 error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1691 if (NFSHASNFSV4(nmp))
1692 NFSFREECRED(newcred);
1697 * The actual write RPC.
1700 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1701 int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1702 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1705 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1706 struct nfsnode *np = VTONFS(vp);
1707 int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1708 int wccflag = 0, wsize;
1710 struct nfsrv_descript nfsd;
1711 struct nfsrv_descript *nd = &nfsd;
1712 nfsattrbit_t attrbits;
1715 KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1717 tsiz = uio_uio_resid(uiop);
1718 tmp_off = uiop->uio_offset + tsiz;
1720 if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1724 wsize = nmp->nm_wsize;
1726 nd->nd_mrep = NULL; /* NFSv2 sometimes does a write with */
1727 nd->nd_repstat = 0; /* uio_resid == 0, so the while is not done */
1730 len = (tsiz > wsize) ? wsize : tsiz;
1731 NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1732 if (nd->nd_flag & ND_NFSV4) {
1733 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1734 NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1735 txdr_hyper(uiop->uio_offset, tl);
1737 *tl++ = txdr_unsigned(*iomode);
1738 *tl = txdr_unsigned(len);
1739 } else if (nd->nd_flag & ND_NFSV3) {
1740 NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1741 txdr_hyper(uiop->uio_offset, tl);
1743 *tl++ = txdr_unsigned(len);
1744 *tl++ = txdr_unsigned(*iomode);
1745 *tl = txdr_unsigned(len);
1749 NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1751 * Not sure why someone changed this, since the
1752 * RFC clearly states that "beginoffset" and
1753 * "totalcount" are ignored, but it wouldn't
1754 * surprise me if there's a busted server out there.
1756 /* Set both "begin" and "current" to non-garbage. */
1757 x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1758 *tl++ = x; /* "begin offset" */
1759 *tl++ = x; /* "current offset" */
1760 x = txdr_unsigned(len);
1761 *tl++ = x; /* total to this offset */
1762 *tl = x; /* size of this write */
1765 nfsm_uiombuf(nd, uiop, len);
1767 * Although it is tempting to do a normal Getattr Op in the
1768 * NFSv4 compound, the result can be a nearly hung client
1769 * system if the Getattr asks for Owner and/or OwnerGroup.
1770 * It occurs when the client can't map either the Owner or
1771 * Owner_group name in the Getattr reply to a uid/gid. When
1772 * there is a cache miss, the kernel does an upcall to the
1773 * nfsuserd. Then, it can try and read the local /etc/passwd
1774 * or /etc/group file. It can then block in getnewbuf(),
1775 * waiting for dirty writes to be pushed to the NFS server.
1776 * The only reason this doesn't result in a complete
1777 * deadlock, is that the upcall times out and allows
1778 * the write to complete. However, progress is so slow
1779 * that it might just as well be deadlocked.
1780 * As such, we get the rest of the attributes, but not
1781 * Owner or Owner_group.
1782 * nb: nfscl_loadattrcache() needs to be told that these
1783 * partial attributes from a write rpc are being
1784 * passed in, via a argument flag.
1786 if (nd->nd_flag & ND_NFSV4) {
1787 NFSWRITEGETATTR_ATTRBIT(&attrbits);
1788 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1789 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1790 (void) nfsrv_putattrbit(nd, &attrbits);
1792 error = nfscl_request(nd, vp, p, cred, stuff);
1795 if (nd->nd_repstat) {
1797 * In case the rpc gets retried, roll
1798 * the uio fileds changed by nfsm_uiombuf()
1801 uiop->uio_offset -= len;
1802 uio_uio_resid_add(uiop, len);
1803 uio_iov_base_add(uiop, -len);
1804 uio_iov_len_add(uiop, len);
1806 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1807 error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1812 if (!nd->nd_repstat) {
1813 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1814 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1816 rlen = fxdr_unsigned(int, *tl++);
1820 } else if (rlen < len) {
1821 backup = len - rlen;
1822 uio_iov_base_add(uiop, -(backup));
1823 uio_iov_len_add(uiop, backup);
1824 uiop->uio_offset -= backup;
1825 uio_uio_resid_add(uiop, backup);
1828 commit = fxdr_unsigned(int, *tl++);
1831 * Return the lowest commitment level
1832 * obtained by any of the RPCs.
1834 if (committed == NFSWRITE_FILESYNC)
1836 else if (committed == NFSWRITE_DATASYNC &&
1837 commit == NFSWRITE_UNSTABLE)
1840 if (!NFSHASWRITEVERF(nmp)) {
1841 NFSBCOPY((caddr_t)tl,
1842 (caddr_t)&nmp->nm_verf[0],
1844 NFSSETWRITEVERF(nmp);
1845 } else if (NFSBCMP(tl, nmp->nm_verf,
1848 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1852 if (nd->nd_flag & ND_NFSV4)
1853 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1854 if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1855 error = nfsm_loadattr(nd, nap);
1857 *attrflagp = NFS_LATTR_NOSHRINK;
1860 error = nd->nd_repstat;
1864 NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1865 mbuf_freem(nd->nd_mrep);
1870 if (nd->nd_mrep != NULL)
1871 mbuf_freem(nd->nd_mrep);
1872 *iomode = committed;
1873 if (nd->nd_repstat && !error)
1874 error = nd->nd_repstat;
1880 * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1881 * mode set to specify the file type and the size field for rdev.
1884 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1885 u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1886 struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1887 int *attrflagp, int *dattrflagp, void *dstuff)
1891 struct nfsrv_descript nfsd, *nd = &nfsd;
1892 nfsattrbit_t attrbits;
1897 if (namelen > NFS_MAXNAMLEN)
1898 return (ENAMETOOLONG);
1899 NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1900 if (nd->nd_flag & ND_NFSV4) {
1901 if (vtyp == VBLK || vtyp == VCHR) {
1902 NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1903 *tl++ = vtonfsv34_type(vtyp);
1904 *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1905 *tl = txdr_unsigned(NFSMINOR(rdev));
1907 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1908 *tl = vtonfsv34_type(vtyp);
1911 (void) nfsm_strtom(nd, name, namelen);
1912 if (nd->nd_flag & ND_NFSV3) {
1913 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1914 *tl = vtonfsv34_type(vtyp);
1916 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1917 nfscl_fillsattr(nd, vap, dvp, 0, 0);
1918 if ((nd->nd_flag & ND_NFSV3) &&
1919 (vtyp == VCHR || vtyp == VBLK)) {
1920 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1921 *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1922 *tl = txdr_unsigned(NFSMINOR(rdev));
1924 if (nd->nd_flag & ND_NFSV4) {
1925 NFSGETATTR_ATTRBIT(&attrbits);
1926 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1927 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1928 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1929 (void) nfsrv_putattrbit(nd, &attrbits);
1931 if (nd->nd_flag & ND_NFSV2)
1932 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
1933 error = nfscl_request(nd, dvp, p, cred, dstuff);
1936 if (nd->nd_flag & ND_NFSV4)
1937 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1938 if (!nd->nd_repstat) {
1939 if (nd->nd_flag & ND_NFSV4) {
1940 NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1941 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1945 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1949 if (nd->nd_flag & ND_NFSV3)
1950 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1951 if (!error && nd->nd_repstat)
1952 error = nd->nd_repstat;
1954 mbuf_freem(nd->nd_mrep);
1959 * nfs file create call
1960 * Mostly just call the approriate routine. (I separated out v4, so that
1961 * error recovery wouldn't be as difficult.)
1964 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1965 nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1966 struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1967 int *attrflagp, int *dattrflagp, void *dstuff)
1969 int error = 0, newone, expireret = 0, retrycnt, unlocked;
1970 struct nfsclowner *owp;
1971 struct nfscldeleg *dp;
1972 struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
1975 if (NFSHASNFSV4(nmp)) {
1979 error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
1980 NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
1984 if (nmp->nm_clp != NULL)
1985 clidrev = nmp->nm_clp->nfsc_clientidrev;
1988 if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
1989 nfs_numnfscbd == 0 || retrycnt > 0)
1990 error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
1991 fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
1992 attrflagp, dattrflagp, dstuff, &unlocked);
1994 error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
1995 cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
1996 attrflagp, dattrflagp, dstuff, &unlocked);
1998 * There is no need to invalidate cached attributes here,
1999 * since new post-delegation issue attributes are always
2000 * returned by nfsrpc_createv4() and these will update the
2004 (void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2005 (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2006 nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2007 if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2008 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2009 error == NFSERR_BADSESSION) {
2010 (void) nfs_catnap(PZERO, error, "nfs_open");
2011 } else if ((error == NFSERR_EXPIRED ||
2012 error == NFSERR_BADSTATEID) && clidrev != 0) {
2013 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2016 } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2017 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2018 error == NFSERR_BADSESSION ||
2019 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2020 expireret == 0 && clidrev != 0 && retrycnt < 4));
2021 if (error && retrycnt >= 4)
2024 error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2025 fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2032 * The create rpc for v2 and 3.
2035 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2036 nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2037 struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2038 int *attrflagp, int *dattrflagp, void *dstuff)
2042 struct nfsrv_descript nfsd, *nd = &nfsd;
2047 if (namelen > NFS_MAXNAMLEN)
2048 return (ENAMETOOLONG);
2049 NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2050 (void) nfsm_strtom(nd, name, namelen);
2051 if (nd->nd_flag & ND_NFSV3) {
2052 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2053 if (fmode & O_EXCL) {
2054 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2055 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2056 *tl++ = cverf.lval[0];
2057 *tl = cverf.lval[1];
2059 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2060 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2063 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2065 error = nfscl_request(nd, dvp, p, cred, dstuff);
2068 if (nd->nd_repstat == 0) {
2069 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2073 if (nd->nd_flag & ND_NFSV3)
2074 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2075 if (nd->nd_repstat != 0 && error == 0)
2076 error = nd->nd_repstat;
2078 mbuf_freem(nd->nd_mrep);
2083 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2084 nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2085 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2086 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2087 int *dattrflagp, void *dstuff, int *unlockedp)
2090 int error = 0, deleg, newone, ret, acesize, limitby;
2091 struct nfsrv_descript nfsd, *nd = &nfsd;
2092 struct nfsclopen *op;
2093 struct nfscldeleg *dp = NULL;
2096 nfsattrbit_t attrbits;
2097 nfsv4stateid_t stateid;
2099 struct nfsmount *nmp;
2100 struct nfsclsession *tsep;
2102 nmp = VFSTONFS(dvp->v_mount);
2109 if (namelen > NFS_MAXNAMLEN)
2110 return (ENAMETOOLONG);
2111 NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2113 * For V4, this is actually an Open op.
2115 NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2116 *tl++ = txdr_unsigned(owp->nfsow_seqid);
2117 *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2118 NFSV4OPEN_ACCESSREAD);
2119 *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2120 tsep = nfsmnt_mdssession(nmp);
2121 *tl++ = tsep->nfsess_clientid.lval[0];
2122 *tl = tsep->nfsess_clientid.lval[1];
2123 (void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2124 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2125 *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2126 if (fmode & O_EXCL) {
2127 if (NFSHASNFSV4N(nmp)) {
2128 if (NFSHASSESSPERSIST(nmp)) {
2129 /* Use GUARDED for persistent sessions. */
2130 *tl = txdr_unsigned(NFSCREATE_GUARDED);
2131 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2133 /* Otherwise, use EXCLUSIVE4_1. */
2134 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2135 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2136 *tl++ = cverf.lval[0];
2137 *tl = cverf.lval[1];
2138 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2142 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2143 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2144 *tl++ = cverf.lval[0];
2145 *tl = cverf.lval[1];
2148 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2149 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2151 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2152 *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2153 (void) nfsm_strtom(nd, name, namelen);
2154 /* Get the new file's handle and attributes. */
2155 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2156 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2157 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2158 NFSGETATTR_ATTRBIT(&attrbits);
2159 (void) nfsrv_putattrbit(nd, &attrbits);
2160 /* Get the directory's post-op attributes. */
2161 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2162 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2163 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2164 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2165 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2166 (void) nfsrv_putattrbit(nd, &attrbits);
2167 error = nfscl_request(nd, dvp, p, cred, dstuff);
2170 NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2171 if (nd->nd_repstat == 0) {
2172 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2174 stateid.seqid = *tl++;
2175 stateid.other[0] = *tl++;
2176 stateid.other[1] = *tl++;
2177 stateid.other[2] = *tl;
2178 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2179 (void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2180 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2181 deleg = fxdr_unsigned(int, *tl);
2182 if (deleg == NFSV4OPEN_DELEGATEREAD ||
2183 deleg == NFSV4OPEN_DELEGATEWRITE) {
2184 if (!(owp->nfsow_clp->nfsc_flags &
2185 NFSCLFLAGS_FIRSTDELEG))
2186 owp->nfsow_clp->nfsc_flags |=
2187 (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2189 sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2190 M_NFSCLDELEG, M_WAITOK);
2191 LIST_INIT(&dp->nfsdl_owner);
2192 LIST_INIT(&dp->nfsdl_lock);
2193 dp->nfsdl_clp = owp->nfsow_clp;
2194 newnfs_copyincred(cred, &dp->nfsdl_cred);
2195 nfscl_lockinit(&dp->nfsdl_rwlock);
2196 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2198 dp->nfsdl_stateid.seqid = *tl++;
2199 dp->nfsdl_stateid.other[0] = *tl++;
2200 dp->nfsdl_stateid.other[1] = *tl++;
2201 dp->nfsdl_stateid.other[2] = *tl++;
2202 ret = fxdr_unsigned(int, *tl);
2203 if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2204 dp->nfsdl_flags = NFSCLDL_WRITE;
2206 * Indicates how much the file can grow.
2208 NFSM_DISSECT(tl, u_int32_t *,
2210 limitby = fxdr_unsigned(int, *tl++);
2212 case NFSV4OPEN_LIMITSIZE:
2213 dp->nfsdl_sizelimit = fxdr_hyper(tl);
2215 case NFSV4OPEN_LIMITBLOCKS:
2216 dp->nfsdl_sizelimit =
2217 fxdr_unsigned(u_int64_t, *tl++);
2218 dp->nfsdl_sizelimit *=
2219 fxdr_unsigned(u_int64_t, *tl);
2222 error = NFSERR_BADXDR;
2226 dp->nfsdl_flags = NFSCLDL_READ;
2229 dp->nfsdl_flags |= NFSCLDL_RECALL;
2230 error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2234 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
2235 error = NFSERR_BADXDR;
2238 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2241 /* Get rid of the PutFH and Getattr status values. */
2242 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2243 /* Load the directory attributes. */
2244 error = nfsm_loadattr(nd, dnap);
2248 if (dp != NULL && *attrflagp) {
2249 dp->nfsdl_change = nnap->na_filerev;
2250 dp->nfsdl_modtime = nnap->na_mtime;
2251 dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2254 * We can now complete the Open state.
2258 dp->nfsdl_fhlen = nfhp->nfh_len;
2259 NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2262 * Get an Open structure that will be
2263 * attached to the OpenOwner, acquired already.
2265 error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2266 (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2267 cred, p, NULL, &op, &newone, NULL, 0);
2270 op->nfso_stateid = stateid;
2271 newnfs_copyincred(cred, &op->nfso_cred);
2272 if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2274 ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2275 nfhp->nfh_len, op, cred, p);
2276 if (ret == NFSERR_DELAY)
2277 (void) nfs_catnap(PZERO, ret, "nfs_create");
2278 } while (ret == NFSERR_DELAY);
2283 * If the server is handing out delegations, but we didn't
2284 * get one because an OpenConfirm was required, try the
2285 * Open again, to get a delegation. This is a harmless no-op,
2286 * from a server's point of view.
2288 if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2289 (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2290 !error && dp == NULL) {
2292 ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2293 np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2294 nfhp->nfh_fh, nfhp->nfh_len,
2295 (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2296 name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2297 if (ret == NFSERR_DELAY)
2298 (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2299 } while (ret == NFSERR_DELAY);
2302 free(dp, M_NFSCLDELEG);
2305 if (ret == NFSERR_STALECLIENTID ||
2306 ret == NFSERR_STALEDONTRECOVER ||
2307 ret == NFSERR_BADSESSION)
2311 nfscl_openrelease(nmp, op, error, newone);
2314 if (nd->nd_repstat != 0 && error == 0)
2315 error = nd->nd_repstat;
2316 if (error == NFSERR_STALECLIENTID)
2317 nfscl_initiate_recovery(owp->nfsow_clp);
2321 else if (dp != NULL)
2322 free(dp, M_NFSCLDELEG);
2323 mbuf_freem(nd->nd_mrep);
2331 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2332 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2336 struct nfsrv_descript nfsd, *nd = &nfsd;
2338 struct nfsmount *nmp;
2339 nfsv4stateid_t dstateid;
2340 int error, ret = 0, i;
2343 if (namelen > NFS_MAXNAMLEN)
2344 return (ENAMETOOLONG);
2345 nmp = VFSTONFS(vnode_mount(dvp));
2347 if (NFSHASNFSV4(nmp) && ret == 0) {
2348 ret = nfscl_removedeleg(vp, p, &dstateid);
2350 NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2351 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2353 if (NFSHASNFSV4N(nmp))
2356 *tl++ = dstateid.seqid;
2357 *tl++ = dstateid.other[0];
2358 *tl++ = dstateid.other[1];
2359 *tl++ = dstateid.other[2];
2360 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2362 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2363 np->n_fhp->nfh_len, 0);
2364 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2365 *tl = txdr_unsigned(NFSV4OP_REMOVE);
2371 NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2372 (void) nfsm_strtom(nd, name, namelen);
2373 error = nfscl_request(nd, dvp, p, cred, dstuff);
2376 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2377 /* For NFSv4, parse out any Delereturn replies. */
2378 if (ret > 0 && nd->nd_repstat != 0 &&
2379 (nd->nd_flag & ND_NOMOREDATA)) {
2381 * If the Delegreturn failed, try again without
2382 * it. The server will Recall, as required.
2384 mbuf_freem(nd->nd_mrep);
2387 for (i = 0; i < (ret * 2); i++) {
2388 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2390 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2392 nd->nd_flag |= ND_NOMOREDATA;
2395 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2397 if (nd->nd_repstat && !error)
2398 error = nd->nd_repstat;
2400 mbuf_freem(nd->nd_mrep);
2405 * Do an nfs rename rpc.
