/* * 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_vnops.c 8.16 (Berkeley) 5/27/95 */ #include __FBSDID("$FreeBSD: src/sys/nfsclient/nfs_vnops.c,v 1.218 2004/04/07 04:59:57 imp Exp $"); /* * vnode op calls for Sun NFS version 2 and 3 */ #include "opt_inet.h" #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 #include #include #include #include #include #include #include #include /* Defs */ #define TRUE 1 #define FALSE 0 /* * Ifdef for FreeBSD-current merged buffer cache. It is unfortunate that these * calls are not in getblk() and brelse() so that they would not be necessary * here. */ #ifndef B_VMIO #define vfs_busy_pages(bp, f) #endif int nfsx_fsync(struct vop_fsync_args *); int nfsx_close(struct vop_close_args *); int nfsx_lookup(struct vop_lookup_args *); int nfsx_mknod(struct vop_mknod_args *); int nfsx_print(struct vop_print_args *); int nfsx_read(struct vop_read_args *); int nfsx_readdir(struct vop_readdir_args *); int nfsx_readlink(struct vop_readlink_args *); int nfsx_rename(struct vop_rename_args *); int nfsx_setattr(struct vop_setattr_args *); int nfsx_lookitup(struct vnode *, const char *, int, struct ucred *, struct thread *, struct nfsnode **); int nfsx_sillyrename(struct vnode *, struct vnode *, struct componentname *); int nfsx_remove(struct vop_remove_args *); int nfsx_renameit(struct vnode *sdvp, struct componentname *scnp, struct sillyrename *sp); int nfsx_strategy(struct vop_strategy_args *); int nfsx_advlock(struct vop_advlock_args *); static int nfsx_removeit(struct sillyrename *); static int nfsx_flush(struct vnode *, struct ucred *, int, struct thread *, int); /* * Global variables */ struct proc *nfsx_iodwant[NFS_MAXASYNCDAEMON]; struct nfsmount *nfsx_iodmount[NFS_MAXASYNCDAEMON]; int nfsx_numasync = 0; #define DIRHDSIZ (sizeof (struct dirent) - (MAXNAMLEN + 1)) SYSCTL_DECL(_vfs_nfsx); static int nfsaccess_cache_timeout = NFS_MAXATTRTIMO; SYSCTL_INT(_vfs_nfsx, OID_AUTO, access_cache_timeout, CTLFLAG_RW, &nfsaccess_cache_timeout, 0, "NFS ACCESS cache timeout"); static int nfsv3_commit_on_close = 0; SYSCTL_INT(_vfs_nfsx, OID_AUTO, nfsv3_commit_on_close, CTLFLAG_RW, &nfsv3_commit_on_close, 0, "write+commit on close, else only write"); #if 0 SYSCTL_INT(_vfs_nfsx, OID_AUTO, access_cache_hits, CTLFLAG_RD, &nfsstats.accesscache_hits, 0, "NFS ACCESS cache hit count"); SYSCTL_INT(_vfs_nfsx, OID_AUTO, access_cache_misses, CTLFLAG_RD, &nfsstats.accesscache_misses, 0, "NFS ACCESS cache miss count"); #endif #define NFSV3ACCESS_ALL (NFSV3ACCESS_READ | NFSV3ACCESS_MODIFY \ | NFSV3ACCESS_EXTEND | NFSV3ACCESS_EXECUTE \ | NFSV3ACCESS_DELETE | NFSV3ACCESS_LOOKUP) /* * nfs close vnode op * What an NFS client should do upon close after writing is a debatable issue. * Most NFS clients push delayed writes to the server upon close, basically for * two reasons: * 1 - So that any write errors may be reported back to the client process * doing the close system call. By far the two most likely errors are * NFSERR_NOSPC and NFSERR_DQUOT to indicate space allocation failure. * 2 - To put a worst case upper bound on cache inconsistency between * multiple clients for the file. * There is also a consistency problem for Version 2 of the protocol w.r.t. * not being able to tell if other clients are writing a file concurrently, * since there is no way of knowing if the changed modify time in the reply * is only due to the write for this client. * (NFS Version 3 provides weak cache consistency data in the reply that * should be sufficient to detect and handle this case.) * * The current code does the following: * for NFS Version 2 - play it safe and flush/invalidate all dirty buffers * for NFS Version 3 - flush dirty buffers to the server but don't invalidate * or commit them (this satisfies 1 and 2 except for the * case where the server crashes after this close but * before the commit RPC, which is felt to be "good * enough". Changing the last argument to nfs_flush() to * a 1 would force a commit operation, if it is felt a * commit is necessary now. */ /* ARGSUSED */ int nfsx_close(struct vop_close_args *ap) { struct vnode *vp = ap->a_vp; struct nfsnode *np = VTONFS(vp); int error = 0; if (vp->v_type == VREG) { if (np->n_flag & NMODIFIED) { if (NFS_ISV3(vp)) { /* * Under NFSv3 we have dirty buffers to dispose of. We * must flush them to the NFS server. We have the option * of waiting all the way through the commit rpc or just * waiting for the initial write. The default is to only * wait through the initial write so the data is in the * server's cache, which is roughly similar to the state * a standard disk subsystem leaves the file in on close(). * * We cannot clear the NMODIFIED bit in np->n_flag due to * potential races with other processes, and certainly * cannot clear it if we don't commit. */ int cm = nfsv3_commit_on_close ? 1 : 0; error = nfsx_flush(vp, ap->a_cred, MNT_WAIT, ap->a_td, cm); /* np->n_flag &= ~NMODIFIED; */ } else { vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, ap->a_td); error = nfsx_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_td, 1); VOP_UNLOCK(vp, 0, ap->a_td); } np->n_attrstamp = 0; } if (np->n_flag & NWRITEERR) { np->n_flag &= ~NWRITEERR; error = np->n_error; } } return (error); } /* * nfs setattr call. */ int nfsx_setattr(struct vop_setattr_args *ap) { struct vnode *vp = ap->a_vp; struct nfsnode *np = VTONFS(vp); struct vattr *vap = ap->a_vap; int error = 0; u_quad_t tsize; struct nfsmount *nmp = VFSTONFS(vp->v_mount); #ifndef nolint tsize = (u_quad_t)0; #endif /* * Setting of flags is not supported. */ if (vap->va_flags != VNOVAL) return (EOPNOTSUPP); /* * Disallow write attempts if the filesystem is mounted read-only. */ if ((vap->va_flags != VNOVAL || vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL || vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL || vap->va_mode != (mode_t)VNOVAL) && (vp->v_mount->mnt_flag & MNT_RDONLY)) return (EROFS); if (vap->va_size != VNOVAL) { switch (vp->v_type) { case VDIR: return (EISDIR); case VCHR: case VBLK: case VSOCK: case VFIFO: if (vap->va_mtime.tv_sec == VNOVAL && vap->va_atime.tv_sec == VNOVAL && vap->va_mode == (mode_t)VNOVAL && vap->va_uid == (uid_t)VNOVAL && vap->va_gid == (gid_t)VNOVAL) return (0); vap->va_size = VNOVAL; break; default: /* * Disallow write attempts if the filesystem is * mounted read-only. */ if (vp->v_mount->mnt_flag & MNT_RDONLY) return (EROFS); /* * We run vnode_pager_setsize() early (why?), * we must set np->n_size now to avoid vinvalbuf * V_SAVE races that might setsize a lower * value. */ tsize = np->n_size; error = nfsx_meta_setsize(vp, ap->a_cred, ap->a_td, vap->va_size); if (np->n_flag & NMODIFIED) { if (vap->va_size == 0) error = nfsx_vinvalbuf(vp, 0, ap->a_cred, ap->a_td, 1); else error = nfsx_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_td, 1); if (error) { vnode_pager_setsize(vp, np->n_size); return (error); } } /* * np->n_size has already been set to vap->va_size * in nfsx_meta_setsize(). We must set it again since * nfs_loadattrcache() could be called through * nfsx_meta_setsize() and could modify np->n_size. */ np->n_vattr.va_size = np->n_size = vap->va_size; }; } else if ((vap->va_mtime.tv_sec != VNOVAL || vap->va_atime.tv_sec != VNOVAL) && (np->n_flag & NMODIFIED) && vp->v_type == VREG && (error = nfsx_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_td, 1)) == EINTR) return (error); error = NFSOP(nmp, setattrrpc)(vp, vap, ap->a_cred, ap->a_td); if (error && vap->va_size != VNOVAL) { np->n_size = np->n_vattr.va_size = tsize; vnode_pager_setsize(vp, np->n_size); } return (error); } /* * nfs lookup call, one step at a time... * First look in cache * If not found, unlock the directory nfsnode and do the rpc */ int nfsx_lookup(struct vop_lookup_args *ap) { struct componentname *cnp = ap->a_cnp; struct vnode *dvp = ap->a_dvp; struct vnode **vpp = ap->a_vpp; int flags = cnp->cn_flags; struct vnode *newvp; struct nfsmount *nmp; nfsfh_t *fhp; struct nfsnode *np; int lockparent, wantparent, error = 0, fhsize; struct thread *td = cnp->cn_thread; u_char stp[NFSX_STATESIZ]; *vpp = NULLVP; cnp->cn_flags &= ~PDIRUNLOCK; if ((flags & ISLASTCN) && (dvp->v_mount->mnt_flag & MNT_RDONLY) && (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) return (EROFS); if (dvp->v_type != VDIR) return (ENOTDIR); lockparent = flags & LOCKPARENT; wantparent = flags & (LOCKPARENT|WANTPARENT); nmp = VFSTONFS(dvp->v_mount); np = VTONFS(dvp); if ((error = cache_lookup(dvp, vpp, cnp)) && error != ENOENT) { struct vattr vattr; int vpid; if ((error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, td)) != 0) { *vpp = NULLVP; return (error); } vhold(*vpp); newvp = *vpp; vpid = newvp->v_id; /* * See the comment starting `Step through' in ufs/ufs_lookup.c * for an explanation of the locking protocol */ if (dvp == newvp) { VREF(newvp); error = 0; } else if (flags & ISDOTDOT) { VOP_UNLOCK(dvp, 0, td); cnp->cn_flags |= PDIRUNLOCK; error = vget(newvp, LK_EXCLUSIVE, td); if (!error && lockparent && (flags & ISLASTCN)) { error = vn_lock(dvp, LK_EXCLUSIVE, td); if (error == 0) cnp->cn_flags &= ~PDIRUNLOCK; } } else { error = vget(newvp, LK_EXCLUSIVE, td); if (!lockparent || error || !(flags & ISLASTCN)) { VOP_UNLOCK(dvp, 0, td); cnp->cn_flags |= PDIRUNLOCK; } } if (!error) { if (vpid == newvp->v_id) { if (!VOP_GETATTR(newvp, &vattr, cnp->cn_cred, td) && vattr.va_ctime.tv_sec == VTONFS(newvp)->n_ctime) { nfsxstats.lookupcache_hits++; if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) cnp->cn_flags |= SAVENAME; vdrop(newvp); return (0); } cache_purge(newvp); } vput(newvp); if (lockparent && dvp != newvp && (flags & ISLASTCN)) VOP_UNLOCK(dvp, 0, td); } vdrop(newvp); error = vn_lock(dvp, LK_EXCLUSIVE, td); *vpp = NULLVP; if (error) { cnp->cn_flags |= PDIRUNLOCK; return (error); } cnp->cn_flags &= ~PDIRUNLOCK; } newvp = NULLVP; nfsxstats.lookupcache_misses++; error = NFSOP(nmp, lookuprpc)(dvp, cnp, flags & ISDOTDOT, &fhp, &fhsize, stp); if (error != 0) goto nfsmout; /* * Handle RENAME case... */ if (cnp->cn_nameiop == RENAME && wantparent && (flags & ISLASTCN)) { if (NFS_CMPFH(np, fhp, fhsize)) return (EISDIR); error = nfsx_nget(dvp->v_mount, fhp, fhsize, &np); if (error) return (error); newvp = NFSTOV(np); NFSOP(nmp, lookuploadattr)(newvp, dvp, stp); *vpp = newvp; cnp->cn_flags |= SAVENAME; if (!lockparent) { VOP_UNLOCK(dvp, 0, td); cnp->cn_flags |= PDIRUNLOCK; } return (0); } if (flags & ISDOTDOT) { VOP_UNLOCK(dvp, 0, td); error = nfsx_nget(dvp->v_mount, fhp, fhsize, &np); if (error) { vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td); return (error); } newvp = NFSTOV(np); if (lockparent && (flags & ISLASTCN)) { error = vn_lock(dvp, LK_EXCLUSIVE, td); if (error) { cnp->cn_flags |= PDIRUNLOCK; vput(newvp); return (error); } } else cnp->cn_flags |= PDIRUNLOCK; } else if (NFS_CMPFH(np, fhp, fhsize)) { VREF(dvp); newvp = dvp; } else { error = nfsx_nget(dvp->v_mount, fhp, fhsize, &np); if (error) return (error); if (!lockparent || !(flags & ISLASTCN)) { cnp->cn_flags |= PDIRUNLOCK; VOP_UNLOCK(dvp, 0, td); } newvp = NFSTOV(np); /* * XXXMARIUS: this is v4 specific; should have a * version handler here. */ /* Copy name for open() */ np->n_dvp = dvp; np->n_namelen = cnp->cn_namelen; if (np->n_name != NULL) FREE(np->n_name, M_NFSREQ); MALLOC(np->n_name, u_char *, np->n_namelen + 1, M_NFSREQ, M_WAITOK); bcopy(cnp->cn_nameptr, np->n_name, np->n_namelen); np->n_name[np->n_namelen] = '\0'; NFSOP(nmp, lookuploadattr)(newvp, dvp, stp); } if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) cnp->cn_flags |= SAVENAME; if ((cnp->cn_flags & MAKEENTRY) && (cnp->cn_nameiop != DELETE || !