diff options
Diffstat (limited to 'sys/fs/nfsclient')
| -rw-r--r-- | sys/fs/nfsclient/nfs_clcomsubs.c | 2 | ||||
| -rw-r--r-- | sys/fs/nfsclient/nfs_clport.c | 26 | ||||
| -rw-r--r-- | sys/fs/nfsclient/nfs_clrpcops.c | 284 | ||||
| -rw-r--r-- | sys/fs/nfsclient/nfs_clstate.c | 3 | ||||
| -rw-r--r-- | sys/fs/nfsclient/nfs_clvfsops.c | 18 | ||||
| -rw-r--r-- | sys/fs/nfsclient/nfs_clvnops.c | 196 | ||||
| -rw-r--r-- | sys/fs/nfsclient/nfsmount.h | 1 |
7 files changed, 424 insertions, 106 deletions
diff --git a/sys/fs/nfsclient/nfs_clcomsubs.c b/sys/fs/nfsclient/nfs_clcomsubs.c index bca0bdcd0df1..05963074e53d 100644 --- a/sys/fs/nfsclient/nfs_clcomsubs.c +++ b/sys/fs/nfsclient/nfs_clcomsubs.c @@ -272,7 +272,7 @@ nfsm_loadattr(struct nfsrv_descript *nd, struct nfsvattr *nap) if (nd->nd_flag & ND_NFSV4) { error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, - NULL); + NULL, NULL); } else if (nd->nd_flag & ND_NFSV3) { NFSM_DISSECT(fp, struct nfs_fattr *, NFSX_V3FATTR); nap->na_type = nfsv34tov_type(fp->fa_type); diff --git a/sys/fs/nfsclient/nfs_clport.c b/sys/fs/nfsclient/nfs_clport.c index b25d967982a1..77e71d4153c9 100644 --- a/sys/fs/nfsclient/nfs_clport.c +++ b/sys/fs/nfsclient/nfs_clport.c @@ -412,7 +412,7 @@ nfscl_warn_fileid(struct nfsmount *nmp, struct nfsvattr *oldnap, } void -ncl_copy_vattr(struct vattr *dst, struct vattr *src) +ncl_copy_vattr(struct vnode *vp, struct vattr *dst, struct vattr *src) { dst->va_type = src->va_type; dst->va_mode = src->va_mode; @@ -429,7 +429,7 @@ ncl_copy_vattr(struct vattr *dst, struct vattr *src) dst->va_birthtime = src->va_birthtime; dst->va_gen = src->va_gen; dst->va_flags = src->va_flags; - dst->va_rdev = src->va_rdev; + dst->va_rdev = VN_ISDEV(vp) ? src->va_rdev : NODEV; dst->va_bytes = src->va_bytes; dst->va_filerev = src->va_filerev; } @@ -595,7 +595,7 @@ nfscl_loadattrcache(struct vnode **vpp, struct nfsvattr *nap, void *nvaper, KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp); } if (vaper != NULL) { - ncl_copy_vattr(vaper, vap); + ncl_copy_vattr(vp, vaper, vap); if (np->n_flag & NCHG) { if (np->n_flag & NACC) vaper->va_atime = np->n_atim; @@ -828,7 +828,7 @@ nfscl_wcc_data(struct nfsrv_descript *nd, struct vnode *vp, == (ND_NFSV4 | ND_V4WCCATTR)) { error = nfsv4_loadattr(nd, NULL, &nfsva, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, - NULL, NULL, NULL, NULL, NULL, NULL); + NULL, NULL, NULL, NULL, NULL, NULL, NULL); if (error) return (error); /* @@ -963,7 +963,8 @@ nfscl_loadsbinfo(struct nfsmount *nmp, struct nfsstatfs *sfp, void *statfs) * Use the fsinfo stuff to update the mount point. */ void -nfscl_loadfsinfo(struct nfsmount *nmp, struct nfsfsinfo *fsp) +nfscl_loadfsinfo(struct nfsmount *nmp, struct nfsfsinfo *fsp, + uint32_t clone_blksize) { if ((nmp->nm_wsize == 0 || fsp->fs_wtpref < nmp->nm_wsize) && @@ -1003,6 +1004,14 @@ nfscl_loadfsinfo(struct nfsmount *nmp, struct nfsfsinfo *fsp) fsp->fs_maxfilesize < nmp->nm_maxfilesize) nmp->nm_maxfilesize = fsp->fs_maxfilesize; nmp->nm_mountp->mnt_stat.f_iosize = newnfs_iosize(nmp); + + /* + * Although ZFS reports a clone_blksize of 16Mbytes, + * 128Kbytes usually works, so set it to that. + */ + if (clone_blksize > 128 * 1024) + clone_blksize = 128 * 1024; + nmp->nm_cloneblksize = clone_blksize; nmp->nm_state |= NFSSTA_GOTFSINFO; } @@ -1089,9 +1098,10 @@ newnfs_copyincred(struct ucred *cr, struct nfscred *nfscr) KASSERT(cr->cr_ngroups >= 0, ("newnfs_copyincred: negative cr_ngroups")); nfscr->nfsc_uid = cr->cr_uid; - nfscr->nfsc_ngroups = MIN(cr->cr_ngroups, NFS_MAXGRPS + 1); - for (i = 0; i < nfscr->nfsc_ngroups; i++) - nfscr->nfsc_groups[i] = cr->cr_groups[i]; + nfscr->nfsc_ngroups = MIN(cr->cr_ngroups + 1, NFS_MAXGRPS + 1); + nfscr->nfsc_groups[0] = cr->cr_gid; + for (i = 1; i < nfscr->nfsc_ngroups; i++) + nfscr->nfsc_groups[i] = cr->cr_groups[i - 1]; } /* diff --git a/sys/fs/nfsclient/nfs_clrpcops.c b/sys/fs/nfsclient/nfs_clrpcops.c index 920fcf7b8c61..f5deef183efb 100644 --- a/sys/fs/nfsclient/nfs_clrpcops.c +++ b/sys/fs/nfsclient/nfs_clrpcops.c @@ -225,6 +225,9 @@ static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *, static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *, nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *, struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *); +static int nfsrpc_clonerpc(vnode_t, off_t, vnode_t, off_t, size_t *, bool, + nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *, + struct nfsvattr *, int *, struct ucred *, NFSPROC_T *); static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *, int, struct nfsvattr *, int *, struct ucred *); static struct mbuf *nfsm_split(struct mbuf *, uint64_t); @@ -696,7 +699,7 @@ nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen, ("nfsrpc_openrpc: Getattr repstat")); error = nfsv4_loadattr(nd, NULL, &nfsva, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, - NULL, NULL, NULL, NULL, p, cred); + NULL, NULL, NULL, NULL, NULL, p, cred); if (error) goto nfsmout; } @@ -1355,7 +1358,7 @@ nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred, if ((nd->nd_flag & ND_NFSV4) != 0) error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL, - NULL, NULL, NULL); + NULL, NULL, NULL, NULL); else error = nfsm_loadattr(nd, nap); } else @@ -2209,7 +2212,7 @@ nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode, NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF); rlen = fxdr_unsigned(int, *tl++); - if (rlen == 0) { + if (rlen <= 0 || rlen > len) { error = NFSERR_IO; goto nfsmout; } else if (rlen < len) { @@ -3597,7 +3600,7 @@ nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, nfsva.na_mntonfileno = UINT64_MAX; error = nfsv4_loadattr(nd, NULL, &nfsva, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, - NULL, NULL, NULL, NULL, p, cred); + NULL, NULL, NULL, NULL, NULL, p, cred); if (error) { dotdotfileid = dotfileid; } else if (gotmnton) { @@ -3847,7 +3850,7 @@ nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, nfsva.na_mntonfileno = UINT64_MAX; error = nfsv4_loadattr(nd, NULL, &nfsva, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, - NULL, NULL, &rderr, NULL, p, cred); + NULL, NULL, &rderr, NULL, NULL, p, cred); if (error) goto nfsmout; NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED); @@ -3978,7 +3981,7 @@ nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, int len, left; struct dirent *dp = NULL; u_int32_t *tl; - vnode_t newvp = NULLVP; + vnode_t newvp = NULL; struct nfsrv_descript nfsd, *nd = &nfsd; struct nameidata nami, *ndp = &nami; struct componentname *cnp = &ndp->ni_cnd; @@ -4072,7 +4075,7 @@ nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, nfsva.na_mntonfileno = UINT64_MAX; error = nfsv4_loadattr(nd, NULL, &nfsva, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, - NULL, NULL, NULL, NULL, p, cred); + NULL, NULL, NULL, NULL, NULL, p, cred); if (error) { dotdotfileid = dotfileid; } else if (gotmnton) { @@ -4159,9 +4162,12 @@ nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, NFSATTRBIT_TIMECREATE)) NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE); if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr, + NFSATTRBIT_ARCHIVE) || + !NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr, NFSATTRBIT_HIDDEN) || !NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr, NFSATTRBIT_SYSTEM)) { + NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_ARCHIVE); NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_HIDDEN); NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_SYSTEM); } @@ -4346,7 +4352,7 @@ nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, nfsva.na_mntonfileno = 0xffffffff; error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, - NULL, NULL, &rderr, NULL, p, cred); + NULL, NULL, &rderr, NULL, NULL, p, cred); if (error) goto nfsmout; } @@ -4384,7 +4390,7 @@ nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, attr_ok = true; if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len, dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) { - VREF(vp); + vref(vp); newvp = vp; unlocknewvp = 0; free(nfhp, M_NFSFH); @@ -4433,7 +4439,7 @@ nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, } } nfhp = NULL; - if (newvp != NULLVP) { + if (newvp != NULL) { if (attr_ok) error = nfscl_loadattrcache(&newvp, &nfsva, NULL, 0, 0); @@ -4463,7 +4469,7 @@ nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep, vput(newvp); else vrele(newvp); - newvp = NULLVP; + newvp = NULL; } } } else if (nfhp != NULL) { @@ -4981,8 +4987,8 @@ nfsmout: */ int nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp, - uint32_t *leasep, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, - int *attrflagp) + uint32_t *leasep, uint32_t *cloneblksizep, struct ucred *cred, NFSPROC_T *p, + struct nfsvattr *nap, int *attrflagp) { u_int32_t *tl = NULL; struct nfsrv_descript nfsd, *nd = &nfsd; @@ -4991,6 +4997,8 @@ nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp, int error; *attrflagp = 0; + if (cloneblksizep != NULL) + *cloneblksizep = 0; nmp = VFSTONFS(vp->v_mount); if (NFSHASNFSV4(nmp)) { /* @@ -5009,7 +5017,7 @@ nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp, if (nd->nd_repstat == 0) { error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL, sbp, fsp, NULL, 0, NULL, leasep, NULL, - NULL, p, cred); + NULL, cloneblksizep, p, cred); if (!error) { nmp->nm_fsid[0] = nap->na_filesid[0]; nmp->nm_fsid[1] = nap->na_filesid[1]; @@ -5063,7 +5071,8 @@ nfsmout: */ int nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc, bool *has_namedattrp, - struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp) + uint32_t *clone_blksizep, struct ucred *cred, NFSPROC_T *p, + struct nfsvattr *nap, int *attrflagp) { struct nfsrv_descript nfsd, *nd = &nfsd; struct nfsmount *nmp; @@ -5074,6 +5083,7 @@ nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc, bool *has_namedattrp, *has_namedattrp = false; *attrflagp = 0; + *clone_blksizep = 0; nmp = VFSTONFS(vp->v_mount); if (NFSHASNFSV4(nmp)) { np = VTONFS(vp); @@ -5100,7 +5110,7 @@ nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc, bool *has_namedattrp, if (nd->nd_repstat == 0) { error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, - has_namedattrp, p, cred); + has_namedattrp, clone_blksizep, p, cred); if (!error) *attrflagp = 1; } else { @@ -5274,7 +5284,7 @@ nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred, struct nfsrv_descript nfsd; struct nfsrv_descript *nd = &nfsd; u_char *cp, *cp2, *fhp; - int error, cnt, len, setnil; + int error, cnt, i, len, setnil; u_int32_t *opcntp; nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0, @@ -5315,8 +5325,12 @@ nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred, if (error) return (error); if (nd->nd_repstat == 0) { - NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED); - tl += (2 + 2 * cnt); + NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED); + tl += 2; + for (i = 0; i < cnt; i++) { + NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED); + tl++; + } if ((len = fxdr_unsigned(int, *tl)) <= 