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.
48 #include "opt_inet6.h"
50 #include <fs/nfs/nfsport.h>
51 #include <sys/sysctl.h>
52 #include <sys/taskqueue.h>
54 SYSCTL_DECL(_vfs_nfs);
56 static int nfsignore_eexist = 0;
57 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
58 &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
60 static int nfscl_dssameconn = 0;
61 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
62 &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
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;
81 #endif /* !APPLEKEXT */
83 #define DIRHDSIZ offsetof(struct dirent, d_name)
86 * nfscl_getsameserver() can return one of three values:
87 * NFSDSP_USETHISSESSION - Use this session for the DS.
88 * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
90 * NFSDSP_NOTFOUND - No matching server was found.
93 NFSDSP_USETHISSESSION = 0,
94 NFSDSP_SEQTHISSESSION = 1,
99 * Do a write RPC on a DS data file, using this structure for the arguments,
100 * so that this function can be executed by a separate kernel process.
102 struct nfsclwritedsdorpc {
109 nfsv4stateid_t *stateidp;
122 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
123 struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
124 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
125 nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
126 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
127 struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
129 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
130 nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
131 struct nfsvattr *, struct nfsfh **, int *, int *, void *);
132 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
133 nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
134 NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
135 int *, void *, int *);
136 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
137 struct nfscllockowner *, u_int64_t, u_int64_t,
138 u_int32_t, struct ucred *, NFSPROC_T *, int);
139 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
140 struct acl *, nfsv4stateid_t *, void *);
141 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
142 uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
143 struct ucred *, NFSPROC_T *);
144 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
145 struct sockaddr_in6 *, sa_family_t, int, struct nfsclds **, NFSPROC_T *);
146 static void nfscl_initsessionslots(struct nfsclsession *);
147 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
148 nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
149 struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
151 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
152 nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
153 struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
154 struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
155 static struct mbuf *nfsm_copym(struct mbuf *, int, int);
156 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
157 struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
158 struct ucred *, NFSPROC_T *);
159 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
160 nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
161 struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
162 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
163 struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
164 struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
165 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
166 struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
167 struct ucred *, NFSPROC_T *);
168 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
169 struct nfsclds *, struct nfsclds **, uint32_t *);
170 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
171 struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
173 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
174 struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
175 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
176 uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
177 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
179 static int nfsrv_parselayoutget(struct nfsrv_descript *, nfsv4stateid_t *,
180 int *, struct nfsclflayouthead *);
181 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
182 int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
183 struct nfscldeleg **, struct ucred *, NFSPROC_T *);
184 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
185 nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
186 struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
187 struct nfsfh **, int *, int *, void *, int *);
188 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
189 int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
190 struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
191 struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
192 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
193 nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
194 struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
195 struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
196 int, int, int, int *, struct nfsclflayouthead *, int *);
197 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
198 uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
199 struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
200 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
201 int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
202 struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
204 int nfs_pnfsio(task_fn_t *, void *);
207 * nfs null call from vfs.
210 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
213 struct nfsrv_descript nfsd, *nd = &nfsd;
215 NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
216 error = nfscl_request(nd, vp, p, cred, NULL);
217 if (nd->nd_repstat && !error)
218 error = nd->nd_repstat;
219 mbuf_freem(nd->nd_mrep);
225 * For nfs version 3 and 4, use the access rpc to check accessibility. If file
226 * modes are changed on the server, accesses might still fail later.
229 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
230 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
233 u_int32_t mode, rmode;
236 mode = NFSACCESS_READ;
239 if (vnode_vtype(vp) == VDIR) {
241 mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
244 mode |= NFSACCESS_LOOKUP;
247 mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
249 mode |= NFSACCESS_EXECUTE;
253 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
255 error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
259 * The NFS V3 spec does not clarify whether or not
260 * the returned access bits can be a superset of
261 * the ones requested, so...
263 if (!error && (rmode & mode) != mode)
269 * The actual rpc, separated out for Darwin.
272 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
273 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
277 u_int32_t supported, rmode;
279 struct nfsrv_descript nfsd, *nd = &nfsd;
280 nfsattrbit_t attrbits;
284 NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
285 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
286 *tl = txdr_unsigned(mode);
287 if (nd->nd_flag & ND_NFSV4) {
289 * And do a Getattr op.
291 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
292 *tl = txdr_unsigned(NFSV4OP_GETATTR);
293 NFSGETATTR_ATTRBIT(&attrbits);
294 (void) nfsrv_putattrbit(nd, &attrbits);
296 error = nfscl_request(nd, vp, p, cred, stuff);
299 if (nd->nd_flag & ND_NFSV3) {
300 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
304 if (!nd->nd_repstat) {
305 if (nd->nd_flag & ND_NFSV4) {
306 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
307 supported = fxdr_unsigned(u_int32_t, *tl++);
309 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
311 rmode = fxdr_unsigned(u_int32_t, *tl);
312 if (nd->nd_flag & ND_NFSV4)
313 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
316 * It's not obvious what should be done about
317 * unsupported access modes. For now, be paranoid
318 * and clear the unsupported ones.
323 error = nd->nd_repstat;
325 mbuf_freem(nd->nd_mrep);
333 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
335 struct nfsclopen *op;
336 struct nfscldeleg *dp;
338 struct nfsnode *np = VTONFS(vp);
339 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
340 u_int32_t mode, clidrev;
341 int ret, newone, error, expireret = 0, retrycnt;
344 * For NFSv4, Open Ops are only done on Regular Files.
346 if (vnode_vtype(vp) != VREG)
350 mode |= NFSV4OPEN_ACCESSREAD;
352 mode |= NFSV4OPEN_ACCESSWRITE;
357 { char name[100]; int namel;
358 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
359 bcopy(NFS4NODENAME(np->n_v4), name, namel);
361 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
362 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
363 else printf(" fhl=0\n");
368 error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
369 cred, p, NULL, &op, &newone, &ret, 1);
373 if (nmp->nm_clp != NULL)
374 clidrev = nmp->nm_clp->nfsc_clientidrev;
377 if (ret == NFSCLOPEN_DOOPEN) {
378 if (np->n_v4 != NULL) {
380 * For the first attempt, try and get a layout, if
381 * pNFS is enabled for the mount.
383 if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
384 nfs_numnfscbd == 0 ||
385 (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
386 error = nfsrpc_openrpc(nmp, vp,
388 np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
389 np->n_fhp->nfh_len, mode, op,
390 NFS4NODENAME(np->n_v4),
391 np->n_v4->n4_namelen,
392 &dp, 0, 0x0, cred, p, 0, 0);
394 error = nfsrpc_getopenlayout(nmp, vp,
396 np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
397 np->n_fhp->nfh_len, mode, op,
398 NFS4NODENAME(np->n_v4),
399 np->n_v4->n4_namelen, &dp, cred, p);
402 OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
405 np->n_flag &= ~NDELEGMOD;
407 * Invalidate the attribute cache, so that
408 * attributes that pre-date the issue of a
409 * delegation are not cached, since the
410 * cached attributes will remain valid while
411 * the delegation is held.
413 NFSINVALATTRCACHE(np);
416 (void) nfscl_deleg(nmp->nm_mountp,
417 op->nfso_own->nfsow_clp,
418 nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
423 newnfs_copyincred(cred, &op->nfso_cred);
424 } else if (ret == NFSCLOPEN_SETCRED)
426 * This is a new local open on a delegation. It needs
427 * to have credentials so that an open can be done
428 * against the server during recovery.
430 newnfs_copyincred(cred, &op->nfso_cred);
433 * nfso_opencnt is the count of how many VOP_OPEN()s have
434 * been done on this Open successfully and a VOP_CLOSE()
435 * is expected for each of these.
436 * If error is non-zero, don't increment it, since the Open
437 * hasn't succeeded yet.
441 nfscl_openrelease(nmp, op, error, newone);
442 if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
443 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
444 error == NFSERR_BADSESSION) {
445 (void) nfs_catnap(PZERO, error, "nfs_open");
446 } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
448 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
451 } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
452 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
453 error == NFSERR_BADSESSION ||
454 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
455 expireret == 0 && clidrev != 0 && retrycnt < 4));
456 if (error && retrycnt >= 4)
462 * the actual open rpc
465 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
466 u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
467 u_int8_t *name, int namelen, struct nfscldeleg **dpp,
468 int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
469 int syscred, int recursed)
472 struct nfsrv_descript nfsd, *nd = &nfsd;
473 struct nfscldeleg *dp, *ndp = NULL;
474 struct nfsvattr nfsva;
475 u_int32_t rflags, deleg;
476 nfsattrbit_t attrbits;
477 int error, ret, acesize, limitby;
478 struct nfsclsession *tsep;
482 nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
483 NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
484 *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
485 *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
486 *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
487 tsep = nfsmnt_mdssession(nmp);
488 *tl++ = tsep->nfsess_clientid.lval[0];
489 *tl = tsep->nfsess_clientid.lval[1];
490 (void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
491 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
492 *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
494 *tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
495 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
496 *tl = txdr_unsigned(delegtype);
499 *tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
500 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
501 if (NFSHASNFSV4N(nmp))
504 *tl++ = dp->nfsdl_stateid.seqid;
505 *tl++ = dp->nfsdl_stateid.other[0];
506 *tl++ = dp->nfsdl_stateid.other[1];
507 *tl = dp->nfsdl_stateid.other[2];
509 *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
511 (void) nfsm_strtom(nd, name, namelen);
513 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
514 *tl = txdr_unsigned(NFSV4OP_GETATTR);
515 NFSZERO_ATTRBIT(&attrbits);
516 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
517 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
518 (void) nfsrv_putattrbit(nd, &attrbits);
520 nd->nd_flag |= ND_USEGSSNAME;
521 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
522 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
525 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
526 if (!nd->nd_repstat) {
527 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
529 op->nfso_stateid.seqid = *tl++;
530 op->nfso_stateid.other[0] = *tl++;
531 op->nfso_stateid.other[1] = *tl++;
532 op->nfso_stateid.other[2] = *tl;
533 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
534 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
537 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
538 deleg = fxdr_unsigned(u_int32_t, *tl);
539 if (deleg == NFSV4OPEN_DELEGATEREAD ||
540 deleg == NFSV4OPEN_DELEGATEWRITE) {
541 if (!(op->nfso_own->nfsow_clp->nfsc_flags &
542 NFSCLFLAGS_FIRSTDELEG))
543 op->nfso_own->nfsow_clp->nfsc_flags |=
544 (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
546 sizeof (struct nfscldeleg) + newfhlen,
547 M_NFSCLDELEG, M_WAITOK);
548 LIST_INIT(&ndp->nfsdl_owner);
549 LIST_INIT(&ndp->nfsdl_lock);
550 ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
551 ndp->nfsdl_fhlen = newfhlen;
552 NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
553 newnfs_copyincred(cred, &ndp->nfsdl_cred);
554 nfscl_lockinit(&ndp->nfsdl_rwlock);
555 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
557 ndp->nfsdl_stateid.seqid = *tl++;
558 ndp->nfsdl_stateid.other[0] = *tl++;
559 ndp->nfsdl_stateid.other[1] = *tl++;
560 ndp->nfsdl_stateid.other[2] = *tl++;
561 ret = fxdr_unsigned(int, *tl);
562 if (deleg == NFSV4OPEN_DELEGATEWRITE) {
563 ndp->nfsdl_flags = NFSCLDL_WRITE;
565 * Indicates how much the file can grow.
567 NFSM_DISSECT(tl, u_int32_t *,
569 limitby = fxdr_unsigned(int, *tl++);
571 case NFSV4OPEN_LIMITSIZE:
572 ndp->nfsdl_sizelimit = fxdr_hyper(tl);
574 case NFSV4OPEN_LIMITBLOCKS:
575 ndp->nfsdl_sizelimit =
576 fxdr_unsigned(u_int64_t, *tl++);
577 ndp->nfsdl_sizelimit *=
578 fxdr_unsigned(u_int64_t, *tl);
581 error = NFSERR_BADXDR;
585 ndp->nfsdl_flags = NFSCLDL_READ;
588 ndp->nfsdl_flags |= NFSCLDL_RECALL;
589 error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
593 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
594 error = NFSERR_BADXDR;
597 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
598 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
599 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
600 NULL, NULL, NULL, p, cred);
604 ndp->nfsdl_change = nfsva.na_filerev;
605 ndp->nfsdl_modtime = nfsva.na_mtime;
606 ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
608 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
610 ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
612 if (ret == NFSERR_DELAY)
613 (void) nfs_catnap(PZERO, ret, "nfs_open");
614 } while (ret == NFSERR_DELAY);
617 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
618 nfscl_assumeposixlocks)
619 op->nfso_posixlock = 1;
621 op->nfso_posixlock = 0;
624 * If the server is handing out delegations, but we didn't
625 * get one because an OpenConfirm was required, try the
626 * Open again, to get a delegation. This is a harmless no-op,
627 * from a server's point of view.
629 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
630 (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
631 && !error && dp == NULL && ndp == NULL && !recursed) {
633 ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
634 newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
635 cred, p, syscred, 1);
636 if (ret == NFSERR_DELAY)
637 (void) nfs_catnap(PZERO, ret, "nfs_open2");
638 } while (ret == NFSERR_DELAY);
641 free(ndp, M_NFSCLDELEG);
644 if (ret == NFSERR_STALECLIENTID ||
645 ret == NFSERR_STALEDONTRECOVER ||
646 ret == NFSERR_BADSESSION)
651 if (nd->nd_repstat != 0 && error == 0)
652 error = nd->nd_repstat;
653 if (error == NFSERR_STALECLIENTID)
654 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
658 else if (ndp != NULL)
659 free(ndp, M_NFSCLDELEG);
660 mbuf_freem(nd->nd_mrep);
668 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
669 struct ucred *cred, NFSPROC_T *p)
672 struct nfsrv_descript nfsd, *nd = &nfsd;
675 NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
676 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
677 if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
680 *tl++ = op->nfso_stateid.seqid;
681 *tl++ = op->nfso_stateid.other[0];
682 *tl++ = op->nfso_stateid.other[1];
683 *tl++ = op->nfso_stateid.other[2];
684 *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
685 *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
686 *tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
687 error = nfscl_request(nd, vp, p, cred, NULL);
690 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
691 if (!nd->nd_repstat) {
692 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
693 op->nfso_stateid.seqid = *tl++;
694 op->nfso_stateid.other[0] = *tl++;
695 op->nfso_stateid.other[1] = *tl++;
696 op->nfso_stateid.other[2] = *tl;
698 if (nd->nd_repstat && error == 0)
699 error = nd->nd_repstat;
700 if (error == NFSERR_STALESTATEID)
701 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
703 mbuf_freem(nd->nd_mrep);
708 * V4 Close operation.
711 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
713 struct nfsclclient *clp;
716 if (vnode_vtype(vp) != VREG)
719 error = nfscl_doclose(vp, &clp, p);
721 error = nfscl_getclose(vp, &clp);
725 nfscl_clientrelease(clp);
733 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
735 struct nfsrv_descript nfsd, *nd = &nfsd;
736 struct nfscllockowner *lp, *nlp;
737 struct nfscllock *lop, *nlop;
739 u_int64_t off = 0, len = 0;
740 u_int32_t type = NFSV4LOCKT_READ;
741 int error, do_unlock, trycnt;
743 tcred = newnfs_getcred();
744 newnfs_copycred(&op->nfso_cred, tcred);
746 * (Theoretically this could be done in the same
747 * compound as the close, but having multiple
748 * sequenced Ops in the same compound might be
749 * too scary for some servers.)
751 if (op->nfso_posixlock) {
754 type = NFSV4LOCKT_READ;
758 * Since this function is only called from VOP_INACTIVE(), no
759 * other thread will be manipulating this Open. As such, the
760 * lock lists are not being changed by other threads, so it should
761 * be safe to do this without locking.