2408 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2409 vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2410 NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2411 int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2414 struct nfsrv_descript nfsd, *nd = &nfsd;
2415 struct nfsmount *nmp;
2417 nfsattrbit_t attrbits;
2418 nfsv4stateid_t fdstateid, tdstateid;
2419 int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2423 nmp = VFSTONFS(vnode_mount(fdvp));
2424 if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2425 return (ENAMETOOLONG);
2427 if (NFSHASNFSV4(nmp) && ret == 0) {
2428 ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2429 &tdstateid, &gottd, p);
2430 if (gotfd && gottd) {
2431 NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2433 NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2435 NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2438 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2439 if (NFSHASNFSV4N(nmp))
2442 *tl++ = fdstateid.seqid;
2443 *tl++ = fdstateid.other[0];
2444 *tl++ = fdstateid.other[1];
2445 *tl = fdstateid.other[2];
2447 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2448 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2450 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2451 np->n_fhp->nfh_len, 0);
2452 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2453 *tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2457 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2458 if (NFSHASNFSV4N(nmp))
2461 *tl++ = tdstateid.seqid;
2462 *tl++ = tdstateid.other[0];
2463 *tl++ = tdstateid.other[1];
2464 *tl = tdstateid.other[2];
2467 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2468 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2470 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2471 np->n_fhp->nfh_len, 0);
2472 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2473 *tl = txdr_unsigned(NFSV4OP_SAVEFH);
2479 NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2480 if (nd->nd_flag & ND_NFSV4) {
2481 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2482 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2483 NFSWCCATTR_ATTRBIT(&attrbits);
2484 (void) nfsrv_putattrbit(nd, &attrbits);
2485 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2486 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2487 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2488 VTONFS(tdvp)->n_fhp->nfh_len, 0);
2489 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2490 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2491 (void) nfsrv_putattrbit(nd, &attrbits);
2492 nd->nd_flag |= ND_V4WCCATTR;
2493 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2494 *tl = txdr_unsigned(NFSV4OP_RENAME);
2496 (void) nfsm_strtom(nd, fnameptr, fnamelen);
2497 if (!(nd->nd_flag & ND_NFSV4))
2498 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2499 VTONFS(tdvp)->n_fhp->nfh_len, 0);
2500 (void) nfsm_strtom(nd, tnameptr, tnamelen);
2501 error = nfscl_request(nd, fdvp, p, cred, fstuff);
2504 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2505 /* For NFSv4, parse out any Delereturn replies. */
2506 if (ret > 0 && nd->nd_repstat != 0 &&
2507 (nd->nd_flag & ND_NOMOREDATA)) {
2509 * If the Delegreturn failed, try again without
2510 * it. The server will Recall, as required.
2512 mbuf_freem(nd->nd_mrep);
2515 for (i = 0; i < (ret * 2); i++) {
2516 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2518 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2520 if (i == 0 && ret > 1) {
2522 * If the Delegreturn failed, try again
2523 * without it. The server will Recall, as
2525 * If ret > 1, the first iteration of this
2526 * loop is the second DelegReturn result.
2528 mbuf_freem(nd->nd_mrep);
2531 nd->nd_flag |= ND_NOMOREDATA;
2536 /* Now, the first wcc attribute reply. */
2537 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2538 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2540 nd->nd_flag |= ND_NOMOREDATA;
2542 error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2544 /* and the second wcc attribute reply. */
2545 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2547 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2549 nd->nd_flag |= ND_NOMOREDATA;
2552 error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2555 if (nd->nd_repstat && !error)
2556 error = nd->nd_repstat;
2558 mbuf_freem(nd->nd_mrep);
2563 * nfs hard link create rpc
2566 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2567 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2568 struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2571 struct nfsrv_descript nfsd, *nd = &nfsd;
2572 nfsattrbit_t attrbits;
2577 if (namelen > NFS_MAXNAMLEN)
2578 return (ENAMETOOLONG);
2579 NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2580 if (nd->nd_flag & ND_NFSV4) {
2581 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2582 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2584 (void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2585 VTONFS(dvp)->n_fhp->nfh_len, 0);
2586 if (nd->nd_flag & ND_NFSV4) {
2587 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2588 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2589 NFSWCCATTR_ATTRBIT(&attrbits);
2590 (void) nfsrv_putattrbit(nd, &attrbits);
2591 nd->nd_flag |= ND_V4WCCATTR;
2592 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2593 *tl = txdr_unsigned(NFSV4OP_LINK);
2595 (void) nfsm_strtom(nd, name, namelen);
2596 error = nfscl_request(nd, vp, p, cred, dstuff);
2599 if (nd->nd_flag & ND_NFSV3) {
2600 error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2602 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2604 } else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2606 * First, parse out the PutFH and Getattr result.
2608 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2610 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2612 nd->nd_flag |= ND_NOMOREDATA;
2614 * Get the pre-op attributes.
2616 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2618 if (nd->nd_repstat && !error)
2619 error = nd->nd_repstat;
2621 mbuf_freem(nd->nd_mrep);
2626 * nfs symbolic link create rpc
2629 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, char *target,
2630 struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2631 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2632 int *dattrflagp, void *dstuff)
2635 struct nfsrv_descript nfsd, *nd = &nfsd;
2636 struct nfsmount *nmp;
2637 int slen, error = 0;
2642 nmp = VFSTONFS(vnode_mount(dvp));
2643 slen = strlen(target);
2644 if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2645 return (ENAMETOOLONG);
2646 NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2647 if (nd->nd_flag & ND_NFSV4) {
2648 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2649 *tl = txdr_unsigned(NFLNK);
2650 (void) nfsm_strtom(nd, target, slen);
2652 (void) nfsm_strtom(nd, name, namelen);
2653 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2654 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2655 if (!(nd->nd_flag & ND_NFSV4))
2656 (void) nfsm_strtom(nd, target, slen);
2657 if (nd->nd_flag & ND_NFSV2)
2658 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2659 error = nfscl_request(nd, dvp, p, cred, dstuff);
2662 if (nd->nd_flag & ND_NFSV4)
2663 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2664 if ((nd->nd_flag & ND_NFSV3) && !error) {
2665 if (!nd->nd_repstat)
2666 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2668 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2671 if (nd->nd_repstat && !error)
2672 error = nd->nd_repstat;
2673 mbuf_freem(nd->nd_mrep);
2675 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2676 * Only do this if vfs.nfs.ignore_eexist is set.
2677 * Never do this for NFSv4.1 or later minor versions, since sessions
2678 * should guarantee "exactly once" RPC semantics.
2680 if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2681 nmp->nm_minorvers == 0))
2690 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2691 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2692 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2693 int *dattrflagp, void *dstuff)
2696 struct nfsrv_descript nfsd, *nd = &nfsd;
2697 nfsattrbit_t attrbits;
2700 struct nfsmount *nmp;
2705 nmp = VFSTONFS(vnode_mount(dvp));
2706 fhp = VTONFS(dvp)->n_fhp;
2707 if (namelen > NFS_MAXNAMLEN)
2708 return (ENAMETOOLONG);
2709 NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2710 if (nd->nd_flag & ND_NFSV4) {
2711 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2712 *tl = txdr_unsigned(NFDIR);
2714 (void) nfsm_strtom(nd, name, namelen);
2715 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2716 if (nd->nd_flag & ND_NFSV4) {
2717 NFSGETATTR_ATTRBIT(&attrbits);
2718 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2719 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2720 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2721 (void) nfsrv_putattrbit(nd, &attrbits);
2722 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2723 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2724 (void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2725 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2726 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2727 (void) nfsrv_putattrbit(nd, &attrbits);
2729 error = nfscl_request(nd, dvp, p, cred, dstuff);
2732 if (nd->nd_flag & ND_NFSV4)
2733 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2734 if (!nd->nd_repstat && !error) {
2735 if (nd->nd_flag & ND_NFSV4) {
2736 NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2737 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2740 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2741 if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2742 /* Get rid of the PutFH and Getattr status values. */
2743 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2744 /* Load the directory attributes. */
2745 error = nfsm_loadattr(nd, dnap);
2750 if ((nd->nd_flag & ND_NFSV3) && !error)
2751 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2752 if (nd->nd_repstat && !error)
2753 error = nd->nd_repstat;
2755 mbuf_freem(nd->nd_mrep);
2757 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2758 * Only do this if vfs.nfs.ignore_eexist is set.
2759 * Never do this for NFSv4.1 or later minor versions, since sessions
2760 * should guarantee "exactly once" RPC semantics.
2762 if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2763 nmp->nm_minorvers == 0))
2769 * nfs remove directory call
2772 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2773 NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2775 struct nfsrv_descript nfsd, *nd = &nfsd;
2779 if (namelen > NFS_MAXNAMLEN)
2780 return (ENAMETOOLONG);
2781 NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2782 (void) nfsm_strtom(nd, name, namelen);
2783 error = nfscl_request(nd, dvp, p, cred, dstuff);
2786 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2787 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2788 if (nd->nd_repstat && !error)
2789 error = nd->nd_repstat;
2790 mbuf_freem(nd->nd_mrep);
2792 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2794 if (error == ENOENT)
2801 * Always returns with either uio_resid unchanged, if you are at the
2802 * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2804 * I felt this would allow caching of directory blocks more easily
2805 * than returning a pertially filled block.
2806 * Directory offset cookies:
2807 * Oh my, what to do with them...
2808 * I can think of three ways to deal with them:
2809 * 1 - have the layer above these RPCs maintain a map between logical
2810 * directory byte offsets and the NFS directory offset cookies
2811 * 2 - pass the opaque directory offset cookies up into userland
2812 * and let the libc functions deal with them, via the system call
2813 * 3 - return them to userland in the "struct dirent", so future versions
2814 * of libc can use them and do whatever is necessary to make things work
2815 * above these rpc calls, in the meantime
2816 * For now, I do #3 by "hiding" the directory offset cookies after the
2817 * d_name field in struct dirent. This is space inside d_reclen that
2818 * will be ignored by anything that doesn't know about them.
2819 * The directory offset cookies are filled in as the last 8 bytes of
2820 * each directory entry, after d_name. Someday, the userland libc
2821 * functions may be able to use these. In the meantime, it satisfies
2822 * OpenBSD's requirements for cookies being returned.
2823 * If expects the directory offset cookie for the read to be in uio_offset
2824 * and returns the one for the next entry after this directory block in
2828 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2829 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2830 int *eofp, void *stuff)
2833 struct dirent *dp = NULL;
2835 nfsquad_t cookie, ncookie;
2836 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2837 struct nfsnode *dnp = VTONFS(vp);
2838 struct nfsvattr nfsva;
2839 struct nfsrv_descript nfsd, *nd = &nfsd;
2840 int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2841 int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2842 u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2844 nfsattrbit_t attrbits, dattrbits;
2845 u_int32_t rderr, *tl2 = NULL;
2848 KASSERT(uiop->uio_iovcnt == 1 &&
2849 (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2850 ("nfs readdirrpc bad uio"));
2851 ncookie.lval[0] = ncookie.lval[1] = 0;
2853 * There is no point in reading a lot more than uio_resid, however
2854 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2855 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2856 * will never make readsize > nm_readdirsize.
2858 readsize = nmp->nm_readdirsize;
2859 if (readsize > uio_uio_resid(uiop))
2860 readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2865 tresid = uio_uio_resid(uiop);
2866 cookie.lval[0] = cookiep->nfsuquad[0];
2867 cookie.lval[1] = cookiep->nfsuquad[1];
2871 * For NFSv4, first create the "." and ".." entries.
2873 if (NFSHASNFSV4(nmp)) {
2874 reqsize = 6 * NFSX_UNSIGNED;
2875 NFSGETATTR_ATTRBIT(&dattrbits);
2876 NFSZERO_ATTRBIT(&attrbits);
2877 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2878 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2879 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2880 NFSATTRBIT_MOUNTEDONFILEID)) {
2881 NFSSETBIT_ATTRBIT(&attrbits,
2882 NFSATTRBIT_MOUNTEDONFILEID);
2886 * Must fake it. Use the fileno, except when the
2887 * fsid is != to that of the directory. For that
2888 * case, generate a fake fileno that is not the same.
2890 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2895 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2897 if (uiop->uio_offset == 0) {
2898 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2899 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2900 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2901 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2902 (void) nfsrv_putattrbit(nd, &attrbits);
2903 error = nfscl_request(nd, vp, p, cred, stuff);
2906 dotfileid = 0; /* Fake out the compiler. */
2907 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2908 error = nfsm_loadattr(nd, &nfsva);
2911 dotfileid = nfsva.na_fileid;
2913 if (nd->nd_repstat == 0) {
2914 NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2915 len = fxdr_unsigned(int, *(tl + 4));
2916 if (len > 0 && len <= NFSX_V4FHMAX)
2917 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2921 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2922 nfsva.na_mntonfileno = UINT64_MAX;
2923 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2924 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2925 NULL, NULL, NULL, p, cred);
2927 dotdotfileid = dotfileid;
2928 } else if (gotmnton) {
2929 if (nfsva.na_mntonfileno != UINT64_MAX)
2930 dotdotfileid = nfsva.na_mntonfileno;
2932 dotdotfileid = nfsva.na_fileid;
2933 } else if (nfsva.na_filesid[0] ==
2934 dnp->n_vattr.na_filesid[0] &&
2935 nfsva.na_filesid[1] ==
2936 dnp->n_vattr.na_filesid[1]) {
2937 dotdotfileid = nfsva.na_fileid;
2941 } while (fakefileno ==
2943 dotdotfileid = fakefileno;
2946 } else if (nd->nd_repstat == NFSERR_NOENT) {
2948 * Lookupp returns NFSERR_NOENT when we are
2949 * at the root, so just use the current dir.
2952 dotdotfileid = dotfileid;
2954 error = nd->nd_repstat;
2956 mbuf_freem(nd->nd_mrep);
2960 dp = (struct dirent *)uio_iov_base(uiop);
2961 dp->d_pad0 = dp->d_pad1 = 0;
2963 dp->d_type = DT_DIR;
2964 dp->d_fileno = dotfileid;
2966 *((uint64_t *)dp->d_name) = 0; /* Zero pad it. */
2967 dp->d_name[0] = '.';
2968 dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
2970 * Just make these offset cookie 0.
2972 tl = (u_int32_t *)&dp->d_name[8];
2975 blksiz += dp->d_reclen;
2976 uio_uio_resid_add(uiop, -(dp->d_reclen));
2977 uiop->uio_offset += dp->d_reclen;
2978 uio_iov_base_add(uiop, dp->d_reclen);
2979 uio_iov_len_add(uiop, -(dp->d_reclen));
2980 dp = (struct dirent *)uio_iov_base(uiop);
2981 dp->d_pad0 = dp->d_pad1 = 0;
2983 dp->d_type = DT_DIR;
2984 dp->d_fileno = dotdotfileid;
2986 *((uint64_t *)dp->d_name) = 0;
2987 dp->d_name[0] = '.';
2988 dp->d_name[1] = '.';
2989 dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
2991 * Just make these offset cookie 0.
2993 tl = (u_int32_t *)&dp->d_name[8];
2996 blksiz += dp->d_reclen;
2997 uio_uio_resid_add(uiop, -(dp->d_reclen));
2998 uiop->uio_offset += dp->d_reclen;
2999 uio_iov_base_add(uiop, dp->d_reclen);
3000 uio_iov_len_add(uiop, -(dp->d_reclen));
3002 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3004 reqsize = 5 * NFSX_UNSIGNED;
3009 * Loop around doing readdir rpc's of size readsize.
3010 * The stopping criteria is EOF or buffer full.
3012 while (more_dirs && bigenough) {
3014 NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3015 if (nd->nd_flag & ND_NFSV2) {
3016 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3017 *tl++ = cookie.lval[1];
3018 *tl = txdr_unsigned(readsize);
3020 NFSM_BUILD(tl, u_int32_t *, reqsize);
3021 *tl++ = cookie.lval[0];
3022 *tl++ = cookie.lval[1];
3023 if (cookie.qval == 0) {
3028 *tl++ = dnp->n_cookieverf.nfsuquad[0];
3029 *tl++ = dnp->n_cookieverf.nfsuquad[1];
3032 if (nd->nd_flag & ND_NFSV4) {
3033 *tl++ = txdr_unsigned(readsize);
3034 *tl = txdr_unsigned(readsize);
3035 (void) nfsrv_putattrbit(nd, &attrbits);
3036 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3037 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3038 (void) nfsrv_putattrbit(nd, &dattrbits);
3040 *tl = txdr_unsigned(readsize);
3043 error = nfscl_request(nd, vp, p, cred, stuff);
3046 if (!(nd->nd_flag & ND_NFSV2)) {
3047 if (nd->nd_flag & ND_NFSV3)
3048 error = nfscl_postop_attr(nd, nap, attrflagp,
3050 if (!nd->nd_repstat && !error) {
3051 NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3053 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3054 dnp->n_cookieverf.nfsuquad[1] = *tl;
3058 if (nd->nd_repstat || error) {
3060 error = nd->nd_repstat;
3063 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3064 more_dirs = fxdr_unsigned(int, *tl);
3068 /* loop through the dir entries, doctoring them to 4bsd form */
3069 while (more_dirs && bigenough) {
3070 if (nd->nd_flag & ND_NFSV4) {
3071 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3072 ncookie.lval[0] = *tl++;
3073 ncookie.lval[1] = *tl++;
3074 len = fxdr_unsigned(int, *tl);
3075 } else if (nd->nd_flag & ND_NFSV3) {
3076 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3077 nfsva.na_fileid = fxdr_hyper(tl);
3079 len = fxdr_unsigned(int, *tl);
3081 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3082 nfsva.na_fileid = fxdr_unsigned(uint64_t,
3084 len = fxdr_unsigned(int, *tl);
3086 if (len <= 0 || len > NFS_MAXNAMLEN) {
3090 tlen = roundup2(len, 8);
3092 tlen += 8; /* To ensure null termination. */
3093 left = DIRBLKSIZ - blksiz;
3094 if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3095 NFSBZERO(uio_iov_base(uiop), left);
3096 dp->d_reclen += left;
3097 uio_iov_base_add(uiop, left);
3098 uio_iov_len_add(uiop, -(left));
3099 uio_uio_resid_add(uiop, -(left));
3100 uiop->uio_offset += left;
3103 if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3104 uio_uio_resid(uiop))
3107 dp = (struct dirent *)uio_iov_base(uiop);
3108 dp->d_pad0 = dp->d_pad1 = 0;
3111 dp->d_reclen = _GENERIC_DIRLEN(len) +
3113 dp->d_type = DT_UNKNOWN;
3114 blksiz += dp->d_reclen;
3115 if (blksiz == DIRBLKSIZ)
3117 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3118 uiop->uio_offset += DIRHDSIZ;
3119 uio_iov_base_add(uiop, DIRHDSIZ);
3120 uio_iov_len_add(uiop, -(DIRHDSIZ));
3121 error = nfsm_mbufuio(nd, uiop, len);
3124 cp = uio_iov_base(uiop);
3127 cp += tlen; /* points to cookie storage */
3128 tl2 = (u_int32_t *)cp;
3129 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3130 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3131 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3132 uiop->uio_offset += (tlen + NFSX_HYPER);
3134 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3138 if (nd->nd_flag & ND_NFSV4) {
3140 nfsva.na_mntonfileno = UINT64_MAX;
3141 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3142 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3143 NULL, NULL, &rderr, p, cred);
3146 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3147 } else if (nd->nd_flag & ND_NFSV3) {
3148 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3149 ncookie.lval[0] = *tl++;
3150 ncookie.lval[1] = *tl++;
3152 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3153 ncookie.lval[0] = 0;
3154 ncookie.lval[1] = *tl++;
3157 if (nd->nd_flag & ND_NFSV4) {
3162 if (nfsva.na_mntonfileno != UINT64_MAX)
3163 dp->d_fileno = nfsva.na_mntonfileno;
3165 dp->d_fileno = nfsva.na_fileid;
3166 } else if (nfsva.na_filesid[0] ==
3167 dnp->n_vattr.na_filesid[0] &&
3168 nfsva.na_filesid[1] ==
3169 dnp->n_vattr.na_filesid[1]) {
3170 dp->d_fileno = nfsva.na_fileid;
3174 } while (fakefileno ==
3176 dp->d_fileno = fakefileno;
3178 dp->d_type = vtonfs_dtype(nfsva.na_type);
3181 dp->d_fileno = nfsva.na_fileid;
3183 *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3185 *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3188 more_dirs = fxdr_unsigned(int, *tl);
3191 * If at end of rpc data, get the eof boolean
3194 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3195 eof = fxdr_unsigned(int, *tl);
3198 if (nd->nd_flag & ND_NFSV4) {
3199 error = nfscl_postop_attr(nd, nap, attrflagp,
3205 mbuf_freem(nd->nd_mrep);
3209 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3210 * by increasing d_reclen for the last record.
