1191783Srmacklem/*- 2191783Srmacklem * Copyright (c) 1989, 1993 3191783Srmacklem * The Regents of the University of California. All rights reserved. 4191783Srmacklem * 5191783Srmacklem * This code is derived from software contributed to Berkeley by 6191783Srmacklem * Rick Macklem at The University of Guelph. 7191783Srmacklem * 8191783Srmacklem * Redistribution and use in source and binary forms, with or without 9191783Srmacklem * modification, are permitted provided that the following conditions 10191783Srmacklem * are met: 11191783Srmacklem * 1. Redistributions of source code must retain the above copyright 12191783Srmacklem * notice, this list of conditions and the following disclaimer. 13191783Srmacklem * 2. Redistributions in binary form must reproduce the above copyright 14191783Srmacklem * notice, this list of conditions and the following disclaimer in the 15191783Srmacklem * documentation and/or other materials provided with the distribution. 16191783Srmacklem * 4. Neither the name of the University nor the names of its contributors 17191783Srmacklem * may be used to endorse or promote products derived from this software 18191783Srmacklem * without specific prior written permission. 19191783Srmacklem * 20191783Srmacklem * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21191783Srmacklem * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22191783Srmacklem * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23191783Srmacklem * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24191783Srmacklem * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25191783Srmacklem * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26191783Srmacklem * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27191783Srmacklem * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28191783Srmacklem * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29191783Srmacklem * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30191783Srmacklem * SUCH DAMAGE. 31191783Srmacklem * 32191783Srmacklem */ 33191783Srmacklem 34191783Srmacklem#include <sys/cdefs.h> 35191783Srmacklem__FBSDID("$FreeBSD$"); 36191783Srmacklem 37224778Srwatson#include <sys/capability.h> 38224778Srwatson 39191783Srmacklem/* 40191783Srmacklem * Functions that perform the vfs operations required by the routines in 41191783Srmacklem * nfsd_serv.c. It is hoped that this change will make the server more 42191783Srmacklem * portable. 43191783Srmacklem */ 44191783Srmacklem 45191783Srmacklem#include <fs/nfs/nfsport.h> 46214255Srmacklem#include <sys/hash.h> 47191783Srmacklem#include <sys/sysctl.h> 48192503Srmacklem#include <nlm/nlm_prot.h> 49192503Srmacklem#include <nlm/nlm.h> 50191783Srmacklem 51219028SnetchildFEATURE(nfsd, "NFSv4 server"); 52219028Snetchild 53191783Srmacklemextern u_int32_t newnfs_true, newnfs_false, newnfs_xdrneg1; 54191783Srmacklemextern int nfsrv_useacl; 55191783Srmacklemextern int newnfs_numnfsd; 56191783Srmacklemextern struct mount nfsv4root_mnt; 57191783Srmacklemextern struct nfsrv_stablefirst nfsrv_stablefirst; 58191783Srmacklemextern void (*nfsd_call_servertimer)(void); 59220530Srmacklemextern SVCPOOL *nfsrvd_pool; 60243738Srmacklemextern struct nfsv4lock nfsd_suspend_lock; 61191783Srmacklemstruct vfsoptlist nfsv4root_opt, nfsv4root_newopt; 62191783SrmacklemNFSDLOCKMUTEX; 63255532Srmacklemstruct nfsrchash_bucket nfsrchash_table[NFSRVCACHE_HASHSIZE]; 64261067Smavstruct nfsrchash_bucket nfsrcahash_table[NFSRVCACHE_HASHSIZE]; 65255532Srmacklemstruct mtx nfsrc_udpmtx; 66191783Srmacklemstruct mtx nfs_v4root_mutex; 67191783Srmacklemstruct nfsrvfh nfs_rootfh, nfs_pubfh; 68191783Srmacklemint nfs_pubfhset = 0, nfs_rootfhset = 0; 69217432Srmacklemstruct proc *nfsd_master_proc = NULL; 70217432Srmacklemstatic pid_t nfsd_master_pid = (pid_t)-1; 71217432Srmacklemstatic char nfsd_master_comm[MAXCOMLEN + 1]; 72217432Srmacklemstatic struct timeval nfsd_master_start; 73192503Srmacklemstatic uint32_t nfsv4_sysid = 0; 74191783Srmacklem 75192255Srmacklemstatic int nfssvc_srvcall(struct thread *, struct nfssvc_args *, 76192255Srmacklem struct ucred *); 77191783Srmacklem 78220645Srmacklemint nfsrv_enable_crossmntpt = 1; 79191783Srmacklemstatic int nfs_commit_blks; 80191783Srmacklemstatic int nfs_commit_miss; 81191783Srmacklemextern int nfsrv_issuedelegs; 82192255Srmacklemextern int nfsrv_dolocallocks; 83265724Srmacklemextern int nfsd_enable_stringtouid; 84192255Srmacklem 85221615SrmacklemSYSCTL_NODE(_vfs, OID_AUTO, nfsd, CTLFLAG_RW, 0, "New NFS server"); 86221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, mirrormnt, CTLFLAG_RW, 87220645Srmacklem &nfsrv_enable_crossmntpt, 0, "Enable nfsd to cross mount points"); 88221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_blks, CTLFLAG_RW, &nfs_commit_blks, 89192255Srmacklem 0, ""); 90221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_miss, CTLFLAG_RW, &nfs_commit_miss, 91192255Srmacklem 0, ""); 92221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, issue_delegations, CTLFLAG_RW, 93192255Srmacklem &nfsrv_issuedelegs, 0, "Enable nfsd to issue delegations"); 94221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, enable_locallocks, CTLFLAG_RW, 95192255Srmacklem &nfsrv_dolocallocks, 0, "Enable nfsd to acquire local locks on files"); 96265724SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, enable_stringtouid, CTLFLAG_RW, 97265724Srmacklem &nfsd_enable_stringtouid, 0, "Enable nfsd to accept numeric owner_names"); 98191783Srmacklem 99229617Sjhb#define MAX_REORDERED_RPC 16 100229617Sjhb#define NUM_HEURISTIC 1031 101191783Srmacklem#define NHUSE_INIT 64 102191783Srmacklem#define NHUSE_INC 16 103191783Srmacklem#define NHUSE_MAX 2048 104191783Srmacklem 105191783Srmacklemstatic struct nfsheur { 106191783Srmacklem struct vnode *nh_vp; /* vp to match (unreferenced pointer) */ 107229617Sjhb off_t nh_nextoff; /* next offset for sequential detection */ 108191783Srmacklem int nh_use; /* use count for selection */ 109191783Srmacklem int nh_seqcount; /* heuristic */ 110191783Srmacklem} nfsheur[NUM_HEURISTIC]; 111191783Srmacklem 112191783Srmacklem 113191783Srmacklem/* 114229617Sjhb * Heuristic to detect sequential operation. 115229617Sjhb */ 116229617Sjhbstatic struct nfsheur * 117229617Sjhbnfsrv_sequential_heuristic(struct uio *uio, struct vnode *vp) 118229617Sjhb{ 119229617Sjhb struct nfsheur *nh; 120229617Sjhb int hi, try; 121229617Sjhb 122229617Sjhb /* Locate best candidate. */ 123229617Sjhb try = 32; 124229617Sjhb hi = ((int)(vm_offset_t)vp / sizeof(struct vnode)) % NUM_HEURISTIC; 125229617Sjhb nh = &nfsheur[hi]; 126229617Sjhb while (try--) { 127229617Sjhb if (nfsheur[hi].nh_vp == vp) { 128229617Sjhb nh = &nfsheur[hi]; 129229617Sjhb break; 130229617Sjhb } 131229617Sjhb if (nfsheur[hi].nh_use > 0) 132229617Sjhb --nfsheur[hi].nh_use; 133229617Sjhb hi = (hi + 1) % NUM_HEURISTIC; 134229617Sjhb if (nfsheur[hi].nh_use < nh->nh_use) 135229617Sjhb nh = &nfsheur[hi]; 136229617Sjhb } 137229617Sjhb 138229617Sjhb /* Initialize hint if this is a new file. */ 139229617Sjhb if (nh->nh_vp != vp) { 140229617Sjhb nh->nh_vp = vp; 141229617Sjhb nh->nh_nextoff = uio->uio_offset; 142229617Sjhb nh->nh_use = NHUSE_INIT; 143229617Sjhb if (uio->uio_offset == 0) 144229617Sjhb nh->nh_seqcount = 4; 145229617Sjhb else 146229617Sjhb nh->nh_seqcount = 1; 147229617Sjhb } 148229617Sjhb 149229617Sjhb /* Calculate heuristic. */ 150229617Sjhb if ((uio->uio_offset == 0 && nh->nh_seqcount > 0) || 151229617Sjhb uio->uio_offset == nh->nh_nextoff) { 152229617Sjhb /* See comments in vfs_vnops.c:sequential_heuristic(). */ 153229617Sjhb nh->nh_seqcount += howmany(uio->uio_resid, 16384); 154229617Sjhb if (nh->nh_seqcount > IO_SEQMAX) 155229617Sjhb nh->nh_seqcount = IO_SEQMAX; 156229617Sjhb } else if (qabs(uio->uio_offset - nh->nh_nextoff) <= MAX_REORDERED_RPC * 157229617Sjhb imax(vp->v_mount->mnt_stat.f_iosize, uio->uio_resid)) { 158229617Sjhb /* Probably a reordered RPC, leave seqcount alone. */ 159229617Sjhb } else if (nh->nh_seqcount > 1) { 160229617Sjhb nh->nh_seqcount /= 2; 161229617Sjhb } else { 162229617Sjhb nh->nh_seqcount = 0; 163229617Sjhb } 164229617Sjhb nh->nh_use += NHUSE_INC; 165229617Sjhb if (nh->nh_use > NHUSE_MAX) 166229617Sjhb nh->nh_use = NHUSE_MAX; 167229617Sjhb return (nh); 168229617Sjhb} 169229617Sjhb 170229617Sjhb/* 171191783Srmacklem * Get attributes into nfsvattr structure. 172191783Srmacklem */ 173191783Srmacklemint 174191783Srmacklemnfsvno_getattr(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred, 175216693Srmacklem struct thread *p, int vpislocked) 176191783Srmacklem{ 177191783Srmacklem int error, lockedit = 0; 178191783Srmacklem 179216693Srmacklem if (vpislocked == 0) { 180216693Srmacklem /* 181216693Srmacklem * When vpislocked == 0, the vnode is either exclusively 182216693Srmacklem * locked by this thread or not locked by this thread. 183216693Srmacklem * As such, shared lock it, if not exclusively locked. 184216693Srmacklem */ 185224083Szack if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) { 186216693Srmacklem lockedit = 1; 187224081Szack NFSVOPLOCK(vp, LK_SHARED | LK_RETRY); 188216693Srmacklem } 189191783Srmacklem } 190191783Srmacklem error = VOP_GETATTR(vp, &nvap->na_vattr, cred); 191216693Srmacklem if (lockedit != 0) 192224082Szack NFSVOPUNLOCK(vp, 0); 193224086Szack 194224086Szack NFSEXITCODE(error); 195191783Srmacklem return (error); 196191783Srmacklem} 197191783Srmacklem 198191783Srmacklem/* 199191783Srmacklem * Get a file handle for a vnode. 200191783Srmacklem */ 201191783Srmacklemint 202191783Srmacklemnfsvno_getfh(struct vnode *vp, fhandle_t *fhp, struct thread *p) 203191783Srmacklem{ 204191783Srmacklem int error; 205191783Srmacklem 206191783Srmacklem NFSBZERO((caddr_t)fhp, sizeof(fhandle_t)); 207191783Srmacklem fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 208191783Srmacklem error = VOP_VPTOFH(vp, &fhp->fh_fid); 209224086Szack 210224086Szack NFSEXITCODE(error); 211191783Srmacklem return (error); 212191783Srmacklem} 213191783Srmacklem 214191783Srmacklem/* 215191783Srmacklem * Perform access checking for vnodes obtained from file handles that would 216191783Srmacklem * refer to files already opened by a Unix client. You cannot just use 217200999Srmacklem * vn_writechk() and VOP_ACCESSX() for two reasons. 218200999Srmacklem * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write 219200999Srmacklem * case. 220191783Srmacklem * 2 - The owner is to be given access irrespective of mode bits for some 221191783Srmacklem * operations, so that processes that chmod after opening a file don't 222191783Srmacklem * break. 223191783Srmacklem */ 224191783Srmacklemint 225200999Srmacklemnfsvno_accchk(struct vnode *vp, accmode_t accmode, struct ucred *cred, 226200999Srmacklem struct nfsexstuff *exp, struct thread *p, int override, int vpislocked, 227200999Srmacklem u_int32_t *supportedtypep) 228191783Srmacklem{ 229191783Srmacklem struct vattr vattr; 230191783Srmacklem int error = 0, getret = 0; 231191783Srmacklem 232216893Srmacklem if (vpislocked == 0) { 233224086Szack if (NFSVOPLOCK(vp, LK_SHARED) != 0) { 234224086Szack error = EPERM; 235224086Szack goto out; 236224086Szack } 237216893Srmacklem } 238191783Srmacklem if (accmode & VWRITE) { 239191783Srmacklem /* Just vn_writechk() changed to check rdonly */ 240191783Srmacklem /* 241191783Srmacklem * Disallow write attempts on read-only file systems; 242191783Srmacklem * unless the file is a socket or a block or character 243191783Srmacklem * device resident on the file system. 244191783Srmacklem */ 245191783Srmacklem if (NFSVNO_EXRDONLY(exp) || 246200999Srmacklem (vp->v_mount->mnt_flag & MNT_RDONLY)) { 247191783Srmacklem switch (vp->v_type) { 248191783Srmacklem case VREG: 249191783Srmacklem case VDIR: 250191783Srmacklem case VLNK: 251216893Srmacklem error = EROFS; 252191783Srmacklem default: 253191783Srmacklem break; 254191783Srmacklem } 255191783Srmacklem } 256191783Srmacklem /* 257191783Srmacklem * If there's shared text associated with 258191783Srmacklem * the inode, try to free it up once. If 259191783Srmacklem * we fail, we can't allow writing. 260191783Srmacklem */ 261244658Skib if (VOP_IS_TEXT(vp) && error == 0) 262216893Srmacklem error = ETXTBSY; 263191783Srmacklem } 264216893Srmacklem if (error != 0) { 265216893Srmacklem if (vpislocked == 0) 266224082Szack NFSVOPUNLOCK(vp, 0); 267224086Szack goto out; 268216893Srmacklem } 269191783Srmacklem 270191783Srmacklem /* 271191783Srmacklem * Should the override still be applied when ACLs are enabled? 272191783Srmacklem */ 273200999Srmacklem error = VOP_ACCESSX(vp, accmode, cred, p); 274200999Srmacklem if (error != 0 && (accmode & (VDELETE | VDELETE_CHILD))) { 275200999Srmacklem /* 276200999Srmacklem * Try again with VEXPLICIT_DENY, to see if the test for 277200999Srmacklem * deletion is supported. 278200999Srmacklem */ 279200999Srmacklem error = VOP_ACCESSX(vp, accmode | VEXPLICIT_DENY, cred, p); 280200999Srmacklem if (error == 0) { 281200999Srmacklem if (vp->v_type == VDIR) { 282200999Srmacklem accmode &= ~(VDELETE | VDELETE_CHILD); 283200999Srmacklem accmode |= VWRITE; 284200999Srmacklem error = VOP_ACCESSX(vp, accmode, cred, p); 285200999Srmacklem } else if (supportedtypep != NULL) { 286200999Srmacklem *supportedtypep &= ~NFSACCESS_DELETE; 287200999Srmacklem } 288200999Srmacklem } 289200999Srmacklem } 290191783Srmacklem 291191783Srmacklem /* 292191783Srmacklem * Allow certain operations for the owner (reads and writes 293191783Srmacklem * on files that are already open). 294191783Srmacklem */ 295191783Srmacklem if (override != NFSACCCHK_NOOVERRIDE && 296191783Srmacklem (error == EPERM || error == EACCES)) { 297191783Srmacklem if (cred->cr_uid == 0 && (override & NFSACCCHK_ALLOWROOT)) 298191783Srmacklem error = 0; 299191783Srmacklem else if (override & NFSACCCHK_ALLOWOWNER) { 300191783Srmacklem getret = VOP_GETATTR(vp, &vattr, cred); 301191783Srmacklem if (getret == 0 && cred->cr_uid == vattr.va_uid) 302191783Srmacklem error = 0; 303191783Srmacklem } 304191783Srmacklem } 305191783Srmacklem if (vpislocked == 0) 306224082Szack NFSVOPUNLOCK(vp, 0); 307224086Szack 308224086Szackout: 309224086Szack NFSEXITCODE(error); 310191783Srmacklem return (error); 311191783Srmacklem} 312191783Srmacklem 313191783Srmacklem/* 314191783Srmacklem * Set attribute(s) vnop. 315191783Srmacklem */ 316191783Srmacklemint 317191783Srmacklemnfsvno_setattr(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred, 318191783Srmacklem struct thread *p, struct nfsexstuff *exp) 319191783Srmacklem{ 320191783Srmacklem int error; 321191783Srmacklem 322191783Srmacklem error = VOP_SETATTR(vp, &nvap->na_vattr, cred); 323224086Szack NFSEXITCODE(error); 324191783Srmacklem return (error); 325191783Srmacklem} 326191783Srmacklem 327191783Srmacklem/* 328232018Srmacklem * Set up nameidata for a lookup() call and do it. 329191783Srmacklem */ 330191783Srmacklemint 331191783Srmacklemnfsvno_namei(struct nfsrv_descript *nd, struct nameidata *ndp, 332191783Srmacklem struct vnode *dp, int islocked, struct nfsexstuff *exp, struct thread *p, 333191783Srmacklem struct vnode **retdirp) 334191783Srmacklem{ 335191783Srmacklem struct componentname *cnp = &ndp->ni_cnd; 336191783Srmacklem int i; 337191783Srmacklem struct iovec aiov; 338191783Srmacklem struct uio auio; 339191783Srmacklem int lockleaf = (cnp->cn_flags & LOCKLEAF) != 0, linklen; 340191783Srmacklem int error = 0, crossmnt; 341191783Srmacklem char *cp; 342191783Srmacklem 343191783Srmacklem *retdirp = NULL; 344191783Srmacklem cnp->cn_nameptr = cnp->cn_pnbuf; 345225356Srmacklem ndp->ni_strictrelative = 0; 346191783Srmacklem /* 347191783Srmacklem * Extract and set starting directory. 348191783Srmacklem */ 349191783Srmacklem if (dp->v_type != VDIR) { 350191783Srmacklem if (islocked) 351191783Srmacklem vput(dp); 352191783Srmacklem else 353191783Srmacklem vrele(dp); 354191783Srmacklem nfsvno_relpathbuf(ndp); 355224086Szack error = ENOTDIR; 356224086Szack goto out1; 357191783Srmacklem } 358191783Srmacklem if (islocked) 359224080Szack NFSVOPUNLOCK(dp, 0); 360191783Srmacklem VREF(dp); 361191783Srmacklem *retdirp = dp; 362191783Srmacklem if (NFSVNO_EXRDONLY(exp)) 363191783Srmacklem cnp->cn_flags |= RDONLY; 364191783Srmacklem ndp->ni_segflg = UIO_SYSSPACE; 365191783Srmacklem crossmnt = 1; 366191783Srmacklem 367191783Srmacklem if (nd->nd_flag & ND_PUBLOOKUP) { 368191783Srmacklem ndp->ni_loopcnt = 0; 369191783Srmacklem if (cnp->cn_pnbuf[0] == '/') { 370191783Srmacklem vrele(dp); 371191783Srmacklem /* 372191783Srmacklem * Check for degenerate pathnames here, since lookup() 373191783Srmacklem * panics on them. 374191783Srmacklem */ 375191783Srmacklem for (i = 1; i < ndp->ni_pathlen; i++) 376191783Srmacklem if (cnp->cn_pnbuf[i] != '/') 377191783Srmacklem break; 378191783Srmacklem if (i == ndp->ni_pathlen) { 379191783Srmacklem error = NFSERR_ACCES; 380191783Srmacklem goto out; 381191783Srmacklem } 382191783Srmacklem dp = rootvnode; 383191783Srmacklem VREF(dp); 384191783Srmacklem } 385220645Srmacklem } else if ((nfsrv_enable_crossmntpt == 0 && NFSVNO_EXPORTED(exp)) || 386191783Srmacklem (nd->nd_flag & ND_NFSV4) == 0) { 387191783Srmacklem /* 388191783Srmacklem * Only cross mount points for NFSv4 when doing a 389191783Srmacklem * mount while traversing the file system above 390220645Srmacklem * the mount point, unless nfsrv_enable_crossmntpt is set. 391191783Srmacklem */ 392191783Srmacklem cnp->cn_flags |= NOCROSSMOUNT; 393191783Srmacklem crossmnt = 0; 394191783Srmacklem } 395191783Srmacklem 396191783Srmacklem /* 397191783Srmacklem * Initialize for scan, set ni_startdir and bump ref on dp again 398191783Srmacklem * becuase lookup() will dereference ni_startdir. 399191783Srmacklem */ 400191783Srmacklem 401191783Srmacklem cnp->cn_thread = p; 402191783Srmacklem ndp->ni_startdir = dp; 403191783Srmacklem ndp->ni_rootdir = rootvnode; 404232683Srmacklem ndp->ni_topdir = NULL; 405191783Srmacklem 406191783Srmacklem if (!lockleaf) 407191783Srmacklem cnp->cn_flags |= LOCKLEAF; 408191783Srmacklem for (;;) { 409191783Srmacklem cnp->cn_nameptr = cnp->cn_pnbuf; 410191783Srmacklem /* 411191783Srmacklem * Call lookup() to do the real work. If an error occurs, 412191783Srmacklem * ndp->ni_vp and ni_dvp are left uninitialized or NULL and 413191783Srmacklem * we do not have to dereference anything before returning. 