(flags & ISLASTCN))) { np->n_ctime = np->n_vattr.va_ctime.tv_sec; cache_enter(dvp, newvp, cnp); } *vpp = newvp; nfsmout: if (error) { if (newvp != NULLVP) { vrele(newvp); *vpp = NULLVP; } if ((cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME) && (flags & ISLASTCN) && error == ENOENT) { if (!lockparent) { VOP_UNLOCK(dvp, 0, td); cnp->cn_flags |= PDIRUNLOCK; } if (dvp->v_mount->mnt_flag & MNT_RDONLY) error = EROFS; else error = EJUSTRETURN; } if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) cnp->cn_flags |= SAVENAME; } return (error); } /* * nfs read call. * Just call nfs_bioread() to do the work. */ int nfsx_read(struct vop_read_args *ap) { struct vnode *vp = ap->a_vp; if (vp->v_type == VREG) return (nfsx_bioread(vp, ap->a_uio, ap->a_ioflag, ap->a_cred)); return (vp->v_type == VDIR ? EISDIR : EOPNOTSUPP); } /* * nfs readlink call */ int nfsx_readlink(struct vop_readlink_args *ap) { struct vnode *vp = ap->a_vp; if (vp->v_type != VLNK) return (EINVAL); return (nfsx_bioread(vp, ap->a_uio, 0, ap->a_cred)); } /* * nfs mknod vop * just call nfs_mknodrpc() to do the work. */ /* ARGSUSED */ int nfsx_mknod(struct vop_mknod_args *ap) { struct vnode *newvp; struct nfsmount *nmp = VFSTONFS(ap->a_dvp->v_mount); int error; error = NFSOP(nmp, mknodrpc)(ap->a_dvp, &newvp, ap->a_cnp, ap->a_vap); /* XXX - is this actually referenced here? */ if (error == 0) { *ap->a_vpp = newvp; vrele(newvp); } return (error); } /* * nfs file remove call * To try and make nfs semantics closer to ufs semantics, a file that has * other processes using the vnode is renamed instead of removed and then * removed later on the last close. * - If v_usecount > 1 * If a rename is not already in the works * call nfsx_sillyrename() to set it up * else * do the remove rpc */ int nfsx_remove(struct vop_remove_args *ap) { struct vnode *vp = ap->a_vp; struct vnode *dvp = ap->a_dvp; struct componentname *cnp = ap->a_cnp; struct nfsnode *np = VTONFS(vp); int error = 0; struct vattr vattr; struct nfsmount *nmp = VFSTONFS(ap->a_vp->v_mount); #ifndef DIAGNOSTIC if ((cnp->cn_flags & HASBUF) == 0) panic("nfs_remove: no name"); if (vrefcnt(vp) < 1) panic("nfs_remove: bad v_usecount"); #endif if (vp->v_type == VDIR) error = EPERM; else if (vrefcnt(vp) == 1 || (np->n_sillyrename && VOP_GETATTR(vp, &vattr, cnp->cn_cred, cnp->cn_thread) == 0 && vattr.va_nlink > 1)) { /* * Purge the name cache so that the chance of a lookup for * the name succeeding while the remove is in progress is * minimized. Without node locking it can still happen, such * that an I/O op returns ESTALE, but since you get this if * another host removes the file.. */ cache_purge(vp); /* * throw away biocache buffers, mainly to avoid * unnecessary delayed writes later. */ error = nfsx_vinvalbuf(vp, 0, cnp->cn_cred, cnp->cn_thread, 1); /* Do the rpc */ if (error != EINTR) error = NFSOP(nmp, removerpc)(dvp, cnp->cn_nameptr, cnp->cn_namelen, cnp->cn_cred, cnp->cn_thread); /* * Kludge City: If the first reply to the remove rpc is lost.. * the reply to the retransmitted request will be ENOENT * since the file was in fact removed * Therefore, we cheat and return success. */ if (error == ENOENT) error = 0; } else if (!np->n_sillyrename) error = nfsx_sillyrename(dvp, vp, cnp); np->n_attrstamp = 0; return (error); } /* * nfs file remove rpc called from nfs_inactive */ static int nfsx_removeit(struct sillyrename *sp) { struct nfsmount *nmp = VFSTONFS(sp->s_dvp->v_mount); /* * Make sure that the directory vnode is still valid. * XXX we should lock sp->s_dvp here. */ if (sp->s_dvp->v_type == VBAD) return (0); return (NFSOP(nmp, removerpc)(sp->s_dvp, sp->s_name, sp->s_namlen, sp->s_cred, NULL)); } /* * nfs file rename call */ int nfsx_rename(struct vop_rename_args *ap) { struct vnode *fvp = ap->a_fvp; struct vnode *tvp = ap->a_tvp; struct vnode *fdvp = ap->a_fdvp; struct vnode *tdvp = ap->a_tdvp; struct componentname *tcnp = ap->a_tcnp; struct