3213 left = DIRBLKSIZ - blksiz;
3214 NFSBZERO(uio_iov_base(uiop), left);
3215 dp->d_reclen += left;
3216 uio_iov_base_add(uiop, left);
3217 uio_iov_len_add(uiop, -(left));
3218 uio_uio_resid_add(uiop, -(left));
3219 uiop->uio_offset += left;
3223 * If returning no data, assume end of file.
3224 * If not bigenough, return not end of file, since you aren't
3225 * returning all the data
3226 * Otherwise, return the eof flag from the server.
3229 if (tresid == ((size_t)(uio_uio_resid(uiop))))
3231 else if (!bigenough)
3238 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3240 while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3241 dp = (struct dirent *)uio_iov_base(uiop);
3242 NFSBZERO(dp, DIRBLKSIZ);
3243 dp->d_type = DT_UNKNOWN;
3244 tl = (u_int32_t *)&dp->d_name[4];
3245 *tl++ = cookie.lval[0];
3246 *tl = cookie.lval[1];
3247 dp->d_reclen = DIRBLKSIZ;
3248 uio_iov_base_add(uiop, DIRBLKSIZ);
3249 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3250 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3251 uiop->uio_offset += DIRBLKSIZ;
3255 if (nd->nd_mrep != NULL)
3256 mbuf_freem(nd->nd_mrep);
3262 * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3263 * (Also used for NFS V4 when mount flag set.)
3264 * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3267 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3268 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3269 int *eofp, void *stuff)
3272 struct dirent *dp = NULL;
3274 vnode_t newvp = NULLVP;
3275 struct nfsrv_descript nfsd, *nd = &nfsd;
3276 struct nameidata nami, *ndp = &nami;
3277 struct componentname *cnp = &ndp->ni_cnd;
3278 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3279 struct nfsnode *dnp = VTONFS(vp), *np;
3280 struct nfsvattr nfsva;
3282 nfsquad_t cookie, ncookie;
3283 int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3284 int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3285 int isdotdot = 0, unlocknewvp = 0;
3286 u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3287 u_int64_t fileno = 0;
3289 nfsattrbit_t attrbits, dattrbits;
3291 u_int32_t *tl2 = NULL, rderr;
3292 struct timespec dctime;
3294 KASSERT(uiop->uio_iovcnt == 1 &&
3295 (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3296 ("nfs readdirplusrpc bad uio"));
3297 ncookie.lval[0] = ncookie.lval[1] = 0;
3298 timespecclear(&dctime);
3304 cookie.lval[0] = cookiep->nfsuquad[0];
3305 cookie.lval[1] = cookiep->nfsuquad[1];
3306 tresid = uio_uio_resid(uiop);
3309 * For NFSv4, first create the "." and ".." entries.
3311 if (NFSHASNFSV4(nmp)) {
3312 NFSGETATTR_ATTRBIT(&dattrbits);
3313 NFSZERO_ATTRBIT(&attrbits);
3314 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3315 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3316 NFSATTRBIT_MOUNTEDONFILEID)) {
3317 NFSSETBIT_ATTRBIT(&attrbits,
3318 NFSATTRBIT_MOUNTEDONFILEID);
3322 * Must fake it. Use the fileno, except when the
3323 * fsid is != to that of the directory. For that
3324 * case, generate a fake fileno that is not the same.
3326 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3331 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3333 if (uiop->uio_offset == 0) {
3334 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3335 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3336 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
3337 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3338 (void) nfsrv_putattrbit(nd, &attrbits);
3339 error = nfscl_request(nd, vp, p, cred, stuff);
3342 dotfileid = 0; /* Fake out the compiler. */
3343 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3344 error = nfsm_loadattr(nd, &nfsva);
3347 dctime = nfsva.na_ctime;
3348 dotfileid = nfsva.na_fileid;
3350 if (nd->nd_repstat == 0) {
3351 NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3352 len = fxdr_unsigned(int, *(tl + 4));
3353 if (len > 0 && len <= NFSX_V4FHMAX)
3354 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3358 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3359 nfsva.na_mntonfileno = UINT64_MAX;
3360 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3361 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3362 NULL, NULL, NULL, p, cred);
3364 dotdotfileid = dotfileid;
3365 } else if (gotmnton) {
3366 if (nfsva.na_mntonfileno != UINT64_MAX)
3367 dotdotfileid = nfsva.na_mntonfileno;
3369 dotdotfileid = nfsva.na_fileid;
3370 } else if (nfsva.na_filesid[0] ==
3371 dnp->n_vattr.na_filesid[0] &&
3372 nfsva.na_filesid[1] ==
3373 dnp->n_vattr.na_filesid[1]) {
3374 dotdotfileid = nfsva.na_fileid;
3378 } while (fakefileno ==
3380 dotdotfileid = fakefileno;
3383 } else if (nd->nd_repstat == NFSERR_NOENT) {
3385 * Lookupp returns NFSERR_NOENT when we are
3386 * at the root, so just use the current dir.
3389 dotdotfileid = dotfileid;
3391 error = nd->nd_repstat;
3393 mbuf_freem(nd->nd_mrep);
3397 dp = (struct dirent *)uio_iov_base(uiop);
3398 dp->d_pad0 = dp->d_pad1 = 0;
3400 dp->d_type = DT_DIR;
3401 dp->d_fileno = dotfileid;
3403 *((uint64_t *)dp->d_name) = 0; /* Zero pad it. */
3404 dp->d_name[0] = '.';
3405 dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3407 * Just make these offset cookie 0.
3409 tl = (u_int32_t *)&dp->d_name[8];
3412 blksiz += dp->d_reclen;
3413 uio_uio_resid_add(uiop, -(dp->d_reclen));
3414 uiop->uio_offset += dp->d_reclen;
3415 uio_iov_base_add(uiop, dp->d_reclen);
3416 uio_iov_len_add(uiop, -(dp->d_reclen));
3417 dp = (struct dirent *)uio_iov_base(uiop);
3418 dp->d_pad0 = dp->d_pad1 = 0;
3420 dp->d_type = DT_DIR;
3421 dp->d_fileno = dotdotfileid;
3423 *((uint64_t *)dp->d_name) = 0;
3424 dp->d_name[0] = '.';
3425 dp->d_name[1] = '.';
3426 dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3428 * Just make these offset cookie 0.
3430 tl = (u_int32_t *)&dp->d_name[8];
3433 blksiz += dp->d_reclen;
3434 uio_uio_resid_add(uiop, -(dp->d_reclen));
3435 uiop->uio_offset += dp->d_reclen;
3436 uio_iov_base_add(uiop, dp->d_reclen);
3437 uio_iov_len_add(uiop, -(dp->d_reclen));
3439 NFSREADDIRPLUS_ATTRBIT(&attrbits);
3441 NFSSETBIT_ATTRBIT(&attrbits,
3442 NFSATTRBIT_MOUNTEDONFILEID);
3446 * Loop around doing readdir rpc's of size nm_readdirsize.
3447 * The stopping criteria is EOF or buffer full.
3449 while (more_dirs && bigenough) {
3451 NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3452 NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3453 *tl++ = cookie.lval[0];
3454 *tl++ = cookie.lval[1];
3455 if (cookie.qval == 0) {
3460 *tl++ = dnp->n_cookieverf.nfsuquad[0];
3461 *tl++ = dnp->n_cookieverf.nfsuquad[1];
3464 *tl++ = txdr_unsigned(nmp->nm_readdirsize);
3465 *tl = txdr_unsigned(nmp->nm_readdirsize);
3466 if (nd->nd_flag & ND_NFSV4) {
3467 (void) nfsrv_putattrbit(nd, &attrbits);
3468 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3469 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3470 (void) nfsrv_putattrbit(nd, &dattrbits);
3472 error = nfscl_request(nd, vp, p, cred, stuff);
3475 if (nd->nd_flag & ND_NFSV3)
3476 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3477 if (nd->nd_repstat || error) {
3479 error = nd->nd_repstat;
3482 if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3483 dctime = nap->na_ctime;
3484 NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3486 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3487 dnp->n_cookieverf.nfsuquad[1] = *tl++;
3489 more_dirs = fxdr_unsigned(int, *tl);
3493 /* loop through the dir entries, doctoring them to 4bsd form */
3494 while (more_dirs && bigenough) {
3495 NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3496 if (nd->nd_flag & ND_NFSV4) {
3497 ncookie.lval[0] = *tl++;
3498 ncookie.lval[1] = *tl++;
3500 fileno = fxdr_hyper(tl);
3503 len = fxdr_unsigned(int, *tl);
3504 if (len <= 0 || len > NFS_MAXNAMLEN) {
3508 tlen = roundup2(len, 8);
3510 tlen += 8; /* To ensure null termination. */
3511 left = DIRBLKSIZ - blksiz;
3512 if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3513 NFSBZERO(uio_iov_base(uiop), left);
3514 dp->d_reclen += left;
3515 uio_iov_base_add(uiop, left);
3516 uio_iov_len_add(uiop, -(left));
3517 uio_uio_resid_add(uiop, -(left));
3518 uiop->uio_offset += left;
3521 if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3522 uio_uio_resid(uiop))
3525 dp = (struct dirent *)uio_iov_base(uiop);
3526 dp->d_pad0 = dp->d_pad1 = 0;
3529 dp->d_reclen = _GENERIC_DIRLEN(len) +
3531 dp->d_type = DT_UNKNOWN;
3532 blksiz += dp->d_reclen;
3533 if (blksiz == DIRBLKSIZ)
3535 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3536 uiop->uio_offset += DIRHDSIZ;
3537 uio_iov_base_add(uiop, DIRHDSIZ);
3538 uio_iov_len_add(uiop, -(DIRHDSIZ));
3539 cnp->cn_nameptr = uio_iov_base(uiop);
3540 cnp->cn_namelen = len;
3542 error = nfsm_mbufuio(nd, uiop, len);
3545 cp = uio_iov_base(uiop);
3548 cp += tlen; /* points to cookie storage */
3549 tl2 = (u_int32_t *)cp;
3550 if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3551 cnp->cn_nameptr[1] == '.')
3555 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3556 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3557 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3558 uiop->uio_offset += (tlen + NFSX_HYPER);
3560 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3565 if (nd->nd_flag & ND_NFSV3) {
3566 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3567 ncookie.lval[0] = *tl++;
3568 ncookie.lval[1] = *tl++;
3569 attrflag = fxdr_unsigned(int, *tl);
3571 error = nfsm_loadattr(nd, &nfsva);
3575 NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3577 error = nfsm_getfh(nd, &nfhp);
3581 if (!attrflag && nfhp != NULL) {
3582 free(nfhp, M_NFSFH);
3587 nfsva.na_mntonfileno = 0xffffffff;
3588 error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3589 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3590 NULL, NULL, &rderr, p, cred);
3596 if (nd->nd_flag & ND_NFSV4) {
3599 } else if (gotmnton) {
3600 if (nfsva.na_mntonfileno != 0xffffffff)
3601 dp->d_fileno = nfsva.na_mntonfileno;
3603 dp->d_fileno = nfsva.na_fileid;
3604 } else if (nfsva.na_filesid[0] ==
3605 dnp->n_vattr.na_filesid[0] &&
3606 nfsva.na_filesid[1] ==
3607 dnp->n_vattr.na_filesid[1]) {
3608 dp->d_fileno = nfsva.na_fileid;
3612 } while (fakefileno ==
3614 dp->d_fileno = fakefileno;
3617 dp->d_fileno = fileno;
3619 *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3621 *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3625 if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3626 dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3630 free(nfhp, M_NFSFH);
3632 } else if (isdotdot != 0) {
3634 * Skip doing a nfscl_nget() call for "..".
3635 * There's a race between acquiring the nfs
3636 * node here and lookups that look for the
3637 * directory being read (in the parent).
3638 * It would try to get a lock on ".." here,
3639 * owning the lock on the directory being
3640 * read. Lookup will hold the lock on ".."
3641 * and try to acquire the lock on the
3642 * directory being read.
3643 * If the directory is unlocked/relocked,
3644 * then there is a LOR with the buflock
3647 free(nfhp, M_NFSFH);
3649 error = nfscl_nget(vnode_mount(vp), vp,
3650 nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3657 if (newvp != NULLVP) {
3658 error = nfscl_loadattrcache(&newvp,
3659 &nfsva, NULL, NULL, 0, 0);
3668 vtonfs_dtype(np->n_vattr.na_type);
3670 NFSCNHASH(cnp, HASHINIT);
3671 if (cnp->cn_namelen <= NCHNAMLEN &&
3672 ndp->ni_dvp != ndp->ni_vp &&
3673 (newvp->v_type != VDIR ||
3674 dctime.tv_sec != 0)) {
3675 cache_enter_time(ndp->ni_dvp,
3678 newvp->v_type != VDIR ? NULL :
3688 } else if (nfhp != NULL) {
3689 free(nfhp, M_NFSFH);
3691 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3692 more_dirs = fxdr_unsigned(int, *tl);
3695 * If at end of rpc data, get the eof boolean
3698 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3699 eof = fxdr_unsigned(int, *tl);
3702 if (nd->nd_flag & ND_NFSV4) {
3703 error = nfscl_postop_attr(nd, nap, attrflagp,
3709 mbuf_freem(nd->nd_mrep);
3713 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3714 * by increasing d_reclen for the last record.
3717 left = DIRBLKSIZ - blksiz;
3718 NFSBZERO(uio_iov_base(uiop), left);
3719 dp->d_reclen += left;
3720 uio_iov_base_add(uiop, left);
3721 uio_iov_len_add(uiop, -(left));
3722 uio_uio_resid_add(uiop, -(left));
3723 uiop->uio_offset += left;
3727 * If returning no data, assume end of file.
3728 * If not bigenough, return not end of file, since you aren't
3729 * returning all the data
3730 * Otherwise, return the eof flag from the server.
3733 if (tresid == uio_uio_resid(uiop))
3735 else if (!bigenough)
3742 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3744 while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3745 dp = (struct dirent *)uio_iov_base(uiop);
3746 NFSBZERO(dp, DIRBLKSIZ);
3747 dp->d_type = DT_UNKNOWN;
3748 tl = (u_int32_t *)&dp->d_name[4];
3749 *tl++ = cookie.lval[0];
3750 *tl = cookie.lval[1];
3751 dp->d_reclen = DIRBLKSIZ;
3752 uio_iov_base_add(uiop, DIRBLKSIZ);
3753 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3754 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3755 uiop->uio_offset += DIRBLKSIZ;
3759 if (nd->nd_mrep != NULL)
3760 mbuf_freem(nd->nd_mrep);
3769 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3770 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3773 struct nfsrv_descript nfsd, *nd = &nfsd;
3774 nfsattrbit_t attrbits;
3776 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3779 NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3780 NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3781 txdr_hyper(offset, tl);
3783 *tl = txdr_unsigned(cnt);
3784 if (nd->nd_flag & ND_NFSV4) {
3786 * And do a Getattr op.
3788 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3789 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3790 NFSGETATTR_ATTRBIT(&attrbits);
3791 (void) nfsrv_putattrbit(nd, &attrbits);
3793 error = nfscl_request(nd, vp, p, cred, stuff);
3796 error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3797 if (!error && !nd->nd_repstat) {
3798 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3800 if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3801 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3802 nd->nd_repstat = NFSERR_STALEWRITEVERF;
3805 if (nd->nd_flag & ND_NFSV4)
3806 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3809 if (!error && nd->nd_repstat)
3810 error = nd->nd_repstat;
3811 mbuf_freem(nd->nd_mrep);
3816 * NFS byte range lock rpc.
3817 * (Mostly just calls one of the three lower level RPC routines.)
3820 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3821 int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3823 struct nfscllockowner *lp;
3824 struct nfsclclient *clp;
3826 struct nfsrv_descript nfsd, *nd = &nfsd;
3827 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3830 u_int32_t clidrev = 0;
3831 int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3835 * Convert the flock structure into a start and end and do POSIX
3838 switch (fl->l_whence) {
3842 * Caller is responsible for adding any necessary offset
3843 * when SEEK_CUR is used.