414191783Srmacklem * In either case ni_startdir will be dereferenced and NULLed 415191783Srmacklem * out. 416191783Srmacklem */ 417191783Srmacklem error = lookup(ndp); 418191783Srmacklem if (error) 419191783Srmacklem break; 420191783Srmacklem 421191783Srmacklem /* 422191783Srmacklem * Check for encountering a symbolic link. Trivial 423191783Srmacklem * termination occurs if no symlink encountered. 424191783Srmacklem */ 425191783Srmacklem if ((cnp->cn_flags & ISSYMLINK) == 0) { 426191783Srmacklem if ((cnp->cn_flags & (SAVENAME | SAVESTART)) == 0) 427191783Srmacklem nfsvno_relpathbuf(ndp); 428191783Srmacklem if (ndp->ni_vp && !lockleaf) 429224080Szack NFSVOPUNLOCK(ndp->ni_vp, 0); 430191783Srmacklem break; 431191783Srmacklem } 432191783Srmacklem 433191783Srmacklem /* 434191783Srmacklem * Validate symlink 435191783Srmacklem */ 436191783Srmacklem if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1) 437224080Szack NFSVOPUNLOCK(ndp->ni_dvp, 0); 438191783Srmacklem if (!(nd->nd_flag & ND_PUBLOOKUP)) { 439191783Srmacklem error = EINVAL; 440191783Srmacklem goto badlink2; 441191783Srmacklem } 442191783Srmacklem 443191783Srmacklem if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 444191783Srmacklem error = ELOOP; 445191783Srmacklem goto badlink2; 446191783Srmacklem } 447191783Srmacklem if (ndp->ni_pathlen > 1) 448191783Srmacklem cp = uma_zalloc(namei_zone, M_WAITOK); 449191783Srmacklem else 450191783Srmacklem cp = cnp->cn_pnbuf; 451191783Srmacklem aiov.iov_base = cp; 452191783Srmacklem aiov.iov_len = MAXPATHLEN; 453191783Srmacklem auio.uio_iov = &aiov; 454191783Srmacklem auio.uio_iovcnt = 1; 455191783Srmacklem auio.uio_offset = 0; 456191783Srmacklem auio.uio_rw = UIO_READ; 457191783Srmacklem auio.uio_segflg = UIO_SYSSPACE; 458191783Srmacklem auio.uio_td = NULL; 459191783Srmacklem auio.uio_resid = MAXPATHLEN; 460191783Srmacklem error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 461191783Srmacklem if (error) { 462191783Srmacklem badlink1: 463191783Srmacklem if (ndp->ni_pathlen > 1) 464191783Srmacklem uma_zfree(namei_zone, cp); 465191783Srmacklem badlink2: 466191783Srmacklem vrele(ndp->ni_dvp); 467191783Srmacklem vput(ndp->ni_vp); 468191783Srmacklem break; 469191783Srmacklem } 470191783Srmacklem linklen = MAXPATHLEN - auio.uio_resid; 471191783Srmacklem if (linklen == 0) { 472191783Srmacklem error = ENOENT; 473191783Srmacklem goto badlink1; 474191783Srmacklem } 475191783Srmacklem if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 476191783Srmacklem error = ENAMETOOLONG; 477191783Srmacklem goto badlink1; 478191783Srmacklem } 479191783Srmacklem 480191783Srmacklem /* 481191783Srmacklem * Adjust or replace path 482191783Srmacklem */ 483191783Srmacklem if (ndp->ni_pathlen > 1) { 484191783Srmacklem NFSBCOPY(ndp->ni_next, cp + linklen, ndp->ni_pathlen); 485191783Srmacklem uma_zfree(namei_zone, cnp->cn_pnbuf); 486191783Srmacklem cnp->cn_pnbuf = cp; 487191783Srmacklem } else 488191783Srmacklem cnp->cn_pnbuf[linklen] = '\0'; 489191783Srmacklem ndp->ni_pathlen += linklen; 490191783Srmacklem 491191783Srmacklem /* 492191783Srmacklem * Cleanup refs for next loop and check if root directory 493191783Srmacklem * should replace current directory. Normally ni_dvp 494191783Srmacklem * becomes the new base directory and is cleaned up when 495191783Srmacklem * we loop. Explicitly null pointers after invalidation 496191783Srmacklem * to clarify operation. 497191783Srmacklem */ 498191783Srmacklem vput(ndp->ni_vp); 499191783Srmacklem ndp->ni_vp = NULL; 500191783Srmacklem 501191783Srmacklem if (cnp->cn_pnbuf[0] == '/') { 502191783Srmacklem vrele(ndp->ni_dvp); 503191783Srmacklem ndp->ni_dvp = ndp->ni_rootdir; 504191783Srmacklem VREF(ndp->ni_dvp); 505191783Srmacklem } 506191783Srmacklem ndp->ni_startdir = ndp->ni_dvp; 507191783Srmacklem ndp->ni_dvp = NULL; 508191783Srmacklem } 509191783Srmacklem if (!lockleaf) 510191783Srmacklem cnp->cn_flags &= ~LOCKLEAF; 511191783Srmacklem 512191783Srmacklemout: 513191783Srmacklem if (error) { 514191783Srmacklem uma_zfree(namei_zone, cnp->cn_pnbuf); 515191783Srmacklem ndp->ni_vp = NULL; 516191783Srmacklem ndp->ni_dvp = NULL; 517191783Srmacklem ndp->ni_startdir = NULL; 518191783Srmacklem cnp->cn_flags &= ~HASBUF; 519191783Srmacklem } else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) { 520191783Srmacklem ndp->ni_dvp = NULL; 521191783Srmacklem } 522224086Szack 523224086Szackout1: 524224086Szack NFSEXITCODE2(error, nd); 525191783Srmacklem return (error); 526191783Srmacklem} 527191783Srmacklem 528191783Srmacklem/* 529191783Srmacklem * Set up a pathname buffer and return a pointer to it and, optionally 530191783Srmacklem * set a hash pointer. 531191783Srmacklem */ 532191783Srmacklemvoid 533191783Srmacklemnfsvno_setpathbuf(struct nameidata *ndp, char **bufpp, u_long **hashpp) 534191783Srmacklem{ 535191783Srmacklem struct componentname *cnp = &ndp->ni_cnd; 536191783Srmacklem 537191783Srmacklem cnp->cn_flags |= (NOMACCHECK | HASBUF); 538191783Srmacklem cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK); 539191783Srmacklem if (hashpp != NULL) 540191783Srmacklem *hashpp = NULL; 541191783Srmacklem *bufpp = cnp->cn_pnbuf; 542191783Srmacklem} 543191783Srmacklem 544191783Srmacklem/* 545191783Srmacklem * Release the above path buffer, if not released by nfsvno_namei(). 546191783Srmacklem */ 547191783Srmacklemvoid 548191783Srmacklemnfsvno_relpathbuf(struct nameidata *ndp) 549191783Srmacklem{ 550191783Srmacklem 551191783Srmacklem if ((ndp->ni_cnd.cn_flags & HASBUF) == 0) 552191783Srmacklem panic("nfsrelpath"); 553191783Srmacklem uma_zfree(namei_zone, ndp->ni_cnd.cn_pnbuf); 554191783Srmacklem ndp->ni_cnd.cn_flags &= ~HASBUF; 555191783Srmacklem} 556191783Srmacklem 557191783Srmacklem/* 558191783Srmacklem * Readlink vnode op into an mbuf list. 559191783Srmacklem */ 560191783Srmacklemint 561191783Srmacklemnfsvno_readlink(struct vnode *vp, struct ucred *cred, struct thread *p, 562191783Srmacklem struct mbuf **mpp, struct mbuf **mpendp, int *lenp) 563191783Srmacklem{ 564191783Srmacklem struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 565191783Srmacklem struct iovec *ivp = iv; 566191783Srmacklem struct uio io, *uiop = &io; 567191783Srmacklem struct mbuf *mp, *mp2 = NULL, *mp3 = NULL; 568224086Szack int i, len, tlen, error = 0; 569191783Srmacklem 570191783Srmacklem len = 0; 571191783Srmacklem i = 0; 572191783Srmacklem while (len < NFS_MAXPATHLEN) { 573191783Srmacklem NFSMGET(mp); 574191783Srmacklem MCLGET(mp, M_WAIT); 575191783Srmacklem mp->m_len = NFSMSIZ(mp); 576191783Srmacklem if (len == 0) { 577191783Srmacklem mp3 = mp2 = mp; 578191783Srmacklem } else { 579191783Srmacklem mp2->m_next = mp; 580191783Srmacklem mp2 = mp; 581191783Srmacklem } 582191783Srmacklem if ((len + mp->m_len) > NFS_MAXPATHLEN) { 583191783Srmacklem mp->m_len = NFS_MAXPATHLEN - len; 584191783Srmacklem len = NFS_MAXPATHLEN; 585191783Srmacklem } else { 586191783Srmacklem len += mp->m_len; 587191783Srmacklem } 588191783Srmacklem ivp->iov_base = mtod(mp, caddr_t); 589191783Srmacklem ivp->iov_len = mp->m_len; 590191783Srmacklem i++; 591191783Srmacklem ivp++; 592191783Srmacklem } 593191783Srmacklem uiop->uio_iov = iv; 594191783Srmacklem uiop->uio_iovcnt = i; 595191783Srmacklem uiop->uio_offset = 0; 596191783Srmacklem uiop->uio_resid = len; 597191783Srmacklem uiop->uio_rw = UIO_READ; 598191783Srmacklem uiop->uio_segflg = UIO_SYSSPACE; 599191783Srmacklem uiop->uio_td = NULL; 600191783Srmacklem error = VOP_READLINK(vp, uiop, cred); 601191783Srmacklem if (error) { 602191783Srmacklem m_freem(mp3); 603191783Srmacklem *lenp = 0; 604224086Szack goto out; 605191783Srmacklem } 606191783Srmacklem if (uiop->uio_resid > 0) { 607191783Srmacklem len -= uiop->uio_resid; 608191783Srmacklem tlen = NFSM_RNDUP(len); 609191783Srmacklem nfsrv_adj(mp3, NFS_MAXPATHLEN - tlen, tlen - len); 610191783Srmacklem } 611191783Srmacklem *lenp = len; 612191783Srmacklem *mpp = mp3; 613191783Srmacklem *mpendp = mp; 614224086Szack 615224086Szackout: 616224086Szack NFSEXITCODE(error); 617224086Szack return (error); 618191783Srmacklem} 619191783Srmacklem 620191783Srmacklem/* 621191783Srmacklem * Read vnode op call into mbuf list. 622191783Srmacklem */ 623191783Srmacklemint 624191783Srmacklemnfsvno_read(struct vnode *vp, off_t off, int cnt, struct ucred *cred, 625191783Srmacklem struct thread *p, struct mbuf **mpp, struct mbuf **mpendp) 626191783Srmacklem{ 627191783Srmacklem struct mbuf *m; 628191783Srmacklem int i; 629191783Srmacklem struct iovec *iv; 630191783Srmacklem struct iovec *iv2; 631229617Sjhb int error = 0, len, left, siz, tlen, ioflag = 0; 632191783Srmacklem struct mbuf *m2 = NULL, *m3; 633191783Srmacklem struct uio io, *uiop = &io; 634191783Srmacklem struct nfsheur *nh; 635191783Srmacklem 636191783Srmacklem len = left = NFSM_RNDUP(cnt); 637191783Srmacklem m3 = NULL; 638191783Srmacklem /* 639191783Srmacklem * Generate the mbuf list with the uio_iov ref. to it. 640191783Srmacklem */ 641191783Srmacklem i = 0; 642191783Srmacklem while (left > 0) { 643191783Srmacklem NFSMGET(m); 644191783Srmacklem MCLGET(m, M_WAIT); 645191783Srmacklem m->m_len = 0; 646191783Srmacklem siz = min(M_TRAILINGSPACE(m), left); 647191783Srmacklem left -= siz; 648191783Srmacklem i++; 649191783Srmacklem if (m3) 650191783Srmacklem m2->m_next = m; 651191783Srmacklem else 652191783Srmacklem m3 = m; 653191783Srmacklem m2 = m; 654191783Srmacklem } 655191783Srmacklem MALLOC(iv, struct iovec *, i * sizeof (struct iovec), 656191783Srmacklem M_TEMP, M_WAITOK); 657191783Srmacklem uiop->uio_iov = iv2 = iv; 658191783Srmacklem m = m3; 659191783Srmacklem left = len; 660191783Srmacklem i = 0; 661191783Srmacklem while (left > 0) { 662191783Srmacklem if (m == NULL) 663191783Srmacklem panic("nfsvno_read iov"); 664191783Srmacklem siz = min(M_TRAILINGSPACE(m), left); 665191783Srmacklem if (siz > 0) { 666191783Srmacklem iv->iov_base = mtod(m, caddr_t) + m->m_len; 667191783Srmacklem iv->iov_len = siz; 668191783Srmacklem m->m_len += siz; 669191783Srmacklem left -= siz; 670191783Srmacklem iv++; 671191783Srmacklem i++; 672191783Srmacklem } 673191783Srmacklem m = m->m_next; 674191783Srmacklem } 675191783Srmacklem uiop->uio_iovcnt = i; 676191783Srmacklem uiop->uio_offset = off; 677191783Srmacklem uiop->uio_resid = len; 678191783Srmacklem uiop->uio_rw = UIO_READ; 679191783Srmacklem uiop->uio_segflg = UIO_SYSSPACE; 680265723Srmacklem uiop->uio_td = NULL; 681229617Sjhb nh = nfsrv_sequential_heuristic(uiop, vp); 682229617Sjhb ioflag |= nh->nh_seqcount << IO_SEQSHIFT; 683191783Srmacklem error = VOP_READ(vp, uiop, IO_NODELOCKED | ioflag, cred); 684191783Srmacklem FREE((caddr_t)iv2, M_TEMP); 685191783Srmacklem if (error) { 686191783Srmacklem m_freem(m3); 687191783Srmacklem *mpp = NULL; 688224086Szack goto out; 689191783Srmacklem } 690229617Sjhb nh->nh_nextoff = uiop->uio_offset; 691191783Srmacklem tlen = len - uiop->uio_resid; 692191783Srmacklem cnt = cnt < tlen ? cnt : tlen; 693191783Srmacklem tlen = NFSM_RNDUP(cnt); 694191783Srmacklem if (tlen == 0) { 695191783Srmacklem m_freem(m3); 696191783Srmacklem m3 = NULL; 697191783Srmacklem } else if (len != tlen || tlen != cnt) 698191783Srmacklem nfsrv_adj(m3, len - tlen, tlen - cnt); 699191783Srmacklem *mpp = m3; 700191783Srmacklem *mpendp = m2; 701224086Szack 702224086Szackout: 703224086Szack NFSEXITCODE(error); 704224086Szack return (error); 705191783Srmacklem} 706191783Srmacklem 707191783Srmacklem/* 708191783Srmacklem * Write vnode op from an mbuf list. 709191783Srmacklem */ 710191783Srmacklemint 711191783Srmacklemnfsvno_write(struct vnode *vp, off_t off, int retlen, int cnt, int stable, 712191783Srmacklem struct mbuf *mp, char *cp, struct ucred *cred, struct thread *p) 713191783Srmacklem{ 714191783Srmacklem struct iovec *ivp; 715191783Srmacklem int i, len; 716191783Srmacklem struct iovec *iv; 717191783Srmacklem int ioflags, error; 718191783Srmacklem struct uio io, *uiop = &io; 719229617Sjhb struct nfsheur *nh; 720191783Srmacklem 721191783Srmacklem MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP, 722191783Srmacklem M_WAITOK); 723191783Srmacklem uiop->uio_iov = iv = ivp; 724191783Srmacklem uiop->uio_iovcnt = cnt; 725191783Srmacklem i = mtod(mp, caddr_t) + mp->m_len - cp; 726191783Srmacklem len = retlen; 727191783Srmacklem while (len > 0) { 728191783Srmacklem if (mp == NULL) 729191783Srmacklem panic("nfsvno_write"); 730191783Srmacklem if (i > 0) { 731191783Srmacklem i = min(i, len); 732191783Srmacklem ivp->iov_base = cp; 733191783Srmacklem ivp->iov_len = i; 734191783Srmacklem ivp++; 735191783Srmacklem len -= i; 736191783Srmacklem } 737191783Srmacklem mp = mp->m_next; 738191783Srmacklem if (mp) { 739191783Srmacklem i = mp->m_len; 740191783Srmacklem cp = mtod(mp, caddr_t); 741191783Srmacklem } 742191783Srmacklem } 743191783Srmacklem 744191783Srmacklem if (stable == NFSWRITE_UNSTABLE) 745191783Srmacklem ioflags = IO_NODELOCKED; 746191783Srmacklem else 747191783Srmacklem ioflags = (IO_SYNC | IO_NODELOCKED); 748191783Srmacklem uiop->uio_resid = retlen; 749191783Srmacklem uiop->uio_rw = UIO_WRITE; 750191783Srmacklem uiop->uio_segflg = UIO_SYSSPACE; 751191783Srmacklem NFSUIOPROC(uiop, p); 752191783Srmacklem uiop->uio_offset = off; 753229617Sjhb nh = nfsrv_sequential_heuristic(uiop, vp); 754229617Sjhb ioflags |= nh->nh_seqcount << IO_SEQSHIFT; 755191783Srmacklem error = VOP_WRITE(vp, uiop, ioflags, cred); 756229617Sjhb if (error == 0) 757229617Sjhb nh->nh_nextoff = uiop->uio_offset; 758191783Srmacklem FREE((caddr_t)iv, M_TEMP); 759224086Szack 760224086Szack NFSEXITCODE(error); 761191783Srmacklem return (error); 762191783Srmacklem} 763191783Srmacklem 764191783Srmacklem/* 765191783Srmacklem * Common code for creating a regular file (plus special files for V2). 766191783Srmacklem */ 767191783Srmacklemint 768191783Srmacklemnfsvno_createsub(struct nfsrv_descript *nd, struct nameidata *ndp, 769191783Srmacklem struct vnode **vpp, struct nfsvattr *nvap, int *exclusive_flagp, 770199616Srmacklem int32_t *cverf, NFSDEV_T rdev, struct thread *p, struct nfsexstuff *exp) 771191783Srmacklem{ 772191783Srmacklem u_quad_t tempsize; 773191783Srmacklem int error; 774191783Srmacklem 775191783Srmacklem error = nd->nd_repstat; 776191783Srmacklem if (!error && ndp->ni_vp == NULL) { 777191783Srmacklem if (nvap->na_type == VREG || nvap->na_type == VSOCK) { 778191783Srmacklem vrele(ndp->ni_startdir); 779191783Srmacklem error = VOP_CREATE(ndp->ni_dvp, 780191783Srmacklem &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); 781191783Srmacklem vput(ndp->ni_dvp); 782191783Srmacklem nfsvno_relpathbuf(ndp); 783191783Srmacklem if (!error) { 784191783Srmacklem if (*exclusive_flagp) { 785191783Srmacklem *exclusive_flagp = 0; 786191783Srmacklem NFSVNO_ATTRINIT(nvap); 787199616Srmacklem nvap->na_atime.tv_sec = cverf[0]; 788199616Srmacklem nvap->na_atime.tv_nsec = cverf[1]; 789191783Srmacklem error = VOP_SETATTR(ndp->ni_vp, 790191783Srmacklem &nvap->na_vattr, nd->nd_cred); 791191783Srmacklem } 792191783Srmacklem } 793191783Srmacklem /* 794191783Srmacklem * NFS V2 Only. nfsrvd_mknod() does this for V3. 795191783Srmacklem * (This implies, just get out on an error.) 796191783Srmacklem */ 797191783Srmacklem } else if (nvap->na_type == VCHR || nvap->na_type == VBLK || 798191783Srmacklem nvap->na_type == VFIFO) { 799191783Srmacklem if (nvap->na_type == VCHR && rdev == 0xffffffff) 800191783Srmacklem nvap->na_type = VFIFO; 801191783Srmacklem if (nvap->na_type != VFIFO && 802191783Srmacklem (error = priv_check_cred(nd->nd_cred, 803191783Srmacklem PRIV_VFS_MKNOD_DEV, 0))) { 804191783Srmacklem vrele(ndp->ni_startdir); 805191783Srmacklem nfsvno_relpathbuf(ndp); 806191783Srmacklem vput(ndp->ni_dvp); 807224086Szack goto out; 808191783Srmacklem } 809191783Srmacklem nvap->na_rdev = rdev; 810191783Srmacklem error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp, 811191783Srmacklem &ndp->ni_cnd, &nvap->na_vattr); 812191783Srmacklem vput(ndp->ni_dvp); 813191783Srmacklem nfsvno_relpathbuf(ndp); 814220546Srmacklem vrele(ndp->ni_startdir); 815220546Srmacklem if (error) 816224086Szack goto out; 817191783Srmacklem } else { 818191783Srmacklem vrele(ndp->ni_startdir); 819191783Srmacklem nfsvno_relpathbuf(ndp); 820191783Srmacklem vput(ndp->ni_dvp); 821224086Szack error = ENXIO; 822224086Szack goto out; 823191783Srmacklem } 824191783Srmacklem *vpp = ndp->ni_vp; 825191783Srmacklem } else { 826191783Srmacklem /* 827191783Srmacklem * Handle cases where error is already set and/or 828191783Srmacklem * the file exists. 