763 LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
765 LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
766 if (op->nfso_posixlock == 0) {
767 off = lop->nfslo_first;
768 len = lop->nfslo_end - lop->nfslo_first;
769 if (lop->nfslo_type == F_WRLCK)
770 type = NFSV4LOCKT_WRITE;
772 type = NFSV4LOCKT_READ;
777 error = nfsrpc_locku(nd, nmp, lp, off,
778 len, type, tcred, p, 0);
779 if ((nd->nd_repstat == NFSERR_GRACE ||
780 nd->nd_repstat == NFSERR_DELAY) &&
782 (void) nfs_catnap(PZERO,
785 } while ((nd->nd_repstat == NFSERR_GRACE ||
786 nd->nd_repstat == NFSERR_DELAY) &&
787 error == 0 && trycnt++ < 5);
788 if (op->nfso_posixlock)
791 nfscl_freelock(lop, 0);
794 * Do a ReleaseLockOwner.
795 * The lock owner name nfsl_owner may be used by other opens for
796 * other files but the lock_owner4 name that nfsrpc_rellockown()
797 * puts on the wire has the file handle for this file appended
798 * to it, so it can be done now.
800 (void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
801 lp->nfsl_open->nfso_fhlen, tcred, p);
805 * There could be other Opens for different files on the same
806 * OpenOwner, so locking is required.
809 nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
812 error = nfscl_tryclose(op, tcred, nmp, p);
813 if (error == NFSERR_GRACE)
814 (void) nfs_catnap(PZERO, error, "nfs_close");
815 } while (error == NFSERR_GRACE);
817 nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
819 LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
820 nfscl_freelockowner(lp, 0);
821 nfscl_freeopen(op, 0);
827 * The actual Close RPC.
830 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
831 struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
837 nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
838 op->nfso_fhlen, NULL, NULL, 0, 0);
839 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
840 *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
841 if (NFSHASNFSV4N(nmp))
844 *tl++ = op->nfso_stateid.seqid;
845 *tl++ = op->nfso_stateid.other[0];
846 *tl++ = op->nfso_stateid.other[1];
847 *tl = op->nfso_stateid.other[2];
849 nd->nd_flag |= ND_USEGSSNAME;
850 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
851 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
854 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
855 if (nd->nd_repstat == 0)
856 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
857 error = nd->nd_repstat;
858 if (error == NFSERR_STALESTATEID)
859 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
861 mbuf_freem(nd->nd_mrep);
866 * V4 Open Confirm RPC.
869 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
870 struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
873 struct nfsrv_descript nfsd, *nd = &nfsd;
874 struct nfsmount *nmp;
877 nmp = VFSTONFS(vnode_mount(vp));
878 if (NFSHASNFSV4N(nmp))
879 return (0); /* No confirmation for NFSv4.1. */
880 nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
882 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
883 *tl++ = op->nfso_stateid.seqid;
884 *tl++ = op->nfso_stateid.other[0];
885 *tl++ = op->nfso_stateid.other[1];
886 *tl++ = op->nfso_stateid.other[2];
887 *tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
888 error = nfscl_request(nd, vp, p, cred, NULL);
891 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
892 if (!nd->nd_repstat) {
893 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
894 op->nfso_stateid.seqid = *tl++;
895 op->nfso_stateid.other[0] = *tl++;
896 op->nfso_stateid.other[1] = *tl++;
897 op->nfso_stateid.other[2] = *tl;
899 error = nd->nd_repstat;
900 if (error == NFSERR_STALESTATEID)
901 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
903 mbuf_freem(nd->nd_mrep);
908 * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
909 * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
912 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
913 struct ucred *cred, NFSPROC_T *p)
916 struct nfsrv_descript nfsd;
917 struct nfsrv_descript *nd = &nfsd;
918 nfsattrbit_t attrbits;
919 u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
921 int error, isinet6 = 0, callblen;
924 static u_int32_t rev = 0;
927 struct nfsclsession *tsep;
929 if (nfsboottime.tv_sec == 0)
930 NFSSETBOOTTIME(nfsboottime);
931 clp->nfsc_rev = rev++;
932 if (NFSHASNFSV4N(nmp)) {
934 * Either there was no previous session or the
935 * previous session has failed, so...
936 * do an ExchangeID followed by the CreateSession.
938 error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq,
939 NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp, cred, p);
940 NFSCL_DEBUG(1, "aft exch=%d\n", error);
942 error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
944 dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
948 * The old sessions cannot be safely free'd
949 * here, since they may still be used by
953 if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
954 tsep = NFSMNT_MDSSESSION(nmp);
955 TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
958 * Wake up RPCs waiting for a slot on the
959 * old session. These will then fail with
960 * NFSERR_BADSESSION and be retried with the
961 * new session by nfsv4_setsequence().
962 * Also wakeup() processes waiting for the
966 wakeup(&tsep->nfsess_slots);
967 wakeup(&nmp->nm_sess);
970 nfscl_freenfsclds(dsp);
971 NFSCL_DEBUG(1, "aft createsess=%d\n", error);
972 if (error == 0 && reclaim == 0) {
973 error = nfsrpc_reclaimcomplete(nmp, cred, p);
974 NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
975 if (error == NFSERR_COMPLETEALREADY ||
976 error == NFSERR_NOTSUPP)
977 /* Ignore this error. */
984 * Allocate a single session structure for NFSv4.0, because some of
985 * the fields are used by NFSv4.0 although it doesn't do a session.
987 dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
988 mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
989 mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
991 TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
992 tsep = NFSMNT_MDSSESSION(nmp);
995 nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
996 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
997 *tl++ = txdr_unsigned(nfsboottime.tv_sec);
998 *tl = txdr_unsigned(clp->nfsc_rev);
999 (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1002 * set up the callback address
1004 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1005 *tl = txdr_unsigned(NFS_CALLBCKPROG);
1006 callblen = strlen(nfsv4_callbackaddr);
1008 cp = nfscl_getmyip(nmp, &a6, &isinet6);
1009 if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1010 (callblen > 0 || cp != NULL)) {
1011 port = htons(nfsv4_cbport);
1012 cp2 = (u_int8_t *)&port;
1014 if ((callblen > 0 &&
1015 strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1016 char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1018 (void) nfsm_strtom(nd, "tcp6", 4);
1019 if (callblen == 0) {
1020 ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1023 ip6add = nfsv4_callbackaddr;
1025 snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1026 ip6add, cp2[0], cp2[1]);
1030 (void) nfsm_strtom(nd, "tcp", 3);
1032 snprintf(addr, INET6_ADDRSTRLEN + 9,
1033 "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1034 cp[2], cp[3], cp2[0], cp2[1]);
1036 snprintf(addr, INET6_ADDRSTRLEN + 9,
1037 "%s.%d.%d", nfsv4_callbackaddr,
1040 (void) nfsm_strtom(nd, addr, strlen(addr));
1042 (void) nfsm_strtom(nd, "tcp", 3);
1043 (void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1045 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1046 *tl = txdr_unsigned(clp->nfsc_cbident);
1047 nd->nd_flag |= ND_USEGSSNAME;
1048 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1049 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1052 if (nd->nd_repstat == 0) {
1053 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1054 tsep->nfsess_clientid.lval[0] = *tl++;
1055 tsep->nfsess_clientid.lval[1] = *tl++;
1056 confirm.lval[0] = *tl++;
1057 confirm.lval[1] = *tl;
1058 mbuf_freem(nd->nd_mrep);
1064 nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1066 NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1067 *tl++ = tsep->nfsess_clientid.lval[0];
1068 *tl++ = tsep->nfsess_clientid.lval[1];
1069 *tl++ = confirm.lval[0];
1070 *tl = confirm.lval[1];
1071 nd->nd_flag |= ND_USEGSSNAME;
1072 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1073 cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1076 mbuf_freem(nd->nd_mrep);
1078 if (nd->nd_repstat == 0) {
1079 nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
1080 nmp->nm_fhsize, NULL, NULL, 0, 0);
1081 NFSZERO_ATTRBIT(&attrbits);
1082 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1083 (void) nfsrv_putattrbit(nd, &attrbits);
1084 nd->nd_flag |= ND_USEGSSNAME;
1085 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1086 cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1089 if (nd->nd_repstat == 0) {
1090 error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
1091 NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
1094 clp->nfsc_renew = NFSCL_RENEW(lease);
1095 clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
1096 clp->nfsc_clientidrev++;
1097 if (clp->nfsc_clientidrev == 0)
1098 clp->nfsc_clientidrev++;
1102 error = nd->nd_repstat;
1104 mbuf_freem(nd->nd_mrep);
1112 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1113 struct nfsvattr *nap, void *stuff)
1115 struct nfsrv_descript nfsd, *nd = &nfsd;
1117 nfsattrbit_t attrbits;
1119 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1120 if (nd->nd_flag & ND_NFSV4) {
1121 NFSGETATTR_ATTRBIT(&attrbits);
1122 (void) nfsrv_putattrbit(nd, &attrbits);
1124 error = nfscl_request(nd, vp, p, cred, stuff);
1127 if (!nd->nd_repstat)
1128 error = nfsm_loadattr(nd, nap);
1130 error = nd->nd_repstat;
1131 mbuf_freem(nd->nd_mrep);
1136 * nfs getattr call with non-vnode arguemnts.
1139 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1140 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1143 struct nfsrv_descript nfsd, *nd = &nfsd;
1144 int error, vers = NFS_VER2;
1145 nfsattrbit_t attrbits;
1147 nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1148 if (nd->nd_flag & ND_NFSV4) {
1150 NFSGETATTR_ATTRBIT(&attrbits);
1151 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1152 (void) nfsrv_putattrbit(nd, &attrbits);
1153 } else if (nd->nd_flag & ND_NFSV3) {
1157 nd->nd_flag |= ND_USEGSSNAME;
1158 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1159 NFS_PROG, vers, NULL, 1, xidp, NULL);
1162 if (nd->nd_repstat == 0) {
1163 if ((nd->nd_flag & ND_NFSV4) != 0)
1164 error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1165 NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1168 error = nfsm_loadattr(nd, nap);
1170 error = nd->nd_repstat;
1171 mbuf_freem(nd->nd_mrep);
1176 * Do an nfs setattr operation.
1179 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1180 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1183 int error, expireret = 0, openerr, retrycnt;
1184 u_int32_t clidrev = 0, mode;
1185 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1187 nfsv4stateid_t stateid;
1190 if (nmp->nm_clp != NULL)
1191 clidrev = nmp->nm_clp->nfsc_clientidrev;
1192 if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1193 mode = NFSV4OPEN_ACCESSWRITE;
1195 mode = NFSV4OPEN_ACCESSREAD;
1200 if (NFSHASNFSV4(nmp)) {
1201 nfhp = VTONFS(vp)->n_fhp;
1202 error = nfscl_getstateid(vp, nfhp->nfh_fh,
1203 nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1204 if (error && vnode_vtype(vp) == VREG &&
1205 (mode == NFSV4OPEN_ACCESSWRITE ||
1206 nfstest_openallsetattr)) {
1208 * No Open stateid, so try and open the file
1211 if (mode == NFSV4OPEN_ACCESSWRITE)
1212 openerr = nfsrpc_open(vp, FWRITE, cred,
1215 openerr = nfsrpc_open(vp, FREAD, cred,
1218 (void) nfscl_getstateid(vp,
1219 nfhp->nfh_fh, nfhp->nfh_len,
1220 mode, 0, cred, p, &stateid, &lckp);
1224 error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1225 rnap, attrflagp, stuff);
1227 error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1229 if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1231 nmp->nm_state |= NFSSTA_OPENMODE;
1234 if (error == NFSERR_STALESTATEID)
1235 nfscl_initiate_recovery(nmp->nm_clp);
1237 nfscl_lockderef(lckp);
1239 (void) nfsrpc_close(vp, 0, p);
1240 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1241 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1242 error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1243 (void) nfs_catnap(PZERO, error, "nfs_setattr");
1244 } else if ((error == NFSERR_EXPIRED ||
1245 error == NFSERR_BADSTATEID) && clidrev != 0) {
1246 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1249 } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1250 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1251 error == NFSERR_BADSESSION ||
1252 (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1253 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1254 expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1255 (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1257 if (error && retrycnt >= 4)
1263 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1264 nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1265 struct nfsvattr *rnap, int *attrflagp, void *stuff)
1268 struct nfsrv_descript nfsd, *nd = &nfsd;
1270 nfsattrbit_t attrbits;
1273 NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1274 if (nd->nd_flag & ND_NFSV4)
1275 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1276 vap->va_type = vnode_vtype(vp);
1277 nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1278 if (nd->nd_flag & ND_NFSV3) {
1279 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1281 } else if (nd->nd_flag & ND_NFSV4) {
1282 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1283 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1284 NFSGETATTR_ATTRBIT(&attrbits);
1285 (void) nfsrv_putattrbit(nd, &attrbits);
1287 error = nfscl_request(nd, vp, p, cred, stuff);
1290 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1291 error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1292 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1293 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1294 if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1295 error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1296 mbuf_freem(nd->nd_mrep);
1297 if (nd->nd_repstat && !error)
1298 error = nd->nd_repstat;
1306 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1307 NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1308 struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1311 struct nfsrv_descript nfsd, *nd = &nfsd;
1312 struct nfsmount *nmp;
1315 nfsattrbit_t attrbits;
1316 int error = 0, lookupp = 0;
1320 if (vnode_vtype(dvp) != VDIR)
1322 nmp = VFSTONFS(vnode_mount(dvp));
1323 if (len > NFS_MAXNAMLEN)
1324 return (ENAMETOOLONG);
1325 if (NFSHASNFSV4(nmp) && len == 1 &&
1328 * Just return the current dir's fh.
1331 nfhp = malloc(sizeof (struct nfsfh) +
1332 np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1333 nfhp->nfh_len = np->n_fhp->nfh_len;
1334 NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1338 if (NFSHASNFSV4(nmp) && len == 2 &&
1339 name[0] == '.' && name[1] == '.') {
1341 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1343 NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1344 (void) nfsm_strtom(nd, name, len);
1346 if (nd->nd_flag & ND_NFSV4) {
1347 NFSGETATTR_ATTRBIT(&attrbits);
1348 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1349 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1350 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1351 (void) nfsrv_putattrbit(nd, &attrbits);
1353 error = nfscl_request(nd, dvp, p, cred, stuff);
1356 if (nd->nd_repstat) {
1358 * When an NFSv4 Lookupp returns ENOENT, it means that
1359 * the lookup is at the root of an fs, so return this dir.
1361 if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1363 nfhp = malloc(sizeof (struct nfsfh) +
1364 np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1365 nfhp->nfh_len = np->n_fhp->nfh_len;
1366 NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1368 mbuf_freem(nd->nd_mrep);
1371 if (nd->nd_flag & ND_NFSV3)
1372 error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1373 else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1375 /* Load the directory attributes. */
1376 error = nfsm_loadattr(nd, dnap);
1382 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1383 /* Load the directory attributes. */
1384 error = nfsm_loadattr(nd, dnap);
1388 /* Skip over the Lookup and GetFH operation status values. */
1389 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1391 error = nfsm_getfh(nd, nfhpp);
1395 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1396 if ((nd->nd_flag & ND_NFSV3) && !error)
1397 error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1399 mbuf_freem(nd->nd_mrep);
1400 if (!error && nd->nd_repstat)
1401 error = nd->nd_repstat;
1406 * Do a readlink rpc.
1409 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1410 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1413 struct nfsrv_descript nfsd, *nd = &nfsd;
1414 struct nfsnode *np = VTONFS(vp);
1415 nfsattrbit_t attrbits;
1416 int error, len, cangetattr = 1;
1419 NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1420 if (nd->nd_flag & ND_NFSV4) {
1422 * And do a Getattr op.