3845 start = fl->l_start;
3849 start = size + fl->l_start;
3850 off = size + fl->l_start;
3857 if (fl->l_len != 0) {
3858 end = start + fl->l_len - 1;
3869 if (op == F_GETLK) {
3870 error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3873 error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3875 clidrev = clp->nfsc_clientidrev;
3876 error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3878 } else if (error == -1) {
3881 nfscl_clientrelease(clp);
3882 } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3884 * We must loop around for all lockowner cases.
3887 error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3891 error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3892 clp, id, flags, &lp, &dorpc);
3894 * If it returns a NULL lp, we're done.
3898 nfscl_clientrelease(clp);
3900 nfscl_releasealllocks(clp, vp, p, id, flags);
3903 if (nmp->nm_clp != NULL)
3904 clidrev = nmp->nm_clp->nfsc_clientidrev;
3908 * If the server doesn't support Posix lock semantics,
3909 * only allow locks on the entire file, since it won't
3910 * handle overlapping byte ranges.
3911 * There might still be a problem when a lock
3912 * upgrade/downgrade (read<->write) occurs, since the
3913 * server "might" expect an unlock first?
3915 if (dorpc && (lp->nfsl_open->nfso_posixlock ||
3916 (off == 0 && len == NFS64BITSSET))) {
3918 * Since the lock records will go away, we must
3919 * wait for grace and delay here.
3922 error = nfsrpc_locku(nd, nmp, lp, off, len,
3923 NFSV4LOCKT_READ, cred, p, 0);
3924 if ((nd->nd_repstat == NFSERR_GRACE ||
3925 nd->nd_repstat == NFSERR_DELAY) &&
3927 (void) nfs_catnap(PZERO, (int)nd->nd_repstat,
3929 } while ((nd->nd_repstat == NFSERR_GRACE ||
3930 nd->nd_repstat == NFSERR_DELAY) && error == 0);
3933 } while (error == 0 && nd->nd_repstat == 0);
3934 nfscl_releasealllocks(clp, vp, p, id, flags);
3935 } else if (op == F_SETLK) {
3936 error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
3937 NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
3938 if (error || donelocally) {
3941 if (nmp->nm_clp != NULL)
3942 clidrev = nmp->nm_clp->nfsc_clientidrev;
3945 nfhp = VTONFS(vp)->n_fhp;
3946 if (!lp->nfsl_open->nfso_posixlock &&
3947 (off != 0 || len != NFS64BITSSET)) {
3950 error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
3951 nfhp->nfh_len, lp, newone, reclaim, off,
3952 len, fl->l_type, cred, p, 0);
3955 error = nd->nd_repstat;
3956 nfscl_lockrelease(lp, error, newone);
3961 error = nd->nd_repstat;
3962 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
3963 error == NFSERR_STALEDONTRECOVER ||
3964 error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3965 error == NFSERR_BADSESSION) {
3966 (void) nfs_catnap(PZERO, error, "nfs_advlock");
3967 } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
3969 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
3972 } while (error == NFSERR_GRACE ||
3973 error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3974 error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
3975 error == NFSERR_BADSESSION ||
3976 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
3977 expireret == 0 && clidrev != 0 && retrycnt < 4));
3978 if (error && retrycnt >= 4)
3984 * The lower level routine for the LockT case.
3987 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
3988 struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
3989 struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3992 int error, type, size;
3993 uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3995 struct nfsmount *nmp;
3996 struct nfsclsession *tsep;
3998 nmp = VFSTONFS(vp->v_mount);
3999 NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4000 NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4001 if (fl->l_type == F_RDLCK)
4002 *tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4004 *tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4005 txdr_hyper(off, tl);
4007 txdr_hyper(len, tl);
4009 tsep = nfsmnt_mdssession(nmp);
4010 *tl++ = tsep->nfsess_clientid.lval[0];
4011 *tl = tsep->nfsess_clientid.lval[1];
4012 nfscl_filllockowner(id, own, flags);
4014 NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4015 np->n_fhp->nfh_len);
4016 (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4017 error = nfscl_request(nd, vp, p, cred, NULL);
4020 if (nd->nd_repstat == 0) {
4021 fl->l_type = F_UNLCK;
4022 } else if (nd->nd_repstat == NFSERR_DENIED) {
4024 fl->l_whence = SEEK_SET;
4025 NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4026 fl->l_start = fxdr_hyper(tl);
4028 len = fxdr_hyper(tl);
4030 if (len == NFS64BITSSET)
4034 type = fxdr_unsigned(int, *tl++);
4035 if (type == NFSV4LOCKT_WRITE)
4036 fl->l_type = F_WRLCK;
4038 fl->l_type = F_RDLCK;
4040 * XXX For now, I have no idea what to do with the
4041 * conflicting lock_owner, so I'll just set the pid == 0
4042 * and skip over the lock_owner.
4044 fl->l_pid = (pid_t)0;
4046 size = fxdr_unsigned(int, *tl);
4047 if (size < 0 || size > NFSV4_OPAQUELIMIT)
4050 error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4051 } else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4052 nfscl_initiate_recovery(clp);
4054 mbuf_freem(nd->nd_mrep);
4059 * Lower level function that performs the LockU RPC.
4062 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4063 struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4064 u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4069 nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4070 lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4071 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4072 *tl++ = txdr_unsigned(type);
4073 *tl = txdr_unsigned(lp->nfsl_seqid);
4074 if (nfstest_outofseq &&
4075 (arc4random() % nfstest_outofseq) == 0)
4076 *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4078 if (NFSHASNFSV4N(nmp))
4081 *tl++ = lp->nfsl_stateid.seqid;
4082 *tl++ = lp->nfsl_stateid.other[0];
4083 *tl++ = lp->nfsl_stateid.other[1];
4084 *tl++ = lp->nfsl_stateid.other[2];
4085 txdr_hyper(off, tl);
4087 txdr_hyper(len, tl);
4089 nd->nd_flag |= ND_USEGSSNAME;
4090 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4091 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4092 NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4095 if (nd->nd_repstat == 0) {
4096 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4097 lp->nfsl_stateid.seqid = *tl++;
4098 lp->nfsl_stateid.other[0] = *tl++;
4099 lp->nfsl_stateid.other[1] = *tl++;
4100 lp->nfsl_stateid.other[2] = *tl;
4101 } else if (nd->nd_repstat == NFSERR_STALESTATEID)
4102 nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4104 mbuf_freem(nd->nd_mrep);
4109 * The actual Lock RPC.
4112 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4113 u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4114 int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4115 NFSPROC_T *p, int syscred)
4119 uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4120 struct nfsclsession *tsep;
4122 nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4123 NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4124 if (type == F_RDLCK)
4125 *tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4127 *tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4128 *tl++ = txdr_unsigned(reclaim);
4129 txdr_hyper(off, tl);
4131 txdr_hyper(len, tl);
4135 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4136 2 * NFSX_UNSIGNED + NFSX_HYPER);
4137 *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4138 if (NFSHASNFSV4N(nmp))
4141 *tl++ = lp->nfsl_open->nfso_stateid.seqid;
4142 *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4143 *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4144 *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4145 *tl++ = txdr_unsigned(lp->nfsl_seqid);
4146 tsep = nfsmnt_mdssession(nmp);
4147 *tl++ = tsep->nfsess_clientid.lval[0];
4148 *tl = tsep->nfsess_clientid.lval[1];
4149 NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4150 NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4151 (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4154 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4155 if (NFSHASNFSV4N(nmp))
4158 *tl++ = lp->nfsl_stateid.seqid;
4159 *tl++ = lp->nfsl_stateid.other[0];
4160 *tl++ = lp->nfsl_stateid.other[1];
4161 *tl++ = lp->nfsl_stateid.other[2];
4162 *tl = txdr_unsigned(lp->nfsl_seqid);
4163 if (nfstest_outofseq &&
4164 (arc4random() % nfstest_outofseq) == 0)
4165 *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4168 nd->nd_flag |= ND_USEGSSNAME;
4169 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4170 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4174 NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4175 NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4176 if (nd->nd_repstat == 0) {
4177 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4178 lp->nfsl_stateid.seqid = *tl++;
4179 lp->nfsl_stateid.other[0] = *tl++;
4180 lp->nfsl_stateid.other[1] = *tl++;
4181 lp->nfsl_stateid.other[2] = *tl;
4182 } else if (nd->nd_repstat == NFSERR_DENIED) {
4183 NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4184 size = fxdr_unsigned(int, *(tl + 7));
4185 if (size < 0 || size > NFSV4_OPAQUELIMIT)
4188 error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4189 } else if (nd->nd_repstat == NFSERR_STALESTATEID)
4190 nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4192 mbuf_freem(nd->nd_mrep);
4198 * (always called with the vp for the mount point)
4201 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4202 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4205 u_int32_t *tl = NULL;
4206 struct nfsrv_descript nfsd, *nd = &nfsd;
4207 struct nfsmount *nmp;
4208 nfsattrbit_t attrbits;
4212 nmp = VFSTONFS(vnode_mount(vp));
4213 if (NFSHASNFSV4(nmp)) {
4215 * For V4, you actually do a getattr.
4217 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4218 NFSSTATFS_GETATTRBIT(&attrbits);
4219 (void) nfsrv_putattrbit(nd, &attrbits);
4220 nd->nd_flag |= ND_USEGSSNAME;
4221 error = nfscl_request(nd, vp, p, cred, stuff);
4224 if (nd->nd_repstat == 0) {
4225 error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4226 NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4229 nmp->nm_fsid[0] = nap->na_filesid[0];
4230 nmp->nm_fsid[1] = nap->na_filesid[1];
4231 NFSSETHASSETFSID(nmp);
4235 error = nd->nd_repstat;
4240 NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4241 error = nfscl_request(nd, vp, p, cred, stuff);
4244 if (nd->nd_flag & ND_NFSV3) {
4245 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4249 if (nd->nd_repstat) {
4250 error = nd->nd_repstat;
4253 NFSM_DISSECT(tl, u_int32_t *,
4254 NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4256 if (NFSHASNFSV3(nmp)) {
4257 sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4258 sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4259 sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4260 sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4261 sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4262 sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4263 sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4264 } else if (NFSHASNFSV4(nmp) == 0) {
4265 sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4266 sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4267 sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4268 sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4269 sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4272 mbuf_freem(nd->nd_mrep);
4280 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4281 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4284 struct nfsrv_descript nfsd, *nd = &nfsd;
4285 struct nfsmount *nmp;
4287 nfsattrbit_t attrbits;
4291 nmp = VFSTONFS(vnode_mount(vp));
4292 if (NFSHASNFSV4(nmp)) {
4294 * For V4, you actually do a getattr.
4296 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4297 NFSPATHCONF_GETATTRBIT(&attrbits);
4298 (void) nfsrv_putattrbit(nd, &attrbits);
4299 nd->nd_flag |= ND_USEGSSNAME;
4300 error = nfscl_request(nd, vp, p, cred, stuff);
4303 if (nd->nd_repstat == 0) {
4304 error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4305 pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4310 error = nd->nd_repstat;
4313 NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4314 error = nfscl_request(nd, vp, p, cred, stuff);
4317 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4318 if (nd->nd_repstat && !error)
4319 error = nd->nd_repstat;
4321 NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4322 pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4323 pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4324 pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4325 pc->pc_chownrestricted =
4326 fxdr_unsigned(u_int32_t, *tl++);
4327 pc->pc_caseinsensitive =
4328 fxdr_unsigned(u_int32_t, *tl++);
4329 pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4333 mbuf_freem(nd->nd_mrep);
4338 * nfs version 3 fsinfo rpc call
4341 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4342 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4345 struct nfsrv_descript nfsd, *nd = &nfsd;
4349 NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4350 error = nfscl_request(nd, vp, p, cred, stuff);
4353 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4354 if (nd->nd_repstat && !error)
4355 error = nd->nd_repstat;
4357 NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4358 fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4359 fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4360 fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4361 fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4362 fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4363 fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4364 fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4365 fsp->fs_maxfilesize = fxdr_hyper(tl);
4367 fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4369 fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4372 mbuf_freem(nd->nd_mrep);
4377 * This function performs the Renew RPC.
4380 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4384 struct nfsrv_descript nfsd;
4385 struct nfsrv_descript *nd = &nfsd;
4386 struct nfsmount *nmp;
4388 struct nfssockreq *nrp;
4389 struct nfsclsession *tsep;
4391 nmp = clp->nfsc_nmp;
4395 nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4398 nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4399 &dsp->nfsclds_sess, 0, 0);
4400 if (!NFSHASNFSV4N(nmp)) {
4401 /* NFSv4.1 just uses a Sequence Op and not a Renew. */
4402 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4403 tsep = nfsmnt_mdssession(nmp);
4404 *tl++ = tsep->nfsess_clientid.lval[0];
4405 *tl = tsep->nfsess_clientid.lval[1];
4409 nrp = dsp->nfsclds_sockp;
4411 /* If NULL, use the MDS socket. */
4412 nrp = &nmp->nm_sockreq;
4413 nd->nd_flag |= ND_USEGSSNAME;
4415 error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4416 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4418 error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4419 NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4421 nfscl_cancelreqs(dsp);
4425 error = nd->nd_repstat;
4426 mbuf_freem(nd->nd_mrep);
4431 * This function performs the Releaselockowner RPC.
4434 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4435 uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4437 struct nfsrv_descript nfsd, *nd = &nfsd;
4440 uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4441 struct nfsclsession *tsep;
4443 if (NFSHASNFSV4N(nmp)) {
4444 /* For NFSv4.1, do a FreeStateID. */
4445 nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4447 nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4449 nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4451 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4452 tsep = nfsmnt_mdssession(nmp);
4453 *tl++ = tsep->nfsess_clientid.lval[0];
4454 *tl = tsep->nfsess_clientid.lval[1];
4455 NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4456 NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4457 (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4459 nd->nd_flag |= ND_USEGSSNAME;
4460 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4461 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4464 error = nd->nd_repstat;
4465 mbuf_freem(nd->nd_mrep);
4470 * This function performs the Compound to get the mount pt FH.
4473 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4477 struct nfsrv_descript nfsd;
4478 struct nfsrv_descript *nd = &nfsd;
4480 int error, cnt, len, setnil;
4483 nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4492 while (*cp2 != '\0' && *cp2 != '/')
4499 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4500 *tl = txdr_unsigned(NFSV4OP_LOOKUP);
4501 nfsm_strtom(nd, cp, strlen(cp));
4507 } while (*cp != '\0');
4508 if (NFSHASNFSV4N(nmp))
4509 /* Has a Sequence Op done by nfscl_reqstart(). */
4510 *opcntp = txdr_unsigned(3 + cnt);
4512 *opcntp = txdr_unsigned(2 + cnt);
4513 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4514 *tl = txdr_unsigned(NFSV4OP_GETFH);
4515 nd->nd_flag |= ND_USEGSSNAME;
4516 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4517 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4520 if (nd->nd_repstat == 0) {
4521 NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4522 tl += (2 + 2 * cnt);
4523 if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4525 nd->nd_repstat = NFSERR_BADXDR;
4527 nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4528 if (nd->nd_repstat == 0)
4529 nmp->nm_fhsize = len;
4532 error = nd->nd_repstat;
4534 mbuf_freem(nd->nd_mrep);
4539 * This function performs the Delegreturn RPC.
4542 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4543 struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4546 struct nfsrv_descript nfsd;
4547 struct nfsrv_descript *nd = &nfsd;
4550 nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4551 dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4552 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4553 if (NFSHASNFSV4N(nmp))
4556 *tl++ = dp->nfsdl_stateid.seqid;
4557 *tl++ = dp->nfsdl_stateid.other[0];
4558 *tl++ = dp->nfsdl_stateid.other[1];
4559 *tl = dp->nfsdl_stateid.other[2];
4561 nd->nd_flag |= ND_USEGSSNAME;
4562 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4563 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4566 error = nd->nd_repstat;
4567 mbuf_freem(nd->nd_mrep);
4575 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4576 struct acl *aclp, void *stuff)
4578 struct nfsrv_descript nfsd, *nd = &nfsd;
4580 nfsattrbit_t attrbits;
4581 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4583 if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4584 return (EOPNOTSUPP);
4585 NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4586 NFSZERO_ATTRBIT(&attrbits);
4587 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4588 (void) nfsrv_putattrbit(nd, &attrbits);
4589 error = nfscl_request(nd, vp, p, cred, stuff);
4592 if (!nd->nd_repstat)
4593 error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4594 NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4596 error = nd->nd_repstat;
4597 mbuf_freem(nd->nd_mrep);
4605 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4606 struct acl *aclp, void *stuff)
4609 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4611 if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4612 return (EOPNOTSUPP);
4613 error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4621 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4622 struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4624 struct nfsrv_descript nfsd, *nd = &nfsd;
4626 nfsattrbit_t attrbits;
4627 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4629 if (!NFSHASNFSV4(nmp))
4630 return (EOPNOTSUPP);
4631 NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4632 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4633 NFSZERO_ATTRBIT(&attrbits);
4634 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4635 (void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4636 &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
4637 error = nfscl_request(nd, vp, p, cred, stuff);
4640 /* Don't care about the pre/postop attributes */
4641 mbuf_freem(nd->nd_mrep);
4642 return (nd->nd_repstat);
4646 * Do the NFSv4.1 Exchange ID.