829191783Srmacklem * 1 - clean up the lookup 830191783Srmacklem * 2 - iff !error and na_size set, truncate it 831191783Srmacklem */ 832191783Srmacklem vrele(ndp->ni_startdir); 833191783Srmacklem nfsvno_relpathbuf(ndp); 834191783Srmacklem *vpp = ndp->ni_vp; 835191783Srmacklem if (ndp->ni_dvp == *vpp) 836191783Srmacklem vrele(ndp->ni_dvp); 837191783Srmacklem else 838191783Srmacklem vput(ndp->ni_dvp); 839191783Srmacklem if (!error && nvap->na_size != VNOVAL) { 840200999Srmacklem error = nfsvno_accchk(*vpp, VWRITE, 841191783Srmacklem nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, 842200999Srmacklem NFSACCCHK_VPISLOCKED, NULL); 843191783Srmacklem if (!error) { 844191783Srmacklem tempsize = nvap->na_size; 845191783Srmacklem NFSVNO_ATTRINIT(nvap); 846191783Srmacklem nvap->na_size = tempsize; 847191783Srmacklem error = VOP_SETATTR(*vpp, 848191783Srmacklem &nvap->na_vattr, nd->nd_cred); 849191783Srmacklem } 850191783Srmacklem } 851191783Srmacklem if (error) 852191783Srmacklem vput(*vpp); 853191783Srmacklem } 854224086Szack 855224086Szackout: 856224086Szack NFSEXITCODE(error); 857191783Srmacklem return (error); 858191783Srmacklem} 859191783Srmacklem 860191783Srmacklem/* 861191783Srmacklem * Do a mknod vnode op. 862191783Srmacklem */ 863191783Srmacklemint 864191783Srmacklemnfsvno_mknod(struct nameidata *ndp, struct nfsvattr *nvap, struct ucred *cred, 865191783Srmacklem struct thread *p) 866191783Srmacklem{ 867191783Srmacklem int error = 0; 868191783Srmacklem enum vtype vtyp; 869191783Srmacklem 870191783Srmacklem vtyp = nvap->na_type; 871191783Srmacklem /* 872191783Srmacklem * Iff doesn't exist, create it. 873191783Srmacklem */ 874191783Srmacklem if (ndp->ni_vp) { 875191783Srmacklem vrele(ndp->ni_startdir); 876191783Srmacklem nfsvno_relpathbuf(ndp); 877191783Srmacklem vput(ndp->ni_dvp); 878191783Srmacklem vrele(ndp->ni_vp); 879224086Szack error = EEXIST; 880224086Szack goto out; 881191783Srmacklem } 882191783Srmacklem if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 883191783Srmacklem vrele(ndp->ni_startdir); 884191783Srmacklem nfsvno_relpathbuf(ndp); 885191783Srmacklem vput(ndp->ni_dvp); 886224086Szack error = NFSERR_BADTYPE; 887224086Szack goto out; 888191783Srmacklem } 889191783Srmacklem if (vtyp == VSOCK) { 890191783Srmacklem vrele(ndp->ni_startdir); 891191783Srmacklem error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 892191783Srmacklem &ndp->ni_cnd, &nvap->na_vattr); 893191783Srmacklem vput(ndp->ni_dvp); 894191783Srmacklem nfsvno_relpathbuf(ndp); 895191783Srmacklem } else { 896191783Srmacklem if (nvap->na_type != VFIFO && 897191783Srmacklem (error = priv_check_cred(cred, PRIV_VFS_MKNOD_DEV, 0))) { 898191783Srmacklem vrele(ndp->ni_startdir); 899191783Srmacklem nfsvno_relpathbuf(ndp); 900191783Srmacklem vput(ndp->ni_dvp); 901224086Szack goto out; 902191783Srmacklem } 903191783Srmacklem error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp, 904191783Srmacklem &ndp->ni_cnd, &nvap->na_vattr); 905191783Srmacklem vput(ndp->ni_dvp); 906191783Srmacklem nfsvno_relpathbuf(ndp); 907216898Srmacklem vrele(ndp->ni_startdir); 908191783Srmacklem /* 909191783Srmacklem * Since VOP_MKNOD returns the ni_vp, I can't 910191783Srmacklem * see any reason to do the lookup. 911191783Srmacklem */ 912191783Srmacklem } 913224086Szack 914224086Szackout: 915224086Szack NFSEXITCODE(error); 916191783Srmacklem return (error); 917191783Srmacklem} 918191783Srmacklem 919191783Srmacklem/* 920191783Srmacklem * Mkdir vnode op. 921191783Srmacklem */ 922191783Srmacklemint 923191783Srmacklemnfsvno_mkdir(struct nameidata *ndp, struct nfsvattr *nvap, uid_t saved_uid, 924191783Srmacklem struct ucred *cred, struct thread *p, struct nfsexstuff *exp) 925191783Srmacklem{ 926191783Srmacklem int error = 0; 927191783Srmacklem 928191783Srmacklem if (ndp->ni_vp != NULL) { 929191783Srmacklem if (ndp->ni_dvp == ndp->ni_vp) 930191783Srmacklem vrele(ndp->ni_dvp); 931191783Srmacklem else 932191783Srmacklem vput(ndp->ni_dvp); 933191783Srmacklem vrele(ndp->ni_vp); 934206063Srmacklem nfsvno_relpathbuf(ndp); 935224086Szack error = EEXIST; 936224086Szack goto out; 937191783Srmacklem } 938191783Srmacklem error = VOP_MKDIR(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, 939191783Srmacklem &nvap->na_vattr); 940191783Srmacklem vput(ndp->ni_dvp); 941191783Srmacklem nfsvno_relpathbuf(ndp); 942224086Szack 943224086Szackout: 944224086Szack NFSEXITCODE(error); 945191783Srmacklem return (error); 946191783Srmacklem} 947191783Srmacklem 948191783Srmacklem/* 949191783Srmacklem * symlink vnode op. 950191783Srmacklem */ 951191783Srmacklemint 952191783Srmacklemnfsvno_symlink(struct nameidata *ndp, struct nfsvattr *nvap, char *pathcp, 953191783Srmacklem int pathlen, int not_v2, uid_t saved_uid, struct ucred *cred, struct thread *p, 954191783Srmacklem struct nfsexstuff *exp) 955191783Srmacklem{ 956191783Srmacklem int error = 0; 957191783Srmacklem 958191783Srmacklem if (ndp->ni_vp) { 959191783Srmacklem vrele(ndp->ni_startdir); 960191783Srmacklem nfsvno_relpathbuf(ndp); 961191783Srmacklem if (ndp->ni_dvp == ndp->ni_vp) 962191783Srmacklem vrele(ndp->ni_dvp); 963191783Srmacklem else 964191783Srmacklem vput(ndp->ni_dvp); 965191783Srmacklem vrele(ndp->ni_vp); 966224086Szack error = EEXIST; 967224086Szack goto out; 968191783Srmacklem } 969191783Srmacklem 970191783Srmacklem error = VOP_SYMLINK(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, 971191783Srmacklem &nvap->na_vattr, pathcp); 972191783Srmacklem vput(ndp->ni_dvp); 973191783Srmacklem vrele(ndp->ni_startdir); 974191783Srmacklem nfsvno_relpathbuf(ndp); 975191783Srmacklem /* 976191783Srmacklem * Although FreeBSD still had the lookup code in 977191783Srmacklem * it for 7/current, there doesn't seem to be any 978191783Srmacklem * point, since VOP_SYMLINK() returns the ni_vp. 979191783Srmacklem * Just vput it for v2. 980191783Srmacklem */ 981191783Srmacklem if (!not_v2 && !error) 982191783Srmacklem vput(ndp->ni_vp); 983224086Szack 984224086Szackout: 985224086Szack NFSEXITCODE(error); 986191783Srmacklem return (error); 987191783Srmacklem} 988191783Srmacklem 989191783Srmacklem/* 990191783Srmacklem * Parse symbolic link arguments. 991191783Srmacklem * This function has an ugly side effect. It will MALLOC() an area for 992191783Srmacklem * the symlink and set iov_base to point to it, only if it succeeds. 993191783Srmacklem * So, if it returns with uiop->uio_iov->iov_base != NULL, that must 994191783Srmacklem * be FREE'd later. 995191783Srmacklem */ 996191783Srmacklemint 997191783Srmacklemnfsvno_getsymlink(struct nfsrv_descript *nd, struct nfsvattr *nvap, 998191783Srmacklem struct thread *p, char **pathcpp, int *lenp) 999191783Srmacklem{ 1000191783Srmacklem u_int32_t *tl; 1001191783Srmacklem char *pathcp = NULL; 1002191783Srmacklem int error = 0, len; 1003191783Srmacklem struct nfsv2_sattr *sp; 1004191783Srmacklem 1005191783Srmacklem *pathcpp = NULL; 1006191783Srmacklem *lenp = 0; 1007191783Srmacklem if ((nd->nd_flag & ND_NFSV3) && 1008191783Srmacklem (error = nfsrv_sattr(nd, nvap, NULL, NULL, p))) 1009191783Srmacklem goto nfsmout; 1010191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 1011191783Srmacklem len = fxdr_unsigned(int, *tl); 1012191783Srmacklem if (len > NFS_MAXPATHLEN || len <= 0) { 1013191783Srmacklem error = EBADRPC; 1014191783Srmacklem goto nfsmout; 1015191783Srmacklem } 1016191783Srmacklem MALLOC(pathcp, caddr_t, len + 1, M_TEMP, M_WAITOK); 1017191783Srmacklem error = nfsrv_mtostr(nd, pathcp, len); 1018191783Srmacklem if (error) 1019191783Srmacklem goto nfsmout; 1020191783Srmacklem if (nd->nd_flag & ND_NFSV2) { 1021191783Srmacklem NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1022191783Srmacklem nvap->na_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 1023191783Srmacklem } 1024191783Srmacklem *pathcpp = pathcp; 1025191783Srmacklem *lenp = len; 1026224086Szack NFSEXITCODE2(0, nd); 1027191783Srmacklem return (0); 1028191783Srmacklemnfsmout: 1029191783Srmacklem if (pathcp) 1030191783Srmacklem free(pathcp, M_TEMP); 1031224086Szack NFSEXITCODE2(error, nd); 1032191783Srmacklem return (error); 1033191783Srmacklem} 1034191783Srmacklem 1035191783Srmacklem/* 1036191783Srmacklem * Remove a non-directory object. 1037191783Srmacklem */ 1038191783Srmacklemint 1039191783Srmacklemnfsvno_removesub(struct nameidata *ndp, int is_v4, struct ucred *cred, 1040191783Srmacklem struct thread *p, struct nfsexstuff *exp) 1041191783Srmacklem{ 1042191783Srmacklem struct vnode *vp; 1043191783Srmacklem int error = 0; 1044191783Srmacklem 1045191783Srmacklem vp = ndp->ni_vp; 1046191783Srmacklem if (vp->v_type == VDIR) 1047191783Srmacklem error = NFSERR_ISDIR; 1048191783Srmacklem else if (is_v4) 1049191783Srmacklem error = nfsrv_checkremove(vp, 1, p); 1050191783Srmacklem if (!error) 1051191783Srmacklem error = VOP_REMOVE(ndp->ni_dvp, vp, &ndp->ni_cnd); 1052191783Srmacklem if (ndp->ni_dvp == vp) 1053191783Srmacklem vrele(ndp->ni_dvp); 1054191783Srmacklem else 1055191783Srmacklem vput(ndp->ni_dvp); 1056191783Srmacklem vput(vp); 1057236134Srmacklem if ((ndp->ni_cnd.cn_flags & SAVENAME) != 0) 1058236134Srmacklem nfsvno_relpathbuf(ndp); 1059224086Szack NFSEXITCODE(error); 1060191783Srmacklem return (error); 1061191783Srmacklem} 1062191783Srmacklem 1063191783Srmacklem/* 1064191783Srmacklem * Remove a directory. 1065191783Srmacklem */ 1066191783Srmacklemint 1067191783Srmacklemnfsvno_rmdirsub(struct nameidata *ndp, int is_v4, struct ucred *cred, 1068191783Srmacklem struct thread *p, struct nfsexstuff *exp) 1069191783Srmacklem{ 1070191783Srmacklem struct vnode *vp; 1071191783Srmacklem int error = 0; 1072191783Srmacklem 1073191783Srmacklem vp = ndp->ni_vp; 1074191783Srmacklem if (vp->v_type != VDIR) { 1075191783Srmacklem error = ENOTDIR; 1076191783Srmacklem goto out; 1077191783Srmacklem } 1078191783Srmacklem /* 1079191783Srmacklem * No rmdir "." please. 1080191783Srmacklem */ 1081191783Srmacklem if (ndp->ni_dvp == vp) { 1082191783Srmacklem error = EINVAL; 1083191783Srmacklem goto out; 1084191783Srmacklem } 1085191783Srmacklem /* 1086191783Srmacklem * The root of a mounted filesystem cannot be deleted. 1087191783Srmacklem */ 1088191783Srmacklem if (vp->v_vflag & VV_ROOT) 1089191783Srmacklem error = EBUSY; 1090191783Srmacklemout: 1091191783Srmacklem if (!error) 1092191783Srmacklem error = VOP_RMDIR(ndp->ni_dvp, vp, &ndp->ni_cnd); 1093191783Srmacklem if (ndp->ni_dvp == vp) 1094191783Srmacklem vrele(ndp->ni_dvp); 1095191783Srmacklem else 1096191783Srmacklem vput(ndp->ni_dvp); 1097191783Srmacklem vput(vp); 1098236134Srmacklem if ((ndp->ni_cnd.cn_flags & SAVENAME) != 0) 1099236134Srmacklem nfsvno_relpathbuf(ndp); 1100224086Szack NFSEXITCODE(error); 1101191783Srmacklem return (error); 1102191783Srmacklem} 1103191783Srmacklem 1104191783Srmacklem/* 1105191783Srmacklem * Rename vnode op. 1106191783Srmacklem */ 1107191783Srmacklemint 1108191783Srmacklemnfsvno_rename(struct nameidata *fromndp, struct nameidata *tondp, 1109191783Srmacklem u_int32_t ndstat, u_int32_t ndflag, struct ucred *cred, struct thread *p) 1110191783Srmacklem{ 1111191783Srmacklem struct vnode *fvp, *tvp, *tdvp; 1112191783Srmacklem int error = 0; 1113191783Srmacklem 1114191783Srmacklem fvp = fromndp->ni_vp; 1115191783Srmacklem if (ndstat) { 1116191783Srmacklem vrele(fromndp->ni_dvp); 1117191783Srmacklem vrele(fvp); 1118191783Srmacklem error = ndstat; 1119191783Srmacklem goto out1; 1120191783Srmacklem } 1121191783Srmacklem tdvp = tondp->ni_dvp; 1122191783Srmacklem tvp = tondp->ni_vp; 1123191783Srmacklem if (tvp != NULL) { 1124191783Srmacklem if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1125191783Srmacklem error = (ndflag & ND_NFSV2) ? EISDIR : EEXIST; 1126191783Srmacklem goto out; 1127191783Srmacklem } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1128191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTDIR : EEXIST; 1129191783Srmacklem goto out; 1130191783Srmacklem } 1131191783Srmacklem if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1132191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; 1133191783Srmacklem goto out; 1134191783Srmacklem } 1135191783Srmacklem 1136191783Srmacklem /* 1137191783Srmacklem * A rename to '.' or '..' results in a prematurely 1138191783Srmacklem * unlocked vnode on FreeBSD5, so I'm just going to fail that 1139191783Srmacklem * here. 1140191783Srmacklem */ 1141191783Srmacklem if ((tondp->ni_cnd.cn_namelen == 1 && 1142191783Srmacklem tondp->ni_cnd.cn_nameptr[0] == '.') || 1143191783Srmacklem (tondp->ni_cnd.cn_namelen == 2 && 1144191783Srmacklem tondp->ni_cnd.cn_nameptr[0] == '.' && 1145191783Srmacklem tondp->ni_cnd.cn_nameptr[1] == '.')) { 1146191783Srmacklem error = EINVAL; 1147191783Srmacklem goto out; 1148191783Srmacklem } 1149191783Srmacklem } 1150191783Srmacklem if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1151191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; 1152191783Srmacklem goto out; 1153191783Srmacklem } 1154191783Srmacklem if (fvp->v_mount != tdvp->v_mount) { 1155191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; 1156191783Srmacklem goto out; 1157191783Srmacklem } 1158191783Srmacklem if (fvp == tdvp) { 1159191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EINVAL; 1160191783Srmacklem goto out; 1161191783Srmacklem } 1162191783Srmacklem if (fvp == tvp) { 1163191783Srmacklem /* 1164191783Srmacklem * If source and destination are the same, there is nothing to 1165191783Srmacklem * do. Set error to -1 to indicate this. 1166191783Srmacklem */ 1167191783Srmacklem error = -1; 1168191783Srmacklem goto out; 1169191783Srmacklem } 1170191783Srmacklem if (ndflag & ND_NFSV4) { 1171224081Szack if (NFSVOPLOCK(fvp, LK_EXCLUSIVE) == 0) { 1172216893Srmacklem error = nfsrv_checkremove(fvp, 0, p); 1173224082Szack NFSVOPUNLOCK(fvp, 0); 1174216893Srmacklem } else 1175216893Srmacklem error = EPERM; 1176191783Srmacklem if (tvp && !error) 1177191783Srmacklem error = nfsrv_checkremove(tvp, 1, p); 1178191783Srmacklem } else { 1179191783Srmacklem /* 1180191783Srmacklem * For NFSv2 and NFSv3, try to get rid of the delegation, so 1181191783Srmacklem * that the NFSv4 client won't be confused by the rename. 1182191783Srmacklem * Since nfsd_recalldelegation() can only be called on an 1183191783Srmacklem * unlocked vnode at this point and fvp is the file that will 1184191783Srmacklem * still exist after the rename, just do fvp. 1185191783Srmacklem */ 1186191783Srmacklem nfsd_recalldelegation(fvp, p); 1187191783Srmacklem } 1188191783Srmacklemout: 1189191783Srmacklem if (!error) { 1190191783Srmacklem error = VOP_RENAME(fromndp->ni_dvp, fromndp->ni_vp, 1191191783Srmacklem &fromndp->ni_cnd, tondp->ni_dvp, tondp->ni_vp, 1192191783Srmacklem &tondp->ni_cnd); 1193191783Srmacklem } else { 1194191783Srmacklem if (tdvp == tvp) 1195191783Srmacklem vrele(tdvp); 1196191783Srmacklem else 1197191783Srmacklem vput(tdvp); 1198191783Srmacklem if (tvp) 1199191783Srmacklem vput(tvp); 1200191783Srmacklem vrele(fromndp->ni_dvp); 1201191783Srmacklem vrele(fvp); 1202191783Srmacklem if (error == -1) 1203191783Srmacklem error = 0; 1204191783Srmacklem } 1205191783Srmacklem vrele(tondp->ni_startdir); 1206191783Srmacklem nfsvno_relpathbuf(tondp); 1207191783Srmacklemout1: 1208191783Srmacklem vrele(fromndp->ni_startdir); 1209191783Srmacklem nfsvno_relpathbuf(fromndp); 1210224086Szack NFSEXITCODE(error); 1211191783Srmacklem return (error); 1212191783Srmacklem} 1213191783Srmacklem 1214191783Srmacklem/* 1215191783Srmacklem * Link vnode op. 1216191783Srmacklem */ 1217191783Srmacklemint 1218191783Srmacklemnfsvno_link(struct nameidata *ndp, struct vnode *vp, struct ucred *cred, 1219191783Srmacklem struct thread *p, struct nfsexstuff *exp) 1220191783Srmacklem{ 1221191783Srmacklem struct vnode *xp; 1222191783Srmacklem int error = 0; 1223191783Srmacklem 1224191783Srmacklem xp = ndp->ni_vp; 1225191783Srmacklem if (xp != NULL) { 1226191783Srmacklem error = EEXIST; 1227191783Srmacklem } else { 1228191783Srmacklem xp = ndp->ni_dvp; 1229191783Srmacklem if (vp->v_mount != xp->v_mount) 1230191783Srmacklem error = EXDEV; 1231191783Srmacklem } 1232191783Srmacklem if (!error) { 1233224081Szack NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY); 1234216893Srmacklem if ((vp->v_iflag & VI_DOOMED) == 0) 1235216893Srmacklem error = VOP_LINK(ndp->ni_dvp, vp, &ndp->ni_cnd); 1236216893Srmacklem else 1237216893Srmacklem error = EPERM; 1238191783Srmacklem if (ndp->ni_dvp == vp) 1239191783Srmacklem vrele(ndp->ni_dvp); 1240191783Srmacklem else 1241191783Srmacklem vput(ndp->ni_dvp); 1242224082Szack NFSVOPUNLOCK(vp, 0); 1243191783Srmacklem } else { 1244191783Srmacklem if (ndp->ni_dvp == ndp->ni_vp) 1245191783Srmacklem vrele(ndp->ni_dvp); 1246191783Srmacklem else 1247191783Srmacklem vput(ndp->ni_dvp); 1248191783Srmacklem if (ndp->ni_vp) 1249191783Srmacklem vrele(ndp->ni_vp); 1250191783Srmacklem } 1251191783Srmacklem nfsvno_relpathbuf(ndp); 1252224086Szack NFSEXITCODE(error); 1253191783Srmacklem return (error); 1254191783Srmacklem} 1255191783Srmacklem 1256191783Srmacklem/* 1257191783Srmacklem * Do the fsync() appropriate for the commit. 1258191783Srmacklem */ 1259191783Srmacklemint 1260191783Srmacklemnfsvno_fsync(struct vnode *vp, u_int64_t off, int cnt, struct ucred *cred, 1261191783Srmacklem struct thread *td) 1262191783Srmacklem{ 1263191783Srmacklem int error = 0; 1264191783Srmacklem 1265229827Srmacklem /* 1266229827Srmacklem * RFC 1813 3.3.21: if count is 0, a flush from offset to the end of 1267229827Srmacklem * file is done. At this time VOP_FSYNC does not accept offset and 1268229827Srmacklem * byte count parameters so call VOP_FSYNC the whole file for now. 1269229827Srmacklem * The same is true for NFSv4: RFC 3530 Sec. 14.2.3. 1270229827Srmacklem */ 1271229827Srmacklem if (cnt == 0 || cnt > MAX_COMMIT_COUNT) { 1272191783Srmacklem /* 1273191783Srmacklem * Give up and do the whole thing 1274191783Srmacklem */ 1275191783Srmacklem if (vp->v_object && 1276191783Srmacklem (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) { 1277191783Srmacklem VM_OBJECT_LOCK(vp->v_object); 1278191783Srmacklem vm_object_page_clean(vp->v_object, 0, 0, OBJPC_SYNC); 1279191783Srmacklem VM_OBJECT_UNLOCK(vp->v_object); 1280191783Srmacklem } 1281191783Srmacklem error = VOP_FSYNC(vp, MNT_WAIT, td); 1282191783Srmacklem } else { 1283191783Srmacklem /* 1284191783Srmacklem * Locate and synchronously write any buffers that fall 1285191783Srmacklem * into the requested range. Note: we are assuming that 1286191783Srmacklem * f_iosize is a power of 2. 1287191783Srmacklem */ 1288191783Srmacklem int iosize = vp->v_mount->mnt_stat.f_iosize; 1289191783Srmacklem int iomask = iosize - 1; 1290191783Srmacklem struct bufobj *bo; 1291191783Srmacklem daddr_t lblkno; 1292191783Srmacklem 1293191783Srmacklem /* 1294191783Srmacklem * Align to iosize boundry, super-align to page boundry. 1295191783Srmacklem */ 1296191783Srmacklem if (off & iomask) { 1297191783Srmacklem cnt += off & iomask; 1298191783Srmacklem off &= ~(u_quad_t)iomask; 1299191783Srmacklem } 1300191783Srmacklem if (off & PAGE_MASK) { 1301191783Srmacklem cnt += off & PAGE_MASK; 1302191783Srmacklem off &= ~(u_quad_t)PAGE_MASK; 1303191783Srmacklem } 1304191783Srmacklem lblkno = off / iosize; 1305191783Srmacklem 1306191783Srmacklem if (vp->v_object && 1307191783Srmacklem (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) { 1308191783Srmacklem VM_OBJECT_LOCK(vp->v_object); 1309218345Salc vm_object_page_clean(vp->v_object, off, off + cnt, 1310218345Salc OBJPC_SYNC); 1311191783Srmacklem VM_OBJECT_UNLOCK(vp->v_object); 1312191783Srmacklem } 1313191783Srmacklem 1314191783Srmacklem bo = &vp->v_bufobj; 1315191783Srmacklem BO_LOCK(bo); 1316191783Srmacklem while (cnt > 0) { 1317191783Srmacklem struct buf *bp; 1318191783Srmacklem 1319191783Srmacklem /* 1320191783Srmacklem * If we have a buffer and it is marked B_DELWRI we 1321191783Srmacklem * have to lock and write it. Otherwise the prior 1322191783Srmacklem * write is assumed to have already been committed. 