componentname *fcnp = ap->a_fcnp; int error; struct nfsmount *nmp = VFSTONFS(ap->a_fvp->v_mount); #ifndef DIAGNOSTIC if ((tcnp->cn_flags & HASBUF) == 0 || (fcnp->cn_flags & HASBUF) == 0) panic("nfs_rename: no name"); #endif /* Check for cross-device rename */ if ((fvp->v_mount != tdvp->v_mount) || (tvp && (fvp->v_mount != tvp->v_mount))) { error = EXDEV; goto out; } if (fvp == tvp) { printf("nfs_rename: fvp == tvp (can't happen)\n"); error = 0; goto out; } if ((error = vn_lock(fvp, LK_EXCLUSIVE, fcnp->cn_thread)) != 0) goto out; /* * We have to flush B_DELWRI data prior to renaming * the file. If we don't, the delayed-write buffers * can be flushed out later after the file has gone stale * under NFSV3. NFSV2 does not have this problem because * ( as far as I can tell ) it flushes dirty buffers more * often. */ VOP_FSYNC(fvp, fcnp->cn_cred, MNT_WAIT, fcnp->cn_thread); VOP_UNLOCK(fvp, 0, fcnp->cn_thread); if (tvp) VOP_FSYNC(tvp, tcnp->cn_cred, MNT_WAIT, tcnp->cn_thread); /* * If the tvp exists and is in use, sillyrename it before doing the * rename of the new file over it. * XXX Can't sillyrename a directory. */ if (tvp && vrefcnt(tvp) > 1 && !VTONFS(tvp)->n_sillyrename && tvp->v_type != VDIR && !nfsx_sillyrename(tdvp, tvp, tcnp)) { vput(tvp); tvp = NULL; } error = NFSOP(nmp, renamerpc)(fdvp, fcnp->cn_nameptr, fcnp->cn_namelen, tdvp, tcnp->cn_nameptr, tcnp->cn_namelen, tcnp->cn_cred, tcnp->cn_thread); if (fvp->v_type == VDIR) { if (tvp != NULL && tvp->v_type == VDIR) cache_purge(tdvp); cache_purge(fdvp); } out: if (tdvp == tvp) vrele(tdvp); else vput(tdvp); if (tvp) vput(tvp); vrele(fdvp); vrele(fvp); /* * Kludge: Map ENOENT => 0 assuming that it is a reply to a retry. */ if (error == ENOENT) error = 0; return (error); } /* * nfs file rename rpc called from nfs_remove() above */ int nfsx_renameit(struct vnode *sdvp, struct componentname *scnp, struct sillyrename *sp) { struct nfsmount *nmp = VFSTONFS(sdvp->v_mount); return (NFSOP(nmp, renamerpc)(sdvp, scnp->cn_nameptr, scnp->cn_namelen, sdvp, sp->s_name, sp->s_namlen, scnp->cn_cred, scnp->cn_thread)); } /* * nfs readdir call */ int nfsx_readdir(struct vop_readdir_args *ap) { struct vnode *vp = ap->a_vp; struct nfsnode *np = VTONFS(vp); struct uio *uio = ap->a_uio; int tresid, error; struct vattr vattr; if (vp->v_type != VDIR) return (EPERM); /* * first, check for hit on the EOF offset cache */ if (np->n_direofoffset > 0 && uio->uio_offset >= np->n_direofoffset && (np->n_flag & NMODIFIED) == 0) { if (VOP_GETATTR(vp, &vattr, ap->a_cred, uio->uio_td) == 0 && np->n_mtime == vattr.va_mtime.tv_sec) { nfsxstats.direofcache_hits++; return (0); } } /* * Call nfs_bioread() to do the real work. */ tresid = uio->uio_resid; error = nfsx_bioread(vp, uio, 0, ap->a_cred); if (!error && uio->uio_resid == tresid) nfsxstats.direofcache_misses++; return (error); } /* * Silly rename. To make the NFS filesystem that is stateless look a little * more like the "ufs" a remove of an active vnode is translated to a rename * to a funny looking filename that is removed by nfs_inactive on the * nfsnode. There is the potential for another process on a different client * to create the same funny name between the nfs_lookitup() fails and the * nfs_rename() completes, but... */ int nfsx_sillyrename(struct vnode *dvp, struct vnode *vp, struct componentname *cnp) { struct sillyrename *sp; struct nfsnode *np; int error; short pid; cache_purge(dvp); np = VTONFS(vp); #ifndef DIAGNOSTIC if (vp->v_type == VDIR) panic("nfs: sillyrename dir"); #endif MALLOC(sp, struct sillyrename *, sizeof (struct sillyrename), M_NFSREQ, M_WAITOK); sp->s_cred = crhold(cnp->cn_cred); sp->s_dvp = dvp; sp->s_removeit = nfsx_removeit; VREF(dvp); /* Fudge together a funny name */ pid = cnp->cn_thread->td_proc->p_pid; sp->s_namlen = sprintf(sp->s_name, ".nfsA%04x4.4", pid); /* Try lookitups until we get one that isn't there */ while (nfsx_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred, cnp->cn_thread, NULL) == 0) { sp->s_name[4]++; if (sp->s_name[4] > 'z') { error = EINVAL; goto bad; } } error = nfsx_renameit(dvp, cnp, sp); if (error) goto bad; error = nfsx_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred, cnp->cn_thread, &np); np->n_sillyrename = sp; return (0); bad: vrele(sp->s_dvp); crfree(sp->s_cred); free((caddr_t)sp, M_NFSREQ); return (error); } /* * Look up a file name and optionally either update the file handle or * allocate an nfsnode, depending on the value of npp. * npp == NULL --> just do the lookup * *npp == NULL --> allocate a new nfsnode and make sure attributes are * handled too * *npp != NULL --> update the file handle in the vnode */ int nfsx_lookitup(struct vnode *dvp, const char *name, int len, struct ucred *cred, struct thread *td, struct nfsnode **npp) { struct vnode *newvp = NULL; struct nfsnode *np, *dnp = VTONFS(dvp); int error = 0, fhlen; nfsfh_t *nfhp; struct componentname cn; struct nfsmount *nmp = VFSTONFS(dvp->v_mount); u_char stp[NFSX_STATESIZ]; nfsxstats.rpccnt[NFSPROC_LOOKUP]++; (const char *)cn.cn_nameptr = name; cn.cn_namelen = len; error = NFSOP(nmp, lookuprpc)(dvp, &cn, 0, &nfhp, &fhlen, stp); if (npp && !error) { if (*npp) { np = *npp; if (np->n_fhsize > NFS_SMALLFH && fhlen <= NFS_SMALLFH) { free((caddr_t)np->n_fhp, M_NFSBIGFH); np->n_fhp = &np->n_fh; } else if (np->n_fhsize <= NFS_SMALLFH && fhlen>NFS_SMALLFH) np->n_fhp =(nfsfh_t *)malloc(fhlen, M_NFSBIGFH, M_WAITOK); bcopy((caddr_t)nfhp, (caddr_t)np->n_fhp, fhlen); np->n_fhsize = fhlen; newvp = NFSTOV(np); } else if (NFS_CMPFH(dnp, nfhp, fhlen)) { VREF(dvp); newvp = dvp; } else { error = nfsx_nget(dvp->v_mount, nfhp, fhlen, &np); if (error) return (error); newvp = NFSTOV(np); } if (newvp != dvp) { np->n_dvp = dvp; np->n_namelen = len; if (np->n_name != NULL) FREE(np->n_name, M_NFSREQ); MALLOC(np->n_name, u_char *, np->n_namelen + 1, M_NFSREQ, M_WAITOK); memcpy(np->n_name, name, len); np->n_name[len] = '\0'; } NFSOP(nmp, lookuploadattr)(newvp, dvp, stp); } /* nfsmout: */ if (npp && *npp == NULL) { if (error) { if (newvp) { if (newvp == dvp) vrele(newvp); else vput(newvp); } } else *npp = np; } return (error); } /* * Strategy routine. * For async requests when nfsiod(s) are running, queue the request by * calling nfsx_asyncio(), otherwise just all nfsx_doio() to do the * request. */ int nfsx_strategy(struct vop_strategy_args *ap) { struct buf *bp = ap->a_bp; struct ucred *cr; struct thread *td; int error = 0; KASSERT(ap->a_vp == ap->a_bp->b_vp, ("%s(%p != %p)", __func__, ap->a_vp, ap->a_bp->b_vp)); KASSERT(!(bp->b_flags & B_DONE), ("nfs_strategy: buffer %p unexpectedly marked B_DONE", bp)); KASSERT(BUF_REFCNT(bp) > 0, ("nfs_strategy: buffer %p not locked", bp)); if (bp->b_flags & B_ASYNC) td = NULL; else td = curthread; /* XXX */ if (bp->b_iocmd == BIO_READ) cr = bp->b_rcred; else cr = bp->b_wcred; /* * If the op is asynchronous and an i/o daemon is waiting * queue the request, wake it up and wait for completion * otherwise just do it ourselves. */ if ((bp->b_flags & B_ASYNC) == 0 || nfsx_asyncio(bp, NOCRED, td)) error = nfsx_doio(bp, cr, td); return (error); } /* * fsync vnode op. Just call nfs_flush() with commit == 1. */ /* ARGSUSED */ int nfsx_fsync(struct vop_fsync_args *ap) { return (nfsx_flush(ap->a_vp, ap->a_cred, ap->a_waitfor, ap->a_td, 1)); } /* * Flush all the blocks associated with a vnode. * Walk through the buffer pool and push any dirty pages * associated with the vnode. */ static int nfsx_flush(struct vnode *vp, struct ucred *cred, int waitfor, struct thread *td, int commit) { struct nfsnode *np = VTONFS(vp); struct buf *bp; int i; struct buf *nbp; struct nfsmount *nmp = VFSTONFS(vp->v_mount); int s, error = 0, slptimeo = 0, slpflag = 0, retv, bvecpos; int passone = 1; u_quad_t off, endoff, toff; struct ucred* wcred = NULL; struct buf **bvec = NULL; #ifndef NFS_COMMITBVECSIZ #define NFS_COMMITBVECSIZ 20 #endif struct buf *bvec_on_stack[NFS_COMMITBVECSIZ]; int bvecsize = 0, bveccount; if (nmp->nm_flag & NFSMNT_INT) slpflag = PCATCH; if (!commit) passone = 0; /* * A b_flags == (B_DELWRI | B_NEEDCOMMIT) block has been written to the * server, but nas not been committed to stable storage on the server * yet. On the first pass, the byte range is worked out and the commit * rpc is done. On the second pass, nfs_writebp() is called to do the * job. */ again: off = (u_quad_t)-1; endoff = 0; bvecpos = 0; if (NFS_ISV3(vp) && commit) { s = splbio(); if (bvec != NULL && bvec != bvec_on_stack) free(bvec, M_TEMP); /* * Count up how many buffers waiting for a commit. */ bveccount = 0; VI_LOCK(vp); for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) { nbp = TAILQ_NEXT(bp, b_vnbufs); if (BUF_REFCNT(bp) == 0 && (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT)) == (B_DELWRI | B_NEEDCOMMIT)) bveccount++; } /* * Allocate space to remember the list of bufs to commit. It is * important to use M_NOWAIT here to avoid a race with nfs_write. * If we can't get memory (for whatever reason), we will end up * committing the buffers one-by-one in the loop below. */ if (bveccount > NFS_COMMITBVECSIZ) { /* * Release the vnode interlock to avoid a lock * order reversal. */ VI_UNLOCK(vp); bvec = (struct buf **) malloc(bveccount * sizeof(struct buf *), M_TEMP, M_NOWAIT); VI_LOCK(vp); if (bvec == NULL) { bvec = bvec_on_stack; bvecsize = NFS_COMMITBVECSIZ; } else bvecsize = bveccount; } else { bvec = bvec_on_stack; bvecsize = NFS_COMMITBVECSIZ; } for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) { if (bvecpos >= bvecsize) break; if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) { nbp = TAILQ_NEXT(bp, b_vnbufs); continue; } if ((bp->b_flags & (B_DELWRI | B_NEEDCOMMIT)) != (B_DELWRI | B_NEEDCOMMIT)) { BUF_UNLOCK(bp); nbp = TAILQ_NEXT(bp, b_vnbufs); continue; } VI_UNLOCK(vp); bremfree(bp); /* * Work out if all buffers are using the same cred * so we can deal with them all with one commit. * * NOTE: we are not clearing B_DONE here, so we have * to do it later on in this routine if we intend to * initiate I/O on the bp. * * Note: to avoid loopback deadlocks, we do not * assign b_runningbufspace. */ if (wcred == NULL) wcred = bp->b_wcred; else if (wcred != bp->b_wcred) wcred = NOCRED; bp->b_flags |= B_WRITEINPROG; vfs_busy_pages(bp, 1); VI_LOCK(vp); /* * bp is protected by being locked, but nbp is not * and vfs_busy_pages() may sleep. We have to * recalculate nbp. */ nbp = TAILQ_NEXT(bp, b_vnbufs); /* * A list of these buffers is kept so that the * second loop knows which buffers have actually * been committed. This is necessary, since there * may be a race between the commit rpc and new * uncommitted writes on the file. */ bvec[bvecpos++] = bp; toff = ((u_quad_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff; if (toff < off) off = toff; toff += (u_quad_t)(bp->b_dirtyend - bp->b_dirtyoff); if (toff > endoff) endoff = toff; } splx(s); VI_UNLOCK(vp); } if (bvecpos > 0) { /* * Commit data on the server, as required. * If all bufs are using the same wcred, then use that with * one call for all of them, otherwise commit each one * separately. */ if (wcred != NOCRED) retv = NFSOP(nmp, commit)(vp, off, (int)(endoff - off), wcred, td); else { retv = 0; for (i = 0; i < bvecpos; i++) { off_t off, size; bp = bvec[i]; off = ((u_quad_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff; size = (u_quad_t)(bp->b_dirtyend - bp->b_dirtyoff); retv = NFSOP(nmp, commit)(vp, off, (int)size, bp->b_wcred, td); if (retv) break; } } if (retv == NFSERR_STALEWRITEVERF) nfsx_clearcommit(vp->v_mount); /* * Now, either mark the blocks I/O done or mark the * blocks dirty, depending on whether the commit * succeeded. */ for (i = 0; i < bvecpos; i++) { bp = bvec[i]; bp->b_flags &= ~(B_NEEDCOMMIT | B_WRITEINPROG | B_CLUSTEROK); if (retv) { /* * Error, leave B_DELWRI intact */ vfs_unbusy_pages(bp); brelse(bp); } else { /* * Success, remove B_DELWRI ( bundirty() ). * * b_dirtyoff/b_dirtyend seem to be NFS * specific. We should probably move that * into bundirty(). XXX */ s = splbio(); VI_LOCK(vp); vp->v_numoutput++; VI_UNLOCK(vp); bp->b_flags |= B_ASYNC; bundirty(bp); bp->b_flags &= ~B_DONE; bp->b_ioflags &= ~BIO_ERROR; bp->b_dirtyoff = bp->b_dirtyend = 0; splx(s); bufdone(bp); } } } /* * Start/do any write(s) that are required. */ loop: s = splbio(); VI_LOCK(vp); for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) { nbp = TAILQ_NEXT(bp, b_vnbufs); if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) { if (waitfor != MNT_WAIT || passone) continue; error = BUF_TIMELOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK, VI_MTX(vp), "nfsfsync", slpflag, slptimeo); splx(s); if (error == 0) panic("nfs_fsync: inconsistent lock"); if (error == ENOLCK) goto loop; if (nfsx_sigintr(nmp, NULL, td)) { error = EINTR; goto done; } if (slpflag == PCATCH) { slpflag = 0; slptimeo = 2 * hz; } goto loop; } if ((bp->b_flags & B_DELWRI) == 0) panic("nfs_fsync: not dirty"); if ((passone || !commit) && (bp->b_flags & B_NEEDCOMMIT)) { BUF_UNLOCK(bp); continue; } VI_UNLOCK(vp); bremfree(bp); if (passone || !commit) bp->b_flags |= B_ASYNC; else bp->b_flags |= B_ASYNC | B_WRITEINPROG; splx(s); bwrite(bp); goto loop; } splx(s); if (passone) { passone = 0; VI_UNLOCK(vp); goto again; } if (waitfor == MNT_WAIT) { while (vp->v_numoutput) { vp->v_iflag |= VI_BWAIT; error = msleep((caddr_t)&vp->v_numoutput, VI_MTX(vp), slpflag | (PRIBIO + 1), "nfsfsync", slptimeo); if (error) { VI_UNLOCK(vp); if (nfsx_sigintr(nmp, NULL, td)) { error = EINTR; goto done; } if (slpflag == PCATCH) { slpflag = 0; slptimeo = 2 * hz; } VI_LOCK(vp); } } if (!TAILQ_EMPTY(&vp->v_dirtyblkhd) && commit) { VI_UNLOCK(vp); goto loop; } } VI_UNLOCK(vp); if (np->n_flag & NWRITEERR) { error = np->n_error; np->n_flag &= ~NWRITEERR; } done: if (bvec != NULL && bvec != bvec_on_stack) free(bvec, M_TEMP); return (error); } /* * NFS advisory byte-level locks. */ int nfsx_advlock(struct vop_advlock_args *ap) { if ((VFSTONFS(ap->a_vp->v_mount)->nm_flag & NFSMNT_NOLOCKD) != 0) { struct nfsnode *np = VTONFS(ap->a_vp); return (lf_advlock(ap, &(np->n_lockf), np->n_size)); } return (nfsx_dolock(ap)); } /* * Print out the contents of an nfsnode. */ int nfsx_print(struct vop_print_args *ap) { struct vnode *vp = ap->a_vp; struct nfsnode *np = VTONFS(vp); printf("\tfileid %ld fsid 0x%x", np->n_vattr.va_fileid, np->n_vattr.va_fsid); if (vp->v_type == VFIFO) fifo_printinfo(vp); printf("\n"); return (0); } /* * This is the "real" nfs::bwrite(struct buf*). * B_WRITEINPROG isn't set unless the force flag is one and it * handles the B_NEEDCOMMIT flag. * We set B_CACHE if this is a VMIO buffer. */ int nfsx_writebp(struct buf *bp, int force, struct thread *td) { int s; int oldflags = bp->b_flags; #if 0 int retv = 1; off_t off; #endif if (BUF_REFCNT(bp) == 0) panic("bwrite: buffer is not locked???"); if (bp->b_flags & B_INVAL) { brelse(bp); return(0); } bp->b_flags |= B_CACHE; /* * Undirty the bp. We will redirty it later if the I/O fails. */ s = splbio(); bundirty(bp); bp->b_flags &= ~B_DONE; bp->b_ioflags &= ~BIO_ERROR; bp->b_iocmd = BIO_WRITE; VI_LOCK(bp->b_vp); bp->b_vp->v_numoutput++; VI_UNLOCK(bp->b_vp); curthread->td_proc->p_stats->p_ru.ru_oublock++; splx(s); /* * Note: to avoid loopback deadlocks, we do not * assign b_runningbufspace. */ vfs_busy_pages(bp, 1); if (force) bp->b_flags |= B_WRITEINPROG; BUF_KERNPROC(bp); bp->b_iooffset = dbtob(bp->b_blkno); VOP_STRATEGY(bp->b_vp, bp); if((oldflags & B_ASYNC) == 0) { int rtval = bufwait(bp); if (oldflags & B_DELWRI) { s = splbio(); reassignbuf(bp, bp->b_vp); splx(s); } brelse(bp); return (rtval); } return (0); }