4649 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4650 struct nfssockreq *nrp, uint32_t exchflags, struct nfsclds **dspp,
4651 struct ucred *cred, NFSPROC_T *p)
4653 uint32_t *tl, v41flags;
4654 struct nfsrv_descript nfsd;
4655 struct nfsrv_descript *nd = &nfsd;
4656 struct nfsclds *dsp;
4657 struct timespec verstime;
4661 nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL, 0, 0);
4662 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4663 *tl++ = txdr_unsigned(nfsboottime.tv_sec); /* Client owner */
4664 *tl = txdr_unsigned(clp->nfsc_rev);
4665 (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4667 NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4668 *tl++ = txdr_unsigned(exchflags);
4669 *tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4671 /* Set the implementation id4 */
4672 *tl = txdr_unsigned(1);
4673 (void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4674 (void) nfsm_strtom(nd, version, strlen(version));
4675 NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4676 verstime.tv_sec = 1293840000; /* Jan 1, 2011 */
4677 verstime.tv_nsec = 0;
4678 txdr_nfsv4time(&verstime, tl);
4679 nd->nd_flag |= ND_USEGSSNAME;
4680 error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4681 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4682 NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4683 (int)nd->nd_repstat);
4686 if (nd->nd_repstat == 0) {
4687 NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4688 len = fxdr_unsigned(int, *(tl + 7));
4689 if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4690 error = NFSERR_BADXDR;
4693 dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4695 dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4696 dsp->nfsclds_servownlen = len;
4697 dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4698 dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4699 dsp->nfsclds_sess.nfsess_sequenceid =
4700 fxdr_unsigned(uint32_t, *tl++);
4701 v41flags = fxdr_unsigned(uint32_t, *tl);
4702 if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4703 NFSHASPNFSOPT(nmp)) {
4704 NFSCL_DEBUG(1, "set PNFS\n");
4706 nmp->nm_state |= NFSSTA_PNFS;
4708 dsp->nfsclds_flags |= NFSCLDS_MDS;
4710 if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4711 dsp->nfsclds_flags |= NFSCLDS_DS;
4713 nd->nd_repstat = nfsrv_mtostr(nd,
4714 dsp->nfsclds_serverown, len);
4715 if (nd->nd_repstat == 0) {
4716 mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4717 mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4719 nfscl_initsessionslots(&dsp->nfsclds_sess);
4722 free(dsp, M_NFSCLDS);
4724 error = nd->nd_repstat;
4726 mbuf_freem(nd->nd_mrep);
4731 * Do the NFSv4.1 Create Session.
4734 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4735 struct nfssockreq *nrp, uint32_t sequenceid, int mds, struct ucred *cred,
4738 uint32_t crflags, maxval, *tl;
4739 struct nfsrv_descript nfsd;
4740 struct nfsrv_descript *nd = &nfsd;
4743 /* Make sure nm_rsize, nm_wsize is set. */
4744 if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
4745 nmp->nm_rsize = NFS_MAXBSIZE;
4746 if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
4747 nmp->nm_wsize = NFS_MAXBSIZE;
4748 nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL, 0,
4750 NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4751 *tl++ = sep->nfsess_clientid.lval[0];
4752 *tl++ = sep->nfsess_clientid.lval[1];
4753 *tl++ = txdr_unsigned(sequenceid);
4754 crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4755 if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4756 crflags |= NFSV4CRSESS_CONNBACKCHAN;
4757 *tl = txdr_unsigned(crflags);
4759 /* Fill in fore channel attributes. */
4760 NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4761 *tl++ = 0; /* Header pad size */
4762 *tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);/* Max request size */
4763 *tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);/* Max reply size */
4764 *tl++ = txdr_unsigned(4096); /* Max response size cached */
4765 *tl++ = txdr_unsigned(20); /* Max operations */
4766 *tl++ = txdr_unsigned(64); /* Max slots */
4767 *tl = 0; /* No rdma ird */
4769 /* Fill in back channel attributes. */
4770 NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4771 *tl++ = 0; /* Header pad size */
4772 *tl++ = txdr_unsigned(10000); /* Max request size */
4773 *tl++ = txdr_unsigned(10000); /* Max response size */
4774 *tl++ = txdr_unsigned(4096); /* Max response size cached */
4775 *tl++ = txdr_unsigned(4); /* Max operations */
4776 *tl++ = txdr_unsigned(NFSV4_CBSLOTS); /* Max slots */
4777 *tl = 0; /* No rdma ird */
4779 NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4780 *tl++ = txdr_unsigned(NFS_CALLBCKPROG); /* Call back prog # */
4782 /* Allow AUTH_SYS callbacks as uid, gid == 0. */
4783 *tl++ = txdr_unsigned(1); /* Auth_sys only */
4784 *tl++ = txdr_unsigned(AUTH_SYS); /* AUTH_SYS type */
4785 *tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4786 *tl++ = 0; /* Null machine name */
4787 *tl++ = 0; /* Uid == 0 */
4788 *tl++ = 0; /* Gid == 0 */
4789 *tl = 0; /* No additional gids */
4790 nd->nd_flag |= ND_USEGSSNAME;
4791 error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4792 NFS_VER4, NULL, 1, NULL, NULL);
4795 if (nd->nd_repstat == 0) {
4796 NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4798 bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4799 tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4800 sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4801 crflags = fxdr_unsigned(uint32_t, *tl);
4802 if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4804 nmp->nm_state |= NFSSTA_SESSPERSIST;
4808 /* Get the fore channel slot count. */
4809 NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4810 tl++; /* Skip the header pad size. */
4812 /* Make sure nm_wsize is small enough. */
4813 maxval = fxdr_unsigned(uint32_t, *tl++);
4814 while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
4815 if (nmp->nm_wsize > 8096)
4821 /* Make sure nm_rsize is small enough. */
4822 maxval = fxdr_unsigned(uint32_t, *tl++);
4823 while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
4824 if (nmp->nm_rsize > 8096)
4830 sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4832 sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4833 NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4834 irdcnt = fxdr_unsigned(int, *tl);
4836 NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4838 /* and the back channel slot count. */
4839 NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4841 sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4842 NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4844 error = nd->nd_repstat;
4846 mbuf_freem(nd->nd_mrep);
4851 * Do the NFSv4.1 Destroy Session.
4854 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4855 struct ucred *cred, NFSPROC_T *p)
4858 struct nfsrv_descript nfsd;
4859 struct nfsrv_descript *nd = &nfsd;
4861 struct nfsclsession *tsep;
4863 nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
4865 NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4866 tsep = nfsmnt_mdssession(nmp);
4867 bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4868 nd->nd_flag |= ND_USEGSSNAME;
4869 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4870 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4873 error = nd->nd_repstat;
4874 mbuf_freem(nd->nd_mrep);
4879 * Do the NFSv4.1 Destroy Client.
4882 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4883 struct ucred *cred, NFSPROC_T *p)
4886 struct nfsrv_descript nfsd;
4887 struct nfsrv_descript *nd = &nfsd;
4889 struct nfsclsession *tsep;
4891 nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
4893 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4894 tsep = nfsmnt_mdssession(nmp);
4895 *tl++ = tsep->nfsess_clientid.lval[0];
4896 *tl = tsep->nfsess_clientid.lval[1];
4897 nd->nd_flag |= ND_USEGSSNAME;
4898 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4899 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4902 error = nd->nd_repstat;
4903 mbuf_freem(nd->nd_mrep);
4908 * Do the NFSv4.1 LayoutGet.
4911 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
4912 uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
4913 int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
4914 struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
4917 struct nfsrv_descript nfsd, *nd = &nfsd;
4920 nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
4922 nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
4923 layouttype, layoutlen, 0);
4924 nd->nd_flag |= ND_USEGSSNAME;
4925 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4926 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4927 NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
4930 if (nd->nd_repstat == 0)
4931 error = nfsrv_parselayoutget(nd, stateidp, retonclosep, flhp);
4932 if (error == 0 && nd->nd_repstat != 0)
4933 error = nd->nd_repstat;
4934 mbuf_freem(nd->nd_mrep);
4939 * Do the NFSv4.1 Get Device Info.
4942 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
4943 uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
4946 uint32_t cnt, *tl, vers, minorvers;
4947 struct nfsrv_descript nfsd;
4948 struct nfsrv_descript *nd = &nfsd;
4949 struct sockaddr_in sin, ssin;
4950 struct sockaddr_in6 sin6, ssin6;
4951 struct nfsclds *dsp = NULL, **dspp, **gotdspp;
4952 struct nfscldevinfo *ndi;
4953 int addrcnt = 0, bitcnt, error, gotvers, i, isudp, j, stripecnt;
4954 uint8_t stripeindex;
4955 sa_family_t af, safilled;
4960 nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
4962 NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
4963 NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
4964 tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4965 *tl++ = txdr_unsigned(layouttype);
4966 *tl++ = txdr_unsigned(100000);
4967 if (notifybitsp != NULL && *notifybitsp != 0) {
4968 *tl = txdr_unsigned(1); /* One word of bits. */
4969 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
4970 *tl = txdr_unsigned(*notifybitsp);
4972 *tl = txdr_unsigned(0);
4973 nd->nd_flag |= ND_USEGSSNAME;
4974 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4975 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4978 if (nd->nd_repstat == 0) {
4979 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4980 if (layouttype != fxdr_unsigned(int, *tl))
4981 printf("EEK! devinfo layout type not same!\n");
4982 if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
4983 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4984 stripecnt = fxdr_unsigned(int, *tl);
4985 NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
4986 if (stripecnt < 1 || stripecnt > 4096) {
4987 printf("pNFS File layout devinfo stripecnt %d:"
4988 " out of range\n", stripecnt);
4989 error = NFSERR_BADXDR;
4992 NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
4994 addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
4995 NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
4996 if (addrcnt < 1 || addrcnt > 128) {
4997 printf("NFS devinfo addrcnt %d: out of range\n",
4999 error = NFSERR_BADXDR;
5004 * Now we know how many stripe indices and addresses, so
5005 * we can allocate the structure the correct size.
5007 i = (stripecnt * sizeof(uint8_t)) /
5008 sizeof(struct nfsclds *) + 1;
5009 NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5010 ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5011 sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5013 NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5015 ndi->nfsdi_refcnt = 0;
5016 ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5017 ndi->nfsdi_stripecnt = stripecnt;
5018 ndi->nfsdi_addrcnt = addrcnt;
5019 /* Fill in the stripe indices. */
5020 for (i = 0; i < stripecnt; i++) {
5021 stripeindex = fxdr_unsigned(uint8_t, *tl++);
5022 NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5023 if (stripeindex >= addrcnt) {
5024 printf("pNFS File Layout devinfo"
5025 " stripeindex %d: too big\n",
5027 error = NFSERR_BADXDR;
5030 nfsfldi_setstripeindex(ndi, i, stripeindex);
5032 } else if (layouttype == NFSLAYOUT_FLEXFILE) {
5033 /* For Flex File, we only get one address list. */
5034 ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5035 M_NFSDEVINFO, M_WAITOK | M_ZERO);
5036 NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5038 ndi->nfsdi_refcnt = 0;
5039 ndi->nfsdi_flags = NFSDI_FLEXFILE;
5040 addrcnt = ndi->nfsdi_addrcnt = 1;
5043 /* Now, dissect the server address(es). */
5044 safilled = AF_UNSPEC;
5045 for (i = 0; i < addrcnt; i++) {
5046 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5047 cnt = fxdr_unsigned(uint32_t, *tl);
5049 printf("NFS devinfo 0 len addrlist\n");
5050 error = NFSERR_BADXDR;
5053 dspp = nfsfldi_addr(ndi, i);
5054 safilled = AF_UNSPEC;
5055 for (j = 0; j < cnt; j++) {
5056 error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5058 if (error != 0 && error != EPERM) {
5059 error = NFSERR_BADXDR;
5062 if (error == 0 && isudp == 0) {
5065 * - Same address family.
5066 * Save the address and dspp, so that
5067 * the connection can be done after
5068 * parsing is complete.
5070 if (safilled == AF_UNSPEC ||
5071 (af == nmp->nm_nam->sa_family &&
5072 safilled != nmp->nm_nam->sa_family)
5085 gotvers = NFS_VER4; /* Always NFSv4 for File Layout. */
5086 /* For Flex File, we will take one of the versions to use. */
5087 if (layouttype == NFSLAYOUT_FLEXFILE) {
5088 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5089 j = fxdr_unsigned(int, *tl);
5090 if (j < 1 || j > NFSDEV_MAXVERS) {
5091 printf("pNFS: too many versions\n");
5092 error = NFSERR_BADXDR;
5096 for (i = 0; i < j; i++) {
5097 NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5098 vers = fxdr_unsigned(uint32_t, *tl++);
5099 minorvers = fxdr_unsigned(uint32_t, *tl++);
5100 if ((vers == NFS_VER4 && minorvers ==
5101 NFSV41_MINORVERSION) || (vers == NFS_VER3 &&
5104 /* We'll take this one. */
5105 ndi->nfsdi_versindex = i;
5106 ndi->nfsdi_vers = vers;
5107 ndi->nfsdi_minorvers = minorvers;
5108 ndi->nfsdi_rsize = fxdr_unsigned(
5110 ndi->nfsdi_wsize = fxdr_unsigned(
5112 if (*tl == newnfs_true)
5117 ~NFSDI_TIGHTCOUPLED;
5121 printf("pNFS: no NFSv3 or NFSv4.1\n");
5122 error = NFSERR_BADXDR;
5127 /* And the notify bits. */
5128 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5129 bitcnt = fxdr_unsigned(int, *tl);
5131 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5132 if (notifybitsp != NULL)
5134 fxdr_unsigned(uint32_t, *tl);
5136 if (safilled != AF_UNSPEC) {
5137 KASSERT(ndi != NULL, ("ndi is NULL"));
5143 * Now we can do a TCP connection for the correct
5144 * NFS version and IP address.
5146 error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5150 KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5154 if (nd->nd_repstat != 0 && error == 0)
5155 error = nd->nd_repstat;
5157 if (error != 0 && ndi != NULL)
5158 nfscl_freedevinfo(ndi);
5159 mbuf_freem(nd->nd_mrep);
5164 * Do the NFSv4.1 LayoutCommit.
5167 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5168 uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5169 int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5172 struct nfsrv_descript nfsd, *nd = &nfsd;
5175 nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5177 NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5179 txdr_hyper(off, tl);
5181 txdr_hyper(len, tl);
5184 *tl++ = newnfs_true;
5186 *tl++ = newnfs_false;
5187 *tl++ = txdr_unsigned(stateidp->seqid);
5188 *tl++ = stateidp->other[0];
5189 *tl++ = stateidp->other[1];
5190 *tl++ = stateidp->other[2];
5191 *tl++ = newnfs_true;
5194 else if (lastbyte >= (off + len))
5195 lastbyte = off + len - 1;
5196 txdr_hyper(lastbyte, tl);
5198 *tl++ = newnfs_false;
5199 *tl++ = txdr_unsigned(layouttype);
5200 /* All supported layouts are 0 length. */
5201 *tl = txdr_unsigned(0);
5202 nd->nd_flag |= ND_USEGSSNAME;
5203 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5204 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5207 error = nd->nd_repstat;
5208 mbuf_freem(nd->nd_mrep);
5213 * Do the NFSv4.1 LayoutReturn.
5216 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5217 int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5218 uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5219 uint32_t stat, uint32_t op, char *devid)
5222 struct nfsrv_descript nfsd, *nd = &nfsd;
5226 nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5228 NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5230 *tl++ = newnfs_true;
5232 *tl++ = newnfs_false;
5233 *tl++ = txdr_unsigned(layouttype);
5234 *tl++ = txdr_unsigned(iomode);
5235 *tl = txdr_unsigned(layoutreturn);
5236 if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5237 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5239 txdr_hyper(offset, tl);
5241 txdr_hyper(len, tl);
5243 NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5244 *tl++ = txdr_unsigned(stateidp->seqid);
5245 *tl++ = stateidp->other[0];
5246 *tl++ = stateidp->other[1];
5247 *tl++ = stateidp->other[2];
5248 if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5249 *tl = txdr_unsigned(0);
5250 else if (layouttype == NFSLAYOUT_FLEXFILE) {
5252 *tl = txdr_unsigned(2 * NFSX_HYPER +
5253 NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5255 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5256 NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5258 *tl++ = txdr_unsigned(1); /* One error. */
5259 tu64 = 0; /* Offset. */
5260 txdr_hyper(tu64, tl); tl += 2;
5261 tu64 = UINT64_MAX; /* Length. */
5262 txdr_hyper(tu64, tl); tl += 2;
5263 NFSBCOPY(stateidp, tl, NFSX_STATEID);
5264 tl += (NFSX_STATEID / NFSX_UNSIGNED);
5265 *tl++ = txdr_unsigned(1); /* One error. */
5266 NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5267 tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5268 *tl++ = txdr_unsigned(stat);
5269 *tl++ = txdr_unsigned(op);
5271 *tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5272 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5276 *tl = 0; /* No stats yet. */
5279 nd->nd_flag |= ND_USEGSSNAME;
5280 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5281 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5284 if (nd->nd_repstat == 0) {
5285 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5287 NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5288 stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5289 stateidp->other[0] = *tl++;
5290 stateidp->other[1] = *tl++;
5291 stateidp->other[2] = *tl;
5294 error = nd->nd_repstat;
5296 mbuf_freem(nd->nd_mrep);
5301 * Acquire a layout and devinfo, if possible. The caller must have acquired
5302 * a reference count on the nfsclclient structure before calling this.
5303 * Return the layout in lypp with a reference count on it, if successful.
5306 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5307 int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5308 struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5310 struct nfscllayout *lyp;
5311 struct nfsclflayout *flp;
5312 struct nfsclflayouthead flh;
5313 int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5314 nfsv4stateid_t stateid;
5315 struct nfsclsession *tsep;
5318 if (NFSHASFLEXFILE(nmp))
5319 layouttype = NFSLAYOUT_FLEXFILE;
5321 layouttype = NFSLAYOUT_NFSV4_1_FILES;
5323 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5324 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5327 lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5328 off, &flp, &recalled);
5330 if (lyp == NULL || flp == NULL) {
5334 tsep = nfsmnt_mdssession(nmp);
5335 layoutlen = tsep->nfsess_maxcache -
5336 (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5339 stateid.other[0] = stateidp->other[0];
5340 stateid.other[1] = stateidp->other[1];
5341 stateid.other[2] = stateidp->other[2];
5342 error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5343 nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5344 (uint64_t)0, layouttype, layoutlen, &stateid,
5345 &retonclose, &flh, cred, p, NULL);
5348 stateid.seqid = lyp->nfsly_stateid.seqid;
5349 stateid.other[0] = lyp->nfsly_stateid.other[0];
5350 stateid.other[1] = lyp->nfsly_stateid.other[1];
5351 stateid.other[2] = lyp->nfsly_stateid.other[2];
5352 error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5353 nfhp->nfh_len, iomode, off, UINT64_MAX,
5354 (uint64_t)0, layouttype, layoutlen, &stateid,
5355 &retonclose, &flh, cred, p, NULL);
5357 error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5358 nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5359 &flh, layouttype, error, NULL, cred, p);
5362 else if (islocked != 0)
5363 nfscl_rellayout(lyp, 1);
5370 * Do a TCP connection plus exchange id and create session.
