Index: head/sys/fs/nfsserver/nfs_nfsdport.c =================================================================== --- head/sys/fs/nfsserver/nfs_nfsdport.c (revision 264887) +++ head/sys/fs/nfsserver/nfs_nfsdport.c (revision 264888) @@ -1,3393 +1,3394 @@ /*- * Copyright (c) 1989, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Rick Macklem at The University of Guelph. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * */ #include __FBSDID("$FreeBSD$"); #include /* * Functions that perform the vfs operations required by the routines in * nfsd_serv.c. It is hoped that this change will make the server more * portable. */ #include #include #include #include #include FEATURE(nfsd, "NFSv4 server"); extern u_int32_t newnfs_true, newnfs_false, newnfs_xdrneg1; extern int nfsrv_useacl; extern int newnfs_numnfsd; extern struct mount nfsv4root_mnt; extern struct nfsrv_stablefirst nfsrv_stablefirst; extern void (*nfsd_call_servertimer)(void); extern SVCPOOL *nfsrvd_pool; extern struct nfsv4lock nfsd_suspend_lock; struct vfsoptlist nfsv4root_opt, nfsv4root_newopt; NFSDLOCKMUTEX; struct nfsrchash_bucket nfsrchash_table[NFSRVCACHE_HASHSIZE]; struct nfsrchash_bucket nfsrcahash_table[NFSRVCACHE_HASHSIZE]; struct mtx nfsrc_udpmtx; struct mtx nfs_v4root_mutex; struct nfsrvfh nfs_rootfh, nfs_pubfh; int nfs_pubfhset = 0, nfs_rootfhset = 0; struct proc *nfsd_master_proc = NULL; static pid_t nfsd_master_pid = (pid_t)-1; static char nfsd_master_comm[MAXCOMLEN + 1]; static struct timeval nfsd_master_start; static uint32_t nfsv4_sysid = 0; static int nfssvc_srvcall(struct thread *, struct nfssvc_args *, struct ucred *); int nfsrv_enable_crossmntpt = 1; static int nfs_commit_blks; static int nfs_commit_miss; extern int nfsrv_issuedelegs; extern int nfsrv_dolocallocks; SYSCTL_NODE(_vfs, OID_AUTO, nfsd, CTLFLAG_RW, 0, "New NFS server"); SYSCTL_INT(_vfs_nfsd, OID_AUTO, mirrormnt, CTLFLAG_RW, &nfsrv_enable_crossmntpt, 0, "Enable nfsd to cross mount points"); SYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_blks, CTLFLAG_RW, &nfs_commit_blks, 0, ""); SYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_miss, CTLFLAG_RW, &nfs_commit_miss, 0, ""); SYSCTL_INT(_vfs_nfsd, OID_AUTO, issue_delegations, CTLFLAG_RW, &nfsrv_issuedelegs, 0, "Enable nfsd to issue delegations"); SYSCTL_INT(_vfs_nfsd, OID_AUTO, enable_locallocks, CTLFLAG_RW, &nfsrv_dolocallocks, 0, "Enable nfsd to acquire local locks on files"); #define MAX_REORDERED_RPC 16 #define NUM_HEURISTIC 1031 #define NHUSE_INIT 64 #define NHUSE_INC 16 #define NHUSE_MAX 2048 static struct nfsheur { struct vnode *nh_vp; /* vp to match (unreferenced pointer) */ off_t nh_nextoff; /* next offset for sequential detection */ int nh_use; /* use count for selection */ int nh_seqcount; /* heuristic */ } nfsheur[NUM_HEURISTIC]; /* * Heuristic to detect sequential operation. */ static struct nfsheur * nfsrv_sequential_heuristic(struct uio *uio, struct vnode *vp) { struct nfsheur *nh; int hi, try; /* Locate best candidate. */ try = 32; hi = ((int)(vm_offset_t)vp / sizeof(struct vnode)) % NUM_HEURISTIC; nh = &nfsheur[hi]; while (try--) { if (nfsheur[hi].nh_vp == vp) { nh = &nfsheur[hi]; break; } if (nfsheur[hi].nh_use > 0) --nfsheur[hi].nh_use; hi = (hi + 1) % NUM_HEURISTIC; if (nfsheur[hi].nh_use < nh->nh_use) nh = &nfsheur[hi]; } /* Initialize hint if this is a new file. */ if (nh->nh_vp != vp) { nh->nh_vp = vp; nh->nh_nextoff = uio->uio_offset; nh->nh_use = NHUSE_INIT; if (uio->uio_offset == 0) nh->nh_seqcount = 4; else nh->nh_seqcount = 1; } /* Calculate heuristic. */ if ((uio->uio_offset == 0 && nh->nh_seqcount > 0) || uio->uio_offset == nh->nh_nextoff) { /* See comments in vfs_vnops.c:sequential_heuristic(). */ nh->nh_seqcount += howmany(uio->uio_resid, 16384); if (nh->nh_seqcount > IO_SEQMAX) nh->nh_seqcount = IO_SEQMAX; } else if (qabs(uio->uio_offset - nh->nh_nextoff) <= MAX_REORDERED_RPC * imax(vp->v_mount->mnt_stat.f_iosize, uio->uio_resid)) { /* Probably a reordered RPC, leave seqcount alone. */ } else if (nh->nh_seqcount > 1) { nh->nh_seqcount /= 2; } else { nh->nh_seqcount = 0; } nh->nh_use += NHUSE_INC; if (nh->nh_use > NHUSE_MAX) nh->nh_use = NHUSE_MAX; return (nh); } /* * Get attributes into nfsvattr structure. */ int nfsvno_getattr(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred, struct thread *p, int vpislocked) { int error, lockedit = 0; if (vpislocked == 0) { /* * When vpislocked == 0, the vnode is either exclusively * locked by this thread or not locked by this thread. * As such, shared lock it, if not exclusively locked. */ if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) { lockedit = 1; NFSVOPLOCK(vp, LK_SHARED | LK_RETRY); } } error = VOP_GETATTR(vp, &nvap->na_vattr, cred); if (lockedit != 0) NFSVOPUNLOCK(vp, 0); NFSEXITCODE(error); return (error); } /* * Get a file handle for a vnode. */ int nfsvno_getfh(struct vnode *vp, fhandle_t *fhp, struct thread *p) { int error; NFSBZERO((caddr_t)fhp, sizeof(fhandle_t)); fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; error = VOP_VPTOFH(vp, &fhp->fh_fid); NFSEXITCODE(error); return (error); } /* * Perform access checking for vnodes obtained from file handles that would * refer to files already opened by a Unix client. You cannot just use * vn_writechk() and VOP_ACCESSX() for two reasons. * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write * case. * 2 - The owner is to be given access irrespective of mode bits for some * operations, so that processes that chmod after opening a file don't * break. */ int nfsvno_accchk(struct vnode *vp, accmode_t accmode, struct ucred *cred, struct nfsexstuff *exp, struct thread *p, int override, int vpislocked, u_int32_t *supportedtypep) { struct vattr vattr; int error = 0, getret = 0; if (vpislocked == 0) { if (NFSVOPLOCK(vp, LK_SHARED) != 0) { error = EPERM; goto out; } } if (accmode & VWRITE) { /* Just vn_writechk() changed to check rdonly */ /* * Disallow write attempts on read-only file systems; * unless the file is a socket or a block or character * device resident on the file system. */ if (NFSVNO_EXRDONLY(exp) || (vp->v_mount->mnt_flag & MNT_RDONLY)) { switch (vp->v_type) { case VREG: case VDIR: case VLNK: error = EROFS; default: break; } } /* * If there's shared text associated with * the inode, try to free it up once. If * we fail, we can't allow writing. */ if (VOP_IS_TEXT(vp) && error == 0) error = ETXTBSY; } if (error != 0) { if (vpislocked == 0) NFSVOPUNLOCK(vp, 0); goto out; } /* * Should the override still be applied when ACLs are enabled? */ error = VOP_ACCESSX(vp, accmode, cred, p); if (error != 0 && (accmode & (VDELETE | VDELETE_CHILD))) { /* * Try again with VEXPLICIT_DENY, to see if the test for * deletion is supported. */ error = VOP_ACCESSX(vp, accmode | VEXPLICIT_DENY, cred, p); if (error == 0) { if (vp->v_type == VDIR) { accmode &= ~(VDELETE | VDELETE_CHILD); accmode |= VWRITE; error = VOP_ACCESSX(vp, accmode, cred, p); } else if (supportedtypep != NULL) { *supportedtypep &= ~NFSACCESS_DELETE; } } } /* * Allow certain operations for the owner (reads and writes * on files that are already open). */ if (override != NFSACCCHK_NOOVERRIDE && (error == EPERM || error == EACCES)) { if (cred->cr_uid == 0 && (override & NFSACCCHK_ALLOWROOT)) error = 0; else if (override & NFSACCCHK_ALLOWOWNER) { getret = VOP_GETATTR(vp, &vattr, cred); if (getret == 0 && cred->cr_uid == vattr.va_uid) error = 0; } } if (vpislocked == 0) NFSVOPUNLOCK(vp, 0); out: NFSEXITCODE(error); return (error); } /* * Set attribute(s) vnop. */ int nfsvno_setattr(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred, struct thread *p, struct nfsexstuff *exp) { int error; error = VOP_SETATTR(vp, &nvap->na_vattr, cred); NFSEXITCODE(error); return (error); } /* * Set up nameidata for a lookup() call and do it. */ int nfsvno_namei(struct nfsrv_descript *nd, struct nameidata *ndp, struct vnode *dp, int islocked, struct nfsexstuff *exp, struct thread *p, struct vnode **retdirp) { struct componentname *cnp = &ndp->ni_cnd; int i; struct iovec aiov; struct uio auio; int lockleaf = (cnp->cn_flags & LOCKLEAF) != 0, linklen; int error = 0, crossmnt; char *cp; *retdirp = NULL; cnp->cn_nameptr = cnp->cn_pnbuf; ndp->ni_strictrelative = 0; /* * Extract and set starting directory. */ if (dp->v_type != VDIR) { if (islocked) vput(dp); else vrele(dp); nfsvno_relpathbuf(ndp); error = ENOTDIR; goto out1; } if (islocked) NFSVOPUNLOCK(dp, 0); VREF(dp); *retdirp = dp; if (NFSVNO_EXRDONLY(exp)) cnp->cn_flags |= RDONLY; ndp->ni_segflg = UIO_SYSSPACE; crossmnt = 1; if (nd->nd_flag & ND_PUBLOOKUP) { ndp->ni_loopcnt = 0; if (cnp->cn_pnbuf[0] == '/') { vrele(dp); /* * Check for degenerate pathnames here, since lookup() * panics on them. */ for (i = 1; i < ndp->ni_pathlen; i++) if (cnp->cn_pnbuf[i] != '/') break; if (i == ndp->ni_pathlen) { error = NFSERR_ACCES; goto out; } dp = rootvnode; VREF(dp); } } else if ((nfsrv_enable_crossmntpt == 0 && NFSVNO_EXPORTED(exp)) || (nd->nd_flag & ND_NFSV4) == 0) { /* * Only cross mount points for NFSv4 when doing a * mount while traversing the file system above * the mount point, unless nfsrv_enable_crossmntpt is set. */ cnp->cn_flags |= NOCROSSMOUNT; crossmnt = 0; } /* * Initialize for scan, set ni_startdir and bump ref on dp again * because lookup() will dereference ni_startdir. */ cnp->cn_thread = p; ndp->ni_startdir = dp; ndp->ni_rootdir = rootvnode; ndp->ni_topdir = NULL; if (!lockleaf) cnp->cn_flags |= LOCKLEAF; for (;;) { cnp->cn_nameptr = cnp->cn_pnbuf; /* * Call lookup() to do the real work. If an error occurs, * ndp->ni_vp and ni_dvp are left uninitialized or NULL and * we do not have to dereference anything before returning. * In either case ni_startdir will be dereferenced and NULLed * out. */ error = lookup(ndp); if (error) break; /* * Check for encountering a symbolic link. Trivial * termination occurs if no symlink encountered. */ if ((cnp->cn_flags & ISSYMLINK) == 0) { if ((cnp->cn_flags & (SAVENAME | SAVESTART)) == 0) nfsvno_relpathbuf(ndp); if (ndp->ni_vp && !lockleaf) NFSVOPUNLOCK(ndp->ni_vp, 0); break; } /* * Validate symlink */ if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1) NFSVOPUNLOCK(ndp->ni_dvp, 0); if (!(nd->nd_flag & ND_PUBLOOKUP)) { error = EINVAL; goto badlink2; } if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { error = ELOOP; goto badlink2; } if (ndp->ni_pathlen > 1) cp = uma_zalloc(namei_zone, M_WAITOK); else cp = cnp->cn_pnbuf; aiov.iov_base = cp; aiov.iov_len = MAXPATHLEN; auio.uio_iov = &aiov; auio.uio_iovcnt = 1; auio.uio_offset = 0; auio.uio_rw = UIO_READ; auio.uio_segflg = UIO_SYSSPACE; auio.uio_td = NULL; auio.uio_resid = MAXPATHLEN; error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); if (error) { badlink1: if (ndp->ni_pathlen > 1) uma_zfree(namei_zone, cp); badlink2: vrele(ndp->ni_dvp); vput(ndp->ni_vp); break; } linklen = MAXPATHLEN - auio.uio_resid; if (linklen == 0) { error = ENOENT; goto badlink1; } if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { error = ENAMETOOLONG; goto badlink1; } /* * Adjust or replace path */ if (ndp->ni_pathlen > 1) { NFSBCOPY(ndp->ni_next, cp + linklen, ndp->ni_pathlen); uma_zfree(namei_zone, cnp->cn_pnbuf); cnp->cn_pnbuf = cp; } else cnp->cn_pnbuf[linklen] = '\0'; ndp->ni_pathlen += linklen; /* * Cleanup refs for next loop and check if root directory * should replace current directory. Normally ni_dvp * becomes the new base directory and is cleaned up when * we loop. Explicitly null pointers after invalidation * to clarify operation. */ vput(ndp->ni_vp); ndp->ni_vp = NULL; if (cnp->cn_pnbuf[0] == '/') { vrele(ndp->ni_dvp); ndp->ni_dvp = ndp->ni_rootdir; VREF(ndp->ni_dvp); } ndp->ni_startdir = ndp->ni_dvp; ndp->ni_dvp = NULL; } if (!lockleaf) cnp->cn_flags &= ~LOCKLEAF; out: if (error) { nfsvno_relpathbuf(ndp); ndp->ni_vp = NULL; ndp->ni_dvp = NULL; ndp->ni_startdir = NULL; } else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) { ndp->ni_dvp = NULL; } out1: NFSEXITCODE2(error, nd); return (error); } /* * Set up a pathname buffer and return a pointer to it and, optionally * set a hash pointer. */ void nfsvno_setpathbuf(struct nameidata *ndp, char **bufpp, u_long **hashpp) { struct componentname *cnp = &ndp->ni_cnd; cnp->cn_flags |= (NOMACCHECK | HASBUF); cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK); if (hashpp != NULL) *hashpp = NULL; *bufpp = cnp->cn_pnbuf; } /* * Release the above path buffer, if not released by nfsvno_namei(). */ void nfsvno_relpathbuf(struct nameidata *ndp) { if ((ndp->ni_cnd.cn_flags & HASBUF) == 0) panic("nfsrelpath"); uma_zfree(namei_zone, ndp->ni_cnd.cn_pnbuf); ndp->ni_cnd.cn_flags &= ~HASBUF; } /* * Readlink vnode op into an mbuf list. */ int nfsvno_readlink(struct vnode *vp, struct ucred *cred, struct thread *p, struct mbuf **mpp, struct mbuf **mpendp, int *lenp) { struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; struct iovec *ivp = iv; struct uio io, *uiop = &io; struct mbuf *mp, *mp2 = NULL, *mp3 = NULL; int i, len, tlen, error = 0; len = 0; i = 0; while (len < NFS_MAXPATHLEN) { NFSMGET(mp); MCLGET(mp, M_WAITOK); mp->m_len = NFSMSIZ(mp); if (len == 0) { mp3 = mp2 = mp; } else { mp2->m_next = mp; mp2 = mp; } if ((len + mp->m_len) > NFS_MAXPATHLEN) { mp->m_len = NFS_MAXPATHLEN - len; len = NFS_MAXPATHLEN; } else { len += mp->m_len; } ivp->iov_base = mtod(mp, caddr_t); ivp->iov_len = mp->m_len; i++; ivp++; } uiop->uio_iov = iv; uiop->uio_iovcnt = i; uiop->uio_offset = 0; uiop->uio_resid = len; uiop->uio_rw = UIO_READ; uiop->uio_segflg = UIO_SYSSPACE; uiop->uio_td = NULL; error = VOP_READLINK(vp, uiop, cred); if (error) { m_freem(mp3); *lenp = 0; goto out; } if (uiop->uio_resid > 0) { len -= uiop->uio_resid; tlen = NFSM_RNDUP(len); nfsrv_adj(mp3, NFS_MAXPATHLEN - tlen, tlen - len); } *lenp = len; *mpp = mp3; *mpendp = mp; out: NFSEXITCODE(error); return (error); } /* * Read vnode op call into mbuf list. */ int nfsvno_read(struct vnode *vp, off_t off, int cnt, struct ucred *cred, struct thread *p, struct mbuf **mpp, struct mbuf **mpendp) { struct mbuf *m; int i; struct iovec *iv; struct iovec *iv2; int error = 0, len, left, siz, tlen, ioflag = 0; struct mbuf *m2 = NULL, *m3; struct uio io, *uiop = &io; struct nfsheur *nh; len = left = NFSM_RNDUP(cnt); m3 = NULL; /* * Generate the mbuf list with the uio_iov ref. to it. */ i = 0; while (left > 0) { NFSMGET(m); MCLGET(m, M_WAITOK); m->m_len = 0; siz = min(M_TRAILINGSPACE(m), left); left -= siz; i++; if (m3) m2->m_next = m; else m3 = m; m2 = m; } MALLOC(iv, struct iovec *, i * sizeof (struct iovec), M_TEMP, M_WAITOK); uiop->uio_iov = iv2 = iv; m = m3; left = len; i = 0; while (left > 0) { if (m == NULL) panic("nfsvno_read iov"); siz = min(M_TRAILINGSPACE(m), left); if (siz > 0) { iv->iov_base = mtod(m, caddr_t) + m->m_len; iv->iov_len = siz; m->m_len += siz; left -= siz; iv++; i++; } m = m->m_next; } uiop->uio_iovcnt = i; uiop->uio_offset = off; uiop->uio_resid = len; uiop->uio_rw = UIO_READ; uiop->uio_segflg = UIO_SYSSPACE; + uiop->uio_td = NULL; nh = nfsrv_sequential_heuristic(uiop, vp); ioflag |= nh->nh_seqcount << IO_SEQSHIFT; error = VOP_READ(vp, uiop, IO_NODELOCKED | ioflag, cred); FREE((caddr_t)iv2, M_TEMP); if (error) { m_freem(m3); *mpp = NULL; goto out; } nh->nh_nextoff = uiop->uio_offset; tlen = len - uiop->uio_resid; cnt = cnt < tlen ? cnt : tlen; tlen = NFSM_RNDUP(cnt); if (tlen == 0) { m_freem(m3); m3 = NULL; } else if (len != tlen || tlen != cnt) nfsrv_adj(m3, len - tlen, tlen - cnt); *mpp = m3; *mpendp = m2; out: NFSEXITCODE(error); return (error); } /* * Write vnode op from an mbuf list. */ int nfsvno_write(struct vnode *vp, off_t off, int retlen, int cnt, int stable, struct mbuf *mp, char *cp, struct ucred *cred, struct thread *p) { struct iovec *ivp; int i, len; struct iovec *iv; int ioflags, error; struct uio io, *uiop = &io; struct nfsheur *nh; MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); uiop->uio_iov = iv = ivp; uiop->uio_iovcnt = cnt; i = mtod(mp, caddr_t) + mp->m_len - cp; len = retlen; while (len > 0) { if (mp == NULL) panic("nfsvno_write"); if (i > 0) { i = min(i, len); ivp->iov_base = cp; ivp->iov_len = i; ivp++; len -= i; } mp = mp->m_next; if (mp) { i = mp->m_len; cp = mtod(mp, caddr_t); } } if (stable == NFSWRITE_UNSTABLE) ioflags = IO_NODELOCKED; else ioflags = (IO_SYNC | IO_NODELOCKED); uiop->uio_resid = retlen; uiop->uio_rw = UIO_WRITE; uiop->uio_segflg = UIO_SYSSPACE; NFSUIOPROC(uiop, p); uiop->uio_offset = off; nh = nfsrv_sequential_heuristic(uiop, vp); ioflags |= nh->nh_seqcount << IO_SEQSHIFT; error = VOP_WRITE(vp, uiop, ioflags, cred); if (error == 0) nh->nh_nextoff = uiop->uio_offset; FREE((caddr_t)iv, M_TEMP); NFSEXITCODE(error); return (error); } /* * Common code for creating a regular file (plus special files for V2). */ int nfsvno_createsub(struct nfsrv_descript *nd, struct nameidata *ndp, struct vnode **vpp, struct nfsvattr *nvap, int *exclusive_flagp, int32_t *cverf, NFSDEV_T rdev, struct thread *p, struct nfsexstuff *exp) { u_quad_t tempsize; int error; error = nd->nd_repstat; if (!error && ndp->ni_vp == NULL) { if (nvap->na_type == VREG || nvap->na_type == VSOCK) { vrele(ndp->ni_startdir); error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); vput(ndp->ni_dvp); nfsvno_relpathbuf(ndp); if (!error) { if (*exclusive_flagp) { *exclusive_flagp = 0; NFSVNO_ATTRINIT(nvap); nvap->na_atime.tv_sec = cverf[0]; nvap->na_atime.tv_nsec = cverf[1]; error = VOP_SETATTR(ndp->ni_vp, &nvap->na_vattr, nd->nd_cred); } } /* * NFS V2 Only. nfsrvd_mknod() does this for V3. * (This implies, just get out on an error.) */ } else if (nvap->na_type == VCHR || nvap->na_type == VBLK || nvap->na_type == VFIFO) { if (nvap->na_type == VCHR && rdev == 0xffffffff) nvap->na_type = VFIFO; if (nvap->na_type != VFIFO && (error = priv_check_cred(nd->nd_cred, PRIV_VFS_MKNOD_DEV, 0))) { vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); vput(ndp->ni_dvp); goto out; } nvap->na_rdev = rdev; error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); vput(ndp->ni_dvp); nfsvno_relpathbuf(ndp); vrele(ndp->ni_startdir); if (error) goto out; } else { vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); vput(ndp->ni_dvp); error = ENXIO; goto out; } *vpp = ndp->ni_vp; } else { /* * Handle cases where error is already set and/or * the file exists. * 1 - clean up the lookup * 2 - iff !error and na_size set, truncate it */ vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); *vpp = ndp->ni_vp; if (ndp->ni_dvp == *vpp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); if (!error && nvap->na_size != VNOVAL) { error = nfsvno_accchk(*vpp, VWRITE, nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, NFSACCCHK_VPISLOCKED, NULL); if (!error) { tempsize = nvap->na_size; NFSVNO_ATTRINIT(nvap); nvap->na_size = tempsize; error = VOP_SETATTR(*vpp, &nvap->na_vattr, nd->nd_cred); } } if (error) vput(*vpp); } out: NFSEXITCODE(error); return (error); } /* * Do a mknod vnode op. */ int nfsvno_mknod(struct nameidata *ndp, struct nfsvattr *nvap, struct ucred *cred, struct thread *p) { int error = 0; enum vtype vtyp; vtyp = nvap->na_type; /* * Iff doesn't exist, create it. */ if (ndp->ni_vp) { vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); vput(ndp->ni_dvp); vrele(ndp->ni_vp); error = EEXIST; goto out; } if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); vput(ndp->ni_dvp); error = NFSERR_BADTYPE; goto out; } if (vtyp == VSOCK) { vrele(ndp->ni_startdir); error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); vput(ndp->ni_dvp); nfsvno_relpathbuf(ndp); } else { if (nvap->na_type != VFIFO && (error = priv_check_cred(cred, PRIV_VFS_MKNOD_DEV, 0))) { vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); vput(ndp->ni_dvp); goto out; } error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); vput(ndp->ni_dvp); nfsvno_relpathbuf(ndp); vrele(ndp->ni_startdir); /* * Since VOP_MKNOD returns the ni_vp, I can't * see any reason to do the lookup. */ } out: NFSEXITCODE(error); return (error); } /* * Mkdir vnode op. */ int nfsvno_mkdir(struct nameidata *ndp, struct nfsvattr *nvap, uid_t saved_uid, struct ucred *cred, struct thread *p, struct nfsexstuff *exp) { int error = 0; if (ndp->ni_vp != NULL) { if (ndp->ni_dvp == ndp->ni_vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); vrele(ndp->ni_vp); nfsvno_relpathbuf(ndp); error = EEXIST; goto out; } error = VOP_MKDIR(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); vput(ndp->ni_dvp); nfsvno_relpathbuf(ndp); out: NFSEXITCODE(error); return (error); } /* * symlink vnode op. */ int nfsvno_symlink(struct nameidata *ndp, struct nfsvattr *nvap, char *pathcp, int pathlen, int not_v2, uid_t saved_uid, struct ucred *cred, struct thread *p, struct nfsexstuff *exp) { int error = 0; if (ndp->ni_vp) { vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); if (ndp->ni_dvp == ndp->ni_vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); vrele(ndp->ni_vp); error = EEXIST; goto out; } error = VOP_SYMLINK(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr, pathcp); vput(ndp->ni_dvp); vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); /* * Although FreeBSD still had the lookup code in * it for 7/current, there doesn't seem to be any * point, since VOP_SYMLINK() returns the ni_vp. * Just vput it for v2. */ if (!not_v2 && !error) vput(ndp->ni_vp); out: NFSEXITCODE(error); return (error); } /* * Parse symbolic link arguments. * This function has an ugly side effect. It will MALLOC() an area for * the symlink and set iov_base to point to it, only if it succeeds. * So, if it returns with uiop->uio_iov->iov_base != NULL, that must * be FREE'd later. */ int nfsvno_getsymlink(struct nfsrv_descript *nd, struct nfsvattr *nvap, struct thread *p, char **pathcpp, int *lenp) { u_int32_t *tl; char *pathcp = NULL; int error = 0, len; struct nfsv2_sattr *sp; *pathcpp = NULL; *lenp = 0; if ((nd->nd_flag & ND_NFSV3) && (error = nfsrv_sattr(nd, nvap, NULL, NULL, p))) goto nfsmout; NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); len = fxdr_unsigned(int, *tl); if (len > NFS_MAXPATHLEN || len <= 0) { error = EBADRPC; goto nfsmout; } MALLOC(pathcp, caddr_t, len + 1, M_TEMP, M_WAITOK); error = nfsrv_mtostr(nd, pathcp, len); if (error) goto nfsmout; if (nd->nd_flag & ND_NFSV2) { NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR); nvap->na_mode = fxdr_unsigned(u_int16_t, sp->sa_mode); } *pathcpp = pathcp; *lenp = len; NFSEXITCODE2(0, nd); return (0); nfsmout: if (pathcp) free(pathcp, M_TEMP); NFSEXITCODE2(error, nd); return (error); } /* * Remove a non-directory object. */ int nfsvno_removesub(struct nameidata *ndp, int is_v4, struct ucred *cred, struct thread *p, struct nfsexstuff *exp) { struct vnode *vp; int error = 0; vp = ndp->ni_vp; if (vp->v_type == VDIR) error = NFSERR_ISDIR; else if (is_v4) error = nfsrv_checkremove(vp, 1, p); if (!error) error = VOP_REMOVE(ndp->ni_dvp, vp, &ndp->ni_cnd); if (ndp->ni_dvp == vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); vput(vp); if ((ndp->ni_cnd.cn_flags & SAVENAME) != 0) nfsvno_relpathbuf(ndp); NFSEXITCODE(error); return (error); } /* * Remove a directory. */ int nfsvno_rmdirsub(struct nameidata *ndp, int is_v4, struct ucred *cred, struct thread *p, struct nfsexstuff *exp) { struct vnode *vp; int error = 0; vp = ndp->ni_vp; if (vp->v_type != VDIR) { error = ENOTDIR; goto out; } /* * No rmdir "." please. */ if (ndp->ni_dvp == vp) { error = EINVAL; goto out; } /* * The root of a mounted filesystem cannot be deleted. */ if (vp->v_vflag & VV_ROOT) error = EBUSY; out: if (!error) error = VOP_RMDIR(ndp->ni_dvp, vp, &ndp->ni_cnd); if (ndp->ni_dvp == vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); vput(vp); if ((ndp->ni_cnd.cn_flags & SAVENAME) != 0) nfsvno_relpathbuf(ndp); NFSEXITCODE(error); return (error); } /* * Rename vnode op. */ int nfsvno_rename(struct nameidata *fromndp, struct nameidata *tondp, u_int32_t ndstat, u_int32_t ndflag, struct ucred *cred, struct thread *p) { struct vnode *fvp, *tvp, *tdvp; int error = 0; fvp = fromndp->ni_vp; if (ndstat) { vrele(fromndp->ni_dvp); vrele(fvp); error = ndstat; goto out1; } tdvp = tondp->ni_dvp; tvp = tondp->ni_vp; if (tvp != NULL) { if (fvp->v_type == VDIR && tvp->v_type != VDIR) { error = (ndflag & ND_NFSV2) ? EISDIR : EEXIST; goto out; } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { error = (ndflag & ND_NFSV2) ? ENOTDIR : EEXIST; goto out; } if (tvp->v_type == VDIR && tvp->v_mountedhere) { error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; goto out; } /* * A rename to '.' or '..' results in a prematurely * unlocked vnode on FreeBSD5, so I'm just going to fail that * here. */ if ((tondp->ni_cnd.cn_namelen == 1 && tondp->ni_cnd.cn_nameptr[0] == '.') || (tondp->ni_cnd.cn_namelen == 2 && tondp->ni_cnd.cn_nameptr[0] == '.' && tondp->ni_cnd.cn_nameptr[1] == '.')) { error = EINVAL; goto out; } } if (fvp->v_type == VDIR && fvp->v_mountedhere) { error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; goto out; } if (fvp->v_mount != tdvp->v_mount) { error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV; goto out; } if (fvp == tdvp) { error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EINVAL; goto out; } if (fvp == tvp) { /* * If source and destination are the same, there is nothing to * do. Set error to -1 to indicate this. */ error = -1; goto out; } if (ndflag & ND_NFSV4) { if (NFSVOPLOCK(fvp, LK_EXCLUSIVE) == 0) { error = nfsrv_checkremove(fvp, 0, p); NFSVOPUNLOCK(fvp, 0); } else error = EPERM; if (tvp && !error) error = nfsrv_checkremove(tvp, 1, p); } else { /* * For NFSv2 and NFSv3, try to get rid of the delegation, so * that the NFSv4 client won't be confused by the rename. * Since nfsd_recalldelegation() can only be called on an * unlocked vnode at this point and fvp is the file that will * still exist after the rename, just do fvp. */ nfsd_recalldelegation(fvp, p); } out: if (!error) { error = VOP_RENAME(fromndp->ni_dvp, fromndp->ni_vp, &fromndp->ni_cnd, tondp->ni_dvp, tondp->ni_vp, &tondp->ni_cnd); } else { if (tdvp == tvp) vrele(tdvp); else vput(tdvp); if (tvp) vput(tvp); vrele(fromndp->ni_dvp); vrele(fvp); if (error == -1) error = 0; } vrele(tondp->ni_startdir); nfsvno_relpathbuf(tondp); out1: vrele(fromndp->ni_startdir); nfsvno_relpathbuf(fromndp); NFSEXITCODE(error); return (error); } /* * Link vnode op. */ int nfsvno_link(struct nameidata *ndp, struct vnode *vp, struct ucred *cred, struct thread *p, struct nfsexstuff *exp) { struct vnode *xp; int error = 0; xp = ndp->ni_vp; if (xp != NULL) { error = EEXIST; } else { xp = ndp->ni_dvp; if (vp->v_mount != xp->v_mount) error = EXDEV; } if (!error) { NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY); if ((vp->v_iflag & VI_DOOMED) == 0) error = VOP_LINK(ndp->ni_dvp, vp, &ndp->ni_cnd); else error = EPERM; if (ndp->ni_dvp == vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); NFSVOPUNLOCK(vp, 0); } else { if (ndp->ni_dvp == ndp->ni_vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); if (ndp->ni_vp) vrele(ndp->ni_vp); } nfsvno_relpathbuf(ndp); NFSEXITCODE(error); return (error); } /* * Do the fsync() appropriate for the commit. */ int nfsvno_fsync(struct vnode *vp, u_int64_t off, int cnt, struct ucred *cred, struct thread *td) { int error = 0; /* * RFC 1813 3.3.21: if count is 0, a flush from offset to the end of * file is done. At this time VOP_FSYNC does not accept offset and * byte count parameters so call VOP_FSYNC the whole file for now. * The same is true for NFSv4: RFC 3530 Sec. 14.2.3. */ if (cnt == 0 || cnt > MAX_COMMIT_COUNT) { /* * Give up and do the whole thing */ if (vp->v_object && (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) { VM_OBJECT_WLOCK(vp->v_object); vm_object_page_clean(vp->v_object, 0, 0, OBJPC_SYNC); VM_OBJECT_WUNLOCK(vp->v_object); } error = VOP_FSYNC(vp, MNT_WAIT, td); } else { /* * Locate and synchronously write any buffers that fall * into the requested range. Note: we are assuming that * f_iosize is a power of 2. */ int iosize = vp->v_mount->mnt_stat.f_iosize; int iomask = iosize - 1; struct bufobj *bo; daddr_t lblkno; /* * Align to iosize boundry, super-align to page boundry. */ if (off & iomask) { cnt += off & iomask; off &= ~(u_quad_t)iomask; } if (off & PAGE_MASK) { cnt += off & PAGE_MASK; off &= ~(u_quad_t)PAGE_MASK; } lblkno = off / iosize; if (vp->v_object && (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) { VM_OBJECT_WLOCK(vp->v_object); vm_object_page_clean(vp->v_object, off, off + cnt, OBJPC_SYNC); VM_OBJECT_WUNLOCK(vp->v_object); } bo = &vp->v_bufobj; BO_LOCK(bo); while (cnt > 0) { struct buf *bp; /* * If we have a buffer and it is marked B_DELWRI we * have to lock and write it. Otherwise the prior * write is assumed to have already been committed. * * gbincore() can return invalid buffers now so we * have to check that bit as well (though B_DELWRI * should not be set if B_INVAL is set there could be * a race here since we haven't locked the buffer). */ if ((bp = gbincore(&vp->v_bufobj, lblkno)) != NULL) { if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK, BO_LOCKPTR(bo)) == ENOLCK) { BO_LOCK(bo); continue; /* retry */ } if ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI) { bremfree(bp); bp->b_flags &= ~B_ASYNC; bwrite(bp); ++nfs_commit_miss; } else BUF_UNLOCK(bp); BO_LOCK(bo); } ++nfs_commit_blks; if (cnt < iosize) break; cnt -= iosize; ++lblkno; } BO_UNLOCK(bo); } NFSEXITCODE(error); return (error); } /* * Statfs vnode op. */ int nfsvno_statfs(struct vnode *vp, struct statfs *sf) { int error; error = VFS_STATFS(vp->v_mount, sf); if (error == 0) { /* * Since NFS handles these values as unsigned on the * wire, there is no way to represent negative values, * so set them to 0. Without this, they will appear * to be very large positive values for clients like * Solaris10. */ if (sf->f_bavail < 0) sf->f_bavail = 0; if (sf->f_ffree < 0) sf->f_ffree = 0; } NFSEXITCODE(error); return (error); } /* * Do the vnode op stuff for Open. Similar to nfsvno_createsub(), but * must handle nfsrv_opencheck() calls after any other access checks. */ void nfsvno_open(struct nfsrv_descript *nd, struct nameidata *ndp, nfsquad_t clientid, nfsv4stateid_t *stateidp, struct nfsstate *stp, int *exclusive_flagp, struct nfsvattr *nvap, int32_t *cverf, int create, NFSACL_T *aclp, nfsattrbit_t *attrbitp, struct ucred *cred, struct thread *p, struct nfsexstuff *exp, struct vnode **vpp) { struct vnode *vp = NULL; u_quad_t tempsize; struct nfsexstuff nes; if (ndp->ni_vp == NULL) nd->nd_repstat = nfsrv_opencheck(clientid, stateidp, stp, NULL, nd, p, nd->nd_repstat); if (!nd->nd_repstat) { if (ndp->ni_vp == NULL) { vrele(ndp->ni_startdir); nd->nd_repstat = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr); vput(ndp->ni_dvp); nfsvno_relpathbuf(ndp); if (!nd->nd_repstat) { if (*exclusive_flagp) { *exclusive_flagp = 0; NFSVNO_ATTRINIT(nvap); nvap->na_atime.tv_sec = cverf[0]; nvap->na_atime.tv_nsec = cverf[1]; nd->nd_repstat = VOP_SETATTR(ndp->ni_vp, &nvap->na_vattr, cred); } else { nfsrv_fixattr(nd, ndp->ni_vp, nvap, aclp, p, attrbitp, exp); } } vp = ndp->ni_vp; } else { if (ndp->ni_startdir) vrele(ndp->ni_startdir); nfsvno_relpathbuf(ndp); vp = ndp->ni_vp; if (create == NFSV4OPEN_CREATE) { if (ndp->ni_dvp == vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); } if (NFSVNO_ISSETSIZE(nvap) && vp->v_type == VREG) { if (ndp->ni_cnd.cn_flags & RDONLY) NFSVNO_SETEXRDONLY(&nes); else NFSVNO_EXINIT(&nes); nd->nd_repstat = nfsvno_accchk(vp, VWRITE, cred, &nes, p, NFSACCCHK_NOOVERRIDE, NFSACCCHK_VPISLOCKED, NULL); nd->nd_repstat = nfsrv_opencheck(clientid, stateidp, stp, vp, nd, p, nd->nd_repstat); if (!nd->nd_repstat) { tempsize = nvap->na_size; NFSVNO_ATTRINIT(nvap); nvap->na_size = tempsize; nd->nd_repstat = VOP_SETATTR(vp, &nvap->na_vattr, cred); } } else if (vp->v_type == VREG) { nd->nd_repstat = nfsrv_opencheck(clientid, stateidp, stp, vp, nd, p, nd->nd_repstat); } } } else { if (ndp->ni_cnd.cn_flags & HASBUF) nfsvno_relpathbuf(ndp); if (ndp->ni_startdir && create == NFSV4OPEN_CREATE) { vrele(ndp->ni_startdir); if (ndp->ni_dvp == ndp->ni_vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); if (ndp->ni_vp) vput(ndp->ni_vp); } } *vpp = vp; NFSEXITCODE2(0, nd); } /* * Updates the file rev and sets the mtime and ctime * to the current clock time, returning the va_filerev and va_Xtime * values. * Return ESTALE to indicate the vnode is VI_DOOMED. */ int nfsvno_updfilerev(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred, struct thread *p) { struct vattr va; VATTR_NULL(&va); vfs_timestamp(&va.va_mtime); if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) { NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY); if ((vp->v_iflag & VI_DOOMED) != 0) return (ESTALE); } (void) VOP_SETATTR(vp, &va, cred); (void) nfsvno_getattr(vp, nvap, cred, p, 1); return (0); } /* * Glue routine to nfsv4_fillattr(). */ int nfsvno_fillattr(struct nfsrv_descript *nd, struct mount *mp, struct vnode *vp, struct nfsvattr *nvap, fhandle_t *fhp, int rderror, nfsattrbit_t *attrbitp, struct ucred *cred, struct thread *p, int isdgram, int reterr, int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno) { int error; error = nfsv4_fillattr(nd, mp, vp, NULL, &nvap->na_vattr, fhp, rderror, attrbitp, cred, p, isdgram, reterr, supports_nfsv4acls, at_root, mounted_on_fileno); NFSEXITCODE2(0, nd); return (error); } /* Since the Readdir vnode ops vary, put the entire functions in here. */ /* * nfs readdir service * - mallocs what it thinks is enough to read * count rounded up to a multiple of DIRBLKSIZ <= NFS_MAXREADDIR * - calls VOP_READDIR() * - loops around building the reply * if the output generated exceeds count break out of loop * The NFSM_CLGET macro is used here so that the reply will be packed * tightly in mbuf clusters. * - it trims out records with d_fileno == 0 * this doesn't matter for Unix clients, but they might confuse clients * for other os'. * - it trims out records with d_type == DT_WHT * these cannot be seen through NFS (unless we extend the protocol) * The alternate call nfsrvd_readdirplus() does lookups as well. * PS: The NFS protocol spec. does not clarify what the "count" byte * argument is a count of.. just name strings and file id's or the * entire reply rpc or ... * I tried just file name and id sizes and it confused the Sun client, * so I am using the full rpc size now. The "paranoia.." comment refers * to including the status longwords that are not a part of the dir. * "entry" structures, but are in the rpc. */ int nfsrvd_readdir(struct nfsrv_descript *nd, int isdgram, struct vnode *vp, struct thread *p, struct nfsexstuff *exp) { struct dirent *dp; u_int32_t *tl; int dirlen; char *cpos, *cend, *rbuf; struct nfsvattr at; int nlen, error = 0, getret = 1; int siz, cnt, fullsiz, eofflag, ncookies; u_int64_t off, toff, verf; u_long *cookies = NULL, *cookiep; struct uio io; struct iovec iv; int not_zfs; if (nd->nd_repstat) { nfsrv_postopattr(nd, getret, &at); goto out; } if (nd->nd_flag & ND_NFSV2) { NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); off = fxdr_unsigned(u_quad_t, *tl++); } else { NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED); off = fxdr_hyper(tl); tl += 2; verf = fxdr_hyper(tl); tl += 2; } toff = off; cnt = fxdr_unsigned(int, *tl); if (cnt > NFS_SRVMAXDATA(nd) || cnt < 0) cnt = NFS_SRVMAXDATA(nd); siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); fullsiz = siz; if (nd->nd_flag & ND_NFSV3) { nd->nd_repstat = getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); #if 0 /* * va_filerev is not sufficient as a cookie verifier, * since it is not supposed to change when entries are * removed/added unless that offset cookies returned to * the client are no longer valid. */ if (!nd->nd_repstat && toff && verf != at.na_filerev) nd->nd_repstat = NFSERR_BAD_COOKIE; #endif } if (!nd->nd_repstat && vp->v_type != VDIR) nd->nd_repstat = NFSERR_NOTDIR; if (nd->nd_repstat == 0 && cnt == 0) { if (nd->nd_flag & ND_NFSV2) /* NFSv2 does not have NFSERR_TOOSMALL */ nd->nd_repstat = EPERM; else nd->nd_repstat = NFSERR_TOOSMALL; } if (!nd->nd_repstat) nd->nd_repstat = nfsvno_accchk(vp, VEXEC, nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, NFSACCCHK_VPISLOCKED, NULL); if (nd->nd_repstat) { vput(vp); if (nd->nd_flag & ND_NFSV3) nfsrv_postopattr(nd, getret, &at); goto out; } not_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs"); MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); again: eofflag = 0; if (cookies) { free((caddr_t)cookies, M_TEMP); cookies = NULL; } iv.iov_base = rbuf; iv.iov_len = siz; io.uio_iov = &iv; io.uio_iovcnt = 1; io.uio_offset = (off_t)off; io.uio_resid = siz; io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_td = NULL; nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies, &cookies); off = (u_int64_t)io.uio_offset; if (io.uio_resid) siz -= io.uio_resid; if (!cookies && !nd->nd_repstat) nd->nd_repstat = NFSERR_PERM; if (nd->nd_flag & ND_NFSV3) { getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); if (!nd->nd_repstat) nd->nd_repstat = getret; } /* * Handles the failed cases. nd->nd_repstat == 0 past here. */ if (nd->nd_repstat) { vput(vp); free((caddr_t)rbuf, M_TEMP); if (cookies) free((caddr_t)cookies, M_TEMP); if (nd->nd_flag & ND_NFSV3) nfsrv_postopattr(nd, getret, &at); goto out; } /* * If nothing read, return eof * rpc reply */ if (siz == 0) { vput(vp); if (nd->nd_flag & ND_NFSV2) { NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); } else { nfsrv_postopattr(nd, getret, &at); NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED); txdr_hyper(at.na_filerev, tl); tl += 2; } *tl++ = newnfs_false; *tl = newnfs_true; FREE((caddr_t)rbuf, M_TEMP); FREE((caddr_t)cookies, M_TEMP); goto out; } /* * Check for degenerate cases of nothing useful read. * If so go try again */ cpos = rbuf; cend = rbuf + siz; dp = (struct dirent *)cpos; cookiep = cookies; /* * For some reason FreeBSD's ufs_readdir() chooses to back the * directory offset up to a block boundary, so it is necessary to * skip over the records that precede the requested offset. This * requires the assumption that file offset cookies monotonically * increase. * Since the offset cookies don't monotonically increase for ZFS, * this is not done when ZFS is the file system. */ while (cpos < cend && ncookies > 0 && (dp->d_fileno == 0 || dp->d_type == DT_WHT || (not_zfs != 0 && ((u_quad_t)(*cookiep)) <= toff))) { cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } if (cpos >= cend || ncookies == 0) { siz = fullsiz; toff = off; goto again; } vput(vp); /* * dirlen is the size of the reply, including all XDR and must * not exceed cnt. For NFSv2, RFC1094 didn't clearly indicate * if the XDR should be included in "count", but to be safe, we do. * (Include the two booleans at the end of the reply in dirlen now.) */ if (nd->nd_flag & ND_NFSV3) { nfsrv_postopattr(nd, getret, &at); NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); txdr_hyper(at.na_filerev, tl); dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED; } else { dirlen = 2 * NFSX_UNSIGNED; } /* Loop through the records and build reply */ while (cpos < cend && ncookies > 0) { nlen = dp->d_namlen; if (dp->d_fileno != 0 && dp->d_type != DT_WHT && nlen <= NFS_MAXNAMLEN) { if (nd->nd_flag & ND_NFSV3) dirlen += (6*NFSX_UNSIGNED + NFSM_RNDUP(nlen)); else dirlen += (4*NFSX_UNSIGNED + NFSM_RNDUP(nlen)); if (dirlen > cnt) { eofflag = 0; break; } /* * Build the directory record xdr from * the dirent entry. */ if (nd->nd_flag & ND_NFSV3) { NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); *tl++ = newnfs_true; *tl++ = 0; } else { NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); *tl++ = newnfs_true; } *tl = txdr_unsigned(dp->d_fileno); (void) nfsm_strtom(nd, dp->d_name, nlen); if (nd->nd_flag & ND_NFSV3) { NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); *tl++ = 0; } else NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED); *tl = txdr_unsigned(*cookiep); } cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } if (cpos < cend) eofflag = 0; NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); *tl++ = newnfs_false; if (eofflag) *tl = newnfs_true; else *tl = newnfs_false; FREE((caddr_t)rbuf, M_TEMP); FREE((caddr_t)cookies, M_TEMP); out: NFSEXITCODE2(0, nd); return (0); nfsmout: vput(vp); NFSEXITCODE2(error, nd); return (error); } /* * Readdirplus for V3 and Readdir for V4. */ int nfsrvd_readdirplus(struct nfsrv_descript *nd, int isdgram, struct vnode *vp, struct thread *p, struct nfsexstuff *exp) { struct dirent *dp; u_int32_t *tl; int dirlen; char *cpos, *cend, *rbuf; struct vnode *nvp; fhandle_t nfh; struct nfsvattr nva, at, *nvap = &nva; struct mbuf *mb0, *mb1; struct nfsreferral *refp; int nlen, r, error = 0, getret = 1, usevget = 1; int siz, cnt, fullsiz, eofflag, ncookies, entrycnt; caddr_t bpos0, bpos1; u_int64_t off, toff, verf; u_long *cookies = NULL, *cookiep; nfsattrbit_t attrbits, rderrbits, savbits; struct uio io; struct iovec iv; struct componentname cn; int at_root, needs_unbusy, not_zfs, supports_nfsv4acls; struct mount *mp, *new_mp; uint64_t mounted_on_fileno; if (nd->nd_repstat) { nfsrv_postopattr(nd, getret, &at); goto out; } NFSM_DISSECT(tl, u_int32_t *, 6 * NFSX_UNSIGNED); off = fxdr_hyper(tl); toff = off; tl += 2; verf = fxdr_hyper(tl); tl += 2; siz = fxdr_unsigned(int, *tl++); cnt = fxdr_unsigned(int, *tl); /* * Use the server's maximum data transfer size as the upper bound * on reply datalen. */ if (cnt > NFS_SRVMAXDATA(nd) || cnt < 0) cnt = NFS_SRVMAXDATA(nd); /* * siz is a "hint" of how much directory information (name, fileid, * cookie) should be in the reply. At least one client "hints" 0, * so I set it to cnt for that case. I also round it up to the * next multiple of DIRBLKSIZ. */ if (siz <= 0) siz = cnt; siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); if (nd->nd_flag & ND_NFSV4) { error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL); if (error) goto nfsmout; NFSSET_ATTRBIT(&savbits, &attrbits); NFSCLRNOTFILLABLE_ATTRBIT(&attrbits); NFSZERO_ATTRBIT(&rderrbits); NFSSETBIT_ATTRBIT(&rderrbits, NFSATTRBIT_RDATTRERROR); } else { NFSZERO_ATTRBIT(&attrbits); } fullsiz = siz; nd->nd_repstat = getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); if (!nd->nd_repstat) { if (off && verf != at.na_filerev) { /* * va_filerev is not sufficient as a cookie verifier, * since it is not supposed to change when entries are * removed/added unless that offset cookies returned to * the client are no longer valid. */ #if 0 if (nd->nd_flag & ND_NFSV4) { nd->nd_repstat = NFSERR_NOTSAME; } else { nd->nd_repstat = NFSERR_BAD_COOKIE; } #endif } else if ((nd->nd_flag & ND_NFSV4) && off == 0 && verf != 0) { nd->nd_repstat = NFSERR_BAD_COOKIE; } } if (!nd->nd_repstat && vp->v_type != VDIR) nd->nd_repstat = NFSERR_NOTDIR; if (!nd->nd_repstat && cnt == 0) nd->nd_repstat = NFSERR_TOOSMALL; if (!nd->nd_repstat) nd->nd_repstat = nfsvno_accchk(vp, VEXEC, nd->nd_cred, exp, p, NFSACCCHK_NOOVERRIDE, NFSACCCHK_VPISLOCKED, NULL); if (nd->nd_repstat) { vput(vp); if (nd->nd_flag & ND_NFSV3) nfsrv_postopattr(nd, getret, &at); goto out; } not_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs"); MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); again: eofflag = 0; if (cookies) { free((caddr_t)cookies, M_TEMP); cookies = NULL; } iv.iov_base = rbuf; iv.iov_len = siz; io.uio_iov = &iv; io.uio_iovcnt = 1; io.uio_offset = (off_t)off; io.uio_resid = siz; io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_td = NULL; nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies, &cookies); off = (u_int64_t)io.uio_offset; if (io.uio_resid) siz -= io.uio_resid; getret = nfsvno_getattr(vp, &at, nd->nd_cred, p, 1); if (!cookies && !nd->nd_repstat) nd->nd_repstat = NFSERR_PERM; if (!nd->nd_repstat) nd->nd_repstat = getret; if (nd->nd_repstat) { vput(vp); if (cookies) free((caddr_t)cookies, M_TEMP); free((caddr_t)rbuf, M_TEMP); if (nd->nd_flag & ND_NFSV3) nfsrv_postopattr(nd, getret, &at); goto out; } /* * If nothing read, return eof * rpc reply */ if (siz == 0) { vput(vp); if (nd->nd_flag & ND_NFSV3) nfsrv_postopattr(nd, getret, &at); NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED); txdr_hyper(at.na_filerev, tl); tl += 2; *tl++ = newnfs_false; *tl = newnfs_true; free((caddr_t)cookies, M_TEMP); free((caddr_t)rbuf, M_TEMP); goto out; } /* * Check for degenerate cases of nothing useful read. * If so go try again */ cpos = rbuf; cend = rbuf + siz; dp = (struct dirent *)cpos; cookiep = cookies; /* * For some reason FreeBSD's ufs_readdir() chooses to back the * directory offset up to a block boundary, so it is necessary to * skip over the records that precede the requested offset. This * requires the assumption that file offset cookies monotonically * increase. * Since the offset cookies don't monotonically increase for ZFS, * this is not done when ZFS is the file system. */ while (cpos < cend && ncookies > 0 && (dp->d_fileno == 0 || dp->d_type == DT_WHT || (not_zfs != 0 && ((u_quad_t)(*cookiep)) <= toff) || ((nd->nd_flag & ND_NFSV4) && ((dp->d_namlen == 1 && dp->d_name[0] == '.') || (dp->d_namlen==2 && dp->d_name[0]=='.' && dp->d_name[1]=='.'))))) { cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } if (cpos >= cend || ncookies == 0) { siz = fullsiz; toff = off; goto again; } /* * Busy the file system so that the mount point won't go away * and, as such, VFS_VGET() can be used safely. */ mp = vp->v_mount; vfs_ref(mp); NFSVOPUNLOCK(vp, 0); nd->nd_repstat = vfs_busy(mp, 0); vfs_rel(mp); if (nd->nd_repstat != 0) { vrele(vp); free(cookies, M_TEMP); free(rbuf, M_TEMP); if (nd->nd_flag & ND_NFSV3) nfsrv_postopattr(nd, getret, &at); goto out; } /* * Check to see if entries in this directory can be safely acquired * via VFS_VGET() or if a switch to VOP_LOOKUP() is required. * ZFS snapshot directories need VOP_LOOKUP(), so that any * automount of the snapshot directory that is required will * be done. * This needs to be done here for NFSv4, since NFSv4 never does * a VFS_VGET() for "." or "..". */ if (not_zfs == 0) { r = VFS_VGET(mp, at.na_fileid, LK_SHARED, &nvp); if (r == EOPNOTSUPP) { usevget = 0; cn.cn_nameiop = LOOKUP; cn.cn_lkflags = LK_SHARED | LK_RETRY; cn.cn_cred = nd->nd_cred; cn.cn_thread = p; } else if (r == 0) vput(nvp); } /* * Save this position, in case there is an error before one entry * is created. */ mb0 = nd->nd_mb; bpos0 = nd->nd_bpos; /* * Fill in the first part of the reply. * dirlen is the reply length in bytes and cannot exceed cnt. * (Include the two booleans at the end of the reply in dirlen now, * so we recognize when we have exceeded cnt.) */ if (nd->nd_flag & ND_NFSV3) { dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED; nfsrv_postopattr(nd, getret, &at); } else { dirlen = NFSX_VERF + 2 * NFSX_UNSIGNED; } NFSM_BUILD(tl, u_int32_t *, NFSX_VERF); txdr_hyper(at.na_filerev, tl); /* * Save this position, in case there is an empty reply needed. */ mb1 = nd->nd_mb; bpos1 = nd->nd_bpos; /* Loop through the records and build reply */ entrycnt = 0; while (cpos < cend && ncookies > 0 && dirlen < cnt) { nlen = dp->d_namlen; if (dp->d_fileno != 0 && dp->d_type != DT_WHT && nlen <= NFS_MAXNAMLEN && ((nd->nd_flag & ND_NFSV3) || nlen > 2 || (nlen==2 && (dp->d_name[0]!='.' || dp->d_name[1]!='.')) || (nlen == 1 && dp->d_name[0] != '.'))) { /* * Save the current position in the reply, in case * this entry exceeds cnt. */ mb1 = nd->nd_mb; bpos1 = nd->nd_bpos; /* * For readdir_and_lookup get the vnode using * the file number. */ nvp = NULL; refp = NULL; r = 0; at_root = 0; needs_unbusy = 0; new_mp = mp; mounted_on_fileno = (uint64_t)dp->d_fileno; if ((nd->nd_flag & ND_NFSV3) || NFSNONZERO_ATTRBIT(&savbits)) { if (nd->nd_flag & ND_NFSV4) refp = nfsv4root_getreferral(NULL, vp, dp->d_fileno); if (refp == NULL) { if (usevget) r = VFS_VGET(mp, dp->d_fileno, LK_SHARED, &nvp); else r = EOPNOTSUPP; if (r == EOPNOTSUPP) { if (usevget) { usevget = 0; cn.cn_nameiop = LOOKUP; cn.cn_lkflags = LK_SHARED | LK_RETRY; cn.cn_cred = nd->nd_cred; cn.cn_thread = p; } cn.cn_nameptr = dp->d_name; cn.cn_namelen = nlen; cn.cn_flags = ISLASTCN | NOFOLLOW | LOCKLEAF; if (nlen == 2 && dp->d_name[0] == '.' && dp->d_name[1] == '.') cn.cn_flags |= ISDOTDOT; if (NFSVOPLOCK(vp, LK_SHARED) != 0) { nd->nd_repstat = EPERM; break; } if ((vp->v_vflag & VV_ROOT) != 0 && (cn.cn_flags & ISDOTDOT) != 0) { vref(vp); nvp = vp; r = 0; } else { r = VOP_LOOKUP(vp, &nvp, &cn); if (vp != nvp) NFSVOPUNLOCK(vp, 0); } } /* * For NFSv4, check to see if nvp is * a mount point and get the mount * point vnode, as required. */ if (r == 0 && nfsrv_enable_crossmntpt != 0 && (nd->nd_flag & ND_NFSV4) != 0 && nvp->v_type == VDIR && nvp->v_mountedhere != NULL) { new_mp = nvp->v_mountedhere; r = vfs_busy(new_mp, 0); vput(nvp); nvp = NULL; if (r == 0) { r = VFS_ROOT(new_mp, LK_SHARED, &nvp); needs_unbusy = 1; if (r == 0) at_root = 1; } } } if (!r) { if (refp == NULL && ((nd->nd_flag & ND_NFSV3) || NFSNONZERO_ATTRBIT(&attrbits))) { r = nfsvno_getfh(nvp, &nfh, p); if (!r) r = nfsvno_getattr(nvp, nvap, nd->nd_cred, p, 1); if (r == 0 && not_zfs == 0 && nfsrv_enable_crossmntpt != 0 && (nd->nd_flag & ND_NFSV4) != 0 && nvp->v_type == VDIR && vp->v_mount != nvp->v_mount) { /* * For a ZFS snapshot, there is a * pseudo mount that does not set * v_mountedhere, so it needs to * be detected via a different * mount structure. */ at_root = 1; if (new_mp == mp) new_mp = nvp->v_mount; } } } else { nvp = NULL; } if (r) { if (!NFSISSET_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR)) { if (nvp != NULL) vput(nvp); if (needs_unbusy != 0) vfs_unbusy(new_mp); nd->nd_repstat = r; break; } } } /* * Build the directory record xdr */ if (nd->nd_flag & ND_NFSV3) { NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); *tl++ = newnfs_true; *tl++ = 0; *tl = txdr_unsigned(dp->d_fileno); dirlen += nfsm_strtom(nd, dp->d_name, nlen); NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); *tl++ = 0; *tl = txdr_unsigned(*cookiep); nfsrv_postopattr(nd, 0, nvap); dirlen += nfsm_fhtom(nd,(u_int8_t *)&nfh,0,1); dirlen += (5*NFSX_UNSIGNED+NFSX_V3POSTOPATTR); if (nvp != NULL) vput(nvp); } else { NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED); *tl++ = newnfs_true; *tl++ = 0; *tl = txdr_unsigned(*cookiep); dirlen += nfsm_strtom(nd, dp->d_name, nlen); if (nvp != NULL) { supports_nfsv4acls = nfs_supportsnfsv4acls(nvp); NFSVOPUNLOCK(nvp, 0); } else supports_nfsv4acls = 0; if (refp != NULL) { dirlen += nfsrv_putreferralattr(nd, &savbits, refp, 0, &nd->nd_repstat); if (nd->nd_repstat) { if (nvp != NULL) vrele(nvp); if (needs_unbusy != 0) vfs_unbusy(new_mp); break; } } else if (r) { dirlen += nfsvno_fillattr(nd, new_mp, nvp, nvap, &nfh, r, &rderrbits, nd->nd_cred, p, isdgram, 0, supports_nfsv4acls, at_root, mounted_on_fileno); } else { dirlen += nfsvno_fillattr(nd, new_mp, nvp, nvap, &nfh, r, &attrbits, nd->nd_cred, p, isdgram, 0, supports_nfsv4acls, at_root, mounted_on_fileno); } if (nvp != NULL) vrele(nvp); dirlen += (3 * NFSX_UNSIGNED); } if (needs_unbusy != 0) vfs_unbusy(new_mp); if (dirlen <= cnt) entrycnt++; } cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } vrele(vp); vfs_unbusy(mp); /* * If dirlen > cnt, we must strip off the last entry. If that * results in an empty reply, report NFSERR_TOOSMALL. */ if (dirlen > cnt || nd->nd_repstat) { if (!nd->nd_repstat && entrycnt == 0) nd->nd_repstat = NFSERR_TOOSMALL; if (nd->nd_repstat) newnfs_trimtrailing(nd, mb0, bpos0); else newnfs_trimtrailing(nd, mb1, bpos1); eofflag = 0; } else if (cpos < cend) eofflag = 0; if (!nd->nd_repstat) { NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED); *tl++ = newnfs_false; if (eofflag) *tl = newnfs_true; else *tl = newnfs_false; } FREE((caddr_t)cookies, M_TEMP); FREE((caddr_t)rbuf, M_TEMP); out: NFSEXITCODE2(0, nd); return (0); nfsmout: vput(vp); NFSEXITCODE2(error, nd); return (error); } /* * Get the settable attributes out of the mbuf list. * (Return 0 or EBADRPC) */ int nfsrv_sattr(struct nfsrv_descript *nd, struct nfsvattr *nvap, nfsattrbit_t *attrbitp, NFSACL_T *aclp, struct thread *p) { u_int32_t *tl; struct nfsv2_sattr *sp; int error = 0, toclient = 0; switch (nd->nd_flag & (ND_NFSV2 | ND_NFSV3 | ND_NFSV4)) { case ND_NFSV2: NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR); /* * Some old clients didn't fill in the high order 16bits. * --> check the low order 2 bytes for 0xffff */ if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) nvap->na_mode = nfstov_mode(sp->sa_mode); if (sp->sa_uid != newnfs_xdrneg1) nvap->na_uid = fxdr_unsigned(uid_t, sp->sa_uid); if (sp->sa_gid != newnfs_xdrneg1) nvap->na_gid = fxdr_unsigned(gid_t, sp->sa_gid); if (sp->sa_size != newnfs_xdrneg1) nvap->na_size = fxdr_unsigned(u_quad_t, sp->sa_size); if (sp->sa_atime.nfsv2_sec != newnfs_xdrneg1) { #ifdef notyet fxdr_nfsv2time(&sp->sa_atime, &nvap->na_atime); #else nvap->na_atime.tv_sec = fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec); nvap->na_atime.tv_nsec = 0; #endif } if (sp->sa_mtime.nfsv2_sec != newnfs_xdrneg1) fxdr_nfsv2time(&sp->sa_mtime, &nvap->na_mtime); break; case ND_NFSV3: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); if (*tl == newnfs_true) { NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); nvap->na_mode = nfstov_mode(*tl); } NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); if (*tl == newnfs_true) { NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); nvap->na_uid = fxdr_unsigned(uid_t, *tl); } NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); if (*tl == newnfs_true) { NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); nvap->na_gid = fxdr_unsigned(gid_t, *tl); } NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); if (*tl == newnfs_true) { NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); nvap->na_size = fxdr_hyper(tl); } NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); switch (fxdr_unsigned(int, *tl)) { case NFSV3SATTRTIME_TOCLIENT: NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); fxdr_nfsv3time(tl, &nvap->na_atime); toclient = 1; break; case NFSV3SATTRTIME_TOSERVER: vfs_timestamp(&nvap->na_atime); nvap->na_vaflags |= VA_UTIMES_NULL; break; }; NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); switch (fxdr_unsigned(int, *tl)) { case NFSV3SATTRTIME_TOCLIENT: NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED); fxdr_nfsv3time(tl, &nvap->na_mtime); nvap->na_vaflags &= ~VA_UTIMES_NULL; break; case NFSV3SATTRTIME_TOSERVER: vfs_timestamp(&nvap->na_mtime); if (!toclient) nvap->na_vaflags |= VA_UTIMES_NULL; break; }; break; case ND_NFSV4: error = nfsv4_sattr(nd, nvap, attrbitp, aclp, p); }; nfsmout: NFSEXITCODE2(error, nd); return (error); } /* * Handle the setable attributes for V4. * Returns NFSERR_BADXDR if it can't be parsed, 0 otherwise. */ int nfsv4_sattr(struct nfsrv_descript *nd, struct nfsvattr *nvap, nfsattrbit_t *attrbitp, NFSACL_T *aclp, struct thread *p) { u_int32_t *tl; int attrsum = 0; int i, j; int error, attrsize, bitpos, aclsize, aceerr, retnotsup = 0; int toclient = 0; u_char *cp, namestr[NFSV4_SMALLSTR + 1]; uid_t uid; gid_t gid; error = nfsrv_getattrbits(nd, attrbitp, NULL, &retnotsup); if (error) goto nfsmout; NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); attrsize = fxdr_unsigned(int, *tl); /* * Loop around getting the setable attributes. If an unsupported * one is found, set nd_repstat == NFSERR_ATTRNOTSUPP and return. */ if (retnotsup) { nd->nd_repstat = NFSERR_ATTRNOTSUPP; bitpos = NFSATTRBIT_MAX; } else { bitpos = 0; } for (; bitpos < NFSATTRBIT_MAX; bitpos++) { if (attrsum > attrsize) { error = NFSERR_BADXDR; goto nfsmout; } if (NFSISSET_ATTRBIT(attrbitp, bitpos)) switch (bitpos) { case NFSATTRBIT_SIZE: NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER); nvap->na_size = fxdr_hyper(tl); attrsum += NFSX_HYPER; break; case NFSATTRBIT_ACL: error = nfsrv_dissectacl(nd, aclp, &aceerr, &aclsize, p); if (error) goto nfsmout; if (aceerr && !nd->nd_repstat) nd->nd_repstat = aceerr; attrsum += aclsize; break; case NFSATTRBIT_ARCHIVE: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); if (!nd->nd_repstat) nd->nd_repstat = NFSERR_ATTRNOTSUPP; attrsum += NFSX_UNSIGNED; break; case NFSATTRBIT_HIDDEN: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); if (!nd->nd_repstat) nd->nd_repstat = NFSERR_ATTRNOTSUPP; attrsum += NFSX_UNSIGNED; break; case NFSATTRBIT_MIMETYPE: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); i = fxdr_unsigned(int, *tl); error = nfsm_advance(nd, NFSM_RNDUP(i), -1); if (error) goto nfsmout; if (!nd->nd_repstat) nd->nd_repstat = NFSERR_ATTRNOTSUPP; attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(i)); break; case NFSATTRBIT_MODE: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); nvap->na_mode = nfstov_mode(*tl); attrsum += NFSX_UNSIGNED; break; case NFSATTRBIT_OWNER: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); j = fxdr_unsigned(int, *tl); if (j < 0) { error = NFSERR_BADXDR; goto nfsmout; } if (j > NFSV4_SMALLSTR) cp = malloc(j + 1, M_NFSSTRING, M_WAITOK); else cp = namestr; error = nfsrv_mtostr(nd, cp, j); if (error) { if (j > NFSV4_SMALLSTR) free(cp, M_NFSSTRING); goto nfsmout; } if (!nd->nd_repstat) { nd->nd_repstat = nfsv4_strtouid(nd, cp, j, &uid, p); if (!nd->nd_repstat) nvap->na_uid = uid; } if (j > NFSV4_SMALLSTR) free(cp, M_NFSSTRING); attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j)); break; case NFSATTRBIT_OWNERGROUP: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); j = fxdr_unsigned(int, *tl); if (j < 0) { error = NFSERR_BADXDR; goto nfsmout; } if (j > NFSV4_SMALLSTR) cp = malloc(j + 1, M_NFSSTRING, M_WAITOK); else cp = namestr; error = nfsrv_mtostr(nd, cp, j); if (error) { if (j > NFSV4_SMALLSTR) free(cp, M_NFSSTRING); goto nfsmout; } if (!nd->nd_repstat) { nd->nd_repstat = nfsv4_strtogid(nd, cp, j, &gid, p); if (!nd->nd_repstat) nvap->na_gid = gid; } if (j > NFSV4_SMALLSTR) free(cp, M_NFSSTRING); attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j)); break; case NFSATTRBIT_SYSTEM: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); if (!nd->nd_repstat) nd->nd_repstat = NFSERR_ATTRNOTSUPP; attrsum += NFSX_UNSIGNED; break; case NFSATTRBIT_TIMEACCESSSET: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); attrsum += NFSX_UNSIGNED; if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) { NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); fxdr_nfsv4time(tl, &nvap->na_atime); toclient = 1; attrsum += NFSX_V4TIME; } else { vfs_timestamp(&nvap->na_atime); nvap->na_vaflags |= VA_UTIMES_NULL; } break; case NFSATTRBIT_TIMEBACKUP: NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); if (!nd->nd_repstat) nd->nd_repstat = NFSERR_ATTRNOTSUPP; attrsum += NFSX_V4TIME; break; case NFSATTRBIT_TIMECREATE: NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); if (!nd->nd_repstat) nd->nd_repstat = NFSERR_ATTRNOTSUPP; attrsum += NFSX_V4TIME; break; case NFSATTRBIT_TIMEMODIFYSET: NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); attrsum += NFSX_UNSIGNED; if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) { NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME); fxdr_nfsv4time(tl, &nvap->na_mtime); nvap->na_vaflags &= ~VA_UTIMES_NULL; attrsum += NFSX_V4TIME; } else { vfs_timestamp(&nvap->na_mtime); if (!toclient) nvap->na_vaflags |= VA_UTIMES_NULL; } break; default: nd->nd_repstat = NFSERR_ATTRNOTSUPP; /* * set bitpos so we drop out of the loop. */ bitpos = NFSATTRBIT_MAX; break; }; } /* * some clients pad the attrlist, so we need to skip over the * padding. */ if (attrsum > attrsize) { error = NFSERR_BADXDR; } else { attrsize = NFSM_RNDUP(attrsize); if (attrsum < attrsize) error = nfsm_advance(nd, attrsize - attrsum, -1); } nfsmout: NFSEXITCODE2(error, nd); return (error); } /* * Check/setup export credentials. */ int nfsd_excred(struct nfsrv_descript *nd, struct nfsexstuff *exp, struct ucred *credanon) { int error = 0; /* * Check/setup credentials. */ if (nd->nd_flag & ND_GSS) exp->nes_exflag &= ~MNT_EXPORTANON; /* * Check to see if the operation is allowed for this security flavor. * RFC2623 suggests that the NFSv3 Fsinfo RPC be allowed to * AUTH_NONE or AUTH_SYS for file systems requiring RPCSEC_GSS. * Also, allow Secinfo, so that it can acquire the correct flavor(s). */ if (nfsvno_testexp(nd, exp) && nd->nd_procnum != NFSV4OP_SECINFO && nd->nd_procnum != NFSPROC_FSINFO) { if (nd->nd_flag & ND_NFSV4) error = NFSERR_WRONGSEC; else error = (NFSERR_AUTHERR | AUTH_TOOWEAK); goto out; } /* * Check to see if the file system is exported V4 only. */ if (NFSVNO_EXV4ONLY(exp) && !