0 || len > NFSX_FHMAX) { nd->nd_repstat = NFSERR_BADXDR; @@ -5395,7 +5409,8 @@ nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp) return (error); if (!nd->nd_repstat) error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL, - NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, NULL, p, cred); + NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, NULL, NULL, p, + cred); else error = nd->nd_repstat; m_freem(nd->nd_mrep); @@ -5437,7 +5452,7 @@ nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p, NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL); (void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0, &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL, false, false, - false); + false, 0, NULL, false); error = nfscl_request(nd, vp, p, cred); if (error) return (error); @@ -5588,7 +5603,7 @@ nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep, } *tl++ = txdr_unsigned(4096); /* Max response size cached */ *tl++ = txdr_unsigned(20); /* Max operations */ - *tl++ = txdr_unsigned(64); /* Max slots */ + *tl++ = txdr_unsigned(NFSV4_SLOTS); /* Max slots */ *tl = 0; /* No rdma ird */ /* Fill in back channel attributes. */ @@ -5657,6 +5672,11 @@ nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep, sep->nfsess_maxcache = fxdr_unsigned(int, *tl++); tl++; sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++); + if (sep->nfsess_foreslots == 0) { + error = NFSERR_BADXDR; + goto nfsmout; + } else if (sep->nfsess_foreslots > NFSV4_SLOTS) + sep->nfsess_foreslots = NFSV4_SLOTS; NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots); irdcnt = fxdr_unsigned(int, *tl); if (irdcnt < 0 || irdcnt > 1) { @@ -5670,6 +5690,8 @@ nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep, NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED); tl += 5; sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl); + if (sep->nfsess_backslots > NFSV4_CBSLOTS) + sep->nfsess_backslots = NFSV4_CBSLOTS; NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots); } error = nd->nd_repstat; @@ -5789,7 +5811,8 @@ nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype, NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED); stripecnt = fxdr_unsigned(int, *tl); NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt); - if (stripecnt < 1 || stripecnt > 4096) { + if (stripecnt >= MHLEN / NFSX_UNSIGNED || + stripecnt < 1) { printf("pNFS File layout devinfo stripecnt %d:" " out of range\n", stripecnt); error = NFSERR_BADXDR; @@ -7238,7 +7261,7 @@ nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit, NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF); rlen = fxdr_unsigned(int, *tl++); NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen); - if (rlen == 0) { + if (rlen <= 0 || rlen > len) { error = NFSERR_IO; goto nfsmout; } else if (rlen < len) { @@ -8235,7 +8258,7 @@ nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp, NFSPROC_T *p) { uint32_t *tl; - char *cp, *str, str0[NFSV4_SMALLSTR + 1]; + char *str, str0[NFSV4_SMALLSTR + 1]; uint32_t len = 0; int error = 0; @@ -8258,9 +8281,9 @@ nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp, str = malloc(len + 1, M_TEMP, M_WAITOK); else str = str0; - NFSM_DISSECT(cp, char *, NFSM_RNDUP(len)); - NFSBCOPY(cp, str, len); - str[len] = '\0'; + error = nfsrv_mtostr(nd, str, len); + if (error != 0) + goto nfsmout; NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str); if (dogrp != 0) error = nfsv4_strtogid(nd, str, len, gidp); @@ -8496,7 +8519,7 @@ nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, if (*++tl == 0) { error = nfsv4_loadattr(nd, NULL, &nfsva, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL, 