5371 * If successful, a "struct nfsclds" is linked into the list for the
5372 * mount point and a pointer to it is returned.
5375 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5376 struct sockaddr_in6 *sin6, sa_family_t af, int vers, struct nfsclds **dspp,
5379 struct sockaddr_in *msad, *sad;
5380 struct sockaddr_in6 *msad6, *sad6;
5381 struct nfsclclient *clp;
5382 struct nfssockreq *nrp;
5383 struct nfsclds *dsp, *tdsp;
5385 enum nfsclds_state retv;
5386 uint32_t sequenceid;
5388 KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5389 ("nfsrpc_fillsa: NULL nr_cred"));
5395 if (af == AF_INET) {
5398 * Check to see if we already have a session for this
5399 * address that is usable for a DS.
5400 * Note that the MDS's address is in a different place
5401 * than the sessions already acquired for DS's.
5403 msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5404 tdsp = TAILQ_FIRST(&nmp->nm_sess);
5405 while (tdsp != NULL) {
5406 if (msad != NULL && msad->sin_family == AF_INET &&
5407 sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5408 sin->sin_port == msad->sin_port &&
5409 (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5410 tdsp->nfsclds_sess.nfsess_defunct == 0) {
5413 NFSCL_DEBUG(4, "fnd same addr\n");
5416 tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5417 if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5418 msad = (struct sockaddr_in *)
5419 tdsp->nfsclds_sockp->nr_nam;
5425 /* No IP address match, so look for new/trunked one. */
5426 sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5427 sad->sin_len = sizeof(*sad);
5428 sad->sin_family = AF_INET;
5429 sad->sin_port = sin->sin_port;
5430 sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5431 nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5432 nrp->nr_nam = (struct sockaddr *)sad;
5433 } else if (af == AF_INET6) {
5436 * Check to see if we already have a session for this
5437 * address that is usable for a DS.
5438 * Note that the MDS's address is in a different place
5439 * than the sessions already acquired for DS's.
5441 msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5442 tdsp = TAILQ_FIRST(&nmp->nm_sess);
5443 while (tdsp != NULL) {
5444 if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5445 IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5446 &msad6->sin6_addr) &&
5447 sin6->sin6_port == msad6->sin6_port &&
5448 (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5449 tdsp->nfsclds_sess.nfsess_defunct == 0) {
5454 tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5455 if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5456 msad6 = (struct sockaddr_in6 *)
5457 tdsp->nfsclds_sockp->nr_nam;
5463 /* No IP address match, so look for new/trunked one. */
5464 sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5465 sad6->sin6_len = sizeof(*sad6);
5466 sad6->sin6_family = AF_INET6;
5467 sad6->sin6_port = sin6->sin6_port;
5468 NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5469 sizeof(struct in6_addr));
5470 nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5471 nrp->nr_nam = (struct sockaddr *)sad6;
5475 nrp->nr_sotype = SOCK_STREAM;
5476 mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5477 nrp->nr_prog = NFS_PROG;
5478 nrp->nr_vers = vers;
5481 * Use the credentials that were used for the mount, which are
5482 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5483 * Ref. counting the credentials with crhold() is probably not
5484 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5485 * unmount, but I did it anyhow.
5487 nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5488 error = newnfs_connect(nmp, nrp, NULL, p, 0);
5489 NFSCL_DEBUG(3, "DS connect=%d\n", error);
5492 /* Now, do the exchangeid and create session. */
5494 if (vers == NFS_VER4) {
5495 error = nfsrpc_exchangeid(nmp, clp, nrp,
5496 NFSV4EXCH_USEPNFSDS, &dsp, nrp->nr_cred, p);
5497 NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5499 newnfs_disconnect(nrp);
5501 dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5503 dsp->nfsclds_flags |= NFSCLDS_DS;
5504 dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5505 mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5506 mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5511 dsp->nfsclds_sockp = nrp;
5512 if (vers == NFS_VER4) {
5514 retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5516 NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5517 if (retv == NFSDSP_USETHISSESSION &&
5518 nfscl_dssameconn != 0) {
5520 tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
5524 * If there is already a session for this
5527 (void)newnfs_disconnect(nrp);
5528 nfscl_freenfsclds(dsp);
5532 if (retv == NFSDSP_NOTFOUND)
5534 dsp->nfsclds_sess.nfsess_sequenceid;
5536 error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5537 nrp, sequenceid, 0, nrp->nr_cred, p);
5538 NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5541 NFSFREECRED(nrp->nr_cred);
5542 NFSFREEMUTEX(&nrp->nr_mtx);
5543 free(nrp->nr_nam, M_SONAME);
5544 free(nrp, M_NFSSOCKREQ);
5547 NFSCL_DEBUG(3, "add DS session\n");
5549 * Put it at the end of the list. That way the list
5550 * is ordered by when the entry was added. This matters
5551 * since the one done first is the one that should be
5552 * used for sequencid'ing any subsequent create sessions.
5555 TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5558 } else if (dsp != NULL) {
5559 newnfs_disconnect(nrp);
5560 nfscl_freenfsclds(dsp);
5566 * Do the NFSv4.1 Reclaim Complete.
5569 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5572 struct nfsrv_descript nfsd;
5573 struct nfsrv_descript *nd = &nfsd;
5576 nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5578 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5580 nd->nd_flag |= ND_USEGSSNAME;
5581 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5582 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5585 error = nd->nd_repstat;
5586 mbuf_freem(nd->nd_mrep);
5591 * Initialize the slot tables for a session.
5594 nfscl_initsessionslots(struct nfsclsession *sep)
5598 for (i = 0; i < NFSV4_CBSLOTS; i++) {
5599 if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5600 m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5601 NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5603 for (i = 0; i < 64; i++)
5604 sep->nfsess_slotseq[i] = 0;
5605 sep->nfsess_slots = 0;
5609 * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5612 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5613 uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
5615 struct nfsnode *np = VTONFS(vp);
5616 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5617 struct nfscllayout *layp;
5618 struct nfscldevinfo *dip;
5619 struct nfsclflayout *rflp;
5621 struct nfsclwritedsdorpc *drpc, *tdrpc;
5622 nfsv4stateid_t stateid;
5623 struct ucred *newcred;
5624 uint64_t lastbyte, len, off, oresid, xfer;
5625 int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
5628 void *iovbase = NULL;
5633 if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5634 (np->n_flag & NNOLAYOUT) != 0)
5636 /* Now, get a reference cnt on the clientid for this mount. */
5637 if (nfscl_getref(nmp) == 0)
5640 /* Find an appropriate stateid. */
5641 newcred = NFSNEWCRED(cred);
5642 error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5643 rwaccess, 1, newcred, p, &stateid, &lckp);
5645 NFSFREECRED(newcred);
5649 /* Search for a layout for this file. */
5650 off = uiop->uio_offset;
5651 layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5652 np->n_fhp->nfh_len, off, &rflp, &recalled);
5653 if (layp == NULL || rflp == NULL) {
5654 if (recalled != 0) {
5655 NFSFREECRED(newcred);
5660 nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5663 /* Try and get a Layout, if it is supported. */
5664 if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5665 (np->n_flag & NWRITEOPENED) != 0)
5666 iolaymode = NFSLAYOUTIOMODE_RW;
5668 iolaymode = NFSLAYOUTIOMODE_READ;
5669 error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5670 NULL, &stateid, off, &layp, newcred, p);
5673 np->n_flag |= NNOLAYOUT;
5676 nfscl_lockderef(lckp);
5677 NFSFREECRED(newcred);
5679 nfscl_rellayout(layp, 0);
5686 * Loop around finding a layout that works for the first part of
5687 * this I/O operation, and then call the function that actually
5691 len = (uint64_t)uiop->uio_resid;
5692 while (len > 0 && error == 0 && eof == 0) {
5693 off = uiop->uio_offset;
5694 error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5696 oresid = xfer = (uint64_t)uiop->uio_resid;
5697 if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5698 xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5700 * For Flex File layout with mirrored DSs, select one
5701 * of them at random for reads. For writes and commits,
5705 tdrpc = drpc = NULL;
5708 if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
5709 (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
5710 if (rwaccess == NFSV4OPEN_ACCESSREAD) {
5711 firstmirror = arc4random() % mirrorcnt;
5712 mirrorcnt = firstmirror + 1;
5714 if (docommit == 0) {
5716 * Save values, so uiop can be
5717 * rolled back upon a write
5720 offs = uiop->uio_offset;
5721 resid = uiop->uio_resid;
5723 uiop->uio_iov->iov_base;
5724 iovlen = uiop->uio_iov->iov_len;
5725 m = nfsm_uiombuflist(uiop, len,
5728 tdrpc = drpc = malloc(sizeof(*drpc) *
5729 (mirrorcnt - 1), M_TEMP, M_WAITOK |
5733 for (i = firstmirror; i < mirrorcnt && error == 0; i++){
5734 if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
5735 dev = rflp->nfsfl_ffm[i].dev;
5736 dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5737 rflp->nfsfl_ffm[i].devp);
5739 dev = rflp->nfsfl_dev;
5740 dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5744 if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
5746 error = nfscl_dofflayoutio(vp,
5747 uiop, iomode, must_commit,
5748 &eof, &stateid, rwaccess,
5749 dip, layp, rflp, off, xfer,
5750 i, docommit, m, tdrpc,
5753 error = nfscl_doflayoutio(vp,
5754 uiop, iomode, must_commit,
5755 &eof, &stateid, rwaccess,
5756 dip, layp, rflp, off, xfer,
5757 docommit, newcred, p);
5758 nfscl_reldevinfo(dip);
5766 timo = hz / 50; /* Wait for 20msec. */
5769 for (i = firstmirror; i < mirrorcnt - 1 &&
5770 tdrpc != NULL; i++, tdrpc++) {
5772 * For the unused drpc entries, both inprog and
5773 * err == 0, so this loop won't break.
5775 while (tdrpc->inprog != 0 && tdrpc->done == 0)
5776 tsleep(&tdrpc->tsk, PVFS, "clrpcio",
5778 if (error == 0 && tdrpc->err != 0)
5783 if (mirrorcnt > 1 && rwaccess ==
5784 NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5786 layp->nfsly_flags |= NFSLY_WRITTEN;
5789 lastbyte = off + xfer - 1;
5791 if (lastbyte > layp->nfsly_lastbyte)
5792 layp->nfsly_lastbyte = lastbyte;
5794 } else if (error == NFSERR_OPENMODE &&
5795 rwaccess == NFSV4OPEN_ACCESSREAD) {
5797 nmp->nm_state |= NFSSTA_OPENMODE;
5802 len -= (oresid - (uint64_t)uiop->uio_resid);
5803 else if (mirrorcnt > 1 && rwaccess ==
5804 NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5806 * In case the rpc gets retried, roll the
5807 * uio fields changed by nfsm_uiombuflist()
5810 uiop->uio_offset = offs;
5811 uiop->uio_resid = resid;
5812 uiop->uio_iov->iov_base = iovbase;
5813 uiop->uio_iov->iov_len = iovlen;
5818 nfscl_lockderef(lckp);
5819 NFSFREECRED(newcred);
5820 nfscl_rellayout(layp, 0);
5826 * Make a copy of the mbuf chain and add an mbuf for null padding, as required.
5828 static struct mbuf *
5829 nfsm_copym(struct mbuf *m, int off, int xfer)
5831 struct mbuf *m2, *m3, *m4;
5835 m2 = m_copym(m, off, xfer, M_WAITOK);
5836 rem = NFSM_RNDUP(xfer) - xfer;
5839 * The zero padding to a multiple of 4 bytes is required by
5840 * the XDR. So that the mbufs copied by reference aren't
5841 * modified, add an mbuf with the zero'd bytes to the list.
5842 * rem will be a maximum of 3, so one zero'd uint32_t is
5846 while (m3->m_next != NULL)
5849 tl = NFSMTOD(m4, uint32_t *);
5851 mbuf_setlen(m4, rem);
5852 mbuf_setnext(m3, m4);
5858 * Find a file layout that will handle the first bytes of the requested
5859 * range and return the information from it needed to the I/O operation.
5862 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5863 struct nfsclflayout **retflpp)
5865 struct nfsclflayout *flp, *nflp, *rflp;
5870 /* For reading, do the Read list first and then the Write list. */
5872 if (rw == NFSV4OPEN_ACCESSREAD)
5873 flp = LIST_FIRST(&lyp->nfsly_flayread);
5875 flp = LIST_FIRST(&lyp->nfsly_flayrw);
5876 while (flp != NULL) {
5877 nflp = LIST_NEXT(flp, nfsfl_list);
5878 if (flp->nfsfl_off > off)
5880 if (flp->nfsfl_end > off &&
5881 (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
5885 if (rw == NFSV4OPEN_ACCESSREAD)
5886 rw = NFSV4OPEN_ACCESSWRITE;
5891 /* This one covers the most bytes starting at off. */
5899 * Do I/O using an NFSv4.1 file layout.
5902 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5903 int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5904 struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5905 uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
5907 uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
5908 int commit_thru_mds, error, stripe_index, stripe_pos;
5911 struct nfsclds **dspp;
5914 rel_off = off - flp->nfsfl_patoff;
5915 stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
5916 stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
5917 dp->nfsdi_stripecnt;
5918 transfer = stripe_unit_size - (rel_off % stripe_unit_size);
5921 /* Loop around, doing I/O for each stripe unit. */
5922 while (len > 0 && error == 0) {
5923 stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
5924 dspp = nfsfldi_addr(dp, stripe_index);
5925 if (len > transfer && docommit == 0)
5929 if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
5931 if (stripe_pos >= flp->nfsfl_fhcnt)
5933 fhp = flp->nfsfl_fh[stripe_pos];
5934 io_off = (rel_off / (stripe_unit_size *
5935 dp->nfsdi_stripecnt)) * stripe_unit_size +
5936 rel_off % stripe_unit_size;
5938 /* Sparse layout. */
5939 if (flp->nfsfl_fhcnt > 1) {
5940 if (stripe_index >= flp->nfsfl_fhcnt)
5942 fhp = flp->nfsfl_fh[stripe_index];
5943 } else if (flp->nfsfl_fhcnt == 1)
5944 fhp = flp->nfsfl_fh[0];
5949 if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
5950 commit_thru_mds = 1;
5954 commit_thru_mds = 0;
5956 np->n_flag |= NDSCOMMIT;
5959 if (docommit != 0) {
5961 error = nfsrpc_commitds(vp, io_off, xfer,
5962 *dspp, fhp, 0, 0, cred, p);
5965 * Set both eof and uio_resid = 0 to end any
5969 uiop->uio_resid = 0;
5972 np->n_flag &= ~NDSCOMMIT;
5975 } else if (rwflag == NFSV4OPEN_ACCESSREAD)
5976 error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
5977 io_off, xfer, fhp, 0, 0, 0, cred, p);
5979 error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
5980 stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
5984 lyp->nfsly_flags |= NFSLY_WRITTEN;
5989 transfer = stripe_unit_size;
5990 stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
5999 * Do I/O using an NFSv4.1 flex file layout.
6002 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6003 int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6004 struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6005 uint64_t len, int mirror, int docommit, struct mbuf *mp,
6006 struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6008 uint64_t transfer, xfer;
6012 struct nfsclds **dspp;
6013 struct ucred *tcred;
6019 NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6021 /* Loop around, doing I/O for each stripe unit. */
6022 while (len > 0 && error == 0) {
6023 dspp = nfsfldi_addr(dp, 0);
6024 fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6025 stateidp = &flp->nfsfl_ffm[mirror].st;
6026 NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6027 mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6028 if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6029 tcred = NFSNEWCRED(cred);
6030 tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6031 tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6032 tcred->cr_ngroups = 1;
6035 if (rwflag == NFSV4OPEN_ACCESSREAD)
6036 transfer = dp->nfsdi_rsize;
6038 transfer = dp->nfsdi_wsize;
6040 np->n_flag |= NDSCOMMIT;
6042 if (len > transfer && docommit == 0)
6046 if (docommit != 0) {
6049 * Do last mirrored DS commit with this thread.
6051 if (mirror < flp->nfsfl_mirrorcnt - 1)
6052 error = nfsio_commitds(vp, off, xfer,
6053 *dspp, fhp, dp->nfsdi_vers,
6054 dp->nfsdi_minorvers, drpc, tcred,
6057 error = nfsrpc_commitds(vp, off, xfer,
6058 *dspp, fhp, dp->nfsdi_vers,
6059 dp->nfsdi_minorvers, tcred, p);
6060 NFSCL_DEBUG(4, "commitds=%d\n", error);
6061 if (error != 0 && error != EACCES && error !=
6064 "DS layreterr for commit\n");
6065 nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6069 NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6072 * Set both eof and uio_resid = 0 to end any
6076 uiop->uio_resid = 0;
6079 np->n_flag &= ~NDSCOMMIT;
6082 } else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6083 error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6084 off, xfer, fhp, 1, dp->nfsdi_vers,
6085 dp->nfsdi_minorvers, tcred, p);
6086 NFSCL_DEBUG(4, "readds=%d\n", error);
6087 if (error != 0 && error != EACCES && error != ESTALE) {
6088 NFSCL_DEBUG(4, "DS layreterr for read\n");
6089 nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6093 if (flp->nfsfl_mirrorcnt == 1) {
6094 error = nfsrpc_writeds(vp, uiop, iomode,
6095 must_commit, stateidp, *dspp, off, xfer,
6096 fhp, 0, 1, dp->nfsdi_vers,
6097 dp->nfsdi_minorvers, tcred, p);
6100 lyp->nfsly_flags |= NFSLY_WRITTEN;
6104 m = nfsm_copym(mp, rel_off, xfer);
6105 NFSCL_DEBUG(4, "mcopy reloff=%d xfer=%jd\n",
6106 rel_off, (uintmax_t)xfer);
6108 * Do the writes after the first loop iteration
6109 * and the write for the last mirror via this
6111 * This loop only iterates for small values
6112 * of nfsdi_wsize, which may never occur in
6113 * practice. However, the drpc is completely
6114 * used by the first iteration and, as such,
6115 * cannot be used after that.