1424 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1425 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1426 NFSGETATTR_ATTRBIT(&attrbits);
1427 (void) nfsrv_putattrbit(nd, &attrbits);
1429 error = nfscl_request(nd, vp, p, cred, stuff);
1432 if (nd->nd_flag & ND_NFSV3)
1433 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1434 if (!nd->nd_repstat && !error) {
1435 NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1437 * This seems weird to me, but must have been added to
1438 * FreeBSD for some reason. The only thing I can think of
1439 * is that there was/is some server that replies with
1440 * more link data than it should?
1442 if (len == NFS_MAXPATHLEN) {
1444 if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1450 error = nfsm_mbufuio(nd, uiop, len);
1451 if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1452 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1454 if (nd->nd_repstat && !error)
1455 error = nd->nd_repstat;
1457 mbuf_freem(nd->nd_mrep);
1465 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1466 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1468 int error, expireret = 0, retrycnt;
1469 u_int32_t clidrev = 0;
1470 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1471 struct nfsnode *np = VTONFS(vp);
1472 struct ucred *newcred;
1473 struct nfsfh *nfhp = NULL;
1474 nfsv4stateid_t stateid;
1477 if (nmp->nm_clp != NULL)
1478 clidrev = nmp->nm_clp->nfsc_clientidrev;
1480 if (NFSHASNFSV4(nmp)) {
1482 newcred = NFSNEWCRED(cred);
1487 if (NFSHASNFSV4(nmp))
1488 (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1489 NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1491 error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1493 if (error == NFSERR_OPENMODE) {
1495 nmp->nm_state |= NFSSTA_OPENMODE;
1498 if (error == NFSERR_STALESTATEID)
1499 nfscl_initiate_recovery(nmp->nm_clp);
1501 nfscl_lockderef(lckp);
1502 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1503 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1504 error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1505 (void) nfs_catnap(PZERO, error, "nfs_read");
1506 } else if ((error == NFSERR_EXPIRED ||
1507 error == NFSERR_BADSTATEID) && clidrev != 0) {
1508 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1511 } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1512 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1513 error == NFSERR_BADSESSION ||
1514 (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1515 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1516 expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1517 (error == NFSERR_OPENMODE && retrycnt < 4));
1518 if (error && retrycnt >= 4)
1520 if (NFSHASNFSV4(nmp))
1521 NFSFREECRED(newcred);
1526 * The actual read RPC.
1529 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1530 nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1531 int *attrflagp, void *stuff)
1534 int error = 0, len, retlen, tsiz, eof = 0;
1535 struct nfsrv_descript nfsd;
1536 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1537 struct nfsrv_descript *nd = &nfsd;
1542 tsiz = uio_uio_resid(uiop);
1543 tmp_off = uiop->uio_offset + tsiz;
1545 if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1549 rsize = nmp->nm_rsize;
1554 len = (tsiz > rsize) ? rsize : tsiz;
1555 NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1556 if (nd->nd_flag & ND_NFSV4)
1557 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1558 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1559 if (nd->nd_flag & ND_NFSV2) {
1560 *tl++ = txdr_unsigned(uiop->uio_offset);
1561 *tl++ = txdr_unsigned(len);
1564 txdr_hyper(uiop->uio_offset, tl);
1565 *(tl + 2) = txdr_unsigned(len);
1568 * Since I can't do a Getattr for NFSv4 for Write, there
1569 * doesn't seem any point in doing one here, either.
1570 * (See the comment in nfsrpc_writerpc() for more info.)
1572 error = nfscl_request(nd, vp, p, cred, stuff);
1575 if (nd->nd_flag & ND_NFSV3) {
1576 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1577 } else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1578 error = nfsm_loadattr(nd, nap);
1582 if (nd->nd_repstat || error) {
1584 error = nd->nd_repstat;
1587 if (nd->nd_flag & ND_NFSV3) {
1588 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1589 eof = fxdr_unsigned(int, *(tl + 1));
1590 } else if (nd->nd_flag & ND_NFSV4) {
1591 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1592 eof = fxdr_unsigned(int, *tl);
1594 NFSM_STRSIZ(retlen, len);
1595 error = nfsm_mbufuio(nd, uiop, retlen);
1598 mbuf_freem(nd->nd_mrep);
1601 if (!(nd->nd_flag & ND_NFSV2)) {
1602 if (eof || retlen == 0)
1604 } else if (retlen < len)
1609 if (nd->nd_mrep != NULL)
1610 mbuf_freem(nd->nd_mrep);
1615 * nfs write operation
1616 * When called_from_strategy != 0, it should return EIO for an error that
1617 * indicates recovery is in progress, so that the buffer will be left
1618 * dirty and be written back to the server later. If it loops around,
1619 * the recovery thread could get stuck waiting for the buffer and recovery
1620 * will then deadlock.
1623 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1624 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1625 void *stuff, int called_from_strategy)
1627 int error, expireret = 0, retrycnt, nostateid;
1628 u_int32_t clidrev = 0;
1629 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1630 struct nfsnode *np = VTONFS(vp);
1631 struct ucred *newcred;
1632 struct nfsfh *nfhp = NULL;
1633 nfsv4stateid_t stateid;
1637 if (nmp->nm_clp != NULL)
1638 clidrev = nmp->nm_clp->nfsc_clientidrev;
1640 if (NFSHASNFSV4(nmp)) {
1641 newcred = NFSNEWCRED(cred);
1648 if (NFSHASNFSV4(nmp)) {
1649 (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1650 NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1652 if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1653 stateid.other[2] == 0) {
1655 NFSCL_DEBUG(1, "stateid0 in write\n");
1660 * If there is no stateid for NFSv4, it means this is an
1661 * extraneous write after close. Basically a poorly
1662 * implemented buffer cache. Just don't do the write.
1667 error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1668 newcred, &stateid, p, nap, attrflagp, stuff);
1669 if (error == NFSERR_STALESTATEID)
1670 nfscl_initiate_recovery(nmp->nm_clp);
1672 nfscl_lockderef(lckp);
1673 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1674 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1675 error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1676 (void) nfs_catnap(PZERO, error, "nfs_write");
1677 } else if ((error == NFSERR_EXPIRED ||
1678 error == NFSERR_BADSTATEID) && clidrev != 0) {
1679 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1682 } while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1683 ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1684 error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1685 (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1686 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1687 expireret == 0 && clidrev != 0 && retrycnt < 4));
1688 if (error != 0 && (retrycnt >= 4 ||
1689 ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1690 error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1692 if (NFSHASNFSV4(nmp))
1693 NFSFREECRED(newcred);
1698 * The actual write RPC.
1701 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1702 int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1703 NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1706 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1707 struct nfsnode *np = VTONFS(vp);
1708 int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1709 int wccflag = 0, wsize;
1711 struct nfsrv_descript nfsd;
1712 struct nfsrv_descript *nd = &nfsd;
1713 nfsattrbit_t attrbits;
1716 KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1718 tsiz = uio_uio_resid(uiop);
1719 tmp_off = uiop->uio_offset + tsiz;
1721 if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1725 wsize = nmp->nm_wsize;
1727 nd->nd_mrep = NULL; /* NFSv2 sometimes does a write with */
1728 nd->nd_repstat = 0; /* uio_resid == 0, so the while is not done */
1731 len = (tsiz > wsize) ? wsize : tsiz;
1732 NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1733 if (nd->nd_flag & ND_NFSV4) {
1734 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1735 NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1736 txdr_hyper(uiop->uio_offset, tl);
1738 *tl++ = txdr_unsigned(*iomode);
1739 *tl = txdr_unsigned(len);
1740 } else if (nd->nd_flag & ND_NFSV3) {
1741 NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1742 txdr_hyper(uiop->uio_offset, tl);
1744 *tl++ = txdr_unsigned(len);
1745 *tl++ = txdr_unsigned(*iomode);
1746 *tl = txdr_unsigned(len);
1750 NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1752 * Not sure why someone changed this, since the
1753 * RFC clearly states that "beginoffset" and
1754 * "totalcount" are ignored, but it wouldn't
1755 * surprise me if there's a busted server out there.
1757 /* Set both "begin" and "current" to non-garbage. */
1758 x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1759 *tl++ = x; /* "begin offset" */
1760 *tl++ = x; /* "current offset" */
1761 x = txdr_unsigned(len);
1762 *tl++ = x; /* total to this offset */
1763 *tl = x; /* size of this write */
1766 nfsm_uiombuf(nd, uiop, len);
1768 * Although it is tempting to do a normal Getattr Op in the
1769 * NFSv4 compound, the result can be a nearly hung client
1770 * system if the Getattr asks for Owner and/or OwnerGroup.
1771 * It occurs when the client can't map either the Owner or
1772 * Owner_group name in the Getattr reply to a uid/gid. When
1773 * there is a cache miss, the kernel does an upcall to the
1774 * nfsuserd. Then, it can try and read the local /etc/passwd
1775 * or /etc/group file. It can then block in getnewbuf(),
1776 * waiting for dirty writes to be pushed to the NFS server.
1777 * The only reason this doesn't result in a complete
1778 * deadlock, is that the upcall times out and allows
1779 * the write to complete. However, progress is so slow
1780 * that it might just as well be deadlocked.
1781 * As such, we get the rest of the attributes, but not
1782 * Owner or Owner_group.
1783 * nb: nfscl_loadattrcache() needs to be told that these
1784 * partial attributes from a write rpc are being
1785 * passed in, via a argument flag.
1787 if (nd->nd_flag & ND_NFSV4) {
1788 NFSWRITEGETATTR_ATTRBIT(&attrbits);
1789 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1790 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1791 (void) nfsrv_putattrbit(nd, &attrbits);
1793 error = nfscl_request(nd, vp, p, cred, stuff);
1796 if (nd->nd_repstat) {
1798 * In case the rpc gets retried, roll
1799 * the uio fileds changed by nfsm_uiombuf()
1802 uiop->uio_offset -= len;
1803 uio_uio_resid_add(uiop, len);
1804 uio_iov_base_add(uiop, -len);
1805 uio_iov_len_add(uiop, len);
1807 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1808 error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1813 if (!nd->nd_repstat) {
1814 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1815 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1817 rlen = fxdr_unsigned(int, *tl++);
1821 } else if (rlen < len) {
1822 backup = len - rlen;
1823 uio_iov_base_add(uiop, -(backup));
1824 uio_iov_len_add(uiop, backup);
1825 uiop->uio_offset -= backup;
1826 uio_uio_resid_add(uiop, backup);
1829 commit = fxdr_unsigned(int, *tl++);
1832 * Return the lowest commitment level
1833 * obtained by any of the RPCs.
1835 if (committed == NFSWRITE_FILESYNC)
1837 else if (committed == NFSWRITE_DATASYNC &&
1838 commit == NFSWRITE_UNSTABLE)
1841 if (!NFSHASWRITEVERF(nmp)) {
1842 NFSBCOPY((caddr_t)tl,
1843 (caddr_t)&nmp->nm_verf[0],
1845 NFSSETWRITEVERF(nmp);
1846 } else if (NFSBCMP(tl, nmp->nm_verf,
1849 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1853 if (nd->nd_flag & ND_NFSV4)
1854 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1855 if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1856 error = nfsm_loadattr(nd, nap);
1858 *attrflagp = NFS_LATTR_NOSHRINK;
1861 error = nd->nd_repstat;
1865 NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1866 mbuf_freem(nd->nd_mrep);
1871 if (nd->nd_mrep != NULL)
1872 mbuf_freem(nd->nd_mrep);
1873 *iomode = committed;
1874 if (nd->nd_repstat && !error)
1875 error = nd->nd_repstat;
1881 * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1882 * mode set to specify the file type and the size field for rdev.
1885 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1886 u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1887 struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1888 int *attrflagp, int *dattrflagp, void *dstuff)
1892 struct nfsrv_descript nfsd, *nd = &nfsd;
1893 nfsattrbit_t attrbits;
1898 if (namelen > NFS_MAXNAMLEN)
1899 return (ENAMETOOLONG);
1900 NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1901 if (nd->nd_flag & ND_NFSV4) {
1902 if (vtyp == VBLK || vtyp == VCHR) {
1903 NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1904 *tl++ = vtonfsv34_type(vtyp);
1905 *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1906 *tl = txdr_unsigned(NFSMINOR(rdev));
1908 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1909 *tl = vtonfsv34_type(vtyp);
1912 (void) nfsm_strtom(nd, name, namelen);
1913 if (nd->nd_flag & ND_NFSV3) {
1914 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1915 *tl = vtonfsv34_type(vtyp);
1917 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1918 nfscl_fillsattr(nd, vap, dvp, 0, 0);
1919 if ((nd->nd_flag & ND_NFSV3) &&
1920 (vtyp == VCHR || vtyp == VBLK)) {
1921 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1922 *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1923 *tl = txdr_unsigned(NFSMINOR(rdev));
1925 if (nd->nd_flag & ND_NFSV4) {
1926 NFSGETATTR_ATTRBIT(&attrbits);
1927 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1928 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1929 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1930 (void) nfsrv_putattrbit(nd, &attrbits);
1932 if (nd->nd_flag & ND_NFSV2)
1933 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
1934 error = nfscl_request(nd, dvp, p, cred, dstuff);
1937 if (nd->nd_flag & ND_NFSV4)
1938 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1939 if (!nd->nd_repstat) {
1940 if (nd->nd_flag & ND_NFSV4) {
1941 NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1942 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1946 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1950 if (nd->nd_flag & ND_NFSV3)
1951 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1952 if (!error && nd->nd_repstat)
1953 error = nd->nd_repstat;
1955 mbuf_freem(nd->nd_mrep);
1960 * nfs file create call
1961 * Mostly just call the approriate routine. (I separated out v4, so that
1962 * error recovery wouldn't be as difficult.)
1965 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1966 nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1967 struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1968 int *attrflagp, int *dattrflagp, void *dstuff)
1970 int error = 0, newone, expireret = 0, retrycnt, unlocked;
1971 struct nfsclowner *owp;
1972 struct nfscldeleg *dp;
1973 struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
1976 if (NFSHASNFSV4(nmp)) {
1980 error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
1981 NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
1985 if (nmp->nm_clp != NULL)
1986 clidrev = nmp->nm_clp->nfsc_clientidrev;
1989 if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
1990 nfs_numnfscbd == 0 || retrycnt > 0)
1991 error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
1992 fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
1993 attrflagp, dattrflagp, dstuff, &unlocked);
1995 error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
1996 cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
1997 attrflagp, dattrflagp, dstuff, &unlocked);
1999 * There is no need to invalidate cached attributes here,
2000 * since new post-delegation issue attributes are always
2001 * returned by nfsrpc_createv4() and these will update the
2005 (void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2006 (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2007 nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2008 if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2009 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2010 error == NFSERR_BADSESSION) {
2011 (void) nfs_catnap(PZERO, error, "nfs_open");
2012 } else if ((error == NFSERR_EXPIRED ||
2013 error == NFSERR_BADSTATEID) && clidrev != 0) {
2014 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2017 } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2018 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2019 error == NFSERR_BADSESSION ||
2020 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2021 expireret == 0 && clidrev != 0 && retrycnt < 4));
2022 if (error && retrycnt >= 4)
2025 error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2026 fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2033 * The create rpc for v2 and 3.
