nfs_nfsdport.c revision 232683
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: stable/9/sys/fs/nfsserver/nfs_nfsdport.c 232683 2012-03-08 03:02:48Z rmacklem $"); 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; 60191783Srmacklemstruct vfsoptlist nfsv4root_opt, nfsv4root_newopt; 61191783SrmacklemNFSDLOCKMUTEX; 62191783Srmacklemstruct mtx nfs_cache_mutex; 63191783Srmacklemstruct mtx nfs_v4root_mutex; 64191783Srmacklemstruct nfsrvfh nfs_rootfh, nfs_pubfh; 65191783Srmacklemint nfs_pubfhset = 0, nfs_rootfhset = 0; 66217432Srmacklemstruct proc *nfsd_master_proc = NULL; 67217432Srmacklemstatic pid_t nfsd_master_pid = (pid_t)-1; 68217432Srmacklemstatic char nfsd_master_comm[MAXCOMLEN + 1]; 69217432Srmacklemstatic struct timeval nfsd_master_start; 70192503Srmacklemstatic uint32_t nfsv4_sysid = 0; 71191783Srmacklem 72192255Srmacklemstatic int nfssvc_srvcall(struct thread *, struct nfssvc_args *, 73192255Srmacklem struct ucred *); 74191783Srmacklem 75220645Srmacklemint nfsrv_enable_crossmntpt = 1; 76191783Srmacklemstatic int nfs_commit_blks; 77191783Srmacklemstatic int nfs_commit_miss; 78191783Srmacklemextern int nfsrv_issuedelegs; 79192255Srmacklemextern int nfsrv_dolocallocks; 80192255Srmacklem 81221615SrmacklemSYSCTL_NODE(_vfs, OID_AUTO, nfsd, CTLFLAG_RW, 0, "New NFS server"); 82221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, mirrormnt, CTLFLAG_RW, 83220645Srmacklem &nfsrv_enable_crossmntpt, 0, "Enable nfsd to cross mount points"); 84221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_blks, CTLFLAG_RW, &nfs_commit_blks, 85192255Srmacklem 0, ""); 86221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_miss, CTLFLAG_RW, &nfs_commit_miss, 87192255Srmacklem 0, ""); 88221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, issue_delegations, CTLFLAG_RW, 89192255Srmacklem &nfsrv_issuedelegs, 0, "Enable nfsd to issue delegations"); 90221615SrmacklemSYSCTL_INT(_vfs_nfsd, OID_AUTO, enable_locallocks, CTLFLAG_RW, 91192255Srmacklem &nfsrv_dolocallocks, 0, "Enable nfsd to acquire local locks on files"); 92191783Srmacklem 93229617Sjhb#define MAX_REORDERED_RPC 16 94229617Sjhb#define NUM_HEURISTIC 1031 95191783Srmacklem#define NHUSE_INIT 64 96191783Srmacklem#define NHUSE_INC 16 97191783Srmacklem#define NHUSE_MAX 2048 98191783Srmacklem 99191783Srmacklemstatic struct nfsheur { 100191783Srmacklem struct vnode *nh_vp; /* vp to match (unreferenced pointer) */ 101229617Sjhb off_t nh_nextoff; /* next offset for sequential detection */ 102191783Srmacklem int nh_use; /* use count for selection */ 103191783Srmacklem int nh_seqcount; /* heuristic */ 104191783Srmacklem} nfsheur[NUM_HEURISTIC]; 105191783Srmacklem 106191783Srmacklem 107191783Srmacklem/* 108229617Sjhb * Heuristic to detect sequential operation. 109229617Sjhb */ 110229617Sjhbstatic struct nfsheur * 111229617Sjhbnfsrv_sequential_heuristic(struct uio *uio, struct vnode *vp) 112229617Sjhb{ 113229617Sjhb struct nfsheur *nh; 114229617Sjhb int hi, try; 115229617Sjhb 116229617Sjhb /* Locate best candidate. */ 117229617Sjhb try = 32; 118229617Sjhb hi = ((int)(vm_offset_t)vp / sizeof(struct vnode)) % NUM_HEURISTIC; 119229617Sjhb nh = &nfsheur[hi]; 120229617Sjhb while (try--) { 121229617Sjhb if (nfsheur[hi].nh_vp == vp) { 122229617Sjhb nh = &nfsheur[hi]; 123229617Sjhb break; 124229617Sjhb } 125229617Sjhb if (nfsheur[hi].nh_use > 0) 126229617Sjhb --nfsheur[hi].nh_use; 127229617Sjhb hi = (hi + 1) % NUM_HEURISTIC; 128229617Sjhb if (nfsheur[hi].nh_use < nh->nh_use) 129229617Sjhb nh = &nfsheur[hi]; 130229617Sjhb } 131229617Sjhb 132229617Sjhb /* Initialize hint if this is a new file. */ 133229617Sjhb if (nh->nh_vp != vp) { 134229617Sjhb nh->nh_vp = vp; 135229617Sjhb nh->nh_nextoff = uio->uio_offset; 136229617Sjhb nh->nh_use = NHUSE_INIT; 137229617Sjhb if (uio->uio_offset == 0) 138229617Sjhb nh->nh_seqcount = 4; 139229617Sjhb else 140229617Sjhb nh->nh_seqcount = 1; 141229617Sjhb } 142229617Sjhb 143229617Sjhb /* Calculate heuristic. */ 144229617Sjhb if ((uio->uio_offset == 0 && nh->nh_seqcount > 0) || 145229617Sjhb uio->uio_offset == nh->nh_nextoff) { 146229617Sjhb /* See comments in vfs_vnops.c:sequential_heuristic(). */ 147229617Sjhb nh->nh_seqcount += howmany(uio->uio_resid, 16384); 148229617Sjhb if (nh->nh_seqcount > IO_SEQMAX) 149229617Sjhb nh->nh_seqcount = IO_SEQMAX; 150229617Sjhb } else if (qabs(uio->uio_offset - nh->nh_nextoff) <= MAX_REORDERED_RPC * 151229617Sjhb imax(vp->v_mount->mnt_stat.f_iosize, uio->uio_resid)) { 152229617Sjhb /* Probably a reordered RPC, leave seqcount alone. */ 153229617Sjhb } else if (nh->nh_seqcount > 1) { 154229617Sjhb nh->nh_seqcount /= 2; 155229617Sjhb } else { 156229617Sjhb nh->nh_seqcount = 0; 157229617Sjhb } 158229617Sjhb nh->nh_use += NHUSE_INC; 159229617Sjhb if (nh->nh_use > NHUSE_MAX) 160229617Sjhb nh->nh_use = NHUSE_MAX; 161229617Sjhb return (nh); 162229617Sjhb} 163229617Sjhb 164229617Sjhb/* 165191783Srmacklem * Get attributes into nfsvattr structure. 166191783Srmacklem */ 167191783Srmacklemint 168191783Srmacklemnfsvno_getattr(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred, 169216693Srmacklem struct thread *p, int vpislocked) 170191783Srmacklem{ 171191783Srmacklem int error, lockedit = 0; 172191783Srmacklem 173216693Srmacklem if (vpislocked == 0) { 174216693Srmacklem /* 175216693Srmacklem * When vpislocked == 0, the vnode is either exclusively 176216693Srmacklem * locked by this thread or not locked by this thread. 177216693Srmacklem * As such, shared lock it, if not exclusively locked. 178216693Srmacklem */ 179224083Szack if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) { 180216693Srmacklem lockedit = 1; 181224081Szack NFSVOPLOCK(vp, LK_SHARED | LK_RETRY); 182216693Srmacklem } 183191783Srmacklem } 184191783Srmacklem error = VOP_GETATTR(vp, &nvap->na_vattr, cred); 185216693Srmacklem if (lockedit != 0) 186224082Szack NFSVOPUNLOCK(vp, 0); 187224086Szack 188224086Szack NFSEXITCODE(error); 189191783Srmacklem return (error); 190191783Srmacklem} 191191783Srmacklem 192191783Srmacklem/* 193191783Srmacklem * Get a file handle for a vnode. 194191783Srmacklem */ 195191783Srmacklemint 196191783Srmacklemnfsvno_getfh(struct vnode *vp, fhandle_t *fhp, struct thread *p) 197191783Srmacklem{ 198191783Srmacklem int error; 199191783Srmacklem 200191783Srmacklem NFSBZERO((caddr_t)fhp, sizeof(fhandle_t)); 201191783Srmacklem fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; 202191783Srmacklem error = VOP_VPTOFH(vp, &fhp->fh_fid); 203224086Szack 204224086Szack NFSEXITCODE(error); 205191783Srmacklem return (error); 206191783Srmacklem} 207191783Srmacklem 208191783Srmacklem/* 209191783Srmacklem * Perform access checking for vnodes obtained from file handles that would 210191783Srmacklem * refer to files already opened by a Unix client. You cannot just use 211200999Srmacklem * vn_writechk() and VOP_ACCESSX() for two reasons. 212200999Srmacklem * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write 213200999Srmacklem * case. 214191783Srmacklem * 2 - The owner is to be given access irrespective of mode bits for some 215191783Srmacklem * operations, so that processes that chmod after opening a file don't 216191783Srmacklem * break. 217191783Srmacklem */ 218191783Srmacklemint 219200999Srmacklemnfsvno_accchk(struct vnode *vp, accmode_t accmode, struct ucred *cred, 220200999Srmacklem struct nfsexstuff *exp, struct thread *p, int override, int vpislocked, 221200999Srmacklem u_int32_t *supportedtypep) 222191783Srmacklem{ 223191783Srmacklem struct vattr vattr; 224191783Srmacklem int error = 0, getret = 0; 225191783Srmacklem 226216893Srmacklem if (vpislocked == 0) { 227224086Szack if (NFSVOPLOCK(vp, LK_SHARED) != 0) { 228224086Szack error = EPERM; 229224086Szack goto out; 230224086Szack } 231216893Srmacklem } 232191783Srmacklem if (accmode & VWRITE) { 233191783Srmacklem /* Just vn_writechk() changed to check rdonly */ 234191783Srmacklem /* 235191783Srmacklem * Disallow write attempts on read-only file systems; 236191783Srmacklem * unless the file is a socket or a block or character 237191783Srmacklem * device resident on the file system. 238191783Srmacklem */ 239191783Srmacklem if (NFSVNO_EXRDONLY(exp) || 240200999Srmacklem (vp->v_mount->mnt_flag & MNT_RDONLY)) { 241191783Srmacklem switch (vp->v_type) { 242191783Srmacklem case VREG: 243191783Srmacklem case VDIR: 244191783Srmacklem case VLNK: 245216893Srmacklem error = EROFS; 246191783Srmacklem default: 247191783Srmacklem break; 248191783Srmacklem } 249191783Srmacklem } 250191783Srmacklem /* 251191783Srmacklem * If there's shared text associated with 252191783Srmacklem * the inode, try to free it up once. If 253191783Srmacklem * we fail, we can't allow writing. 254191783Srmacklem */ 255216893Srmacklem if ((vp->v_vflag & VV_TEXT) != 0 && error == 0) 256216893Srmacklem error = ETXTBSY; 257191783Srmacklem } 258216893Srmacklem if (error != 0) { 259216893Srmacklem if (vpislocked == 0) 260224082Szack NFSVOPUNLOCK(vp, 0); 261224086Szack goto out; 262216893Srmacklem } 263191783Srmacklem 264191783Srmacklem /* 265191783Srmacklem * Should the override still be applied when ACLs are enabled? 266191783Srmacklem */ 267200999Srmacklem error = VOP_ACCESSX(vp, accmode, cred, p); 268200999Srmacklem if (error != 0 && (accmode & (VDELETE | VDELETE_CHILD))) { 269200999Srmacklem /* 270200999Srmacklem * Try again with VEXPLICIT_DENY, to see if the test for 271200999Srmacklem * deletion is supported. 272200999Srmacklem */ 273200999Srmacklem error = VOP_ACCESSX(vp, accmode | VEXPLICIT_DENY, cred, p); 274200999Srmacklem if (error == 0) { 275200999Srmacklem if (vp->v_type == VDIR) { 276200999Srmacklem accmode &= ~(VDELETE | VDELETE_CHILD); 277200999Srmacklem accmode |= VWRITE; 278200999Srmacklem error = VOP_ACCESSX(vp, accmode, cred, p); 279200999Srmacklem } else if (supportedtypep != NULL) { 280200999Srmacklem *supportedtypep &= ~NFSACCESS_DELETE; 281200999Srmacklem } 282200999Srmacklem } 283200999Srmacklem } 284191783Srmacklem 285191783Srmacklem /* 286191783Srmacklem * Allow certain operations for the owner (reads and writes 287191783Srmacklem * on files that are already open). 288191783Srmacklem */ 289191783Srmacklem if (override != NFSACCCHK_NOOVERRIDE && 290191783Srmacklem (error == EPERM || error == EACCES)) { 291191783Srmacklem if (cred->cr_uid == 0 && (override & NFSACCCHK_ALLOWROOT)) 292191783Srmacklem error = 0; 293191783Srmacklem else if (override & NFSACCCHK_ALLOWOWNER) { 294191783Srmacklem getret = VOP_GETATTR(vp, &vattr, cred); 295191783Srmacklem if (getret == 0 && cred->cr_uid == vattr.va_uid) 296191783Srmacklem error = 0; 297191783Srmacklem } 298191783Srmacklem } 299191783Srmacklem if (vpislocked == 0) 300224082Szack NFSVOPUNLOCK(vp, 0); 301224086Szack 302224086Szackout: 303224086Szack NFSEXITCODE(error); 304191783Srmacklem return (error); 305191783Srmacklem} 306191783Srmacklem 307191783Srmacklem/* 308191783Srmacklem * Set attribute(s) vnop. 309191783Srmacklem */ 310191783Srmacklemint 311191783Srmacklemnfsvno_setattr(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred, 312191783Srmacklem struct thread *p, struct nfsexstuff *exp) 313191783Srmacklem{ 314191783Srmacklem int error; 315191783Srmacklem 316191783Srmacklem error = VOP_SETATTR(vp, &nvap->na_vattr, cred); 317224086Szack NFSEXITCODE(error); 318191783Srmacklem return (error); 319191783Srmacklem} 320191783Srmacklem 321191783Srmacklem/* 322232018Srmacklem * Set up nameidata for a lookup() call and do it. 323191783Srmacklem */ 324191783Srmacklemint 325191783Srmacklemnfsvno_namei(struct nfsrv_descript *nd, struct nameidata *ndp, 326191783Srmacklem struct vnode *dp, int islocked, struct nfsexstuff *exp, struct thread *p, 327191783Srmacklem struct vnode **retdirp) 328191783Srmacklem{ 329191783Srmacklem struct componentname *cnp = &ndp->ni_cnd; 330191783Srmacklem int i; 331191783Srmacklem struct iovec aiov; 332191783Srmacklem struct uio auio; 333191783Srmacklem int lockleaf = (cnp->cn_flags & LOCKLEAF) != 0, linklen; 334191783Srmacklem int error = 0, crossmnt; 335191783Srmacklem char *cp; 336191783Srmacklem 337191783Srmacklem *retdirp = NULL; 338191783Srmacklem cnp->cn_nameptr = cnp->cn_pnbuf; 339225356Srmacklem ndp->ni_strictrelative = 0; 340191783Srmacklem /* 341191783Srmacklem * Extract and set starting directory. 342191783Srmacklem */ 343191783Srmacklem if (dp->v_type != VDIR) { 344191783Srmacklem if (islocked) 345191783Srmacklem vput(dp); 346191783Srmacklem else 347191783Srmacklem vrele(dp); 348191783Srmacklem nfsvno_relpathbuf(ndp); 349224086Szack error = ENOTDIR; 350224086Szack goto out1; 351191783Srmacklem } 352191783Srmacklem if (islocked) 353224080Szack NFSVOPUNLOCK(dp, 0); 354191783Srmacklem VREF(dp); 355191783Srmacklem *retdirp = dp; 356191783Srmacklem if (NFSVNO_EXRDONLY(exp)) 357191783Srmacklem cnp->cn_flags |= RDONLY; 358191783Srmacklem ndp->ni_segflg = UIO_SYSSPACE; 359191783Srmacklem crossmnt = 1; 360191783Srmacklem 361191783Srmacklem if (nd->nd_flag & ND_PUBLOOKUP) { 362191783Srmacklem ndp->ni_loopcnt = 0; 363191783Srmacklem if (cnp->cn_pnbuf[0] == '/') { 364191783Srmacklem vrele(dp); 365191783Srmacklem /* 366191783Srmacklem * Check for degenerate pathnames here, since lookup() 367191783Srmacklem * panics on them. 368191783Srmacklem */ 369191783Srmacklem for (i = 1; i < ndp->ni_pathlen; i++) 370191783Srmacklem if (cnp->cn_pnbuf[i] != '/') 371191783Srmacklem break; 372191783Srmacklem if (i == ndp->ni_pathlen) { 373191783Srmacklem error = NFSERR_ACCES; 374191783Srmacklem goto out; 375191783Srmacklem } 376191783Srmacklem dp = rootvnode; 377191783Srmacklem VREF(dp); 378191783Srmacklem } 379220645Srmacklem } else if ((nfsrv_enable_crossmntpt == 0 && NFSVNO_EXPORTED(exp)) || 380191783Srmacklem (nd->nd_flag & ND_NFSV4) == 0) { 381191783Srmacklem /* 382191783Srmacklem * Only cross mount points for NFSv4 when doing a 383191783Srmacklem * mount while traversing the file system above 384220645Srmacklem * the mount point, unless nfsrv_enable_crossmntpt is set. 385191783Srmacklem */ 386191783Srmacklem cnp->cn_flags |= NOCROSSMOUNT; 387191783Srmacklem crossmnt = 0; 388191783Srmacklem } 389191783Srmacklem 390191783Srmacklem /* 391191783Srmacklem * Initialize for scan, set ni_startdir and bump ref on dp again 392191783Srmacklem * becuase lookup() will dereference ni_startdir. 393191783Srmacklem */ 394191783Srmacklem 395191783Srmacklem cnp->cn_thread = p; 396191783Srmacklem ndp->ni_startdir = dp; 397191783Srmacklem ndp->ni_rootdir = rootvnode; 398232683Srmacklem ndp->ni_topdir = NULL; 399191783Srmacklem 400191783Srmacklem if (!lockleaf) 401191783Srmacklem cnp->cn_flags |= LOCKLEAF; 402191783Srmacklem for (;;) { 403191783Srmacklem cnp->cn_nameptr = cnp->cn_pnbuf; 404191783Srmacklem /* 405191783Srmacklem * Call lookup() to do the real work. If an error occurs, 406191783Srmacklem * ndp->ni_vp and ni_dvp are left uninitialized or NULL and 407191783Srmacklem * we do not have to dereference anything before returning. 408191783Srmacklem * In either case ni_startdir will be dereferenced and NULLed 409191783Srmacklem * out. 410191783Srmacklem */ 411191783Srmacklem error = lookup(ndp); 412191783Srmacklem if (error) 413191783Srmacklem break; 414191783Srmacklem 415191783Srmacklem /* 416191783Srmacklem * Check for encountering a symbolic link. Trivial 417191783Srmacklem * termination occurs if no symlink encountered. 418191783Srmacklem */ 419191783Srmacklem if ((cnp->cn_flags & ISSYMLINK) == 0) { 420191783Srmacklem if ((cnp->cn_flags & (SAVENAME | SAVESTART)) == 0) 421191783Srmacklem nfsvno_relpathbuf(ndp); 422191783Srmacklem if (ndp->ni_vp && !lockleaf) 423224080Szack NFSVOPUNLOCK(ndp->ni_vp, 0); 424191783Srmacklem break; 425191783Srmacklem } 426191783Srmacklem 427191783Srmacklem /* 428191783Srmacklem * Validate symlink 429191783Srmacklem */ 430191783Srmacklem if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1) 431224080Szack NFSVOPUNLOCK(ndp->ni_dvp, 0); 432191783Srmacklem if (!(nd->nd_flag & ND_PUBLOOKUP)) { 433191783Srmacklem error = EINVAL; 434191783Srmacklem goto badlink2; 435191783Srmacklem } 436191783Srmacklem 437191783Srmacklem if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 438191783Srmacklem error = ELOOP; 439191783Srmacklem goto badlink2; 440191783Srmacklem } 441191783Srmacklem if (ndp->ni_pathlen > 1) 442191783Srmacklem cp = uma_zalloc(namei_zone, M_WAITOK); 443191783Srmacklem else 444191783Srmacklem cp = cnp->cn_pnbuf; 445191783Srmacklem aiov.iov_base = cp; 446191783Srmacklem aiov.iov_len = MAXPATHLEN; 447191783Srmacklem auio.uio_iov = &aiov; 448191783Srmacklem auio.uio_iovcnt = 1; 449191783Srmacklem auio.uio_offset = 0; 450191783Srmacklem auio.uio_rw = UIO_READ; 451191783Srmacklem auio.uio_segflg = UIO_SYSSPACE; 452191783Srmacklem auio.uio_td = NULL; 453191783Srmacklem auio.uio_resid = MAXPATHLEN; 454191783Srmacklem error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 455191783Srmacklem if (error) { 456191783Srmacklem badlink1: 457191783Srmacklem if (ndp->ni_pathlen > 1) 458191783Srmacklem uma_zfree(namei_zone, cp); 459191783Srmacklem badlink2: 460191783Srmacklem vrele(ndp->ni_dvp); 461191783Srmacklem vput(ndp->ni_vp); 462191783Srmacklem break; 463191783Srmacklem } 464191783Srmacklem linklen = MAXPATHLEN - auio.uio_resid; 465191783Srmacklem if (linklen == 0) { 466191783Srmacklem error = ENOENT; 467191783Srmacklem goto badlink1; 468191783Srmacklem } 469191783Srmacklem if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 470191783Srmacklem error = ENAMETOOLONG; 471191783Srmacklem goto badlink1; 472191783Srmacklem } 473191783Srmacklem 474191783Srmacklem /* 475191783Srmacklem * Adjust or replace path 476191783Srmacklem */ 477191783Srmacklem if (ndp->ni_pathlen > 1) { 478191783Srmacklem NFSBCOPY(ndp->ni_next, cp + linklen, ndp->ni_pathlen); 479191783Srmacklem uma_zfree(namei_zone, cnp->cn_pnbuf); 480191783Srmacklem cnp->cn_pnbuf = cp; 481191783Srmacklem } else 482191783Srmacklem cnp->cn_pnbuf[linklen] = '\0'; 483191783Srmacklem ndp->ni_pathlen += linklen; 484191783Srmacklem 485191783Srmacklem /* 486191783Srmacklem * Cleanup refs for next loop and check if root directory 487191783Srmacklem * should replace current directory. Normally ni_dvp 488191783Srmacklem * becomes the new base directory and is cleaned up when 489191783Srmacklem * we loop. Explicitly null pointers after invalidation 490191783Srmacklem * to clarify operation. 