1323191783Srmacklem * 1324191783Srmacklem * gbincore() can return invalid buffers now so we 1325191783Srmacklem * have to check that bit as well (though B_DELWRI 1326191783Srmacklem * should not be set if B_INVAL is set there could be 1327191783Srmacklem * a race here since we haven't locked the buffer). 1328191783Srmacklem */ 1329191783Srmacklem if ((bp = gbincore(&vp->v_bufobj, lblkno)) != NULL) { 1330191783Srmacklem if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | 1331191783Srmacklem LK_INTERLOCK, BO_MTX(bo)) == ENOLCK) { 1332191783Srmacklem BO_LOCK(bo); 1333191783Srmacklem continue; /* retry */ 1334191783Srmacklem } 1335191783Srmacklem if ((bp->b_flags & (B_DELWRI|B_INVAL)) == 1336191783Srmacklem B_DELWRI) { 1337191783Srmacklem bremfree(bp); 1338191783Srmacklem bp->b_flags &= ~B_ASYNC; 1339191783Srmacklem bwrite(bp); 1340191783Srmacklem ++nfs_commit_miss; 1341191783Srmacklem } else 1342191783Srmacklem BUF_UNLOCK(bp); 1343191783Srmacklem BO_LOCK(bo); 1344191783Srmacklem } 1345191783Srmacklem ++nfs_commit_blks; 1346191783Srmacklem if (cnt < iosize) 1347191783Srmacklem break; 1348191783Srmacklem cnt -= iosize; 1349191783Srmacklem ++lblkno; 1350191783Srmacklem } 1351191783Srmacklem BO_UNLOCK(bo); 1352191783Srmacklem } 1353224086Szack NFSEXITCODE(error); 1354191783Srmacklem return (error); 1355191783Srmacklem} 1356191783Srmacklem 1357191783Srmacklem/* 1358191783Srmacklem * Statfs vnode op. 1359191783Srmacklem */ 1360191783Srmacklemint 1361191990Sattilionfsvno_statfs(struct vnode *vp, struct statfs *sf) 1362191783Srmacklem{ 1363221517Srmacklem int error; 1364191783Srmacklem 1365221517Srmacklem error = VFS_STATFS(vp->v_mount, sf); 1366221517Srmacklem if (error == 0) { 1367221517Srmacklem /* 1368221517Srmacklem * Since NFS handles these values as unsigned on the 1369221517Srmacklem * wire, there is no way to represent negative values, 1370221517Srmacklem * so set them to 0. Without this, they will appear 1371221517Srmacklem * to be very large positive values for clients like 1372221517Srmacklem * Solaris10. 1373221517Srmacklem */ 1374221517Srmacklem if (sf->f_bavail < 0) 1375221517Srmacklem sf->f_bavail = 0; 1376221517Srmacklem if (sf->f_ffree < 0) 1377221517Srmacklem sf->f_ffree = 0; 1378221517Srmacklem } 1379224086Szack NFSEXITCODE(error); 1380221517Srmacklem return (error); 1381191783Srmacklem} 1382191783Srmacklem 1383191783Srmacklem/* 1384191783Srmacklem * Do the vnode op stuff for Open. Similar to nfsvno_createsub(), but 1385191783Srmacklem * must handle nfsrv_opencheck() calls after any other access checks. 1386191783Srmacklem */ 1387191783Srmacklemvoid 1388191783Srmacklemnfsvno_open(struct nfsrv_descript *nd, struct nameidata *ndp, 1389191783Srmacklem nfsquad_t clientid, nfsv4stateid_t *stateidp, struct nfsstate *stp, 1390199616Srmacklem int *exclusive_flagp, struct nfsvattr *nvap, int32_t *cverf, int create, 1391191783Srmacklem NFSACL_T *aclp, nfsattrbit_t *attrbitp, struct ucred *cred, struct thread *p, 1392191783Srmacklem struct nfsexstuff *exp, struct vnode **vpp) 1393191783Srmacklem{ 1394191783Srmacklem struct vnode *vp = NULL; 1395191783Srmacklem u_quad_t tempsize; 1396191783Srmacklem struct nfsexstuff nes; 1397191783Srmacklem 1398191783Srmacklem if (ndp->ni_vp == NULL) 1399191783Srmacklem nd->nd_repstat = nfsrv_opencheck(clientid, 1400191783Srmacklem stateidp, stp, NULL, nd, p, nd->nd_repstat); 1401191783Srmacklem if (!nd->nd_repstat) { 1402191783Srmacklem if (ndp->ni_vp == NULL) { 1403191783Srmacklem vrele(ndp->ni_startdir); 1404191783Srmacklem nd->nd_repstat = VOP_CREATE(ndp->ni_dvp, 1405191783Srmacklem &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); 1406191783Srmacklem vput(ndp->ni_dvp); 1407191783Srmacklem nfsvno_relpathbuf(ndp); 1408191783Srmacklem if (!nd->nd_repstat) { 1409191783Srmacklem if (*exclusive_flagp) { 1410191783Srmacklem *exclusive_flagp = 0; 1411191783Srmacklem NFSVNO_ATTRINIT(nvap); 1412199616Srmacklem nvap->na_atime.tv_sec = cverf[0]; 1413199616Srmacklem nvap->na_atime.tv_nsec = cverf[1]; 1414191783Srmacklem nd->nd_repstat = VOP_SETATTR(ndp->ni_vp, 1415191783Srmacklem &nvap->na_vattr, cred); 1416191783Srmacklem } else { 1417191783Srmacklem nfsrv_fixattr(nd, ndp->ni_vp, nvap, 1418191783Srmacklem aclp, p, attrbitp, exp); 1419191783Srmacklem } 1420191783Srmacklem } 1421191783Srmacklem vp = ndp->ni_vp; 1422191783Srmacklem } else { 1423191783Srmacklem if (ndp->ni_startdir) 1424191783Srmacklem vrele(ndp->ni_startdir); 1425191783Srmacklem nfsvno_relpathbuf(ndp); 1426191783Srmacklem vp = ndp->ni_vp; 1427191783Srmacklem if (create == NFSV4OPEN_CREATE) { 1428191783Srmacklem if (ndp->ni_dvp == vp) 1429191783Srmacklem vrele(ndp->ni_dvp); 1430191783Srmacklem else 1431191783Srmacklem vput(ndp->ni_dvp); 1432191783Srmacklem } 1433191783Srmacklem if (NFSVNO_ISSETSIZE(nvap) && vp->v_type == VREG) { 1434191783Srmacklem if (ndp->ni_cnd.cn_flags & RDONLY) 1435191783Srmacklem NFSVNO_SETEXRDONLY(&nes); 1436191783Srmacklem else 1437191783Srmacklem NFSVNO_EXINIT(&nes); 1438191783Srmacklem nd->nd_repstat = nfsvno_accchk(vp, 1439200999Srmacklem VWRITE, cred, &nes, p, 1440200999Srmacklem NFSACCCHK_NOOVERRIDE, 1441200999Srmacklem NFSACCCHK_VPISLOCKED, NULL); 1442191783Srmacklem nd->nd_repstat = nfsrv_opencheck(clientid, 1443191783Srmacklem stateidp, stp, vp, nd, p, nd->nd_repstat); 1444191783Srmacklem if (!nd->nd_repstat) { 1445191783Srmacklem tempsize = nvap->na_size; 1446191783Srmacklem NFSVNO_ATTRINIT(nvap); 1447191783Srmacklem nvap->na_size = tempsize; 1448191783Srmacklem nd->nd_repstat = VOP_SETATTR(vp, 1449191783Srmacklem &nvap->na_vattr, cred); 1450191783Srmacklem } 1451191783Srmacklem } else if (vp->v_type == VREG) { 1452191783Srmacklem nd->nd_repstat = nfsrv_opencheck(clientid, 1453191783Srmacklem stateidp, stp, vp, nd, p, nd->nd_repstat); 1454191783Srmacklem } 1455191783Srmacklem } 1456191783Srmacklem } else { 1457191783Srmacklem if (ndp->ni_cnd.cn_flags & HASBUF) 1458191783Srmacklem nfsvno_relpathbuf(ndp); 1459191783Srmacklem if (ndp->ni_startdir && create == NFSV4OPEN_CREATE) { 1460191783Srmacklem vrele(ndp->ni_startdir); 1461191783Srmacklem if (ndp->ni_dvp == ndp->ni_vp) 1462191783Srmacklem vrele(ndp->ni_dvp); 1463191783Srmacklem else 1464191783Srmacklem vput(ndp->ni_dvp); 1465191783Srmacklem if (ndp->ni_vp) 1466191783Srmacklem vput(ndp->ni_vp); 1467191783Srmacklem } 1468191783Srmacklem } 1469191783Srmacklem *vpp = vp; 1470224086Szack 1471224086Szack NFSEXITCODE2(0, nd); 1472191783Srmacklem} 1473191783Srmacklem 1474191783Srmacklem/* 1475191783Srmacklem * Updates the file rev and sets the mtime and ctime 1476191783Srmacklem * to the current clock time, returning the va_filerev and va_Xtime 1477191783Srmacklem * values. 1478260174Srmacklem * Return ESTALE to indicate the vnode is VI_DOOMED. 1479191783Srmacklem */ 1480260174Srmacklemint 1481191783Srmacklemnfsvno_updfilerev(struct vnode *vp, struct nfsvattr *nvap, 1482191783Srmacklem struct ucred *cred, struct thread *p) 1483191783Srmacklem{ 1484191783Srmacklem struct vattr va; 1485191783Srmacklem 1486191783Srmacklem VATTR_NULL(&va); 1487247502Sjhb vfs_timestamp(&va.va_mtime); 1488260174Srmacklem if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) { 1489260174Srmacklem NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY); 1490260174Srmacklem if ((vp->v_iflag & VI_DOOMED) != 0) 1491260174Srmacklem return (ESTALE); 1492260174Srmacklem } 1493191783Srmacklem (void) VOP_SETATTR(vp, &va, cred); 1494216693Srmacklem (void) nfsvno_getattr(vp, nvap, cred, p, 1); 1495260174Srmacklem return (0); 1496191783Srmacklem} 1497191783Srmacklem 1498191783Srmacklem/* 1499191783Srmacklem * Glue routine to nfsv4_fillattr(). 1500191783Srmacklem */ 1501191783Srmacklemint 1502220645Srmacklemnfsvno_fillattr(struct nfsrv_descript *nd, struct mount *mp, struct vnode *vp, 1503191783Srmacklem struct nfsvattr *nvap, fhandle_t *fhp, int rderror, nfsattrbit_t *attrbitp, 1504220648Srmacklem struct ucred *cred, struct thread *p, int isdgram, int reterr, 1505220648Srmacklem int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno) 1506191783Srmacklem{ 1507191783Srmacklem int error; 1508191783Srmacklem 1509220645Srmacklem error = nfsv4_fillattr(nd, mp, vp, NULL, &nvap->na_vattr, fhp, rderror, 1510220648Srmacklem attrbitp, cred, p, isdgram, reterr, supports_nfsv4acls, at_root, 1511220648Srmacklem mounted_on_fileno); 1512224086Szack NFSEXITCODE2(0, nd); 1513191783Srmacklem return (error); 1514191783Srmacklem} 1515191783Srmacklem 1516191783Srmacklem/* Since the Readdir vnode ops vary, put the entire functions in here. */ 1517191783Srmacklem/* 1518191783Srmacklem * nfs readdir service 1519191783Srmacklem * - mallocs what it thinks is enough to read 1520191783Srmacklem * count rounded up to a multiple of DIRBLKSIZ <= NFS_MAXREADDIR 1521206170Srmacklem * - calls VOP_READDIR() 1522191783Srmacklem * - loops around building the reply 1523191783Srmacklem * if the output generated exceeds count break out of loop 1524191783Srmacklem * The NFSM_CLGET macro is used here so that the reply will be packed 1525191783Srmacklem * tightly in mbuf clusters. 1526191783Srmacklem * - it trims out records with d_fileno == 0 1527191783Srmacklem * this doesn't matter for Unix clients, but they might confuse clients 1528191783Srmacklem * for other os'. 1529191783Srmacklem * - it trims out records with d_type == DT_WHT 1530191783Srmacklem * these cannot be seen through NFS (unless we extend the protocol) 1531191783Srmacklem * The alternate call nfsrvd_readdirplus() does lookups as well. 1532191783Srmacklem * PS: The NFS protocol spec. does not clarify what the "count" byte 1533191783Srmacklem * argument is a count of.. just name strings and file id's or the 1534191783Srmacklem * entire reply rpc or ... 1535191783Srmacklem * I tried just file name and id sizes and it confused the Sun client, 1536191783Srmacklem * so I am using the full rpc size now. The "paranoia.." comment refers 1537191783Srmacklem * to including the status longwords that are not a part of the dir. 1538191783Srmacklem * "entry" structures, but are in the rpc. 1539191783Srmacklem */ 1540191783Srmacklemint 1541191783Srmacklemnfsrvd_readdir(struct nfsrv_descript *nd, int isdgram, 1542191783Srmacklem struct vnode *vp, struct thread *p, struct nfsexstuff *exp) 1543191783Srmacklem{ 1544191783Srmacklem struct dirent *dp; 1545191783Srmacklem u_int32_t *tl; 1546191783Srmacklem int dirlen; 1547191783Srmacklem char *cpos, *cend, *rbuf; 1548191783Srmacklem struct nfsvattr at; 1549191783Srmacklem int nlen, error = 0, getret = 1; 1550191783Srmacklem int siz, cnt, fullsiz, eofflag, ncookies; 1551191783Srmacklem u_int64_t off, toff, verf; 1552191783Srmacklem u_long *cookies = NULL, *cookiep; 1553191783Srmacklem struct uio io; 1554191783Srmacklem struct iovec iv; 1555216691Srmacklem int not_zfs; 1556191783Srmacklem 1557191783Srmacklem if (nd->nd_repstat) { 1558191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1559224086Szack goto out; 1560191783Srmacklem } 1561191783Srmacklem if (nd->nd_flag & ND_NFSV2) { 1562191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1563191783Srmacklem off = fxdr_unsigned(u_quad_t, *tl++); 1564191783Srmacklem } else { 1565191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1566191783Srmacklem off = fxdr_hyper(tl); 1567191783Srmacklem tl += 2; 1568191783Srmacklem verf = fxdr_hyper(tl); 1569191783Srmacklem tl += 2; 1570191783Srmacklem } 1571191783Srmacklem toff = off; 1572191783Srmacklem cnt = fxdr_unsigned(int, *tl); 1573206170Srmacklem if (cnt > NFS_SRVMAXDATA(nd) || cnt < 0) 1574191783Srmacklem cnt = NFS_SRVMAXDATA(nd); 1575191783Srmacklem siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 1576191783Srmacklem fullsiz = siz; 1577191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1578191783Srmacklem nd->nd_repstat = getret = nfsvno_getattr(vp, &at, nd->nd_cred, 1579216693Srmacklem p, 1); 1580191783Srmacklem#if 0 1581191783Srmacklem /* 1582191783Srmacklem * va_filerev is not sufficient as a cookie verifier, 1583191783Srmacklem * since it is not supposed to change when entries are 1584191783Srmacklem * removed/added unless that offset cookies returned to 1585191783Srmacklem * the client are no longer valid. 1586191783Srmacklem */ 1587191783Srmacklem if (!nd->nd_repstat && toff && verf != at.na_filerev) 1588191783Srmacklem nd->nd_repstat = NFSERR_BAD_COOKIE; 1589191783Srmacklem#endif 1590191783Srmacklem } 1591250060Sdes if (!nd->nd_repstat && vp->v_type != VDIR) 1592250060Sdes nd->nd_repstat = NFSERR_NOTDIR; 1593206170Srmacklem if (nd->nd_repstat == 0 && cnt == 0) { 1594206170Srmacklem if (nd->nd_flag & ND_NFSV2) 1595206170Srmacklem /* NFSv2 does not have NFSERR_TOOSMALL */ 1596206170Srmacklem nd->nd_repstat = EPERM; 1597206170Srmacklem else 1598206170Srmacklem nd->nd_repstat = NFSERR_TOOSMALL; 1599206170Srmacklem } 1600191783Srmacklem if (!nd->nd_repstat) 1601200999Srmacklem nd->nd_repstat = nfsvno_accchk(vp, VEXEC, 1602191783Srmacklem nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, 1603200999Srmacklem NFSACCCHK_VPISLOCKED, NULL); 1604191783Srmacklem if (nd->nd_repstat) { 1605191783Srmacklem vput(vp); 1606191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1607191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1608224086Szack goto out; 1609191783Srmacklem } 1610216691Srmacklem not_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs"); 1611191783Srmacklem MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); 1612191783Srmacklemagain: 1613191783Srmacklem eofflag = 0; 1614191783Srmacklem if (cookies) { 1615191783Srmacklem free((caddr_t)cookies, M_TEMP); 1616191783Srmacklem cookies = NULL; 1617191783Srmacklem } 1618191783Srmacklem 1619191783Srmacklem iv.iov_base = rbuf; 1620191783Srmacklem iv.iov_len = siz; 1621191783Srmacklem io.uio_iov = &iv; 1622191783Srmacklem io.uio_iovcnt = 1; 1623191783Srmacklem io.uio_offset = (off_t)off; 1624191783Srmacklem io.uio_resid = siz; 1625191783Srmacklem io.uio_segflg = UIO_SYSSPACE; 1626191783Srmacklem io.uio_rw = UIO_READ; 1627191783Srmacklem io.uio_td = NULL; 1628191783Srmacklem nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies, 1629191783Srmacklem &cookies); 1630191783Srmacklem off = (u_int64_t)io.uio_offset; 1631191783Srmacklem if (io.uio_resid) 1632191783Srmacklem siz -= io.uio_resid; 1633191783Srmacklem 1634191783Srmacklem if (!cookies && !nd->nd_repstat) 1635191783Srmacklem nd->nd_repstat = NFSERR_PERM; 1636191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1637216693Srmacklem getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); 1638191783Srmacklem if (!nd->nd_repstat) 1639191783Srmacklem nd->nd_repstat = getret; 1640191783Srmacklem } 1641191783Srmacklem 1642191783Srmacklem /* 1643191783Srmacklem * Handles the failed cases. nd->nd_repstat == 0 past here. 1644191783Srmacklem */ 1645191783Srmacklem if (nd->nd_repstat) { 1646216692Srmacklem vput(vp); 1647191783Srmacklem free((caddr_t)rbuf, M_TEMP); 1648191783Srmacklem if (cookies) 1649191783Srmacklem free((caddr_t)cookies, M_TEMP); 1650191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1651191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1652224086Szack goto out; 1653191783Srmacklem } 1654191783Srmacklem /* 1655191783Srmacklem * If nothing read, return eof 1656191783Srmacklem * rpc reply 1657191783Srmacklem */ 1658191783Srmacklem if (siz == 0) { 1659216692Srmacklem vput(vp); 1660191783Srmacklem if (nd->nd_flag & ND_NFSV2) { 1661191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1662191783Srmacklem } else { 1663191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1664191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1665191783Srmacklem txdr_hyper(at.na_filerev, tl); 1666191783Srmacklem tl += 2; 1667191783Srmacklem } 1668191783Srmacklem *tl++ = newnfs_false; 1669191783Srmacklem *tl = newnfs_true; 1670191783Srmacklem FREE((caddr_t)rbuf, M_TEMP); 1671191783Srmacklem FREE((caddr_t)cookies, M_TEMP); 1672224086Szack goto out; 1673191783Srmacklem } 1674191783Srmacklem 1675191783Srmacklem /* 1676191783Srmacklem * Check for degenerate cases of nothing useful read. 1677191783Srmacklem * If so go try again 1678191783Srmacklem */ 1679191783Srmacklem cpos = rbuf; 1680191783Srmacklem cend = rbuf + siz; 1681191783Srmacklem dp = (struct dirent *)cpos; 1682191783Srmacklem cookiep = cookies; 1683191783Srmacklem 1684191783Srmacklem /* 1685191783Srmacklem * For some reason FreeBSD's ufs_readdir() chooses to back the 1686191783Srmacklem * directory offset up to a block boundary, so it is necessary to 1687191783Srmacklem * skip over the records that precede the requested offset. This 1688191783Srmacklem * requires the assumption that file offset cookies monotonically 1689191783Srmacklem * increase. 1690216691Srmacklem * Since the offset cookies don't monotonically increase for ZFS, 1691216691Srmacklem * this is not done when ZFS is the file system. 