2036 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2037 nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2038 struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2039 int *attrflagp, int *dattrflagp, void *dstuff)
2043 struct nfsrv_descript nfsd, *nd = &nfsd;
2048 if (namelen > NFS_MAXNAMLEN)
2049 return (ENAMETOOLONG);
2050 NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2051 (void) nfsm_strtom(nd, name, namelen);
2052 if (nd->nd_flag & ND_NFSV3) {
2053 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2054 if (fmode & O_EXCL) {
2055 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2056 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2057 *tl++ = cverf.lval[0];
2058 *tl = cverf.lval[1];
2060 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2061 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2064 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2066 error = nfscl_request(nd, dvp, p, cred, dstuff);
2069 if (nd->nd_repstat == 0) {
2070 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2074 if (nd->nd_flag & ND_NFSV3)
2075 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2076 if (nd->nd_repstat != 0 && error == 0)
2077 error = nd->nd_repstat;
2079 mbuf_freem(nd->nd_mrep);
2084 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2085 nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2086 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2087 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2088 int *dattrflagp, void *dstuff, int *unlockedp)
2091 int error = 0, deleg, newone, ret, acesize, limitby;
2092 struct nfsrv_descript nfsd, *nd = &nfsd;
2093 struct nfsclopen *op;
2094 struct nfscldeleg *dp = NULL;
2097 nfsattrbit_t attrbits;
2098 nfsv4stateid_t stateid;
2100 struct nfsmount *nmp;
2101 struct nfsclsession *tsep;
2103 nmp = VFSTONFS(dvp->v_mount);
2110 if (namelen > NFS_MAXNAMLEN)
2111 return (ENAMETOOLONG);
2112 NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2114 * For V4, this is actually an Open op.
2116 NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2117 *tl++ = txdr_unsigned(owp->nfsow_seqid);
2118 *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2119 NFSV4OPEN_ACCESSREAD);
2120 *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2121 tsep = nfsmnt_mdssession(nmp);
2122 *tl++ = tsep->nfsess_clientid.lval[0];
2123 *tl = tsep->nfsess_clientid.lval[1];
2124 (void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2125 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2126 *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2127 if (fmode & O_EXCL) {
2128 if (NFSHASNFSV4N(nmp)) {
2129 if (NFSHASSESSPERSIST(nmp)) {
2130 /* Use GUARDED for persistent sessions. */
2131 *tl = txdr_unsigned(NFSCREATE_GUARDED);
2132 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2134 /* Otherwise, use EXCLUSIVE4_1. */
2135 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2136 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2137 *tl++ = cverf.lval[0];
2138 *tl = cverf.lval[1];
2139 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2143 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2144 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2145 *tl++ = cverf.lval[0];
2146 *tl = cverf.lval[1];
2149 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2150 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2152 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2153 *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2154 (void) nfsm_strtom(nd, name, namelen);
2155 /* Get the new file's handle and attributes. */
2156 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2157 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2158 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2159 NFSGETATTR_ATTRBIT(&attrbits);
2160 (void) nfsrv_putattrbit(nd, &attrbits);
2161 /* Get the directory's post-op attributes. */
2162 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2163 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2164 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2165 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2166 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2167 (void) nfsrv_putattrbit(nd, &attrbits);
2168 error = nfscl_request(nd, dvp, p, cred, dstuff);
2171 NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2172 if (nd->nd_repstat == 0) {
2173 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2175 stateid.seqid = *tl++;
2176 stateid.other[0] = *tl++;
2177 stateid.other[1] = *tl++;
2178 stateid.other[2] = *tl;
2179 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2180 (void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2181 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2182 deleg = fxdr_unsigned(int, *tl);
2183 if (deleg == NFSV4OPEN_DELEGATEREAD ||
2184 deleg == NFSV4OPEN_DELEGATEWRITE) {
2185 if (!(owp->nfsow_clp->nfsc_flags &
2186 NFSCLFLAGS_FIRSTDELEG))
2187 owp->nfsow_clp->nfsc_flags |=
2188 (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2190 sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2191 M_NFSCLDELEG, M_WAITOK);
2192 LIST_INIT(&dp->nfsdl_owner);
2193 LIST_INIT(&dp->nfsdl_lock);
2194 dp->nfsdl_clp = owp->nfsow_clp;
2195 newnfs_copyincred(cred, &dp->nfsdl_cred);
2196 nfscl_lockinit(&dp->nfsdl_rwlock);
2197 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2199 dp->nfsdl_stateid.seqid = *tl++;
2200 dp->nfsdl_stateid.other[0] = *tl++;
2201 dp->nfsdl_stateid.other[1] = *tl++;
2202 dp->nfsdl_stateid.other[2] = *tl++;
2203 ret = fxdr_unsigned(int, *tl);
2204 if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2205 dp->nfsdl_flags = NFSCLDL_WRITE;
2207 * Indicates how much the file can grow.
2209 NFSM_DISSECT(tl, u_int32_t *,
2211 limitby = fxdr_unsigned(int, *tl++);
2213 case NFSV4OPEN_LIMITSIZE:
2214 dp->nfsdl_sizelimit = fxdr_hyper(tl);
2216 case NFSV4OPEN_LIMITBLOCKS:
2217 dp->nfsdl_sizelimit =
2218 fxdr_unsigned(u_int64_t, *tl++);
2219 dp->nfsdl_sizelimit *=
2220 fxdr_unsigned(u_int64_t, *tl);
2223 error = NFSERR_BADXDR;
2227 dp->nfsdl_flags = NFSCLDL_READ;
2230 dp->nfsdl_flags |= NFSCLDL_RECALL;
2231 error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2235 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
2236 error = NFSERR_BADXDR;
2239 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2242 /* Get rid of the PutFH and Getattr status values. */
2243 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2244 /* Load the directory attributes. */
2245 error = nfsm_loadattr(nd, dnap);
2249 if (dp != NULL && *attrflagp) {
2250 dp->nfsdl_change = nnap->na_filerev;
2251 dp->nfsdl_modtime = nnap->na_mtime;
2252 dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2255 * We can now complete the Open state.
2259 dp->nfsdl_fhlen = nfhp->nfh_len;
2260 NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2263 * Get an Open structure that will be
2264 * attached to the OpenOwner, acquired already.
2266 error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2267 (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2268 cred, p, NULL, &op, &newone, NULL, 0);
2271 op->nfso_stateid = stateid;
2272 newnfs_copyincred(cred, &op->nfso_cred);
2273 if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2275 ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2276 nfhp->nfh_len, op, cred, p);
2277 if (ret == NFSERR_DELAY)
2278 (void) nfs_catnap(PZERO, ret, "nfs_create");
2279 } while (ret == NFSERR_DELAY);
2284 * If the server is handing out delegations, but we didn't
2285 * get one because an OpenConfirm was required, try the
2286 * Open again, to get a delegation. This is a harmless no-op,
2287 * from a server's point of view.
2289 if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2290 (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2291 !error && dp == NULL) {
2293 ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2294 np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2295 nfhp->nfh_fh, nfhp->nfh_len,
2296 (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2297 name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2298 if (ret == NFSERR_DELAY)
2299 (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2300 } while (ret == NFSERR_DELAY);
2303 free(dp, M_NFSCLDELEG);
2306 if (ret == NFSERR_STALECLIENTID ||
2307 ret == NFSERR_STALEDONTRECOVER ||
2308 ret == NFSERR_BADSESSION)
2312 nfscl_openrelease(nmp, op, error, newone);
2315 if (nd->nd_repstat != 0 && error == 0)
2316 error = nd->nd_repstat;
2317 if (error == NFSERR_STALECLIENTID)
2318 nfscl_initiate_recovery(owp->nfsow_clp);
2322 else if (dp != NULL)
2323 free(dp, M_NFSCLDELEG);
2324 mbuf_freem(nd->nd_mrep);
2332 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2333 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2337 struct nfsrv_descript nfsd, *nd = &nfsd;
2339 struct nfsmount *nmp;
2340 nfsv4stateid_t dstateid;
2341 int error, ret = 0, i;
2344 if (namelen > NFS_MAXNAMLEN)
2345 return (ENAMETOOLONG);
2346 nmp = VFSTONFS(vnode_mount(dvp));
2348 if (NFSHASNFSV4(nmp) && ret == 0) {
2349 ret = nfscl_removedeleg(vp, p, &dstateid);
2351 NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2352 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2354 if (NFSHASNFSV4N(nmp))
2357 *tl++ = dstateid.seqid;
2358 *tl++ = dstateid.other[0];
2359 *tl++ = dstateid.other[1];
2360 *tl++ = dstateid.other[2];
2361 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2363 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2364 np->n_fhp->nfh_len, 0);
2365 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2366 *tl = txdr_unsigned(NFSV4OP_REMOVE);
2372 NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2373 (void) nfsm_strtom(nd, name, namelen);
2374 error = nfscl_request(nd, dvp, p, cred, dstuff);
2377 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2378 /* For NFSv4, parse out any Delereturn replies. */
2379 if (ret > 0 && nd->nd_repstat != 0 &&
2380 (nd->nd_flag & ND_NOMOREDATA)) {
2382 * If the Delegreturn failed, try again without
2383 * it. The server will Recall, as required.
2385 mbuf_freem(nd->nd_mrep);
2388 for (i = 0; i < (ret * 2); i++) {
2389 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2391 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2393 nd->nd_flag |= ND_NOMOREDATA;
2396 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2398 if (nd->nd_repstat && !error)
2399 error = nd->nd_repstat;
2401 mbuf_freem(nd->nd_mrep);
2406 * Do an nfs rename rpc.
2409 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2410 vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2411 NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2412 int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2415 struct nfsrv_descript nfsd, *nd = &nfsd;
2416 struct nfsmount *nmp;
2418 nfsattrbit_t attrbits;
2419 nfsv4stateid_t fdstateid, tdstateid;
2420 int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2424 nmp = VFSTONFS(vnode_mount(fdvp));
2425 if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2426 return (ENAMETOOLONG);
2428 if (NFSHASNFSV4(nmp) && ret == 0) {
2429 ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2430 &tdstateid, &gottd, p);
2431 if (gotfd && gottd) {
2432 NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2434 NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2436 NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2439 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2440 if (NFSHASNFSV4N(nmp))
2443 *tl++ = fdstateid.seqid;
2444 *tl++ = fdstateid.other[0];
2445 *tl++ = fdstateid.other[1];
2446 *tl = fdstateid.other[2];
2448 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2449 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2451 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2452 np->n_fhp->nfh_len, 0);
2453 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2454 *tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2458 NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2459 if (NFSHASNFSV4N(nmp))
2462 *tl++ = tdstateid.seqid;
2463 *tl++ = tdstateid.other[0];
2464 *tl++ = tdstateid.other[1];
2465 *tl = tdstateid.other[2];
2468 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2469 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2471 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2472 np->n_fhp->nfh_len, 0);
2473 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2474 *tl = txdr_unsigned(NFSV4OP_SAVEFH);
2480 NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2481 if (nd->nd_flag & ND_NFSV4) {
2482 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2483 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2484 NFSWCCATTR_ATTRBIT(&attrbits);
2485 (void) nfsrv_putattrbit(nd, &attrbits);
2486 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2487 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2488 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2489 VTONFS(tdvp)->n_fhp->nfh_len, 0);
2490 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2491 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2492 (void) nfsrv_putattrbit(nd, &attrbits);
2493 nd->nd_flag |= ND_V4WCCATTR;
2494 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2495 *tl = txdr_unsigned(NFSV4OP_RENAME);
2497 (void) nfsm_strtom(nd, fnameptr, fnamelen);
2498 if (!(nd->nd_flag & ND_NFSV4))
2499 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2500 VTONFS(tdvp)->n_fhp->nfh_len, 0);
2501 (void) nfsm_strtom(nd, tnameptr, tnamelen);
2502 error = nfscl_request(nd, fdvp, p, cred, fstuff);
2505 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2506 /* For NFSv4, parse out any Delereturn replies. */
2507 if (ret > 0 && nd->nd_repstat != 0 &&
2508 (nd->nd_flag & ND_NOMOREDATA)) {
2510 * If the Delegreturn failed, try again without
2511 * it. The server will Recall, as required.
2513 mbuf_freem(nd->nd_mrep);
2516 for (i = 0; i < (ret * 2); i++) {
2517 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2519 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2521 if (i == 0 && ret > 1) {
2523 * If the Delegreturn failed, try again
2524 * without it. The server will Recall, as
2526 * If ret > 1, the first iteration of this
2527 * loop is the second DelegReturn result.
2529 mbuf_freem(nd->nd_mrep);
2532 nd->nd_flag |= ND_NOMOREDATA;
2537 /* Now, the first wcc attribute reply. */
2538 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2539 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2541 nd->nd_flag |= ND_NOMOREDATA;
2543 error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2545 /* and the second wcc attribute reply. */
2546 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2548 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2550 nd->nd_flag |= ND_NOMOREDATA;
2553 error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2556 if (nd->nd_repstat && !error)
2557 error = nd->nd_repstat;
2559 mbuf_freem(nd->nd_mrep);
2564 * nfs hard link create rpc
2567 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2568 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2569 struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2572 struct nfsrv_descript nfsd, *nd = &nfsd;
2573 nfsattrbit_t attrbits;
2578 if (namelen > NFS_MAXNAMLEN)
2579 return (ENAMETOOLONG);
2580 NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2581 if (nd->nd_flag & ND_NFSV4) {
2582 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2583 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2585 (void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2586 VTONFS(dvp)->n_fhp->nfh_len, 0);
2587 if (nd->nd_flag & ND_NFSV4) {
2588 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2589 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2590 NFSWCCATTR_ATTRBIT(&attrbits);
2591 (void) nfsrv_putattrbit(nd, &attrbits);
2592 nd->nd_flag |= ND_V4WCCATTR;
2593 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2594 *tl = txdr_unsigned(NFSV4OP_LINK);
2596 (void) nfsm_strtom(nd, name, namelen);
2597 error = nfscl_request(nd, vp, p, cred, dstuff);
2600 if (nd->nd_flag & ND_NFSV3) {
2601 error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2603 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2605 } else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2607 * First, parse out the PutFH and Getattr result.
2609 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2611 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2613 nd->nd_flag |= ND_NOMOREDATA;
2615 * Get the pre-op attributes.
2617 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2619 if (nd->nd_repstat && !error)
2620 error = nd->nd_repstat;
2622 mbuf_freem(nd->nd_mrep);
2627 * nfs symbolic link create rpc
2630 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
2631 struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2632 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2633 int *dattrflagp, void *dstuff)
2636 struct nfsrv_descript nfsd, *nd = &nfsd;
2637 struct nfsmount *nmp;
2638 int slen, error = 0;
2643 nmp = VFSTONFS(vnode_mount(dvp));
2644 slen = strlen(target);
2645 if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2646 return (ENAMETOOLONG);
2647 NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2648 if (nd->nd_flag & ND_NFSV4) {
2649 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2650 *tl = txdr_unsigned(NFLNK);
2651 (void) nfsm_strtom(nd, target, slen);
2653 (void) nfsm_strtom(nd, name, namelen);
2654 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2655 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2656 if (!(nd->nd_flag & ND_NFSV4))
2657 (void) nfsm_strtom(nd, target, slen);
2658 if (nd->nd_flag & ND_NFSV2)
2659 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2660 error = nfscl_request(nd, dvp, p, cred, dstuff);
2663 if (nd->nd_flag & ND_NFSV4)
2664 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2665 if ((nd->nd_flag & ND_NFSV3) && !error) {
2666 if (!nd->nd_repstat)
2667 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2669 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2672 if (nd->nd_repstat && !error)
2673 error = nd->nd_repstat;
2674 mbuf_freem(nd->nd_mrep);
2676 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2677 * Only do this if vfs.nfs.ignore_eexist is set.
2678 * Never do this for NFSv4.1 or later minor versions, since sessions
2679 * should guarantee "exactly once" RPC semantics.