491191783Srmacklem */ 492191783Srmacklem vput(ndp->ni_vp); 493191783Srmacklem ndp->ni_vp = NULL; 494191783Srmacklem 495191783Srmacklem if (cnp->cn_pnbuf[0] == '/') { 496191783Srmacklem vrele(ndp->ni_dvp); 497191783Srmacklem ndp->ni_dvp = ndp->ni_rootdir; 498191783Srmacklem VREF(ndp->ni_dvp); 499191783Srmacklem } 500191783Srmacklem ndp->ni_startdir = ndp->ni_dvp; 501191783Srmacklem ndp->ni_dvp = NULL; 502191783Srmacklem } 503191783Srmacklem if (!lockleaf) 504191783Srmacklem cnp->cn_flags &= ~LOCKLEAF; 505191783Srmacklem 506191783Srmacklemout: 507191783Srmacklem if (error) { 508191783Srmacklem uma_zfree(namei_zone, cnp->cn_pnbuf); 509191783Srmacklem ndp->ni_vp = NULL; 510191783Srmacklem ndp->ni_dvp = NULL; 511191783Srmacklem ndp->ni_startdir = NULL; 512191783Srmacklem cnp->cn_flags &= ~HASBUF; 513191783Srmacklem } else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) { 514191783Srmacklem ndp->ni_dvp = NULL; 515191783Srmacklem } 516224086Szack 517224086Szackout1: 518224086Szack NFSEXITCODE2(error, nd); 519191783Srmacklem return (error); 520191783Srmacklem} 521191783Srmacklem 522191783Srmacklem/* 523191783Srmacklem * Set up a pathname buffer and return a pointer to it and, optionally 524191783Srmacklem * set a hash pointer. 525191783Srmacklem */ 526191783Srmacklemvoid 527191783Srmacklemnfsvno_setpathbuf(struct nameidata *ndp, char **bufpp, u_long **hashpp) 528191783Srmacklem{ 529191783Srmacklem struct componentname *cnp = &ndp->ni_cnd; 530191783Srmacklem 531191783Srmacklem cnp->cn_flags |= (NOMACCHECK | HASBUF); 532191783Srmacklem cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK); 533191783Srmacklem if (hashpp != NULL) 534191783Srmacklem *hashpp = NULL; 535191783Srmacklem *bufpp = cnp->cn_pnbuf; 536191783Srmacklem} 537191783Srmacklem 538191783Srmacklem/* 539191783Srmacklem * Release the above path buffer, if not released by nfsvno_namei(). 540191783Srmacklem */ 541191783Srmacklemvoid 542191783Srmacklemnfsvno_relpathbuf(struct nameidata *ndp) 543191783Srmacklem{ 544191783Srmacklem 545191783Srmacklem if ((ndp->ni_cnd.cn_flags & HASBUF) == 0) 546191783Srmacklem panic("nfsrelpath"); 547191783Srmacklem uma_zfree(namei_zone, ndp->ni_cnd.cn_pnbuf); 548191783Srmacklem ndp->ni_cnd.cn_flags &= ~HASBUF; 549191783Srmacklem} 550191783Srmacklem 551191783Srmacklem/* 552191783Srmacklem * Readlink vnode op into an mbuf list. 553191783Srmacklem */ 554191783Srmacklemint 555191783Srmacklemnfsvno_readlink(struct vnode *vp, struct ucred *cred, struct thread *p, 556191783Srmacklem struct mbuf **mpp, struct mbuf **mpendp, int *lenp) 557191783Srmacklem{ 558191783Srmacklem struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; 559191783Srmacklem struct iovec *ivp = iv; 560191783Srmacklem struct uio io, *uiop = &io; 561191783Srmacklem struct mbuf *mp, *mp2 = NULL, *mp3 = NULL; 562224086Szack int i, len, tlen, error = 0; 563191783Srmacklem 564191783Srmacklem len = 0; 565191783Srmacklem i = 0; 566191783Srmacklem while (len < NFS_MAXPATHLEN) { 567191783Srmacklem NFSMGET(mp); 568191783Srmacklem MCLGET(mp, M_WAIT); 569191783Srmacklem mp->m_len = NFSMSIZ(mp); 570191783Srmacklem if (len == 0) { 571191783Srmacklem mp3 = mp2 = mp; 572191783Srmacklem } else { 573191783Srmacklem mp2->m_next = mp; 574191783Srmacklem mp2 = mp; 575191783Srmacklem } 576191783Srmacklem if ((len + mp->m_len) > NFS_MAXPATHLEN) { 577191783Srmacklem mp->m_len = NFS_MAXPATHLEN - len; 578191783Srmacklem len = NFS_MAXPATHLEN; 579191783Srmacklem } else { 580191783Srmacklem len += mp->m_len; 581191783Srmacklem } 582191783Srmacklem ivp->iov_base = mtod(mp, caddr_t); 583191783Srmacklem ivp->iov_len = mp->m_len; 584191783Srmacklem i++; 585191783Srmacklem ivp++; 586191783Srmacklem } 587191783Srmacklem uiop->uio_iov = iv; 588191783Srmacklem uiop->uio_iovcnt = i; 589191783Srmacklem uiop->uio_offset = 0; 590191783Srmacklem uiop->uio_resid = len; 591191783Srmacklem uiop->uio_rw = UIO_READ; 592191783Srmacklem uiop->uio_segflg = UIO_SYSSPACE; 593191783Srmacklem uiop->uio_td = NULL; 594191783Srmacklem error = VOP_READLINK(vp, uiop, cred); 595191783Srmacklem if (error) { 596191783Srmacklem m_freem(mp3); 597191783Srmacklem *lenp = 0; 598224086Szack goto out; 599191783Srmacklem } 600191783Srmacklem if (uiop->uio_resid > 0) { 601191783Srmacklem len -= uiop->uio_resid; 602191783Srmacklem tlen = NFSM_RNDUP(len); 603191783Srmacklem nfsrv_adj(mp3, NFS_MAXPATHLEN - tlen, tlen - len); 604191783Srmacklem } 605191783Srmacklem *lenp = len; 606191783Srmacklem *mpp = mp3; 607191783Srmacklem *mpendp = mp; 608224086Szack 609224086Szackout: 610224086Szack NFSEXITCODE(error); 611224086Szack return (error); 612191783Srmacklem} 613191783Srmacklem 614191783Srmacklem/* 615191783Srmacklem * Read vnode op call into mbuf list. 616191783Srmacklem */ 617191783Srmacklemint 618191783Srmacklemnfsvno_read(struct vnode *vp, off_t off, int cnt, struct ucred *cred, 619191783Srmacklem struct thread *p, struct mbuf **mpp, struct mbuf **mpendp) 620191783Srmacklem{ 621191783Srmacklem struct mbuf *m; 622191783Srmacklem int i; 623191783Srmacklem struct iovec *iv; 624191783Srmacklem struct iovec *iv2; 625229617Sjhb int error = 0, len, left, siz, tlen, ioflag = 0; 626191783Srmacklem struct mbuf *m2 = NULL, *m3; 627191783Srmacklem struct uio io, *uiop = &io; 628191783Srmacklem struct nfsheur *nh; 629191783Srmacklem 630191783Srmacklem len = left = NFSM_RNDUP(cnt); 631191783Srmacklem m3 = NULL; 632191783Srmacklem /* 633191783Srmacklem * Generate the mbuf list with the uio_iov ref. to it. 634191783Srmacklem */ 635191783Srmacklem i = 0; 636191783Srmacklem while (left > 0) { 637191783Srmacklem NFSMGET(m); 638191783Srmacklem MCLGET(m, M_WAIT); 639191783Srmacklem m->m_len = 0; 640191783Srmacklem siz = min(M_TRAILINGSPACE(m), left); 641191783Srmacklem left -= siz; 642191783Srmacklem i++; 643191783Srmacklem if (m3) 644191783Srmacklem m2->m_next = m; 645191783Srmacklem else 646191783Srmacklem m3 = m; 647191783Srmacklem m2 = m; 648191783Srmacklem } 649191783Srmacklem MALLOC(iv, struct iovec *, i * sizeof (struct iovec), 650191783Srmacklem M_TEMP, M_WAITOK); 651191783Srmacklem uiop->uio_iov = iv2 = iv; 652191783Srmacklem m = m3; 653191783Srmacklem left = len; 654191783Srmacklem i = 0; 655191783Srmacklem while (left > 0) { 656191783Srmacklem if (m == NULL) 657191783Srmacklem panic("nfsvno_read iov"); 658191783Srmacklem siz = min(M_TRAILINGSPACE(m), left); 659191783Srmacklem if (siz > 0) { 660191783Srmacklem iv->iov_base = mtod(m, caddr_t) + m->m_len; 661191783Srmacklem iv->iov_len = siz; 662191783Srmacklem m->m_len += siz; 663191783Srmacklem left -= siz; 664191783Srmacklem iv++; 665191783Srmacklem i++; 666191783Srmacklem } 667191783Srmacklem m = m->m_next; 668191783Srmacklem } 669191783Srmacklem uiop->uio_iovcnt = i; 670191783Srmacklem uiop->uio_offset = off; 671191783Srmacklem uiop->uio_resid = len; 672191783Srmacklem uiop->uio_rw = UIO_READ; 673191783Srmacklem uiop->uio_segflg = UIO_SYSSPACE; 674229617Sjhb nh = nfsrv_sequential_heuristic(uiop, vp); 675229617Sjhb ioflag |= nh->nh_seqcount << IO_SEQSHIFT; 676191783Srmacklem error = VOP_READ(vp, uiop, IO_NODELOCKED | ioflag, cred); 677191783Srmacklem FREE((caddr_t)iv2, M_TEMP); 678191783Srmacklem if (error) { 679191783Srmacklem m_freem(m3); 680191783Srmacklem *mpp = NULL; 681224086Szack goto out; 682191783Srmacklem } 683229617Sjhb nh->nh_nextoff = uiop->uio_offset; 684191783Srmacklem tlen = len - uiop->uio_resid; 685191783Srmacklem cnt = cnt < tlen ? cnt : tlen; 686191783Srmacklem tlen = NFSM_RNDUP(cnt); 687191783Srmacklem if (tlen == 0) { 688191783Srmacklem m_freem(m3); 689191783Srmacklem m3 = NULL; 690191783Srmacklem } else if (len != tlen || tlen != cnt) 691191783Srmacklem nfsrv_adj(m3, len - tlen, tlen - cnt); 692191783Srmacklem *mpp = m3; 693191783Srmacklem *mpendp = m2; 694224086Szack 695224086Szackout: 696224086Szack NFSEXITCODE(error); 697224086Szack return (error); 698191783Srmacklem} 699191783Srmacklem 700191783Srmacklem/* 701191783Srmacklem * Write vnode op from an mbuf list. 702191783Srmacklem */ 703191783Srmacklemint 704191783Srmacklemnfsvno_write(struct vnode *vp, off_t off, int retlen, int cnt, int stable, 705191783Srmacklem struct mbuf *mp, char *cp, struct ucred *cred, struct thread *p) 706191783Srmacklem{ 707191783Srmacklem struct iovec *ivp; 708191783Srmacklem int i, len; 709191783Srmacklem struct iovec *iv; 710191783Srmacklem int ioflags, error; 711191783Srmacklem struct uio io, *uiop = &io; 712229617Sjhb struct nfsheur *nh; 713191783Srmacklem 714191783Srmacklem MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP, 715191783Srmacklem M_WAITOK); 716191783Srmacklem uiop->uio_iov = iv = ivp; 717191783Srmacklem uiop->uio_iovcnt = cnt; 718191783Srmacklem i = mtod(mp, caddr_t) + mp->m_len - cp; 719191783Srmacklem len = retlen; 720191783Srmacklem while (len > 0) { 721191783Srmacklem if (mp == NULL) 722191783Srmacklem panic("nfsvno_write"); 723191783Srmacklem if (i > 0) { 724191783Srmacklem i = min(i, len); 725191783Srmacklem ivp->iov_base = cp; 726191783Srmacklem ivp->iov_len = i; 727191783Srmacklem ivp++; 728191783Srmacklem len -= i; 729191783Srmacklem } 730191783Srmacklem mp = mp->m_next; 731191783Srmacklem if (mp) { 732191783Srmacklem i = mp->m_len; 733191783Srmacklem cp = mtod(mp, caddr_t); 734191783Srmacklem } 735191783Srmacklem } 736191783Srmacklem 737191783Srmacklem if (stable == NFSWRITE_UNSTABLE) 738191783Srmacklem ioflags = IO_NODELOCKED; 739191783Srmacklem else 740191783Srmacklem ioflags = (IO_SYNC | IO_NODELOCKED); 741191783Srmacklem uiop->uio_resid = retlen; 742191783Srmacklem uiop->uio_rw = UIO_WRITE; 743191783Srmacklem uiop->uio_segflg = UIO_SYSSPACE; 744191783Srmacklem NFSUIOPROC(uiop, p); 745191783Srmacklem uiop->uio_offset = off; 746229617Sjhb nh = nfsrv_sequential_heuristic(uiop, vp); 747229617Sjhb ioflags |= nh->nh_seqcount << IO_SEQSHIFT; 748191783Srmacklem error = VOP_WRITE(vp, uiop, ioflags, cred); 749229617Sjhb if (error == 0) 750229617Sjhb nh->nh_nextoff = uiop->uio_offset; 751191783Srmacklem FREE((caddr_t)iv, M_TEMP); 752224086Szack 753224086Szack NFSEXITCODE(error); 754191783Srmacklem return (error); 755191783Srmacklem} 756191783Srmacklem 757191783Srmacklem/* 758191783Srmacklem * Common code for creating a regular file (plus special files for V2). 759191783Srmacklem */ 760191783Srmacklemint 761191783Srmacklemnfsvno_createsub(struct nfsrv_descript *nd, struct nameidata *ndp, 762191783Srmacklem struct vnode **vpp, struct nfsvattr *nvap, int *exclusive_flagp, 763199616Srmacklem int32_t *cverf, NFSDEV_T rdev, struct thread *p, struct nfsexstuff *exp) 764191783Srmacklem{ 765191783Srmacklem u_quad_t tempsize; 766191783Srmacklem int error; 767191783Srmacklem 768191783Srmacklem error = nd->nd_repstat; 769191783Srmacklem if (!error && ndp->ni_vp == NULL) { 770191783Srmacklem if (nvap->na_type == VREG || nvap->na_type == VSOCK) { 771191783Srmacklem vrele(ndp->ni_startdir); 772191783Srmacklem error = VOP_CREATE(ndp->ni_dvp, 773191783Srmacklem &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); 774191783Srmacklem vput(ndp->ni_dvp); 775191783Srmacklem nfsvno_relpathbuf(ndp); 776191783Srmacklem if (!error) { 777191783Srmacklem if (*exclusive_flagp) { 778191783Srmacklem *exclusive_flagp = 0; 779191783Srmacklem NFSVNO_ATTRINIT(nvap); 780199616Srmacklem nvap->na_atime.tv_sec = cverf[0]; 781199616Srmacklem nvap->na_atime.tv_nsec = cverf[1]; 782191783Srmacklem error = VOP_SETATTR(ndp->ni_vp, 783191783Srmacklem &nvap->na_vattr, nd->nd_cred); 784191783Srmacklem } 785191783Srmacklem } 786191783Srmacklem /* 787191783Srmacklem * NFS V2 Only. nfsrvd_mknod() does this for V3. 788191783Srmacklem * (This implies, just get out on an error.) 789191783Srmacklem */ 790191783Srmacklem } else if (nvap->na_type == VCHR || nvap->na_type == VBLK || 791191783Srmacklem nvap->na_type == VFIFO) { 792191783Srmacklem if (nvap->na_type == VCHR && rdev == 0xffffffff) 793191783Srmacklem nvap->na_type = VFIFO; 794191783Srmacklem if (nvap->na_type != VFIFO && 795191783Srmacklem (error = priv_check_cred(nd->nd_cred, 796191783Srmacklem PRIV_VFS_MKNOD_DEV, 0))) { 797191783Srmacklem vrele(ndp->ni_startdir); 798191783Srmacklem nfsvno_relpathbuf(ndp); 799191783Srmacklem vput(ndp->ni_dvp); 800224086Szack goto out; 801191783Srmacklem } 802191783Srmacklem nvap->na_rdev = rdev; 803191783Srmacklem error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp, 804191783Srmacklem &ndp->ni_cnd, &nvap->na_vattr); 805191783Srmacklem vput(ndp->ni_dvp); 806191783Srmacklem nfsvno_relpathbuf(ndp); 807220546Srmacklem vrele(ndp->ni_startdir); 808220546Srmacklem if (error) 809224086Szack goto out; 810191783Srmacklem } else { 811191783Srmacklem vrele(ndp->ni_startdir); 812191783Srmacklem nfsvno_relpathbuf(ndp); 813191783Srmacklem vput(ndp->ni_dvp); 814224086Szack error = ENXIO; 815224086Szack goto out; 816191783Srmacklem } 817191783Srmacklem *vpp = ndp->ni_vp; 818191783Srmacklem } else { 819191783Srmacklem /* 820191783Srmacklem * Handle cases where error is already set and/or 821191783Srmacklem * the file exists. 822191783Srmacklem * 1 - clean up the lookup 823191783Srmacklem * 2 - iff !error and na_size set, truncate it 824191783Srmacklem */ 825191783Srmacklem vrele(ndp->ni_startdir); 826191783Srmacklem nfsvno_relpathbuf(ndp); 827191783Srmacklem *vpp = ndp->ni_vp; 828191783Srmacklem if (ndp->ni_dvp == *vpp) 829191783Srmacklem vrele(ndp->ni_dvp); 830191783Srmacklem else 831191783Srmacklem vput(ndp->ni_dvp); 832191783Srmacklem if (!error && nvap->na_size != VNOVAL) { 833200999Srmacklem error = nfsvno_accchk(*vpp, VWRITE, 834191783Srmacklem nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, 835200999Srmacklem NFSACCCHK_VPISLOCKED, NULL); 836191783Srmacklem if (!error) { 837191783Srmacklem tempsize = nvap->na_size; 838191783Srmacklem NFSVNO_ATTRINIT(nvap); 839191783Srmacklem nvap->na_size = tempsize; 840191783Srmacklem error = VOP_SETATTR(*vpp, 841191783Srmacklem &nvap->na_vattr, nd->nd_cred); 842191783Srmacklem } 843191783Srmacklem } 844191783Srmacklem if (error) 845191783Srmacklem vput(*vpp); 846191783Srmacklem } 847224086Szack 848224086Szackout: 849224086Szack NFSEXITCODE(error); 850191783Srmacklem return (error); 851191783Srmacklem} 852191783Srmacklem 853191783Srmacklem/* 854191783Srmacklem * Do a mknod vnode op. 855191783Srmacklem */ 856191783Srmacklemint 857191783Srmacklemnfsvno_mknod(struct nameidata *ndp, struct nfsvattr *nvap, struct ucred *cred, 858191783Srmacklem struct thread *p) 859191783Srmacklem{ 860191783Srmacklem int error = 0; 861191783Srmacklem enum vtype vtyp; 862191783Srmacklem 863191783Srmacklem vtyp = nvap->na_type; 864191783Srmacklem /* 865191783Srmacklem * Iff doesn't exist, create it. 866191783Srmacklem */ 867191783Srmacklem if (ndp->ni_vp) { 868191783Srmacklem vrele(ndp->ni_startdir); 869191783Srmacklem nfsvno_relpathbuf(ndp); 870191783Srmacklem vput(ndp->ni_dvp); 871191783Srmacklem vrele(ndp->ni_vp); 872224086Szack error = EEXIST; 873224086Szack goto out; 874191783Srmacklem } 875191783Srmacklem if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { 876191783Srmacklem vrele(ndp->ni_startdir); 877191783Srmacklem nfsvno_relpathbuf(ndp); 878191783Srmacklem vput(ndp->ni_dvp); 879224086Szack error = NFSERR_BADTYPE; 880224086Szack goto out; 881191783Srmacklem } 882191783Srmacklem if (vtyp == VSOCK) { 883191783Srmacklem vrele(ndp->ni_startdir); 884191783Srmacklem error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 885191783Srmacklem &ndp->ni_cnd, &nvap->na_vattr); 886191783Srmacklem vput(ndp->ni_dvp); 887191783Srmacklem nfsvno_relpathbuf(ndp); 888191783Srmacklem } else { 889191783Srmacklem if (nvap->na_type != VFIFO && 890191783Srmacklem (error = priv_check_cred(cred, PRIV_VFS_MKNOD_DEV, 0))) { 891191783Srmacklem vrele(ndp->ni_startdir); 892191783Srmacklem nfsvno_relpathbuf(ndp); 893191783Srmacklem vput(ndp->ni_dvp); 894224086Szack goto out; 895191783Srmacklem } 896191783Srmacklem error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp, 897191783Srmacklem &ndp->ni_cnd, &nvap->na_vattr); 898191783Srmacklem vput(ndp->ni_dvp); 899191783Srmacklem nfsvno_relpathbuf(ndp); 900216898Srmacklem vrele(ndp->ni_startdir); 901191783Srmacklem /* 902191783Srmacklem * Since VOP_MKNOD returns the ni_vp, I can't 903191783Srmacklem * see any reason to do the lookup. 