1692191783Srmacklem */ 1693191783Srmacklem while (cpos < cend && ncookies > 0 && 1694191783Srmacklem (dp->d_fileno == 0 || dp->d_type == DT_WHT || 1695216691Srmacklem (not_zfs != 0 && ((u_quad_t)(*cookiep)) <= toff))) { 1696191783Srmacklem cpos += dp->d_reclen; 1697191783Srmacklem dp = (struct dirent *)cpos; 1698191783Srmacklem cookiep++; 1699191783Srmacklem ncookies--; 1700191783Srmacklem } 1701191783Srmacklem if (cpos >= cend || ncookies == 0) { 1702191783Srmacklem siz = fullsiz; 1703191783Srmacklem toff = off; 1704191783Srmacklem goto again; 1705191783Srmacklem } 1706216692Srmacklem vput(vp); 1707191783Srmacklem 1708191783Srmacklem /* 1709191783Srmacklem * dirlen is the size of the reply, including all XDR and must 1710191783Srmacklem * not exceed cnt. For NFSv2, RFC1094 didn't clearly indicate 1711191783Srmacklem * if the XDR should be included in "count", but to be safe, we do. 1712191783Srmacklem * (Include the two booleans at the end of the reply in dirlen now.) 1713191783Srmacklem */ 1714191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1715191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1716191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1717191783Srmacklem txdr_hyper(at.na_filerev, tl); 1718191783Srmacklem dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED; 1719191783Srmacklem } else { 1720191783Srmacklem dirlen = 2 * NFSX_UNSIGNED; 1721191783Srmacklem } 1722191783Srmacklem 1723191783Srmacklem /* Loop through the records and build reply */ 1724191783Srmacklem while (cpos < cend && ncookies > 0) { 1725191783Srmacklem nlen = dp->d_namlen; 1726191783Srmacklem if (dp->d_fileno != 0 && dp->d_type != DT_WHT && 1727191783Srmacklem nlen <= NFS_MAXNAMLEN) { 1728191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1729191783Srmacklem dirlen += (6*NFSX_UNSIGNED + NFSM_RNDUP(nlen)); 1730191783Srmacklem else 1731191783Srmacklem dirlen += (4*NFSX_UNSIGNED + NFSM_RNDUP(nlen)); 1732191783Srmacklem if (dirlen > cnt) { 1733191783Srmacklem eofflag = 0; 1734191783Srmacklem break; 1735191783Srmacklem } 1736191783Srmacklem 1737191783Srmacklem /* 1738191783Srmacklem * Build the directory record xdr from 1739191783Srmacklem * the dirent entry. 1740191783Srmacklem */ 1741191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1742191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 1743191783Srmacklem *tl++ = newnfs_true; 1744191783Srmacklem *tl++ = 0; 1745191783Srmacklem } else { 1746191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1747191783Srmacklem *tl++ = newnfs_true; 1748191783Srmacklem } 1749191783Srmacklem *tl = txdr_unsigned(dp->d_fileno); 1750191783Srmacklem (void) nfsm_strtom(nd, dp->d_name, nlen); 1751191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1752191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1753191783Srmacklem *tl++ = 0; 1754191783Srmacklem } else 1755191783Srmacklem NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED); 1756191783Srmacklem *tl = txdr_unsigned(*cookiep); 1757191783Srmacklem } 1758191783Srmacklem cpos += dp->d_reclen; 1759191783Srmacklem dp = (struct dirent *)cpos; 1760191783Srmacklem cookiep++; 1761191783Srmacklem ncookies--; 1762191783Srmacklem } 1763191783Srmacklem if (cpos < cend) 1764191783Srmacklem eofflag = 0; 1765191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1766191783Srmacklem *tl++ = newnfs_false; 1767191783Srmacklem if (eofflag) 1768191783Srmacklem *tl = newnfs_true; 1769191783Srmacklem else 1770191783Srmacklem *tl = newnfs_false; 1771191783Srmacklem FREE((caddr_t)rbuf, M_TEMP); 1772191783Srmacklem FREE((caddr_t)cookies, M_TEMP); 1773224086Szack 1774224086Szackout: 1775224086Szack NFSEXITCODE2(0, nd); 1776191783Srmacklem return (0); 1777191783Srmacklemnfsmout: 1778191783Srmacklem vput(vp); 1779224086Szack NFSEXITCODE2(error, nd); 1780191783Srmacklem return (error); 1781191783Srmacklem} 1782191783Srmacklem 1783191783Srmacklem/* 1784191783Srmacklem * Readdirplus for V3 and Readdir for V4. 1785191783Srmacklem */ 1786191783Srmacklemint 1787191783Srmacklemnfsrvd_readdirplus(struct nfsrv_descript *nd, int isdgram, 1788191783Srmacklem struct vnode *vp, struct thread *p, struct nfsexstuff *exp) 1789191783Srmacklem{ 1790191783Srmacklem struct dirent *dp; 1791191783Srmacklem u_int32_t *tl; 1792191783Srmacklem int dirlen; 1793191783Srmacklem char *cpos, *cend, *rbuf; 1794191783Srmacklem struct vnode *nvp; 1795191783Srmacklem fhandle_t nfh; 1796191783Srmacklem struct nfsvattr nva, at, *nvap = &nva; 1797191783Srmacklem struct mbuf *mb0, *mb1; 1798191783Srmacklem struct nfsreferral *refp; 1799199715Srmacklem int nlen, r, error = 0, getret = 1, usevget = 1; 1800191783Srmacklem int siz, cnt, fullsiz, eofflag, ncookies, entrycnt; 1801191783Srmacklem caddr_t bpos0, bpos1; 1802191783Srmacklem u_int64_t off, toff, verf; 1803191783Srmacklem u_long *cookies = NULL, *cookiep; 1804191783Srmacklem nfsattrbit_t attrbits, rderrbits, savbits; 1805191783Srmacklem struct uio io; 1806191783Srmacklem struct iovec iv; 1807199715Srmacklem struct componentname cn; 1808220648Srmacklem int at_root, needs_unbusy, not_zfs, supports_nfsv4acls; 1809220645Srmacklem struct mount *mp, *new_mp; 1810220645Srmacklem uint64_t mounted_on_fileno; 1811191783Srmacklem 1812191783Srmacklem if (nd->nd_repstat) { 1813191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1814224086Szack goto out; 1815191783Srmacklem } 1816191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 1817191783Srmacklem off = fxdr_hyper(tl); 1818191783Srmacklem toff = off; 1819191783Srmacklem tl += 2; 1820191783Srmacklem verf = fxdr_hyper(tl); 1821191783Srmacklem tl += 2; 1822191783Srmacklem siz = fxdr_unsigned(int, *tl++); 1823191783Srmacklem cnt = fxdr_unsigned(int, *tl); 1824191783Srmacklem 1825191783Srmacklem /* 1826191783Srmacklem * Use the server's maximum data transfer size as the upper bound 1827191783Srmacklem * on reply datalen. 1828191783Srmacklem */ 1829206170Srmacklem if (cnt > NFS_SRVMAXDATA(nd) || cnt < 0) 1830191783Srmacklem cnt = NFS_SRVMAXDATA(nd); 1831191783Srmacklem 1832191783Srmacklem /* 1833191783Srmacklem * siz is a "hint" of how much directory information (name, fileid, 1834191783Srmacklem * cookie) should be in the reply. At least one client "hints" 0, 1835191783Srmacklem * so I set it to cnt for that case. I also round it up to the 1836191783Srmacklem * next multiple of DIRBLKSIZ. 1837191783Srmacklem */ 1838206170Srmacklem if (siz <= 0) 1839191783Srmacklem siz = cnt; 1840191783Srmacklem siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 1841191783Srmacklem 1842191783Srmacklem if (nd->nd_flag & ND_NFSV4) { 1843191783Srmacklem error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL); 1844191783Srmacklem if (error) 1845191783Srmacklem goto nfsmout; 1846191783Srmacklem NFSSET_ATTRBIT(&savbits, &attrbits); 1847191783Srmacklem NFSCLRNOTFILLABLE_ATTRBIT(&attrbits); 1848191783Srmacklem NFSZERO_ATTRBIT(&rderrbits); 1849191783Srmacklem NFSSETBIT_ATTRBIT(&rderrbits, NFSATTRBIT_RDATTRERROR); 1850191783Srmacklem } else { 1851191783Srmacklem NFSZERO_ATTRBIT(&attrbits); 1852191783Srmacklem } 1853191783Srmacklem fullsiz = siz; 1854216693Srmacklem nd->nd_repstat = getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); 1855191783Srmacklem if (!nd->nd_repstat) { 1856191783Srmacklem if (off && verf != at.na_filerev) { 1857191783Srmacklem /* 1858191783Srmacklem * va_filerev is not sufficient as a cookie verifier, 1859191783Srmacklem * since it is not supposed to change when entries are 1860191783Srmacklem * removed/added unless that offset cookies returned to 1861191783Srmacklem * the client are no longer valid. 1862191783Srmacklem */ 1863191783Srmacklem#if 0 1864191783Srmacklem if (nd->nd_flag & ND_NFSV4) { 1865191783Srmacklem nd->nd_repstat = NFSERR_NOTSAME; 1866191783Srmacklem } else { 1867191783Srmacklem nd->nd_repstat = NFSERR_BAD_COOKIE; 1868191783Srmacklem } 1869191783Srmacklem#endif 1870191783Srmacklem } else if ((nd->nd_flag & ND_NFSV4) && off == 0 && verf != 0) { 1871191783Srmacklem nd->nd_repstat = NFSERR_BAD_COOKIE; 1872191783Srmacklem } 1873191783Srmacklem } 1874191783Srmacklem if (!nd->nd_repstat && vp->v_type != VDIR) 1875191783Srmacklem nd->nd_repstat = NFSERR_NOTDIR; 1876191783Srmacklem if (!nd->nd_repstat && cnt == 0) 1877191783Srmacklem nd->nd_repstat = NFSERR_TOOSMALL; 1878191783Srmacklem if (!nd->nd_repstat) 1879200999Srmacklem nd->nd_repstat = nfsvno_accchk(vp, VEXEC, 1880191783Srmacklem nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, 1881200999Srmacklem NFSACCCHK_VPISLOCKED, NULL); 1882191783Srmacklem if (nd->nd_repstat) { 1883191783Srmacklem vput(vp); 1884191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1885191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1886224086Szack goto out; 1887191783Srmacklem } 1888216691Srmacklem not_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs"); 1889191783Srmacklem 1890191783Srmacklem MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); 1891191783Srmacklemagain: 1892191783Srmacklem eofflag = 0; 1893191783Srmacklem if (cookies) { 1894191783Srmacklem free((caddr_t)cookies, M_TEMP); 1895191783Srmacklem cookies = NULL; 1896191783Srmacklem } 1897191783Srmacklem 1898191783Srmacklem iv.iov_base = rbuf; 1899191783Srmacklem iv.iov_len = siz; 1900191783Srmacklem io.uio_iov = &iv; 1901191783Srmacklem io.uio_iovcnt = 1; 1902191783Srmacklem io.uio_offset = (off_t)off; 1903191783Srmacklem io.uio_resid = siz; 1904191783Srmacklem io.uio_segflg = UIO_SYSSPACE; 1905191783Srmacklem io.uio_rw = UIO_READ; 1906191783Srmacklem io.uio_td = NULL; 1907191783Srmacklem nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies, 1908191783Srmacklem &cookies); 1909191783Srmacklem off = (u_int64_t)io.uio_offset; 1910191783Srmacklem if (io.uio_resid) 1911191783Srmacklem siz -= io.uio_resid; 1912191783Srmacklem 1913216693Srmacklem getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); 1914191783Srmacklem 1915191783Srmacklem if (!cookies && !nd->nd_repstat) 1916191783Srmacklem nd->nd_repstat = NFSERR_PERM; 1917191783Srmacklem if (!nd->nd_repstat) 1918191783Srmacklem nd->nd_repstat = getret; 1919191783Srmacklem if (nd->nd_repstat) { 1920199715Srmacklem vput(vp); 1921191783Srmacklem if (cookies) 1922191783Srmacklem free((caddr_t)cookies, M_TEMP); 1923191783Srmacklem free((caddr_t)rbuf, M_TEMP); 1924191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1925191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1926224086Szack goto out; 1927191783Srmacklem } 1928191783Srmacklem /* 1929191783Srmacklem * If nothing read, return eof 1930191783Srmacklem * rpc reply 1931191783Srmacklem */ 1932191783Srmacklem if (siz == 0) { 1933199715Srmacklem vput(vp); 1934191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1935191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1936191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1937191783Srmacklem txdr_hyper(at.na_filerev, tl); 1938191783Srmacklem tl += 2; 1939191783Srmacklem *tl++ = newnfs_false; 1940191783Srmacklem *tl = newnfs_true; 1941191783Srmacklem free((caddr_t)cookies, M_TEMP); 1942191783Srmacklem free((caddr_t)rbuf, M_TEMP); 1943224086Szack goto out; 1944191783Srmacklem } 1945191783Srmacklem 1946191783Srmacklem /* 1947191783Srmacklem * Check for degenerate cases of nothing useful read. 1948191783Srmacklem * If so go try again 1949191783Srmacklem */ 1950191783Srmacklem cpos = rbuf; 1951191783Srmacklem cend = rbuf + siz; 1952191783Srmacklem dp = (struct dirent *)cpos; 1953191783Srmacklem cookiep = cookies; 1954191783Srmacklem 1955191783Srmacklem /* 1956191783Srmacklem * For some reason FreeBSD's ufs_readdir() chooses to back the 1957191783Srmacklem * directory offset up to a block boundary, so it is necessary to 1958191783Srmacklem * skip over the records that precede the requested offset. This 1959191783Srmacklem * requires the assumption that file offset cookies monotonically 1960191783Srmacklem * increase. 1961216691Srmacklem * Since the offset cookies don't monotonically increase for ZFS, 1962216691Srmacklem * this is not done when ZFS is the file system. 1963191783Srmacklem */ 1964191783Srmacklem while (cpos < cend && ncookies > 0 && 1965191783Srmacklem (dp->d_fileno == 0 || dp->d_type == DT_WHT || 1966216691Srmacklem (not_zfs != 0 && ((u_quad_t)(*cookiep)) <= toff) || 1967191783Srmacklem ((nd->nd_flag & ND_NFSV4) && 1968191783Srmacklem ((dp->d_namlen == 1 && dp->d_name[0] == '.') || 1969191783Srmacklem (dp->d_namlen==2 && dp->d_name[0]=='.' && dp->d_name[1]=='.'))))) { 1970191783Srmacklem cpos += dp->d_reclen; 1971191783Srmacklem dp = (struct dirent *)cpos; 1972191783Srmacklem cookiep++; 1973191783Srmacklem ncookies--; 1974191783Srmacklem } 1975191783Srmacklem if (cpos >= cend || ncookies == 0) { 1976191783Srmacklem siz = fullsiz; 1977191783Srmacklem toff = off; 1978191783Srmacklem goto again; 1979191783Srmacklem } 1980217176Srmacklem 1981217176Srmacklem /* 1982217176Srmacklem * Busy the file system so that the mount point won't go away 1983217176Srmacklem * and, as such, VFS_VGET() can be used safely. 1984217176Srmacklem */ 1985217176Srmacklem mp = vp->v_mount; 1986217176Srmacklem vfs_ref(mp); 1987224082Szack NFSVOPUNLOCK(vp, 0); 1988217176Srmacklem nd->nd_repstat = vfs_busy(mp, 0); 1989217176Srmacklem vfs_rel(mp); 1990217176Srmacklem if (nd->nd_repstat != 0) { 1991217176Srmacklem vrele(vp); 1992217176Srmacklem free(cookies, M_TEMP); 1993217176Srmacklem free(rbuf, M_TEMP); 1994217176Srmacklem if (nd->nd_flag & ND_NFSV3) 1995217176Srmacklem nfsrv_postopattr(nd, getret, &at); 1996224086Szack goto out; 1997217176Srmacklem } 1998191783Srmacklem 1999191783Srmacklem /* 2000260173Srmacklem * Check to see if entries in this directory can be safely acquired 2001260173Srmacklem * via VFS_VGET() or if a switch to VOP_LOOKUP() is required. 2002260173Srmacklem * ZFS snapshot directories need VOP_LOOKUP(), so that any 2003260173Srmacklem * automount of the snapshot directory that is required will 2004260173Srmacklem * be done. 2005260173Srmacklem * This needs to be done here for NFSv4, since NFSv4 never does 2006260173Srmacklem * a VFS_VGET() for "." or "..". 2007260173Srmacklem */ 2008260173Srmacklem if (not_zfs == 0) { 2009260173Srmacklem r = VFS_VGET(mp, at.na_fileid, LK_SHARED, &nvp); 2010260173Srmacklem if (r == EOPNOTSUPP) { 2011260173Srmacklem usevget = 0; 2012260173Srmacklem cn.cn_nameiop = LOOKUP; 2013260173Srmacklem cn.cn_lkflags = LK_SHARED | LK_RETRY; 2014260173Srmacklem cn.cn_cred = nd->nd_cred; 2015260173Srmacklem cn.cn_thread = p; 2016260173Srmacklem } else if (r == 0) 2017260173Srmacklem vput(nvp); 2018260173Srmacklem } 2019260173Srmacklem 2020260173Srmacklem /* 2021191783Srmacklem * Save this position, in case there is an error before one entry 2022191783Srmacklem * is created. 2023191783Srmacklem */ 2024191783Srmacklem mb0 = nd->nd_mb; 2025191783Srmacklem bpos0 = nd->nd_bpos; 2026191783Srmacklem 2027191783Srmacklem /* 2028191783Srmacklem * Fill in the first part of the reply. 2029191783Srmacklem * dirlen is the reply length in bytes and cannot exceed cnt. 2030191783Srmacklem * (Include the two booleans at the end of the reply in dirlen now, 2031191783Srmacklem * so we recognize when we have exceeded cnt.) 2032191783Srmacklem */ 2033191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 2034191783Srmacklem dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED; 2035191783Srmacklem nfsrv_postopattr(nd, getret, &at); 2036191783Srmacklem } else { 2037191783Srmacklem dirlen = NFSX_VERF + 2 * NFSX_UNSIGNED; 2038191783Srmacklem } 2039191783Srmacklem NFSM_BUILD(tl, u_int32_t *, NFSX_VERF); 2040191783Srmacklem txdr_hyper(at.na_filerev, tl); 2041191783Srmacklem 2042191783Srmacklem /* 2043191783Srmacklem * Save this position, in case there is an empty reply needed. 2044191783Srmacklem */ 2045191783Srmacklem mb1 = nd->nd_mb; 2046191783Srmacklem bpos1 = nd->nd_bpos; 2047191783Srmacklem 2048191783Srmacklem /* Loop through the records and build reply */ 2049191783Srmacklem entrycnt = 0; 2050191783Srmacklem while (cpos < cend && ncookies > 0 && dirlen < cnt) { 2051191783Srmacklem nlen = dp->d_namlen; 2052191783Srmacklem if (dp->d_fileno != 0 && dp->d_type != DT_WHT && 2053191783Srmacklem nlen <= NFS_MAXNAMLEN && 2054191783Srmacklem ((nd->nd_flag & ND_NFSV3) || nlen > 2 || 2055191783Srmacklem (nlen==2 && (dp->d_name[0]!='.' || dp->d_name[1]!='.')) 2056191783Srmacklem || (nlen == 1 && dp->d_name[0] != '.'))) { 2057191783Srmacklem /* 2058191783Srmacklem * Save the current position in the reply, in case 2059191783Srmacklem * this entry exceeds cnt. 2060191783Srmacklem */ 2061191783Srmacklem mb1 = nd->nd_mb; 2062191783Srmacklem bpos1 = nd->nd_bpos; 2063191783Srmacklem 2064191783Srmacklem /* 2065191783Srmacklem * For readdir_and_lookup get the vnode using 2066191783Srmacklem * the file number. 2067191783Srmacklem */ 2068191783Srmacklem nvp = NULL; 2069191783Srmacklem refp = NULL; 2070191783Srmacklem r = 0; 2071220645Srmacklem at_root = 0; 2072220645Srmacklem needs_unbusy = 0; 2073220645Srmacklem new_mp = mp; 2074220645Srmacklem mounted_on_fileno = (uint64_t)dp->d_fileno; 2075191783Srmacklem if ((nd->nd_flag & ND_NFSV3) || 2076191783Srmacklem NFSNONZERO_ATTRBIT(&savbits)) { 2077191783Srmacklem if (nd->nd_flag & ND_NFSV4) 2078191783Srmacklem refp = nfsv4root_getreferral(NULL, 2079191783Srmacklem vp, dp->d_fileno); 2080199715Srmacklem if (refp == NULL) { 2081199715Srmacklem if (usevget) 2082217176Srmacklem r = VFS_VGET(mp, dp->d_fileno, 2083217176Srmacklem LK_SHARED, &nvp); 2084199715Srmacklem else 2085199715Srmacklem r = EOPNOTSUPP; 2086199715Srmacklem if (r == EOPNOTSUPP) { 2087199715Srmacklem if (usevget) { 2088199715Srmacklem usevget = 0; 2089199715Srmacklem cn.cn_nameiop = LOOKUP; 2090199715Srmacklem cn.cn_lkflags = 2091216700Srmacklem LK_SHARED | 2092199715Srmacklem LK_RETRY; 2093199715Srmacklem cn.cn_cred = 2094199715Srmacklem nd->nd_cred; 2095199715Srmacklem cn.cn_thread = p; 2096199715Srmacklem } 2097199715Srmacklem cn.cn_nameptr = dp->d_name; 2098199715Srmacklem cn.cn_namelen = nlen; 2099199715Srmacklem cn.cn_flags = ISLASTCN | 2100199715Srmacklem NOFOLLOW | LOCKLEAF | 2101199715Srmacklem MPSAFE; 2102199715Srmacklem if (nlen == 2 && 2103199715Srmacklem dp->d_name[0] == '.' && 2104199715Srmacklem dp->d_name[1] == '.') 2105199715Srmacklem cn.cn_flags |= 2106199715Srmacklem ISDOTDOT; 2107224081Szack if (NFSVOPLOCK(vp, LK_SHARED) 2108216692Srmacklem != 0) { 2109216692Srmacklem nd->nd_repstat = EPERM; 2110216692Srmacklem break; 2111216692Srmacklem } 2112214149Srmacklem if ((vp->v_vflag & VV_ROOT) != 0 2113214149Srmacklem && (cn.cn_flags & ISDOTDOT) 2114214149Srmacklem != 0) { 2115214149Srmacklem vref(vp); 2116214149Srmacklem nvp = vp; 2117214149Srmacklem r = 0; 2118220507Srmacklem } else { 2119214149Srmacklem r = VOP_LOOKUP(vp, &nvp, 2120214149Srmacklem &cn); 2121220507Srmacklem if (vp != nvp) 2122224082Szack NFSVOPUNLOCK(vp, 2123220507Srmacklem 0); 2124220507Srmacklem } 2125199715Srmacklem } 2126220645Srmacklem 2127220645Srmacklem /* 2128220645Srmacklem * For NFSv4, check to see if nvp is 2129220645Srmacklem * a mount point and get the mount 2130220645Srmacklem * point vnode, as required. 