2681 if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2682 nmp->nm_minorvers == 0))
2691 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2692 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2693 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2694 int *dattrflagp, void *dstuff)
2697 struct nfsrv_descript nfsd, *nd = &nfsd;
2698 nfsattrbit_t attrbits;
2701 struct nfsmount *nmp;
2706 nmp = VFSTONFS(vnode_mount(dvp));
2707 fhp = VTONFS(dvp)->n_fhp;
2708 if (namelen > NFS_MAXNAMLEN)
2709 return (ENAMETOOLONG);
2710 NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2711 if (nd->nd_flag & ND_NFSV4) {
2712 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2713 *tl = txdr_unsigned(NFDIR);
2715 (void) nfsm_strtom(nd, name, namelen);
2716 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2717 if (nd->nd_flag & ND_NFSV4) {
2718 NFSGETATTR_ATTRBIT(&attrbits);
2719 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2720 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2721 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2722 (void) nfsrv_putattrbit(nd, &attrbits);
2723 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2724 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2725 (void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2726 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2727 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2728 (void) nfsrv_putattrbit(nd, &attrbits);
2730 error = nfscl_request(nd, dvp, p, cred, dstuff);
2733 if (nd->nd_flag & ND_NFSV4)
2734 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2735 if (!nd->nd_repstat && !error) {
2736 if (nd->nd_flag & ND_NFSV4) {
2737 NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2738 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2741 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2742 if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2743 /* Get rid of the PutFH and Getattr status values. */
2744 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2745 /* Load the directory attributes. */
2746 error = nfsm_loadattr(nd, dnap);
2751 if ((nd->nd_flag & ND_NFSV3) && !error)
2752 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2753 if (nd->nd_repstat && !error)
2754 error = nd->nd_repstat;
2756 mbuf_freem(nd->nd_mrep);
2758 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2759 * Only do this if vfs.nfs.ignore_eexist is set.
2760 * Never do this for NFSv4.1 or later minor versions, since sessions
2761 * should guarantee "exactly once" RPC semantics.
2763 if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2764 nmp->nm_minorvers == 0))
2770 * nfs remove directory call
2773 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2774 NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2776 struct nfsrv_descript nfsd, *nd = &nfsd;
2780 if (namelen > NFS_MAXNAMLEN)
2781 return (ENAMETOOLONG);
2782 NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2783 (void) nfsm_strtom(nd, name, namelen);
2784 error = nfscl_request(nd, dvp, p, cred, dstuff);
2787 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2788 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2789 if (nd->nd_repstat && !error)
2790 error = nd->nd_repstat;
2791 mbuf_freem(nd->nd_mrep);
2793 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2795 if (error == ENOENT)
2802 * Always returns with either uio_resid unchanged, if you are at the
2803 * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2805 * I felt this would allow caching of directory blocks more easily
2806 * than returning a pertially filled block.
2807 * Directory offset cookies:
2808 * Oh my, what to do with them...
2809 * I can think of three ways to deal with them:
2810 * 1 - have the layer above these RPCs maintain a map between logical
2811 * directory byte offsets and the NFS directory offset cookies
2812 * 2 - pass the opaque directory offset cookies up into userland
2813 * and let the libc functions deal with them, via the system call
2814 * 3 - return them to userland in the "struct dirent", so future versions
2815 * of libc can use them and do whatever is necessary to make things work
2816 * above these rpc calls, in the meantime
2817 * For now, I do #3 by "hiding" the directory offset cookies after the
2818 * d_name field in struct dirent. This is space inside d_reclen that
2819 * will be ignored by anything that doesn't know about them.
2820 * The directory offset cookies are filled in as the last 8 bytes of
2821 * each directory entry, after d_name. Someday, the userland libc
2822 * functions may be able to use these. In the meantime, it satisfies
2823 * OpenBSD's requirements for cookies being returned.
2824 * If expects the directory offset cookie for the read to be in uio_offset
2825 * and returns the one for the next entry after this directory block in
2829 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2830 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2831 int *eofp, void *stuff)
2834 struct dirent *dp = NULL;
2836 nfsquad_t cookie, ncookie;
2837 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2838 struct nfsnode *dnp = VTONFS(vp);
2839 struct nfsvattr nfsva;
2840 struct nfsrv_descript nfsd, *nd = &nfsd;
2841 int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2842 int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2843 u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2845 nfsattrbit_t attrbits, dattrbits;
2846 u_int32_t rderr, *tl2 = NULL;
2849 KASSERT(uiop->uio_iovcnt == 1 &&
2850 (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2851 ("nfs readdirrpc bad uio"));
2852 ncookie.lval[0] = ncookie.lval[1] = 0;
2854 * There is no point in reading a lot more than uio_resid, however
2855 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2856 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2857 * will never make readsize > nm_readdirsize.
2859 readsize = nmp->nm_readdirsize;
2860 if (readsize > uio_uio_resid(uiop))
2861 readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2866 tresid = uio_uio_resid(uiop);
2867 cookie.lval[0] = cookiep->nfsuquad[0];
2868 cookie.lval[1] = cookiep->nfsuquad[1];
2872 * For NFSv4, first create the "." and ".." entries.
2874 if (NFSHASNFSV4(nmp)) {
2875 reqsize = 6 * NFSX_UNSIGNED;
2876 NFSGETATTR_ATTRBIT(&dattrbits);
2877 NFSZERO_ATTRBIT(&attrbits);
2878 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2879 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2880 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2881 NFSATTRBIT_MOUNTEDONFILEID)) {
2882 NFSSETBIT_ATTRBIT(&attrbits,
2883 NFSATTRBIT_MOUNTEDONFILEID);
2887 * Must fake it. Use the fileno, except when the
2888 * fsid is != to that of the directory. For that
2889 * case, generate a fake fileno that is not the same.
2891 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2896 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2898 if (uiop->uio_offset == 0) {
2899 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2900 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2901 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2902 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2903 (void) nfsrv_putattrbit(nd, &attrbits);
2904 error = nfscl_request(nd, vp, p, cred, stuff);
2907 dotfileid = 0; /* Fake out the compiler. */
2908 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2909 error = nfsm_loadattr(nd, &nfsva);
2912 dotfileid = nfsva.na_fileid;
2914 if (nd->nd_repstat == 0) {
2915 NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2916 len = fxdr_unsigned(int, *(tl + 4));
2917 if (len > 0 && len <= NFSX_V4FHMAX)
2918 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2922 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2923 nfsva.na_mntonfileno = UINT64_MAX;
2924 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2925 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2926 NULL, NULL, NULL, p, cred);
2928 dotdotfileid = dotfileid;
2929 } else if (gotmnton) {
2930 if (nfsva.na_mntonfileno != UINT64_MAX)
2931 dotdotfileid = nfsva.na_mntonfileno;
2933 dotdotfileid = nfsva.na_fileid;
2934 } else if (nfsva.na_filesid[0] ==
2935 dnp->n_vattr.na_filesid[0] &&
2936 nfsva.na_filesid[1] ==
2937 dnp->n_vattr.na_filesid[1]) {
2938 dotdotfileid = nfsva.na_fileid;
2942 } while (fakefileno ==
2944 dotdotfileid = fakefileno;
2947 } else if (nd->nd_repstat == NFSERR_NOENT) {
2949 * Lookupp returns NFSERR_NOENT when we are
2950 * at the root, so just use the current dir.
2953 dotdotfileid = dotfileid;
2955 error = nd->nd_repstat;
2957 mbuf_freem(nd->nd_mrep);
2961 dp = (struct dirent *)uio_iov_base(uiop);
2962 dp->d_pad0 = dp->d_pad1 = 0;
2964 dp->d_type = DT_DIR;
2965 dp->d_fileno = dotfileid;
2967 *((uint64_t *)dp->d_name) = 0; /* Zero pad it. */
2968 dp->d_name[0] = '.';
2969 dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
2971 * Just make these offset cookie 0.
2973 tl = (u_int32_t *)&dp->d_name[8];
2976 blksiz += dp->d_reclen;
2977 uio_uio_resid_add(uiop, -(dp->d_reclen));
2978 uiop->uio_offset += dp->d_reclen;
2979 uio_iov_base_add(uiop, dp->d_reclen);
2980 uio_iov_len_add(uiop, -(dp->d_reclen));
2981 dp = (struct dirent *)uio_iov_base(uiop);
2982 dp->d_pad0 = dp->d_pad1 = 0;
2984 dp->d_type = DT_DIR;
2985 dp->d_fileno = dotdotfileid;
2987 *((uint64_t *)dp->d_name) = 0;
2988 dp->d_name[0] = '.';
2989 dp->d_name[1] = '.';
2990 dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
2992 * Just make these offset cookie 0.
2994 tl = (u_int32_t *)&dp->d_name[8];
2997 blksiz += dp->d_reclen;
2998 uio_uio_resid_add(uiop, -(dp->d_reclen));
2999 uiop->uio_offset += dp->d_reclen;
3000 uio_iov_base_add(uiop, dp->d_reclen);
3001 uio_iov_len_add(uiop, -(dp->d_reclen));
3003 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3005 reqsize = 5 * NFSX_UNSIGNED;
3010 * Loop around doing readdir rpc's of size readsize.
3011 * The stopping criteria is EOF or buffer full.
3013 while (more_dirs && bigenough) {
3015 NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3016 if (nd->nd_flag & ND_NFSV2) {
3017 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3018 *tl++ = cookie.lval[1];
3019 *tl = txdr_unsigned(readsize);
3021 NFSM_BUILD(tl, u_int32_t *, reqsize);
3022 *tl++ = cookie.lval[0];
3023 *tl++ = cookie.lval[1];
3024 if (cookie.qval == 0) {
3029 *tl++ = dnp->n_cookieverf.nfsuquad[0];
3030 *tl++ = dnp->n_cookieverf.nfsuquad[1];
3033 if (nd->nd_flag & ND_NFSV4) {
3034 *tl++ = txdr_unsigned(readsize);
3035 *tl = txdr_unsigned(readsize);
3036 (void) nfsrv_putattrbit(nd, &attrbits);
3037 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3038 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3039 (void) nfsrv_putattrbit(nd, &dattrbits);
3041 *tl = txdr_unsigned(readsize);
3044 error = nfscl_request(nd, vp, p, cred, stuff);
3047 if (!(nd->nd_flag & ND_NFSV2)) {
3048 if (nd->nd_flag & ND_NFSV3)
3049 error = nfscl_postop_attr(nd, nap, attrflagp,
3051 if (!nd->nd_repstat && !error) {
3052 NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3054 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3055 dnp->n_cookieverf.nfsuquad[1] = *tl;
3059 if (nd->nd_repstat || error) {
3061 error = nd->nd_repstat;
3064 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3065 more_dirs = fxdr_unsigned(int, *tl);
3069 /* loop through the dir entries, doctoring them to 4bsd form */
3070 while (more_dirs && bigenough) {
3071 if (nd->nd_flag & ND_NFSV4) {
3072 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3073 ncookie.lval[0] = *tl++;
3074 ncookie.lval[1] = *tl++;
3075 len = fxdr_unsigned(int, *tl);
3076 } else if (nd->nd_flag & ND_NFSV3) {
3077 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3078 nfsva.na_fileid = fxdr_hyper(tl);
3080 len = fxdr_unsigned(int, *tl);
3082 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3083 nfsva.na_fileid = fxdr_unsigned(uint64_t,
3085 len = fxdr_unsigned(int, *tl);
3087 if (len <= 0 || len > NFS_MAXNAMLEN) {
3091 tlen = roundup2(len, 8);
3093 tlen += 8; /* To ensure null termination. */
3094 left = DIRBLKSIZ - blksiz;
3095 if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3096 NFSBZERO(uio_iov_base(uiop), left);
3097 dp->d_reclen += left;
3098 uio_iov_base_add(uiop, left);
3099 uio_iov_len_add(uiop, -(left));
3100 uio_uio_resid_add(uiop, -(left));
3101 uiop->uio_offset += left;
3104 if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3105 uio_uio_resid(uiop))
3108 dp = (struct dirent *)uio_iov_base(uiop);
3109 dp->d_pad0 = dp->d_pad1 = 0;
3112 dp->d_reclen = _GENERIC_DIRLEN(len) +
3114 dp->d_type = DT_UNKNOWN;
3115 blksiz += dp->d_reclen;
3116 if (blksiz == DIRBLKSIZ)
3118 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3119 uiop->uio_offset += DIRHDSIZ;
3120 uio_iov_base_add(uiop, DIRHDSIZ);
3121 uio_iov_len_add(uiop, -(DIRHDSIZ));
3122 error = nfsm_mbufuio(nd, uiop, len);
3125 cp = uio_iov_base(uiop);
3128 cp += tlen; /* points to cookie storage */
3129 tl2 = (u_int32_t *)cp;
3130 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3131 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3132 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3133 uiop->uio_offset += (tlen + NFSX_HYPER);
3135 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3139 if (nd->nd_flag & ND_NFSV4) {
3141 nfsva.na_mntonfileno = UINT64_MAX;
3142 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3143 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3144 NULL, NULL, &rderr, p, cred);
3147 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3148 } else if (nd->nd_flag & ND_NFSV3) {
3149 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3150 ncookie.lval[0] = *tl++;
3151 ncookie.lval[1] = *tl++;
3153 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3154 ncookie.lval[0] = 0;
3155 ncookie.lval[1] = *tl++;
3158 if (nd->nd_flag & ND_NFSV4) {
3163 if (nfsva.na_mntonfileno != UINT64_MAX)
3164 dp->d_fileno = nfsva.na_mntonfileno;
3166 dp->d_fileno = nfsva.na_fileid;
3167 } else if (nfsva.na_filesid[0] ==
3168 dnp->n_vattr.na_filesid[0] &&
3169 nfsva.na_filesid[1] ==
3170 dnp->n_vattr.na_filesid[1]) {
3171 dp->d_fileno = nfsva.na_fileid;
3175 } while (fakefileno ==
3177 dp->d_fileno = fakefileno;
3179 dp->d_type = vtonfs_dtype(nfsva.na_type);
3182 dp->d_fileno = nfsva.na_fileid;
3184 *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3186 *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3189 more_dirs = fxdr_unsigned(int, *tl);
3192 * If at end of rpc data, get the eof boolean
3195 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3196 eof = fxdr_unsigned(int, *tl);
3199 if (nd->nd_flag & ND_NFSV4) {
3200 error = nfscl_postop_attr(nd, nap, attrflagp,
3206 mbuf_freem(nd->nd_mrep);
3210 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3211 * by increasing d_reclen for the last record.
3214 left = DIRBLKSIZ - blksiz;
3215 NFSBZERO(uio_iov_base(uiop), left);
3216 dp->d_reclen += left;
3217 uio_iov_base_add(uiop, left);
3218 uio_iov_len_add(uiop, -(left));
3219 uio_uio_resid_add(uiop, -(left));
3220 uiop->uio_offset += left;
3224 * If returning no data, assume end of file.
3225 * If not bigenough, return not end of file, since you aren't
3226 * returning all the data
3227 * Otherwise, return the eof flag from the server.
3230 if (tresid == ((size_t)(uio_uio_resid(uiop))))
3232 else if (!bigenough)
3239 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3241 while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3242 dp = (struct dirent *)uio_iov_base(uiop);
3243 NFSBZERO(dp, DIRBLKSIZ);
3244 dp->d_type = DT_UNKNOWN;
3245 tl = (u_int32_t *)&dp->d_name[4];
3246 *tl++ = cookie.lval[0];
3247 *tl = cookie.lval[1];
3248 dp->d_reclen = DIRBLKSIZ;
3249 uio_iov_base_add(uiop, DIRBLKSIZ);
3250 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3251 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3252 uiop->uio_offset += DIRBLKSIZ;
3256 if (nd->nd_mrep != NULL)
3257 mbuf_freem(nd->nd_mrep);
3263 * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3264 * (Also used for NFS V4 when mount flag set.)