904191783Srmacklem */ 905191783Srmacklem } 906224086Szack 907224086Szackout: 908224086Szack NFSEXITCODE(error); 909191783Srmacklem return (error); 910191783Srmacklem} 911191783Srmacklem 912191783Srmacklem/* 913191783Srmacklem * Mkdir vnode op. 914191783Srmacklem */ 915191783Srmacklemint 916191783Srmacklemnfsvno_mkdir(struct nameidata *ndp, struct nfsvattr *nvap, uid_t saved_uid, 917191783Srmacklem struct ucred *cred, struct thread *p, struct nfsexstuff *exp) 918191783Srmacklem{ 919191783Srmacklem int error = 0; 920191783Srmacklem 921191783Srmacklem if (ndp->ni_vp != NULL) { 922191783Srmacklem if (ndp->ni_dvp == ndp->ni_vp) 923191783Srmacklem vrele(ndp->ni_dvp); 924191783Srmacklem else 925191783Srmacklem vput(ndp->ni_dvp); 926191783Srmacklem vrele(ndp->ni_vp); 927206063Srmacklem nfsvno_relpathbuf(ndp); 928224086Szack error = EEXIST; 929224086Szack goto out; 930191783Srmacklem } 931191783Srmacklem error = VOP_MKDIR(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, 932191783Srmacklem &nvap->na_vattr); 933191783Srmacklem vput(ndp->ni_dvp); 934191783Srmacklem nfsvno_relpathbuf(ndp); 935224086Szack 936224086Szackout: 937224086Szack NFSEXITCODE(error); 938191783Srmacklem return (error); 939191783Srmacklem} 940191783Srmacklem 941191783Srmacklem/* 942191783Srmacklem * symlink vnode op. 943191783Srmacklem */ 944191783Srmacklemint 945191783Srmacklemnfsvno_symlink(struct nameidata *ndp, struct nfsvattr *nvap, char *pathcp, 946191783Srmacklem int pathlen, int not_v2, uid_t saved_uid, struct ucred *cred, struct thread *p, 947191783Srmacklem struct nfsexstuff *exp) 948191783Srmacklem{ 949191783Srmacklem int error = 0; 950191783Srmacklem 951191783Srmacklem if (ndp->ni_vp) { 952191783Srmacklem vrele(ndp->ni_startdir); 953191783Srmacklem nfsvno_relpathbuf(ndp); 954191783Srmacklem if (ndp->ni_dvp == ndp->ni_vp) 955191783Srmacklem vrele(ndp->ni_dvp); 956191783Srmacklem else 957191783Srmacklem vput(ndp->ni_dvp); 958191783Srmacklem vrele(ndp->ni_vp); 959224086Szack error = EEXIST; 960224086Szack goto out; 961191783Srmacklem } 962191783Srmacklem 963191783Srmacklem error = VOP_SYMLINK(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, 964191783Srmacklem &nvap->na_vattr, pathcp); 965191783Srmacklem vput(ndp->ni_dvp); 966191783Srmacklem vrele(ndp->ni_startdir); 967191783Srmacklem nfsvno_relpathbuf(ndp); 968191783Srmacklem /* 969191783Srmacklem * Although FreeBSD still had the lookup code in 970191783Srmacklem * it for 7/current, there doesn't seem to be any 971191783Srmacklem * point, since VOP_SYMLINK() returns the ni_vp. 972191783Srmacklem * Just vput it for v2. 973191783Srmacklem */ 974191783Srmacklem if (!not_v2 && !error) 975191783Srmacklem vput(ndp->ni_vp); 976224086Szack 977224086Szackout: 978224086Szack NFSEXITCODE(error); 979191783Srmacklem return (error); 980191783Srmacklem} 981191783Srmacklem 982191783Srmacklem/* 983191783Srmacklem * Parse symbolic link arguments. 984191783Srmacklem * This function has an ugly side effect. It will MALLOC() an area for 985191783Srmacklem * the symlink and set iov_base to point to it, only if it succeeds. 986191783Srmacklem * So, if it returns with uiop->uio_iov->iov_base != NULL, that must 987191783Srmacklem * be FREE'd later. 988191783Srmacklem */ 989191783Srmacklemint 990191783Srmacklemnfsvno_getsymlink(struct nfsrv_descript *nd, struct nfsvattr *nvap, 991191783Srmacklem struct thread *p, char **pathcpp, int *lenp) 992191783Srmacklem{ 993191783Srmacklem u_int32_t *tl; 994191783Srmacklem char *pathcp = NULL; 995191783Srmacklem int error = 0, len; 996191783Srmacklem struct nfsv2_sattr *sp; 997191783Srmacklem 998191783Srmacklem *pathcpp = NULL; 999191783Srmacklem *lenp = 0; 1000191783Srmacklem if ((nd->nd_flag & ND_NFSV3) && 1001191783Srmacklem (error = nfsrv_sattr(nd, nvap, NULL, NULL, p))) 1002191783Srmacklem goto nfsmout; 1003191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 1004191783Srmacklem len = fxdr_unsigned(int, *tl); 1005191783Srmacklem if (len > NFS_MAXPATHLEN || len <= 0) { 1006191783Srmacklem error = EBADRPC; 1007191783Srmacklem goto nfsmout; 1008191783Srmacklem } 1009191783Srmacklem MALLOC(pathcp, caddr_t, len + 1, M_TEMP, M_WAITOK); 1010191783Srmacklem error = nfsrv_mtostr(nd, pathcp, len); 1011191783Srmacklem if (error) 1012191783Srmacklem goto nfsmout; 1013191783Srmacklem if (nd->nd_flag & ND_NFSV2) { 1014191783Srmacklem NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 1015191783Srmacklem nvap->na_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); 1016191783Srmacklem } 1017191783Srmacklem *pathcpp = pathcp; 1018191783Srmacklem *lenp = len; 1019224086Szack NFSEXITCODE2(0, nd); 1020191783Srmacklem return (0); 1021191783Srmacklemnfsmout: 1022191783Srmacklem if (pathcp) 1023191783Srmacklem free(pathcp, M_TEMP); 1024224086Szack NFSEXITCODE2(error, nd); 1025191783Srmacklem return (error); 1026191783Srmacklem} 1027191783Srmacklem 1028191783Srmacklem/* 1029191783Srmacklem * Remove a non-directory object. 1030191783Srmacklem */ 1031191783Srmacklemint 1032191783Srmacklemnfsvno_removesub(struct nameidata *ndp, int is_v4, struct ucred *cred, 1033191783Srmacklem struct thread *p, struct nfsexstuff *exp) 1034191783Srmacklem{ 1035191783Srmacklem struct vnode *vp; 1036191783Srmacklem int error = 0; 1037191783Srmacklem 1038191783Srmacklem vp = ndp->ni_vp; 1039191783Srmacklem if (vp->v_type == VDIR) 1040191783Srmacklem error = NFSERR_ISDIR; 1041191783Srmacklem else if (is_v4) 1042191783Srmacklem error = nfsrv_checkremove(vp, 1, p); 1043191783Srmacklem if (!error) 1044191783Srmacklem error = VOP_REMOVE(ndp->ni_dvp, vp, &ndp->ni_cnd); 1045191783Srmacklem if (ndp->ni_dvp == vp) 1046191783Srmacklem vrele(ndp->ni_dvp); 1047191783Srmacklem else 1048191783Srmacklem vput(ndp->ni_dvp); 1049191783Srmacklem vput(vp); 1050224086Szack NFSEXITCODE(error); 1051191783Srmacklem return (error); 1052191783Srmacklem} 1053191783Srmacklem 1054191783Srmacklem/* 1055191783Srmacklem * Remove a directory. 1056191783Srmacklem */ 1057191783Srmacklemint 1058191783Srmacklemnfsvno_rmdirsub(struct nameidata *ndp, int is_v4, struct ucred *cred, 1059191783Srmacklem struct thread *p, struct nfsexstuff *exp) 1060191783Srmacklem{ 1061191783Srmacklem struct vnode *vp; 1062191783Srmacklem int error = 0; 1063191783Srmacklem 1064191783Srmacklem vp = ndp->ni_vp; 1065191783Srmacklem if (vp->v_type != VDIR) { 1066191783Srmacklem error = ENOTDIR; 1067191783Srmacklem goto out; 1068191783Srmacklem } 1069191783Srmacklem /* 1070191783Srmacklem * No rmdir "." please. 1071191783Srmacklem */ 1072191783Srmacklem if (ndp->ni_dvp == vp) { 1073191783Srmacklem error = EINVAL; 1074191783Srmacklem goto out; 1075191783Srmacklem } 1076191783Srmacklem /* 1077191783Srmacklem * The root of a mounted filesystem cannot be deleted. 1078191783Srmacklem */ 1079191783Srmacklem if (vp->v_vflag & VV_ROOT) 1080191783Srmacklem error = EBUSY; 1081191783Srmacklemout: 1082191783Srmacklem if (!error) 1083191783Srmacklem error = VOP_RMDIR(ndp->ni_dvp, vp, &ndp->ni_cnd); 1084191783Srmacklem if (ndp->ni_dvp == vp) 1085191783Srmacklem vrele(ndp->ni_dvp); 1086191783Srmacklem else 1087191783Srmacklem vput(ndp->ni_dvp); 1088191783Srmacklem vput(vp); 1089224086Szack NFSEXITCODE(error); 1090191783Srmacklem return (error); 1091191783Srmacklem} 1092191783Srmacklem 1093191783Srmacklem/* 1094191783Srmacklem * Rename vnode op. 1095191783Srmacklem */ 1096191783Srmacklemint 1097191783Srmacklemnfsvno_rename(struct nameidata *fromndp, struct nameidata *tondp, 1098191783Srmacklem u_int32_t ndstat, u_int32_t ndflag, struct ucred *cred, struct thread *p) 1099191783Srmacklem{ 1100191783Srmacklem struct vnode *fvp, *tvp, *tdvp; 1101191783Srmacklem int error = 0; 1102191783Srmacklem 1103191783Srmacklem fvp = fromndp->ni_vp; 1104191783Srmacklem if (ndstat) { 1105191783Srmacklem vrele(fromndp->ni_dvp); 1106191783Srmacklem vrele(fvp); 1107191783Srmacklem error = ndstat; 1108191783Srmacklem goto out1; 1109191783Srmacklem } 1110191783Srmacklem tdvp = tondp->ni_dvp; 1111191783Srmacklem tvp = tondp->ni_vp; 1112191783Srmacklem if (tvp != NULL) { 1113191783Srmacklem if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 1114191783Srmacklem error = (ndflag & ND_NFSV2) ? EISDIR : EEXIST; 1115191783Srmacklem goto out; 1116191783Srmacklem } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 1117191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTDIR : EEXIST; 1118191783Srmacklem goto out; 1119191783Srmacklem } 1120191783Srmacklem if (tvp->v_type == VDIR && tvp->v_mountedhere) { 1121191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; 1122191783Srmacklem goto out; 1123191783Srmacklem } 1124191783Srmacklem 1125191783Srmacklem /* 1126191783Srmacklem * A rename to '.' or '..' results in a prematurely 1127191783Srmacklem * unlocked vnode on FreeBSD5, so I'm just going to fail that 1128191783Srmacklem * here. 1129191783Srmacklem */ 1130191783Srmacklem if ((tondp->ni_cnd.cn_namelen == 1 && 1131191783Srmacklem tondp->ni_cnd.cn_nameptr[0] == '.') || 1132191783Srmacklem (tondp->ni_cnd.cn_namelen == 2 && 1133191783Srmacklem tondp->ni_cnd.cn_nameptr[0] == '.' && 1134191783Srmacklem tondp->ni_cnd.cn_nameptr[1] == '.')) { 1135191783Srmacklem error = EINVAL; 1136191783Srmacklem goto out; 1137191783Srmacklem } 1138191783Srmacklem } 1139191783Srmacklem if (fvp->v_type == VDIR && fvp->v_mountedhere) { 1140191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; 1141191783Srmacklem goto out; 1142191783Srmacklem } 1143191783Srmacklem if (fvp->v_mount != tdvp->v_mount) { 1144191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; 1145191783Srmacklem goto out; 1146191783Srmacklem } 1147191783Srmacklem if (fvp == tdvp) { 1148191783Srmacklem error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EINVAL; 1149191783Srmacklem goto out; 1150191783Srmacklem } 1151191783Srmacklem if (fvp == tvp) { 1152191783Srmacklem /* 1153191783Srmacklem * If source and destination are the same, there is nothing to 1154191783Srmacklem * do. Set error to -1 to indicate this. 1155191783Srmacklem */ 1156191783Srmacklem error = -1; 1157191783Srmacklem goto out; 1158191783Srmacklem } 1159191783Srmacklem if (ndflag & ND_NFSV4) { 1160224081Szack if (NFSVOPLOCK(fvp, LK_EXCLUSIVE) == 0) { 1161216893Srmacklem error = nfsrv_checkremove(fvp, 0, p); 1162224082Szack NFSVOPUNLOCK(fvp, 0); 1163216893Srmacklem } else 1164216893Srmacklem error = EPERM; 1165191783Srmacklem if (tvp && !error) 1166191783Srmacklem error = nfsrv_checkremove(tvp, 1, p); 1167191783Srmacklem } else { 1168191783Srmacklem /* 1169191783Srmacklem * For NFSv2 and NFSv3, try to get rid of the delegation, so 1170191783Srmacklem * that the NFSv4 client won't be confused by the rename. 1171191783Srmacklem * Since nfsd_recalldelegation() can only be called on an 1172191783Srmacklem * unlocked vnode at this point and fvp is the file that will 1173191783Srmacklem * still exist after the rename, just do fvp. 1174191783Srmacklem */ 1175191783Srmacklem nfsd_recalldelegation(fvp, p); 1176191783Srmacklem } 1177191783Srmacklemout: 1178191783Srmacklem if (!error) { 1179191783Srmacklem error = VOP_RENAME(fromndp->ni_dvp, fromndp->ni_vp, 1180191783Srmacklem &fromndp->ni_cnd, tondp->ni_dvp, tondp->ni_vp, 1181191783Srmacklem &tondp->ni_cnd); 1182191783Srmacklem } else { 1183191783Srmacklem if (tdvp == tvp) 1184191783Srmacklem vrele(tdvp); 1185191783Srmacklem else 1186191783Srmacklem vput(tdvp); 1187191783Srmacklem if (tvp) 1188191783Srmacklem vput(tvp); 1189191783Srmacklem vrele(fromndp->ni_dvp); 1190191783Srmacklem vrele(fvp); 1191191783Srmacklem if (error == -1) 1192191783Srmacklem error = 0; 1193191783Srmacklem } 1194191783Srmacklem vrele(tondp->ni_startdir); 1195191783Srmacklem nfsvno_relpathbuf(tondp); 1196191783Srmacklemout1: 1197191783Srmacklem vrele(fromndp->ni_startdir); 1198191783Srmacklem nfsvno_relpathbuf(fromndp); 1199224086Szack NFSEXITCODE(error); 1200191783Srmacklem return (error); 1201191783Srmacklem} 1202191783Srmacklem 1203191783Srmacklem/* 1204191783Srmacklem * Link vnode op. 1205191783Srmacklem */ 1206191783Srmacklemint 1207191783Srmacklemnfsvno_link(struct nameidata *ndp, struct vnode *vp, struct ucred *cred, 1208191783Srmacklem struct thread *p, struct nfsexstuff *exp) 1209191783Srmacklem{ 1210191783Srmacklem struct vnode *xp; 1211191783Srmacklem int error = 0; 1212191783Srmacklem 1213191783Srmacklem xp = ndp->ni_vp; 1214191783Srmacklem if (xp != NULL) { 1215191783Srmacklem error = EEXIST; 1216191783Srmacklem } else { 1217191783Srmacklem xp = ndp->ni_dvp; 1218191783Srmacklem if (vp->v_mount != xp->v_mount) 1219191783Srmacklem error = EXDEV; 1220191783Srmacklem } 1221191783Srmacklem if (!error) { 1222224081Szack NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY); 1223216893Srmacklem if ((vp->v_iflag & VI_DOOMED) == 0) 1224216893Srmacklem error = VOP_LINK(ndp->ni_dvp, vp, &ndp->ni_cnd); 1225216893Srmacklem else 1226216893Srmacklem error = EPERM; 1227191783Srmacklem if (ndp->ni_dvp == vp) 1228191783Srmacklem vrele(ndp->ni_dvp); 1229191783Srmacklem else 1230191783Srmacklem vput(ndp->ni_dvp); 1231224082Szack NFSVOPUNLOCK(vp, 0); 1232191783Srmacklem } else { 1233191783Srmacklem if (ndp->ni_dvp == ndp->ni_vp) 1234191783Srmacklem vrele(ndp->ni_dvp); 1235191783Srmacklem else 1236191783Srmacklem vput(ndp->ni_dvp); 1237191783Srmacklem if (ndp->ni_vp) 1238191783Srmacklem vrele(ndp->ni_vp); 1239191783Srmacklem } 1240191783Srmacklem nfsvno_relpathbuf(ndp); 1241224086Szack NFSEXITCODE(error); 1242191783Srmacklem return (error); 1243191783Srmacklem} 1244191783Srmacklem 1245191783Srmacklem/* 1246191783Srmacklem * Do the fsync() appropriate for the commit. 1247191783Srmacklem */ 1248191783Srmacklemint 1249191783Srmacklemnfsvno_fsync(struct vnode *vp, u_int64_t off, int cnt, struct ucred *cred, 1250191783Srmacklem struct thread *td) 1251191783Srmacklem{ 1252191783Srmacklem int error = 0; 1253191783Srmacklem 1254229827Srmacklem /* 1255229827Srmacklem * RFC 1813 3.3.21: if count is 0, a flush from offset to the end of 1256229827Srmacklem * file is done. At this time VOP_FSYNC does not accept offset and 1257229827Srmacklem * byte count parameters so call VOP_FSYNC the whole file for now. 1258229827Srmacklem * The same is true for NFSv4: RFC 3530 Sec. 14.2.3. 1259229827Srmacklem */ 1260229827Srmacklem if (cnt == 0 || cnt > MAX_COMMIT_COUNT) { 1261191783Srmacklem /* 1262191783Srmacklem * Give up and do the whole thing 1263191783Srmacklem */ 1264191783Srmacklem if (vp->v_object && 1265191783Srmacklem (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) { 1266191783Srmacklem VM_OBJECT_LOCK(vp->v_object); 1267191783Srmacklem vm_object_page_clean(vp->v_object, 0, 0, OBJPC_SYNC); 1268191783Srmacklem VM_OBJECT_UNLOCK(vp->v_object); 1269191783Srmacklem } 1270191783Srmacklem error = VOP_FSYNC(vp, MNT_WAIT, td); 1271191783Srmacklem } else { 1272191783Srmacklem /* 1273191783Srmacklem * Locate and synchronously write any buffers that fall 1274191783Srmacklem * into the requested range. Note: we are assuming that 1275191783Srmacklem * f_iosize is a power of 2. 1276191783Srmacklem */ 1277191783Srmacklem int iosize = vp->v_mount->mnt_stat.f_iosize; 1278191783Srmacklem int iomask = iosize - 1; 1279191783Srmacklem struct bufobj *bo; 1280191783Srmacklem daddr_t lblkno; 1281191783Srmacklem 1282191783Srmacklem /* 1283191783Srmacklem * Align to iosize boundry, super-align to page boundry. 1284191783Srmacklem */ 1285191783Srmacklem if (off & iomask) { 1286191783Srmacklem cnt += off & iomask; 1287191783Srmacklem off &= ~(u_quad_t)iomask; 1288191783Srmacklem } 1289191783Srmacklem if (off & PAGE_MASK) { 1290191783Srmacklem cnt += off & PAGE_MASK; 1291191783Srmacklem off &= ~(u_quad_t)PAGE_MASK; 1292191783Srmacklem } 1293191783Srmacklem lblkno = off / iosize; 1294191783Srmacklem 1295191783Srmacklem if (vp->v_object && 1296191783Srmacklem (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) { 1297191783Srmacklem VM_OBJECT_LOCK(vp->v_object); 1298218345Salc vm_object_page_clean(vp->v_object, off, off + cnt, 1299218345Salc OBJPC_SYNC); 1300191783Srmacklem VM_OBJECT_UNLOCK(vp->v_object); 1301191783Srmacklem } 1302191783Srmacklem 1303191783Srmacklem bo = &vp->v_bufobj; 1304191783Srmacklem BO_LOCK(bo); 1305191783Srmacklem while (cnt > 0) { 1306191783Srmacklem struct buf *bp; 1307191783Srmacklem 1308191783Srmacklem /* 1309191783Srmacklem * If we have a buffer and it is marked B_DELWRI we 1310191783Srmacklem * have to lock and write it. Otherwise the prior 1311191783Srmacklem * write is assumed to have already been committed. 1312191783Srmacklem * 1313191783Srmacklem * gbincore() can return invalid buffers now so we 1314191783Srmacklem * have to check that bit as well (though B_DELWRI 1315191783Srmacklem * should not be set if B_INVAL is set there could be 1316191783Srmacklem * a race here since we haven't locked the buffer). 1317191783Srmacklem */ 1318191783Srmacklem if ((bp = gbincore(&vp->v_bufobj, lblkno)) != NULL) { 1319191783Srmacklem if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | 1320191783Srmacklem LK_INTERLOCK, BO_MTX(bo)) == ENOLCK) { 1321191783Srmacklem BO_LOCK(bo); 1322191783Srmacklem continue; /* retry */ 1323191783Srmacklem } 1324191783Srmacklem if ((bp->b_flags & (B_DELWRI|B_INVAL)) == 1325191783Srmacklem B_DELWRI) { 1326191783Srmacklem bremfree(bp); 1327191783Srmacklem bp->b_flags &= ~B_ASYNC; 1328191783Srmacklem bwrite(bp); 1329191783Srmacklem ++nfs_commit_miss; 1330191783Srmacklem } else 1331191783Srmacklem BUF_UNLOCK(bp); 1332191783Srmacklem BO_LOCK(bo); 1333191783Srmacklem } 1334191783Srmacklem ++nfs_commit_blks; 1335191783Srmacklem if (cnt < iosize) 1336191783Srmacklem break; 1337191783Srmacklem cnt -= iosize; 1338191783Srmacklem ++lblkno; 1339191783Srmacklem } 1340191783Srmacklem BO_UNLOCK(bo); 1341191783Srmacklem } 1342224086Szack NFSEXITCODE(error); 1343191783Srmacklem return (error); 1344191783Srmacklem} 1345191783Srmacklem 1346191783Srmacklem/* 1347191783Srmacklem * Statfs vnode op. 1348191783Srmacklem */ 1349191783Srmacklemint 1350191990Sattilionfsvno_statfs(struct vnode *vp, struct statfs *sf) 1351191783Srmacklem{ 1352221517Srmacklem int error; 1353191783Srmacklem 1354221517Srmacklem error = VFS_STATFS(vp->v_mount, sf); 1355221517Srmacklem if (error == 0) { 1356221517Srmacklem /* 1357221517Srmacklem * Since NFS handles these values as unsigned on the 1358221517Srmacklem * wire, there is no way to represent negative values, 1359221517Srmacklem * so set them to 0. Without this, they will appear 1360221517Srmacklem * to be very large positive values for clients like 1361221517Srmacklem * Solaris10. 