2131220645Srmacklem */ 2132220645Srmacklem if (r == 0 && 2133220645Srmacklem nfsrv_enable_crossmntpt != 0 && 2134220645Srmacklem (nd->nd_flag & ND_NFSV4) != 0 && 2135220645Srmacklem nvp->v_type == VDIR && 2136220645Srmacklem nvp->v_mountedhere != NULL) { 2137220645Srmacklem new_mp = nvp->v_mountedhere; 2138220645Srmacklem r = vfs_busy(new_mp, 0); 2139220645Srmacklem vput(nvp); 2140220645Srmacklem nvp = NULL; 2141220645Srmacklem if (r == 0) { 2142220645Srmacklem r = VFS_ROOT(new_mp, 2143220645Srmacklem LK_SHARED, &nvp); 2144220645Srmacklem needs_unbusy = 1; 2145220645Srmacklem if (r == 0) 2146220645Srmacklem at_root = 1; 2147220645Srmacklem } 2148220645Srmacklem } 2149199715Srmacklem } 2150191783Srmacklem if (!r) { 2151191783Srmacklem if (refp == NULL && 2152191783Srmacklem ((nd->nd_flag & ND_NFSV3) || 2153191783Srmacklem NFSNONZERO_ATTRBIT(&attrbits))) { 2154191783Srmacklem r = nfsvno_getfh(nvp, &nfh, p); 2155191783Srmacklem if (!r) 2156191783Srmacklem r = nfsvno_getattr(nvp, nvap, 2157216693Srmacklem nd->nd_cred, p, 1); 2158260173Srmacklem if (r == 0 && not_zfs == 0 && 2159260173Srmacklem nfsrv_enable_crossmntpt != 0 && 2160260173Srmacklem (nd->nd_flag & ND_NFSV4) != 0 && 2161260173Srmacklem nvp->v_type == VDIR && 2162260173Srmacklem vp->v_mount != nvp->v_mount) { 2163260173Srmacklem /* 2164260173Srmacklem * For a ZFS snapshot, there is a 2165260173Srmacklem * pseudo mount that does not set 2166260173Srmacklem * v_mountedhere, so it needs to 2167260173Srmacklem * be detected via a different 2168260173Srmacklem * mount structure. 2169260173Srmacklem */ 2170260173Srmacklem at_root = 1; 2171260173Srmacklem if (new_mp == mp) 2172260173Srmacklem new_mp = nvp->v_mount; 2173260173Srmacklem } 2174191783Srmacklem } 2175191783Srmacklem } else { 2176191783Srmacklem nvp = NULL; 2177191783Srmacklem } 2178191783Srmacklem if (r) { 2179191783Srmacklem if (!NFSISSET_ATTRBIT(&attrbits, 2180191783Srmacklem NFSATTRBIT_RDATTRERROR)) { 2181191783Srmacklem if (nvp != NULL) 2182191783Srmacklem vput(nvp); 2183220645Srmacklem if (needs_unbusy != 0) 2184220645Srmacklem vfs_unbusy(new_mp); 2185191783Srmacklem nd->nd_repstat = r; 2186191783Srmacklem break; 2187191783Srmacklem } 2188191783Srmacklem } 2189191783Srmacklem } 2190191783Srmacklem 2191191783Srmacklem /* 2192191783Srmacklem * Build the directory record xdr 2193191783Srmacklem */ 2194191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 2195191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 2196191783Srmacklem *tl++ = newnfs_true; 2197191783Srmacklem *tl++ = 0; 2198191783Srmacklem *tl = txdr_unsigned(dp->d_fileno); 2199191783Srmacklem dirlen += nfsm_strtom(nd, dp->d_name, nlen); 2200191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2201191783Srmacklem *tl++ = 0; 2202191783Srmacklem *tl = txdr_unsigned(*cookiep); 2203191783Srmacklem nfsrv_postopattr(nd, 0, nvap); 2204191783Srmacklem dirlen += nfsm_fhtom(nd,(u_int8_t *)&nfh,0,1); 2205191783Srmacklem dirlen += (5*NFSX_UNSIGNED+NFSX_V3POSTOPATTR); 2206191783Srmacklem if (nvp != NULL) 2207191783Srmacklem vput(nvp); 2208191783Srmacklem } else { 2209191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 2210191783Srmacklem *tl++ = newnfs_true; 2211191783Srmacklem *tl++ = 0; 2212191783Srmacklem *tl = txdr_unsigned(*cookiep); 2213191783Srmacklem dirlen += nfsm_strtom(nd, dp->d_name, nlen); 2214220648Srmacklem if (nvp != NULL) { 2215220648Srmacklem supports_nfsv4acls = 2216220648Srmacklem nfs_supportsnfsv4acls(nvp); 2217224082Szack NFSVOPUNLOCK(nvp, 0); 2218220648Srmacklem } else 2219220648Srmacklem supports_nfsv4acls = 0; 2220191783Srmacklem if (refp != NULL) { 2221191783Srmacklem dirlen += nfsrv_putreferralattr(nd, 2222191783Srmacklem &savbits, refp, 0, 2223191783Srmacklem &nd->nd_repstat); 2224191783Srmacklem if (nd->nd_repstat) { 2225191783Srmacklem if (nvp != NULL) 2226191783Srmacklem vrele(nvp); 2227220645Srmacklem if (needs_unbusy != 0) 2228220645Srmacklem vfs_unbusy(new_mp); 2229191783Srmacklem break; 2230191783Srmacklem } 2231191783Srmacklem } else if (r) { 2232220645Srmacklem dirlen += nfsvno_fillattr(nd, new_mp, 2233220645Srmacklem nvp, nvap, &nfh, r, &rderrbits, 2234220648Srmacklem nd->nd_cred, p, isdgram, 0, 2235220648Srmacklem supports_nfsv4acls, at_root, 2236220645Srmacklem mounted_on_fileno); 2237191783Srmacklem } else { 2238220645Srmacklem dirlen += nfsvno_fillattr(nd, new_mp, 2239220645Srmacklem nvp, nvap, &nfh, r, &attrbits, 2240220648Srmacklem nd->nd_cred, p, isdgram, 0, 2241220648Srmacklem supports_nfsv4acls, at_root, 2242220645Srmacklem mounted_on_fileno); 2243191783Srmacklem } 2244191783Srmacklem if (nvp != NULL) 2245191783Srmacklem vrele(nvp); 2246191783Srmacklem dirlen += (3 * NFSX_UNSIGNED); 2247191783Srmacklem } 2248220645Srmacklem if (needs_unbusy != 0) 2249220645Srmacklem vfs_unbusy(new_mp); 2250191783Srmacklem if (dirlen <= cnt) 2251191783Srmacklem entrycnt++; 2252191783Srmacklem } 2253191783Srmacklem cpos += dp->d_reclen; 2254191783Srmacklem dp = (struct dirent *)cpos; 2255191783Srmacklem cookiep++; 2256191783Srmacklem ncookies--; 2257191783Srmacklem } 2258216692Srmacklem vrele(vp); 2259217176Srmacklem vfs_unbusy(mp); 2260191783Srmacklem 2261191783Srmacklem /* 2262191783Srmacklem * If dirlen > cnt, we must strip off the last entry. If that 2263191783Srmacklem * results in an empty reply, report NFSERR_TOOSMALL. 2264191783Srmacklem */ 2265191783Srmacklem if (dirlen > cnt || nd->nd_repstat) { 2266191783Srmacklem if (!nd->nd_repstat && entrycnt == 0) 2267191783Srmacklem nd->nd_repstat = NFSERR_TOOSMALL; 2268191783Srmacklem if (nd->nd_repstat) 2269191783Srmacklem newnfs_trimtrailing(nd, mb0, bpos0); 2270191783Srmacklem else 2271191783Srmacklem newnfs_trimtrailing(nd, mb1, bpos1); 2272191783Srmacklem eofflag = 0; 2273191783Srmacklem } else if (cpos < cend) 2274191783Srmacklem eofflag = 0; 2275191783Srmacklem if (!nd->nd_repstat) { 2276191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2277191783Srmacklem *tl++ = newnfs_false; 2278191783Srmacklem if (eofflag) 2279191783Srmacklem *tl = newnfs_true; 2280191783Srmacklem else 2281191783Srmacklem *tl = newnfs_false; 2282191783Srmacklem } 2283191783Srmacklem FREE((caddr_t)cookies, M_TEMP); 2284191783Srmacklem FREE((caddr_t)rbuf, M_TEMP); 2285224086Szack 2286224086Szackout: 2287224086Szack NFSEXITCODE2(0, nd); 2288191783Srmacklem return (0); 2289191783Srmacklemnfsmout: 2290191783Srmacklem vput(vp); 2291224086Szack NFSEXITCODE2(error, nd); 2292191783Srmacklem return (error); 2293191783Srmacklem} 2294191783Srmacklem 2295191783Srmacklem/* 2296191783Srmacklem * Get the settable attributes out of the mbuf list. 2297191783Srmacklem * (Return 0 or EBADRPC) 2298191783Srmacklem */ 2299191783Srmacklemint 2300191783Srmacklemnfsrv_sattr(struct nfsrv_descript *nd, struct nfsvattr *nvap, 2301191783Srmacklem nfsattrbit_t *attrbitp, NFSACL_T *aclp, struct thread *p) 2302191783Srmacklem{ 2303191783Srmacklem u_int32_t *tl; 2304191783Srmacklem struct nfsv2_sattr *sp; 2305191783Srmacklem int error = 0, toclient = 0; 2306191783Srmacklem 2307191783Srmacklem switch (nd->nd_flag & (ND_NFSV2 | ND_NFSV3 | ND_NFSV4)) { 2308191783Srmacklem case ND_NFSV2: 2309191783Srmacklem NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2310191783Srmacklem /* 2311191783Srmacklem * Some old clients didn't fill in the high order 16bits. 2312191783Srmacklem * --> check the low order 2 bytes for 0xffff 2313191783Srmacklem */ 2314191783Srmacklem if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 2315191783Srmacklem nvap->na_mode = nfstov_mode(sp->sa_mode); 2316191783Srmacklem if (sp->sa_uid != newnfs_xdrneg1) 2317191783Srmacklem nvap->na_uid = fxdr_unsigned(uid_t, sp->sa_uid); 2318191783Srmacklem if (sp->sa_gid != newnfs_xdrneg1) 2319191783Srmacklem nvap->na_gid = fxdr_unsigned(gid_t, sp->sa_gid); 2320191783Srmacklem if (sp->sa_size != newnfs_xdrneg1) 2321191783Srmacklem nvap->na_size = fxdr_unsigned(u_quad_t, sp->sa_size); 2322191783Srmacklem if (sp->sa_atime.nfsv2_sec != newnfs_xdrneg1) { 2323191783Srmacklem#ifdef notyet 2324191783Srmacklem fxdr_nfsv2time(&sp->sa_atime, &nvap->na_atime); 2325191783Srmacklem#else 2326191783Srmacklem nvap->na_atime.tv_sec = 2327191783Srmacklem fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 2328191783Srmacklem nvap->na_atime.tv_nsec = 0; 2329191783Srmacklem#endif 2330191783Srmacklem } 2331191783Srmacklem if (sp->sa_mtime.nfsv2_sec != newnfs_xdrneg1) 2332191783Srmacklem fxdr_nfsv2time(&sp->sa_mtime, &nvap->na_mtime); 2333191783Srmacklem break; 2334191783Srmacklem case ND_NFSV3: 2335191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2336191783Srmacklem if (*tl == newnfs_true) { 2337191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2338191783Srmacklem nvap->na_mode = nfstov_mode(*tl); 2339191783Srmacklem } 2340191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2341191783Srmacklem if (*tl == newnfs_true) { 2342191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2343191783Srmacklem nvap->na_uid = fxdr_unsigned(uid_t, *tl); 2344191783Srmacklem } 2345191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2346191783Srmacklem if (*tl == newnfs_true) { 2347191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2348191783Srmacklem nvap->na_gid = fxdr_unsigned(gid_t, *tl); 2349191783Srmacklem } 2350191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2351191783Srmacklem if (*tl == newnfs_true) { 2352191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2353191783Srmacklem nvap->na_size = fxdr_hyper(tl); 2354191783Srmacklem } 2355191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2356191783Srmacklem switch (fxdr_unsigned(int, *tl)) { 2357191783Srmacklem case NFSV3SATTRTIME_TOCLIENT: 2358191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2359191783Srmacklem fxdr_nfsv3time(tl, &nvap->na_atime); 2360191783Srmacklem toclient = 1; 2361191783Srmacklem break; 2362191783Srmacklem case NFSV3SATTRTIME_TOSERVER: 2363247502Sjhb vfs_timestamp(&nvap->na_atime); 2364191783Srmacklem nvap->na_vaflags |= VA_UTIMES_NULL; 2365191783Srmacklem break; 2366191783Srmacklem }; 2367191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2368191783Srmacklem switch (fxdr_unsigned(int, *tl)) { 2369191783Srmacklem case NFSV3SATTRTIME_TOCLIENT: 2370191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2371191783Srmacklem fxdr_nfsv3time(tl, &nvap->na_mtime); 2372191783Srmacklem nvap->na_vaflags &= ~VA_UTIMES_NULL; 2373191783Srmacklem break; 2374191783Srmacklem case NFSV3SATTRTIME_TOSERVER: 2375247502Sjhb vfs_timestamp(&nvap->na_mtime); 2376191783Srmacklem if (!toclient) 2377191783Srmacklem nvap->na_vaflags |= VA_UTIMES_NULL; 2378191783Srmacklem break; 2379191783Srmacklem }; 2380191783Srmacklem break; 2381191783Srmacklem case ND_NFSV4: 2382191783Srmacklem error = nfsv4_sattr(nd, nvap, attrbitp, aclp, p); 2383191783Srmacklem }; 2384191783Srmacklemnfsmout: 2385224086Szack NFSEXITCODE2(error, nd); 2386191783Srmacklem return (error); 2387191783Srmacklem} 2388191783Srmacklem 2389191783Srmacklem/* 2390191783Srmacklem * Handle the setable attributes for V4. 2391191783Srmacklem * Returns NFSERR_BADXDR if it can't be parsed, 0 otherwise. 2392191783Srmacklem */ 2393191783Srmacklemint 2394191783Srmacklemnfsv4_sattr(struct nfsrv_descript *nd, struct nfsvattr *nvap, 2395191783Srmacklem nfsattrbit_t *attrbitp, NFSACL_T *aclp, struct thread *p) 2396191783Srmacklem{ 2397191783Srmacklem u_int32_t *tl; 2398191783Srmacklem int attrsum = 0; 2399191783Srmacklem int i, j; 2400191783Srmacklem int error, attrsize, bitpos, aclsize, aceerr, retnotsup = 0; 2401191783Srmacklem int toclient = 0; 2402191783Srmacklem u_char *cp, namestr[NFSV4_SMALLSTR + 1]; 2403191783Srmacklem uid_t uid; 2404191783Srmacklem gid_t gid; 2405191783Srmacklem 2406191783Srmacklem error = nfsrv_getattrbits(nd, attrbitp, NULL, &retnotsup); 2407191783Srmacklem if (error) 2408224086Szack goto nfsmout; 2409191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2410191783Srmacklem attrsize = fxdr_unsigned(int, *tl); 2411191783Srmacklem 2412191783Srmacklem /* 2413191783Srmacklem * Loop around getting the setable attributes. If an unsupported 2414191783Srmacklem * one is found, set nd_repstat == NFSERR_ATTRNOTSUPP and return. 2415191783Srmacklem */ 2416191783Srmacklem if (retnotsup) { 2417191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2418191783Srmacklem bitpos = NFSATTRBIT_MAX; 2419191783Srmacklem } else { 2420191783Srmacklem bitpos = 0; 2421191783Srmacklem } 2422191783Srmacklem for (; bitpos < NFSATTRBIT_MAX; bitpos++) { 2423191783Srmacklem if (attrsum > attrsize) { 2424191783Srmacklem error = NFSERR_BADXDR; 2425191783Srmacklem goto nfsmout; 2426191783Srmacklem } 2427191783Srmacklem if (NFSISSET_ATTRBIT(attrbitp, bitpos)) 2428191783Srmacklem switch (bitpos) { 2429191783Srmacklem case NFSATTRBIT_SIZE: 2430191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER); 2431191783Srmacklem nvap->na_size = fxdr_hyper(tl); 2432191783Srmacklem attrsum += NFSX_HYPER; 2433191783Srmacklem break; 2434191783Srmacklem case NFSATTRBIT_ACL: 2435191783Srmacklem error = nfsrv_dissectacl(nd, aclp, &aceerr, &aclsize, 2436191783Srmacklem p); 2437191783Srmacklem if (error) 2438191783Srmacklem goto nfsmout; 2439191783Srmacklem if (aceerr && !nd->nd_repstat) 2440224077Szack nd->nd_repstat = aceerr; 2441191783Srmacklem attrsum += aclsize; 2442191783Srmacklem break; 2443191783Srmacklem case NFSATTRBIT_ARCHIVE: 2444191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2445191783Srmacklem if (!nd->nd_repstat) 2446191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2447191783Srmacklem attrsum += NFSX_UNSIGNED; 2448191783Srmacklem break; 2449191783Srmacklem case NFSATTRBIT_HIDDEN: 2450191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2451191783Srmacklem if (!nd->nd_repstat) 2452191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2453191783Srmacklem attrsum += NFSX_UNSIGNED; 2454191783Srmacklem break; 2455191783Srmacklem case NFSATTRBIT_MIMETYPE: 2456191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2457191783Srmacklem i = fxdr_unsigned(int, *tl); 2458191783Srmacklem error = nfsm_advance(nd, NFSM_RNDUP(i), -1); 2459191783Srmacklem if (error) 2460191783Srmacklem goto nfsmout; 2461191783Srmacklem if (!nd->nd_repstat) 2462191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2463191783Srmacklem attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(i)); 2464191783Srmacklem break; 2465191783Srmacklem case NFSATTRBIT_MODE: 2466191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2467191783Srmacklem nvap->na_mode = nfstov_mode(*tl); 2468191783Srmacklem attrsum += NFSX_UNSIGNED; 2469191783Srmacklem break; 2470191783Srmacklem case NFSATTRBIT_OWNER: 2471191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2472191783Srmacklem j = fxdr_unsigned(int, *tl); 2473224086Szack if (j < 0) { 2474224086Szack error = NFSERR_BADXDR; 2475224086Szack goto nfsmout; 2476224086Szack } 2477191783Srmacklem if (j > NFSV4_SMALLSTR) 2478191783Srmacklem cp = malloc(j + 1, M_NFSSTRING, M_WAITOK); 2479191783Srmacklem else 2480191783Srmacklem cp = namestr; 2481191783Srmacklem error = nfsrv_mtostr(nd, cp, j); 2482191783Srmacklem if (error) { 2483191783Srmacklem if (j > NFSV4_SMALLSTR) 2484191783Srmacklem free(cp, M_NFSSTRING); 2485224086Szack goto nfsmout; 2486191783Srmacklem } 2487191783Srmacklem if (!nd->nd_repstat) { 2488241194Srmacklem nd->nd_repstat = nfsv4_strtouid(nd, cp, j, &uid, 2489241194Srmacklem p); 2490191783Srmacklem if (!nd->nd_repstat) 2491191783Srmacklem nvap->na_uid = uid; 2492191783Srmacklem } 2493191783Srmacklem if (j > NFSV4_SMALLSTR) 2494191783Srmacklem free(cp, M_NFSSTRING); 2495191783Srmacklem attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j)); 2496191783Srmacklem break; 2497191783Srmacklem case NFSATTRBIT_OWNERGROUP: 2498191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2499191783Srmacklem j = fxdr_unsigned(int, *tl); 2500224086Szack if (j < 0) { 2501224086Szack error = NFSERR_BADXDR; 2502224086Szack goto nfsmout; 2503224086Szack } 2504191783Srmacklem if (j > NFSV4_SMALLSTR) 2505191783Srmacklem cp = malloc(j + 1, M_NFSSTRING, M_WAITOK); 2506191783Srmacklem else 2507191783Srmacklem cp = namestr; 2508191783Srmacklem error = nfsrv_mtostr(nd, cp, j); 2509191783Srmacklem if (error) { 2510191783Srmacklem if (j > NFSV4_SMALLSTR) 2511191783Srmacklem free(cp, M_NFSSTRING); 2512224086Szack goto nfsmout; 2513191783Srmacklem } 2514191783Srmacklem if (!nd->nd_repstat) { 2515241194Srmacklem nd->nd_repstat = nfsv4_strtogid(nd, cp, j, &gid, 2516241194Srmacklem p); 2517191783Srmacklem if (!nd->nd_repstat) 2518191783Srmacklem nvap->na_gid = gid; 2519191783Srmacklem } 2520191783Srmacklem if (j > NFSV4_SMALLSTR) 2521191783Srmacklem free(cp, M_NFSSTRING); 2522191783Srmacklem attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j)); 2523191783Srmacklem break; 2524191783Srmacklem case NFSATTRBIT_SYSTEM: 2525191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2526191783Srmacklem if (!nd->nd_repstat) 2527191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2528191783Srmacklem attrsum += NFSX_UNSIGNED; 2529191783Srmacklem break; 2530191783Srmacklem case NFSATTRBIT_TIMEACCESSSET: 2531191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2532191783Srmacklem attrsum += NFSX_UNSIGNED; 2533191783Srmacklem if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) { 2534191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); 2535191783Srmacklem fxdr_nfsv4time(tl, &nvap->na_atime); 2536191783Srmacklem toclient = 1; 2537191783Srmacklem attrsum += NFSX_V4TIME; 2538191783Srmacklem } else { 2539247502Sjhb vfs_timestamp(&nvap->na_atime); 2540191783Srmacklem nvap->na_vaflags |= VA_UTIMES_NULL; 2541191783Srmacklem } 2542191783Srmacklem break; 2543191783Srmacklem case NFSATTRBIT_TIMEBACKUP: 2544191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); 2545191783Srmacklem if (!nd->nd_repstat) 2546191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2547191783Srmacklem attrsum += NFSX_V4TIME; 2548191783Srmacklem break; 2549191783Srmacklem case NFSATTRBIT_TIMECREATE: 2550191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); 2551191783Srmacklem if (!nd->nd_repstat) 2552191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2553191783Srmacklem attrsum += NFSX_V4TIME; 2554191783Srmacklem break; 2555191783Srmacklem case NFSATTRBIT_TIMEMODIFYSET: 2556191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2557191783Srmacklem attrsum += NFSX_UNSIGNED; 2558191783Srmacklem if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) { 2559191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); 2560191783Srmacklem fxdr_nfsv4time(tl, &nvap->na_mtime); 2561191783Srmacklem nvap->na_vaflags &= ~VA_UTIMES_NULL; 2562191783Srmacklem attrsum += NFSX_V4TIME; 2563191783Srmacklem } else { 2564247502Sjhb vfs_timestamp(&nvap->na_mtime); 2565191783Srmacklem if (!toclient) 2566191783Srmacklem nvap->na_vaflags |= VA_UTIMES_NULL; 2567191783Srmacklem } 2568191783Srmacklem break; 2569191783Srmacklem default: 2570191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2571191783Srmacklem /* 2572191783Srmacklem * set bitpos so we drop out of the loop. 2573191783Srmacklem */ 2574191783Srmacklem bitpos = NFSATTRBIT_MAX; 2575191783Srmacklem break; 2576191783Srmacklem }; 2577191783Srmacklem } 2578191783Srmacklem 2579191783Srmacklem /* 2580191783Srmacklem * some clients pad the attrlist, so we need to skip over the 2581191783Srmacklem * padding. 