6117 if (mirror < flp->nfsfl_mirrorcnt - 1 &&
6119 error = nfsio_writedsmir(vp, iomode,
6120 must_commit, stateidp, *dspp, off,
6121 xfer, fhp, m, dp->nfsdi_vers,
6122 dp->nfsdi_minorvers, drpc, tcred,
6125 error = nfsrpc_writedsmir(vp, iomode,
6126 must_commit, stateidp, *dspp, off,
6127 xfer, fhp, m, dp->nfsdi_vers,
6128 dp->nfsdi_minorvers, tcred, p);
6129 NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6130 if (error != 0 && error != EACCES && error !=
6133 "DS layreterr for write\n");
6134 nfscl_dserr(NFSV4OP_WRITE, error, dp,
6139 NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6145 if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6148 NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6153 * The actual read RPC done to a DS.
6156 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6157 struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6158 int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6161 int attrflag, error, retlen;
6162 struct nfsrv_descript nfsd;
6163 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6164 struct nfsrv_descript *nd = &nfsd;
6165 struct nfssockreq *nrp;
6169 if (vers == 0 || vers == NFS_VER4) {
6170 nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6171 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6173 NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6175 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6177 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6179 nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6180 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6181 NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6182 NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6183 NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6185 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6186 txdr_hyper(io_off, tl);
6187 *(tl + 2) = txdr_unsigned(len);
6188 nrp = dsp->nfsclds_sockp;
6189 NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6191 /* If NULL, use the MDS socket. */
6192 nrp = &nmp->nm_sockreq;
6193 error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6194 NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6195 NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6199 if (vers == NFS_VER3) {
6200 error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6201 NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6205 if (nd->nd_repstat != 0) {
6206 error = nd->nd_repstat;
6209 if (vers == NFS_VER3) {
6210 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6211 *eofp = fxdr_unsigned(int, *(tl + 1));
6213 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6214 *eofp = fxdr_unsigned(int, *tl);
6216 NFSM_STRSIZ(retlen, len);
6217 NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6218 error = nfsm_mbufuio(nd, uiop, retlen);
6220 if (nd->nd_mrep != NULL)
6221 mbuf_freem(nd->nd_mrep);
6226 * The actual write RPC done to a DS.
6229 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6230 nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6231 struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6232 struct ucred *cred, NFSPROC_T *p)
6235 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6236 int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6238 struct nfsrv_descript nfsd;
6239 struct nfsrv_descript *nd = &nfsd;
6240 struct nfssockreq *nrp;
6243 KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6245 if (vers == 0 || vers == NFS_VER4) {
6246 nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6247 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6248 NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6251 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6253 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6254 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6256 nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6257 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6258 NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6259 NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6260 NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6261 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6263 txdr_hyper(io_off, tl);
6265 if (vers == NFS_VER3)
6266 *tl++ = txdr_unsigned(len);
6267 *tl++ = txdr_unsigned(*iomode);
6268 *tl = txdr_unsigned(len);
6269 nfsm_uiombuf(nd, uiop, len);
6270 nrp = dsp->nfsclds_sockp;
6272 /* If NULL, use the MDS socket. */
6273 nrp = &nmp->nm_sockreq;
6274 error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6275 NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6276 NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6280 if (nd->nd_repstat != 0) {
6282 * In case the rpc gets retried, roll
6283 * the uio fileds changed by nfsm_uiombuf()
6286 uiop->uio_offset -= len;
6287 uio_uio_resid_add(uiop, len);
6288 uio_iov_base_add(uiop, -len);
6289 uio_iov_len_add(uiop, len);
6290 error = nd->nd_repstat;
6292 if (vers == NFS_VER3) {
6293 error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6295 NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6299 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6300 rlen = fxdr_unsigned(int, *tl++);
6301 NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6305 } else if (rlen < len) {
6306 backup = len - rlen;
6307 uio_iov_base_add(uiop, -(backup));
6308 uio_iov_len_add(uiop, backup);
6309 uiop->uio_offset -= backup;
6310 uio_uio_resid_add(uiop, backup);
6313 commit = fxdr_unsigned(int, *tl++);
6316 * Return the lowest commitment level
6317 * obtained by any of the RPCs.
6319 if (committed == NFSWRITE_FILESYNC)
6321 else if (committed == NFSWRITE_DATASYNC &&
6322 commit == NFSWRITE_UNSTABLE)
6324 if (commit_thru_mds != 0) {
6326 if (!NFSHASWRITEVERF(nmp)) {
6327 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6328 NFSSETWRITEVERF(nmp);
6329 } else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6331 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6336 if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6337 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6338 dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6339 } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6341 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6347 if (nd->nd_mrep != NULL)
6348 mbuf_freem(nd->nd_mrep);
6349 *iomode = committed;
6350 if (nd->nd_repstat != 0 && error == 0)
6351 error = nd->nd_repstat;
6356 * The actual write RPC done to a DS.
6357 * This variant is called from a separate kernel process for mirrors.
6358 * Any short write is considered an IO error.
6361 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6362 nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6363 struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6364 struct ucred *cred, NFSPROC_T *p)
6367 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6368 int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6369 struct nfsrv_descript nfsd;
6370 struct nfsrv_descript *nd = &nfsd;
6371 struct nfssockreq *nrp;
6375 if (vers == 0 || vers == NFS_VER4) {
6376 nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6377 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6379 NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6381 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6382 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6384 nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6385 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6386 NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6387 NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6388 NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6389 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6391 txdr_hyper(io_off, tl);
6393 if (vers == NFS_VER3)
6394 *tl++ = txdr_unsigned(len);
6395 *tl++ = txdr_unsigned(*iomode);
6396 *tl = txdr_unsigned(len);
6398 /* Put data in mbuf chain. */
6399 nd->nd_mb->m_next = m;
6400 /* Set nd_mb and nd_bpos to end of data. */
6401 while (m->m_next != NULL)
6404 nd->nd_bpos = mtod(m, char *) + m->m_len;
6405 NFSCL_DEBUG(4, "nfsrpc_writedsmir: lastmb len=%d\n", m->m_len);
6407 nrp = dsp->nfsclds_sockp;
6409 /* If NULL, use the MDS socket. */
6410 nrp = &nmp->nm_sockreq;
6411 error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6412 NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6413 NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6417 if (nd->nd_repstat != 0)
6418 error = nd->nd_repstat;
6420 if (vers == NFS_VER3) {
6421 error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6423 NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6428 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6429 rlen = fxdr_unsigned(int, *tl++);
6430 NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6434 NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6438 commit = fxdr_unsigned(int, *tl++);
6441 * Return the lowest commitment level
6442 * obtained by any of the RPCs.
6444 if (committed == NFSWRITE_FILESYNC)
6446 else if (committed == NFSWRITE_DATASYNC &&
6447 commit == NFSWRITE_UNSTABLE)
6450 if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6451 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6452 dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6453 } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6455 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6460 if (nd->nd_mrep != NULL)
6461 mbuf_freem(nd->nd_mrep);
6462 *iomode = committed;
6463 if (nd->nd_repstat != 0 && error == 0)
6464 error = nd->nd_repstat;
6469 * Start up the thread that will execute nfsrpc_writedsmir().
6472 start_writedsmir(void *arg, int pending)
6474 struct nfsclwritedsdorpc *drpc;
6476 drpc = (struct nfsclwritedsdorpc *)arg;
6477 drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6478 &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6479 drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6482 NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6486 * Set up the write DS mirror call for the pNFS I/O thread.
6489 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6490 nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6491 struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6492 struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6499 drpc->iomode = *iomode;
6500 drpc->must_commit = *must_commit;
6501 drpc->stateidp = stateidp;
6508 drpc->minorvers = minorvers;
6513 if (nfs_pnfsiothreads != 0) {
6514 ret = nfs_pnfsio(start_writedsmir, drpc);
6515 NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6518 error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6519 dsp, off, len, fhp, m, vers, minorvers, cred, p);
6520 NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6525 * Free up the nfsclds structure.
6528 nfscl_freenfsclds(struct nfsclds *dsp)
6534 if (dsp->nfsclds_sockp != NULL) {
6535 NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6536 NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6537 free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6538 free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6540 NFSFREEMUTEX(&dsp->nfsclds_mtx);
6541 NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6542 for (i = 0; i < NFSV4_CBSLOTS; i++) {
6543 if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6545 dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6547 free(dsp, M_NFSCLDS);
6550 static enum nfsclds_state
6551 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6552 struct nfsclds **retdspp, uint32_t *sequencep)
6554 struct nfsclds *dsp;
6558 * Search the list of nfsclds structures for one with the same
6562 TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6563 if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6564 dsp->nfsclds_servownlen != 0 &&
6565 !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6566 dsp->nfsclds_servownlen) &&
6567 dsp->nfsclds_sess.nfsess_defunct == 0) {
6568 NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6569 TAILQ_FIRST(&nmp->nm_sess), dsp,
6570 dsp->nfsclds_flags);
6572 /* Get sequenceid# from first entry. */
6574 dsp->nfsclds_sess.nfsess_sequenceid;
6577 /* Server major id matches. */
6578 if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6580 return (NFSDSP_USETHISSESSION);
6586 return (NFSDSP_SEQTHISSESSION);
6587 return (NFSDSP_NOTFOUND);
6591 * NFS commit rpc to a NFSv4.1 DS.
6594 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6595 struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6599 struct nfsrv_descript nfsd, *nd = &nfsd;
6600 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6601 struct nfssockreq *nrp;
6603 int attrflag, error;
6606 if (vers == 0 || vers == NFS_VER4) {
6607 nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6608 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6611 nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6612 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6613 NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
6614 NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
6616 NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6618 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6619 txdr_hyper(offset, tl);
6621 *tl = txdr_unsigned(cnt);
6622 nrp = dsp->nfsclds_sockp;
6624 /* If NULL, use the MDS socket. */
6625 nrp = &nmp->nm_sockreq;
6626 error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6627 NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6628 NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6632 if (nd->nd_repstat == 0) {
6633 if (vers == NFS_VER3) {
6634 error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6636 NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6640 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6642 if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6643 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6644 error = NFSERR_STALEWRITEVERF;
6649 if (error == 0 && nd->nd_repstat != 0)
6650 error = nd->nd_repstat;
6651 mbuf_freem(nd->nd_mrep);
6656 * Start up the thread that will execute nfsrpc_commitds().
6659 start_commitds(void *arg, int pending)
6661 struct nfsclwritedsdorpc *drpc;
6663 drpc = (struct nfsclwritedsdorpc *)arg;
6664 drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6665 drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6668 NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6672 * Set up the commit DS mirror call for the pNFS I/O thread.
6675 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6676 struct nfsfh *fhp, int vers, int minorvers,
6677 struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6689 drpc->minorvers = minorvers;
6694 if (nfs_pnfsiothreads != 0) {
6695 ret = nfs_pnfsio(start_commitds, drpc);
6696 NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6699 error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6700 minorvers, cred, p);
6701 NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6706 * Set up the XDR arguments for the LayoutGet operation.
6709 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
6710 uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
6711 int layoutlen, int usecurstateid)
6715 NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
6717 *tl++ = newnfs_false; /* Don't signal availability. */
6718 *tl++ = txdr_unsigned(layouttype);
6719 *tl++ = txdr_unsigned(iomode);
6720 txdr_hyper(offset, tl);
6722 txdr_hyper(len, tl);
6724 txdr_hyper(minlen, tl);
6726 if (usecurstateid != 0) {
6727 /* Special stateid for Current stateid. */
6728 *tl++ = txdr_unsigned(1);
6733 *tl++ = txdr_unsigned(stateidp->seqid);
6734 NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
6735 *tl++ = stateidp->other[0];
6736 *tl++ = stateidp->other[1];
6737 *tl++ = stateidp->other[2];
6739 *tl = txdr_unsigned(layoutlen);
6743 * Parse the reply for a successful LayoutGet operation.
6746 nfsrv_parselayoutget(struct nfsrv_descript *nd, nfsv4stateid_t *stateidp,
6747 int *retonclosep, struct nfsclflayouthead *flhp)
6750 struct nfsclflayout *flp, *prevflp, *tflp;
6751 int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
6753 uint64_t retlen, off;
6759 NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
6763 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
6768 stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
6769 NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
6770 (int)stateidp->seqid);
6771 stateidp->other[0] = *tl++;
6772 stateidp->other[1] = *tl++;
6773 stateidp->other[2] = *tl++;
6774 cnt = fxdr_unsigned(int, *tl);
6775 NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
6776 if (cnt <= 0 || cnt > 10000) {
6777 /* Don't accept more than 10000 layouts in reply. */
6778 error = NFSERR_BADXDR;
6781 for (i = 0; i < cnt; i++) {
6782 /* Dissect to the layout type. */
6783 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
6785 off = fxdr_hyper(tl); tl += 2;
6786 retlen = fxdr_hyper(tl); tl += 2;
6787 iomode = fxdr_unsigned(int, *tl++);
6788 laytype = fxdr_unsigned(int, *tl);
6789 NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
6790 (uintmax_t)off, (uintmax_t)retlen, iomode);
6791 /* Ignore length of layout body for now. */
6792 if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
6793 /* Parse the File layout up to fhcnt. */
6794 NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
6795 NFSX_HYPER + NFSX_V4DEVICEID);
6796 fhcnt = fxdr_unsigned(int, *(tl + 4 +
6797 NFSX_V4DEVICEID / NFSX_UNSIGNED));
6798 NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
6799 if (fhcnt < 0 || fhcnt > 100) {
6800 /* Don't accept more than 100 file handles. */
6801 error = NFSERR_BADXDR;
6805 flp = malloc(sizeof(*flp) + fhcnt *
6806 sizeof(struct nfsfh *), M_NFSFLAYOUT,
6809 flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
6811 flp->nfsfl_flags = NFSFL_FILE;
6812 flp->nfsfl_fhcnt = 0;
6813 flp->nfsfl_devp = NULL;
6814 flp->nfsfl_off = off;
6815 if (flp->nfsfl_off + retlen < flp->nfsfl_off)
6816 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
6818 flp->nfsfl_end = flp->nfsfl_off + retlen;
6819 flp->nfsfl_iomode = iomode;
6820 if (gotiomode == -1)
6821 gotiomode = flp->nfsfl_iomode;
6822 /* Ignore layout body length for now. */
6823 NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
6824 tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
6825 flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
6826 NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
6827 flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
6828 flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
6829 NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
6830 flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
6831 for (j = 0; j < fhcnt; j++) {
6832 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6833 nfhlen = fxdr_unsigned(int, *tl);
6834 if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
6835 error = NFSERR_BADXDR;
6838 nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
6840 flp->nfsfl_fh[j] = nfhp;
6842 nfhp->nfh_len = nfhlen;
6843 NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
6844 NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
6846 } else if (laytype == NFSLAYOUT_FLEXFILE) {
6847 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
6849 mirrorcnt = fxdr_unsigned(int, *(tl + 2));
6850 NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
6851 if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
6852 error = NFSERR_BADXDR;
6855 flp = malloc(sizeof(*flp) + mirrorcnt *
6856 sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
6857 flp->nfsfl_flags = NFSFL_FLEXFILE;
6858 flp->nfsfl_mirrorcnt = mirrorcnt;
6859 for (j = 0; j < mirrorcnt; j++)
6860 flp->nfsfl_ffm[j].devp = NULL;
6861 flp->nfsfl_off = off;
6862 if (flp->nfsfl_off + retlen < flp->nfsfl_off)
6863 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
6865 flp->nfsfl_end = flp->nfsfl_off + retlen;
6866 flp->nfsfl_iomode = iomode;
6867 if (gotiomode == -1)
6868 gotiomode = flp->nfsfl_iomode;
6869 flp->nfsfl_stripeunit = fxdr_hyper(tl);
6870 NFSCL_DEBUG(4, "stripeunit=%ju\n",
6871 (uintmax_t)flp->nfsfl_stripeunit);
6872 for (j = 0; j < mirrorcnt; j++) {
6873 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6874 k = fxdr_unsigned(int, *tl);
6875 if (k < 1 || k > 128) {
6876 error = NFSERR_BADXDR;
6879 NFSCL_DEBUG(4, "servercnt=%d\n", k);
6880 for (l = 0; l < k; l++) {
6881 NFSM_DISSECT(tl, uint32_t *,
6882 NFSX_V4DEVICEID + NFSX_STATEID +
6885 /* Just use the first server. */
6887 flp->nfsfl_ffm[j].dev,
6889 tl += (NFSX_V4DEVICEID /
6892 flp->nfsfl_ffm[j].st.seqid =
6894 flp->nfsfl_ffm[j].st.other[0] =
6896 flp->nfsfl_ffm[j].st.other[1] =
6898 flp->nfsfl_ffm[j].st.other[2] =
6900 NFSCL_DEBUG(4, "st.seqid=%u "
6901 "st.o0=0x%x st.o1=0x%x "
6903 flp->nfsfl_ffm[j].st.seqid,
6904 flp->nfsfl_ffm[j].st.other[0],
6905 flp->nfsfl_ffm[j].st.other[1],
6906 flp->nfsfl_ffm[j].st.other[2]);
6908 tl += ((NFSX_V4DEVICEID +
6912 fhcnt = fxdr_unsigned(int, *tl);
6913 NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
6915 fhcnt > NFSDEV_MAXVERS) {
6916 error = NFSERR_BADXDR;
6919 for (m = 0; m < fhcnt; m++) {
6920 NFSM_DISSECT(tl, uint32_t *,
6922 nfhlen = fxdr_unsigned(int,
6924 NFSCL_DEBUG(4, "nfhlen=%d\n",
6926 if (nfhlen <= 0 || nfhlen >
6928 error = NFSERR_BADXDR;
6931 NFSM_DISSECT(cp, uint8_t *,
6932 NFSM_RNDUP(nfhlen));
6934 flp->nfsfl_ffm[j].fhcnt
6938 nfhlen - 1, M_NFSFH,
6940 flp->nfsfl_ffm[j].fh[m]
6942 nfhp->nfh_len = nfhlen;
6950 /* Now, get the ffsd_user/ffds_group. */
6951 error = nfsrv_parseug(nd, 0, &user,
6953 NFSCL_DEBUG(4, "after parseu=%d\n",
6956 error = nfsrv_parseug(nd, 1,
6957 &user, &grp, curthread);
6958 NFSCL_DEBUG(4, "aft parseg=%d\n",
6962 NFSCL_DEBUG(4, "user=%d group=%d\n",
6965 flp->nfsfl_ffm[j].user = user;
6966 flp->nfsfl_ffm[j].group = grp;
6968 "usr=%d grp=%d\n", user,
6973 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6974 flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
6975 flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
6976 NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
6977 flp->nfsfl_fflags, flp->nfsfl_statshint);
6979 error = NFSERR_BADXDR;
6982 if (flp->nfsfl_iomode == gotiomode) {
6983 /* Keep the list in increasing offset order. */
6984 tflp = LIST_FIRST(flhp);
6986 while (tflp != NULL &&
6987 tflp->nfsfl_off < flp->nfsfl_off) {
6989 tflp = LIST_NEXT(tflp, nfsfl_list);
6991 if (prevflp == NULL)
6992 LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
6994 LIST_INSERT_AFTER(prevflp, flp,
6996 NFSCL_DEBUG(4, "flp inserted\n");
6998 printf("nfscl_layoutget(): got wrong iomode\n");
6999 nfscl_freeflayout(flp);
7004 NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7005 if (error != 0 && flp != NULL)
7006 nfscl_freeflayout(flp);
7011 * Parse a user/group digit string.