1362221517Srmacklem */ 1363221517Srmacklem if (sf->f_bavail < 0) 1364221517Srmacklem sf->f_bavail = 0; 1365221517Srmacklem if (sf->f_ffree < 0) 1366221517Srmacklem sf->f_ffree = 0; 1367221517Srmacklem } 1368224086Szack NFSEXITCODE(error); 1369221517Srmacklem return (error); 1370191783Srmacklem} 1371191783Srmacklem 1372191783Srmacklem/* 1373191783Srmacklem * Do the vnode op stuff for Open. Similar to nfsvno_createsub(), but 1374191783Srmacklem * must handle nfsrv_opencheck() calls after any other access checks. 1375191783Srmacklem */ 1376191783Srmacklemvoid 1377191783Srmacklemnfsvno_open(struct nfsrv_descript *nd, struct nameidata *ndp, 1378191783Srmacklem nfsquad_t clientid, nfsv4stateid_t *stateidp, struct nfsstate *stp, 1379199616Srmacklem int *exclusive_flagp, struct nfsvattr *nvap, int32_t *cverf, int create, 1380191783Srmacklem NFSACL_T *aclp, nfsattrbit_t *attrbitp, struct ucred *cred, struct thread *p, 1381191783Srmacklem struct nfsexstuff *exp, struct vnode **vpp) 1382191783Srmacklem{ 1383191783Srmacklem struct vnode *vp = NULL; 1384191783Srmacklem u_quad_t tempsize; 1385191783Srmacklem struct nfsexstuff nes; 1386191783Srmacklem 1387191783Srmacklem if (ndp->ni_vp == NULL) 1388191783Srmacklem nd->nd_repstat = nfsrv_opencheck(clientid, 1389191783Srmacklem stateidp, stp, NULL, nd, p, nd->nd_repstat); 1390191783Srmacklem if (!nd->nd_repstat) { 1391191783Srmacklem if (ndp->ni_vp == NULL) { 1392191783Srmacklem vrele(ndp->ni_startdir); 1393191783Srmacklem nd->nd_repstat = VOP_CREATE(ndp->ni_dvp, 1394191783Srmacklem &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); 1395191783Srmacklem vput(ndp->ni_dvp); 1396191783Srmacklem nfsvno_relpathbuf(ndp); 1397191783Srmacklem if (!nd->nd_repstat) { 1398191783Srmacklem if (*exclusive_flagp) { 1399191783Srmacklem *exclusive_flagp = 0; 1400191783Srmacklem NFSVNO_ATTRINIT(nvap); 1401199616Srmacklem nvap->na_atime.tv_sec = cverf[0]; 1402199616Srmacklem nvap->na_atime.tv_nsec = cverf[1]; 1403191783Srmacklem nd->nd_repstat = VOP_SETATTR(ndp->ni_vp, 1404191783Srmacklem &nvap->na_vattr, cred); 1405191783Srmacklem } else { 1406191783Srmacklem nfsrv_fixattr(nd, ndp->ni_vp, nvap, 1407191783Srmacklem aclp, p, attrbitp, exp); 1408191783Srmacklem } 1409191783Srmacklem } 1410191783Srmacklem vp = ndp->ni_vp; 1411191783Srmacklem } else { 1412191783Srmacklem if (ndp->ni_startdir) 1413191783Srmacklem vrele(ndp->ni_startdir); 1414191783Srmacklem nfsvno_relpathbuf(ndp); 1415191783Srmacklem vp = ndp->ni_vp; 1416191783Srmacklem if (create == NFSV4OPEN_CREATE) { 1417191783Srmacklem if (ndp->ni_dvp == vp) 1418191783Srmacklem vrele(ndp->ni_dvp); 1419191783Srmacklem else 1420191783Srmacklem vput(ndp->ni_dvp); 1421191783Srmacklem } 1422191783Srmacklem if (NFSVNO_ISSETSIZE(nvap) && vp->v_type == VREG) { 1423191783Srmacklem if (ndp->ni_cnd.cn_flags & RDONLY) 1424191783Srmacklem NFSVNO_SETEXRDONLY(&nes); 1425191783Srmacklem else 1426191783Srmacklem NFSVNO_EXINIT(&nes); 1427191783Srmacklem nd->nd_repstat = nfsvno_accchk(vp, 1428200999Srmacklem VWRITE, cred, &nes, p, 1429200999Srmacklem NFSACCCHK_NOOVERRIDE, 1430200999Srmacklem NFSACCCHK_VPISLOCKED, NULL); 1431191783Srmacklem nd->nd_repstat = nfsrv_opencheck(clientid, 1432191783Srmacklem stateidp, stp, vp, nd, p, nd->nd_repstat); 1433191783Srmacklem if (!nd->nd_repstat) { 1434191783Srmacklem tempsize = nvap->na_size; 1435191783Srmacklem NFSVNO_ATTRINIT(nvap); 1436191783Srmacklem nvap->na_size = tempsize; 1437191783Srmacklem nd->nd_repstat = VOP_SETATTR(vp, 1438191783Srmacklem &nvap->na_vattr, cred); 1439191783Srmacklem } 1440191783Srmacklem } else if (vp->v_type == VREG) { 1441191783Srmacklem nd->nd_repstat = nfsrv_opencheck(clientid, 1442191783Srmacklem stateidp, stp, vp, nd, p, nd->nd_repstat); 1443191783Srmacklem } 1444191783Srmacklem } 1445191783Srmacklem } else { 1446191783Srmacklem if (ndp->ni_cnd.cn_flags & HASBUF) 1447191783Srmacklem nfsvno_relpathbuf(ndp); 1448191783Srmacklem if (ndp->ni_startdir && create == NFSV4OPEN_CREATE) { 1449191783Srmacklem vrele(ndp->ni_startdir); 1450191783Srmacklem if (ndp->ni_dvp == ndp->ni_vp) 1451191783Srmacklem vrele(ndp->ni_dvp); 1452191783Srmacklem else 1453191783Srmacklem vput(ndp->ni_dvp); 1454191783Srmacklem if (ndp->ni_vp) 1455191783Srmacklem vput(ndp->ni_vp); 1456191783Srmacklem } 1457191783Srmacklem } 1458191783Srmacklem *vpp = vp; 1459224086Szack 1460224086Szack NFSEXITCODE2(0, nd); 1461191783Srmacklem} 1462191783Srmacklem 1463191783Srmacklem/* 1464191783Srmacklem * Updates the file rev and sets the mtime and ctime 1465191783Srmacklem * to the current clock time, returning the va_filerev and va_Xtime 1466191783Srmacklem * values. 1467191783Srmacklem */ 1468191783Srmacklemvoid 1469191783Srmacklemnfsvno_updfilerev(struct vnode *vp, struct nfsvattr *nvap, 1470191783Srmacklem struct ucred *cred, struct thread *p) 1471191783Srmacklem{ 1472191783Srmacklem struct vattr va; 1473191783Srmacklem 1474191783Srmacklem VATTR_NULL(&va); 1475191783Srmacklem getnanotime(&va.va_mtime); 1476191783Srmacklem (void) VOP_SETATTR(vp, &va, cred); 1477216693Srmacklem (void) nfsvno_getattr(vp, nvap, cred, p, 1); 1478191783Srmacklem} 1479191783Srmacklem 1480191783Srmacklem/* 1481191783Srmacklem * Glue routine to nfsv4_fillattr(). 1482191783Srmacklem */ 1483191783Srmacklemint 1484220645Srmacklemnfsvno_fillattr(struct nfsrv_descript *nd, struct mount *mp, struct vnode *vp, 1485191783Srmacklem struct nfsvattr *nvap, fhandle_t *fhp, int rderror, nfsattrbit_t *attrbitp, 1486220648Srmacklem struct ucred *cred, struct thread *p, int isdgram, int reterr, 1487220648Srmacklem int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno) 1488191783Srmacklem{ 1489191783Srmacklem int error; 1490191783Srmacklem 1491220645Srmacklem error = nfsv4_fillattr(nd, mp, vp, NULL, &nvap->na_vattr, fhp, rderror, 1492220648Srmacklem attrbitp, cred, p, isdgram, reterr, supports_nfsv4acls, at_root, 1493220648Srmacklem mounted_on_fileno); 1494224086Szack NFSEXITCODE2(0, nd); 1495191783Srmacklem return (error); 1496191783Srmacklem} 1497191783Srmacklem 1498191783Srmacklem/* Since the Readdir vnode ops vary, put the entire functions in here. */ 1499191783Srmacklem/* 1500191783Srmacklem * nfs readdir service 1501191783Srmacklem * - mallocs what it thinks is enough to read 1502191783Srmacklem * count rounded up to a multiple of DIRBLKSIZ <= NFS_MAXREADDIR 1503206170Srmacklem * - calls VOP_READDIR() 1504191783Srmacklem * - loops around building the reply 1505191783Srmacklem * if the output generated exceeds count break out of loop 1506191783Srmacklem * The NFSM_CLGET macro is used here so that the reply will be packed 1507191783Srmacklem * tightly in mbuf clusters. 1508191783Srmacklem * - it trims out records with d_fileno == 0 1509191783Srmacklem * this doesn't matter for Unix clients, but they might confuse clients 1510191783Srmacklem * for other os'. 1511191783Srmacklem * - it trims out records with d_type == DT_WHT 1512191783Srmacklem * these cannot be seen through NFS (unless we extend the protocol) 1513191783Srmacklem * The alternate call nfsrvd_readdirplus() does lookups as well. 1514191783Srmacklem * PS: The NFS protocol spec. does not clarify what the "count" byte 1515191783Srmacklem * argument is a count of.. just name strings and file id's or the 1516191783Srmacklem * entire reply rpc or ... 1517191783Srmacklem * I tried just file name and id sizes and it confused the Sun client, 1518191783Srmacklem * so I am using the full rpc size now. The "paranoia.." comment refers 1519191783Srmacklem * to including the status longwords that are not a part of the dir. 1520191783Srmacklem * "entry" structures, but are in the rpc. 1521191783Srmacklem */ 1522191783Srmacklemint 1523191783Srmacklemnfsrvd_readdir(struct nfsrv_descript *nd, int isdgram, 1524191783Srmacklem struct vnode *vp, struct thread *p, struct nfsexstuff *exp) 1525191783Srmacklem{ 1526191783Srmacklem struct dirent *dp; 1527191783Srmacklem u_int32_t *tl; 1528191783Srmacklem int dirlen; 1529191783Srmacklem char *cpos, *cend, *rbuf; 1530191783Srmacklem struct nfsvattr at; 1531191783Srmacklem int nlen, error = 0, getret = 1; 1532191783Srmacklem int siz, cnt, fullsiz, eofflag, ncookies; 1533191783Srmacklem u_int64_t off, toff, verf; 1534191783Srmacklem u_long *cookies = NULL, *cookiep; 1535191783Srmacklem struct uio io; 1536191783Srmacklem struct iovec iv; 1537216691Srmacklem int not_zfs; 1538191783Srmacklem 1539191783Srmacklem if (nd->nd_repstat) { 1540191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1541224086Szack goto out; 1542191783Srmacklem } 1543191783Srmacklem if (nd->nd_flag & ND_NFSV2) { 1544191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1545191783Srmacklem off = fxdr_unsigned(u_quad_t, *tl++); 1546191783Srmacklem } else { 1547191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED); 1548191783Srmacklem off = fxdr_hyper(tl); 1549191783Srmacklem tl += 2; 1550191783Srmacklem verf = fxdr_hyper(tl); 1551191783Srmacklem tl += 2; 1552191783Srmacklem } 1553191783Srmacklem toff = off; 1554191783Srmacklem cnt = fxdr_unsigned(int, *tl); 1555206170Srmacklem if (cnt > NFS_SRVMAXDATA(nd) || cnt < 0) 1556191783Srmacklem cnt = NFS_SRVMAXDATA(nd); 1557191783Srmacklem siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 1558191783Srmacklem fullsiz = siz; 1559191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1560191783Srmacklem nd->nd_repstat = getret = nfsvno_getattr(vp, &at, nd->nd_cred, 1561216693Srmacklem p, 1); 1562191783Srmacklem#if 0 1563191783Srmacklem /* 1564191783Srmacklem * va_filerev is not sufficient as a cookie verifier, 1565191783Srmacklem * since it is not supposed to change when entries are 1566191783Srmacklem * removed/added unless that offset cookies returned to 1567191783Srmacklem * the client are no longer valid. 1568191783Srmacklem */ 1569191783Srmacklem if (!nd->nd_repstat && toff && verf != at.na_filerev) 1570191783Srmacklem nd->nd_repstat = NFSERR_BAD_COOKIE; 1571191783Srmacklem#endif 1572191783Srmacklem } 1573206170Srmacklem if (nd->nd_repstat == 0 && cnt == 0) { 1574206170Srmacklem if (nd->nd_flag & ND_NFSV2) 1575206170Srmacklem /* NFSv2 does not have NFSERR_TOOSMALL */ 1576206170Srmacklem nd->nd_repstat = EPERM; 1577206170Srmacklem else 1578206170Srmacklem nd->nd_repstat = NFSERR_TOOSMALL; 1579206170Srmacklem } 1580191783Srmacklem if (!nd->nd_repstat) 1581200999Srmacklem nd->nd_repstat = nfsvno_accchk(vp, VEXEC, 1582191783Srmacklem nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, 1583200999Srmacklem NFSACCCHK_VPISLOCKED, NULL); 1584191783Srmacklem if (nd->nd_repstat) { 1585191783Srmacklem vput(vp); 1586191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1587191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1588224086Szack goto out; 1589191783Srmacklem } 1590216691Srmacklem not_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs"); 1591191783Srmacklem MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); 1592191783Srmacklemagain: 1593191783Srmacklem eofflag = 0; 1594191783Srmacklem if (cookies) { 1595191783Srmacklem free((caddr_t)cookies, M_TEMP); 1596191783Srmacklem cookies = NULL; 1597191783Srmacklem } 1598191783Srmacklem 1599191783Srmacklem iv.iov_base = rbuf; 1600191783Srmacklem iv.iov_len = siz; 1601191783Srmacklem io.uio_iov = &iv; 1602191783Srmacklem io.uio_iovcnt = 1; 1603191783Srmacklem io.uio_offset = (off_t)off; 1604191783Srmacklem io.uio_resid = siz; 1605191783Srmacklem io.uio_segflg = UIO_SYSSPACE; 1606191783Srmacklem io.uio_rw = UIO_READ; 1607191783Srmacklem io.uio_td = NULL; 1608191783Srmacklem nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies, 1609191783Srmacklem &cookies); 1610191783Srmacklem off = (u_int64_t)io.uio_offset; 1611191783Srmacklem if (io.uio_resid) 1612191783Srmacklem siz -= io.uio_resid; 1613191783Srmacklem 1614191783Srmacklem if (!cookies && !nd->nd_repstat) 1615191783Srmacklem nd->nd_repstat = NFSERR_PERM; 1616191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1617216693Srmacklem getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); 1618191783Srmacklem if (!nd->nd_repstat) 1619191783Srmacklem nd->nd_repstat = getret; 1620191783Srmacklem } 1621191783Srmacklem 1622191783Srmacklem /* 1623191783Srmacklem * Handles the failed cases. nd->nd_repstat == 0 past here. 1624191783Srmacklem */ 1625191783Srmacklem if (nd->nd_repstat) { 1626216692Srmacklem vput(vp); 1627191783Srmacklem free((caddr_t)rbuf, M_TEMP); 1628191783Srmacklem if (cookies) 1629191783Srmacklem free((caddr_t)cookies, M_TEMP); 1630191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1631191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1632224086Szack goto out; 1633191783Srmacklem } 1634191783Srmacklem /* 1635191783Srmacklem * If nothing read, return eof 1636191783Srmacklem * rpc reply 1637191783Srmacklem */ 1638191783Srmacklem if (siz == 0) { 1639216692Srmacklem vput(vp); 1640191783Srmacklem if (nd->nd_flag & ND_NFSV2) { 1641191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1642191783Srmacklem } else { 1643191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1644191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1645191783Srmacklem txdr_hyper(at.na_filerev, tl); 1646191783Srmacklem tl += 2; 1647191783Srmacklem } 1648191783Srmacklem *tl++ = newnfs_false; 1649191783Srmacklem *tl = newnfs_true; 1650191783Srmacklem FREE((caddr_t)rbuf, M_TEMP); 1651191783Srmacklem FREE((caddr_t)cookies, M_TEMP); 1652224086Szack goto out; 1653191783Srmacklem } 1654191783Srmacklem 1655191783Srmacklem /* 1656191783Srmacklem * Check for degenerate cases of nothing useful read. 1657191783Srmacklem * If so go try again 1658191783Srmacklem */ 1659191783Srmacklem cpos = rbuf; 1660191783Srmacklem cend = rbuf + siz; 1661191783Srmacklem dp = (struct dirent *)cpos; 1662191783Srmacklem cookiep = cookies; 1663191783Srmacklem 1664191783Srmacklem /* 1665191783Srmacklem * For some reason FreeBSD's ufs_readdir() chooses to back the 1666191783Srmacklem * directory offset up to a block boundary, so it is necessary to 1667191783Srmacklem * skip over the records that precede the requested offset. This 1668191783Srmacklem * requires the assumption that file offset cookies monotonically 1669191783Srmacklem * increase. 1670216691Srmacklem * Since the offset cookies don't monotonically increase for ZFS, 1671216691Srmacklem * this is not done when ZFS is the file system. 1672191783Srmacklem */ 1673191783Srmacklem while (cpos < cend && ncookies > 0 && 1674191783Srmacklem (dp->d_fileno == 0 || dp->d_type == DT_WHT || 1675216691Srmacklem (not_zfs != 0 && ((u_quad_t)(*cookiep)) <= toff))) { 1676191783Srmacklem cpos += dp->d_reclen; 1677191783Srmacklem dp = (struct dirent *)cpos; 1678191783Srmacklem cookiep++; 1679191783Srmacklem ncookies--; 1680191783Srmacklem } 1681191783Srmacklem if (cpos >= cend || ncookies == 0) { 1682191783Srmacklem siz = fullsiz; 1683191783Srmacklem toff = off; 1684191783Srmacklem goto again; 1685191783Srmacklem } 1686216692Srmacklem vput(vp); 1687191783Srmacklem 1688191783Srmacklem /* 1689191783Srmacklem * dirlen is the size of the reply, including all XDR and must 1690191783Srmacklem * not exceed cnt. For NFSv2, RFC1094 didn't clearly indicate 1691191783Srmacklem * if the XDR should be included in "count", but to be safe, we do. 1692191783Srmacklem * (Include the two booleans at the end of the reply in dirlen now.) 1693191783Srmacklem */ 1694191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1695191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1696191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1697191783Srmacklem txdr_hyper(at.na_filerev, tl); 1698191783Srmacklem dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED; 1699191783Srmacklem } else { 1700191783Srmacklem dirlen = 2 * NFSX_UNSIGNED; 1701191783Srmacklem } 1702191783Srmacklem 1703191783Srmacklem /* Loop through the records and build reply */ 1704191783Srmacklem while (cpos < cend && ncookies > 0) { 1705191783Srmacklem nlen = dp->d_namlen; 1706191783Srmacklem if (dp->d_fileno != 0 && dp->d_type != DT_WHT && 1707191783Srmacklem nlen <= NFS_MAXNAMLEN) { 1708191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1709191783Srmacklem dirlen += (6*NFSX_UNSIGNED + NFSM_RNDUP(nlen)); 1710191783Srmacklem else 1711191783Srmacklem dirlen += (4*NFSX_UNSIGNED + NFSM_RNDUP(nlen)); 1712191783Srmacklem if (dirlen > cnt) { 1713191783Srmacklem eofflag = 0; 1714191783Srmacklem break; 1715191783Srmacklem } 1716191783Srmacklem 1717191783Srmacklem /* 1718191783Srmacklem * Build the directory record xdr from 1719191783Srmacklem * the dirent entry. 