3265 * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3268 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3269 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3270 int *eofp, void *stuff)
3273 struct dirent *dp = NULL;
3275 vnode_t newvp = NULLVP;
3276 struct nfsrv_descript nfsd, *nd = &nfsd;
3277 struct nameidata nami, *ndp = &nami;
3278 struct componentname *cnp = &ndp->ni_cnd;
3279 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3280 struct nfsnode *dnp = VTONFS(vp), *np;
3281 struct nfsvattr nfsva;
3283 nfsquad_t cookie, ncookie;
3284 int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3285 int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3286 int isdotdot = 0, unlocknewvp = 0;
3287 u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3288 u_int64_t fileno = 0;
3290 nfsattrbit_t attrbits, dattrbits;
3292 u_int32_t *tl2 = NULL, rderr;
3293 struct timespec dctime;
3295 KASSERT(uiop->uio_iovcnt == 1 &&
3296 (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3297 ("nfs readdirplusrpc bad uio"));
3298 ncookie.lval[0] = ncookie.lval[1] = 0;
3299 timespecclear(&dctime);
3305 cookie.lval[0] = cookiep->nfsuquad[0];
3306 cookie.lval[1] = cookiep->nfsuquad[1];
3307 tresid = uio_uio_resid(uiop);
3310 * For NFSv4, first create the "." and ".." entries.
3312 if (NFSHASNFSV4(nmp)) {
3313 NFSGETATTR_ATTRBIT(&dattrbits);
3314 NFSZERO_ATTRBIT(&attrbits);
3315 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3316 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3317 NFSATTRBIT_MOUNTEDONFILEID)) {
3318 NFSSETBIT_ATTRBIT(&attrbits,
3319 NFSATTRBIT_MOUNTEDONFILEID);
3323 * Must fake it. Use the fileno, except when the
3324 * fsid is != to that of the directory. For that
3325 * case, generate a fake fileno that is not the same.
3327 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3332 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3334 if (uiop->uio_offset == 0) {
3335 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3336 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3337 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
3338 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3339 (void) nfsrv_putattrbit(nd, &attrbits);
3340 error = nfscl_request(nd, vp, p, cred, stuff);
3343 dotfileid = 0; /* Fake out the compiler. */
3344 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3345 error = nfsm_loadattr(nd, &nfsva);
3348 dctime = nfsva.na_ctime;
3349 dotfileid = nfsva.na_fileid;
3351 if (nd->nd_repstat == 0) {
3352 NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3353 len = fxdr_unsigned(int, *(tl + 4));
3354 if (len > 0 && len <= NFSX_V4FHMAX)
3355 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3359 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3360 nfsva.na_mntonfileno = UINT64_MAX;
3361 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3362 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3363 NULL, NULL, NULL, p, cred);
3365 dotdotfileid = dotfileid;
3366 } else if (gotmnton) {
3367 if (nfsva.na_mntonfileno != UINT64_MAX)
3368 dotdotfileid = nfsva.na_mntonfileno;
3370 dotdotfileid = nfsva.na_fileid;
3371 } else if (nfsva.na_filesid[0] ==
3372 dnp->n_vattr.na_filesid[0] &&
3373 nfsva.na_filesid[1] ==
3374 dnp->n_vattr.na_filesid[1]) {
3375 dotdotfileid = nfsva.na_fileid;
3379 } while (fakefileno ==
3381 dotdotfileid = fakefileno;
3384 } else if (nd->nd_repstat == NFSERR_NOENT) {
3386 * Lookupp returns NFSERR_NOENT when we are
3387 * at the root, so just use the current dir.
3390 dotdotfileid = dotfileid;
3392 error = nd->nd_repstat;
3394 mbuf_freem(nd->nd_mrep);
3398 dp = (struct dirent *)uio_iov_base(uiop);
3399 dp->d_pad0 = dp->d_pad1 = 0;
3401 dp->d_type = DT_DIR;
3402 dp->d_fileno = dotfileid;
3404 *((uint64_t *)dp->d_name) = 0; /* Zero pad it. */
3405 dp->d_name[0] = '.';
3406 dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3408 * Just make these offset cookie 0.
3410 tl = (u_int32_t *)&dp->d_name[8];
3413 blksiz += dp->d_reclen;
3414 uio_uio_resid_add(uiop, -(dp->d_reclen));
3415 uiop->uio_offset += dp->d_reclen;
3416 uio_iov_base_add(uiop, dp->d_reclen);
3417 uio_iov_len_add(uiop, -(dp->d_reclen));
3418 dp = (struct dirent *)uio_iov_base(uiop);
3419 dp->d_pad0 = dp->d_pad1 = 0;
3421 dp->d_type = DT_DIR;
3422 dp->d_fileno = dotdotfileid;
3424 *((uint64_t *)dp->d_name) = 0;
3425 dp->d_name[0] = '.';
3426 dp->d_name[1] = '.';
3427 dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3429 * Just make these offset cookie 0.
3431 tl = (u_int32_t *)&dp->d_name[8];
3434 blksiz += dp->d_reclen;
3435 uio_uio_resid_add(uiop, -(dp->d_reclen));
3436 uiop->uio_offset += dp->d_reclen;
3437 uio_iov_base_add(uiop, dp->d_reclen);
3438 uio_iov_len_add(uiop, -(dp->d_reclen));
3440 NFSREADDIRPLUS_ATTRBIT(&attrbits);
3442 NFSSETBIT_ATTRBIT(&attrbits,
3443 NFSATTRBIT_MOUNTEDONFILEID);
3447 * Loop around doing readdir rpc's of size nm_readdirsize.
3448 * The stopping criteria is EOF or buffer full.
3450 while (more_dirs && bigenough) {
3452 NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3453 NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3454 *tl++ = cookie.lval[0];
3455 *tl++ = cookie.lval[1];
3456 if (cookie.qval == 0) {
3461 *tl++ = dnp->n_cookieverf.nfsuquad[0];
3462 *tl++ = dnp->n_cookieverf.nfsuquad[1];
3465 *tl++ = txdr_unsigned(nmp->nm_readdirsize);
3466 *tl = txdr_unsigned(nmp->nm_readdirsize);
3467 if (nd->nd_flag & ND_NFSV4) {
3468 (void) nfsrv_putattrbit(nd, &attrbits);
3469 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3470 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3471 (void) nfsrv_putattrbit(nd, &dattrbits);
3473 error = nfscl_request(nd, vp, p, cred, stuff);
3476 if (nd->nd_flag & ND_NFSV3)
3477 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3478 if (nd->nd_repstat || error) {
3480 error = nd->nd_repstat;
3483 if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3484 dctime = nap->na_ctime;
3485 NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3487 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3488 dnp->n_cookieverf.nfsuquad[1] = *tl++;
3490 more_dirs = fxdr_unsigned(int, *tl);
3494 /* loop through the dir entries, doctoring them to 4bsd form */
3495 while (more_dirs && bigenough) {
3496 NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3497 if (nd->nd_flag & ND_NFSV4) {
3498 ncookie.lval[0] = *tl++;
3499 ncookie.lval[1] = *tl++;
3501 fileno = fxdr_hyper(tl);
3504 len = fxdr_unsigned(int, *tl);
3505 if (len <= 0 || len > NFS_MAXNAMLEN) {
3509 tlen = roundup2(len, 8);
3511 tlen += 8; /* To ensure null termination. */
3512 left = DIRBLKSIZ - blksiz;
3513 if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3514 NFSBZERO(uio_iov_base(uiop), left);
3515 dp->d_reclen += left;
3516 uio_iov_base_add(uiop, left);
3517 uio_iov_len_add(uiop, -(left));
3518 uio_uio_resid_add(uiop, -(left));
3519 uiop->uio_offset += left;
3522 if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3523 uio_uio_resid(uiop))
3526 dp = (struct dirent *)uio_iov_base(uiop);
3527 dp->d_pad0 = dp->d_pad1 = 0;
3530 dp->d_reclen = _GENERIC_DIRLEN(len) +
3532 dp->d_type = DT_UNKNOWN;
3533 blksiz += dp->d_reclen;
3534 if (blksiz == DIRBLKSIZ)
3536 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3537 uiop->uio_offset += DIRHDSIZ;
3538 uio_iov_base_add(uiop, DIRHDSIZ);
3539 uio_iov_len_add(uiop, -(DIRHDSIZ));
3540 cnp->cn_nameptr = uio_iov_base(uiop);
3541 cnp->cn_namelen = len;
3543 error = nfsm_mbufuio(nd, uiop, len);
3546 cp = uio_iov_base(uiop);
3549 cp += tlen; /* points to cookie storage */
3550 tl2 = (u_int32_t *)cp;
3551 if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3552 cnp->cn_nameptr[1] == '.')
3556 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3557 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3558 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3559 uiop->uio_offset += (tlen + NFSX_HYPER);
3561 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3566 if (nd->nd_flag & ND_NFSV3) {
3567 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3568 ncookie.lval[0] = *tl++;
3569 ncookie.lval[1] = *tl++;
3570 attrflag = fxdr_unsigned(int, *tl);
3572 error = nfsm_loadattr(nd, &nfsva);
3576 NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3578 error = nfsm_getfh(nd, &nfhp);
3582 if (!attrflag && nfhp != NULL) {
3583 free(nfhp, M_NFSFH);
3588 nfsva.na_mntonfileno = 0xffffffff;
3589 error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3590 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3591 NULL, NULL, &rderr, p, cred);
3597 if (nd->nd_flag & ND_NFSV4) {
3600 } else if (gotmnton) {
3601 if (nfsva.na_mntonfileno != 0xffffffff)
3602 dp->d_fileno = nfsva.na_mntonfileno;
3604 dp->d_fileno = nfsva.na_fileid;
3605 } else if (nfsva.na_filesid[0] ==
3606 dnp->n_vattr.na_filesid[0] &&
3607 nfsva.na_filesid[1] ==
3608 dnp->n_vattr.na_filesid[1]) {
3609 dp->d_fileno = nfsva.na_fileid;
3613 } while (fakefileno ==
3615 dp->d_fileno = fakefileno;
3618 dp->d_fileno = fileno;
3620 *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3622 *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3626 if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3627 dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3631 free(nfhp, M_NFSFH);
3633 } else if (isdotdot != 0) {
3635 * Skip doing a nfscl_nget() call for "..".
3636 * There's a race between acquiring the nfs
3637 * node here and lookups that look for the
3638 * directory being read (in the parent).
3639 * It would try to get a lock on ".." here,
3640 * owning the lock on the directory being
3641 * read. Lookup will hold the lock on ".."
3642 * and try to acquire the lock on the
3643 * directory being read.
3644 * If the directory is unlocked/relocked,
3645 * then there is a LOR with the buflock
3648 free(nfhp, M_NFSFH);
3650 error = nfscl_nget(vnode_mount(vp), vp,
3651 nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3658 if (newvp != NULLVP) {
3659 error = nfscl_loadattrcache(&newvp,
3660 &nfsva, NULL, NULL, 0, 0);
3669 vtonfs_dtype(np->n_vattr.na_type);
3671 NFSCNHASH(cnp, HASHINIT);
3672 if (cnp->cn_namelen <= NCHNAMLEN &&
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 >> 6) & 0x3ffffff;
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 last write to a mirrored DS with this
6111 if (mirror < flp->nfsfl_mirrorcnt - 1)
6112 error = nfsio_writedsmir(vp, iomode,
6113 must_commit, stateidp, *dspp, off,
6114 xfer, fhp, m, dp->nfsdi_vers,
6115 dp->nfsdi_minorvers, drpc, tcred,
6118 error = nfsrpc_writedsmir(vp, iomode,
6119 must_commit, stateidp, *dspp, off,
6120 xfer, fhp, m, dp->nfsdi_vers,
6121 dp->nfsdi_minorvers, tcred, p);
6122 NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6123 if (error != 0 && error != EACCES && error !=
6126 "DS layreterr for write\n");
6127 nfscl_dserr(NFSV4OP_WRITE, error, dp,
6132 NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6138 if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6141 NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6146 * The actual read RPC done to a DS.
6149 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6150 struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6151 int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6154 int attrflag, error, retlen;
6155 struct nfsrv_descript nfsd;
6156 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6157 struct nfsrv_descript *nd = &nfsd;
6158 struct nfssockreq *nrp;
6162 if (vers == 0 || vers == NFS_VER4) {
6163 nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6164 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6166 NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6168 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6170 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6172 nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6173 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6174 NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6176 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6177 txdr_hyper(io_off, tl);
6178 *(tl + 2) = txdr_unsigned(len);
6179 nrp = dsp->nfsclds_sockp;
6180 NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6182 /* If NULL, use the MDS socket. */
6183 nrp = &nmp->nm_sockreq;
6184 error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6185 NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6186 NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6190 if (vers == NFS_VER3) {
6191 error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6192 NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6196 if (nd->nd_repstat != 0) {
6197 error = nd->nd_repstat;
6200 if (vers == NFS_VER3) {
6201 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6202 *eofp = fxdr_unsigned(int, *(tl + 1));
6204 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6205 *eofp = fxdr_unsigned(int, *tl);
6207 NFSM_STRSIZ(retlen, len);
6208 NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6209 error = nfsm_mbufuio(nd, uiop, retlen);
6211 if (nd->nd_mrep != NULL)
6212 mbuf_freem(nd->nd_mrep);
6217 * The actual write RPC done to a DS.
6220 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6221 nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6222 struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6223 struct ucred *cred, NFSPROC_T *p)
6226 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6227 int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6229 struct nfsrv_descript nfsd;
6230 struct nfsrv_descript *nd = &nfsd;
6231 struct nfssockreq *nrp;
6234 KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6236 if (vers == 0 || vers == NFS_VER4) {
6237 nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6238 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6239 NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6242 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6244 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6245 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6247 nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6248 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6249 NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6250 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6252 txdr_hyper(io_off, tl);
6254 if (vers == NFS_VER3)
6255 *tl++ = txdr_unsigned(len);
6256 *tl++ = txdr_unsigned(*iomode);
6257 *tl = txdr_unsigned(len);
6258 nfsm_uiombuf(nd, uiop, len);
6259 nrp = dsp->nfsclds_sockp;
6261 /* If NULL, use the MDS socket. */
6262 nrp = &nmp->nm_sockreq;
6263 error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6264 NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6265 NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6269 if (nd->nd_repstat != 0) {
6271 * In case the rpc gets retried, roll
6272 * the uio fileds changed by nfsm_uiombuf()
6275 uiop->uio_offset -= len;
6276 uio_uio_resid_add(uiop, len);
6277 uio_iov_base_add(uiop, -len);
6278 uio_iov_len_add(uiop, len);
6279 error = nd->nd_repstat;
6281 if (vers == NFS_VER3) {
6282 error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6284 NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6288 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6289 rlen = fxdr_unsigned(int, *tl++);
6290 NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6294 } else if (rlen < len) {
6295 backup = len - rlen;
6296 uio_iov_base_add(uiop, -(backup));
6297 uio_iov_len_add(uiop, backup);
6298 uiop->uio_offset -= backup;
6299 uio_uio_resid_add(uiop, backup);
6302 commit = fxdr_unsigned(int, *tl++);
6305 * Return the lowest commitment level
6306 * obtained by any of the RPCs.
6308 if (committed == NFSWRITE_FILESYNC)
6310 else if (committed == NFSWRITE_DATASYNC &&
6311 commit == NFSWRITE_UNSTABLE)
6313 if (commit_thru_mds != 0) {
6315 if (!NFSHASWRITEVERF(nmp)) {
6316 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6317 NFSSETWRITEVERF(nmp);
6318 } else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6320 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6325 if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6326 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6327 dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6328 } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6330 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6336 if (nd->nd_mrep != NULL)
6337 mbuf_freem(nd->nd_mrep);
6338 *iomode = committed;
6339 if (nd->nd_repstat != 0 && error == 0)
6340 error = nd->nd_repstat;
6345 * The actual write RPC done to a DS.
6346 * This variant is called from a separate kernel process for mirrors.
6347 * Any short write is considered an IO error.