7014 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7018 char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7022 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7023 len = fxdr_unsigned(uint32_t, *tl);
7025 if (len > NFSV4_OPAQUELIMIT) {
7026 error = NFSERR_BADXDR;
7029 NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7032 *gidp = GID_NOGROUP;
7037 if (len > NFSV4_SMALLSTR)
7038 str = malloc(len + 1, M_TEMP, M_WAITOK);
7041 NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7042 NFSBCOPY(cp, str, len);
7044 NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7046 error = nfsv4_strtogid(nd, str, len, gidp, p);
7048 error = nfsv4_strtouid(nd, str, len, uidp, p);
7050 if (len > NFSV4_SMALLSTR)
7052 NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7057 * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7058 * so that it does both an Open and a Layoutget.
7061 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7062 int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7063 struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7064 struct ucred *cred, NFSPROC_T *p)
7066 struct nfscllayout *lyp;
7067 struct nfsclflayout *flp;
7068 struct nfsclflayouthead flh;
7069 int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7070 int layouttype, laystat;
7071 nfsv4stateid_t stateid;
7072 struct nfsclsession *tsep;
7075 if (NFSHASFLEXFILE(nmp))
7076 layouttype = NFSLAYOUT_FLEXFILE;
7078 layouttype = NFSLAYOUT_NFSV4_1_FILES;
7080 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7081 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7084 lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7086 NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7089 else if (flp != NULL)
7093 if ((lyp == NULL || flp == NULL) && recalled == 0) {
7095 tsep = nfsmnt_mdssession(nmp);
7096 layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7102 stateid.seqid = lyp->nfsly_stateid.seqid;
7103 stateid.other[0] = lyp->nfsly_stateid.other[0];
7104 stateid.other[1] = lyp->nfsly_stateid.other[1];
7105 stateid.other[2] = lyp->nfsly_stateid.other[2];
7107 error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7108 newfhp, newfhlen, mode, op, name, namelen,
7109 dpp, &stateid, usecurstateid, layouttype, layoutlen,
7110 &retonclose, &flh, &laystat, cred, p);
7111 NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7113 laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7114 &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7115 &islocked, cred, p);
7117 error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7118 mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7120 nfscl_rellayout(lyp, 1);
7121 else if (islocked == 1)
7122 nfscl_rellayout(lyp, 0);
7127 * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7128 * enabled, only for the CLAIM_NULL case. All other NFSv4 Opens are
7129 * handled by nfsrpc_openrpc().
7130 * For the case where op == NULL, dvp is the directory. When op != NULL, it
7134 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7135 int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7136 struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7137 nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7138 int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7139 int *laystatp, struct ucred *cred, NFSPROC_T *p)
7142 struct nfsrv_descript nfsd, *nd = &nfsd;
7143 struct nfscldeleg *ndp = NULL;
7144 struct nfsvattr nfsva;
7145 struct nfsclsession *tsep;
7146 uint32_t rflags, deleg;
7147 nfsattrbit_t attrbits;
7148 int error, ret, acesize, limitby, iomode;
7152 nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7154 NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7155 *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7156 *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7157 *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7158 tsep = nfsmnt_mdssession(nmp);
7159 *tl++ = tsep->nfsess_clientid.lval[0];
7160 *tl = tsep->nfsess_clientid.lval[1];
7161 nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7162 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7163 *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7164 *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7165 nfsm_strtom(nd, name, namelen);
7166 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7167 *tl = txdr_unsigned(NFSV4OP_GETATTR);
7168 NFSZERO_ATTRBIT(&attrbits);
7169 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7170 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7171 nfsrv_putattrbit(nd, &attrbits);
7172 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7173 *tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7174 if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7175 iomode = NFSLAYOUTIOMODE_RW;
7177 iomode = NFSLAYOUTIOMODE_READ;
7178 nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7179 layouttype, layoutlen, usecurstateid);
7180 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7181 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7184 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7185 if (nd->nd_repstat != 0)
7186 *laystatp = nd->nd_repstat;
7187 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7188 /* ND_NOMOREDATA will be set if the Open operation failed. */
7189 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7191 op->nfso_stateid.seqid = *tl++;
7192 op->nfso_stateid.other[0] = *tl++;
7193 op->nfso_stateid.other[1] = *tl++;
7194 op->nfso_stateid.other[2] = *tl;
7195 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7196 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7199 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7200 deleg = fxdr_unsigned(u_int32_t, *tl);
7201 if (deleg == NFSV4OPEN_DELEGATEREAD ||
7202 deleg == NFSV4OPEN_DELEGATEWRITE) {
7203 if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7204 NFSCLFLAGS_FIRSTDELEG))
7205 op->nfso_own->nfsow_clp->nfsc_flags |=
7206 (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7207 ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7208 M_NFSCLDELEG, M_WAITOK);
7209 LIST_INIT(&ndp->nfsdl_owner);
7210 LIST_INIT(&ndp->nfsdl_lock);
7211 ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7212 ndp->nfsdl_fhlen = newfhlen;
7213 NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7214 newnfs_copyincred(cred, &ndp->nfsdl_cred);
7215 nfscl_lockinit(&ndp->nfsdl_rwlock);
7216 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7218 ndp->nfsdl_stateid.seqid = *tl++;
7219 ndp->nfsdl_stateid.other[0] = *tl++;
7220 ndp->nfsdl_stateid.other[1] = *tl++;
7221 ndp->nfsdl_stateid.other[2] = *tl++;
7222 ret = fxdr_unsigned(int, *tl);
7223 if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7224 ndp->nfsdl_flags = NFSCLDL_WRITE;
7226 * Indicates how much the file can grow.
7228 NFSM_DISSECT(tl, u_int32_t *,
7230 limitby = fxdr_unsigned(int, *tl++);
7232 case NFSV4OPEN_LIMITSIZE:
7233 ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7235 case NFSV4OPEN_LIMITBLOCKS:
7236 ndp->nfsdl_sizelimit =
7237 fxdr_unsigned(u_int64_t, *tl++);
7238 ndp->nfsdl_sizelimit *=
7239 fxdr_unsigned(u_int64_t, *tl);
7242 error = NFSERR_BADXDR;
7246 ndp->nfsdl_flags = NFSCLDL_READ;
7248 ndp->nfsdl_flags |= NFSCLDL_RECALL;
7249 error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7253 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
7254 error = NFSERR_BADXDR;
7257 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7258 nfscl_assumeposixlocks)
7259 op->nfso_posixlock = 1;
7261 op->nfso_posixlock = 0;
7262 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7263 /* If the 2nd element == NFS_OK, the Getattr succeeded. */
7265 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7266 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7267 NULL, NULL, NULL, p, cred);
7271 ndp->nfsdl_change = nfsva.na_filerev;
7272 ndp->nfsdl_modtime = nfsva.na_mtime;
7273 ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7278 * At this point, the Open has succeeded, so set
7279 * nd_repstat = NFS_OK. If the Layoutget failed,
7280 * this function just won't return a layout.
7282 if (nd->nd_repstat == 0) {
7283 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7284 *laystatp = fxdr_unsigned(int, *++tl);
7285 if (*laystatp == 0) {
7286 error = nfsrv_parselayoutget(nd,
7287 stateidp, retonclosep, flhp);
7292 nd->nd_repstat = 0; /* Return 0 for Open. */
7295 if (nd->nd_repstat != 0 && error == 0)
7296 error = nd->nd_repstat;
7298 free(ndp, M_NFSCLDELEG);
7299 mbuf_freem(nd->nd_mrep);
7304 * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7305 * Used only for mounts with pNFS enabled.
7308 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7309 nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7310 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7311 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7312 int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7313 int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7314 struct nfsclflayouthead *flhp, int *laystatp)
7317 int error = 0, deleg, newone, ret, acesize, limitby;
7318 struct nfsrv_descript nfsd, *nd = &nfsd;
7319 struct nfsclopen *op;
7320 struct nfscldeleg *dp = NULL;
7323 struct nfsclsession *tsep;
7324 nfsattrbit_t attrbits;
7325 nfsv4stateid_t stateid;
7326 struct nfsmount *nmp;
7328 nmp = VFSTONFS(dvp->v_mount);
7336 if (namelen > NFS_MAXNAMLEN)
7337 return (ENAMETOOLONG);
7338 NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7340 * For V4, this is actually an Open op.
7342 NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7343 *tl++ = txdr_unsigned(owp->nfsow_seqid);
7344 *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7345 NFSV4OPEN_ACCESSREAD);
7346 *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7347 tsep = nfsmnt_mdssession(nmp);
7348 *tl++ = tsep->nfsess_clientid.lval[0];
7349 *tl = tsep->nfsess_clientid.lval[1];
7350 nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7351 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7352 *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7353 if ((fmode & O_EXCL) != 0) {
7354 if (NFSHASSESSPERSIST(nmp)) {
7355 /* Use GUARDED for persistent sessions. */
7356 *tl = txdr_unsigned(NFSCREATE_GUARDED);
7357 nfscl_fillsattr(nd, vap, dvp, 0, 0);
7359 /* Otherwise, use EXCLUSIVE4_1. */
7360 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7361 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7362 *tl++ = cverf.lval[0];
7363 *tl = cverf.lval[1];
7364 nfscl_fillsattr(nd, vap, dvp, 0, 0);
7367 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7368 nfscl_fillsattr(nd, vap, dvp, 0, 0);
7370 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7371 *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7372 nfsm_strtom(nd, name, namelen);
7373 /* Get the new file's handle and attributes, plus save the FH. */
7374 NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7375 *tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7376 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
7377 *tl = txdr_unsigned(NFSV4OP_GETATTR);
7378 NFSGETATTR_ATTRBIT(&attrbits);
7379 nfsrv_putattrbit(nd, &attrbits);
7380 /* Get the directory's post-op attributes. */
7381 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7382 *tl = txdr_unsigned(NFSV4OP_PUTFH);
7383 nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7384 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7385 *tl = txdr_unsigned(NFSV4OP_GETATTR);
7386 nfsrv_putattrbit(nd, &attrbits);
7387 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7388 *tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7389 *tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7390 nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7391 layouttype, layoutlen, usecurstateid);
7392 error = nfscl_request(nd, dvp, p, cred, dstuff);
7395 NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7397 if (nd->nd_repstat != 0)
7398 *laystatp = nd->nd_repstat;
7399 NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7400 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7401 NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7402 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7404 stateid.seqid = *tl++;
7405 stateid.other[0] = *tl++;
7406 stateid.other[1] = *tl++;
7407 stateid.other[2] = *tl;
7408 nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7409 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7410 deleg = fxdr_unsigned(int, *tl);
7411 if (deleg == NFSV4OPEN_DELEGATEREAD ||
7412 deleg == NFSV4OPEN_DELEGATEWRITE) {
7413 if (!(owp->nfsow_clp->nfsc_flags &
7414 NFSCLFLAGS_FIRSTDELEG))
7415 owp->nfsow_clp->nfsc_flags |=
7416 (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7417 dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7418 M_NFSCLDELEG, M_WAITOK);
7419 LIST_INIT(&dp->nfsdl_owner);
7420 LIST_INIT(&dp->nfsdl_lock);
7421 dp->nfsdl_clp = owp->nfsow_clp;
7422 newnfs_copyincred(cred, &dp->nfsdl_cred);
7423 nfscl_lockinit(&dp->nfsdl_rwlock);
7424 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7426 dp->nfsdl_stateid.seqid = *tl++;
7427 dp->nfsdl_stateid.other[0] = *tl++;
7428 dp->nfsdl_stateid.other[1] = *tl++;
7429 dp->nfsdl_stateid.other[2] = *tl++;
7430 ret = fxdr_unsigned(int, *tl);
7431 if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7432 dp->nfsdl_flags = NFSCLDL_WRITE;
7434 * Indicates how much the file can grow.
7436 NFSM_DISSECT(tl, u_int32_t *,
7438 limitby = fxdr_unsigned(int, *tl++);
7440 case NFSV4OPEN_LIMITSIZE:
7441 dp->nfsdl_sizelimit = fxdr_hyper(tl);
7443 case NFSV4OPEN_LIMITBLOCKS:
7444 dp->nfsdl_sizelimit =
7445 fxdr_unsigned(u_int64_t, *tl++);
7446 dp->nfsdl_sizelimit *=
7447 fxdr_unsigned(u_int64_t, *tl);
7450 error = NFSERR_BADXDR;
7454 dp->nfsdl_flags = NFSCLDL_READ;
7457 dp->nfsdl_flags |= NFSCLDL_RECALL;
7458 error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7462 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
7463 error = NFSERR_BADXDR;
7467 /* Now, we should have the status for the SaveFH. */
7468 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7470 NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7472 * Now, process the GetFH and Getattr for the newly
7473 * created file. nfscl_mtofh() will set
7474 * ND_NOMOREDATA if these weren't successful.
7476 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7477 NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7481 nd->nd_flag |= ND_NOMOREDATA;
7482 /* Now we have the PutFH and Getattr for the directory. */
7483 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7484 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7486 nd->nd_flag |= ND_NOMOREDATA;
7488 NFSM_DISSECT(tl, uint32_t *, 2 *
7491 nd->nd_flag |= ND_NOMOREDATA;
7494 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7495 /* Load the directory attributes. */
7496 error = nfsm_loadattr(nd, dnap);
7497 NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7501 if (dp != NULL && *attrflagp != 0) {
7502 dp->nfsdl_change = nnap->na_filerev;
7503 dp->nfsdl_modtime = nnap->na_mtime;
7504 dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7507 * We can now complete the Open state.
7511 dp->nfsdl_fhlen = nfhp->nfh_len;
7512 NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7516 * Get an Open structure that will be
7517 * attached to the OpenOwner, acquired already.
7519 error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7520 (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7521 cred, p, NULL, &op, &newone, NULL, 0);
7524 op->nfso_stateid = stateid;
7525 newnfs_copyincred(cred, &op->nfso_cred);
7527 nfscl_openrelease(nmp, op, error, newone);
7530 /* Now, handle the RestoreFH and LayoutGet. */
7531 if (nd->nd_repstat == 0) {
7532 NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7533 *laystatp = fxdr_unsigned(int, *(tl + 3));
7534 if (*laystatp == 0) {
7535 error = nfsrv_parselayoutget(nd,
7536 stateidp, retonclosep, flhp);
7540 NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7546 if (nd->nd_repstat != 0 && error == 0)
7547 error = nd->nd_repstat;
7548 if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7549 nfscl_initiate_recovery(owp->nfsow_clp);
7551 NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7555 free(dp, M_NFSCLDELEG);
7556 mbuf_freem(nd->nd_mrep);
7561 * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7564 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7565 nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7566 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7567 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7568 int *dattrflagp, void *dstuff, int *unlockedp)
7570 struct nfscllayout *lyp;
7571 struct nfsclflayouthead flh;
7573 struct nfsclsession *tsep;
7574 struct nfsmount *nmp;
7575 nfsv4stateid_t stateid;
7576 int error, layoutlen, layouttype, retonclose, laystat;
7579 nmp = VFSTONFS(dvp->v_mount);
7580 if (NFSHASFLEXFILE(nmp))
7581 layouttype = NFSLAYOUT_FLEXFILE;
7583 layouttype = NFSLAYOUT_NFSV4_1_FILES;
7585 tsep = nfsmnt_mdssession(nmp);
7586 layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
7587 error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
7588 owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
7589 dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
7591 NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
7596 laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
7597 nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
7598 layouttype, laystat, NULL, cred, p);
7600 laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
7601 retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
7604 nfscl_rellayout(lyp, 0);
7609 * Process the results of a layoutget() operation.
7612 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
7613 int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
7614 struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
7615 int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
7617 struct nfsclflayout *tflp;
7618 struct nfscldevinfo *dip;
7622 if (laystat == NFSERR_UNKNLAYOUTTYPE) {
7624 if (!NFSHASFLEXFILE(nmp)) {
7625 /* Switch to using Flex File Layout. */
7626 nmp->nm_state |= NFSSTA_FLEXFILE;
7627 } else if (layouttype == NFSLAYOUT_FLEXFILE) {
7629 NFSCL_DEBUG(1, "disable PNFS\n");
7630 nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
7635 NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
7636 LIST_FOREACH(tflp, flhp, nfsfl_list) {
7637 if (layouttype == NFSLAYOUT_FLEXFILE)
7638 mirrorcnt = tflp->nfsfl_mirrorcnt;
7641 for (i = 0; i < mirrorcnt; i++) {
7642 laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
7643 NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
7645 if (layouttype == NFSLAYOUT_FLEXFILE)
7646 dev = tflp->nfsfl_ffm[i].dev;
7648 dev = tflp->nfsfl_dev;
7649 laystat = nfsrpc_getdeviceinfo(nmp, dev,
7650 layouttype, notifybit, &dip, cred,
7652 NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
7656 laystat = nfscl_adddevinfo(nmp, dip, i,
7659 printf("nfsrpc_layoutgetresout"
7668 * nfscl_layout() always returns with the nfsly_lock
7669 * set to a refcnt (shared lock).
7670 * Passing in dvp is sufficient, since it is only used to
7671 * get the fsid for the file system.
7673 laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
7674 layouttype, retonclose, flhp, lypp, cred, p);
7675 NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
7677 if (laystat == 0 && islockedp != NULL)