1720191783Srmacklem */ 1721191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1722191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 1723191783Srmacklem *tl++ = newnfs_true; 1724191783Srmacklem *tl++ = 0; 1725191783Srmacklem } else { 1726191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1727191783Srmacklem *tl++ = newnfs_true; 1728191783Srmacklem } 1729191783Srmacklem *tl = txdr_unsigned(dp->d_fileno); 1730191783Srmacklem (void) nfsm_strtom(nd, dp->d_name, nlen); 1731191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1732191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1733191783Srmacklem *tl++ = 0; 1734191783Srmacklem } else 1735191783Srmacklem NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED); 1736191783Srmacklem *tl = txdr_unsigned(*cookiep); 1737191783Srmacklem } 1738191783Srmacklem cpos += dp->d_reclen; 1739191783Srmacklem dp = (struct dirent *)cpos; 1740191783Srmacklem cookiep++; 1741191783Srmacklem ncookies--; 1742191783Srmacklem } 1743191783Srmacklem if (cpos < cend) 1744191783Srmacklem eofflag = 0; 1745191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 1746191783Srmacklem *tl++ = newnfs_false; 1747191783Srmacklem if (eofflag) 1748191783Srmacklem *tl = newnfs_true; 1749191783Srmacklem else 1750191783Srmacklem *tl = newnfs_false; 1751191783Srmacklem FREE((caddr_t)rbuf, M_TEMP); 1752191783Srmacklem FREE((caddr_t)cookies, M_TEMP); 1753224086Szack 1754224086Szackout: 1755224086Szack NFSEXITCODE2(0, nd); 1756191783Srmacklem return (0); 1757191783Srmacklemnfsmout: 1758191783Srmacklem vput(vp); 1759224086Szack NFSEXITCODE2(error, nd); 1760191783Srmacklem return (error); 1761191783Srmacklem} 1762191783Srmacklem 1763191783Srmacklem/* 1764191783Srmacklem * Readdirplus for V3 and Readdir for V4. 1765191783Srmacklem */ 1766191783Srmacklemint 1767191783Srmacklemnfsrvd_readdirplus(struct nfsrv_descript *nd, int isdgram, 1768191783Srmacklem struct vnode *vp, struct thread *p, struct nfsexstuff *exp) 1769191783Srmacklem{ 1770191783Srmacklem struct dirent *dp; 1771191783Srmacklem u_int32_t *tl; 1772191783Srmacklem int dirlen; 1773191783Srmacklem char *cpos, *cend, *rbuf; 1774191783Srmacklem struct vnode *nvp; 1775191783Srmacklem fhandle_t nfh; 1776191783Srmacklem struct nfsvattr nva, at, *nvap = &nva; 1777191783Srmacklem struct mbuf *mb0, *mb1; 1778191783Srmacklem struct nfsreferral *refp; 1779199715Srmacklem int nlen, r, error = 0, getret = 1, usevget = 1; 1780191783Srmacklem int siz, cnt, fullsiz, eofflag, ncookies, entrycnt; 1781191783Srmacklem caddr_t bpos0, bpos1; 1782191783Srmacklem u_int64_t off, toff, verf; 1783191783Srmacklem u_long *cookies = NULL, *cookiep; 1784191783Srmacklem nfsattrbit_t attrbits, rderrbits, savbits; 1785191783Srmacklem struct uio io; 1786191783Srmacklem struct iovec iv; 1787199715Srmacklem struct componentname cn; 1788220648Srmacklem int at_root, needs_unbusy, not_zfs, supports_nfsv4acls; 1789220645Srmacklem struct mount *mp, *new_mp; 1790220645Srmacklem uint64_t mounted_on_fileno; 1791191783Srmacklem 1792191783Srmacklem if (nd->nd_repstat) { 1793191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1794224086Szack goto out; 1795191783Srmacklem } 1796191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 6 * NFSX_UNSIGNED); 1797191783Srmacklem off = fxdr_hyper(tl); 1798191783Srmacklem toff = off; 1799191783Srmacklem tl += 2; 1800191783Srmacklem verf = fxdr_hyper(tl); 1801191783Srmacklem tl += 2; 1802191783Srmacklem siz = fxdr_unsigned(int, *tl++); 1803191783Srmacklem cnt = fxdr_unsigned(int, *tl); 1804191783Srmacklem 1805191783Srmacklem /* 1806191783Srmacklem * Use the server's maximum data transfer size as the upper bound 1807191783Srmacklem * on reply datalen. 1808191783Srmacklem */ 1809206170Srmacklem if (cnt > NFS_SRVMAXDATA(nd) || cnt < 0) 1810191783Srmacklem cnt = NFS_SRVMAXDATA(nd); 1811191783Srmacklem 1812191783Srmacklem /* 1813191783Srmacklem * siz is a "hint" of how much directory information (name, fileid, 1814191783Srmacklem * cookie) should be in the reply. At least one client "hints" 0, 1815191783Srmacklem * so I set it to cnt for that case. I also round it up to the 1816191783Srmacklem * next multiple of DIRBLKSIZ. 1817191783Srmacklem */ 1818206170Srmacklem if (siz <= 0) 1819191783Srmacklem siz = cnt; 1820191783Srmacklem siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); 1821191783Srmacklem 1822191783Srmacklem if (nd->nd_flag & ND_NFSV4) { 1823191783Srmacklem error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL); 1824191783Srmacklem if (error) 1825191783Srmacklem goto nfsmout; 1826191783Srmacklem NFSSET_ATTRBIT(&savbits, &attrbits); 1827191783Srmacklem NFSCLRNOTFILLABLE_ATTRBIT(&attrbits); 1828191783Srmacklem NFSZERO_ATTRBIT(&rderrbits); 1829191783Srmacklem NFSSETBIT_ATTRBIT(&rderrbits, NFSATTRBIT_RDATTRERROR); 1830191783Srmacklem } else { 1831191783Srmacklem NFSZERO_ATTRBIT(&attrbits); 1832191783Srmacklem } 1833191783Srmacklem fullsiz = siz; 1834216693Srmacklem nd->nd_repstat = getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); 1835191783Srmacklem if (!nd->nd_repstat) { 1836191783Srmacklem if (off && verf != at.na_filerev) { 1837191783Srmacklem /* 1838191783Srmacklem * va_filerev is not sufficient as a cookie verifier, 1839191783Srmacklem * since it is not supposed to change when entries are 1840191783Srmacklem * removed/added unless that offset cookies returned to 1841191783Srmacklem * the client are no longer valid. 1842191783Srmacklem */ 1843191783Srmacklem#if 0 1844191783Srmacklem if (nd->nd_flag & ND_NFSV4) { 1845191783Srmacklem nd->nd_repstat = NFSERR_NOTSAME; 1846191783Srmacklem } else { 1847191783Srmacklem nd->nd_repstat = NFSERR_BAD_COOKIE; 1848191783Srmacklem } 1849191783Srmacklem#endif 1850191783Srmacklem } else if ((nd->nd_flag & ND_NFSV4) && off == 0 && verf != 0) { 1851191783Srmacklem nd->nd_repstat = NFSERR_BAD_COOKIE; 1852191783Srmacklem } 1853191783Srmacklem } 1854191783Srmacklem if (!nd->nd_repstat && vp->v_type != VDIR) 1855191783Srmacklem nd->nd_repstat = NFSERR_NOTDIR; 1856191783Srmacklem if (!nd->nd_repstat && cnt == 0) 1857191783Srmacklem nd->nd_repstat = NFSERR_TOOSMALL; 1858191783Srmacklem if (!nd->nd_repstat) 1859200999Srmacklem nd->nd_repstat = nfsvno_accchk(vp, VEXEC, 1860191783Srmacklem nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, 1861200999Srmacklem NFSACCCHK_VPISLOCKED, NULL); 1862191783Srmacklem if (nd->nd_repstat) { 1863191783Srmacklem vput(vp); 1864191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1865191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1866224086Szack goto out; 1867191783Srmacklem } 1868216691Srmacklem not_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs"); 1869191783Srmacklem 1870191783Srmacklem MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); 1871191783Srmacklemagain: 1872191783Srmacklem eofflag = 0; 1873191783Srmacklem if (cookies) { 1874191783Srmacklem free((caddr_t)cookies, M_TEMP); 1875191783Srmacklem cookies = NULL; 1876191783Srmacklem } 1877191783Srmacklem 1878191783Srmacklem iv.iov_base = rbuf; 1879191783Srmacklem iv.iov_len = siz; 1880191783Srmacklem io.uio_iov = &iv; 1881191783Srmacklem io.uio_iovcnt = 1; 1882191783Srmacklem io.uio_offset = (off_t)off; 1883191783Srmacklem io.uio_resid = siz; 1884191783Srmacklem io.uio_segflg = UIO_SYSSPACE; 1885191783Srmacklem io.uio_rw = UIO_READ; 1886191783Srmacklem io.uio_td = NULL; 1887191783Srmacklem nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies, 1888191783Srmacklem &cookies); 1889191783Srmacklem off = (u_int64_t)io.uio_offset; 1890191783Srmacklem if (io.uio_resid) 1891191783Srmacklem siz -= io.uio_resid; 1892191783Srmacklem 1893216693Srmacklem getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); 1894191783Srmacklem 1895191783Srmacklem if (!cookies && !nd->nd_repstat) 1896191783Srmacklem nd->nd_repstat = NFSERR_PERM; 1897191783Srmacklem if (!nd->nd_repstat) 1898191783Srmacklem nd->nd_repstat = getret; 1899191783Srmacklem if (nd->nd_repstat) { 1900199715Srmacklem vput(vp); 1901191783Srmacklem if (cookies) 1902191783Srmacklem free((caddr_t)cookies, M_TEMP); 1903191783Srmacklem free((caddr_t)rbuf, M_TEMP); 1904191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1905191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1906224086Szack goto out; 1907191783Srmacklem } 1908191783Srmacklem /* 1909191783Srmacklem * If nothing read, return eof 1910191783Srmacklem * rpc reply 1911191783Srmacklem */ 1912191783Srmacklem if (siz == 0) { 1913199715Srmacklem vput(vp); 1914191783Srmacklem if (nd->nd_flag & ND_NFSV3) 1915191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1916191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED); 1917191783Srmacklem txdr_hyper(at.na_filerev, tl); 1918191783Srmacklem tl += 2; 1919191783Srmacklem *tl++ = newnfs_false; 1920191783Srmacklem *tl = newnfs_true; 1921191783Srmacklem free((caddr_t)cookies, M_TEMP); 1922191783Srmacklem free((caddr_t)rbuf, M_TEMP); 1923224086Szack goto out; 1924191783Srmacklem } 1925191783Srmacklem 1926191783Srmacklem /* 1927191783Srmacklem * Check for degenerate cases of nothing useful read. 1928191783Srmacklem * If so go try again 1929191783Srmacklem */ 1930191783Srmacklem cpos = rbuf; 1931191783Srmacklem cend = rbuf + siz; 1932191783Srmacklem dp = (struct dirent *)cpos; 1933191783Srmacklem cookiep = cookies; 1934191783Srmacklem 1935191783Srmacklem /* 1936191783Srmacklem * For some reason FreeBSD's ufs_readdir() chooses to back the 1937191783Srmacklem * directory offset up to a block boundary, so it is necessary to 1938191783Srmacklem * skip over the records that precede the requested offset. This 1939191783Srmacklem * requires the assumption that file offset cookies monotonically 1940191783Srmacklem * increase. 1941216691Srmacklem * Since the offset cookies don't monotonically increase for ZFS, 1942216691Srmacklem * this is not done when ZFS is the file system. 1943191783Srmacklem */ 1944191783Srmacklem while (cpos < cend && ncookies > 0 && 1945191783Srmacklem (dp->d_fileno == 0 || dp->d_type == DT_WHT || 1946216691Srmacklem (not_zfs != 0 && ((u_quad_t)(*cookiep)) <= toff) || 1947191783Srmacklem ((nd->nd_flag & ND_NFSV4) && 1948191783Srmacklem ((dp->d_namlen == 1 && dp->d_name[0] == '.') || 1949191783Srmacklem (dp->d_namlen==2 && dp->d_name[0]=='.' && dp->d_name[1]=='.'))))) { 1950191783Srmacklem cpos += dp->d_reclen; 1951191783Srmacklem dp = (struct dirent *)cpos; 1952191783Srmacklem cookiep++; 1953191783Srmacklem ncookies--; 1954191783Srmacklem } 1955191783Srmacklem if (cpos >= cend || ncookies == 0) { 1956191783Srmacklem siz = fullsiz; 1957191783Srmacklem toff = off; 1958191783Srmacklem goto again; 1959191783Srmacklem } 1960217176Srmacklem 1961217176Srmacklem /* 1962217176Srmacklem * Busy the file system so that the mount point won't go away 1963217176Srmacklem * and, as such, VFS_VGET() can be used safely. 1964217176Srmacklem */ 1965217176Srmacklem mp = vp->v_mount; 1966217176Srmacklem vfs_ref(mp); 1967224082Szack NFSVOPUNLOCK(vp, 0); 1968217176Srmacklem nd->nd_repstat = vfs_busy(mp, 0); 1969217176Srmacklem vfs_rel(mp); 1970217176Srmacklem if (nd->nd_repstat != 0) { 1971217176Srmacklem vrele(vp); 1972217176Srmacklem free(cookies, M_TEMP); 1973217176Srmacklem free(rbuf, M_TEMP); 1974217176Srmacklem if (nd->nd_flag & ND_NFSV3) 1975217176Srmacklem nfsrv_postopattr(nd, getret, &at); 1976224086Szack goto out; 1977217176Srmacklem } 1978191783Srmacklem 1979191783Srmacklem /* 1980191783Srmacklem * Save this position, in case there is an error before one entry 1981191783Srmacklem * is created. 1982191783Srmacklem */ 1983191783Srmacklem mb0 = nd->nd_mb; 1984191783Srmacklem bpos0 = nd->nd_bpos; 1985191783Srmacklem 1986191783Srmacklem /* 1987191783Srmacklem * Fill in the first part of the reply. 1988191783Srmacklem * dirlen is the reply length in bytes and cannot exceed cnt. 1989191783Srmacklem * (Include the two booleans at the end of the reply in dirlen now, 1990191783Srmacklem * so we recognize when we have exceeded cnt.) 1991191783Srmacklem */ 1992191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 1993191783Srmacklem dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED; 1994191783Srmacklem nfsrv_postopattr(nd, getret, &at); 1995191783Srmacklem } else { 1996191783Srmacklem dirlen = NFSX_VERF + 2 * NFSX_UNSIGNED; 1997191783Srmacklem } 1998191783Srmacklem NFSM_BUILD(tl, u_int32_t *, NFSX_VERF); 1999191783Srmacklem txdr_hyper(at.na_filerev, tl); 2000191783Srmacklem 2001191783Srmacklem /* 2002191783Srmacklem * Save this position, in case there is an empty reply needed. 2003191783Srmacklem */ 2004191783Srmacklem mb1 = nd->nd_mb; 2005191783Srmacklem bpos1 = nd->nd_bpos; 2006191783Srmacklem 2007191783Srmacklem /* Loop through the records and build reply */ 2008191783Srmacklem entrycnt = 0; 2009191783Srmacklem while (cpos < cend && ncookies > 0 && dirlen < cnt) { 2010191783Srmacklem nlen = dp->d_namlen; 2011191783Srmacklem if (dp->d_fileno != 0 && dp->d_type != DT_WHT && 2012191783Srmacklem nlen <= NFS_MAXNAMLEN && 2013191783Srmacklem ((nd->nd_flag & ND_NFSV3) || nlen > 2 || 2014191783Srmacklem (nlen==2 && (dp->d_name[0]!='.' || dp->d_name[1]!='.')) 2015191783Srmacklem || (nlen == 1 && dp->d_name[0] != '.'))) { 2016191783Srmacklem /* 2017191783Srmacklem * Save the current position in the reply, in case 2018191783Srmacklem * this entry exceeds cnt. 2019191783Srmacklem */ 2020191783Srmacklem mb1 = nd->nd_mb; 2021191783Srmacklem bpos1 = nd->nd_bpos; 2022191783Srmacklem 2023191783Srmacklem /* 2024191783Srmacklem * For readdir_and_lookup get the vnode using 2025191783Srmacklem * the file number. 2026191783Srmacklem */ 2027191783Srmacklem nvp = NULL; 2028191783Srmacklem refp = NULL; 2029191783Srmacklem r = 0; 2030220645Srmacklem at_root = 0; 2031220645Srmacklem needs_unbusy = 0; 2032220645Srmacklem new_mp = mp; 2033220645Srmacklem mounted_on_fileno = (uint64_t)dp->d_fileno; 2034191783Srmacklem if ((nd->nd_flag & ND_NFSV3) || 2035191783Srmacklem NFSNONZERO_ATTRBIT(&savbits)) { 2036191783Srmacklem if (nd->nd_flag & ND_NFSV4) 2037191783Srmacklem refp = nfsv4root_getreferral(NULL, 2038191783Srmacklem vp, dp->d_fileno); 2039199715Srmacklem if (refp == NULL) { 2040199715Srmacklem if (usevget) 2041217176Srmacklem r = VFS_VGET(mp, dp->d_fileno, 2042217176Srmacklem LK_SHARED, &nvp); 2043199715Srmacklem else 2044199715Srmacklem r = EOPNOTSUPP; 2045199715Srmacklem if (r == EOPNOTSUPP) { 2046199715Srmacklem if (usevget) { 2047199715Srmacklem usevget = 0; 2048199715Srmacklem cn.cn_nameiop = LOOKUP; 2049199715Srmacklem cn.cn_lkflags = 2050216700Srmacklem LK_SHARED | 2051199715Srmacklem LK_RETRY; 2052199715Srmacklem cn.cn_cred = 2053199715Srmacklem nd->nd_cred; 2054199715Srmacklem cn.cn_thread = p; 2055199715Srmacklem } 2056199715Srmacklem cn.cn_nameptr = dp->d_name; 2057199715Srmacklem cn.cn_namelen = nlen; 2058199715Srmacklem cn.cn_flags = ISLASTCN | 2059199715Srmacklem NOFOLLOW | LOCKLEAF | 2060199715Srmacklem MPSAFE; 2061199715Srmacklem if (nlen == 2 && 2062199715Srmacklem dp->d_name[0] == '.' && 2063199715Srmacklem dp->d_name[1] == '.') 2064199715Srmacklem cn.cn_flags |= 2065199715Srmacklem ISDOTDOT; 2066224081Szack if (NFSVOPLOCK(vp, LK_SHARED) 2067216692Srmacklem != 0) { 2068216692Srmacklem nd->nd_repstat = EPERM; 2069216692Srmacklem break; 2070216692Srmacklem } 2071214149Srmacklem if ((vp->v_vflag & VV_ROOT) != 0 2072214149Srmacklem && (cn.cn_flags & ISDOTDOT) 2073214149Srmacklem != 0) { 2074214149Srmacklem vref(vp); 2075214149Srmacklem nvp = vp; 2076214149Srmacklem r = 0; 2077220507Srmacklem } else { 2078214149Srmacklem r = VOP_LOOKUP(vp, &nvp, 2079214149Srmacklem &cn); 2080220507Srmacklem if (vp != nvp) 2081224082Szack NFSVOPUNLOCK(vp, 2082220507Srmacklem 0); 2083220507Srmacklem } 2084199715Srmacklem } 2085220645Srmacklem 2086220645Srmacklem /* 2087220645Srmacklem * For NFSv4, check to see if nvp is 2088220645Srmacklem * a mount point and get the mount 2089220645Srmacklem * point vnode, as required. 