2582191783Srmacklem */ 2583191783Srmacklem if (attrsum > attrsize) { 2584191783Srmacklem error = NFSERR_BADXDR; 2585191783Srmacklem } else { 2586191783Srmacklem attrsize = NFSM_RNDUP(attrsize); 2587191783Srmacklem if (attrsum < attrsize) 2588191783Srmacklem error = nfsm_advance(nd, attrsize - attrsum, -1); 2589191783Srmacklem } 2590191783Srmacklemnfsmout: 2591224086Szack NFSEXITCODE2(error, nd); 2592191783Srmacklem return (error); 2593191783Srmacklem} 2594191783Srmacklem 2595191783Srmacklem/* 2596191783Srmacklem * Check/setup export credentials. 2597191783Srmacklem */ 2598191783Srmacklemint 2599191783Srmacklemnfsd_excred(struct nfsrv_descript *nd, struct nfsexstuff *exp, 2600191783Srmacklem struct ucred *credanon) 2601191783Srmacklem{ 2602191783Srmacklem int error = 0; 2603191783Srmacklem 2604191783Srmacklem /* 2605191783Srmacklem * Check/setup credentials. 2606191783Srmacklem */ 2607191783Srmacklem if (nd->nd_flag & ND_GSS) 2608192121Srmacklem exp->nes_exflag &= ~MNT_EXPORTANON; 2609191783Srmacklem 2610191783Srmacklem /* 2611192121Srmacklem * Check to see if the operation is allowed for this security flavor. 2612191783Srmacklem * RFC2623 suggests that the NFSv3 Fsinfo RPC be allowed to 2613191783Srmacklem * AUTH_NONE or AUTH_SYS for file systems requiring RPCSEC_GSS. 2614192121Srmacklem * Also, allow Secinfo, so that it can acquire the correct flavor(s). 2615191783Srmacklem */ 2616192121Srmacklem if (nfsvno_testexp(nd, exp) && 2617192121Srmacklem nd->nd_procnum != NFSV4OP_SECINFO && 2618191783Srmacklem nd->nd_procnum != NFSPROC_FSINFO) { 2619191783Srmacklem if (nd->nd_flag & ND_NFSV4) 2620191783Srmacklem error = NFSERR_WRONGSEC; 2621191783Srmacklem else 2622191783Srmacklem error = (NFSERR_AUTHERR | AUTH_TOOWEAK); 2623224086Szack goto out; 2624191783Srmacklem } 2625191783Srmacklem 2626191783Srmacklem /* 2627191783Srmacklem * Check to see if the file system is exported V4 only. 2628191783Srmacklem */ 2629224086Szack if (NFSVNO_EXV4ONLY(exp) && !(nd->nd_flag & ND_NFSV4)) { 2630224086Szack error = NFSERR_PROGNOTV4; 2631224086Szack goto out; 2632224086Szack } 2633191783Srmacklem 2634191783Srmacklem /* 2635191783Srmacklem * Now, map the user credentials. 2636191783Srmacklem * (Note that ND_AUTHNONE will only be set for an NFSv3 2637191783Srmacklem * Fsinfo RPC. If set for anything else, this code might need 2638191783Srmacklem * to change.) 2639191783Srmacklem */ 2640191783Srmacklem if (NFSVNO_EXPORTED(exp) && 2641191783Srmacklem ((!(nd->nd_flag & ND_GSS) && nd->nd_cred->cr_uid == 0) || 2642191783Srmacklem NFSVNO_EXPORTANON(exp) || 2643191783Srmacklem (nd->nd_flag & ND_AUTHNONE))) { 2644191783Srmacklem nd->nd_cred->cr_uid = credanon->cr_uid; 2645191783Srmacklem nd->nd_cred->cr_gid = credanon->cr_gid; 2646194498Sbrooks crsetgroups(nd->nd_cred, credanon->cr_ngroups, 2647194498Sbrooks credanon->cr_groups); 2648191783Srmacklem } 2649224086Szack 2650224086Szackout: 2651224086Szack NFSEXITCODE2(error, nd); 2652224086Szack return (error); 2653191783Srmacklem} 2654191783Srmacklem 2655191783Srmacklem/* 2656191783Srmacklem * Check exports. 2657191783Srmacklem */ 2658191783Srmacklemint 2659191783Srmacklemnfsvno_checkexp(struct mount *mp, struct sockaddr *nam, struct nfsexstuff *exp, 2660191783Srmacklem struct ucred **credp) 2661191783Srmacklem{ 2662192121Srmacklem int i, error, *secflavors; 2663191783Srmacklem 2664191783Srmacklem error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp, 2665192121Srmacklem &exp->nes_numsecflavor, &secflavors); 2666192121Srmacklem if (error) { 2667192121Srmacklem if (nfs_rootfhset) { 2668192121Srmacklem exp->nes_exflag = 0; 2669192121Srmacklem exp->nes_numsecflavor = 0; 2670192121Srmacklem error = 0; 2671192121Srmacklem } 2672192121Srmacklem } else { 2673192121Srmacklem /* Copy the security flavors. */ 2674192121Srmacklem for (i = 0; i < exp->nes_numsecflavor; i++) 2675192121Srmacklem exp->nes_secflavors[i] = secflavors[i]; 2676191783Srmacklem } 2677224086Szack NFSEXITCODE(error); 2678191783Srmacklem return (error); 2679191783Srmacklem} 2680191783Srmacklem 2681191783Srmacklem/* 2682191783Srmacklem * Get a vnode for a file handle and export stuff. 2683191783Srmacklem */ 2684191783Srmacklemint 2685191783Srmacklemnfsvno_fhtovp(struct mount *mp, fhandle_t *fhp, struct sockaddr *nam, 2686216700Srmacklem int lktype, struct vnode **vpp, struct nfsexstuff *exp, 2687216700Srmacklem struct ucred **credp) 2688191783Srmacklem{ 2689192121Srmacklem int i, error, *secflavors; 2690191783Srmacklem 2691191998Srmacklem *credp = NULL; 2692192121Srmacklem exp->nes_numsecflavor = 0; 2693217063Srmacklem if (VFS_NEEDSGIANT(mp)) 2694217063Srmacklem error = ESTALE; 2695217063Srmacklem else 2696266447Srmacklem error = VFS_FHTOVP(mp, &fhp->fh_fid, lktype, vpp); 2697205663Srmacklem if (error != 0) 2698205663Srmacklem /* Make sure the server replies ESTALE to the client. */ 2699205663Srmacklem error = ESTALE; 2700191783Srmacklem if (nam && !error) { 2701191783Srmacklem error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp, 2702192121Srmacklem &exp->nes_numsecflavor, &secflavors); 2703191783Srmacklem if (error) { 2704191783Srmacklem if (nfs_rootfhset) { 2705191783Srmacklem exp->nes_exflag = 0; 2706192121Srmacklem exp->nes_numsecflavor = 0; 2707191783Srmacklem error = 0; 2708191783Srmacklem } else { 2709191783Srmacklem vput(*vpp); 2710191783Srmacklem } 2711192121Srmacklem } else { 2712192121Srmacklem /* Copy the security flavors. */ 2713192121Srmacklem for (i = 0; i < exp->nes_numsecflavor; i++) 2714192121Srmacklem exp->nes_secflavors[i] = secflavors[i]; 2715191783Srmacklem } 2716191783Srmacklem } 2717224086Szack NFSEXITCODE(error); 2718191783Srmacklem return (error); 2719191783Srmacklem} 2720191783Srmacklem 2721191783Srmacklem/* 2722191783Srmacklem * nfsd_fhtovp() - convert a fh to a vnode ptr 2723191783Srmacklem * - look up fsid in mount list (if not found ret error) 2724191783Srmacklem * - get vp and export rights by calling nfsvno_fhtovp() 2725191783Srmacklem * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2726191783Srmacklem * for AUTH_SYS 2727217063Srmacklem * - if mpp != NULL, return the mount point so that it can 2728217063Srmacklem * be used for vn_finished_write() by the caller 2729191783Srmacklem */ 2730191783Srmacklemvoid 2731216700Srmacklemnfsd_fhtovp(struct nfsrv_descript *nd, struct nfsrvfh *nfp, int lktype, 2732191783Srmacklem struct vnode **vpp, struct nfsexstuff *exp, 2733191783Srmacklem struct mount **mpp, int startwrite, struct thread *p) 2734191783Srmacklem{ 2735191783Srmacklem struct mount *mp; 2736191783Srmacklem struct ucred *credanon; 2737191783Srmacklem fhandle_t *fhp; 2738191783Srmacklem 2739191783Srmacklem fhp = (fhandle_t *)nfp->nfsrvfh_data; 2740191783Srmacklem /* 2741191783Srmacklem * Check for the special case of the nfsv4root_fh. 2742191783Srmacklem */ 2743217017Srmacklem mp = vfs_busyfs(&fhp->fh_fsid); 2744217063Srmacklem if (mpp != NULL) 2745217063Srmacklem *mpp = mp; 2746217017Srmacklem if (mp == NULL) { 2747191783Srmacklem *vpp = NULL; 2748191783Srmacklem nd->nd_repstat = ESTALE; 2749224086Szack goto out; 2750191783Srmacklem } 2751191783Srmacklem 2752251641Sken if (startwrite) { 2753191783Srmacklem vn_start_write(NULL, mpp, V_WAIT); 2754251641Sken if (lktype == LK_SHARED && !(MNT_SHARED_WRITES(mp))) 2755251641Sken lktype = LK_EXCLUSIVE; 2756251641Sken } 2757216700Srmacklem nd->nd_repstat = nfsvno_fhtovp(mp, fhp, nd->nd_nam, lktype, vpp, exp, 2758191783Srmacklem &credanon); 2759217017Srmacklem vfs_unbusy(mp); 2760191783Srmacklem 2761191783Srmacklem /* 2762191783Srmacklem * For NFSv4 without a pseudo root fs, unexported file handles 2763191783Srmacklem * can be returned, so that Lookup works everywhere. 2764191783Srmacklem */ 2765191783Srmacklem if (!nd->nd_repstat && exp->nes_exflag == 0 && 2766191783Srmacklem !(nd->nd_flag & ND_NFSV4)) { 2767191783Srmacklem vput(*vpp); 2768191783Srmacklem nd->nd_repstat = EACCES; 2769191783Srmacklem } 2770191783Srmacklem 2771191783Srmacklem /* 2772191783Srmacklem * Personally, I've never seen any point in requiring a 2773191783Srmacklem * reserved port#, since only in the rare case where the 2774191783Srmacklem * clients are all boxes with secure system priviledges, 2775191783Srmacklem * does it provide any enhanced security, but... some people 2776191783Srmacklem * believe it to be useful and keep putting this code back in. 2777191783Srmacklem * (There is also some "security checker" out there that 2778191783Srmacklem * complains if the nfs server doesn't enforce this.) 2779191783Srmacklem * However, note the following: 2780191783Srmacklem * RFC3530 (NFSv4) specifies that a reserved port# not be 2781191783Srmacklem * required. 2782191783Srmacklem * RFC2623 recommends that, if a reserved port# is checked for, 2783191783Srmacklem * that there be a way to turn that off--> ifdef'd. 2784191783Srmacklem */ 2785191783Srmacklem#ifdef NFS_REQRSVPORT 2786191783Srmacklem if (!nd->nd_repstat) { 2787192121Srmacklem struct sockaddr_in *saddr; 2788192121Srmacklem struct sockaddr_in6 *saddr6; 2789192121Srmacklem 2790192121Srmacklem saddr = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in *); 2791192121Srmacklem saddr6 = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in6 *); 2792192121Srmacklem if (!(nd->nd_flag & ND_NFSV4) && 2793192121Srmacklem ((saddr->sin_family == AF_INET && 2794192121Srmacklem ntohs(saddr->sin_port) >= IPPORT_RESERVED) || 2795192121Srmacklem (saddr6->sin6_family == AF_INET6 && 2796192121Srmacklem ntohs(saddr6->sin6_port) >= IPPORT_RESERVED))) { 2797192121Srmacklem vput(*vpp); 2798192121Srmacklem nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); 2799192121Srmacklem } 2800191783Srmacklem } 2801191783Srmacklem#endif /* NFS_REQRSVPORT */ 2802191783Srmacklem 2803191783Srmacklem /* 2804191783Srmacklem * Check/setup credentials. 2805191783Srmacklem */ 2806191783Srmacklem if (!nd->nd_repstat) { 2807191783Srmacklem nd->nd_saveduid = nd->nd_cred->cr_uid; 2808191783Srmacklem nd->nd_repstat = nfsd_excred(nd, exp, credanon); 2809191783Srmacklem if (nd->nd_repstat) 2810191783Srmacklem vput(*vpp); 2811191783Srmacklem } 2812191940Skan if (credanon != NULL) 2813191940Skan crfree(credanon); 2814191783Srmacklem if (nd->nd_repstat) { 2815191783Srmacklem if (startwrite) 2816191783Srmacklem vn_finished_write(mp); 2817191783Srmacklem *vpp = NULL; 2818217063Srmacklem if (mpp != NULL) 2819217063Srmacklem *mpp = NULL; 2820191783Srmacklem } 2821224086Szack 2822224086Szackout: 2823224086Szack NFSEXITCODE2(0, nd); 2824191783Srmacklem} 2825191783Srmacklem 2826191783Srmacklem/* 2827191783Srmacklem * glue for fp. 2828191783Srmacklem */ 2829191783Srmacklemint 2830191783Srmacklemfp_getfvp(struct thread *p, int fd, struct file **fpp, struct vnode **vpp) 2831191783Srmacklem{ 2832191783Srmacklem struct filedesc *fdp; 2833191783Srmacklem struct file *fp; 2834224086Szack int error = 0; 2835191783Srmacklem 2836191783Srmacklem fdp = p->td_proc->p_fd; 2837191783Srmacklem if (fd >= fdp->fd_nfiles || 2838224086Szack (fp = fdp->fd_ofiles[fd]) == NULL) { 2839224086Szack error = EBADF; 2840224086Szack goto out; 2841224086Szack } 2842191783Srmacklem *fpp = fp; 2843224086Szack 2844224086Szackout: 2845224086Szack NFSEXITCODE(error); 2846224086Szack return (error); 2847191783Srmacklem} 2848191783Srmacklem 2849191783Srmacklem/* 2850192121Srmacklem * Called from nfssvc() to update the exports list. Just call 2851191783Srmacklem * vfs_export(). This has to be done, since the v4 root fake fs isn't 2852191783Srmacklem * in the mount list. 2853191783Srmacklem */ 2854191783Srmacklemint 2855191783Srmacklemnfsrv_v4rootexport(void *argp, struct ucred *cred, struct thread *p) 2856191783Srmacklem{ 2857191783Srmacklem struct nfsex_args *nfsexargp = (struct nfsex_args *)argp; 2858224086Szack int error = 0; 2859191783Srmacklem struct nameidata nd; 2860191783Srmacklem fhandle_t fh; 2861191783Srmacklem 2862191783Srmacklem error = vfs_export(&nfsv4root_mnt, &nfsexargp->export); 2863216894Srmacklem if ((nfsexargp->export.ex_flags & MNT_DELEXPORT) != 0) 2864191783Srmacklem nfs_rootfhset = 0; 2865216894Srmacklem else if (error == 0) { 2866224086Szack if (nfsexargp->fspec == NULL) { 2867224086Szack error = EPERM; 2868224086Szack goto out; 2869224086Szack } 2870191783Srmacklem /* 2871191783Srmacklem * If fspec != NULL, this is the v4root path. 2872191783Srmacklem */ 2873191783Srmacklem NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, UIO_USERSPACE, 2874191783Srmacklem nfsexargp->fspec, p); 2875191783Srmacklem if ((error = namei(&nd)) != 0) 2876224086Szack goto out; 2877191783Srmacklem error = nfsvno_getfh(nd.ni_vp, &fh, p); 2878191783Srmacklem vrele(nd.ni_vp); 2879191783Srmacklem if (!error) { 2880191783Srmacklem nfs_rootfh.nfsrvfh_len = NFSX_MYFH; 2881191783Srmacklem NFSBCOPY((caddr_t)&fh, 2882191783Srmacklem nfs_rootfh.nfsrvfh_data, 2883191783Srmacklem sizeof (fhandle_t)); 2884191783Srmacklem nfs_rootfhset = 1; 2885191783Srmacklem } 2886191783Srmacklem } 2887224086Szack 2888224086Szackout: 2889224086Szack NFSEXITCODE(error); 2890191783Srmacklem return (error); 2891191783Srmacklem} 2892191783Srmacklem 2893191783Srmacklem/* 2894191783Srmacklem * This function needs to test to see if the system is near its limit 2895191783Srmacklem * for memory allocation via malloc() or mget() and return True iff 2896191783Srmacklem * either of these resources are near their limit. 2897191783Srmacklem * XXX (For now, this is just a stub.) 2898191783Srmacklem */ 2899191783Srmacklemint nfsrv_testmalloclimit = 0; 2900191783Srmacklemint 2901191783Srmacklemnfsrv_mallocmget_limit(void) 2902191783Srmacklem{ 2903191783Srmacklem static int printmesg = 0; 2904191783Srmacklem static int testval = 1; 2905191783Srmacklem 2906191783Srmacklem if (nfsrv_testmalloclimit && (testval++ % 1000) == 0) { 2907191783Srmacklem if ((printmesg++ % 100) == 0) 2908191783Srmacklem printf("nfsd: malloc/mget near limit\n"); 2909191783Srmacklem return (1); 2910191783Srmacklem } 2911191783Srmacklem return (0); 2912191783Srmacklem} 2913191783Srmacklem 2914191783Srmacklem/* 2915191783Srmacklem * BSD specific initialization of a mount point. 2916191783Srmacklem */ 2917191783Srmacklemvoid 2918191783Srmacklemnfsd_mntinit(void) 2919191783Srmacklem{ 2920191783Srmacklem static int inited = 0; 2921191783Srmacklem 2922191783Srmacklem if (inited) 2923191783Srmacklem return; 2924191783Srmacklem inited = 1; 2925191783Srmacklem nfsv4root_mnt.mnt_flag = (MNT_RDONLY | MNT_EXPORTED); 2926191783Srmacklem TAILQ_INIT(&nfsv4root_mnt.mnt_nvnodelist); 2927235626Smckusick TAILQ_INIT(&nfsv4root_mnt.mnt_activevnodelist); 2928191783Srmacklem nfsv4root_mnt.mnt_export = NULL; 2929191783Srmacklem TAILQ_INIT(&nfsv4root_opt); 2930191783Srmacklem TAILQ_INIT(&nfsv4root_newopt); 2931191783Srmacklem nfsv4root_mnt.mnt_opt = &nfsv4root_opt; 2932191783Srmacklem nfsv4root_mnt.mnt_optnew = &nfsv4root_newopt; 2933191783Srmacklem nfsv4root_mnt.mnt_nvnodelistsize = 0; 2934235626Smckusick nfsv4root_mnt.mnt_activevnodelistsize = 0; 2935191783Srmacklem} 2936191783Srmacklem 2937191783Srmacklem/* 2938191783Srmacklem * Get a vnode for a file handle, without checking exports, etc. 2939191783Srmacklem */ 2940191783Srmacklemstruct vnode * 2941191783Srmacklemnfsvno_getvp(fhandle_t *fhp) 2942191783Srmacklem{ 2943191783Srmacklem struct mount *mp; 2944191783Srmacklem struct vnode *vp; 2945191783Srmacklem int error; 2946191783Srmacklem 2947217017Srmacklem mp = vfs_busyfs(&fhp->fh_fsid); 2948191783Srmacklem if (mp == NULL) 2949191783Srmacklem return (NULL); 2950222167Srmacklem error = VFS_FHTOVP(mp, &fhp->fh_fid, LK_EXCLUSIVE, &vp); 2951217017Srmacklem vfs_unbusy(mp); 2952191783Srmacklem if (error) 2953191783Srmacklem return (NULL); 2954191783Srmacklem return (vp); 2955191783Srmacklem} 2956191783Srmacklem 2957191783Srmacklem/* 2958191783Srmacklem * Do a local VOP_ADVLOCK(). 2959191783Srmacklem */ 2960191783Srmacklemint 2961191783Srmacklemnfsvno_advlock(struct vnode *vp, int ftype, u_int64_t first, 2962192256Srmacklem u_int64_t end, struct thread *td) 2963191783Srmacklem{ 2964224086Szack int error = 0; 2965191783Srmacklem struct flock fl; 2966191783Srmacklem u_int64_t tlen; 2967191783Srmacklem 2968212833Srmacklem if (nfsrv_dolocallocks == 0) 2969224086Szack goto out; 2970216893Srmacklem 2971216893Srmacklem /* Check for VI_DOOMED here, so that VOP_ADVLOCK() isn't performed. */ 2972224086Szack if ((vp->v_iflag & VI_DOOMED) != 0) { 2973224086Szack error = EPERM; 2974224086Szack goto out; 2975224086Szack } 2976216893Srmacklem 2977191783Srmacklem fl.l_whence = SEEK_SET; 2978191783Srmacklem fl.l_type = ftype; 2979191783Srmacklem fl.l_start = (off_t)first; 2980191783Srmacklem if (end == NFS64BITSSET) { 2981191783Srmacklem fl.l_len = 0; 2982191783Srmacklem } else { 2983191783Srmacklem tlen = end - first; 2984191783Srmacklem fl.l_len = (off_t)tlen; 2985191783Srmacklem } 2986191783Srmacklem /* 2987192256Srmacklem * For FreeBSD8, the l_pid and l_sysid must be set to the same 2988192256Srmacklem * values for all calls, so that all locks will be held by the 2989192256Srmacklem * nfsd server. (The nfsd server handles conflicts between the 2990192256Srmacklem * various clients.) 