6350 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6351 nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6352 struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6353 struct ucred *cred, NFSPROC_T *p)
6356 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6357 int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6358 struct nfsrv_descript nfsd;
6359 struct nfsrv_descript *nd = &nfsd;
6360 struct nfssockreq *nrp;
6364 if (vers == 0 || vers == NFS_VER4) {
6365 nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6366 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6368 NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6370 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6371 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6373 nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6374 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6375 NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6376 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6378 txdr_hyper(io_off, tl);
6380 if (vers == NFS_VER3)
6381 *tl++ = txdr_unsigned(len);
6382 *tl++ = txdr_unsigned(*iomode);
6383 *tl = txdr_unsigned(len);
6385 /* Put data in mbuf chain. */
6386 nd->nd_mb->m_next = m;
6387 /* Set nd_mb and nd_bpos to end of data. */
6388 while (m->m_next != NULL)
6391 nd->nd_bpos = mtod(m, char *) + m->m_len;
6392 NFSCL_DEBUG(4, "nfsrpc_writedsmir: lastmb len=%d\n", m->m_len);
6394 nrp = dsp->nfsclds_sockp;
6396 /* If NULL, use the MDS socket. */
6397 nrp = &nmp->nm_sockreq;
6398 error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6399 NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6400 NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6404 if (nd->nd_repstat != 0)
6405 error = nd->nd_repstat;
6407 if (vers == NFS_VER3) {
6408 error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6410 NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6415 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6416 rlen = fxdr_unsigned(int, *tl++);
6417 NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6421 NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6425 commit = fxdr_unsigned(int, *tl++);
6428 * Return the lowest commitment level
6429 * obtained by any of the RPCs.
6431 if (committed == NFSWRITE_FILESYNC)
6433 else if (committed == NFSWRITE_DATASYNC &&
6434 commit == NFSWRITE_UNSTABLE)
6437 if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6438 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6439 dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6440 } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6442 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6447 if (nd->nd_mrep != NULL)
6448 mbuf_freem(nd->nd_mrep);
6449 *iomode = committed;
6450 if (nd->nd_repstat != 0 && error == 0)
6451 error = nd->nd_repstat;
6456 * Start up the thread that will execute nfsrpc_writedsmir().
6459 start_writedsmir(void *arg, int pending)
6461 struct nfsclwritedsdorpc *drpc;
6463 drpc = (struct nfsclwritedsdorpc *)arg;
6464 drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6465 &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6466 drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6469 NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6473 * Set up the write DS mirror call for the pNFS I/O thread.
6476 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6477 nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6478 struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6479 struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6486 drpc->iomode = *iomode;
6487 drpc->must_commit = *must_commit;
6488 drpc->stateidp = stateidp;
6495 drpc->minorvers = minorvers;
6500 if (nfs_pnfsiothreads != 0) {
6501 ret = nfs_pnfsio(start_writedsmir, drpc);
6502 NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6505 error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6506 dsp, off, len, fhp, m, vers, minorvers, cred, p);
6507 NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6512 * Free up the nfsclds structure.
6515 nfscl_freenfsclds(struct nfsclds *dsp)
6521 if (dsp->nfsclds_sockp != NULL) {
6522 NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6523 NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6524 free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6525 free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6527 NFSFREEMUTEX(&dsp->nfsclds_mtx);
6528 NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6529 for (i = 0; i < NFSV4_CBSLOTS; i++) {
6530 if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6532 dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6534 free(dsp, M_NFSCLDS);
6537 static enum nfsclds_state
6538 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6539 struct nfsclds **retdspp, uint32_t *sequencep)
6541 struct nfsclds *dsp;
6545 * Search the list of nfsclds structures for one with the same
6549 TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6550 if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6551 dsp->nfsclds_servownlen != 0 &&
6552 !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6553 dsp->nfsclds_servownlen) &&
6554 dsp->nfsclds_sess.nfsess_defunct == 0) {
6555 NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6556 TAILQ_FIRST(&nmp->nm_sess), dsp,
6557 dsp->nfsclds_flags);
6559 /* Get sequenceid# from first entry. */
6561 dsp->nfsclds_sess.nfsess_sequenceid;
6564 /* Server major id matches. */
6565 if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6567 return (NFSDSP_USETHISSESSION);
6573 return (NFSDSP_SEQTHISSESSION);
6574 return (NFSDSP_NOTFOUND);
6578 * NFS commit rpc to a NFSv4.1 DS.
6581 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6582 struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6586 struct nfsrv_descript nfsd, *nd = &nfsd;
6587 struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6588 struct nfssockreq *nrp;
6590 int attrflag, error;
6593 if (vers == 0 || vers == NFS_VER4) {
6594 nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6595 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6598 nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6599 fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6600 NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6602 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6603 txdr_hyper(offset, tl);
6605 *tl = txdr_unsigned(cnt);
6606 nrp = dsp->nfsclds_sockp;
6608 /* If NULL, use the MDS socket. */
6609 nrp = &nmp->nm_sockreq;
6610 error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6611 NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6612 NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6616 if (nd->nd_repstat == 0) {
6617 if (vers == NFS_VER3) {
6618 error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6620 NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6624 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6626 if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6627 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6628 error = NFSERR_STALEWRITEVERF;
6633 if (error == 0 && nd->nd_repstat != 0)
6634 error = nd->nd_repstat;
6635 mbuf_freem(nd->nd_mrep);
6640 * Start up the thread that will execute nfsrpc_commitds().
6643 start_commitds(void *arg, int pending)
6645 struct nfsclwritedsdorpc *drpc;
6647 drpc = (struct nfsclwritedsdorpc *)arg;
6648 drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6649 drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6652 NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6656 * Set up the commit DS mirror call for the pNFS I/O thread.
6659 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6660 struct nfsfh *fhp, int vers, int minorvers,
6661 struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6673 drpc->minorvers = minorvers;
6678 if (nfs_pnfsiothreads != 0) {
6679 ret = nfs_pnfsio(start_commitds, drpc);
6680 NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6683 error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6684 minorvers, cred, p);
6685 NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6690 * Set up the XDR arguments for the LayoutGet operation.
6693 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
6694 uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
6695 int layoutlen, int usecurstateid)
6699 NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
6701 *tl++ = newnfs_false; /* Don't signal availability. */
6702 *tl++ = txdr_unsigned(layouttype);
6703 *tl++ = txdr_unsigned(iomode);
6704 txdr_hyper(offset, tl);
6706 txdr_hyper(len, tl);
6708 txdr_hyper(minlen, tl);
6710 if (usecurstateid != 0) {
6711 /* Special stateid for Current stateid. */
6712 *tl++ = txdr_unsigned(1);
6717 *tl++ = txdr_unsigned(stateidp->seqid);
6718 NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
6719 *tl++ = stateidp->other[0];
6720 *tl++ = stateidp->other[1];
6721 *tl++ = stateidp->other[2];
6723 *tl = txdr_unsigned(layoutlen);
6727 * Parse the reply for a successful LayoutGet operation.
6730 nfsrv_parselayoutget(struct nfsrv_descript *nd, nfsv4stateid_t *stateidp,
6731 int *retonclosep, struct nfsclflayouthead *flhp)
6734 struct nfsclflayout *flp, *prevflp, *tflp;
6735 int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
6737 uint64_t retlen, off;
6743 NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
6747 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
6752 stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
6753 NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
6754 (int)stateidp->seqid);
6755 stateidp->other[0] = *tl++;
6756 stateidp->other[1] = *tl++;
6757 stateidp->other[2] = *tl++;
6758 cnt = fxdr_unsigned(int, *tl);
6759 NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
6760 if (cnt <= 0 || cnt > 10000) {
6761 /* Don't accept more than 10000 layouts in reply. */
6762 error = NFSERR_BADXDR;
6765 for (i = 0; i < cnt; i++) {
6766 /* Dissect to the layout type. */
6767 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
6769 off = fxdr_hyper(tl); tl += 2;
6770 retlen = fxdr_hyper(tl); tl += 2;
6771 iomode = fxdr_unsigned(int, *tl++);
6772 laytype = fxdr_unsigned(int, *tl);
6773 NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
6774 (uintmax_t)off, (uintmax_t)retlen, iomode);
6775 /* Ignore length of layout body for now. */
6776 if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
6777 /* Parse the File layout up to fhcnt. */
6778 NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
6779 NFSX_HYPER + NFSX_V4DEVICEID);
6780 fhcnt = fxdr_unsigned(int, *(tl + 4 +
6781 NFSX_V4DEVICEID / NFSX_UNSIGNED));
6782 NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
6783 if (fhcnt < 0 || fhcnt > 100) {
6784 /* Don't accept more than 100 file handles. */
6785 error = NFSERR_BADXDR;
6789 flp = malloc(sizeof(*flp) + fhcnt *
6790 sizeof(struct nfsfh *), M_NFSFLAYOUT,
6793 flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
6795 flp->nfsfl_flags = NFSFL_FILE;
6796 flp->nfsfl_fhcnt = 0;
6797 flp->nfsfl_devp = NULL;
6798 flp->nfsfl_off = off;
6799 if (flp->nfsfl_off + retlen < flp->nfsfl_off)
6800 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
6802 flp->nfsfl_end = flp->nfsfl_off + retlen;
6803 flp->nfsfl_iomode = iomode;
6804 if (gotiomode == -1)
6805 gotiomode = flp->nfsfl_iomode;
6806 /* Ignore layout body length for now. */
6807 NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
6808 tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
6809 flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
6810 NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
6811 flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
6812 flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
6813 NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
6814 flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
6815 for (j = 0; j < fhcnt; j++) {
6816 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6817 nfhlen = fxdr_unsigned(int, *tl);
6818 if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
6819 error = NFSERR_BADXDR;
6822 nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
6824 flp->nfsfl_fh[j] = nfhp;
6826 nfhp->nfh_len = nfhlen;
6827 NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
6828 NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
6830 } else if (laytype == NFSLAYOUT_FLEXFILE) {
6831 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
6833 mirrorcnt = fxdr_unsigned(int, *(tl + 2));
6834 NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
6835 if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
6836 error = NFSERR_BADXDR;
6839 flp = malloc(sizeof(*flp) + mirrorcnt *
6840 sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
6841 flp->nfsfl_flags = NFSFL_FLEXFILE;
6842 flp->nfsfl_mirrorcnt = mirrorcnt;
6843 for (j = 0; j < mirrorcnt; j++)
6844 flp->nfsfl_ffm[j].devp = NULL;
6845 flp->nfsfl_off = off;
6846 if (flp->nfsfl_off + retlen < flp->nfsfl_off)
6847 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
6849 flp->nfsfl_end = flp->nfsfl_off + retlen;
6850 flp->nfsfl_iomode = iomode;
6851 if (gotiomode == -1)
6852 gotiomode = flp->nfsfl_iomode;
6853 flp->nfsfl_stripeunit = fxdr_hyper(tl);
6854 NFSCL_DEBUG(4, "stripeunit=%ju\n",
6855 (uintmax_t)flp->nfsfl_stripeunit);
6856 for (j = 0; j < mirrorcnt; j++) {
6857 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6858 k = fxdr_unsigned(int, *tl);
6859 if (k < 1 || k > 128) {
6860 error = NFSERR_BADXDR;
6863 NFSCL_DEBUG(4, "servercnt=%d\n", k);
6864 for (l = 0; l < k; l++) {
6865 NFSM_DISSECT(tl, uint32_t *,
6866 NFSX_V4DEVICEID + NFSX_STATEID +
6869 /* Just use the first server. */
6871 flp->nfsfl_ffm[j].dev,
6873 tl += (NFSX_V4DEVICEID /
6876 flp->nfsfl_ffm[j].st.seqid =
6878 flp->nfsfl_ffm[j].st.other[0] =
6880 flp->nfsfl_ffm[j].st.other[1] =
6882 flp->nfsfl_ffm[j].st.other[2] =
6884 NFSCL_DEBUG(4, "st.seqid=%u "
6885 "st.o0=0x%x st.o1=0x%x "
6887 flp->nfsfl_ffm[j].st.seqid,
6888 flp->nfsfl_ffm[j].st.other[0],
6889 flp->nfsfl_ffm[j].st.other[1],
6890 flp->nfsfl_ffm[j].st.other[2]);
6892 tl += ((NFSX_V4DEVICEID +
6896 fhcnt = fxdr_unsigned(int, *tl);
6897 NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
6899 fhcnt > NFSDEV_MAXVERS) {
6900 error = NFSERR_BADXDR;
6903 for (m = 0; m < fhcnt; m++) {
6904 NFSM_DISSECT(tl, uint32_t *,
6906 nfhlen = fxdr_unsigned(int,
6908 NFSCL_DEBUG(4, "nfhlen=%d\n",
6910 if (nfhlen <= 0 || nfhlen >
6912 error = NFSERR_BADXDR;
6915 NFSM_DISSECT(cp, uint8_t *,
6916 NFSM_RNDUP(nfhlen));
6918 flp->nfsfl_ffm[j].fhcnt
6922 nfhlen - 1, M_NFSFH,
6924 flp->nfsfl_ffm[j].fh[m]
6926 nfhp->nfh_len = nfhlen;
6934 /* Now, get the ffsd_user/ffds_group. */
6935 error = nfsrv_parseug(nd, 0, &user,
6937 NFSCL_DEBUG(4, "after parseu=%d\n",
6940 error = nfsrv_parseug(nd, 1,
6941 &user, &grp, curthread);
6942 NFSCL_DEBUG(4, "aft parseg=%d\n",
6946 NFSCL_DEBUG(4, "user=%d group=%d\n",
6949 flp->nfsfl_ffm[j].user = user;
6950 flp->nfsfl_ffm[j].group = grp;
6952 "usr=%d grp=%d\n", user,
6957 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6958 flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
6959 flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
6960 NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
6961 flp->nfsfl_fflags, flp->nfsfl_statshint);
6963 error = NFSERR_BADXDR;
6966 if (flp->nfsfl_iomode == gotiomode) {
6967 /* Keep the list in increasing offset order. */
6968 tflp = LIST_FIRST(flhp);
6970 while (tflp != NULL &&
6971 tflp->nfsfl_off < flp->nfsfl_off) {
6973 tflp = LIST_NEXT(tflp, nfsfl_list);
6975 if (prevflp == NULL)
6976 LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
6978 LIST_INSERT_AFTER(prevflp, flp,
6980 NFSCL_DEBUG(4, "flp inserted\n");
6982 printf("nfscl_layoutget(): got wrong iomode\n");
6983 nfscl_freeflayout(flp);
6988 NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
6989 if (error != 0 && flp != NULL)
6990 nfscl_freeflayout(flp);
6995 * Parse a user/group digit string.
6998 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7002 char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7006 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7007 len = fxdr_unsigned(uint32_t, *tl);
7009 if (len > NFSV4_OPAQUELIMIT) {
7010 error = NFSERR_BADXDR;
7013 NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7016 *gidp = GID_NOGROUP;
7021 if (len > NFSV4_SMALLSTR)
7022 str = malloc(len + 1, M_TEMP, M_WAITOK);
7025 NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7026 NFSBCOPY(cp, str, len);
7028 NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7030 error = nfsv4_strtogid(nd, str, len, gidp);
7032 error = nfsv4_strtouid(nd, str, len, uidp);
7034 if (len > NFSV4_SMALLSTR)
7036 NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7041 * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7042 * so that it does both an Open and a Layoutget.