2090220645Srmacklem */ 2091220645Srmacklem if (r == 0 && 2092220645Srmacklem nfsrv_enable_crossmntpt != 0 && 2093220645Srmacklem (nd->nd_flag & ND_NFSV4) != 0 && 2094220645Srmacklem nvp->v_type == VDIR && 2095220645Srmacklem nvp->v_mountedhere != NULL) { 2096220645Srmacklem new_mp = nvp->v_mountedhere; 2097220645Srmacklem r = vfs_busy(new_mp, 0); 2098220645Srmacklem vput(nvp); 2099220645Srmacklem nvp = NULL; 2100220645Srmacklem if (r == 0) { 2101220645Srmacklem r = VFS_ROOT(new_mp, 2102220645Srmacklem LK_SHARED, &nvp); 2103220645Srmacklem needs_unbusy = 1; 2104220645Srmacklem if (r == 0) 2105220645Srmacklem at_root = 1; 2106220645Srmacklem } 2107220645Srmacklem } 2108199715Srmacklem } 2109191783Srmacklem if (!r) { 2110191783Srmacklem if (refp == NULL && 2111191783Srmacklem ((nd->nd_flag & ND_NFSV3) || 2112191783Srmacklem NFSNONZERO_ATTRBIT(&attrbits))) { 2113191783Srmacklem r = nfsvno_getfh(nvp, &nfh, p); 2114191783Srmacklem if (!r) 2115191783Srmacklem r = nfsvno_getattr(nvp, nvap, 2116216693Srmacklem nd->nd_cred, p, 1); 2117191783Srmacklem } 2118191783Srmacklem } else { 2119191783Srmacklem nvp = NULL; 2120191783Srmacklem } 2121191783Srmacklem if (r) { 2122191783Srmacklem if (!NFSISSET_ATTRBIT(&attrbits, 2123191783Srmacklem NFSATTRBIT_RDATTRERROR)) { 2124191783Srmacklem if (nvp != NULL) 2125191783Srmacklem vput(nvp); 2126220645Srmacklem if (needs_unbusy != 0) 2127220645Srmacklem vfs_unbusy(new_mp); 2128191783Srmacklem nd->nd_repstat = r; 2129191783Srmacklem break; 2130191783Srmacklem } 2131191783Srmacklem } 2132191783Srmacklem } 2133191783Srmacklem 2134191783Srmacklem /* 2135191783Srmacklem * Build the directory record xdr 2136191783Srmacklem */ 2137191783Srmacklem if (nd->nd_flag & ND_NFSV3) { 2138191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 2139191783Srmacklem *tl++ = newnfs_true; 2140191783Srmacklem *tl++ = 0; 2141191783Srmacklem *tl = txdr_unsigned(dp->d_fileno); 2142191783Srmacklem dirlen += nfsm_strtom(nd, dp->d_name, nlen); 2143191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2144191783Srmacklem *tl++ = 0; 2145191783Srmacklem *tl = txdr_unsigned(*cookiep); 2146191783Srmacklem nfsrv_postopattr(nd, 0, nvap); 2147191783Srmacklem dirlen += nfsm_fhtom(nd,(u_int8_t *)&nfh,0,1); 2148191783Srmacklem dirlen += (5*NFSX_UNSIGNED+NFSX_V3POSTOPATTR); 2149191783Srmacklem if (nvp != NULL) 2150191783Srmacklem vput(nvp); 2151191783Srmacklem } else { 2152191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); 2153191783Srmacklem *tl++ = newnfs_true; 2154191783Srmacklem *tl++ = 0; 2155191783Srmacklem *tl = txdr_unsigned(*cookiep); 2156191783Srmacklem dirlen += nfsm_strtom(nd, dp->d_name, nlen); 2157220648Srmacklem if (nvp != NULL) { 2158220648Srmacklem supports_nfsv4acls = 2159220648Srmacklem nfs_supportsnfsv4acls(nvp); 2160224082Szack NFSVOPUNLOCK(nvp, 0); 2161220648Srmacklem } else 2162220648Srmacklem supports_nfsv4acls = 0; 2163191783Srmacklem if (refp != NULL) { 2164191783Srmacklem dirlen += nfsrv_putreferralattr(nd, 2165191783Srmacklem &savbits, refp, 0, 2166191783Srmacklem &nd->nd_repstat); 2167191783Srmacklem if (nd->nd_repstat) { 2168191783Srmacklem if (nvp != NULL) 2169191783Srmacklem vrele(nvp); 2170220645Srmacklem if (needs_unbusy != 0) 2171220645Srmacklem vfs_unbusy(new_mp); 2172191783Srmacklem break; 2173191783Srmacklem } 2174191783Srmacklem } else if (r) { 2175220645Srmacklem dirlen += nfsvno_fillattr(nd, new_mp, 2176220645Srmacklem nvp, nvap, &nfh, r, &rderrbits, 2177220648Srmacklem nd->nd_cred, p, isdgram, 0, 2178220648Srmacklem supports_nfsv4acls, at_root, 2179220645Srmacklem mounted_on_fileno); 2180191783Srmacklem } else { 2181220645Srmacklem dirlen += nfsvno_fillattr(nd, new_mp, 2182220645Srmacklem nvp, nvap, &nfh, r, &attrbits, 2183220648Srmacklem nd->nd_cred, p, isdgram, 0, 2184220648Srmacklem supports_nfsv4acls, at_root, 2185220645Srmacklem mounted_on_fileno); 2186191783Srmacklem } 2187191783Srmacklem if (nvp != NULL) 2188191783Srmacklem vrele(nvp); 2189191783Srmacklem dirlen += (3 * NFSX_UNSIGNED); 2190191783Srmacklem } 2191220645Srmacklem if (needs_unbusy != 0) 2192220645Srmacklem vfs_unbusy(new_mp); 2193191783Srmacklem if (dirlen <= cnt) 2194191783Srmacklem entrycnt++; 2195191783Srmacklem } 2196191783Srmacklem cpos += dp->d_reclen; 2197191783Srmacklem dp = (struct dirent *)cpos; 2198191783Srmacklem cookiep++; 2199191783Srmacklem ncookies--; 2200191783Srmacklem } 2201216692Srmacklem vrele(vp); 2202217176Srmacklem vfs_unbusy(mp); 2203191783Srmacklem 2204191783Srmacklem /* 2205191783Srmacklem * If dirlen > cnt, we must strip off the last entry. If that 2206191783Srmacklem * results in an empty reply, report NFSERR_TOOSMALL. 2207191783Srmacklem */ 2208191783Srmacklem if (dirlen > cnt || nd->nd_repstat) { 2209191783Srmacklem if (!nd->nd_repstat && entrycnt == 0) 2210191783Srmacklem nd->nd_repstat = NFSERR_TOOSMALL; 2211191783Srmacklem if (nd->nd_repstat) 2212191783Srmacklem newnfs_trimtrailing(nd, mb0, bpos0); 2213191783Srmacklem else 2214191783Srmacklem newnfs_trimtrailing(nd, mb1, bpos1); 2215191783Srmacklem eofflag = 0; 2216191783Srmacklem } else if (cpos < cend) 2217191783Srmacklem eofflag = 0; 2218191783Srmacklem if (!nd->nd_repstat) { 2219191783Srmacklem NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2220191783Srmacklem *tl++ = newnfs_false; 2221191783Srmacklem if (eofflag) 2222191783Srmacklem *tl = newnfs_true; 2223191783Srmacklem else 2224191783Srmacklem *tl = newnfs_false; 2225191783Srmacklem } 2226191783Srmacklem FREE((caddr_t)cookies, M_TEMP); 2227191783Srmacklem FREE((caddr_t)rbuf, M_TEMP); 2228224086Szack 2229224086Szackout: 2230224086Szack NFSEXITCODE2(0, nd); 2231191783Srmacklem return (0); 2232191783Srmacklemnfsmout: 2233191783Srmacklem vput(vp); 2234224086Szack NFSEXITCODE2(error, nd); 2235191783Srmacklem return (error); 2236191783Srmacklem} 2237191783Srmacklem 2238191783Srmacklem/* 2239191783Srmacklem * Get the settable attributes out of the mbuf list. 2240191783Srmacklem * (Return 0 or EBADRPC) 2241191783Srmacklem */ 2242191783Srmacklemint 2243191783Srmacklemnfsrv_sattr(struct nfsrv_descript *nd, struct nfsvattr *nvap, 2244191783Srmacklem nfsattrbit_t *attrbitp, NFSACL_T *aclp, struct thread *p) 2245191783Srmacklem{ 2246191783Srmacklem u_int32_t *tl; 2247191783Srmacklem struct nfsv2_sattr *sp; 2248191783Srmacklem struct timeval curtime; 2249191783Srmacklem int error = 0, toclient = 0; 2250191783Srmacklem 2251191783Srmacklem switch (nd->nd_flag & (ND_NFSV2 | ND_NFSV3 | ND_NFSV4)) { 2252191783Srmacklem case ND_NFSV2: 2253191783Srmacklem NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR); 2254191783Srmacklem /* 2255191783Srmacklem * Some old clients didn't fill in the high order 16bits. 2256191783Srmacklem * --> check the low order 2 bytes for 0xffff 2257191783Srmacklem */ 2258191783Srmacklem if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) 2259191783Srmacklem nvap->na_mode = nfstov_mode(sp->sa_mode); 2260191783Srmacklem if (sp->sa_uid != newnfs_xdrneg1) 2261191783Srmacklem nvap->na_uid = fxdr_unsigned(uid_t, sp->sa_uid); 2262191783Srmacklem if (sp->sa_gid != newnfs_xdrneg1) 2263191783Srmacklem nvap->na_gid = fxdr_unsigned(gid_t, sp->sa_gid); 2264191783Srmacklem if (sp->sa_size != newnfs_xdrneg1) 2265191783Srmacklem nvap->na_size = fxdr_unsigned(u_quad_t, sp->sa_size); 2266191783Srmacklem if (sp->sa_atime.nfsv2_sec != newnfs_xdrneg1) { 2267191783Srmacklem#ifdef notyet 2268191783Srmacklem fxdr_nfsv2time(&sp->sa_atime, &nvap->na_atime); 2269191783Srmacklem#else 2270191783Srmacklem nvap->na_atime.tv_sec = 2271191783Srmacklem fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); 2272191783Srmacklem nvap->na_atime.tv_nsec = 0; 2273191783Srmacklem#endif 2274191783Srmacklem } 2275191783Srmacklem if (sp->sa_mtime.nfsv2_sec != newnfs_xdrneg1) 2276191783Srmacklem fxdr_nfsv2time(&sp->sa_mtime, &nvap->na_mtime); 2277191783Srmacklem break; 2278191783Srmacklem case ND_NFSV3: 2279191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2280191783Srmacklem if (*tl == newnfs_true) { 2281191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2282191783Srmacklem nvap->na_mode = nfstov_mode(*tl); 2283191783Srmacklem } 2284191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2285191783Srmacklem if (*tl == newnfs_true) { 2286191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2287191783Srmacklem nvap->na_uid = fxdr_unsigned(uid_t, *tl); 2288191783Srmacklem } 2289191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2290191783Srmacklem if (*tl == newnfs_true) { 2291191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2292191783Srmacklem nvap->na_gid = fxdr_unsigned(gid_t, *tl); 2293191783Srmacklem } 2294191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2295191783Srmacklem if (*tl == newnfs_true) { 2296191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2297191783Srmacklem nvap->na_size = fxdr_hyper(tl); 2298191783Srmacklem } 2299191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2300191783Srmacklem switch (fxdr_unsigned(int, *tl)) { 2301191783Srmacklem case NFSV3SATTRTIME_TOCLIENT: 2302191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2303191783Srmacklem fxdr_nfsv3time(tl, &nvap->na_atime); 2304191783Srmacklem toclient = 1; 2305191783Srmacklem break; 2306191783Srmacklem case NFSV3SATTRTIME_TOSERVER: 2307191783Srmacklem NFSGETTIME(&curtime); 2308191783Srmacklem nvap->na_atime.tv_sec = curtime.tv_sec; 2309191783Srmacklem nvap->na_atime.tv_nsec = curtime.tv_usec * 1000; 2310191783Srmacklem nvap->na_vaflags |= VA_UTIMES_NULL; 2311191783Srmacklem break; 2312191783Srmacklem }; 2313191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2314191783Srmacklem switch (fxdr_unsigned(int, *tl)) { 2315191783Srmacklem case NFSV3SATTRTIME_TOCLIENT: 2316191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 2317191783Srmacklem fxdr_nfsv3time(tl, &nvap->na_mtime); 2318191783Srmacklem nvap->na_vaflags &= ~VA_UTIMES_NULL; 2319191783Srmacklem break; 2320191783Srmacklem case NFSV3SATTRTIME_TOSERVER: 2321191783Srmacklem NFSGETTIME(&curtime); 2322191783Srmacklem nvap->na_mtime.tv_sec = curtime.tv_sec; 2323191783Srmacklem nvap->na_mtime.tv_nsec = curtime.tv_usec * 1000; 2324191783Srmacklem if (!toclient) 2325191783Srmacklem nvap->na_vaflags |= VA_UTIMES_NULL; 2326191783Srmacklem break; 2327191783Srmacklem }; 2328191783Srmacklem break; 2329191783Srmacklem case ND_NFSV4: 2330191783Srmacklem error = nfsv4_sattr(nd, nvap, attrbitp, aclp, p); 2331191783Srmacklem }; 2332191783Srmacklemnfsmout: 2333224086Szack NFSEXITCODE2(error, nd); 2334191783Srmacklem return (error); 2335191783Srmacklem} 2336191783Srmacklem 2337191783Srmacklem/* 2338191783Srmacklem * Handle the setable attributes for V4. 2339191783Srmacklem * Returns NFSERR_BADXDR if it can't be parsed, 0 otherwise. 2340191783Srmacklem */ 2341191783Srmacklemint 2342191783Srmacklemnfsv4_sattr(struct nfsrv_descript *nd, struct nfsvattr *nvap, 2343191783Srmacklem nfsattrbit_t *attrbitp, NFSACL_T *aclp, struct thread *p) 2344191783Srmacklem{ 2345191783Srmacklem u_int32_t *tl; 2346191783Srmacklem int attrsum = 0; 2347191783Srmacklem int i, j; 2348191783Srmacklem int error, attrsize, bitpos, aclsize, aceerr, retnotsup = 0; 2349191783Srmacklem int toclient = 0; 2350191783Srmacklem u_char *cp, namestr[NFSV4_SMALLSTR + 1]; 2351191783Srmacklem uid_t uid; 2352191783Srmacklem gid_t gid; 2353191783Srmacklem struct timeval curtime; 2354191783Srmacklem 2355191783Srmacklem error = nfsrv_getattrbits(nd, attrbitp, NULL, &retnotsup); 2356191783Srmacklem if (error) 2357224086Szack goto nfsmout; 2358191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2359191783Srmacklem attrsize = fxdr_unsigned(int, *tl); 2360191783Srmacklem 2361191783Srmacklem /* 2362191783Srmacklem * Loop around getting the setable attributes. If an unsupported 2363191783Srmacklem * one is found, set nd_repstat == NFSERR_ATTRNOTSUPP and return. 2364191783Srmacklem */ 2365191783Srmacklem if (retnotsup) { 2366191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2367191783Srmacklem bitpos = NFSATTRBIT_MAX; 2368191783Srmacklem } else { 2369191783Srmacklem bitpos = 0; 2370191783Srmacklem } 2371191783Srmacklem for (; bitpos < NFSATTRBIT_MAX; bitpos++) { 2372191783Srmacklem if (attrsum > attrsize) { 2373191783Srmacklem error = NFSERR_BADXDR; 2374191783Srmacklem goto nfsmout; 2375191783Srmacklem } 2376191783Srmacklem if (NFSISSET_ATTRBIT(attrbitp, bitpos)) 2377191783Srmacklem switch (bitpos) { 2378191783Srmacklem case NFSATTRBIT_SIZE: 2379191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER); 2380191783Srmacklem nvap->na_size = fxdr_hyper(tl); 2381191783Srmacklem attrsum += NFSX_HYPER; 2382191783Srmacklem break; 2383191783Srmacklem case NFSATTRBIT_ACL: 2384191783Srmacklem error = nfsrv_dissectacl(nd, aclp, &aceerr, &aclsize, 2385191783Srmacklem p); 2386191783Srmacklem if (error) 2387191783Srmacklem goto nfsmout; 2388191783Srmacklem if (aceerr && !nd->nd_repstat) 2389224077Szack nd->nd_repstat = aceerr; 2390191783Srmacklem attrsum += aclsize; 2391191783Srmacklem break; 2392191783Srmacklem case NFSATTRBIT_ARCHIVE: 2393191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2394191783Srmacklem if (!nd->nd_repstat) 2395191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2396191783Srmacklem attrsum += NFSX_UNSIGNED; 2397191783Srmacklem break; 2398191783Srmacklem case NFSATTRBIT_HIDDEN: 2399191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2400191783Srmacklem if (!nd->nd_repstat) 2401191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2402191783Srmacklem attrsum += NFSX_UNSIGNED; 2403191783Srmacklem break; 2404191783Srmacklem case NFSATTRBIT_MIMETYPE: 2405191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2406191783Srmacklem i = fxdr_unsigned(int, *tl); 2407191783Srmacklem error = nfsm_advance(nd, NFSM_RNDUP(i), -1); 2408191783Srmacklem if (error) 2409191783Srmacklem goto nfsmout; 2410191783Srmacklem if (!nd->nd_repstat) 2411191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2412191783Srmacklem attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(i)); 2413191783Srmacklem break; 2414191783Srmacklem case NFSATTRBIT_MODE: 2415191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2416191783Srmacklem nvap->na_mode = nfstov_mode(*tl); 2417191783Srmacklem attrsum += NFSX_UNSIGNED; 2418191783Srmacklem break; 2419191783Srmacklem case NFSATTRBIT_OWNER: 2420191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2421191783Srmacklem j = fxdr_unsigned(int, *tl); 2422224086Szack if (j < 0) { 2423224086Szack error = NFSERR_BADXDR; 2424224086Szack goto nfsmout; 2425224086Szack } 2426191783Srmacklem if (j > NFSV4_SMALLSTR) 2427191783Srmacklem cp = malloc(j + 1, M_NFSSTRING, M_WAITOK); 2428191783Srmacklem else 2429191783Srmacklem cp = namestr; 2430191783Srmacklem error = nfsrv_mtostr(nd, cp, j); 2431191783Srmacklem if (error) { 2432191783Srmacklem if (j > NFSV4_SMALLSTR) 2433191783Srmacklem free(cp, M_NFSSTRING); 2434224086Szack goto nfsmout; 2435191783Srmacklem } 2436191783Srmacklem if (!nd->nd_repstat) { 2437191783Srmacklem nd->nd_repstat = nfsv4_strtouid(cp,j,&uid,p); 2438191783Srmacklem if (!nd->nd_repstat) 2439191783Srmacklem nvap->na_uid = uid; 2440191783Srmacklem } 2441191783Srmacklem if (j > NFSV4_SMALLSTR) 2442191783Srmacklem free(cp, M_NFSSTRING); 2443191783Srmacklem attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j)); 2444191783Srmacklem break; 2445191783Srmacklem case NFSATTRBIT_OWNERGROUP: 2446191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2447191783Srmacklem j = fxdr_unsigned(int, *tl); 2448224086Szack if (j < 0) { 2449224086Szack error = NFSERR_BADXDR; 2450224086Szack goto nfsmout; 2451224086Szack } 2452191783Srmacklem if (j > NFSV4_SMALLSTR) 2453191783Srmacklem cp = malloc(j + 1, M_NFSSTRING, M_WAITOK); 2454191783Srmacklem else 2455191783Srmacklem cp = namestr; 2456191783Srmacklem error = nfsrv_mtostr(nd, cp, j); 2457191783Srmacklem if (error) { 2458191783Srmacklem if (j > NFSV4_SMALLSTR) 2459191783Srmacklem free(cp, M_NFSSTRING); 2460224086Szack goto nfsmout; 2461191783Srmacklem } 2462191783Srmacklem if (!nd->nd_repstat) { 2463191783Srmacklem nd->nd_repstat = nfsv4_strtogid(cp,j,&gid,p); 2464191783Srmacklem if (!nd->nd_repstat) 2465191783Srmacklem nvap->na_gid = gid; 2466191783Srmacklem } 2467191783Srmacklem if (j > NFSV4_SMALLSTR) 2468191783Srmacklem free(cp, M_NFSSTRING); 2469191783Srmacklem attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j)); 2470191783Srmacklem break; 2471191783Srmacklem case NFSATTRBIT_SYSTEM: 2472191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2473191783Srmacklem if (!nd->nd_repstat) 2474191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2475191783Srmacklem attrsum += NFSX_UNSIGNED; 2476191783Srmacklem break; 2477191783Srmacklem case NFSATTRBIT_TIMEACCESSSET: 2478191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2479191783Srmacklem attrsum += NFSX_UNSIGNED; 2480191783Srmacklem if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) { 2481191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); 2482191783Srmacklem fxdr_nfsv4time(tl, &nvap->na_atime); 2483191783Srmacklem toclient = 1; 2484191783Srmacklem attrsum += NFSX_V4TIME; 2485191783Srmacklem } else { 2486191783Srmacklem NFSGETTIME(&curtime); 2487191783Srmacklem nvap->na_atime.tv_sec = curtime.tv_sec; 2488191783Srmacklem nvap->na_atime.tv_nsec = curtime.tv_usec * 1000; 2489191783Srmacklem nvap->na_vaflags |= VA_UTIMES_NULL; 2490191783Srmacklem } 2491191783Srmacklem break; 2492191783Srmacklem case NFSATTRBIT_TIMEBACKUP: 2493191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); 2494191783Srmacklem if (!nd->nd_repstat) 2495191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2496191783Srmacklem attrsum += NFSX_V4TIME; 2497191783Srmacklem break; 2498191783Srmacklem case NFSATTRBIT_TIMECREATE: 2499191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); 2500191783Srmacklem if (!nd->nd_repstat) 2501191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2502191783Srmacklem attrsum += NFSX_V4TIME; 2503191783Srmacklem break; 2504191783Srmacklem case NFSATTRBIT_TIMEMODIFYSET: 2505191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); 2506191783Srmacklem attrsum += NFSX_UNSIGNED; 2507191783Srmacklem if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) { 2508191783Srmacklem NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); 2509191783Srmacklem fxdr_nfsv4time(tl, &nvap->na_mtime); 2510191783Srmacklem nvap->na_vaflags &= ~VA_UTIMES_NULL; 2511191783Srmacklem attrsum += NFSX_V4TIME; 2512191783Srmacklem } else { 2513191783Srmacklem NFSGETTIME(&curtime); 2514191783Srmacklem nvap->na_mtime.tv_sec = curtime.tv_sec; 2515191783Srmacklem nvap->na_mtime.tv_nsec = curtime.tv_usec * 1000; 2516191783Srmacklem if (!toclient) 2517191783Srmacklem nvap->na_vaflags |= VA_UTIMES_NULL; 2518191783Srmacklem } 2519191783Srmacklem break; 2520191783Srmacklem default: 2521191783Srmacklem nd->nd_repstat = NFSERR_ATTRNOTSUPP; 2522191783Srmacklem /* 2523191783Srmacklem * set bitpos so we drop out of the loop. 2524191783Srmacklem */ 2525191783Srmacklem bitpos = NFSATTRBIT_MAX; 2526191783Srmacklem break; 2527191783Srmacklem }; 2528191783Srmacklem } 2529191783Srmacklem 2530191783Srmacklem /* 2531191783Srmacklem * some clients pad the attrlist, so we need to skip over the 2532191783Srmacklem * padding. 