2991192256Srmacklem * Since an NFSv4 lockowner is a ClientID plus an array of up to 1024 2992192256Srmacklem * bytes, so it can't be put in l_sysid. 2993191783Srmacklem */ 2994192503Srmacklem if (nfsv4_sysid == 0) 2995192503Srmacklem nfsv4_sysid = nlm_acquire_next_sysid(); 2996192256Srmacklem fl.l_pid = (pid_t)0; 2997192503Srmacklem fl.l_sysid = (int)nfsv4_sysid; 2998192256Srmacklem 2999224080Szack NFSVOPUNLOCK(vp, 0); 3000212833Srmacklem if (ftype == F_UNLCK) 3001212833Srmacklem error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_UNLCK, &fl, 3002212833Srmacklem (F_POSIX | F_REMOTE)); 3003212833Srmacklem else 3004212833Srmacklem error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_SETLK, &fl, 3005212833Srmacklem (F_POSIX | F_REMOTE)); 3006224080Szack NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY); 3007224086Szack 3008224086Szackout: 3009224086Szack NFSEXITCODE(error); 3010191783Srmacklem return (error); 3011191783Srmacklem} 3012191783Srmacklem 3013191783Srmacklem/* 3014191783Srmacklem * Check the nfsv4 root exports. 3015191783Srmacklem */ 3016191783Srmacklemint 3017191783Srmacklemnfsvno_v4rootexport(struct nfsrv_descript *nd) 3018191783Srmacklem{ 3019191783Srmacklem struct ucred *credanon; 3020224086Szack int exflags, error = 0, numsecflavor, *secflavors, i; 3021191783Srmacklem 3022191783Srmacklem error = vfs_stdcheckexp(&nfsv4root_mnt, nd->nd_nam, &exflags, 3023192121Srmacklem &credanon, &numsecflavor, &secflavors); 3024224086Szack if (error) { 3025224086Szack error = NFSERR_PROGUNAVAIL; 3026224086Szack goto out; 3027224086Szack } 3028191940Skan if (credanon != NULL) 3029191940Skan crfree(credanon); 3030192121Srmacklem for (i = 0; i < numsecflavor; i++) { 3031192121Srmacklem if (secflavors[i] == AUTH_SYS) 3032192121Srmacklem nd->nd_flag |= ND_EXAUTHSYS; 3033192121Srmacklem else if (secflavors[i] == RPCSEC_GSS_KRB5) 3034192121Srmacklem nd->nd_flag |= ND_EXGSS; 3035192121Srmacklem else if (secflavors[i] == RPCSEC_GSS_KRB5I) 3036192121Srmacklem nd->nd_flag |= ND_EXGSSINTEGRITY; 3037192121Srmacklem else if (secflavors[i] == RPCSEC_GSS_KRB5P) 3038192121Srmacklem nd->nd_flag |= ND_EXGSSPRIVACY; 3039192121Srmacklem } 3040224086Szack 3041224086Szackout: 3042224086Szack NFSEXITCODE(error); 3043224086Szack return (error); 3044191783Srmacklem} 3045191783Srmacklem 3046191783Srmacklem/* 3047191783Srmacklem * Nfs server psuedo system call for the nfsd's 3048191783Srmacklem */ 3049191783Srmacklem/* 3050191783Srmacklem * MPSAFE 3051191783Srmacklem */ 3052191783Srmacklemstatic int 3053191783Srmacklemnfssvc_nfsd(struct thread *td, struct nfssvc_args *uap) 3054191783Srmacklem{ 3055191783Srmacklem struct file *fp; 3056192017Srmacklem struct nfsd_addsock_args sockarg; 3057192017Srmacklem struct nfsd_nfsd_args nfsdarg; 3058191783Srmacklem int error; 3059191783Srmacklem 3060191783Srmacklem if (uap->flag & NFSSVC_NFSDADDSOCK) { 3061192017Srmacklem error = copyin(uap->argp, (caddr_t)&sockarg, sizeof (sockarg)); 3062191783Srmacklem if (error) 3063224086Szack goto out; 3064224778Srwatson /* 3065224778Srwatson * Since we don't know what rights might be required, 3066224778Srwatson * pretend that we need them all. It is better to be too 3067224778Srwatson * careful than too reckless. 3068224778Srwatson */ 3069224778Srwatson if ((error = fget(td, sockarg.sock, CAP_SOCK_ALL, &fp)) != 0) 3070224086Szack goto out; 3071191783Srmacklem if (fp->f_type != DTYPE_SOCKET) { 3072191783Srmacklem fdrop(fp, td); 3073224086Szack error = EPERM; 3074224086Szack goto out; 3075191783Srmacklem } 3076191783Srmacklem error = nfsrvd_addsock(fp); 3077191783Srmacklem fdrop(fp, td); 3078191783Srmacklem } else if (uap->flag & NFSSVC_NFSDNFSD) { 3079224086Szack if (uap->argp == NULL) { 3080224086Szack error = EINVAL; 3081224086Szack goto out; 3082224086Szack } 3083192017Srmacklem error = copyin(uap->argp, (caddr_t)&nfsdarg, 3084192017Srmacklem sizeof (nfsdarg)); 3085192017Srmacklem if (error) 3086224086Szack goto out; 3087192017Srmacklem error = nfsrvd_nfsd(td, &nfsdarg); 3088191783Srmacklem } else { 3089191783Srmacklem error = nfssvc_srvcall(td, uap, td->td_ucred); 3090191783Srmacklem } 3091224086Szack 3092224086Szackout: 3093224086Szack NFSEXITCODE(error); 3094191783Srmacklem return (error); 3095191783Srmacklem} 3096191783Srmacklem 3097191783Srmacklemstatic int 3098191783Srmacklemnfssvc_srvcall(struct thread *p, struct nfssvc_args *uap, struct ucred *cred) 3099191783Srmacklem{ 3100191783Srmacklem struct nfsex_args export; 3101191783Srmacklem struct file *fp = NULL; 3102191783Srmacklem int stablefd, len; 3103191783Srmacklem struct nfsd_clid adminrevoke; 3104191783Srmacklem struct nfsd_dumplist dumplist; 3105191783Srmacklem struct nfsd_dumpclients *dumpclients; 3106191783Srmacklem struct nfsd_dumplocklist dumplocklist; 3107191783Srmacklem struct nfsd_dumplocks *dumplocks; 3108191783Srmacklem struct nameidata nd; 3109191783Srmacklem vnode_t vp; 3110243738Srmacklem int error = EINVAL, igotlock; 3111217432Srmacklem struct proc *procp; 3112243738Srmacklem static int suspend_nfsd = 0; 3113191783Srmacklem 3114191783Srmacklem if (uap->flag & NFSSVC_PUBLICFH) { 3115191783Srmacklem NFSBZERO((caddr_t)&nfs_pubfh.nfsrvfh_data, 3116191783Srmacklem sizeof (fhandle_t)); 3117191783Srmacklem error = copyin(uap->argp, 3118191783Srmacklem &nfs_pubfh.nfsrvfh_data, sizeof (fhandle_t)); 3119191783Srmacklem if (!error) 3120191783Srmacklem nfs_pubfhset = 1; 3121191783Srmacklem } else if (uap->flag & NFSSVC_V4ROOTEXPORT) { 3122191783Srmacklem error = copyin(uap->argp,(caddr_t)&export, 3123191783Srmacklem sizeof (struct nfsex_args)); 3124191783Srmacklem if (!error) 3125191783Srmacklem error = nfsrv_v4rootexport(&export, cred, p); 3126191783Srmacklem } else if (uap->flag & NFSSVC_NOPUBLICFH) { 3127191783Srmacklem nfs_pubfhset = 0; 3128191783Srmacklem error = 0; 3129191783Srmacklem } else if (uap->flag & NFSSVC_STABLERESTART) { 3130191783Srmacklem error = copyin(uap->argp, (caddr_t)&stablefd, 3131191783Srmacklem sizeof (int)); 3132191783Srmacklem if (!error) 3133191783Srmacklem error = fp_getfvp(p, stablefd, &fp, &vp); 3134191783Srmacklem if (!error && (NFSFPFLAG(fp) & (FREAD | FWRITE)) != (FREAD | FWRITE)) 3135191783Srmacklem error = EBADF; 3136191783Srmacklem if (!error && newnfs_numnfsd != 0) 3137191783Srmacklem error = EPERM; 3138191783Srmacklem if (!error) { 3139191783Srmacklem nfsrv_stablefirst.nsf_fp = fp; 3140191783Srmacklem nfsrv_setupstable(p); 3141191783Srmacklem } 3142191783Srmacklem } else if (uap->flag & NFSSVC_ADMINREVOKE) { 3143191783Srmacklem error = copyin(uap->argp, (caddr_t)&adminrevoke, 3144191783Srmacklem sizeof (struct nfsd_clid)); 3145191783Srmacklem if (!error) 3146191783Srmacklem error = nfsrv_adminrevoke(&adminrevoke, p); 3147191783Srmacklem } else if (uap->flag & NFSSVC_DUMPCLIENTS) { 3148191783Srmacklem error = copyin(uap->argp, (caddr_t)&dumplist, 3149191783Srmacklem sizeof (struct nfsd_dumplist)); 3150191783Srmacklem if (!error && (dumplist.ndl_size < 1 || 3151191783Srmacklem dumplist.ndl_size > NFSRV_MAXDUMPLIST)) 3152191783Srmacklem error = EPERM; 3153191783Srmacklem if (!error) { 3154191783Srmacklem len = sizeof (struct nfsd_dumpclients) * dumplist.ndl_size; 3155191783Srmacklem dumpclients = (struct nfsd_dumpclients *)malloc(len, 3156191783Srmacklem M_TEMP, M_WAITOK); 3157191783Srmacklem nfsrv_dumpclients(dumpclients, dumplist.ndl_size); 3158191783Srmacklem error = copyout(dumpclients, 3159191783Srmacklem CAST_USER_ADDR_T(dumplist.ndl_list), len); 3160191783Srmacklem free((caddr_t)dumpclients, M_TEMP); 3161191783Srmacklem } 3162191783Srmacklem } else if (uap->flag & NFSSVC_DUMPLOCKS) { 3163191783Srmacklem error = copyin(uap->argp, (caddr_t)&dumplocklist, 3164191783Srmacklem sizeof (struct nfsd_dumplocklist)); 3165191783Srmacklem if (!error && (dumplocklist.ndllck_size < 1 || 3166191783Srmacklem dumplocklist.ndllck_size > NFSRV_MAXDUMPLIST)) 3167191783Srmacklem error = EPERM; 3168191783Srmacklem if (!error) 3169191783Srmacklem error = nfsrv_lookupfilename(&nd, 3170191783Srmacklem dumplocklist.ndllck_fname, p); 3171191783Srmacklem if (!error) { 3172191783Srmacklem len = sizeof (struct nfsd_dumplocks) * 3173191783Srmacklem dumplocklist.ndllck_size; 3174191783Srmacklem dumplocks = (struct nfsd_dumplocks *)malloc(len, 3175191783Srmacklem M_TEMP, M_WAITOK); 3176191783Srmacklem nfsrv_dumplocks(nd.ni_vp, dumplocks, 3177191783Srmacklem dumplocklist.ndllck_size, p); 3178191783Srmacklem vput(nd.ni_vp); 3179191783Srmacklem error = copyout(dumplocks, 3180191783Srmacklem CAST_USER_ADDR_T(dumplocklist.ndllck_list), len); 3181191783Srmacklem free((caddr_t)dumplocks, M_TEMP); 3182191783Srmacklem } 3183217432Srmacklem } else if (uap->flag & NFSSVC_BACKUPSTABLE) { 3184217432Srmacklem procp = p->td_proc; 3185217432Srmacklem PROC_LOCK(procp); 3186217432Srmacklem nfsd_master_pid = procp->p_pid; 3187217432Srmacklem bcopy(procp->p_comm, nfsd_master_comm, MAXCOMLEN + 1); 3188217432Srmacklem nfsd_master_start = procp->p_stats->p_start; 3189217432Srmacklem nfsd_master_proc = procp; 3190217432Srmacklem PROC_UNLOCK(procp); 3191243738Srmacklem } else if ((uap->flag & NFSSVC_SUSPENDNFSD) != 0) { 3192243738Srmacklem NFSLOCKV4ROOTMUTEX(); 3193243738Srmacklem if (suspend_nfsd == 0) { 3194243738Srmacklem /* Lock out all nfsd threads */ 3195243738Srmacklem do { 3196243738Srmacklem igotlock = nfsv4_lock(&nfsd_suspend_lock, 1, 3197243738Srmacklem NULL, NFSV4ROOTLOCKMUTEXPTR, NULL); 3198243738Srmacklem } while (igotlock == 0 && suspend_nfsd == 0); 3199243738Srmacklem suspend_nfsd = 1; 3200243738Srmacklem } 3201243738Srmacklem NFSUNLOCKV4ROOTMUTEX(); 3202243738Srmacklem error = 0; 3203243738Srmacklem } else if ((uap->flag & NFSSVC_RESUMENFSD) != 0) { 3204243738Srmacklem NFSLOCKV4ROOTMUTEX(); 3205243738Srmacklem if (suspend_nfsd != 0) { 3206243738Srmacklem nfsv4_unlock(&nfsd_suspend_lock, 0); 3207243738Srmacklem suspend_nfsd = 0; 3208243738Srmacklem } 3209243738Srmacklem NFSUNLOCKV4ROOTMUTEX(); 3210243738Srmacklem error = 0; 3211191783Srmacklem } 3212224086Szack 3213224086Szack NFSEXITCODE(error); 3214191783Srmacklem return (error); 3215191783Srmacklem} 3216191783Srmacklem 3217192121Srmacklem/* 3218192121Srmacklem * Check exports. 3219192121Srmacklem * Returns 0 if ok, 1 otherwise. 3220192121Srmacklem */ 3221192121Srmacklemint 3222192121Srmacklemnfsvno_testexp(struct nfsrv_descript *nd, struct nfsexstuff *exp) 3223192121Srmacklem{ 3224192121Srmacklem int i; 3225192121Srmacklem 3226192121Srmacklem /* 3227192121Srmacklem * This seems odd, but allow the case where the security flavor 3228192121Srmacklem * list is empty. This happens when NFSv4 is traversing non-exported 3229192121Srmacklem * file systems. Exported file systems should always have a non-empty 3230192121Srmacklem * security flavor list. 3231192121Srmacklem */ 3232192121Srmacklem if (exp->nes_numsecflavor == 0) 3233192121Srmacklem return (0); 3234192121Srmacklem 3235192121Srmacklem for (i = 0; i < exp->nes_numsecflavor; i++) { 3236192121Srmacklem /* 3237192121Srmacklem * The tests for privacy and integrity must be first, 3238192121Srmacklem * since ND_GSS is set for everything but AUTH_SYS. 3239192121Srmacklem */ 3240192121Srmacklem if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5P && 3241192121Srmacklem (nd->nd_flag & ND_GSSPRIVACY)) 3242192121Srmacklem return (0); 3243192121Srmacklem if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5I && 3244192121Srmacklem (nd->nd_flag & ND_GSSINTEGRITY)) 3245192121Srmacklem return (0); 3246192121Srmacklem if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5 && 3247192121Srmacklem (nd->nd_flag & ND_GSS)) 3248192121Srmacklem return (0); 3249192121Srmacklem if (exp->nes_secflavors[i] == AUTH_SYS && 3250192121Srmacklem (nd->nd_flag & ND_GSS) == 0) 3251192121Srmacklem return (0); 3252192121Srmacklem } 3253192121Srmacklem return (1); 3254192121Srmacklem} 3255192121Srmacklem 3256214224Srmacklem/* 3257214224Srmacklem * Calculate a hash value for the fid in a file handle. 3258214224Srmacklem */ 3259214255Srmacklemuint32_t 3260214224Srmacklemnfsrv_hashfh(fhandle_t *fhp) 3261214224Srmacklem{ 3262214255Srmacklem uint32_t hashval; 3263214224Srmacklem 3264214255Srmacklem hashval = hash32_buf(&fhp->fh_fid, sizeof(struct fid), 0); 3265214224Srmacklem return (hashval); 3266214224Srmacklem} 3267214224Srmacklem 3268217432Srmacklem/* 3269217432Srmacklem * Signal the userland master nfsd to backup the stable restart file. 3270217432Srmacklem */ 3271217432Srmacklemvoid 3272217432Srmacklemnfsrv_backupstable(void) 3273217432Srmacklem{ 3274217432Srmacklem struct proc *procp; 3275217432Srmacklem 3276217432Srmacklem if (nfsd_master_proc != NULL) { 3277217432Srmacklem procp = pfind(nfsd_master_pid); 3278217432Srmacklem /* Try to make sure it is the correct process. */ 3279217432Srmacklem if (procp == nfsd_master_proc && 3280217432Srmacklem procp->p_stats->p_start.tv_sec == 3281217432Srmacklem nfsd_master_start.tv_sec && 3282217432Srmacklem procp->p_stats->p_start.tv_usec == 3283217432Srmacklem nfsd_master_start.tv_usec && 3284217432Srmacklem strcmp(procp->p_comm, nfsd_master_comm) == 0) 3285225617Skmacy kern_psignal(procp, SIGUSR2); 3286217432Srmacklem else 3287217432Srmacklem nfsd_master_proc = NULL; 3288217432Srmacklem 3289217432Srmacklem if (procp != NULL) 3290217432Srmacklem PROC_UNLOCK(procp); 3291217432Srmacklem } 3292217432Srmacklem} 3293217432Srmacklem 3294191783Srmacklemextern int (*nfsd_call_nfsd)(struct thread *, struct nfssvc_args *); 3295191783Srmacklem 3296191783Srmacklem/* 3297191783Srmacklem * Called once to initialize data structures... 3298191783Srmacklem */ 3299191783Srmacklemstatic int 3300191783Srmacklemnfsd_modevent(module_t mod, int type, void *data) 3301191783Srmacklem{ 3302255532Srmacklem int error = 0, i; 3303191783Srmacklem static int loaded = 0; 3304191783Srmacklem 3305191783Srmacklem switch (type) { 3306191783Srmacklem case MOD_LOAD: 3307191783Srmacklem if (loaded) 3308224086Szack goto out; 3309191783Srmacklem newnfs_portinit(); 3310255532Srmacklem for (i = 0; i < NFSRVCACHE_HASHSIZE; i++) { 3311255532Srmacklem snprintf(nfsrchash_table[i].lock_name, 3312255532Srmacklem sizeof(nfsrchash_table[i].lock_name), "nfsrc_tcp%d", 3313255532Srmacklem i); 3314255532Srmacklem mtx_init(&nfsrchash_table[i].mtx, 3315255532Srmacklem nfsrchash_table[i].lock_name, NULL, MTX_DEF); 3316261067Smav snprintf(nfsrcahash_table[i].lock_name, 3317261067Smav sizeof(nfsrcahash_table[i].lock_name), "nfsrc_tcpa%d", 3318261067Smav i); 3319261067Smav mtx_init(&nfsrcahash_table[i].mtx, 3320261067Smav nfsrcahash_table[i].lock_name, NULL, MTX_DEF); 3321255532Srmacklem } 3322255532Srmacklem mtx_init(&nfsrc_udpmtx, "nfs_udpcache_mutex", NULL, MTX_DEF); 3323191783Srmacklem mtx_init(&nfs_v4root_mutex, "nfs_v4root_mutex", NULL, MTX_DEF); 3324191783Srmacklem mtx_init(&nfsv4root_mnt.mnt_mtx, "struct mount mtx", NULL, 3325191783Srmacklem MTX_DEF); 3326191783Srmacklem lockinit(&nfsv4root_mnt.mnt_explock, PVFS, "explock", 0, 0); 3327191783Srmacklem nfsrvd_initcache(); 3328191783Srmacklem nfsd_init(); 3329191783Srmacklem NFSD_LOCK(); 3330191783Srmacklem nfsrvd_init(0); 3331191783Srmacklem NFSD_UNLOCK(); 3332191783Srmacklem nfsd_mntinit(); 3333191783Srmacklem#ifdef VV_DISABLEDELEG 3334191783Srmacklem vn_deleg_ops.vndeleg_recall = nfsd_recalldelegation; 3335191783Srmacklem vn_deleg_ops.vndeleg_disable = nfsd_disabledelegation; 3336191783Srmacklem#endif 3337191783Srmacklem nfsd_call_servertimer = nfsrv_servertimer; 3338191783Srmacklem nfsd_call_nfsd = nfssvc_nfsd; 3339191783Srmacklem loaded = 1; 3340191783Srmacklem break; 3341191783Srmacklem 3342191783Srmacklem case MOD_UNLOAD: 3343191783Srmacklem if (newnfs_numnfsd != 0) { 3344191783Srmacklem error = EBUSY; 3345191783Srmacklem break; 3346191783Srmacklem } 3347191783Srmacklem 3348191783Srmacklem#ifdef VV_DISABLEDELEG 3349191783Srmacklem vn_deleg_ops.vndeleg_recall = NULL; 3350191783Srmacklem vn_deleg_ops.vndeleg_disable = NULL; 3351191783Srmacklem#endif 3352191783Srmacklem nfsd_call_servertimer = NULL; 3353191783Srmacklem nfsd_call_nfsd = NULL; 3354217335Szack 3355220530Srmacklem /* Clean out all NFSv4 state. */ 3356220530Srmacklem nfsrv_throwawayallstate(curthread); 3357220530Srmacklem 3358217335Szack /* Clean the NFS server reply cache */ 3359217335Szack nfsrvd_cleancache(); 3360217335Szack 3361220530Srmacklem /* Free up the krpc server pool. */ 3362220530Srmacklem if (nfsrvd_pool != NULL) 3363220530Srmacklem svcpool_destroy(nfsrvd_pool); 3364220530Srmacklem 3365191783Srmacklem /* and get rid of the locks */ 3366261067Smav for (i = 0; i < NFSRVCACHE_HASHSIZE; i++) { 3367255532Srmacklem mtx_destroy(&nfsrchash_table[i].mtx); 3368261067Smav mtx_destroy(&nfsrcahash_table[i].mtx); 3369261067Smav } 3370255532Srmacklem mtx_destroy(&nfsrc_udpmtx); 3371191783Srmacklem mtx_destroy(&nfs_v4root_mutex); 3372191783Srmacklem mtx_destroy(&nfsv4root_mnt.mnt_mtx); 3373191783Srmacklem lockdestroy(&nfsv4root_mnt.mnt_explock); 3374191783Srmacklem loaded = 0; 3375191783Srmacklem break; 3376191783Srmacklem default: 3377191783Srmacklem error = EOPNOTSUPP; 3378191783Srmacklem break; 3379191783Srmacklem } 3380224086Szack 3381224086Szackout: 3382224086Szack NFSEXITCODE(error); 3383224086Szack return (error); 3384191783Srmacklem} 3385191783Srmacklemstatic moduledata_t nfsd_mod = { 3386191783Srmacklem "nfsd", 3387191783Srmacklem nfsd_modevent, 3388191783Srmacklem NULL, 3389191783Srmacklem}; 3390191783SrmacklemDECLARE_MODULE(nfsd, nfsd_mod, SI_SUB_VFS, SI_ORDER_ANY); 3391191783Srmacklem 3392191783Srmacklem/* So that loader and kldload(2) can find us, wherever we are.. */ 3393191783SrmacklemMODULE_VERSION(nfsd, 1); 3394191783SrmacklemMODULE_DEPEND(nfsd, nfscommon, 1, 1, 1); 3395216931SrmacklemMODULE_DEPEND(nfsd, nfslock, 1, 1, 1); 3396192574SrmacklemMODULE_DEPEND(nfsd, nfslockd, 1, 1, 1); 3397209191SrmacklemMODULE_DEPEND(nfsd, krpc, 1, 1, 1); 3398209191SrmacklemMODULE_DEPEND(nfsd, nfssvc, 1, 1, 1); 3399191783Srmacklem 3400