(nd->nd_flag & ND_NFSV4)) { error = NFSERR_PROGNOTV4; goto out; } /* * Now, map the user credentials. * (Note that ND_AUTHNONE will only be set for an NFSv3 * Fsinfo RPC. If set for anything else, this code might need * to change.) */ if (NFSVNO_EXPORTED(exp) && ((!(nd->nd_flag & ND_GSS) && nd->nd_cred->cr_uid == 0) || NFSVNO_EXPORTANON(exp) || (nd->nd_flag & ND_AUTHNONE))) { nd->nd_cred->cr_uid = credanon->cr_uid; nd->nd_cred->cr_gid = credanon->cr_gid; crsetgroups(nd->nd_cred, credanon->cr_ngroups, credanon->cr_groups); } out: NFSEXITCODE2(error, nd); return (error); } /* * Check exports. */ int nfsvno_checkexp(struct mount *mp, struct sockaddr *nam, struct nfsexstuff *exp, struct ucred **credp) { int i, error, *secflavors; error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp, &exp->nes_numsecflavor, &secflavors); if (error) { if (nfs_rootfhset) { exp->nes_exflag = 0; exp->nes_numsecflavor = 0; error = 0; } } else { /* Copy the security flavors. */ for (i = 0; i < exp->nes_numsecflavor; i++) exp->nes_secflavors[i] = secflavors[i]; } NFSEXITCODE(error); return (error); } /* * Get a vnode for a file handle and export stuff. */ int nfsvno_fhtovp(struct mount *mp, fhandle_t *fhp, struct sockaddr *nam, int lktype, struct vnode **vpp, struct nfsexstuff *exp, struct ucred **credp) { int i, error, *secflavors; *credp = NULL; exp->nes_numsecflavor = 0; error = VFS_FHTOVP(mp, &fhp->fh_fid, lktype, vpp); if (error != 0) /* Make sure the server replies ESTALE to the client. */ error = ESTALE; if (nam && !error) { error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp, &exp->nes_numsecflavor, &secflavors); if (error) { if (nfs_rootfhset) { exp->nes_exflag = 0; exp->nes_numsecflavor = 0; error = 0; } else { vput(*vpp); } } else { /* Copy the security flavors. */ for (i = 0; i < exp->nes_numsecflavor; i++) exp->nes_secflavors[i] = secflavors[i]; } } NFSEXITCODE(error); return (error); } /* * nfsd_fhtovp() - convert a fh to a vnode ptr * - look up fsid in mount list (if not found ret error) * - get vp and export rights by calling nfsvno_fhtovp() * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon * for AUTH_SYS * - if mpp != NULL, return the mount point so that it can * be used for vn_finished_write() by the caller */ void nfsd_fhtovp(struct nfsrv_descript *nd, struct nfsrvfh *nfp, int lktype, struct vnode **vpp, struct nfsexstuff *exp, struct mount **mpp, int startwrite, struct thread *p) { struct mount *mp; struct ucred *credanon; fhandle_t *fhp; fhp = (fhandle_t *)nfp->nfsrvfh_data; /* * Check for the special case of the nfsv4root_fh. */ mp = vfs_busyfs(&fhp->fh_fsid); if (mpp != NULL) *mpp = mp; if (mp == NULL) { *vpp = NULL; nd->nd_repstat = ESTALE; goto out; } if (startwrite) { vn_start_write(NULL, mpp, V_WAIT); if (lktype == LK_SHARED && !(MNT_SHARED_WRITES(mp))) lktype = LK_EXCLUSIVE; } nd->nd_repstat = nfsvno_fhtovp(mp, fhp, nd->nd_nam, lktype, vpp, exp, &credanon); vfs_unbusy(mp); /* * For NFSv4 without a pseudo root fs, unexported file handles * can be returned, so that Lookup works everywhere. */ if (!nd->nd_repstat && exp->nes_exflag == 0 && !(nd->nd_flag & ND_NFSV4)) { vput(*vpp); nd->nd_repstat = EACCES; } /* * Personally, I've never seen any point in requiring a * reserved port#, since only in the rare case where the * clients are all boxes with secure system priviledges, * does it provide any enhanced security, but... some people * believe it to be useful and keep putting this code back in. * (There is also some "security checker" out there that * complains if the nfs server doesn't enforce this.) * However, note the following: * RFC3530 (NFSv4) specifies that a reserved port# not be * required. * RFC2623 recommends that, if a reserved port# is checked for, * that there be a way to turn that off--> ifdef'd. */ #ifdef NFS_REQRSVPORT if (!nd->nd_repstat) { struct sockaddr_in *saddr; struct sockaddr_in6 *saddr6; saddr = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in *); saddr6 = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in6 *); if (!(nd->nd_flag & ND_NFSV4) && ((saddr->sin_family == AF_INET && ntohs(saddr->sin_port) >= IPPORT_RESERVED) || (saddr6->sin6_family == AF_INET6 && ntohs(saddr6->sin6_port) >= IPPORT_RESERVED))) { vput(*vpp); nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); } } #endif /* NFS_REQRSVPORT */ /* * Check/setup credentials. */ if (!nd->nd_repstat) { nd->nd_saveduid = nd->nd_cred->cr_uid; nd->nd_repstat = nfsd_excred(nd, exp, credanon); if (nd->nd_repstat) vput(*vpp); } if (credanon != NULL) crfree(credanon); if (nd->nd_repstat) { if (startwrite) vn_finished_write(mp); *vpp = NULL; if (mpp != NULL) *mpp = NULL; } out: NFSEXITCODE2(0, nd); } /* * glue for fp. */ static int fp_getfvp(struct thread *p, int fd, struct file **fpp, struct vnode **vpp) { struct filedesc *fdp; struct file *fp; int error = 0; fdp = p->td_proc->p_fd; if (fd < 0 || fd >= fdp->fd_nfiles || (fp = fdp->fd_ofiles[fd].fde_file) == NULL) { error = EBADF; goto out; } *fpp = fp; out: NFSEXITCODE(error); return (error); } /* * Called from nfssvc() to update the exports list. Just call * vfs_export(). This has to be done, since the v4 root fake fs isn't * in the mount list. */ int nfsrv_v4rootexport(void *argp, struct ucred *cred, struct thread *p) { struct nfsex_args *nfsexargp = (struct nfsex_args *)argp; int error = 0; struct nameidata nd; fhandle_t fh; error = vfs_export(&nfsv4root_mnt, &nfsexargp->export); if ((nfsexargp->export.ex_flags & MNT_DELEXPORT) != 0) nfs_rootfhset = 0; else if (error == 0) { if (nfsexargp->fspec == NULL) { error = EPERM; goto out; } /* * If fspec != NULL, this is the v4root path. */ NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, nfsexargp->fspec, p); if ((error = namei(&nd)) != 0) goto out; error = nfsvno_getfh(nd.ni_vp, &fh, p); vrele(nd.ni_vp); if (!error) { nfs_rootfh.nfsrvfh_len = NFSX_MYFH; NFSBCOPY((caddr_t)&fh, nfs_rootfh.nfsrvfh_data, sizeof (fhandle_t)); nfs_rootfhset = 1; } } out: NFSEXITCODE(error); return (error); } /* * This function needs to test to see if the system is near its limit * for memory allocation via malloc() or mget() and return True iff * either of these resources are near their limit. * XXX (For now, this is just a stub.) */ int nfsrv_testmalloclimit = 0; int nfsrv_mallocmget_limit(void) { static int printmesg = 0; static int testval = 1; if (nfsrv_testmalloclimit && (testval++ % 1000) == 0) { if ((printmesg++ % 100) == 0) printf("nfsd: malloc/mget near limit\n"); return (1); } return (0); } /* * BSD specific initialization of a mount point. */ void nfsd_mntinit(void) { static int inited = 0; if (inited) return; inited = 1; nfsv4root_mnt.mnt_flag = (MNT_RDONLY | MNT_EXPORTED); TAILQ_INIT(&nfsv4root_mnt.mnt_nvnodelist); TAILQ_INIT(&nfsv4root_mnt.mnt_activevnodelist); nfsv4root_mnt.mnt_export = NULL; TAILQ_INIT(&nfsv4root_opt); TAILQ_INIT(&nfsv4root_newopt); nfsv4root_mnt.mnt_opt = &nfsv4root_opt; nfsv4root_mnt.mnt_optnew = &nfsv4root_newopt; nfsv4root_mnt.mnt_nvnodelistsize = 0; nfsv4root_mnt.mnt_activevnodelistsize = 0; } /* * Get a vnode for a file handle, without checking exports, etc. */ struct vnode * nfsvno_getvp(fhandle_t *fhp) { struct mount *mp; struct vnode *vp; int error; mp = vfs_busyfs(&fhp->fh_fsid); if (mp == NULL) return (NULL); error = VFS_FHTOVP(mp, &fhp->fh_fid, LK_EXCLUSIVE, &vp); vfs_unbusy(mp); if (error) return (NULL); return (vp); } /* * Do a local VOP_ADVLOCK(). */ int nfsvno_advlock(struct vnode *vp, int ftype, u_int64_t first, u_int64_t end, struct thread *td) { int error = 0; struct flock fl; u_int64_t tlen; if (nfsrv_dolocallocks == 0) goto out; /* Check for VI_DOOMED here, so that VOP_ADVLOCK() isn't performed. */ if ((vp->v_iflag & VI_DOOMED) != 0) { error = EPERM; goto out; } fl.l_whence = SEEK_SET; fl.l_type = ftype; fl.l_start = (off_t)first; if (end == NFS64BITSSET) { fl.l_len = 0; } else { tlen = end - first; fl.l_len = (off_t)tlen; } /* * For FreeBSD8, the l_pid and l_sysid must be set to the same * values for all calls, so that all locks will be held by the * nfsd server. (The nfsd server handles conflicts between the * various clients.) * Since an NFSv4 lockowner is a ClientID plus an array of up to 1024 * bytes, so it can't be put in l_sysid. */ if (nfsv4_sysid == 0) nfsv4_sysid = nlm_acquire_next_sysid(); fl.l_pid = (pid_t)0; fl.l_sysid = (int)nfsv4_sysid; NFSVOPUNLOCK(vp, 0); if (ftype == F_UNLCK) error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_UNLCK, &fl, (F_POSIX | F_REMOTE)); else error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_SETLK, &fl, (F_POSIX | F_REMOTE)); NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY); out: NFSEXITCODE(error); return (error); } /* * Check the nfsv4 root exports. */ int nfsvno_v4rootexport(struct nfsrv_descript *nd) { struct ucred *credanon; int exflags, error = 0, numsecflavor, *secflavors, i; error = vfs_stdcheckexp(&nfsv4root_mnt, nd->nd_nam, &exflags, &credanon, &numsecflavor, &secflavors); if (error) { error = NFSERR_PROGUNAVAIL; goto out; } if (credanon != NULL) crfree(credanon); for (i = 0; i < numsecflavor; i++) { if (secflavors[i] == AUTH_SYS) nd->nd_flag |= ND_EXAUTHSYS; else if (secflavors[i] == RPCSEC_GSS_KRB5) nd->nd_flag |= ND_EXGSS; else if (secflavors[i] == RPCSEC_GSS_KRB5I) nd->nd_flag |= ND_EXGSSINTEGRITY; else if (secflavors[i] == RPCSEC_GSS_KRB5P) nd->nd_flag |= ND_EXGSSPRIVACY; } out: NFSEXITCODE(error); return (error); } /* * Nfs server psuedo system call for the nfsd's */ /* * MPSAFE */ static int nfssvc_nfsd(struct thread *td, struct nfssvc_args *uap) { struct file *fp; struct nfsd_addsock_args sockarg; struct nfsd_nfsd_args nfsdarg; cap_rights_t rights; int error; if (uap->flag & NFSSVC_NFSDADDSOCK) { error = copyin(uap->argp, (caddr_t)&sockarg, sizeof (sockarg)); if (error) goto out; /* * Since we don't know what rights might be required, * pretend that we need them all. It is better to be too * careful than too reckless. */ error = fget(td, sockarg.sock, cap_rights_init(&rights, CAP_SOCK_SERVER), &fp); if (error != 0) goto out; if (fp->f_type != DTYPE_SOCKET) { fdrop(fp, td); error = EPERM; goto out; } error = nfsrvd_addsock(fp); fdrop(fp, td); } else if (uap->flag & NFSSVC_NFSDNFSD) { if (uap->argp == NULL) { error = EINVAL; goto out; } error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof (nfsdarg)); if (error) goto out; error = nfsrvd_nfsd(td, &nfsdarg); } else { error = nfssvc_srvcall(td, uap, td->td_ucred); } out: NFSEXITCODE(error); return (error); } static int nfssvc_srvcall(struct thread *p, struct nfssvc_args *uap, struct ucred *cred) { struct nfsex_args export; struct file *fp = NULL; int stablefd, len; struct nfsd_clid adminrevoke; struct nfsd_dumplist dumplist; struct nfsd_dumpclients *dumpclients; struct nfsd_dumplocklist dumplocklist; struct nfsd_dumplocks *dumplocks; struct nameidata nd; vnode_t vp; int error = EINVAL, igotlock; struct proc *procp; static int suspend_nfsd = 0; if (uap->flag & NFSSVC_PUBLICFH) { NFSBZERO((caddr_t)&nfs_pubfh.nfsrvfh_data, sizeof (fhandle_t)); error = copyin(uap->argp, &nfs_pubfh.nfsrvfh_data, sizeof (fhandle_t)); if (!error) nfs_pubfhset = 1; } else if (uap->flag & NFSSVC_V4ROOTEXPORT) { error = copyin(uap->argp,(caddr_t)&export, sizeof (struct nfsex_args)); if (!error) error = nfsrv_v4rootexport(&export, cred, p); } else if (uap->flag & NFSSVC_NOPUBLICFH) { nfs_pubfhset = 0; error = 0; } else if (uap->flag & NFSSVC_STABLERESTART) { error = copyin(uap->argp, (caddr_t)&stablefd, sizeof (int)); if (!error) error = fp_getfvp(p, stablefd, &fp, &vp); if (!error && (NFSFPFLAG(fp) & (FREAD | FWRITE)) != (FREAD | FWRITE)) error = EBADF; if (!error && newnfs_numnfsd != 0) error = EPERM; if (!error) { nfsrv_stablefirst.nsf_fp = fp; nfsrv_setupstable(p); } } else if (uap->flag & NFSSVC_ADMINREVOKE) { error = copyin(uap->argp, (caddr_t)&adminrevoke, sizeof (struct nfsd_clid)); if (!error) error = nfsrv_adminrevoke(&adminrevoke, p); } else if (uap->flag & NFSSVC_DUMPCLIENTS) { error = copyin(uap->argp, (caddr_t)&dumplist, sizeof (struct nfsd_dumplist)); if (!error && (dumplist.ndl_size < 1 || dumplist.ndl_size > NFSRV_MAXDUMPLIST)) error = EPERM; if (!error) { len = sizeof (struct nfsd_dumpclients) * dumplist.ndl_size; dumpclients = (struct nfsd_dumpclients *)malloc(len, M_TEMP, M_WAITOK); nfsrv_dumpclients(dumpclients, dumplist.ndl_size); error = copyout(dumpclients, CAST_USER_ADDR_T(dumplist.ndl_list), len); free((caddr_t)dumpclients, M_TEMP); } } else if (uap->flag & NFSSVC_DUMPLOCKS) { error = copyin(uap->argp, (caddr_t)&dumplocklist, sizeof (struct nfsd_dumplocklist)); if (!error && (dumplocklist.ndllck_size < 1 || dumplocklist.ndllck_size > NFSRV_MAXDUMPLIST)) error = EPERM; if (!error) error = nfsrv_lookupfilename(&nd, dumplocklist.ndllck_fname, p); if (!error) { len = sizeof (struct nfsd_dumplocks) * dumplocklist.ndllck_size; dumplocks = (struct nfsd_dumplocks *)malloc(len, M_TEMP, M_WAITOK); nfsrv_dumplocks(nd.ni_vp, dumplocks, dumplocklist.ndllck_size, p); vput(nd.ni_vp); error = copyout(dumplocks, CAST_USER_ADDR_T(dumplocklist.ndllck_list), len); free((caddr_t)dumplocks, M_TEMP); } } else if (uap->flag & NFSSVC_BACKUPSTABLE) { procp = p->td_proc; PROC_LOCK(procp); nfsd_master_pid = procp->p_pid; bcopy(procp->p_comm, nfsd_master_comm, MAXCOMLEN + 1); nfsd_master_start = procp->p_stats->p_start; nfsd_master_proc = procp; PROC_UNLOCK(procp); } else if ((uap->flag & NFSSVC_SUSPENDNFSD) != 0) { NFSLOCKV4ROOTMUTEX(); if (suspend_nfsd == 0) { /* Lock out all nfsd threads */ do { igotlock = nfsv4_lock(&nfsd_suspend_lock, 1, NULL, NFSV4ROOTLOCKMUTEXPTR, NULL); } while (igotlock == 0 && suspend_nfsd == 0); suspend_nfsd = 1; } NFSUNLOCKV4ROOTMUTEX(); error = 0; } else if ((uap->flag & NFSSVC_RESUMENFSD) != 0) { NFSLOCKV4ROOTMUTEX(); if (suspend_nfsd != 0) { nfsv4_unlock(&nfsd_suspend_lock, 0); suspend_nfsd = 0; } NFSUNLOCKV4ROOTMUTEX(); error = 0; } NFSEXITCODE(error); return (error); } /* * Check exports. * Returns 0 if ok, 1 otherwise. */ int nfsvno_testexp(struct nfsrv_descript *nd, struct nfsexstuff *exp) { int i; /* * This seems odd, but allow the case where the security flavor * list is empty. This happens when NFSv4 is traversing non-exported * file systems. Exported file systems should always have a non-empty * security flavor list. */ if (exp->nes_numsecflavor == 0) return (0); for (i = 0; i < exp->nes_numsecflavor; i++) { /* * The tests for privacy and integrity must be first, * since ND_GSS is set for everything but AUTH_SYS. */ if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5P && (nd->nd_flag & ND_GSSPRIVACY)) return (0); if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5I && (nd->nd_flag & ND_GSSINTEGRITY)) return (0); if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5 && (nd->nd_flag & ND_GSS)) return (0); if (exp->nes_secflavors[i] == AUTH_SYS && (nd->nd_flag & ND_GSS) == 0) return (0); } return (1); } /* * Calculate a hash value for the fid in a file handle. */ uint32_t nfsrv_hashfh(fhandle_t *fhp) { uint32_t hashval; hashval = hash32_buf(&fhp->fh_fid, sizeof(struct fid), 0); return (hashval); } /* * Signal the userland master nfsd to backup the stable restart file. */ void nfsrv_backupstable(void) { struct proc *procp; if (nfsd_master_proc != NULL) { procp = pfind(nfsd_master_pid); /* Try to make sure it is the correct process. */ if (procp == nfsd_master_proc && procp->p_stats->p_start.tv_sec == nfsd_master_start.tv_sec && procp->p_stats->p_start.tv_usec == nfsd_master_start.tv_usec && strcmp(procp->p_comm, nfsd_master_comm) == 0) kern_psignal(procp, SIGUSR2); else nfsd_master_proc = NULL; if (procp != NULL) PROC_UNLOCK(procp); } } extern int (*nfsd_call_nfsd)(struct thread *, struct nfssvc_args *); /* * Called once to initialize data structures... */ static int nfsd_modevent(module_t mod, int type, void *data) { int error = 0, i; static int loaded = 0; switch (type) { case MOD_LOAD: if (loaded) goto out; newnfs_portinit(); for (i = 0; i < NFSRVCACHE_HASHSIZE; i++) { snprintf(nfsrchash_table[i].lock_name, sizeof(nfsrchash_table[i].lock_name), "nfsrc_tcp%d", i); mtx_init(&nfsrchash_table[i].mtx, nfsrchash_table[i].lock_name, NULL, MTX_DEF); snprintf(nfsrcahash_table[i].lock_name, sizeof(nfsrcahash_table[i].lock_name), "nfsrc_tcpa%d", i); mtx_init(&nfsrcahash_table[i].mtx, nfsrcahash_table[i].lock_name, NULL, MTX_DEF); } mtx_init(&nfsrc_udpmtx, "nfs_udpcache_mutex", NULL, MTX_DEF); mtx_init(&nfs_v4root_mutex, "nfs_v4root_mutex", NULL, MTX_DEF); mtx_init(&nfsv4root_mnt.mnt_mtx, "struct mount mtx", NULL, MTX_DEF); lockinit(&nfsv4root_mnt.mnt_explock, PVFS, "explock", 0, 0); nfsrvd_initcache(); nfsd_init(); NFSD_LOCK(); nfsrvd_init(0); NFSD_UNLOCK(); nfsd_mntinit(); #ifdef VV_DISABLEDELEG vn_deleg_ops.vndeleg_recall = nfsd_recalldelegation; vn_deleg_ops.vndeleg_disable = nfsd_disabledelegation; #endif nfsd_call_servertimer = nfsrv_servertimer; nfsd_call_nfsd = nfssvc_nfsd; loaded = 1; break; case MOD_UNLOAD: if (newnfs_numnfsd != 0) { error = EBUSY; break; } #ifdef VV_DISABLEDELEG vn_deleg_ops.vndeleg_recall = NULL; vn_deleg_ops.vndeleg_disable = NULL; #endif nfsd_call_servertimer = NULL; nfsd_call_nfsd = NULL; /* Clean out all NFSv4 state. */ nfsrv_throwawayallstate(curthread); /* Clean the NFS server reply cache */ nfsrvd_cleancache(); /* Free up the krpc server pool. */ if (nfsrvd_pool != NULL) svcpool_destroy(nfsrvd_pool); /* and get rid of the locks */ for (i = 0; i < NFSRVCACHE_HASHSIZE; i++) { mtx_destroy(&nfsrchash_table[i].mtx); mtx_destroy(&nfsrcahash_table[i].mtx); } mtx_destroy(&nfsrc_udpmtx); mtx_destroy(&nfs_v4root_mutex); mtx_destroy(&nfsv4root_mnt.mnt_mtx); lockdestroy(&nfsv4root_mnt.mnt_explock); loaded = 0; break; default: error = EOPNOTSUPP; break; } out: NFSEXITCODE(error); return (error); } static moduledata_t nfsd_mod = { "nfsd", nfsd_modevent, NULL, }; DECLARE_MODULE(nfsd, nfsd_mod, SI_SUB_VFS, SI_ORDER_ANY); /* So that loader and kldload(2) can find us, wherever we are.. */ MODULE_VERSION(nfsd, 1); MODULE_DEPEND(nfsd, nfscommon, 1, 1, 1); MODULE_DEPEND(nfsd, nfslock, 1, 1, 1); MODULE_DEPEND(nfsd, nfslockd, 1, 1, 1); MODULE_DEPEND(nfsd, krpc, 1, 1, 1); MODULE_DEPEND(nfsd, nfssvc, 1, 1, 1); Index: head/sys/nfsserver/nfs_serv.c =================================================================== --- head/sys/nfsserver/nfs_serv.c (revision 264887) +++ head/sys/nfsserver/nfs_serv.c (revision 264888) @@ -1,3770 +1,3771 @@ /*- * Copyright (c) 1989, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Rick Macklem at The University of Guelph. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)nfs_serv.c 8.8 (Berkeley) 7/31/95 */ #include __FBSDID("$FreeBSD$"); /* * nfs version 2 and 3 server calls to vnode ops * - these routines generally have 3 phases * 1 - break down and validate rpc request in mbuf list * 2 - do the vnode ops for the request * (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c) * 3 - build the rpc reply in an mbuf list * nb: * - do not mix the phases, since the nfsm_?? macros can return failures * on a bad rpc or similar and do not do any vrele() or vput()'s * * - the nfsm_reply() macro generates an nfs rpc reply with the nfs * error number iff error != 0 whereas * returning an error from the server function implies a fatal error * such as a badly constructed rpc request that should be dropped without * a reply. * For nfsm_reply(), the case where error == EBADRPC is treated * specially; after constructing a reply, it does an immediate * `goto nfsmout' to avoid getting any V3 post-op status appended. * * Other notes: * Warning: always pay careful attention to resource cleanup on return * and note that nfsm_*() macros can terminate a procedure on certain * errors. * * lookup() and namei() * may return garbage in various structural fields/return elements * if an error is returned, and may garbage up nd.ni_dvp even if no * error is returned and you did not request LOCKPARENT or WANTPARENT. * * We use the ni_cnd.cn_flags 'HASBUF' flag to track whether the name * buffer has been freed or not. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include FEATURE(nfsserver, "NFS server"); #ifdef NFSRV_DEBUG #define nfsdbprintf(info) printf info #else #define nfsdbprintf(info) #endif #define MAX_COMMIT_COUNT (1024 * 1024) #define MAX_REORDERED_RPC 16 #define NUM_HEURISTIC 1031 #define NHUSE_INIT 64 #define NHUSE_INC 16 #define NHUSE_MAX 2048 static struct nfsheur { struct vnode *nh_vp; /* vp to match (unreferenced pointer) */ off_t nh_nextoff; /* next offset for sequential detection */ int nh_use; /* use count for selection */ int nh_seqcount; /* heuristic */ } nfsheur[NUM_HEURISTIC]; /* Global vars */ int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; int nfsrvw_procrastinate_v3 = 0; static struct timeval nfsver = { 0 }; SYSCTL_NODE(_vfs, OID_AUTO, nfsrv, CTLFLAG_RW, 0, "NFS server"); static int nfs_async; static int nfs_commit_blks; static int nfs_commit_miss; SYSCTL_INT(_vfs_nfsrv, OID_AUTO, async, CTLFLAG_RW, &nfs_async, 0, "Tell client that writes were synced even though they were not"); SYSCTL_INT(_vfs_nfsrv, OID_AUTO, commit_blks, CTLFLAG_RW, &nfs_commit_blks, 0, "Number of completed commits"); SYSCTL_INT(_vfs_nfsrv, OID_AUTO, commit_miss, CTLFLAG_RW, &nfs_commit_miss, 0, ""); struct nfsrvstats nfsrvstats; SYSCTL_STRUCT(_vfs_nfsrv, NFS_NFSRVSTATS, nfsrvstats, CTLFLAG_RW, &nfsrvstats, nfsrvstats, "S,nfsrvstats"); static int nfsrv_access(struct vnode *, accmode_t, struct ucred *, int, int); /* * Clear nameidata fields that are tested in nsfmout cleanup code prior * to using first nfsm macro (that might jump to the cleanup code). */ static __inline void ndclear(struct nameidata *nd) { nd->ni_cnd.cn_flags = 0; nd->ni_vp = NULL; nd->ni_dvp = NULL; nd->ni_startdir = NULL; nd->ni_strictrelative = 0; } /* * Heuristic to detect sequential operation. */ static struct nfsheur * nfsrv_sequential_heuristic(struct uio *uio, struct vnode *vp) { struct nfsheur *nh; int hi, try; /* Locate best candidate. */ try = 32; hi = ((int)(vm_offset_t)vp / sizeof(struct vnode)) % NUM_HEURISTIC; nh = &nfsheur[hi]; while (try--) { if (nfsheur[hi].nh_vp == vp) { nh = &nfsheur[hi]; break; } if (nfsheur[hi].nh_use > 0) --nfsheur[hi].nh_use; hi = (hi + 1) % NUM_HEURISTIC; if (nfsheur[hi].nh_use < nh->nh_use) nh = &nfsheur[hi]; } /* Initialize hint if this is a new file. */ if (nh->nh_vp != vp) { nh->nh_vp = vp; nh->nh_nextoff = uio->uio_offset; nh->nh_use = NHUSE_INIT; if (uio->uio_offset == 0) nh->nh_seqcount = 4; else nh->nh_seqcount = 1; } /* Calculate heuristic. */ if ((uio->uio_offset == 0 && nh->nh_seqcount > 0) || uio->uio_offset == nh->nh_nextoff) { /* See comments in vfs_vnops.c:sequential_heuristic(). */ nh->nh_seqcount += howmany(uio->uio_resid, 16384); if (nh->nh_seqcount > IO_SEQMAX) nh->nh_seqcount = IO_SEQMAX; } else if (qabs(uio->uio_offset - nh->nh_nextoff) <= MAX_REORDERED_RPC * imax(vp->v_mount->mnt_stat.f_iosize, uio->uio_resid)) { /* Probably a reordered RPC, leave seqcount alone. */ } else if (nh->nh_seqcount > 1) { nh->nh_seqcount /= 2; } else { nh->nh_seqcount = 0; } nh->nh_use += NHUSE_INC; if (nh->nh_use > NHUSE_MAX) nh->nh_use = NHUSE_MAX; return (nh); } /* * nfs v3 access service */ int nfsrv3_access(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct vnode *vp = NULL; nfsfh_t nfh; fhandle_t *fhp; u_int32_t *tl; caddr_t bpos; int error = 0, rdonly, getret; struct mbuf *mb, *mreq; struct vattr vattr, *vap = &vattr; u_long testmode, nfsmode; int v3 = (nfsd->nd_flag & ND_NFSV3); nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); if (!v3) panic("nfsrv3_access: v3 proc called on a v2 connection"); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); tl = nfsm_dissect_nonblock(u_int32_t *, NFSX_UNSIGNED); error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(NFSX_UNSIGNED); nfsm_srvpostop_attr(1, NULL); error = 0; goto nfsmout; } nfsmode = fxdr_unsigned(u_int32_t, *tl); if ((nfsmode & NFSV3ACCESS_READ) && nfsrv_access(vp, VREAD, cred, rdonly, 0)) nfsmode &= ~NFSV3ACCESS_READ; if (vp->v_type == VDIR) testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | NFSV3ACCESS_DELETE); else testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); if ((nfsmode & testmode) && nfsrv_access(vp, VWRITE, cred, rdonly, 0)) nfsmode &= ~testmode; if (vp->v_type == VDIR) testmode = NFSV3ACCESS_LOOKUP; else testmode = NFSV3ACCESS_EXECUTE; if ((nfsmode & testmode) && nfsrv_access(vp, VEXEC, cred, rdonly, 0)) nfsmode &= ~testmode; getret = VOP_GETATTR(vp, vap, cred); vput(vp); vp = NULL; nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); nfsm_srvpostop_attr(getret, vap); tl = nfsm_build(u_int32_t *, NFSX_UNSIGNED); *tl = txdr_unsigned(nfsmode); nfsmout: if (vp) vput(vp); return(error); } /* * nfs getattr service */ int nfsrv_getattr(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct nfs_fattr *fp; struct vattr va; struct vattr *vap = &va; struct vnode *vp = NULL; nfsfh_t nfh; fhandle_t *fhp; caddr_t bpos; int error = 0, rdonly; struct mbuf *mb, *mreq; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(0); error = 0; goto nfsmout; } error = VOP_GETATTR(vp, vap, cred); vput(vp); vp = NULL; nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); if (error) { error = 0; goto nfsmout; } fp = nfsm_build(struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); nfsm_srvfillattr(vap, fp); /* fall through */ nfsmout: if (vp) vput(vp); return(error); } /* * nfs setattr service */ int nfsrv_setattr(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct vattr va, preat; struct vattr *vap = &va; struct nfsv2_sattr *sp; struct nfs_fattr *fp; struct vnode *vp = NULL; nfsfh_t nfh; fhandle_t *fhp; u_int32_t *tl; caddr_t bpos; int error = 0, rdonly, preat_ret = 1, postat_ret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; struct mbuf *mb, *mreq; struct timespec guard = { 0, 0 }; struct mount *mp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto out; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ VATTR_NULL(vap); if (v3) { nfsm_srvsattr(vap); tl = nfsm_dissect_nonblock(u_int32_t *, NFSX_UNSIGNED); gcheck = fxdr_unsigned(int, *tl); if (gcheck) { tl = nfsm_dissect_nonblock(u_int32_t *, 2 * NFSX_UNSIGNED); fxdr_nfsv3time(tl, &guard); } } else { sp = nfsm_dissect_nonblock(struct nfsv2_sattr *, NFSX_V2SATTR); /* * Nah nah nah nah na nah * There is a bug in the Sun client that puts 0xffff in the mode * field of sattr when it should put in 0xffffffff. The u_short * doesn't sign extend. * --> check the low order 2 bytes for 0xffff */ if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) vap->va_mode = nfstov_mode(sp->sa_mode); if (sp->sa_uid != nfsrv_nfs_xdrneg1) vap->va_uid = fxdr_unsigned(uid_t, sp->sa_uid); if (sp->sa_gid != nfsrv_nfs_xdrneg1) vap->va_gid = fxdr_unsigned(gid_t, sp->sa_gid); if (sp->sa_size != nfsrv_nfs_xdrneg1) vap->va_size = fxdr_unsigned(u_quad_t, sp->sa_size); if (sp->sa_atime.nfsv2_sec != nfsrv_nfs_xdrneg1) { #ifdef notyet fxdr_nfsv2time(&sp->sa_atime, &vap->va_atime); #else vap->va_atime.tv_sec = fxdr_unsigned(int32_t, sp->sa_atime.nfsv2_sec); vap->va_atime.tv_nsec = 0; #endif } if (sp->sa_mtime.nfsv2_sec != nfsrv_nfs_xdrneg1) fxdr_nfsv2time(&sp->sa_mtime, &vap->va_mtime); } /* * Now that we have all the fields, lets do it. */ error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(2 * NFSX_UNSIGNED); if (v3) nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); error = 0; goto nfsmout; } /* * vp now an active resource, pay careful attention to cleanup */ if (v3) { error = preat_ret = VOP_GETATTR(vp, &preat, cred); if (!error && gcheck && (preat.va_ctime.tv_sec != guard.tv_sec || preat.va_ctime.tv_nsec != guard.tv_nsec)) error = NFSERR_NOT_SYNC; if (error) { vput(vp); vp = NULL; nfsm_reply(NFSX_WCCDATA(v3)); if (v3) nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); error = 0; goto nfsmout; } } /* * If the size is being changed write acces is required, otherwise * just check for a read only filesystem. */ if (vap->va_size == ((u_quad_t)((quad_t) -1))) { if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { error = EROFS; goto out; } } else { if (vp->v_type == VDIR) { error = EISDIR; goto out; } else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly, 0)) != 0) goto out; } error = VOP_SETATTR(vp, vap, cred); postat_ret = VOP_GETATTR(vp, vap, cred); if (!error) error = postat_ret; out: if (vp != NULL) vput(vp); vp = NULL; nfsm_reply(NFSX_WCCORFATTR(v3)); if (v3) { nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); } else if (!error) { /* v2 non-error case. */ fp = nfsm_build(struct nfs_fattr *, NFSX_V2FATTR); nfsm_srvfillattr(vap, fp); } error = 0; /* fall through */ nfsmout: if (vp) vput(vp); vn_finished_write(mp); return(error); } /* * nfs lookup rpc */ int nfsrv_lookup(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct nfs_fattr *fp; struct nameidata nd, ind, *ndp = &nd; struct vnode *vp, *dirp = NULL; nfsfh_t nfh; fhandle_t *fhp; caddr_t bpos; int error = 0, len, dirattr_ret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag; struct mbuf *mb, *mreq; struct vattr va, dirattr, *vap = &va; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); ndclear(&nd); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); nfsm_srvnamesiz(len); pubflag = nfs_ispublicfh(fhp); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = LOOKUP; nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; error = nfs_namei(&nd, nfsd, fhp, len, slp, nam, &md, &dpos, &dirp, v3, &dirattr, &dirattr_ret, pubflag); /* * namei failure, only dirp to cleanup. Clear out garbarge from * structure in case macros jump to nfsmout. */ if (error) { if (dirp) { vrele(dirp); dirp = NULL; } nfsm_reply(NFSX_POSTOPATTR(v3)); if (v3) nfsm_srvpostop_attr(dirattr_ret, &dirattr); error = 0; goto nfsmout; } /* * Locate index file for public filehandle * * error is 0 on entry and 0 on exit from this block. */ if (pubflag) { if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) { /* * Setup call to lookup() to see if we can find * the index file. Arguably, this doesn't belong * in a kernel.. Ugh. If an error occurs, do not * try to install an index file and then clear the * error. * * When we replace nd with ind and redirect ndp, * maintenance of ni_startdir and ni_vp shift to * ind and we have to clean them up in the old nd. * However, the cnd resource continues to be maintained * via the original nd. Confused? You aren't alone! */ ind = nd; VOP_UNLOCK(nd.ni_vp, 0); ind.ni_pathlen = strlen(nfs_pub.np_index); ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf = nfs_pub.np_index; ind.ni_startdir = nd.ni_vp; VREF(ind.ni_startdir); error = lookup(&ind); ind.ni_dvp = NULL; if (error == 0) { /* * Found an index file. Get rid of * the old references. transfer nd.ni_vp' */ if (dirp) vrele(dirp); dirp = nd.ni_vp; nd.ni_vp = NULL; vrele(nd.ni_startdir); nd.ni_startdir = NULL; ndp = &ind; } error = 0; } /* * If the public filehandle was used, check that this lookup * didn't result in a filehandle outside the publicly exported * filesystem. We clear the poor vp here to avoid lockups due * to NFS I/O. */ if (ndp->ni_vp->v_mount != nfs_pub.np_mount) { vput(nd.ni_vp); nd.ni_vp = NULL; error = EPERM; } } /* * Resources at this point: * ndp->ni_vp may not be NULL */ if (error) { nfsm_reply(NFSX_POSTOPATTR(v3)); if (v3) nfsm_srvpostop_attr(dirattr_ret, &dirattr); error = 0; goto nfsmout; } /* * Get underlying attribute, then release remaining resources ( for * the same potential blocking reason ) and reply. */ vp = ndp->ni_vp; bzero((caddr_t)fhp, sizeof(nfh)); fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; error = VOP_VPTOFH(vp, &fhp->fh_fid); if (!error) error = VOP_GETATTR(vp, vap, cred); vput(vp); vrele(ndp->ni_startdir); vrele(dirp); ndp->ni_vp = NULL; ndp->ni_startdir = NULL; dirp = NULL; nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); if (error) { if (v3) nfsm_srvpostop_attr(dirattr_ret, &dirattr); error = 0; goto nfsmout; } nfsm_srvfhtom(fhp, v3); if (v3) { nfsm_srvpostop_attr(0, vap); nfsm_srvpostop_attr(dirattr_ret, &dirattr); } else { fp = nfsm_build(struct nfs_fattr *, NFSX_V2FATTR); nfsm_srvfillattr(vap, fp); } nfsmout: if (ndp->ni_vp || dirp || ndp->ni_startdir) { if (ndp->ni_vp) vput(ndp->ni_vp); if (dirp) vrele(dirp); if (ndp->ni_startdir) vrele(ndp->ni_startdir); } NDFREE(&nd, NDF_ONLY_PNBUF); return (error); } /* * nfs readlink service */ int nfsrv_readlink(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; struct iovec *ivp = iv; struct mbuf *mp; u_int32_t *tl; caddr_t bpos; int error = 0, rdonly, i, tlen, len, getret; int v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mp3, *nmp, *mreq; struct vnode *vp = NULL; struct vattr attr; nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); #ifndef nolint mp = NULL; #endif mp3 = NULL; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); len = 0; i = 0; while (len < NFS_MAXPATHLEN) { MGET(nmp, M_WAITOK, MT_DATA); MCLGET(nmp, M_WAITOK); nmp->m_len = NFSMSIZ(nmp); if (len == 0) mp3 = mp = nmp; else { mp->m_next = nmp; mp = nmp; } if ((len + mp->m_len) > NFS_MAXPATHLEN) { mp->m_len = NFS_MAXPATHLEN - len; len = NFS_MAXPATHLEN; } else len += mp->m_len; ivp->iov_base = mtod(mp, caddr_t); ivp->iov_len = mp->m_len; i++; ivp++; } uiop->uio_iov = iv; uiop->uio_iovcnt = i; uiop->uio_offset = 0; uiop->uio_resid = len; uiop->uio_rw = UIO_READ; uiop->uio_segflg = UIO_SYSSPACE; uiop->uio_td = NULL; error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(2 * NFSX_UNSIGNED); if (v3) nfsm_srvpostop_attr(1, NULL); error = 0; goto nfsmout; } if (vp->v_type != VLNK) { if (v3) error = EINVAL; else error = ENXIO; } else error = VOP_READLINK(vp, uiop, cred); getret = VOP_GETATTR(vp, &attr, cred); vput(vp); vp = NULL; nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); if (v3) nfsm_srvpostop_attr(getret, &attr); if (error) { error = 0; goto nfsmout; } if (uiop->uio_resid > 0) { len -= uiop->uio_resid; tlen = nfsm_rndup(len); nfsm_adj(mp3, NFS_MAXPATHLEN-tlen, tlen-len); } tl = nfsm_build(u_int32_t *, NFSX_UNSIGNED); *tl = txdr_unsigned(len); mb->m_next = mp3; mp3 = NULL; nfsmout: if (mp3) m_freem(mp3); if (vp) vput(vp); return(error); } /* * nfs read service */ int nfsrv_read(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct iovec *iv; struct iovec *iv2; struct mbuf *m; struct nfs_fattr *fp; u_int32_t *tl; int i; caddr_t bpos; int error = 0, rdonly, cnt, len, left, siz, tlen, getret; int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen; struct mbuf *mb, *mreq; struct mbuf *m2; struct vnode *vp = NULL; nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; struct vattr va, *vap = &va; struct nfsheur *nh; off_t off; int ioflag = 0; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if (v3) { tl = nfsm_dissect_nonblock(u_int32_t *, 2 * NFSX_UNSIGNED); off = fxdr_hyper(tl); } else { tl = nfsm_dissect_nonblock(u_int32_t *, NFSX_UNSIGNED); off = (off_t)fxdr_unsigned(u_int32_t, *tl); } nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd)); /* * Reference vp. If an error occurs, vp will be invalid, but we * have to NULL it just in case. The macros might goto nfsmout * as well. */ error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { vp = NULL; nfsm_reply(2 * NFSX_UNSIGNED); if (v3) nfsm_srvpostop_attr(1, NULL); error = 0; goto nfsmout; } if (vp->v_type != VREG) { if (v3) error = EINVAL; else error = (vp->v_type == VDIR) ? EISDIR : EACCES; } if (!error) { if ((error = nfsrv_access(vp, VREAD, cred, rdonly, 1)) != 0) error = nfsrv_access(vp, VEXEC, cred, rdonly, 1); } getret = VOP_GETATTR(vp, vap, cred); if (!error) error = getret; if (error) { vput(vp); vp = NULL; nfsm_reply(NFSX_POSTOPATTR(v3)); if (v3) nfsm_srvpostop_attr(getret, vap); error = 0; goto nfsmout; } /* * Calculate byte count to read */ if (off >= vap->va_size) cnt = 0; else if ((off + reqlen) > vap->va_size) cnt = vap->va_size - off; else cnt = reqlen; nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); if (v3) { tl = nfsm_build(u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); *tl++ = nfsrv_nfs_true; fp = (struct nfs_fattr *)tl; tl += (NFSX_V3FATTR / sizeof (u_int32_t)); } else { tl = nfsm_build(u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED); fp = (struct nfs_fattr *)tl; tl += (NFSX_V2FATTR / sizeof (u_int32_t)); } len = left = nfsm_rndup(cnt); if (cnt > 0) { /* * Generate the mbuf list with the uio_iov ref. to it. */ i = 0; m = m2 = mb; while (left > 0) { siz = min(M_TRAILINGSPACE(m), left); if (siz > 0) { left -= siz; i++; } if (left > 0) { MGET(m, M_WAITOK, MT_DATA); MCLGET(m, M_WAITOK); m->m_len = 0; m2->m_next = m; m2 = m; } } iv = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK); uiop->uio_iov = iv2 = iv; m = mb; left = len; i = 0; while (left > 0) { if (m == NULL) panic("nfsrv_read iov"); siz = min(M_TRAILINGSPACE(m), left); if (siz > 0) { iv->iov_base = mtod(m, caddr_t) + m->m_len; iv->iov_len = siz; m->m_len += siz; left -= siz; iv++; i++; } m = m->m_next; } uiop->uio_iovcnt = i; uiop->uio_offset = off; uiop->uio_resid = len; uiop->uio_rw = UIO_READ; uiop->uio_segflg = UIO_SYSSPACE; + uiop->uio_td = NULL; nh = nfsrv_sequential_heuristic(uiop, vp); ioflag |= nh->nh_seqcount << IO_SEQSHIFT; error = VOP_READ(vp, uiop, IO_NODELOCKED | ioflag, cred); if (error == 0) nh->nh_nextoff = uiop->uio_offset; free((caddr_t)iv2, M_TEMP); if (error || (getret = VOP_GETATTR(vp, vap, cred))) { if (!error) error = getret; m_freem(mreq); vput(vp); vp = NULL; nfsm_reply(NFSX_POSTOPATTR(v3)); if (v3) nfsm_srvpostop_attr(getret, vap); error = 0; goto nfsmout; } } else uiop->uio_resid = 0; vput(vp); vp = NULL; nfsm_srvfillattr(vap, fp); tlen = len - uiop->uio_resid; cnt = cnt < tlen ? cnt : tlen; tlen = nfsm_rndup(cnt); if (len != tlen || tlen != cnt) nfsm_adj(mb, len - tlen, tlen - cnt); if (v3) { *tl++ = txdr_unsigned(cnt); if (cnt < reqlen) *tl++ = nfsrv_nfs_true; else *tl++ = nfsrv_nfs_false; } *tl = txdr_unsigned(cnt); nfsmout: if (vp) vput(vp); return(error); } /* * nfs write service */ int nfsrv_write(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct iovec *ivp; int i, cnt; struct mbuf *mp; struct nfs_fattr *fp; struct iovec *iv; struct vattr va, forat; struct vattr *vap = &va; u_int32_t *tl; caddr_t bpos; int error = 0, rdonly, len, forat_ret = 1; int ioflags, aftat_ret = 1, retlen = 0, zeroing, adjust; int stable = NFSV3WRITE_FILESYNC; int v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mreq; struct vnode *vp = NULL; struct nfsheur *nh; nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; off_t off; struct mount *mntp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); if (mrep == NULL) { *mrq = NULL; error = 0; goto nfsmout; } fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto ereply; } (void) vn_start_write(NULL, &mntp, V_WAIT); vfs_rel(mntp); /* The write holds a ref. */ if (v3) { tl = nfsm_dissect_nonblock(u_int32_t *, 5 * NFSX_UNSIGNED); off = fxdr_hyper(tl); tl += 3; stable = fxdr_unsigned(int, *tl++); } else { tl = nfsm_dissect_nonblock(u_int32_t *, 4 * NFSX_UNSIGNED); off = (off_t)fxdr_unsigned(u_int32_t, *++tl); tl += 2; if (nfs_async) stable = NFSV3WRITE_UNSTABLE; } retlen = len = fxdr_unsigned(int32_t, *tl); cnt = i = 0; /* * For NFS Version 2, it is not obvious what a write of zero length * should do, but I might as well be consistent with Version 3, * which is to return ok so long as there are no permission problems. */ if (len > 0) { zeroing = 1; mp = mrep; while (mp) { if (mp == md) { zeroing = 0; adjust = dpos - mtod(mp, caddr_t); mp->m_len -= adjust; if (mp->m_len > 0 && adjust > 0) mp->m_data += adjust; } if (zeroing) mp->m_len = 0; else if (mp->m_len > 0) { i += mp->m_len; if (i > len) { mp->m_len -= (i - len); zeroing = 1; } if (mp->m_len > 0) cnt++; } mp = mp->m_next; } } if (len > NFS_MAXDATA || len < 0 || i < len) { error = EIO; nfsm_reply(2 * NFSX_UNSIGNED); if (v3) nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); error = 0; goto nfsmout; } error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { vp = NULL; nfsm_reply(2 * NFSX_UNSIGNED); if (v3) nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); error = 0; goto nfsmout; } if (v3) forat_ret = VOP_GETATTR(vp, &forat, cred); if (vp->v_type != VREG) { if (v3) error = EINVAL; else error = (vp->v_type == VDIR) ? EISDIR : EACCES; } if (!error) error = nfsrv_access(vp, VWRITE, cred, rdonly, 1); if (error) { vput(vp); vp = NULL; nfsm_reply(NFSX_WCCDATA(v3)); if (v3) nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); error = 0; goto nfsmout; } if (len > 0) { ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK); uiop->uio_iov = iv = ivp; uiop->uio_iovcnt = cnt; mp = mrep; while (mp) { if (mp->m_len > 0) { ivp->iov_base = mtod(mp, caddr_t); ivp->iov_len = mp->m_len; ivp++; } mp = mp->m_next; } /* * XXX * The IO_METASYNC flag indicates that all metadata (and not just * enough to ensure data integrity) mus be written to stable storage * synchronously. * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) */ if (stable == NFSV3WRITE_UNSTABLE) ioflags = IO_NODELOCKED; else if (stable == NFSV3WRITE_DATASYNC) ioflags = (IO_SYNC | IO_NODELOCKED); else ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); uiop->uio_resid = len; uiop->uio_rw = UIO_WRITE; uiop->uio_segflg = UIO_SYSSPACE; uiop->uio_td = NULL; uiop->uio_offset = off; nh = nfsrv_sequential_heuristic(uiop, vp); ioflags |= nh->nh_seqcount << IO_SEQSHIFT; error = VOP_WRITE(vp, uiop, ioflags, cred); if (error == 0) nh->nh_nextoff = uiop->uio_offset; /* Unlocked write. */ nfsrvstats.srvvop_writes++; free((caddr_t)iv, M_TEMP); } aftat_ret = VOP_GETATTR(vp, vap, cred); vput(vp); vp = NULL; if (!error) error = aftat_ret; ereply: nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); if (v3) { nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); if (error) { error = 0; goto nfsmout; } tl = nfsm_build(u_int32_t *, 4 * NFSX_UNSIGNED); *tl++ = txdr_unsigned(retlen); /* * If nfs_async is set, then pretend the write was FILESYNC. */ if (stable == NFSV3WRITE_UNSTABLE && !nfs_async) *tl++ = txdr_unsigned(stable); else *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); /* * Actually, there is no need to txdr these fields, * but it may make the values more human readable, * for debugging purposes. */ if (nfsver.tv_sec == 0) nfsver = boottime; *tl++ = txdr_unsigned(nfsver.tv_sec); *tl = txdr_unsigned(nfsver.tv_usec); } else if (!error) { /* v2 non-error case. */ fp = nfsm_build(struct nfs_fattr *, NFSX_V2FATTR); nfsm_srvfillattr(vap, fp); } error = 0; nfsmout: if (vp) vput(vp); vn_finished_write(mntp); return(error); } /* * nfs create service * now does a truncate to 0 length via. setattr if it already exists */ int nfsrv_create(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct nfs_fattr *fp; struct vattr va, dirfor, diraft; struct vattr *vap = &va; struct nfsv2_sattr *sp; u_int32_t *tl; struct nameidata nd; caddr_t bpos; int error = 0, rdev, len, tsize, dirfor_ret = 1, diraft_ret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; struct mbuf *mb, *mreq; struct vnode *dirp = NULL; nfsfh_t nfh; fhandle_t *fhp; u_quad_t tempsize; struct timespec cverf; struct mount *mp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); #ifndef nolint rdev = 0; #endif ndclear(&nd); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto ereply; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART; /* * Call namei and do initial cleanup to get a few things * out of the way. If we get an initial error we cleanup * and return here to avoid special-casing the invalid nd * structure through the rest of the case. dirp may be * set even if an error occurs, but the nd structure will not * be valid at all if an error occurs so we have to invalidate it * prior to calling nfsm_reply ( which might goto nfsmout ). */ error = nfs_namei(&nd, nfsd, fhp, len, slp, nam, &md, &dpos, &dirp, v3, &dirfor, &dirfor_ret, FALSE); if (dirp && !v3) { vrele(dirp); dirp = NULL; } if (error) { nfsm_reply(NFSX_WCCDATA(v3)); if (v3) nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); error = 0; goto nfsmout; } /* * No error. Continue. State: * * startdir is valid ( we release this immediately ) * dirp may be valid * nd.ni_vp may be valid * nd.ni_dvp is valid * * The error state is set through the code and we may also do some * opportunistic releasing of vnodes to avoid holding locks through * NFS I/O. The cleanup at the end is a catch-all */ VATTR_NULL(vap); if (v3) { tl = nfsm_dissect_nonblock(u_int32_t *, NFSX_UNSIGNED); how = fxdr_unsigned(int, *tl); switch (how) { case NFSV3CREATE_GUARDED: if (nd.ni_vp) { error = EEXIST; break; } /* fall through */ case NFSV3CREATE_UNCHECKED: nfsm_srvsattr(vap); break; case NFSV3CREATE_EXCLUSIVE: tl = nfsm_dissect_nonblock(u_int32_t *, NFSX_V3CREATEVERF); /* Unique bytes, endianness is not important. */ cverf.tv_sec = (int32_t)tl[0]; cverf.tv_nsec = tl[1]; exclusive_flag = 1; break; }; vap->va_type = VREG; } else { sp = nfsm_dissect_nonblock(struct nfsv2_sattr *, NFSX_V2SATTR); vap->va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode)); if (vap->va_type == VNON) vap->va_type = VREG; vap->va_mode = nfstov_mode(sp->sa_mode); switch (vap->va_type) { case VREG: tsize = fxdr_unsigned(int32_t, sp->sa_size); if (tsize != -1) vap->va_size = (u_quad_t)tsize; break; case VCHR: case VBLK: case VFIFO: rdev = fxdr_unsigned(long, sp->sa_size); break; default: break; }; } /* * Iff doesn't exist, create it * otherwise just truncate to 0 length * should I set the mode too ? * * The only possible error we can have at this point is EEXIST. * nd.ni_vp will also be non-NULL in that case. */ if (nd.ni_vp == NULL) { if (vap->va_mode == (mode_t)VNOVAL) vap->va_mode = 0; if (vap->va_type == VREG || vap->va_type == VSOCK) { error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); if (error) NDFREE(&nd, NDF_ONLY_PNBUF); else { if (exclusive_flag) { exclusive_flag = 0; VATTR_NULL(vap); vap->va_atime = cverf; error = VOP_SETATTR(nd.ni_vp, vap, cred); } } } else if (vap->va_type == VCHR || vap->va_type == VBLK || vap->va_type == VFIFO) { /* * NFSv2-specific code for creating device nodes * and fifos. * * Handle SysV FIFO node special cases. All other * devices require super user to access. */ if (vap->va_type == VCHR && rdev == 0xffffffff) vap->va_type = VFIFO; if (vap->va_type != VFIFO && (error = priv_check_cred(cred, PRIV_VFS_MKNOD_DEV, 0))) { goto ereply; } vap->va_rdev = rdev; error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); if (error) { NDFREE(&nd, NDF_ONLY_PNBUF); goto ereply; } vput(nd.ni_vp); nd.ni_vp = NULL; /* * release dvp prior to lookup */ vput(nd.ni_dvp); nd.ni_dvp = NULL; /* * Setup for lookup. * * Even though LOCKPARENT was cleared, ni_dvp may * be garbage. */ nd.ni_cnd.cn_nameiop = LOOKUP; nd.ni_cnd.cn_flags &= ~(LOCKPARENT); nd.ni_cnd.cn_thread = curthread; nd.ni_cnd.cn_cred = cred; error = lookup(&nd); nd.ni_dvp = NULL; if (error) goto ereply; if (nd.ni_cnd.cn_flags & ISSYMLINK) { error = EINVAL; goto ereply; } } else { error = ENXIO; } } else { if (vap->va_size != -1) { error = nfsrv_access(nd.ni_vp, VWRITE, cred, (nd.ni_cnd.cn_flags & RDONLY), 0); if (!error) { tempsize = vap->va_size; VATTR_NULL(vap); vap->va_size = tempsize; error = VOP_SETATTR(nd.ni_vp, vap, cred); } } } if (!error) { bzero((caddr_t)fhp, sizeof(nfh)); fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid; error = VOP_VPTOFH(nd.ni_vp, &fhp->fh_fid); if (!error) error = VOP_GETATTR(nd.ni_vp, vap, cred); } if (v3) { if (exclusive_flag && !error && bcmp(&cverf, &vap->va_atime, sizeof (cverf))) error = EEXIST; if (dirp == nd.ni_dvp) diraft_ret = VOP_GETATTR(dirp, &diraft, cred); else { /* Drop the other locks to avoid deadlock. */ if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vput(nd.ni_vp); nd.ni_dvp = NULL; nd.ni_vp = NULL; vn_lock(dirp, LK_EXCLUSIVE | LK_RETRY); diraft_ret = VOP_GETATTR(dirp, &diraft, cred); VOP_UNLOCK(dirp, 0); } } ereply: nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); if (v3) { if (!error) { nfsm_srvpostop_fh(fhp); nfsm_srvpostop_attr(0, vap); } nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); } else if (!error) { /* v2 non-error case. */ nfsm_srvfhtom(fhp, v3); fp = nfsm_build(struct nfs_fattr *, NFSX_V2FATTR); nfsm_srvfillattr(vap, fp); } error = 0; nfsmout: if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vput(nd.ni_vp); if (nd.ni_startdir) { vrele(nd.ni_startdir); nd.ni_startdir = NULL; } if (dirp) vrele(dirp); NDFREE(&nd, NDF_ONLY_PNBUF); vn_finished_write(mp); return (error); } /* * nfs v3 mknod service */ int nfsrv_mknod(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct vattr va, dirfor, diraft; struct vattr *vap = &va; struct thread *td = curthread; u_int32_t *tl; struct nameidata nd; caddr_t bpos; int error = 0, len, dirfor_ret = 1, diraft_ret = 1; u_int32_t major, minor; enum vtype vtyp; struct mbuf *mb, *mreq; struct vnode *vp, *dirp = NULL; nfsfh_t nfh; fhandle_t *fhp; struct mount *mp = NULL; int v3 = (nfsd->nd_flag & ND_NFSV3); nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); if (!v3) panic("nfsrv_mknod: v3 proc called on a v2 connection"); ndclear(&nd); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto ereply; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART; /* * Handle nfs_namei() call. If an error occurs, the nd structure * is not valid. However, nfsm_*() routines may still jump to * nfsmout. */ error = nfs_namei(&nd, nfsd, fhp, len, slp, nam, &md, &dpos, &dirp, v3, &dirfor, &dirfor_ret, FALSE); if (error) { nfsm_reply(NFSX_WCCDATA(1)); nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); error = 0; goto nfsmout; } tl = nfsm_dissect_nonblock(u_int32_t *, NFSX_UNSIGNED); vtyp = nfsv3tov_type(*tl); if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { error = NFSERR_BADTYPE; goto out; } VATTR_NULL(vap); nfsm_srvsattr(vap); if (vtyp == VCHR || vtyp == VBLK) { tl = nfsm_dissect_nonblock(u_int32_t *, 2 * NFSX_UNSIGNED); major = fxdr_unsigned(u_int32_t, *tl++); minor = fxdr_unsigned(u_int32_t, *tl); vap->va_rdev = makedev(major, minor); } /* * Iff doesn't exist, create it. */ if (nd.ni_vp) { error = EEXIST; goto out; } vap->va_type = vtyp; if (vap->va_mode == (mode_t)VNOVAL) vap->va_mode = 0; if (vtyp == VSOCK) { vrele(nd.ni_startdir); nd.ni_startdir = NULL; error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); if (error) NDFREE(&nd, NDF_ONLY_PNBUF); } else { if (vtyp != VFIFO && (error = priv_check_cred(cred, PRIV_VFS_MKNOD_DEV, 0))) goto out; error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); if (error) { NDFREE(&nd, NDF_ONLY_PNBUF); goto out; } vput(nd.ni_vp); nd.ni_vp = NULL; /* * Release dvp prior to lookup */ vput(nd.ni_dvp); nd.ni_dvp = NULL; nd.ni_cnd.cn_nameiop = LOOKUP; nd.ni_cnd.cn_flags &= ~(LOCKPARENT); nd.ni_cnd.cn_thread = td; nd.ni_cnd.cn_cred = td->td_ucred; error = lookup(&nd); nd.ni_dvp = NULL; if (error) goto out; if (nd.ni_cnd.cn_flags & ISSYMLINK) error = EINVAL; } /* * send response, cleanup, return. */ out: vp = nd.ni_vp; if (!error) { bzero((caddr_t)fhp, sizeof(nfh)); fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; error = VOP_VPTOFH(vp, &fhp->fh_fid); if (!error) error = VOP_GETATTR(vp, vap, cred); } if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); nd.ni_dvp = NULL; } if (vp) { vput(vp); vp = NULL; nd.ni_vp = NULL; } if (nd.ni_startdir) { vrele(nd.ni_startdir); nd.ni_startdir = NULL; } NDFREE(&nd, NDF_ONLY_PNBUF); if (dirp) { vn_lock(dirp, LK_EXCLUSIVE | LK_RETRY); diraft_ret = VOP_GETATTR(dirp, &diraft, cred); vput(dirp); } ereply: nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); if (v3) { if (!error) { nfsm_srvpostop_fh(fhp); nfsm_srvpostop_attr(0, vap); } nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); } vn_finished_write(mp); return (0); nfsmout: if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vput(nd.ni_vp); if (dirp) vrele(dirp); if (nd.ni_startdir) vrele(nd.ni_startdir); NDFREE(&nd, NDF_ONLY_PNBUF); vn_finished_write(mp); return (error); } /* * nfs remove service */ int nfsrv_remove(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct nameidata nd; caddr_t bpos; int error = 0, len, dirfor_ret = 1, diraft_ret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mreq; struct vnode *dirp; struct vattr dirfor, diraft; nfsfh_t nfh; fhandle_t *fhp; struct mount *mp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); ndclear(&nd); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto ereply; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = DELETE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; error = nfs_namei(&nd, nfsd, fhp, len, slp, nam, &md, &dpos, &dirp, v3, &dirfor, &dirfor_ret, FALSE); if (dirp && !v3) { vrele(dirp); dirp = NULL; } if (error == 0) { if (nd.ni_vp->v_type == VDIR) { error = EPERM; /* POSIX */ goto out; } /* * The root of a mounted filesystem cannot be deleted. */ if (nd.ni_vp->v_vflag & VV_ROOT) { error = EBUSY; goto out; } out: if (!error) { error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); NDFREE(&nd, NDF_ONLY_PNBUF); } } if (dirp && v3) { if (dirp == nd.ni_dvp) diraft_ret = VOP_GETATTR(dirp, &diraft, cred); else { /* Drop the other locks to avoid deadlock. */ if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vput(nd.ni_vp); nd.ni_dvp = NULL; nd.ni_vp = NULL; vn_lock(dirp, LK_EXCLUSIVE | LK_RETRY); diraft_ret = VOP_GETATTR(dirp, &diraft, cred); VOP_UNLOCK(dirp, 0); } vrele(dirp); dirp = NULL; } ereply: nfsm_reply(NFSX_WCCDATA(v3)); if (v3) nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); error = 0; nfsmout: NDFREE(&nd, NDF_ONLY_PNBUF); if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vput(nd.ni_vp); vn_finished_write(mp); return(error); } /* * nfs rename service */ int nfsrv_rename(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; caddr_t bpos; int error = 0, len, len2, fdirfor_ret = 1, fdiraft_ret = 1; int tdirfor_ret = 1, tdiraft_ret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mreq; struct nameidata fromnd, tond; struct vnode *fvp, *tvp, *tdvp, *fdirp = NULL; struct vnode *tdirp = NULL; struct vattr fdirfor, fdiraft, tdirfor, tdiraft; nfsfh_t fnfh, tnfh; fhandle_t *ffhp, *tfhp; uid_t saved_uid; struct mount *mp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); #ifndef nolint fvp = NULL; #endif ffhp = &fnfh.fh_generic; tfhp = &tnfh.fh_generic; /* * Clear fields incase goto nfsmout occurs from macro. */ ndclear(&fromnd); ndclear(&tond); nfsm_srvmtofh(ffhp); if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL) { error = ESTALE; goto out1; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ nfsm_srvnamesiz(len); /* * Remember our original uid so that we can reset cr_uid before * the second nfs_namei() call, in case it is remapped. */ saved_uid = cred->cr_uid; fromnd.ni_cnd.cn_cred = cred; fromnd.ni_cnd.cn_nameiop = DELETE; fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; error = nfs_namei(&fromnd, nfsd, ffhp, len, slp, nam, &md, &dpos, &fdirp, v3, &fdirfor, &fdirfor_ret, FALSE); if (fdirp && !v3) { vrele(fdirp); fdirp = NULL; } if (error) { nfsm_reply(2 * NFSX_WCCDATA(v3)); if (v3) { nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); } error = 0; goto nfsmout; } fvp = fromnd.ni_vp; nfsm_srvmtofh(tfhp); nfsm_srvnamesiz(len2); cred->cr_uid = saved_uid; tond.ni_cnd.cn_cred = cred; tond.ni_cnd.cn_nameiop = RENAME; tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; error = nfs_namei(&tond, nfsd, tfhp, len2, slp, nam, &md, &dpos, &tdirp, v3, &tdirfor, &tdirfor_ret, FALSE); if (tdirp && !v3) { vrele(tdirp); tdirp = NULL; } if (error) goto out1; tdvp = tond.ni_dvp; tvp = tond.ni_vp; if (tvp != NULL) { if (fvp->v_type == VDIR && tvp->v_type != VDIR) { if (v3) error = EEXIST; else error = EISDIR; goto out; } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { if (v3) error = EEXIST; else error = ENOTDIR; goto out; } if (tvp->v_type == VDIR && tvp->v_mountedhere) { if (v3) error = EXDEV; else error = ENOTEMPTY; goto out; } } if (fvp->v_type == VDIR && fvp->v_mountedhere) { if (v3) error = EXDEV; else error = ENOTEMPTY; goto out; } if (fvp->v_mount != tdvp->v_mount) { if (v3) error = EXDEV; else error = ENOTEMPTY; goto out; } if (fvp == tdvp) { if (v3) error = EINVAL; else error = ENOTEMPTY; } /* * If source is the same as the destination (that is the * same vnode with the same name in the same directory), * then there is nothing to do. */ if (fvp == tvp && fromnd.ni_dvp == tdvp && fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && !bcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen)) error = -1; out: if (!error) { /* * The VOP_RENAME function releases all vnode references & * locks prior to returning so we need to clear the pointers * to bypass cleanup code later on. */ error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); fromnd.ni_dvp = NULL; fromnd.ni_vp = NULL; tond.ni_dvp = NULL; tond.ni_vp = NULL; if (error) { NDFREE(&fromnd, NDF_ONLY_PNBUF); NDFREE(&tond, NDF_ONLY_PNBUF); } } else { if (error == -1) error = 0; } /* fall through */ out1: nfsm_reply(2 * NFSX_WCCDATA(v3)); if (v3) { /* Release existing locks to prevent deadlock. */ if (tond.ni_dvp) { if (tond.ni_dvp == tond.ni_vp) vrele(tond.ni_dvp); else vput(tond.ni_dvp); } if (tond.ni_vp) vput(tond.ni_vp); tond.ni_dvp = NULL; tond.ni_vp = NULL; if (fdirp) { vn_lock(fdirp, LK_EXCLUSIVE | LK_RETRY); fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred); VOP_UNLOCK(fdirp, 0); } if (tdirp) { vn_lock(tdirp, LK_EXCLUSIVE | LK_RETRY); tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred); VOP_UNLOCK(tdirp, 0); } nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); } error = 0; /* fall through */ nfsmout: /* * Clear out tond related fields */ if (tond.ni_dvp) { if (tond.ni_dvp == tond.ni_vp) vrele(tond.ni_dvp); else vput(tond.ni_dvp); } if (tond.ni_vp) vput(tond.ni_vp); if (tdirp) vrele(tdirp); if (tond.ni_startdir) vrele(tond.ni_startdir); NDFREE(&tond, NDF_ONLY_PNBUF); /* * Clear out fromnd related fields */ if (fdirp) vrele(fdirp); if (fromnd.ni_startdir) vrele(fromnd.ni_startdir); NDFREE(&fromnd, NDF_ONLY_PNBUF); if (fromnd.ni_dvp) vrele(fromnd.ni_dvp); if (fromnd.ni_vp) vrele(fromnd.ni_vp); vn_finished_write(mp); return (error); } /* * nfs link service */ int nfsrv_link(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct nameidata nd; caddr_t bpos; int error = 0, rdonly, len, dirfor_ret = 1, diraft_ret = 1; int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mreq; struct vnode *vp = NULL, *xp, *dirp = NULL; struct vattr dirfor, diraft, at; nfsfh_t nfh, dnfh; fhandle_t *fhp, *dfhp; struct mount *mp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); ndclear(&nd); fhp = &nfh.fh_generic; dfhp = &dnfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto ereply; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ nfsm_srvmtofh(dfhp); nfsm_srvnamesiz(len); error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); if (v3) { nfsm_srvpostop_attr(getret, &at); nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); } vp = NULL; error = 0; goto nfsmout; } if (v3) getret = VOP_GETATTR(vp, &at, cred); if (vp->v_type == VDIR) { error = EPERM; /* POSIX */ goto out1; } VOP_UNLOCK(vp, 0); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT; error = nfs_namei(&nd, nfsd, dfhp, len, slp, nam, &md, &dpos, &dirp, v3, &dirfor, &dirfor_ret, FALSE); if (dirp && !v3) { vrele(dirp); dirp = NULL; } if (error) { vrele(vp); vp = NULL; goto out2; } xp = nd.ni_vp; if (xp != NULL) { error = EEXIST; vrele(vp); vp = NULL; goto out2; } xp = nd.ni_dvp; if (vp->v_mount != xp->v_mount) { error = EXDEV; vrele(vp); vp = NULL; goto out2; } vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); NDFREE(&nd, NDF_ONLY_PNBUF); /* fall through */ out1: if (v3) getret = VOP_GETATTR(vp, &at, cred); out2: if (dirp) { if (dirp == nd.ni_dvp) diraft_ret = VOP_GETATTR(dirp, &diraft, cred); else { /* Release existing locks to prevent deadlock. */ if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vrele(nd.ni_vp); nd.ni_dvp = NULL; nd.ni_vp = NULL; vn_lock(dirp, LK_EXCLUSIVE | LK_RETRY); diraft_ret = VOP_GETATTR(dirp, &diraft, cred); VOP_UNLOCK(dirp, 0); } } ereply: nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); if (v3) { nfsm_srvpostop_attr(getret, &at); nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); } error = 0; /* fall through */ nfsmout: NDFREE(&nd, NDF_ONLY_PNBUF); if (vp) vput(vp); if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (dirp) vrele(dirp); if (nd.ni_vp) vrele(nd.ni_vp); vn_finished_write(mp); return(error); } /* * nfs symbolic link service */ int nfsrv_symlink(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct vattr va, dirfor, diraft; struct nameidata nd; struct vattr *vap = &va; struct nfsv2_sattr *sp; char *bpos, *pathcp = NULL; struct uio io; struct iovec iv; int error = 0, len, len2, dirfor_ret = 1, diraft_ret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mreq; struct vnode *dirp = NULL; nfsfh_t nfh; fhandle_t *fhp; struct mount *mp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); ndclear(&nd); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto out; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; error = nfs_namei(&nd, nfsd, fhp, len, slp, nam, &md, &dpos, &dirp, v3, &dirfor, &dirfor_ret, FALSE); if (error == 0) { VATTR_NULL(vap); if (v3) nfsm_srvsattr(vap); nfsm_srvpathsiz(len2); } if (dirp && !v3) { vrele(dirp); dirp = NULL; } if (error) goto out; pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK); iv.iov_base = pathcp; iv.iov_len = len2; io.uio_resid = len2; io.uio_offset = 0; io.uio_iov = &iv; io.uio_iovcnt = 1; io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_td = NULL; nfsm_mtouio(&io, len2); if (!v3) { sp = nfsm_dissect_nonblock(struct nfsv2_sattr *, NFSX_V2SATTR); vap->va_mode = nfstov_mode(sp->sa_mode); } *(pathcp + len2) = '\0'; if (nd.ni_vp) { error = EEXIST; goto out; } /* * issue symlink op. SAVESTART is set so the underlying path component * is only freed by the VOP if an error occurs. */ if (vap->va_mode == (mode_t)VNOVAL) vap->va_mode = 0; error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap, pathcp); if (error) NDFREE(&nd, NDF_ONLY_PNBUF); else vput(nd.ni_vp); nd.ni_vp = NULL; /* * releases directory prior to potential lookup op. */ vput(nd.ni_dvp); nd.ni_dvp = NULL; if (error == 0) { if (v3) { /* * Issue lookup. Leave SAVESTART set so we can easily free * the name buffer later on. * * since LOCKPARENT is not set, ni_dvp will be garbage on * return whether an error occurs or not. */ nd.ni_cnd.cn_nameiop = LOOKUP; nd.ni_cnd.cn_flags &= ~(LOCKPARENT | FOLLOW); nd.ni_cnd.cn_flags |= (NOFOLLOW | LOCKLEAF); nd.ni_cnd.cn_thread = curthread; nd.ni_cnd.cn_cred = cred; error = lookup(&nd); nd.ni_dvp = NULL; if (error == 0) { bzero((caddr_t)fhp, sizeof(nfh)); fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid; error = VOP_VPTOFH(nd.ni_vp, &fhp->fh_fid); if (!error) error = VOP_GETATTR(nd.ni_vp, vap, cred); vput(nd.ni_vp); nd.ni_vp = NULL; } } } out: /* * These releases aren't strictly required, does even doing them * make any sense? XXX can nfsm_reply() block? */ if (pathcp) { free(pathcp, M_TEMP); pathcp = NULL; } if (dirp) { vn_lock(dirp, LK_EXCLUSIVE | LK_RETRY); diraft_ret = VOP_GETATTR(dirp, &diraft, cred); VOP_UNLOCK(dirp, 0); } if (nd.ni_startdir) { vrele(nd.ni_startdir); nd.ni_startdir = NULL; } nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); if (v3) { if (!error) { nfsm_srvpostop_fh(fhp); nfsm_srvpostop_attr(0, vap); } nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); } error = 0; /* fall through */ nfsmout: NDFREE(&nd, NDF_ONLY_PNBUF); if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vrele(nd.ni_vp); if (nd.ni_startdir) vrele(nd.ni_startdir); if (dirp) vrele(dirp); if (pathcp) free(pathcp, M_TEMP); vn_finished_write(mp); return (error); } /* * nfs mkdir service */ int nfsrv_mkdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct vattr va, dirfor, diraft; struct vattr *vap = &va; struct nfs_fattr *fp; struct nameidata nd; u_int32_t *tl; caddr_t bpos; int error = 0, len, dirfor_ret = 1, diraft_ret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mreq; struct vnode *dirp = NULL; int vpexcl = 0; nfsfh_t nfh; fhandle_t *fhp; struct mount *mp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); ndclear(&nd); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto out; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT; error = nfs_namei(&nd, nfsd, fhp, len, slp, nam, &md, &dpos, &dirp, v3, &dirfor, &dirfor_ret, FALSE); if (dirp && !v3) { vrele(dirp); dirp = NULL; } if (error) { nfsm_reply(NFSX_WCCDATA(v3)); if (v3) nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); error = 0; goto nfsmout; } VATTR_NULL(vap); if (v3) { nfsm_srvsattr(vap); } else { tl = nfsm_dissect_nonblock(u_int32_t *, NFSX_UNSIGNED); vap->va_mode = nfstov_mode(*tl++); } /* * At this point nd.ni_dvp is referenced and exclusively locked and * nd.ni_vp, if it exists, is referenced but not locked. */ vap->va_type = VDIR; if (nd.ni_vp != NULL) { NDFREE(&nd, NDF_ONLY_PNBUF); error = EEXIST; goto out; } /* * Issue mkdir op. Since SAVESTART is not set, the pathname * component is freed by the VOP call. This will fill-in * nd.ni_vp, reference, and exclusively lock it. */ if (vap->va_mode == (mode_t)VNOVAL) vap->va_mode = 0; error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); NDFREE(&nd, NDF_ONLY_PNBUF); vpexcl = 1; vput(nd.ni_dvp); nd.ni_dvp = NULL; if (!error) { bzero((caddr_t)fhp, sizeof(nfh)); fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid; error = VOP_VPTOFH(nd.ni_vp, &fhp->fh_fid); if (!error) error = VOP_GETATTR(nd.ni_vp, vap, cred); } out: if (dirp) { if (dirp == nd.ni_dvp) { diraft_ret = VOP_GETATTR(dirp, &diraft, cred); } else { /* Release existing locks to prevent deadlock. */ if (nd.ni_dvp) { NDFREE(&nd, NDF_ONLY_PNBUF); if (nd.ni_dvp == nd.ni_vp && vpexcl) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) { if (vpexcl) vput(nd.ni_vp); else vrele(nd.ni_vp); } nd.ni_dvp = NULL; nd.ni_vp = NULL; vn_lock(dirp, LK_EXCLUSIVE | LK_RETRY); diraft_ret = VOP_GETATTR(dirp, &diraft, cred); VOP_UNLOCK(dirp, 0); } } nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); if (v3) { if (!error) { nfsm_srvpostop_fh(fhp); nfsm_srvpostop_attr(0, vap); } nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); } else if (!error) { /* v2 non-error case. */ nfsm_srvfhtom(fhp, v3); fp = nfsm_build(struct nfs_fattr *, NFSX_V2FATTR); nfsm_srvfillattr(vap, fp); } error = 0; /* fall through */ nfsmout: if (nd.ni_dvp) { NDFREE(&nd, NDF_ONLY_PNBUF); if (nd.ni_dvp == nd.ni_vp && vpexcl) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) { if (vpexcl) vput(nd.ni_vp); else vrele(nd.ni_vp); } if (dirp) vrele(dirp); vn_finished_write(mp); return (error); } /* * nfs rmdir service */ int nfsrv_rmdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; caddr_t bpos; int error = 0, len, dirfor_ret = 1, diraft_ret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mreq; struct vnode *vp, *dirp = NULL; struct vattr dirfor, diraft; nfsfh_t nfh; fhandle_t *fhp; struct nameidata nd; struct mount *mp = NULL; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); ndclear(&nd); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto out; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = DELETE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; error = nfs_namei(&nd, nfsd, fhp, len, slp, nam, &md, &dpos, &dirp, v3, &dirfor, &dirfor_ret, FALSE); if (dirp && !v3) { vrele(dirp); dirp = NULL; } if (error) { nfsm_reply(NFSX_WCCDATA(v3)); if (v3) nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); error = 0; goto nfsmout; } vp = nd.ni_vp; if (vp->v_type != VDIR) { error = ENOTDIR; goto out; } /* * No rmdir "." please. */ if (nd.ni_dvp == vp) { error = EINVAL; goto out; } /* * The root of a mounted filesystem cannot be deleted. */ if (vp->v_vflag & VV_ROOT) error = EBUSY; out: /* * Issue or abort op. Since SAVESTART is not set, path name * component is freed by the VOP after either. */ if (!error) error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); NDFREE(&nd, NDF_ONLY_PNBUF); if (dirp) { if (dirp == nd.ni_dvp) diraft_ret = VOP_GETATTR(dirp, &diraft, cred); else { /* Release existing locks to prevent deadlock. */ if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vput(nd.ni_vp); nd.ni_dvp = NULL; nd.ni_vp = NULL; vn_lock(dirp, LK_EXCLUSIVE | LK_RETRY); diraft_ret = VOP_GETATTR(dirp, &diraft, cred); VOP_UNLOCK(dirp, 0); } } nfsm_reply(NFSX_WCCDATA(v3)); error = 0; if (v3) nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); /* fall through */ nfsmout: NDFREE(&nd, NDF_ONLY_PNBUF); if (nd.ni_dvp) { if (nd.ni_dvp == nd.ni_vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); } if (nd.ni_vp) vput(nd.ni_vp); if (dirp) vrele(dirp); vn_finished_write(mp); return(error); } /* * nfs readdir service * - mallocs what it thinks is enough to read * count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR * - calls VOP_READDIR() * - loops around building the reply * if the output generated exceeds count break out of loop * The nfsm_clget macro is used here so that the reply will be packed * tightly in mbuf clusters. * - it only knows that it has encountered eof when the VOP_READDIR() * reads nothing * - as such one readdir rpc will return eof false although you are there * and then the next will return eof * - it trims out records with d_fileno == 0 * this doesn't matter for Unix clients, but they might confuse clients * for other os'. * NB: It is tempting to set eof to true if the VOP_READDIR() reads less * than requested, but this may not apply to all filesystems. For * example, client NFS does not { although it is never remote mounted * anyhow } * The alternate call nfsrv_readdirplus() does lookups as well. * PS: The NFS protocol spec. does not clarify what the "count" byte * argument is a count of.. just name strings and file id's or the * entire reply rpc or ... * I tried just file name and id sizes and it confused the Sun client, * so I am using the full rpc size now. The "paranoia.." comment refers * to including the status longwords that are not a part of the dir. * "entry" structures, but are in the rpc. */ struct flrep { nfsuint64 fl_off; u_int32_t fl_postopok; u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)]; u_int32_t fl_fhok; u_int32_t fl_fhsize; u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)]; }; int nfsrv_readdir(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; char *bp, *be; struct mbuf *mp; struct dirent *dp; caddr_t cp; u_int32_t *tl; caddr_t bpos; struct mbuf *mb, *mreq; char *cpos, *cend, *rbuf; struct vnode *vp = NULL; struct vattr at; nfsfh_t nfh; fhandle_t *fhp; struct uio io; struct iovec iv; int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; int siz, cnt, fullsiz, eofflag, rdonly, ncookies; int v3 = (nfsd->nd_flag & ND_NFSV3); u_quad_t off, toff, verf; u_long *cookies = NULL, *cookiep; /* needs to be int64_t or off_t */ int not_zfs; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if (v3) { tl = nfsm_dissect_nonblock(u_int32_t *, 5 * NFSX_UNSIGNED); toff = fxdr_hyper(tl); tl += 2; verf = fxdr_hyper(tl); tl += 2; } else { tl = nfsm_dissect_nonblock(u_int32_t *, 2 * NFSX_UNSIGNED); toff = fxdr_unsigned(u_quad_t, *tl++); verf = 0; /* shut up gcc */ } off = toff; cnt = fxdr_unsigned(int, *tl); siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); xfer = NFS_SRVMAXDATA(nfsd); if (cnt > xfer) cnt = xfer; if (siz > xfer) siz = xfer; fullsiz = siz; error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (!error && vp->v_type != VDIR) { error = ENOTDIR; vput(vp); vp = NULL; } if (error) { nfsm_reply(NFSX_UNSIGNED); if (v3) nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } /* * Obtain lock on vnode for this section of the code */ if (v3) { error = getret = VOP_GETATTR(vp, &at, cred); #if 0 /* * XXX This check may be too strict for Solaris 2.5 clients. */ if (!error && toff && verf && verf != at.va_filerev) error = NFSERR_BAD_COOKIE; #endif } if (!error) error = nfsrv_access(vp, VEXEC, cred, rdonly, 0); if (error) { vput(vp); vp = NULL; nfsm_reply(NFSX_POSTOPATTR(v3)); if (v3) nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } not_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs") != 0; VOP_UNLOCK(vp, 0); /* * end section. Allocate rbuf and continue */ rbuf = malloc(siz, M_TEMP, M_WAITOK); again: iv.iov_base = rbuf; iv.iov_len = fullsiz; io.uio_iov = &iv; io.uio_iovcnt = 1; io.uio_offset = (off_t)off; io.uio_resid = fullsiz; io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_td = NULL; eofflag = 0; if (cookies) { free((caddr_t)cookies, M_TEMP); cookies = NULL; } vn_lock(vp, LK_SHARED | LK_RETRY); error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies); off = (off_t)io.uio_offset; if (!cookies && !error) error = NFSERR_PERM; if (v3) { getret = VOP_GETATTR(vp, &at, cred); if (!error) error = getret; } VOP_UNLOCK(vp, 0); if (error) { vrele(vp); vp = NULL; free((caddr_t)rbuf, M_TEMP); if (cookies) free((caddr_t)cookies, M_TEMP); nfsm_reply(NFSX_POSTOPATTR(v3)); if (v3) nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } if (io.uio_resid) { siz -= io.uio_resid; /* * If nothing read, return eof * rpc reply */ if (siz == 0) { vrele(vp); vp = NULL; nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + 2 * NFSX_UNSIGNED); if (v3) { nfsm_srvpostop_attr(getret, &at); tl = nfsm_build(u_int32_t *, 4 * NFSX_UNSIGNED); txdr_hyper(at.va_filerev, tl); tl += 2; } else tl = nfsm_build(u_int32_t *, 2 * NFSX_UNSIGNED); *tl++ = nfsrv_nfs_false; *tl = nfsrv_nfs_true; free((caddr_t)rbuf, M_TEMP); free((caddr_t)cookies, M_TEMP); error = 0; goto nfsmout; } } /* * Check for degenerate cases of nothing useful read. * If so go try again */ cpos = rbuf; cend = rbuf + siz; dp = (struct dirent *)cpos; cookiep = cookies; /* * For some reason FreeBSD's ufs_readdir() chooses to back the * directory offset up to a block boundary, so it is necessary to * skip over the records that precede the requested offset. This * requires the assumption that file offset cookies monotonically * increase. * Since the offset cookies don't monotonically increase for ZFS, * this is not done when ZFS is the file system. */ while (cpos < cend && ncookies > 0 && (dp->d_fileno == 0 || dp->d_type == DT_WHT || (not_zfs != 0 && ((u_quad_t)(*cookiep)) <= toff))) { cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } if (cpos >= cend || ncookies == 0) { toff = off; siz = fullsiz; goto again; } len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); if (v3) { nfsm_srvpostop_attr(getret, &at); tl = nfsm_build(u_int32_t *, 2 * NFSX_UNSIGNED); txdr_hyper(at.va_filerev, tl); } mp = mb; bp = bpos; be = bp + M_TRAILINGSPACE(mp); /* Loop through the records and build reply */ while (cpos < cend && ncookies > 0) { if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { nlen = dp->d_namlen; rem = nfsm_rndup(nlen) - nlen; len += (4 * NFSX_UNSIGNED + nlen + rem); if (v3) len += 2 * NFSX_UNSIGNED; if (len > cnt) { eofflag = 0; break; } /* * Build the directory record xdr from * the dirent entry. */ nfsm_clget; *tl = nfsrv_nfs_true; bp += NFSX_UNSIGNED; if (v3) { nfsm_clget; *tl = 0; bp += NFSX_UNSIGNED; } nfsm_clget; *tl = txdr_unsigned(dp->d_fileno); bp += NFSX_UNSIGNED; nfsm_clget; *tl = txdr_unsigned(nlen); bp += NFSX_UNSIGNED; /* And loop around copying the name */ xfer = nlen; cp = dp->d_name; while (xfer > 0) { nfsm_clget; if ((bp+xfer) > be) tsiz = be-bp; else tsiz = xfer; bcopy(cp, bp, tsiz); bp += tsiz; xfer -= tsiz; if (xfer > 0) cp += tsiz; } /* And null pad to an int32_t boundary. */ for (i = 0; i < rem; i++) *bp++ = '\0'; nfsm_clget; /* Finish off the record */ if (v3) { *tl = 0; bp += NFSX_UNSIGNED; nfsm_clget; } *tl = txdr_unsigned(*cookiep); bp += NFSX_UNSIGNED; } cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } vrele(vp); vp = NULL; nfsm_clget; *tl = nfsrv_nfs_false; bp += NFSX_UNSIGNED; nfsm_clget; if (eofflag) *tl = nfsrv_nfs_true; else *tl = nfsrv_nfs_false; bp += NFSX_UNSIGNED; if (mp != mb) { if (bp < be) mp->m_len = bp - mtod(mp, caddr_t); } else mp->m_len += bp - bpos; free((caddr_t)rbuf, M_TEMP); free((caddr_t)cookies, M_TEMP); nfsmout: if (vp) vrele(vp); return(error); } int nfsrv_readdirplus(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; char *bp, *be; struct mbuf *mp; struct dirent *dp; caddr_t cp; u_int32_t *tl; caddr_t bpos; struct mbuf *mb, *mreq; char *cpos, *cend, *rbuf; struct vnode *vp = NULL, *nvp; struct flrep fl; nfsfh_t nfh; fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; struct uio io; struct iovec iv; struct vattr va, at, *vap = &va; struct nfs_fattr *fp; int len, nlen, rem, xfer, tsiz, i, error = 0, error1, getret = 1; int vp_locked; int siz, cnt, fullsiz, eofflag, rdonly, dirlen, ncookies; u_quad_t off, toff, verf; u_long *cookies = NULL, *cookiep; /* needs to be int64_t or off_t */ int v3 = (nfsd->nd_flag & ND_NFSV3); int usevget = 1; struct componentname cn; struct mount *mntp = NULL; int not_zfs; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); vp_locked = 0; if (!v3) panic("nfsrv_readdirplus: v3 proc called on a v2 connection"); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); tl = nfsm_dissect_nonblock(u_int32_t *, 6 * NFSX_UNSIGNED); toff = fxdr_hyper(tl); tl += 2; verf = fxdr_hyper(tl); tl += 2; siz = fxdr_unsigned(int, *tl++); cnt = fxdr_unsigned(int, *tl); off = toff; siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); xfer = NFS_SRVMAXDATA(nfsd); if (cnt > xfer) cnt = xfer; if (siz > xfer) siz = xfer; fullsiz = siz; error = nfsrv_fhtovp(fhp, NFSRV_FLAG_BUSY, &vp, nfsd, slp, nam, &rdonly); if (!error) { vp_locked = 1; mntp = vp->v_mount; if (vp->v_type != VDIR) { error = ENOTDIR; vput(vp); vp = NULL; vp_locked = 0; } } if (error) { nfsm_reply(NFSX_UNSIGNED); nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } error = getret = VOP_GETATTR(vp, &at, cred); #if 0 /* * XXX This check may be too strict for Solaris 2.5 clients. */ if (!error && toff && verf && verf != at.va_filerev) error = NFSERR_BAD_COOKIE; #endif if (!error) error = nfsrv_access(vp, VEXEC, cred, rdonly, 0); if (error) { vput(vp); vp_locked = 0; vp = NULL; nfsm_reply(NFSX_V3POSTOPATTR); nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } not_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs") != 0; VOP_UNLOCK(vp, 0); vp_locked = 0; rbuf = malloc(siz, M_TEMP, M_WAITOK); again: iv.iov_base = rbuf; iv.iov_len = fullsiz; io.uio_iov = &iv; io.uio_iovcnt = 1; io.uio_offset = (off_t)off; io.uio_resid = fullsiz; io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_td = NULL; eofflag = 0; vp_locked = 1; if (cookies) { free((caddr_t)cookies, M_TEMP); cookies = NULL; } vn_lock(vp, LK_SHARED | LK_RETRY); error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies); off = (u_quad_t)io.uio_offset; getret = VOP_GETATTR(vp, &at, cred); VOP_UNLOCK(vp, 0); vp_locked = 0; if (!cookies && !error) error = NFSERR_PERM; if (!error) error = getret; if (error) { vrele(vp); vp = NULL; if (cookies) free((caddr_t)cookies, M_TEMP); free((caddr_t)rbuf, M_TEMP); nfsm_reply(NFSX_V3POSTOPATTR); nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } if (io.uio_resid) { siz -= io.uio_resid; /* * If nothing read, return eof * rpc reply */ if (siz == 0) { vrele(vp); vp = NULL; nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED); nfsm_srvpostop_attr(getret, &at); tl = nfsm_build(u_int32_t *, 4 * NFSX_UNSIGNED); txdr_hyper(at.va_filerev, tl); tl += 2; *tl++ = nfsrv_nfs_false; *tl = nfsrv_nfs_true; free((caddr_t)cookies, M_TEMP); free((caddr_t)rbuf, M_TEMP); error = 0; goto nfsmout; } } /* * Check for degenerate cases of nothing useful read. * If so go try again */ cpos = rbuf; cend = rbuf + siz; dp = (struct dirent *)cpos; cookiep = cookies; /* * For some reason FreeBSD's ufs_readdir() chooses to back the * directory offset up to a block boundary, so it is necessary to * skip over the records that precede the requested offset. This * requires the assumption that file offset cookies monotonically * increase. * Since the offset cookies don't monotonically increase for ZFS, * this is not done when ZFS is the file system. */ while (cpos < cend && ncookies > 0 && (dp->d_fileno == 0 || dp->d_type == DT_WHT || (not_zfs != 0 && ((u_quad_t)(*cookiep)) <= toff))) { cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } if (cpos >= cend || ncookies == 0) { toff = off; siz = fullsiz; goto again; } dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; nfsm_reply(cnt); nfsm_srvpostop_attr(getret, &at); tl = nfsm_build(u_int32_t *, 2 * NFSX_UNSIGNED); txdr_hyper(at.va_filerev, tl); mp = mb; bp = bpos; be = bp + M_TRAILINGSPACE(mp); /* Loop through the records and build reply */ while (cpos < cend && ncookies > 0) { if (dp->d_fileno != 0 && dp->d_type != DT_WHT) { nlen = dp->d_namlen; rem = nfsm_rndup(nlen)-nlen; if (usevget) { /* * For readdir_and_lookup get the vnode using * the file number. */ error = VFS_VGET(mntp, dp->d_fileno, LK_SHARED, &nvp); if (error != 0 && error != EOPNOTSUPP) { error = 0; goto invalid; } else if (error == EOPNOTSUPP) { /* * VFS_VGET() not supported? * Let's switch to VOP_LOOKUP(). */ error = 0; usevget = 0; cn.cn_nameiop = LOOKUP; cn.cn_flags = ISLASTCN | NOFOLLOW | \ LOCKSHARED | LOCKLEAF; cn.cn_lkflags = LK_SHARED | LK_RETRY; cn.cn_cred = cred; cn.cn_thread = curthread; } } if (!usevget) { cn.cn_nameptr = dp->d_name; cn.cn_namelen = dp->d_namlen; if (dp->d_namlen == 2 && dp->d_name[0] == '.' && dp->d_name[1] == '.') { cn.cn_flags |= ISDOTDOT; } else { cn.cn_flags &= ~ISDOTDOT; } if (!vp_locked) { vn_lock(vp, LK_SHARED | LK_RETRY); vp_locked = 1; } if ((vp->v_vflag & VV_ROOT) != 0 && (cn.cn_flags & ISDOTDOT) != 0) { vref(vp); nvp = vp; } else if (VOP_LOOKUP(vp, &nvp, &cn) != 0) goto invalid; } bzero((caddr_t)nfhp, NFSX_V3FH); nfhp->fh_fsid = nvp->v_mount->mnt_stat.f_fsid; if ((error1 = VOP_VPTOFH(nvp, &nfhp->fh_fid)) == 0) error1 = VOP_GETATTR(nvp, vap, cred); if (!usevget && vp == nvp) vunref(nvp); else vput(nvp); nvp = NULL; if (error1 != 0) goto invalid; /* * If either the dircount or maxcount will be * exceeded, get out now. Both of these lengths * are calculated conservatively, including all * XDR overheads. */ len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + NFSX_V3POSTOPATTR); dirlen += (6 * NFSX_UNSIGNED + nlen + rem); if (len > cnt || dirlen > fullsiz) { eofflag = 0; break; } /* * Build the directory record xdr from * the dirent entry. */ fp = (struct nfs_fattr *)&fl.fl_fattr; nfsm_srvfillattr(vap, fp); fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); fl.fl_fhok = nfsrv_nfs_true; fl.fl_postopok = nfsrv_nfs_true; fl.fl_off.nfsuquad[0] = 0; fl.fl_off.nfsuquad[1] = txdr_unsigned(*cookiep); nfsm_clget; *tl = nfsrv_nfs_true; bp += NFSX_UNSIGNED; nfsm_clget; *tl = 0; bp += NFSX_UNSIGNED; nfsm_clget; *tl = txdr_unsigned(dp->d_fileno); bp += NFSX_UNSIGNED; nfsm_clget; *tl = txdr_unsigned(nlen); bp += NFSX_UNSIGNED; /* And loop around copying the name */ xfer = nlen; cp = dp->d_name; while (xfer > 0) { nfsm_clget; if ((bp + xfer) > be) tsiz = be - bp; else tsiz = xfer; bcopy(cp, bp, tsiz); bp += tsiz; xfer -= tsiz; if (xfer > 0) cp += tsiz; } /* And null pad to an int32_t boundary. */ for (i = 0; i < rem; i++) *bp++ = '\0'; /* * Now copy the flrep structure out. */ xfer = sizeof (struct flrep); cp = (caddr_t)&fl; while (xfer > 0) { nfsm_clget; if ((bp + xfer) > be) tsiz = be - bp; else tsiz = xfer; bcopy(cp, bp, tsiz); bp += tsiz; xfer -= tsiz; if (xfer > 0) cp += tsiz; } } invalid: cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } if (!usevget && vp_locked) vput(vp); else vrele(vp); vp = NULL; nfsm_clget; *tl = nfsrv_nfs_false; bp += NFSX_UNSIGNED; nfsm_clget; if (eofflag) *tl = nfsrv_nfs_true; else *tl = nfsrv_nfs_false; bp += NFSX_UNSIGNED; if (mp != mb) { if (bp < be) mp->m_len = bp - mtod(mp, caddr_t); } else mp->m_len += bp - bpos; free((caddr_t)cookies, M_TEMP); free((caddr_t)rbuf, M_TEMP); nfsmout: if (vp) vrele(vp); if (mntp) vfs_unbusy(mntp); return(error); } /* * nfs commit service */ int nfsrv_commit(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct vattr bfor, aft; struct vnode *vp = NULL; nfsfh_t nfh; fhandle_t *fhp; u_int32_t *tl; caddr_t bpos; int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt; struct mbuf *mb, *mreq; u_quad_t off; struct mount *mp = NULL; int v3 = (nfsd->nd_flag & ND_NFSV3); nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); if (!v3) panic("nfsrv_commit: v3 proc called on a v2 connection"); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL) { error = ESTALE; goto ereply; } (void) vn_start_write(NULL, &mp, V_WAIT); vfs_rel(mp); /* The write holds a ref. */ tl = nfsm_dissect_nonblock(u_int32_t *, 3 * NFSX_UNSIGNED); /* * XXX At this time VOP_FSYNC() does not accept offset and byte * count parameters, so these arguments are useless (someday maybe). */ off = fxdr_hyper(tl); tl += 2; cnt = fxdr_unsigned(int, *tl); error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(2 * NFSX_UNSIGNED); nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); error = 0; goto nfsmout; } for_ret = VOP_GETATTR(vp, &bfor, cred); /* * RFC 1813 3.3.21: if count is 0, a flush from offset to the end of file * is done. At this time VOP_FSYNC does not accept offset and byte count * parameters so call VOP_FSYNC the whole file for now. */ if (cnt == 0 || cnt > MAX_COMMIT_COUNT) { /* * Give up and do the whole thing */ if (vp->v_object && (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) { VM_OBJECT_WLOCK(vp->v_object); vm_object_page_clean(vp->v_object, 0, 0, OBJPC_SYNC); VM_OBJECT_WUNLOCK(vp->v_object); } error = VOP_FSYNC(vp, MNT_WAIT, curthread); } else { /* * Locate and synchronously write any buffers that fall * into the requested range. Note: we are assuming that * f_iosize is a power of 2. */ int iosize = vp->v_mount->mnt_stat.f_iosize; int iomask = iosize - 1; struct bufobj *bo; daddr_t lblkno; /* * Align to iosize boundry, super-align to page boundry. */ if (off & iomask) { cnt += off & iomask; off &= ~(u_quad_t)iomask; } if (off & PAGE_MASK) { cnt += off & PAGE_MASK; off &= ~(u_quad_t)PAGE_MASK; } lblkno = off / iosize; if (vp->v_object && (vp->v_object->flags & OBJ_MIGHTBEDIRTY)) { VM_OBJECT_WLOCK(vp->v_object); vm_object_page_clean(vp->v_object, off, off + cnt, OBJPC_SYNC); VM_OBJECT_WUNLOCK(vp->v_object); } bo = &vp->v_bufobj; BO_LOCK(bo); while (cnt > 0) { struct buf *bp; /* * If we have a buffer and it is marked B_DELWRI we * have to lock and write it. Otherwise the prior * write is assumed to have already been committed. * * gbincore() can return invalid buffers now so we * have to check that bit as well (though B_DELWRI * should not be set if B_INVAL is set there could be * a race here since we haven't locked the buffer). */ if ((bp = gbincore(&vp->v_bufobj, lblkno)) != NULL) { if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK, BO_LOCKPTR(bo)) == ENOLCK) { BO_LOCK(bo); continue; /* retry */ } if ((bp->b_flags & (B_DELWRI|B_INVAL)) == B_DELWRI) { bremfree(bp); bp->b_flags &= ~B_ASYNC; bwrite(bp); ++nfs_commit_miss; } else BUF_UNLOCK(bp); BO_LOCK(bo); } ++nfs_commit_blks; if (cnt < iosize) break; cnt -= iosize; ++lblkno; } BO_UNLOCK(bo); } aft_ret = VOP_GETATTR(vp, &aft, cred); vput(vp); vp = NULL; ereply: nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); if (!error) { tl = nfsm_build(u_int32_t *, NFSX_V3WRITEVERF); if (nfsver.tv_sec == 0) nfsver = boottime; *tl++ = txdr_unsigned(nfsver.tv_sec); *tl = txdr_unsigned(nfsver.tv_usec); } else { error = 0; } nfsmout: if (vp) vput(vp); vn_finished_write(mp); return(error); } /* * nfs statfs service */ int nfsrv_statfs(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct statfs *sf; struct nfs_statfs *sfp; caddr_t bpos; int error = 0, rdonly, getret = 1; int v3 = (nfsd->nd_flag & ND_NFSV3); struct mbuf *mb, *mreq; struct vnode *vp = NULL; struct vattr at; nfsfh_t nfh; fhandle_t *fhp; struct statfs statfs; u_quad_t tval; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(NFSX_UNSIGNED); if (v3) nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } sf = &statfs; error = VFS_STATFS(vp->v_mount, sf); getret = VOP_GETATTR(vp, &at, cred); vput(vp); vp = NULL; nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); if (v3) nfsm_srvpostop_attr(getret, &at); if (error) { error = 0; goto nfsmout; } sfp = nfsm_build(struct nfs_statfs *, NFSX_STATFS(v3)); if (v3) { tval = (u_quad_t)sf->f_blocks; tval *= (u_quad_t)sf->f_bsize; txdr_hyper(tval, &sfp->sf_tbytes); tval = (u_quad_t)sf->f_bfree; tval *= (u_quad_t)sf->f_bsize; txdr_hyper(tval, &sfp->sf_fbytes); /* * Don't send negative values for available space, * since this field is unsigned in the NFS protocol. * Otherwise, the client would see absurdly high * numbers for free space. */ if (sf->f_bavail < 0) tval = 0; else tval = (u_quad_t)sf->f_bavail; tval *= (u_quad_t)sf->f_bsize; txdr_hyper(tval, &sfp->sf_abytes); sfp->sf_tfiles.nfsuquad[0] = 0; sfp->sf_tfiles.nfsuquad[1] = txdr_unsigned(sf->f_files); sfp->sf_ffiles.nfsuquad[0] = 0; sfp->sf_ffiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree); sfp->sf_afiles.nfsuquad[0] = 0; sfp->sf_afiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree); sfp->sf_invarsec = 0; } else { sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); sfp->sf_bsize = txdr_unsigned(sf->f_bsize); sfp->sf_blocks = txdr_unsigned(sf->f_blocks); sfp->sf_bfree = txdr_unsigned(sf->f_bfree); if (sf->f_bavail < 0) sfp->sf_bavail = 0; else sfp->sf_bavail = txdr_unsigned(sf->f_bavail); } nfsmout: if (vp) vput(vp); return(error); } /* * nfs fsinfo service */ int nfsrv_fsinfo(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct nfsv3_fsinfo *sip; caddr_t bpos; int error = 0, rdonly, getret = 1, pref; struct mbuf *mb, *mreq; struct vnode *vp = NULL; struct vattr at; nfsfh_t nfh; fhandle_t *fhp; u_quad_t maxfsize; struct statfs sb; int v3 = (nfsd->nd_flag & ND_NFSV3); nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); if (!v3) panic("nfsrv_fsinfo: v3 proc called on a v2 connection"); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(NFSX_UNSIGNED); nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } /* XXX Try to make a guess on the max file size. */ VFS_STATFS(vp->v_mount, &sb); maxfsize = (u_quad_t)0x80000000 * sb.f_bsize - 1; getret = VOP_GETATTR(vp, &at, cred); vput(vp); vp = NULL; nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); nfsm_srvpostop_attr(getret, &at); sip = nfsm_build(struct nfsv3_fsinfo *, NFSX_V3FSINFO); /* * XXX * There should be filesystem VFS OP(s) to get this information. * For now, assume ufs. */ pref = NFS_SRVMAXDATA(nfsd); sip->fs_rtmax = txdr_unsigned(pref); sip->fs_rtpref = txdr_unsigned(pref); sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); sip->fs_wtmax = txdr_unsigned(pref); sip->fs_wtpref = txdr_unsigned(pref); sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); sip->fs_dtpref = txdr_unsigned(pref); txdr_hyper(maxfsize, &sip->fs_maxfilesize); sip->fs_timedelta.nfsv3_sec = 0; sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | NFSV3FSINFO_CANSETTIME); nfsmout: if (vp) vput(vp); return(error); } /* * nfs pathconf service */ int nfsrv_pathconf(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; struct sockaddr *nam = nfsd->nd_nam; caddr_t dpos = nfsd->nd_dpos; struct ucred *cred = nfsd->nd_cr; struct nfsv3_pathconf *pc; caddr_t bpos; int error = 0, rdonly, getret = 1; register_t linkmax, namemax, chownres, notrunc; struct mbuf *mb, *mreq; struct vnode *vp = NULL; struct vattr at; nfsfh_t nfh; fhandle_t *fhp; int v3 = (nfsd->nd_flag & ND_NFSV3); nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); if (!v3) panic("nfsrv_pathconf: v3 proc called on a v2 connection"); fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); error = nfsrv_fhtovp(fhp, 0, &vp, nfsd, slp, nam, &rdonly); if (error) { nfsm_reply(NFSX_UNSIGNED); nfsm_srvpostop_attr(getret, &at); error = 0; goto nfsmout; } error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); if (!error) error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); if (!error) error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); if (!error) error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); getret = VOP_GETATTR(vp, &at, cred); vput(vp); vp = NULL; nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); nfsm_srvpostop_attr(getret, &at); if (error) { error = 0; goto nfsmout; } pc = nfsm_build(struct nfsv3_pathconf *, NFSX_V3PATHCONF); pc->pc_linkmax = txdr_unsigned(linkmax); pc->pc_namemax = txdr_unsigned(namemax); pc->pc_notrunc = txdr_unsigned(notrunc); pc->pc_chownrestricted = txdr_unsigned(chownres); /* * These should probably be supported by VOP_PATHCONF(), but * until msdosfs is exportable (why would you want to?), the * Unix defaults should be ok. */ pc->pc_caseinsensitive = nfsrv_nfs_false; pc->pc_casepreserving = nfsrv_nfs_true; nfsmout: if (vp) vput(vp); return(error); } /* * Null operation, used by clients to ping server */ /* ARGSUSED */ int nfsrv_null(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep; caddr_t bpos; int error = NFSERR_RETVOID; struct mbuf *mb, *mreq; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); nfsm_reply(0); nfsmout: return (error); } /* * No operation, used for obsolete procedures */ /* ARGSUSED */ int nfsrv_noop(struct nfsrv_descript *nfsd, struct nfssvc_sock *slp, struct mbuf **mrq) { struct mbuf *mrep = nfsd->nd_mrep; caddr_t bpos; int error; struct mbuf *mb, *mreq; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); if (nfsd->nd_repstat) error = nfsd->nd_repstat; else error = EPROCUNAVAIL; nfsm_reply(0); error = 0; nfsmout: return (error); } /* * Perform access checking for vnodes obtained from file handles that would * refer to files already opened by a Unix client. You cannot just use * vn_writechk() and VOP_ACCESS() for two reasons. * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write * case. * 2 - The owner is to be given access irrespective of mode bits for some * operations, so that processes that chmod after opening a file don't * break. I don't like this because it opens a security hole, but since * the nfs server opens a security hole the size of a barn door anyhow, * what the heck. * * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS() * will return EPERM instead of EACCES. EPERM is always an error. */ static int nfsrv_access(struct vnode *vp, accmode_t accmode, struct ucred *cred, int rdonly, int override) { struct vattr vattr; int error; nfsdbprintf(("%s %d\n", __FILE__, __LINE__)); if (accmode & VWRITE) { /* Just vn_writechk() changed to check rdonly */ /* * Disallow write attempts on read-only filesystems; * unless the file is a socket or a block or character * device resident on the filesystem. */ if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) { switch (vp->v_type) { case VREG: case VDIR: case VLNK: return (EROFS); default: break; } } /* * If there's shared text associated with * the inode, we can't allow writing. */ if (VOP_IS_TEXT(vp)) return (ETXTBSY); } error = VOP_GETATTR(vp, &vattr, cred); if (error) return (error); error = VOP_ACCESS(vp, accmode, cred, curthread); /* * Allow certain operations for the owner (reads and writes * on files that are already open). */ if (override && error == EACCES && cred->cr_uid == vattr.va_uid) error = 0; return (error); }