2533191783Srmacklem */ 2534191783Srmacklem if (attrsum > attrsize) { 2535191783Srmacklem error = NFSERR_BADXDR; 2536191783Srmacklem } else { 2537191783Srmacklem attrsize = NFSM_RNDUP(attrsize); 2538191783Srmacklem if (attrsum < attrsize) 2539191783Srmacklem error = nfsm_advance(nd, attrsize - attrsum, -1); 2540191783Srmacklem } 2541191783Srmacklemnfsmout: 2542224086Szack NFSEXITCODE2(error, nd); 2543191783Srmacklem return (error); 2544191783Srmacklem} 2545191783Srmacklem 2546191783Srmacklem/* 2547191783Srmacklem * Check/setup export credentials. 2548191783Srmacklem */ 2549191783Srmacklemint 2550191783Srmacklemnfsd_excred(struct nfsrv_descript *nd, struct nfsexstuff *exp, 2551191783Srmacklem struct ucred *credanon) 2552191783Srmacklem{ 2553191783Srmacklem int error = 0; 2554191783Srmacklem 2555191783Srmacklem /* 2556191783Srmacklem * Check/setup credentials. 2557191783Srmacklem */ 2558191783Srmacklem if (nd->nd_flag & ND_GSS) 2559192121Srmacklem exp->nes_exflag &= ~MNT_EXPORTANON; 2560191783Srmacklem 2561191783Srmacklem /* 2562192121Srmacklem * Check to see if the operation is allowed for this security flavor. 2563191783Srmacklem * RFC2623 suggests that the NFSv3 Fsinfo RPC be allowed to 2564191783Srmacklem * AUTH_NONE or AUTH_SYS for file systems requiring RPCSEC_GSS. 2565192121Srmacklem * Also, allow Secinfo, so that it can acquire the correct flavor(s). 2566191783Srmacklem */ 2567192121Srmacklem if (nfsvno_testexp(nd, exp) && 2568192121Srmacklem nd->nd_procnum != NFSV4OP_SECINFO && 2569191783Srmacklem nd->nd_procnum != NFSPROC_FSINFO) { 2570191783Srmacklem if (nd->nd_flag & ND_NFSV4) 2571191783Srmacklem error = NFSERR_WRONGSEC; 2572191783Srmacklem else 2573191783Srmacklem error = (NFSERR_AUTHERR | AUTH_TOOWEAK); 2574224086Szack goto out; 2575191783Srmacklem } 2576191783Srmacklem 2577191783Srmacklem /* 2578191783Srmacklem * Check to see if the file system is exported V4 only. 2579191783Srmacklem */ 2580224086Szack if (NFSVNO_EXV4ONLY(exp) && !(nd->nd_flag & ND_NFSV4)) { 2581224086Szack error = NFSERR_PROGNOTV4; 2582224086Szack goto out; 2583224086Szack } 2584191783Srmacklem 2585191783Srmacklem /* 2586191783Srmacklem * Now, map the user credentials. 2587191783Srmacklem * (Note that ND_AUTHNONE will only be set for an NFSv3 2588191783Srmacklem * Fsinfo RPC. If set for anything else, this code might need 2589191783Srmacklem * to change.) 2590191783Srmacklem */ 2591191783Srmacklem if (NFSVNO_EXPORTED(exp) && 2592191783Srmacklem ((!(nd->nd_flag & ND_GSS) && nd->nd_cred->cr_uid == 0) || 2593191783Srmacklem NFSVNO_EXPORTANON(exp) || 2594191783Srmacklem (nd->nd_flag & ND_AUTHNONE))) { 2595191783Srmacklem nd->nd_cred->cr_uid = credanon->cr_uid; 2596191783Srmacklem nd->nd_cred->cr_gid = credanon->cr_gid; 2597194498Sbrooks crsetgroups(nd->nd_cred, credanon->cr_ngroups, 2598194498Sbrooks credanon->cr_groups); 2599191783Srmacklem } 2600224086Szack 2601224086Szackout: 2602224086Szack NFSEXITCODE2(error, nd); 2603224086Szack return (error); 2604191783Srmacklem} 2605191783Srmacklem 2606191783Srmacklem/* 2607191783Srmacklem * Check exports. 2608191783Srmacklem */ 2609191783Srmacklemint 2610191783Srmacklemnfsvno_checkexp(struct mount *mp, struct sockaddr *nam, struct nfsexstuff *exp, 2611191783Srmacklem struct ucred **credp) 2612191783Srmacklem{ 2613192121Srmacklem int i, error, *secflavors; 2614191783Srmacklem 2615191783Srmacklem error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp, 2616192121Srmacklem &exp->nes_numsecflavor, &secflavors); 2617192121Srmacklem if (error) { 2618192121Srmacklem if (nfs_rootfhset) { 2619192121Srmacklem exp->nes_exflag = 0; 2620192121Srmacklem exp->nes_numsecflavor = 0; 2621192121Srmacklem error = 0; 2622192121Srmacklem } 2623192121Srmacklem } else { 2624192121Srmacklem /* Copy the security flavors. */ 2625192121Srmacklem for (i = 0; i < exp->nes_numsecflavor; i++) 2626192121Srmacklem exp->nes_secflavors[i] = secflavors[i]; 2627191783Srmacklem } 2628224086Szack NFSEXITCODE(error); 2629191783Srmacklem return (error); 2630191783Srmacklem} 2631191783Srmacklem 2632191783Srmacklem/* 2633191783Srmacklem * Get a vnode for a file handle and export stuff. 2634191783Srmacklem */ 2635191783Srmacklemint 2636191783Srmacklemnfsvno_fhtovp(struct mount *mp, fhandle_t *fhp, struct sockaddr *nam, 2637216700Srmacklem int lktype, struct vnode **vpp, struct nfsexstuff *exp, 2638216700Srmacklem struct ucred **credp) 2639191783Srmacklem{ 2640192121Srmacklem int i, error, *secflavors; 2641191783Srmacklem 2642191998Srmacklem *credp = NULL; 2643192121Srmacklem exp->nes_numsecflavor = 0; 2644217063Srmacklem if (VFS_NEEDSGIANT(mp)) 2645217063Srmacklem error = ESTALE; 2646217063Srmacklem else 2647222167Srmacklem error = VFS_FHTOVP(mp, &fhp->fh_fid, LK_EXCLUSIVE, vpp); 2648205663Srmacklem if (error != 0) 2649205663Srmacklem /* Make sure the server replies ESTALE to the client. */ 2650205663Srmacklem error = ESTALE; 2651191783Srmacklem if (nam && !error) { 2652191783Srmacklem error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp, 2653192121Srmacklem &exp->nes_numsecflavor, &secflavors); 2654191783Srmacklem if (error) { 2655191783Srmacklem if (nfs_rootfhset) { 2656191783Srmacklem exp->nes_exflag = 0; 2657192121Srmacklem exp->nes_numsecflavor = 0; 2658191783Srmacklem error = 0; 2659191783Srmacklem } else { 2660191783Srmacklem vput(*vpp); 2661191783Srmacklem } 2662192121Srmacklem } else { 2663192121Srmacklem /* Copy the security flavors. */ 2664192121Srmacklem for (i = 0; i < exp->nes_numsecflavor; i++) 2665192121Srmacklem exp->nes_secflavors[i] = secflavors[i]; 2666191783Srmacklem } 2667191783Srmacklem } 2668216700Srmacklem if (error == 0 && lktype == LK_SHARED) 2669216700Srmacklem /* 2670216700Srmacklem * It would be much better to pass lktype to VFS_FHTOVP(), 2671216700Srmacklem * but this will have to do until VFS_FHTOVP() has a lock 2672216700Srmacklem * type argument like VFS_VGET(). 2673216700Srmacklem */ 2674224081Szack NFSVOPLOCK(*vpp, LK_DOWNGRADE | LK_RETRY); 2675224086Szack 2676224086Szack NFSEXITCODE(error); 2677191783Srmacklem return (error); 2678191783Srmacklem} 2679191783Srmacklem 2680191783Srmacklem/* 2681191783Srmacklem * nfsd_fhtovp() - convert a fh to a vnode ptr 2682191783Srmacklem * - look up fsid in mount list (if not found ret error) 2683191783Srmacklem * - get vp and export rights by calling nfsvno_fhtovp() 2684191783Srmacklem * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2685191783Srmacklem * for AUTH_SYS 2686217063Srmacklem * - if mpp != NULL, return the mount point so that it can 2687217063Srmacklem * be used for vn_finished_write() by the caller 2688191783Srmacklem */ 2689191783Srmacklemvoid 2690216700Srmacklemnfsd_fhtovp(struct nfsrv_descript *nd, struct nfsrvfh *nfp, int lktype, 2691191783Srmacklem struct vnode **vpp, struct nfsexstuff *exp, 2692191783Srmacklem struct mount **mpp, int startwrite, struct thread *p) 2693191783Srmacklem{ 2694191783Srmacklem struct mount *mp; 2695191783Srmacklem struct ucred *credanon; 2696191783Srmacklem fhandle_t *fhp; 2697191783Srmacklem 2698191783Srmacklem fhp = (fhandle_t *)nfp->nfsrvfh_data; 2699191783Srmacklem /* 2700191783Srmacklem * Check for the special case of the nfsv4root_fh. 2701191783Srmacklem */ 2702217017Srmacklem mp = vfs_busyfs(&fhp->fh_fsid); 2703217063Srmacklem if (mpp != NULL) 2704217063Srmacklem *mpp = mp; 2705217017Srmacklem if (mp == NULL) { 2706191783Srmacklem *vpp = NULL; 2707191783Srmacklem nd->nd_repstat = ESTALE; 2708224086Szack goto out; 2709191783Srmacklem } 2710191783Srmacklem 2711191783Srmacklem if (startwrite) 2712191783Srmacklem vn_start_write(NULL, mpp, V_WAIT); 2713191783Srmacklem 2714216700Srmacklem nd->nd_repstat = nfsvno_fhtovp(mp, fhp, nd->nd_nam, lktype, vpp, exp, 2715191783Srmacklem &credanon); 2716217017Srmacklem vfs_unbusy(mp); 2717191783Srmacklem 2718191783Srmacklem /* 2719191783Srmacklem * For NFSv4 without a pseudo root fs, unexported file handles 2720191783Srmacklem * can be returned, so that Lookup works everywhere. 2721191783Srmacklem */ 2722191783Srmacklem if (!nd->nd_repstat && exp->nes_exflag == 0 && 2723191783Srmacklem !(nd->nd_flag & ND_NFSV4)) { 2724191783Srmacklem vput(*vpp); 2725191783Srmacklem nd->nd_repstat = EACCES; 2726191783Srmacklem } 2727191783Srmacklem 2728191783Srmacklem /* 2729191783Srmacklem * Personally, I've never seen any point in requiring a 2730191783Srmacklem * reserved port#, since only in the rare case where the 2731191783Srmacklem * clients are all boxes with secure system priviledges, 2732191783Srmacklem * does it provide any enhanced security, but... some people 2733191783Srmacklem * believe it to be useful and keep putting this code back in. 2734191783Srmacklem * (There is also some "security checker" out there that 2735191783Srmacklem * complains if the nfs server doesn't enforce this.) 2736191783Srmacklem * However, note the following: 2737191783Srmacklem * RFC3530 (NFSv4) specifies that a reserved port# not be 2738191783Srmacklem * required. 2739191783Srmacklem * RFC2623 recommends that, if a reserved port# is checked for, 2740191783Srmacklem * that there be a way to turn that off--> ifdef'd. 2741191783Srmacklem */ 2742191783Srmacklem#ifdef NFS_REQRSVPORT 2743191783Srmacklem if (!nd->nd_repstat) { 2744192121Srmacklem struct sockaddr_in *saddr; 2745192121Srmacklem struct sockaddr_in6 *saddr6; 2746192121Srmacklem 2747192121Srmacklem saddr = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in *); 2748192121Srmacklem saddr6 = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in6 *); 2749192121Srmacklem if (!(nd->nd_flag & ND_NFSV4) && 2750192121Srmacklem ((saddr->sin_family == AF_INET && 2751192121Srmacklem ntohs(saddr->sin_port) >= IPPORT_RESERVED) || 2752192121Srmacklem (saddr6->sin6_family == AF_INET6 && 2753192121Srmacklem ntohs(saddr6->sin6_port) >= IPPORT_RESERVED))) { 2754192121Srmacklem vput(*vpp); 2755192121Srmacklem nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); 2756192121Srmacklem } 2757191783Srmacklem } 2758191783Srmacklem#endif /* NFS_REQRSVPORT */ 2759191783Srmacklem 2760191783Srmacklem /* 2761191783Srmacklem * Check/setup credentials. 2762191783Srmacklem */ 2763191783Srmacklem if (!nd->nd_repstat) { 2764191783Srmacklem nd->nd_saveduid = nd->nd_cred->cr_uid; 2765191783Srmacklem nd->nd_repstat = nfsd_excred(nd, exp, credanon); 2766191783Srmacklem if (nd->nd_repstat) 2767191783Srmacklem vput(*vpp); 2768191783Srmacklem } 2769191940Skan if (credanon != NULL) 2770191940Skan crfree(credanon); 2771191783Srmacklem if (nd->nd_repstat) { 2772191783Srmacklem if (startwrite) 2773191783Srmacklem vn_finished_write(mp); 2774191783Srmacklem *vpp = NULL; 2775217063Srmacklem if (mpp != NULL) 2776217063Srmacklem *mpp = NULL; 2777191783Srmacklem } 2778224086Szack 2779224086Szackout: 2780224086Szack NFSEXITCODE2(0, nd); 2781191783Srmacklem} 2782191783Srmacklem 2783191783Srmacklem/* 2784191783Srmacklem * glue for fp. 2785191783Srmacklem */ 2786191783Srmacklemint 2787191783Srmacklemfp_getfvp(struct thread *p, int fd, struct file **fpp, struct vnode **vpp) 2788191783Srmacklem{ 2789191783Srmacklem struct filedesc *fdp; 2790191783Srmacklem struct file *fp; 2791224086Szack int error = 0; 2792191783Srmacklem 2793191783Srmacklem fdp = p->td_proc->p_fd; 2794191783Srmacklem if (fd >= fdp->fd_nfiles || 2795224086Szack (fp = fdp->fd_ofiles[fd]) == NULL) { 2796224086Szack error = EBADF; 2797224086Szack goto out; 2798224086Szack } 2799191783Srmacklem *fpp = fp; 2800224086Szack 2801224086Szackout: 2802224086Szack NFSEXITCODE(error); 2803224086Szack return (error); 2804191783Srmacklem} 2805191783Srmacklem 2806191783Srmacklem/* 2807192121Srmacklem * Called from nfssvc() to update the exports list. Just call 2808191783Srmacklem * vfs_export(). This has to be done, since the v4 root fake fs isn't 2809191783Srmacklem * in the mount list. 2810191783Srmacklem */ 2811191783Srmacklemint 2812191783Srmacklemnfsrv_v4rootexport(void *argp, struct ucred *cred, struct thread *p) 2813191783Srmacklem{ 2814191783Srmacklem struct nfsex_args *nfsexargp = (struct nfsex_args *)argp; 2815224086Szack int error = 0; 2816191783Srmacklem struct nameidata nd; 2817191783Srmacklem fhandle_t fh; 2818191783Srmacklem 2819191783Srmacklem error = vfs_export(&nfsv4root_mnt, &nfsexargp->export); 2820216894Srmacklem if ((nfsexargp->export.ex_flags & MNT_DELEXPORT) != 0) 2821191783Srmacklem nfs_rootfhset = 0; 2822216894Srmacklem else if (error == 0) { 2823224086Szack if (nfsexargp->fspec == NULL) { 2824224086Szack error = EPERM; 2825224086Szack goto out; 2826224086Szack } 2827191783Srmacklem /* 2828191783Srmacklem * If fspec != NULL, this is the v4root path. 2829191783Srmacklem */ 2830191783Srmacklem NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, UIO_USERSPACE, 2831191783Srmacklem nfsexargp->fspec, p); 2832191783Srmacklem if ((error = namei(&nd)) != 0) 2833224086Szack goto out; 2834191783Srmacklem error = nfsvno_getfh(nd.ni_vp, &fh, p); 2835191783Srmacklem vrele(nd.ni_vp); 2836191783Srmacklem if (!error) { 2837191783Srmacklem nfs_rootfh.nfsrvfh_len = NFSX_MYFH; 2838191783Srmacklem NFSBCOPY((caddr_t)&fh, 2839191783Srmacklem nfs_rootfh.nfsrvfh_data, 2840191783Srmacklem sizeof (fhandle_t)); 2841191783Srmacklem nfs_rootfhset = 1; 2842191783Srmacklem } 2843191783Srmacklem } 2844224086Szack 2845224086Szackout: 2846224086Szack NFSEXITCODE(error); 2847191783Srmacklem return (error); 2848191783Srmacklem} 2849191783Srmacklem 2850191783Srmacklem/* 2851191783Srmacklem * Get the tcp socket sequence numbers we need. 2852191783Srmacklem * (Maybe this should be moved to the tcp sources?) 2853191783Srmacklem */ 2854191783Srmacklemint 2855191783Srmacklemnfsrv_getsocksndseq(struct socket *so, tcp_seq *maxp, tcp_seq *unap) 2856191783Srmacklem{ 2857191783Srmacklem struct inpcb *inp; 2858191783Srmacklem struct tcpcb *tp; 2859224086Szack int error = 0; 2860191783Srmacklem 2861191783Srmacklem inp = sotoinpcb(so); 2862205010Srwatson KASSERT(inp != NULL, ("nfsrv_getsocksndseq: inp == NULL")); 2863205010Srwatson INP_RLOCK(inp); 2864205010Srwatson if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 2865205010Srwatson INP_RUNLOCK(inp); 2866224086Szack error = EPIPE; 2867224086Szack goto out; 2868191783Srmacklem } 2869191783Srmacklem tp = intotcpcb(inp); 2870205010Srwatson if (tp->t_state != TCPS_ESTABLISHED) { 2871205010Srwatson INP_RUNLOCK(inp); 2872224086Szack error = EPIPE; 2873224086Szack goto out; 2874191783Srmacklem } 2875205010Srwatson *maxp = tp->snd_max; 2876205010Srwatson *unap = tp->snd_una; 2877191783Srmacklem INP_RUNLOCK(inp); 2878224086Szack 2879224086Szackout: 2880224086Szack NFSEXITCODE(error); 2881224086Szack return (error); 2882191783Srmacklem} 2883191783Srmacklem 2884191783Srmacklem/* 2885191783Srmacklem * This function needs to test to see if the system is near its limit 2886191783Srmacklem * for memory allocation via malloc() or mget() and return True iff 2887191783Srmacklem * either of these resources are near their limit. 2888191783Srmacklem * XXX (For now, this is just a stub.) 2889191783Srmacklem */ 2890191783Srmacklemint nfsrv_testmalloclimit = 0; 2891191783Srmacklemint 2892191783Srmacklemnfsrv_mallocmget_limit(void) 2893191783Srmacklem{ 2894191783Srmacklem static int printmesg = 0; 2895191783Srmacklem static int testval = 1; 2896191783Srmacklem 2897191783Srmacklem if (nfsrv_testmalloclimit && (testval++ % 1000) == 0) { 2898191783Srmacklem if ((printmesg++ % 100) == 0) 2899191783Srmacklem printf("nfsd: malloc/mget near limit\n"); 2900191783Srmacklem return (1); 2901191783Srmacklem } 2902191783Srmacklem return (0); 2903191783Srmacklem} 2904191783Srmacklem 2905191783Srmacklem/* 2906191783Srmacklem * BSD specific initialization of a mount point. 2907191783Srmacklem */ 2908191783Srmacklemvoid 2909191783Srmacklemnfsd_mntinit(void) 2910191783Srmacklem{ 2911191783Srmacklem static int inited = 0; 2912191783Srmacklem 2913191783Srmacklem if (inited) 2914191783Srmacklem return; 2915191783Srmacklem inited = 1; 2916191783Srmacklem nfsv4root_mnt.mnt_flag = (MNT_RDONLY | MNT_EXPORTED); 2917191783Srmacklem TAILQ_INIT(&nfsv4root_mnt.mnt_nvnodelist); 2918191783Srmacklem nfsv4root_mnt.mnt_export = NULL; 2919191783Srmacklem TAILQ_INIT(&nfsv4root_opt); 2920191783Srmacklem TAILQ_INIT(&nfsv4root_newopt); 2921191783Srmacklem nfsv4root_mnt.mnt_opt = &nfsv4root_opt; 2922191783Srmacklem nfsv4root_mnt.mnt_optnew = &nfsv4root_newopt; 2923191783Srmacklem nfsv4root_mnt.mnt_nvnodelistsize = 0; 2924191783Srmacklem} 2925191783Srmacklem 2926191783Srmacklem/* 2927191783Srmacklem * Get a vnode for a file handle, without checking exports, etc. 2928191783Srmacklem */ 2929191783Srmacklemstruct vnode * 2930191783Srmacklemnfsvno_getvp(fhandle_t *fhp) 2931191783Srmacklem{ 2932191783Srmacklem struct mount *mp; 2933191783Srmacklem struct vnode *vp; 2934191783Srmacklem int error; 2935191783Srmacklem 2936217017Srmacklem mp = vfs_busyfs(&fhp->fh_fsid); 2937191783Srmacklem if (mp == NULL) 2938191783Srmacklem return (NULL); 2939222167Srmacklem error = VFS_FHTOVP(mp, &fhp->fh_fid, LK_EXCLUSIVE, &vp); 2940217017Srmacklem vfs_unbusy(mp); 2941191783Srmacklem if (error) 2942191783Srmacklem return (NULL); 2943191783Srmacklem return (vp); 2944191783Srmacklem} 2945191783Srmacklem 2946191783Srmacklem/* 2947191783Srmacklem * Do a local VOP_ADVLOCK(). 