7045 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7046 int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7047 struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7048 struct ucred *cred, NFSPROC_T *p)
7050 struct nfscllayout *lyp;
7051 struct nfsclflayout *flp;
7052 struct nfsclflayouthead flh;
7053 int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7054 int layouttype, laystat;
7055 nfsv4stateid_t stateid;
7056 struct nfsclsession *tsep;
7059 if (NFSHASFLEXFILE(nmp))
7060 layouttype = NFSLAYOUT_FLEXFILE;
7062 layouttype = NFSLAYOUT_NFSV4_1_FILES;
7064 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7065 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7068 lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7070 NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7073 else if (flp != NULL)
7077 if ((lyp == NULL || flp == NULL) && recalled == 0) {
7079 tsep = nfsmnt_mdssession(nmp);
7080 layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7086 stateid.seqid = lyp->nfsly_stateid.seqid;
7087 stateid.other[0] = lyp->nfsly_stateid.other[0];
7088 stateid.other[1] = lyp->nfsly_stateid.other[1];
7089 stateid.other[2] = lyp->nfsly_stateid.other[2];
7091 error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7092 newfhp, newfhlen, mode, op, name, namelen,
7093 dpp, &stateid, usecurstateid, layouttype, layoutlen,
7094 &retonclose, &flh, &laystat, cred, p);
7095 NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7097 laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7098 &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7099 &islocked, cred, p);
7101 error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7102 mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7104 nfscl_rellayout(lyp, 1);
7105 else if (islocked == 1)
7106 nfscl_rellayout(lyp, 0);
7111 * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7112 * enabled, only for the CLAIM_NULL case. All other NFSv4 Opens are
7113 * handled by nfsrpc_openrpc().
7114 * For the case where op == NULL, dvp is the directory. When op != NULL, it
7118 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7119 int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7120 struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7121 nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7122 int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7123 int *laystatp, struct ucred *cred, NFSPROC_T *p)
7126 struct nfsrv_descript nfsd, *nd = &nfsd;
7127 struct nfscldeleg *ndp = NULL;
7128 struct nfsvattr nfsva;
7129 struct nfsclsession *tsep;
7130 uint32_t rflags, deleg;
7131 nfsattrbit_t attrbits;
7132 int error, ret, acesize, limitby, iomode;
7136 nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7138 NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7139 *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7140 *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7141 *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7142 tsep = nfsmnt_mdssession(nmp);
7143 *tl++ = tsep->nfsess_clientid.lval[0];
7144 *tl = tsep->nfsess_clientid.lval[1];
7145 nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7146 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7147 *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7148 *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7149 nfsm_strtom(nd, name, namelen);
7150 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7151 *tl = txdr_unsigned(NFSV4OP_GETATTR);
7152 NFSZERO_ATTRBIT(&attrbits);
7153 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7154 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7155 nfsrv_putattrbit(nd, &attrbits);
7156 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7157 *tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7158 if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7159 iomode = NFSLAYOUTIOMODE_RW;
7161 iomode = NFSLAYOUTIOMODE_READ;
7162 nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7163 layouttype, layoutlen, usecurstateid);
7164 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7165 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7168 NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7169 if (nd->nd_repstat != 0)
7170 *laystatp = nd->nd_repstat;
7171 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7172 /* ND_NOMOREDATA will be set if the Open operation failed. */
7173 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7175 op->nfso_stateid.seqid = *tl++;
7176 op->nfso_stateid.other[0] = *tl++;
7177 op->nfso_stateid.other[1] = *tl++;
7178 op->nfso_stateid.other[2] = *tl;
7179 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7180 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7183 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7184 deleg = fxdr_unsigned(u_int32_t, *tl);
7185 if (deleg == NFSV4OPEN_DELEGATEREAD ||
7186 deleg == NFSV4OPEN_DELEGATEWRITE) {
7187 if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7188 NFSCLFLAGS_FIRSTDELEG))
7189 op->nfso_own->nfsow_clp->nfsc_flags |=
7190 (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7191 ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7192 M_NFSCLDELEG, M_WAITOK);
7193 LIST_INIT(&ndp->nfsdl_owner);
7194 LIST_INIT(&ndp->nfsdl_lock);
7195 ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7196 ndp->nfsdl_fhlen = newfhlen;
7197 NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7198 newnfs_copyincred(cred, &ndp->nfsdl_cred);
7199 nfscl_lockinit(&ndp->nfsdl_rwlock);
7200 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7202 ndp->nfsdl_stateid.seqid = *tl++;
7203 ndp->nfsdl_stateid.other[0] = *tl++;
7204 ndp->nfsdl_stateid.other[1] = *tl++;
7205 ndp->nfsdl_stateid.other[2] = *tl++;
7206 ret = fxdr_unsigned(int, *tl);
7207 if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7208 ndp->nfsdl_flags = NFSCLDL_WRITE;
7210 * Indicates how much the file can grow.
7212 NFSM_DISSECT(tl, u_int32_t *,
7214 limitby = fxdr_unsigned(int, *tl++);
7216 case NFSV4OPEN_LIMITSIZE:
7217 ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7219 case NFSV4OPEN_LIMITBLOCKS:
7220 ndp->nfsdl_sizelimit =
7221 fxdr_unsigned(u_int64_t, *tl++);
7222 ndp->nfsdl_sizelimit *=
7223 fxdr_unsigned(u_int64_t, *tl);
7226 error = NFSERR_BADXDR;
7230 ndp->nfsdl_flags = NFSCLDL_READ;
7232 ndp->nfsdl_flags |= NFSCLDL_RECALL;
7233 error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7237 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
7238 error = NFSERR_BADXDR;
7241 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7242 nfscl_assumeposixlocks)
7243 op->nfso_posixlock = 1;
7245 op->nfso_posixlock = 0;
7246 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7247 /* If the 2nd element == NFS_OK, the Getattr succeeded. */
7249 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7250 NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7251 NULL, NULL, NULL, p, cred);
7255 ndp->nfsdl_change = nfsva.na_filerev;
7256 ndp->nfsdl_modtime = nfsva.na_mtime;
7257 ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7262 * At this point, the Open has succeeded, so set
7263 * nd_repstat = NFS_OK. If the Layoutget failed,
7264 * this function just won't return a layout.
7266 if (nd->nd_repstat == 0) {
7267 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7268 *laystatp = fxdr_unsigned(int, *++tl);
7269 if (*laystatp == 0) {
7270 error = nfsrv_parselayoutget(nd,
7271 stateidp, retonclosep, flhp);
7276 nd->nd_repstat = 0; /* Return 0 for Open. */
7279 if (nd->nd_repstat != 0 && error == 0)
7280 error = nd->nd_repstat;
7282 free(ndp, M_NFSCLDELEG);
7283 mbuf_freem(nd->nd_mrep);
7288 * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7289 * Used only for mounts with pNFS enabled.
7292 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7293 nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7294 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7295 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7296 int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7297 int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7298 struct nfsclflayouthead *flhp, int *laystatp)
7301 int error = 0, deleg, newone, ret, acesize, limitby;
7302 struct nfsrv_descript nfsd, *nd = &nfsd;
7303 struct nfsclopen *op;
7304 struct nfscldeleg *dp = NULL;
7307 struct nfsclsession *tsep;
7308 nfsattrbit_t attrbits;
7309 nfsv4stateid_t stateid;
7310 struct nfsmount *nmp;
7312 nmp = VFSTONFS(dvp->v_mount);
7320 if (namelen > NFS_MAXNAMLEN)
7321 return (ENAMETOOLONG);
7322 NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7324 * For V4, this is actually an Open op.
7326 NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7327 *tl++ = txdr_unsigned(owp->nfsow_seqid);
7328 *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7329 NFSV4OPEN_ACCESSREAD);
7330 *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7331 tsep = nfsmnt_mdssession(nmp);
7332 *tl++ = tsep->nfsess_clientid.lval[0];
7333 *tl = tsep->nfsess_clientid.lval[1];
7334 nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7335 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7336 *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7337 if ((fmode & O_EXCL) != 0) {
7338 if (NFSHASSESSPERSIST(nmp)) {
7339 /* Use GUARDED for persistent sessions. */
7340 *tl = txdr_unsigned(NFSCREATE_GUARDED);
7341 nfscl_fillsattr(nd, vap, dvp, 0, 0);
7343 /* Otherwise, use EXCLUSIVE4_1. */
7344 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7345 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7346 *tl++ = cverf.lval[0];
7347 *tl = cverf.lval[1];
7348 nfscl_fillsattr(nd, vap, dvp, 0, 0);
7351 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7352 nfscl_fillsattr(nd, vap, dvp, 0, 0);
7354 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7355 *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7356 nfsm_strtom(nd, name, namelen);
7357 /* Get the new file's handle and attributes, plus save the FH. */
7358 NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7359 *tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7360 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
7361 *tl = txdr_unsigned(NFSV4OP_GETATTR);
7362 NFSGETATTR_ATTRBIT(&attrbits);
7363 nfsrv_putattrbit(nd, &attrbits);
7364 /* Get the directory's post-op attributes. */
7365 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7366 *tl = txdr_unsigned(NFSV4OP_PUTFH);
7367 nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7368 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7369 *tl = txdr_unsigned(NFSV4OP_GETATTR);
7370 nfsrv_putattrbit(nd, &attrbits);
7371 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7372 *tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7373 *tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7374 nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7375 layouttype, layoutlen, usecurstateid);
7376 error = nfscl_request(nd, dvp, p, cred, dstuff);
7379 NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7381 if (nd->nd_repstat != 0)
7382 *laystatp = nd->nd_repstat;
7383 NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7384 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7385 NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7386 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7388 stateid.seqid = *tl++;
7389 stateid.other[0] = *tl++;
7390 stateid.other[1] = *tl++;
7391 stateid.other[2] = *tl;
7392 nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7393 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7394 deleg = fxdr_unsigned(int, *tl);
7395 if (deleg == NFSV4OPEN_DELEGATEREAD ||
7396 deleg == NFSV4OPEN_DELEGATEWRITE) {
7397 if (!(owp->nfsow_clp->nfsc_flags &
7398 NFSCLFLAGS_FIRSTDELEG))
7399 owp->nfsow_clp->nfsc_flags |=
7400 (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7401 dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7402 M_NFSCLDELEG, M_WAITOK);
7403 LIST_INIT(&dp->nfsdl_owner);
7404 LIST_INIT(&dp->nfsdl_lock);
7405 dp->nfsdl_clp = owp->nfsow_clp;
7406 newnfs_copyincred(cred, &dp->nfsdl_cred);
7407 nfscl_lockinit(&dp->nfsdl_rwlock);
7408 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7410 dp->nfsdl_stateid.seqid = *tl++;
7411 dp->nfsdl_stateid.other[0] = *tl++;
7412 dp->nfsdl_stateid.other[1] = *tl++;
7413 dp->nfsdl_stateid.other[2] = *tl++;
7414 ret = fxdr_unsigned(int, *tl);
7415 if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7416 dp->nfsdl_flags = NFSCLDL_WRITE;
7418 * Indicates how much the file can grow.
7420 NFSM_DISSECT(tl, u_int32_t *,
7422 limitby = fxdr_unsigned(int, *tl++);
7424 case NFSV4OPEN_LIMITSIZE:
7425 dp->nfsdl_sizelimit = fxdr_hyper(tl);
7427 case NFSV4OPEN_LIMITBLOCKS:
7428 dp->nfsdl_sizelimit =
7429 fxdr_unsigned(u_int64_t, *tl++);
7430 dp->nfsdl_sizelimit *=
7431 fxdr_unsigned(u_int64_t, *tl);
7434 error = NFSERR_BADXDR;
7438 dp->nfsdl_flags = NFSCLDL_READ;
7441 dp->nfsdl_flags |= NFSCLDL_RECALL;
7442 error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7446 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
7447 error = NFSERR_BADXDR;
7451 /* Now, we should have the status for the SaveFH. */
7452 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7454 NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7456 * Now, process the GetFH and Getattr for the newly
7457 * created file. nfscl_mtofh() will set
7458 * ND_NOMOREDATA if these weren't successful.
7460 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7461 NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7465 nd->nd_flag |= ND_NOMOREDATA;
7466 /* Now we have the PutFH and Getattr for the directory. */
7467 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7468 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7470 nd->nd_flag |= ND_NOMOREDATA;
7472 NFSM_DISSECT(tl, uint32_t *, 2 *
7475 nd->nd_flag |= ND_NOMOREDATA;
7478 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7479 /* Load the directory attributes. */
7480 error = nfsm_loadattr(nd, dnap);
7481 NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7485 if (dp != NULL && *attrflagp != 0) {
7486 dp->nfsdl_change = nnap->na_filerev;
7487 dp->nfsdl_modtime = nnap->na_mtime;
7488 dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7491 * We can now complete the Open state.
7495 dp->nfsdl_fhlen = nfhp->nfh_len;
7496 NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7500 * Get an Open structure that will be
7501 * attached to the OpenOwner, acquired already.
7503 error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7504 (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7505 cred, p, NULL, &op, &newone, NULL, 0);
7508 op->nfso_stateid = stateid;
7509 newnfs_copyincred(cred, &op->nfso_cred);
7511 nfscl_openrelease(nmp, op, error, newone);
7514 /* Now, handle the RestoreFH and LayoutGet. */
7515 if (nd->nd_repstat == 0) {
7516 NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7517 *laystatp = fxdr_unsigned(int, *(tl + 3));
7518 if (*laystatp == 0) {
7519 error = nfsrv_parselayoutget(nd,
7520 stateidp, retonclosep, flhp);
7524 NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7530 if (nd->nd_repstat != 0 && error == 0)
7531 error = nd->nd_repstat;
7532 if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7533 nfscl_initiate_recovery(owp->nfsow_clp);
7535 NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7539 free(dp, M_NFSCLDELEG);
7540 mbuf_freem(nd->nd_mrep);
7545 * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7548 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7549 nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7550 struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7551 struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7552 int *dattrflagp, void *dstuff, int *unlockedp)
7554 struct nfscllayout *lyp;
7555 struct nfsclflayouthead flh;
7557 struct nfsclsession *tsep;
7558 struct nfsmount *nmp;
7559 nfsv4stateid_t stateid;
7560 int error, layoutlen, layouttype, retonclose, laystat;
7563 nmp = VFSTONFS(dvp->v_mount);
7564 if (NFSHASFLEXFILE(nmp))
7565 layouttype = NFSLAYOUT_FLEXFILE;
7567 layouttype = NFSLAYOUT_NFSV4_1_FILES;
7569 tsep = nfsmnt_mdssession(nmp);
7570 layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
7571 error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
7572 owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
7573 dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
7575 NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
7580 laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
7581 nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
7582 layouttype, laystat, NULL, cred, p);
7584 laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
7585 retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
7588 nfscl_rellayout(lyp, 0);
7593 * Process the results of a layoutget() operation.
7596 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
7597 int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
7598 struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
7599 int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
7601 struct nfsclflayout *tflp;
7602 struct nfscldevinfo *dip;
7606 if (laystat == NFSERR_UNKNLAYOUTTYPE) {
7608 if (!NFSHASFLEXFILE(nmp)) {
7609 /* Switch to using Flex File Layout. */
7610 nmp->nm_state |= NFSSTA_FLEXFILE;
7611 } else if (layouttype == NFSLAYOUT_FLEXFILE) {
7613 NFSCL_DEBUG(1, "disable PNFS\n");
7614 nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
7619 NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
7620 LIST_FOREACH(tflp, flhp, nfsfl_list) {
7621 if (layouttype == NFSLAYOUT_FLEXFILE)
7622 mirrorcnt = tflp->nfsfl_mirrorcnt;
7625 for (i = 0; i < mirrorcnt; i++) {
7626 laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
7627 NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
7629 if (layouttype == NFSLAYOUT_FLEXFILE)
7630 dev = tflp->nfsfl_ffm[i].dev;
7632 dev = tflp->nfsfl_dev;
7633 laystat = nfsrpc_getdeviceinfo(nmp, dev,
7634 layouttype, notifybit, &dip, cred,
7636 NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
7640 laystat = nfscl_adddevinfo(nmp, dip, i,
7643 printf("nfsrpc_layoutgetresout"
7652 * nfscl_layout() always returns with the nfsly_lock
7653 * set to a refcnt (shared lock).
7654 * Passing in dvp is sufficient, since it is only used to
7655 * get the fsid for the file system.
7657 laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
7658 layouttype, retonclose, flhp, lypp, cred, p);
7659 NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
7661 if (laystat == 0 && islockedp != NULL)