2948191783Srmacklem */ 2949191783Srmacklemint 2950191783Srmacklemnfsvno_advlock(struct vnode *vp, int ftype, u_int64_t first, 2951192256Srmacklem u_int64_t end, struct thread *td) 2952191783Srmacklem{ 2953224086Szack int error = 0; 2954191783Srmacklem struct flock fl; 2955191783Srmacklem u_int64_t tlen; 2956191783Srmacklem 2957212833Srmacklem if (nfsrv_dolocallocks == 0) 2958224086Szack goto out; 2959216893Srmacklem 2960216893Srmacklem /* Check for VI_DOOMED here, so that VOP_ADVLOCK() isn't performed. */ 2961224086Szack if ((vp->v_iflag & VI_DOOMED) != 0) { 2962224086Szack error = EPERM; 2963224086Szack goto out; 2964224086Szack } 2965216893Srmacklem 2966191783Srmacklem fl.l_whence = SEEK_SET; 2967191783Srmacklem fl.l_type = ftype; 2968191783Srmacklem fl.l_start = (off_t)first; 2969191783Srmacklem if (end == NFS64BITSSET) { 2970191783Srmacklem fl.l_len = 0; 2971191783Srmacklem } else { 2972191783Srmacklem tlen = end - first; 2973191783Srmacklem fl.l_len = (off_t)tlen; 2974191783Srmacklem } 2975191783Srmacklem /* 2976192256Srmacklem * For FreeBSD8, the l_pid and l_sysid must be set to the same 2977192256Srmacklem * values for all calls, so that all locks will be held by the 2978192256Srmacklem * nfsd server. (The nfsd server handles conflicts between the 2979192256Srmacklem * various clients.) 2980192256Srmacklem * Since an NFSv4 lockowner is a ClientID plus an array of up to 1024 2981192256Srmacklem * bytes, so it can't be put in l_sysid. 2982191783Srmacklem */ 2983192503Srmacklem if (nfsv4_sysid == 0) 2984192503Srmacklem nfsv4_sysid = nlm_acquire_next_sysid(); 2985192256Srmacklem fl.l_pid = (pid_t)0; 2986192503Srmacklem fl.l_sysid = (int)nfsv4_sysid; 2987192256Srmacklem 2988224080Szack NFSVOPUNLOCK(vp, 0); 2989212833Srmacklem if (ftype == F_UNLCK) 2990212833Srmacklem error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_UNLCK, &fl, 2991212833Srmacklem (F_POSIX | F_REMOTE)); 2992212833Srmacklem else 2993212833Srmacklem error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_SETLK, &fl, 2994212833Srmacklem (F_POSIX | F_REMOTE)); 2995224080Szack NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY); 2996224086Szack 2997224086Szackout: 2998224086Szack NFSEXITCODE(error); 2999191783Srmacklem return (error); 3000191783Srmacklem} 3001191783Srmacklem 3002191783Srmacklem/* 3003191783Srmacklem * Check the nfsv4 root exports. 3004191783Srmacklem */ 3005191783Srmacklemint 3006191783Srmacklemnfsvno_v4rootexport(struct nfsrv_descript *nd) 3007191783Srmacklem{ 3008191783Srmacklem struct ucred *credanon; 3009224086Szack int exflags, error = 0, numsecflavor, *secflavors, i; 3010191783Srmacklem 3011191783Srmacklem error = vfs_stdcheckexp(&nfsv4root_mnt, nd->nd_nam, &exflags, 3012192121Srmacklem &credanon, &numsecflavor, &secflavors); 3013224086Szack if (error) { 3014224086Szack error = NFSERR_PROGUNAVAIL; 3015224086Szack goto out; 3016224086Szack } 3017191940Skan if (credanon != NULL) 3018191940Skan crfree(credanon); 3019192121Srmacklem for (i = 0; i < numsecflavor; i++) { 3020192121Srmacklem if (secflavors[i] == AUTH_SYS) 3021192121Srmacklem nd->nd_flag |= ND_EXAUTHSYS; 3022192121Srmacklem else if (secflavors[i] == RPCSEC_GSS_KRB5) 3023192121Srmacklem nd->nd_flag |= ND_EXGSS; 3024192121Srmacklem else if (secflavors[i] == RPCSEC_GSS_KRB5I) 3025192121Srmacklem nd->nd_flag |= ND_EXGSSINTEGRITY; 3026192121Srmacklem else if (secflavors[i] == RPCSEC_GSS_KRB5P) 3027192121Srmacklem nd->nd_flag |= ND_EXGSSPRIVACY; 3028192121Srmacklem } 3029224086Szack 3030224086Szackout: 3031224086Szack NFSEXITCODE(error); 3032224086Szack return (error); 3033191783Srmacklem} 3034191783Srmacklem 3035191783Srmacklem/* 3036191783Srmacklem * Nfs server psuedo system call for the nfsd's 3037191783Srmacklem */ 3038191783Srmacklem/* 3039191783Srmacklem * MPSAFE 3040191783Srmacklem */ 3041191783Srmacklemstatic int 3042191783Srmacklemnfssvc_nfsd(struct thread *td, struct nfssvc_args *uap) 3043191783Srmacklem{ 3044191783Srmacklem struct file *fp; 3045192017Srmacklem struct nfsd_addsock_args sockarg; 3046192017Srmacklem struct nfsd_nfsd_args nfsdarg; 3047191783Srmacklem int error; 3048191783Srmacklem 3049191783Srmacklem if (uap->flag & NFSSVC_NFSDADDSOCK) { 3050192017Srmacklem error = copyin(uap->argp, (caddr_t)&sockarg, sizeof (sockarg)); 3051191783Srmacklem if (error) 3052224086Szack goto out; 3053224778Srwatson /* 3054224778Srwatson * Since we don't know what rights might be required, 3055224778Srwatson * pretend that we need them all. It is better to be too 3056224778Srwatson * careful than too reckless. 3057224778Srwatson */ 3058224778Srwatson if ((error = fget(td, sockarg.sock, CAP_SOCK_ALL, &fp)) != 0) 3059224086Szack goto out; 3060191783Srmacklem if (fp->f_type != DTYPE_SOCKET) { 3061191783Srmacklem fdrop(fp, td); 3062224086Szack error = EPERM; 3063224086Szack goto out; 3064191783Srmacklem } 3065191783Srmacklem error = nfsrvd_addsock(fp); 3066191783Srmacklem fdrop(fp, td); 3067191783Srmacklem } else if (uap->flag & NFSSVC_NFSDNFSD) { 3068224086Szack if (uap->argp == NULL) { 3069224086Szack error = EINVAL; 3070224086Szack goto out; 3071224086Szack } 3072192017Srmacklem error = copyin(uap->argp, (caddr_t)&nfsdarg, 3073192017Srmacklem sizeof (nfsdarg)); 3074192017Srmacklem if (error) 3075224086Szack goto out; 3076192017Srmacklem error = nfsrvd_nfsd(td, &nfsdarg); 3077191783Srmacklem } else { 3078191783Srmacklem error = nfssvc_srvcall(td, uap, td->td_ucred); 3079191783Srmacklem } 3080224086Szack 3081224086Szackout: 3082224086Szack NFSEXITCODE(error); 3083191783Srmacklem return (error); 3084191783Srmacklem} 3085191783Srmacklem 3086191783Srmacklemstatic int 3087191783Srmacklemnfssvc_srvcall(struct thread *p, struct nfssvc_args *uap, struct ucred *cred) 3088191783Srmacklem{ 3089191783Srmacklem struct nfsex_args export; 3090191783Srmacklem struct file *fp = NULL; 3091191783Srmacklem int stablefd, len; 3092191783Srmacklem struct nfsd_clid adminrevoke; 3093191783Srmacklem struct nfsd_dumplist dumplist; 3094191783Srmacklem struct nfsd_dumpclients *dumpclients; 3095191783Srmacklem struct nfsd_dumplocklist dumplocklist; 3096191783Srmacklem struct nfsd_dumplocks *dumplocks; 3097191783Srmacklem struct nameidata nd; 3098191783Srmacklem vnode_t vp; 3099191783Srmacklem int error = EINVAL; 3100217432Srmacklem struct proc *procp; 3101191783Srmacklem 3102191783Srmacklem if (uap->flag & NFSSVC_PUBLICFH) { 3103191783Srmacklem NFSBZERO((caddr_t)&nfs_pubfh.nfsrvfh_data, 3104191783Srmacklem sizeof (fhandle_t)); 3105191783Srmacklem error = copyin(uap->argp, 3106191783Srmacklem &nfs_pubfh.nfsrvfh_data, sizeof (fhandle_t)); 3107191783Srmacklem if (!error) 3108191783Srmacklem nfs_pubfhset = 1; 3109191783Srmacklem } else if (uap->flag & NFSSVC_V4ROOTEXPORT) { 3110191783Srmacklem error = copyin(uap->argp,(caddr_t)&export, 3111191783Srmacklem sizeof (struct nfsex_args)); 3112191783Srmacklem if (!error) 3113191783Srmacklem error = nfsrv_v4rootexport(&export, cred, p); 3114191783Srmacklem } else if (uap->flag & NFSSVC_NOPUBLICFH) { 3115191783Srmacklem nfs_pubfhset = 0; 3116191783Srmacklem error = 0; 3117191783Srmacklem } else if (uap->flag & NFSSVC_STABLERESTART) { 3118191783Srmacklem error = copyin(uap->argp, (caddr_t)&stablefd, 3119191783Srmacklem sizeof (int)); 3120191783Srmacklem if (!error) 3121191783Srmacklem error = fp_getfvp(p, stablefd, &fp, &vp); 3122191783Srmacklem if (!error && (NFSFPFLAG(fp) & (FREAD | FWRITE)) != (FREAD | FWRITE)) 3123191783Srmacklem error = EBADF; 3124191783Srmacklem if (!error && newnfs_numnfsd != 0) 3125191783Srmacklem error = EPERM; 3126191783Srmacklem if (!error) { 3127191783Srmacklem nfsrv_stablefirst.nsf_fp = fp; 3128191783Srmacklem nfsrv_setupstable(p); 3129191783Srmacklem } 3130191783Srmacklem } else if (uap->flag & NFSSVC_ADMINREVOKE) { 3131191783Srmacklem error = copyin(uap->argp, (caddr_t)&adminrevoke, 3132191783Srmacklem sizeof (struct nfsd_clid)); 3133191783Srmacklem if (!error) 3134191783Srmacklem error = nfsrv_adminrevoke(&adminrevoke, p); 3135191783Srmacklem } else if (uap->flag & NFSSVC_DUMPCLIENTS) { 3136191783Srmacklem error = copyin(uap->argp, (caddr_t)&dumplist, 3137191783Srmacklem sizeof (struct nfsd_dumplist)); 3138191783Srmacklem if (!error && (dumplist.ndl_size < 1 || 3139191783Srmacklem dumplist.ndl_size > NFSRV_MAXDUMPLIST)) 3140191783Srmacklem error = EPERM; 3141191783Srmacklem if (!error) { 3142191783Srmacklem len = sizeof (struct nfsd_dumpclients) * dumplist.ndl_size; 3143191783Srmacklem dumpclients = (struct nfsd_dumpclients *)malloc(len, 3144191783Srmacklem M_TEMP, M_WAITOK); 3145191783Srmacklem nfsrv_dumpclients(dumpclients, dumplist.ndl_size); 3146191783Srmacklem error = copyout(dumpclients, 3147191783Srmacklem CAST_USER_ADDR_T(dumplist.ndl_list), len); 3148191783Srmacklem free((caddr_t)dumpclients, M_TEMP); 3149191783Srmacklem } 3150191783Srmacklem } else if (uap->flag & NFSSVC_DUMPLOCKS) { 3151191783Srmacklem error = copyin(uap->argp, (caddr_t)&dumplocklist, 3152191783Srmacklem sizeof (struct nfsd_dumplocklist)); 3153191783Srmacklem if (!error && (dumplocklist.ndllck_size < 1 || 3154191783Srmacklem dumplocklist.ndllck_size > NFSRV_MAXDUMPLIST)) 3155191783Srmacklem error = EPERM; 3156191783Srmacklem if (!error) 3157191783Srmacklem error = nfsrv_lookupfilename(&nd, 3158191783Srmacklem dumplocklist.ndllck_fname, p); 3159191783Srmacklem if (!error) { 3160191783Srmacklem len = sizeof (struct nfsd_dumplocks) * 3161191783Srmacklem dumplocklist.ndllck_size; 3162191783Srmacklem dumplocks = (struct nfsd_dumplocks *)malloc(len, 3163191783Srmacklem M_TEMP, M_WAITOK); 3164191783Srmacklem nfsrv_dumplocks(nd.ni_vp, dumplocks, 3165191783Srmacklem dumplocklist.ndllck_size, p); 3166191783Srmacklem vput(nd.ni_vp); 3167191783Srmacklem error = copyout(dumplocks, 3168191783Srmacklem CAST_USER_ADDR_T(dumplocklist.ndllck_list), len); 3169191783Srmacklem free((caddr_t)dumplocks, M_TEMP); 3170191783Srmacklem } 3171217432Srmacklem } else if (uap->flag & NFSSVC_BACKUPSTABLE) { 3172217432Srmacklem procp = p->td_proc; 3173217432Srmacklem PROC_LOCK(procp); 3174217432Srmacklem nfsd_master_pid = procp->p_pid; 3175217432Srmacklem bcopy(procp->p_comm, nfsd_master_comm, MAXCOMLEN + 1); 3176217432Srmacklem nfsd_master_start = procp->p_stats->p_start; 3177217432Srmacklem nfsd_master_proc = procp; 3178217432Srmacklem PROC_UNLOCK(procp); 3179191783Srmacklem } 3180224086Szack 3181224086Szack NFSEXITCODE(error); 3182191783Srmacklem return (error); 3183191783Srmacklem} 3184191783Srmacklem 3185192121Srmacklem/* 3186192121Srmacklem * Check exports. 3187192121Srmacklem * Returns 0 if ok, 1 otherwise. 3188192121Srmacklem */ 3189192121Srmacklemint 3190192121Srmacklemnfsvno_testexp(struct nfsrv_descript *nd, struct nfsexstuff *exp) 3191192121Srmacklem{ 3192192121Srmacklem int i; 3193192121Srmacklem 3194192121Srmacklem /* 3195192121Srmacklem * This seems odd, but allow the case where the security flavor 3196192121Srmacklem * list is empty. This happens when NFSv4 is traversing non-exported 3197192121Srmacklem * file systems. Exported file systems should always have a non-empty 3198192121Srmacklem * security flavor list. 3199192121Srmacklem */ 3200192121Srmacklem if (exp->nes_numsecflavor == 0) 3201192121Srmacklem return (0); 3202192121Srmacklem 3203192121Srmacklem for (i = 0; i < exp->nes_numsecflavor; i++) { 3204192121Srmacklem /* 3205192121Srmacklem * The tests for privacy and integrity must be first, 3206192121Srmacklem * since ND_GSS is set for everything but AUTH_SYS. 3207192121Srmacklem */ 3208192121Srmacklem if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5P && 3209192121Srmacklem (nd->nd_flag & ND_GSSPRIVACY)) 3210192121Srmacklem return (0); 3211192121Srmacklem if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5I && 3212192121Srmacklem (nd->nd_flag & ND_GSSINTEGRITY)) 3213192121Srmacklem return (0); 3214192121Srmacklem if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5 && 3215192121Srmacklem (nd->nd_flag & ND_GSS)) 3216192121Srmacklem return (0); 3217192121Srmacklem if (exp->nes_secflavors[i] == AUTH_SYS && 3218192121Srmacklem (nd->nd_flag & ND_GSS) == 0) 3219192121Srmacklem return (0); 3220192121Srmacklem } 3221192121Srmacklem return (1); 3222192121Srmacklem} 3223192121Srmacklem 3224214224Srmacklem/* 3225214224Srmacklem * Calculate a hash value for the fid in a file handle. 3226214224Srmacklem */ 3227214255Srmacklemuint32_t 3228214224Srmacklemnfsrv_hashfh(fhandle_t *fhp) 3229214224Srmacklem{ 3230214255Srmacklem uint32_t hashval; 3231214224Srmacklem 3232214255Srmacklem hashval = hash32_buf(&fhp->fh_fid, sizeof(struct fid), 0); 3233214224Srmacklem return (hashval); 3234214224Srmacklem} 3235214224Srmacklem 3236217432Srmacklem/* 3237217432Srmacklem * Signal the userland master nfsd to backup the stable restart file. 3238217432Srmacklem */ 3239217432Srmacklemvoid 3240217432Srmacklemnfsrv_backupstable(void) 3241217432Srmacklem{ 3242217432Srmacklem struct proc *procp; 3243217432Srmacklem 3244217432Srmacklem if (nfsd_master_proc != NULL) { 3245217432Srmacklem procp = pfind(nfsd_master_pid); 3246217432Srmacklem /* Try to make sure it is the correct process. */ 3247217432Srmacklem if (procp == nfsd_master_proc && 3248217432Srmacklem procp->p_stats->p_start.tv_sec == 3249217432Srmacklem nfsd_master_start.tv_sec && 3250217432Srmacklem procp->p_stats->p_start.tv_usec == 3251217432Srmacklem nfsd_master_start.tv_usec && 3252217432Srmacklem strcmp(procp->p_comm, nfsd_master_comm) == 0) 3253225617Skmacy kern_psignal(procp, SIGUSR2); 3254217432Srmacklem else 3255217432Srmacklem nfsd_master_proc = NULL; 3256217432Srmacklem 3257217432Srmacklem if (procp != NULL) 3258217432Srmacklem PROC_UNLOCK(procp); 3259217432Srmacklem } 3260217432Srmacklem} 3261217432Srmacklem 3262191783Srmacklemextern int (*nfsd_call_nfsd)(struct thread *, struct nfssvc_args *); 3263191783Srmacklem 3264191783Srmacklem/* 3265191783Srmacklem * Called once to initialize data structures... 3266191783Srmacklem */ 3267191783Srmacklemstatic int 3268191783Srmacklemnfsd_modevent(module_t mod, int type, void *data) 3269191783Srmacklem{ 3270191783Srmacklem int error = 0; 3271191783Srmacklem static int loaded = 0; 3272191783Srmacklem 3273191783Srmacklem switch (type) { 3274191783Srmacklem case MOD_LOAD: 3275191783Srmacklem if (loaded) 3276224086Szack goto out; 3277191783Srmacklem newnfs_portinit(); 3278191783Srmacklem mtx_init(&nfs_cache_mutex, "nfs_cache_mutex", NULL, MTX_DEF); 3279191783Srmacklem mtx_init(&nfs_v4root_mutex, "nfs_v4root_mutex", NULL, MTX_DEF); 3280191783Srmacklem mtx_init(&nfsv4root_mnt.mnt_mtx, "struct mount mtx", NULL, 3281191783Srmacklem MTX_DEF); 3282191783Srmacklem lockinit(&nfsv4root_mnt.mnt_explock, PVFS, "explock", 0, 0); 3283191783Srmacklem nfsrvd_initcache(); 3284191783Srmacklem nfsd_init(); 3285191783Srmacklem NFSD_LOCK(); 3286191783Srmacklem nfsrvd_init(0); 3287191783Srmacklem NFSD_UNLOCK(); 3288191783Srmacklem nfsd_mntinit(); 3289191783Srmacklem#ifdef VV_DISABLEDELEG 3290191783Srmacklem vn_deleg_ops.vndeleg_recall = nfsd_recalldelegation; 3291191783Srmacklem vn_deleg_ops.vndeleg_disable = nfsd_disabledelegation; 3292191783Srmacklem#endif 3293191783Srmacklem nfsd_call_servertimer = nfsrv_servertimer; 3294191783Srmacklem nfsd_call_nfsd = nfssvc_nfsd; 3295191783Srmacklem loaded = 1; 3296191783Srmacklem break; 3297191783Srmacklem 3298191783Srmacklem case MOD_UNLOAD: 3299191783Srmacklem if (newnfs_numnfsd != 0) { 3300191783Srmacklem error = EBUSY; 3301191783Srmacklem break; 3302191783Srmacklem } 3303191783Srmacklem 3304191783Srmacklem#ifdef VV_DISABLEDELEG 3305191783Srmacklem vn_deleg_ops.vndeleg_recall = NULL; 3306191783Srmacklem vn_deleg_ops.vndeleg_disable = NULL; 3307191783Srmacklem#endif 3308191783Srmacklem nfsd_call_servertimer = NULL; 3309191783Srmacklem nfsd_call_nfsd = NULL; 3310217335Szack 3311220530Srmacklem /* Clean out all NFSv4 state. */ 3312220530Srmacklem nfsrv_throwawayallstate(curthread); 3313220530Srmacklem 3314217335Szack /* Clean the NFS server reply cache */ 3315217335Szack nfsrvd_cleancache(); 3316217335Szack 3317220530Srmacklem /* Free up the krpc server pool. */ 3318220530Srmacklem if (nfsrvd_pool != NULL) 3319220530Srmacklem svcpool_destroy(nfsrvd_pool); 3320220530Srmacklem 3321191783Srmacklem /* and get rid of the locks */ 3322191783Srmacklem mtx_destroy(&nfs_cache_mutex); 3323191783Srmacklem mtx_destroy(&nfs_v4root_mutex); 3324191783Srmacklem mtx_destroy(&nfsv4root_mnt.mnt_mtx); 3325191783Srmacklem lockdestroy(&nfsv4root_mnt.mnt_explock); 3326191783Srmacklem loaded = 0; 3327191783Srmacklem break; 3328191783Srmacklem default: 3329191783Srmacklem error = EOPNOTSUPP; 3330191783Srmacklem break; 3331191783Srmacklem } 3332224086Szack 3333224086Szackout: 3334224086Szack NFSEXITCODE(error); 3335224086Szack return (error); 3336191783Srmacklem} 3337191783Srmacklemstatic moduledata_t nfsd_mod = { 3338191783Srmacklem "nfsd", 3339191783Srmacklem nfsd_modevent, 3340191783Srmacklem NULL, 3341191783Srmacklem}; 3342191783SrmacklemDECLARE_MODULE(nfsd, nfsd_mod, SI_SUB_VFS, SI_ORDER_ANY); 3343191783Srmacklem 3344191783Srmacklem/* So that loader and kldload(2) can find us, wherever we are.. */ 3345191783SrmacklemMODULE_VERSION(nfsd, 1); 3346191783SrmacklemMODULE_DEPEND(nfsd, nfscommon, 1, 1, 1); 3347216931SrmacklemMODULE_DEPEND(nfsd, nfslock, 1, 1, 1); 3348192574SrmacklemMODULE_DEPEND(nfsd, nfslockd, 1, 1, 1); 3349209191SrmacklemMODULE_DEPEND(nfsd, krpc, 1, 1, 1); 3350209191SrmacklemMODULE_DEPEND(nfsd, nfssvc, 1, 1, 1); 3351191783Srmacklem 3352