0, - NULL, NULL, NULL, NULL, p, cred); + NULL, NULL, NULL, NULL, NULL, p, cred); if (error != 0) goto nfsmout; if (ndp != NULL) { @@ -9168,6 +9191,199 @@ nfsmout: } /* + * nfs clone operation. + */ +int +nfsrpc_clone(vnode_t invp, off_t *inoffp, vnode_t outvp, + off_t *outoffp, size_t *lenp, bool toeof, int *inattrflagp, + struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap, + struct ucred *cred) +{ + int error, expireret = 0, retrycnt; + uint32_t clidrev = 0; + struct nfsmount *nmp = VFSTONFS(invp->v_mount); + struct nfsfh *innfhp = NULL, *outnfhp = NULL; + nfsv4stateid_t instateid, outstateid; + void *inlckp, *outlckp; + + if (nmp->nm_clp != NULL) + clidrev = nmp->nm_clp->nfsc_clientidrev; + innfhp = VTONFS(invp)->n_fhp; + outnfhp = VTONFS(outvp)->n_fhp; + retrycnt = 0; + do { + /* Get both stateids. */ + inlckp = NULL; + nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len, + NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid, + &inlckp); + outlckp = NULL; + nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len, + NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid, + &outlckp); + + error = nfsrpc_clonerpc(invp, *inoffp, outvp, *outoffp, lenp, + toeof, &instateid, &outstateid, innap, inattrflagp, outnap, + outattrflagp, cred, curthread); + if (error == 0) { + *inoffp += *lenp; + *outoffp += *lenp; + } else if (error == NFSERR_STALESTATEID) + nfscl_initiate_recovery(nmp->nm_clp); + if (inlckp != NULL) + nfscl_lockderef(inlckp); + if (outlckp != NULL) + nfscl_lockderef(outlckp); + if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID || + error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY || + error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) { + (void) nfs_catnap(PZERO, error, "nfs_cfr"); + } else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) && + error == NFSERR_BADSTATEID)) && clidrev != 0) { + expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, + curthread); + } else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) { + error = EIO; + } + retrycnt++; + } while (error == NFSERR_GRACE || error == NFSERR_DELAY || + error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION || + error == NFSERR_STALEDONTRECOVER || + (error == NFSERR_OLDSTATEID && retrycnt < 20) || + ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) && + expireret == 0 && clidrev != 0 && retrycnt < 4)); + if (error != 0 && (retrycnt >= 4 || + error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION || + error == NFSERR_STALEDONTRECOVER)) + error = EIO; + return (error); +} + +/* + * The clone RPC. + */ +static int +nfsrpc_clonerpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff, + size_t *lenp, bool toeof, nfsv4stateid_t *instateidp, + nfsv4stateid_t *outstateidp, struct nfsvattr *innap, int *inattrflagp, + struct nfsvattr *outnap, int *outattrflagp, struct ucred *cred, + NFSPROC_T *p) +{ + uint32_t *tl, *opcntp; + int error; + struct nfsrv_descript nfsd; + struct nfsrv_descript *nd = &nfsd; + struct nfsmount *nmp; + nfsattrbit_t attrbits; + struct vattr va; + uint64_t len; + + nmp = VFSTONFS(invp->v_mount); + *inattrflagp = *outattrflagp = 0; + len = *lenp; + if (len == 0) + return (0); + if (toeof) + len = 0; + nfscl_reqstart(nd, NFSPROC_CLONE, nmp, VTONFS(invp)->n_fhp->nfh_fh, + VTONFS(invp)->n_fhp->nfh_len, &opcntp, NULL, 0, 0, cred); + /* + * First do a Setattr of atime to the server's clock + * time. The FreeBSD "collective" was of the opinion + * that setting atime was necessary for this syscall. + * Do the Setattr before the Clone, so that it can be + * handled well if the server replies NFSERR_DELAY to + * the Setattr operation. + */ + if ((nmp->nm_mountp->mnt_flag & MNT_NOATIME) == 0) { + NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV4OP_SETATTR); + nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID); + VATTR_NULL(&va); + va.va_atime.tv_sec = va.va_atime.tv_nsec = 0; + va.va_vaflags = VA_UTIMES_NULL; + nfscl_fillsattr(nd, &va, invp, 0, 0); + /* Bump opcnt from 7 to 8. */ + *opcntp = txdr_unsigned(8); + } + + /* Now Getattr the invp attributes. */ + NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV4OP_GETATTR); + NFSGETATTR_ATTRBIT(&attrbits); + nfsrv_putattrbit(nd, &attrbits); + + /* Set outvp. */ + NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV4OP_PUTFH); + (void)nfsm_fhtom(nmp, nd, VTONFS(outvp)->n_fhp->nfh_fh, + VTONFS(outvp)->n_fhp->nfh_len, 0); + + /* Do the Clone. */ + NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV4OP_CLONE); + nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID); + nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID); + NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + NFSX_UNSIGNED); + txdr_hyper(inoff, tl); tl += 2; + txdr_hyper(outoff, tl); tl += 2; + txdr_hyper(len, tl); tl += 2; + + /* Get the outvp attributes. */ + *tl = txdr_unsigned(NFSV4OP_GETATTR); + NFSWRITEGETATTR_ATTRBIT(&attrbits); + nfsrv_putattrbit(nd, &attrbits); + + error = nfscl_request(nd, invp, p, cred); + if (error != 0) + return (error); + /* Skip over the Setattr reply. */ + if ((nd->nd_flag & ND_NOMOREDATA) == 0 && + (nmp->nm_mountp->mnt_flag & MNT_NOATIME) == 0) { + NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED); + if (*(tl + 1) == 0) { + error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL); + if (error != 0) + goto nfsmout; + } else + nd->nd_flag |= ND_NOMOREDATA; + } + if ((nd->nd_flag & ND_NOMOREDATA) == 0) { + /* Get the input file's attributes. */ + NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED); + if (*(tl + 1) == 0) { + error = nfsm_loadattr(nd, innap); + if (error != 0) + goto nfsmout; + *inattrflagp = 1; + } else + nd->nd_flag |= ND_NOMOREDATA; + } + /* Skip over return stat for PutFH. */ + if ((nd->nd_flag & ND_NOMOREDATA) == 0) { + NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED); + if (*++tl != 0) + nd->nd_flag |= ND_NOMOREDATA; + } + /* Skip over return stat for Clone. */ + if ((nd->nd_flag & ND_NOMOREDATA) == 0) + NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED); + if (nd->nd_repstat == 0) { + NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED); + error = nfsm_loadattr(nd, outnap); + if (error == 0) + *outattrflagp = NFS_LATTR_NOSHRINK; + } else { + *lenp = 0; + } + if (error == 0) + error = nd->nd_repstat; +nfsmout: + m_freem(nd->nd_mrep); + return (error); +} + +/* * Seek operation. */ int @@ -9544,7 +9760,7 @@ nfsm_split(struct mbuf *mp, uint64_t xfer) pgno++; } while (pgno < m->m_epg_npgs); if (pgno == m->m_epg_npgs) - panic("nfsm_split: eroneous ext_pgs mbuf"); + panic("nfsm_split: erroneous ext_pgs mbuf"); m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs, 0); m2->m_epg_flags |= EPG_FLAG_ANON; @@ -9724,13 +9940,13 @@ nfscl_statfs(struct vnode *vp, struct ucred *cred, NFSPROC_T *td) struct nfsstatfs sb; struct mount *mp; struct nfsmount *nmp; - uint32_t lease; + uint32_t clone_blksize, lease; int attrflag, error; mp = vp->v_mount; nmp = VFSTONFS(mp); - error = nfsrpc_statfs(vp, &sb, &fs, &lease, cred, td, &nfsva, - &attrflag); + error = nfsrpc_statfs(vp, &sb, &fs, &lease, &clone_blksize, cred, td, + &nfsva, &attrflag); if (attrflag != 0) (void) nfscl_loadattrcache(&vp, &nfsva, NULL, 0, 1); if (error == 0) { @@ -9739,7 +9955,7 @@ nfscl_statfs(struct vnode *vp, struct ucred *cred, NFSPROC_T *td) nmp->nm_clp->nfsc_renew = NFSCL_RENEW(lease); NFSUNLOCKCLSTATE(); mtx_lock(&nmp->nm_mtx); - nfscl_loadfsinfo(nmp, &fs); + nfscl_loadfsinfo(nmp, &fs, clone_blksize); nfscl_loadsbinfo(nmp, &sb, &mp->mnt_stat); mp->mnt_stat.f_iosize = newnfs_iosize(nmp); mtx_unlock(&nmp->nm_mtx); diff --git a/sys/fs/nfsclient/nfs_clstate.c b/sys/fs/nfsclient/nfs_clstate.c index 99a781640c53..712d49c7160c 100644 --- a/sys/fs/nfsclient/nfs_clstate.c +++ b/sys/fs/nfsclient/nfs_clstate.c @@ -3701,7 +3701,8 @@ nfscl_docb(struct nfsrv_descript *nd, NFSPROC_T *p) if (!error) (void) nfsv4_fillattr(nd, NULL, NULL, NULL, &va, NULL, 0, &rattrbits, NULL, p, 0, 0, 0, 0, - (uint64_t)0, NULL, false, false, false); + (uint64_t)0, NULL, false, false, false, 0, + NULL, false); break; case NFSV4OP_CBRECALL: NFSCL_DEBUG(4, "cbrecall\n"); diff --git a/sys/fs/nfsclient/nfs_clvfsops.c b/sys/fs/nfsclient/nfs_clvfsops.c index 0bd05c03885b..212c88f28930 100644 --- a/sys/fs/nfsclient/nfs_clvfsops.c +++ b/sys/fs/nfsclient/nfs_clvfsops.c @@ -292,8 +292,10 @@ nfs_statfs(struct mount *mp, struct statfs *sbp) int error = 0, attrflag, gotfsinfo = 0, ret; struct nfsnode *np; char *fakefh; + uint32_t clone_blksize; td = curthread; + clone_blksize = 0; error = vfs_busy(mp, MBF_NOWAIT); if (error) @@ -337,8 +339,8 @@ nfs_statfs(struct mount *mp, struct statfs *sbp) } else mtx_unlock(&nmp->nm_mtx); if (!error) - error = nfsrpc_statfs(vp, &sb, &fs, NULL, td->td_ucred, td, - &nfsva, &attrflag); + error = nfsrpc_statfs(vp, &sb, &fs, NULL, &clone_blksize, + td->td_ucred, td, &nfsva, &attrflag); if ((nmp->nm_privflag & NFSMNTP_FAKEROOTFH) != 0 && error == NFSERR_WRONGSEC) { /* Cannot get new stats, so return what is in mnt_stat. */ @@ -375,7 +377,7 @@ nfs_statfs(struct mount *mp, struct statfs *sbp) if (!error) { mtx_lock(&nmp->nm_mtx); if (gotfsinfo || (nmp->nm_flag & NFSMNT_NFSV4)) - nfscl_loadfsinfo(nmp, &fs); + nfscl_loadfsinfo(nmp, &fs, clone_blksize); nfscl_loadsbinfo(nmp, &sb, sbp); sbp->f_iosize = newnfs_iosize(nmp); mtx_unlock(&nmp->nm_mtx); @@ -408,7 +410,7 @@ ncl_fsinfo(struct nfsmount *nmp, struct vnode *vp, struct ucred *cred, if (attrflag) (void) nfscl_loadattrcache(&vp, &nfsva, NULL, 0, 1); mtx_lock(&nmp->nm_mtx); - nfscl_loadfsinfo(nmp, &fs); + nfscl_loadfsinfo(nmp, &fs, 0); mtx_unlock(&nmp->nm_mtx); } return (error); @@ -925,7 +927,7 @@ nfs_mount(struct mount *mp) struct vnode *vp; struct thread *td; char *hst; - u_char nfh[NFSX_FHMAX], krbname[100], dirpath[100], srvkrbname[100]; + u_char nfh[NFSX_FHMAX], krbname[100], *dirpath, srvkrbname[100]; char *cp, *opt, *name, *secname, *tlscertname; int nametimeo = NFS_DEFAULT_NAMETIMEO; int negnametimeo = NFS_DEFAULT_NEGNAMETIMEO; @@ -941,6 +943,7 @@ nfs_mount(struct mount *mp) newflag = 0; tlscertname = NULL; hst = malloc(MNAMELEN, M_TEMP, M_WAITOK); + dirpath = malloc(MNAMELEN, M_TEMP, M_WAITOK); if (vfs_filteropt(mp->mnt_optnew, nfs_opts)) { error = EINVAL; goto out; @@ -1327,7 +1330,7 @@ nfs_mount(struct mount *mp) goto out; } else if (nfs_mount_parse_from(mp->mnt_optnew, &args.hostname, (struct sockaddr_in **)&nam, dirpath, - sizeof(dirpath), &dirlen) == 0) { + MNAMELEN, &dirlen) == 0) { has_nfs_from_opt = 1; bcopy(args.hostname, hst, MNAMELEN); hst[MNAMELEN - 1] = '\0'; @@ -1385,7 +1388,7 @@ nfs_mount(struct mount *mp) if (has_nfs_from_opt == 0) { if (vfs_getopt(mp->mnt_optnew, "dirpath", (void **)&name, NULL) == 0) - strlcpy(dirpath, name, sizeof (dirpath)); + strlcpy(dirpath, name, MNAMELEN); else dirpath[0] = '\0'; dirlen = strlen(dirpath); @@ -1470,6 +1473,7 @@ out: MNT_IUNLOCK(mp); } free(hst, M_TEMP); + free(dirpath, M_TEMP); return (error); } diff --git a/sys/fs/nfsclient/nfs_clvnops.c b/sys/fs/nfsclient/nfs_clvnops.c index fa451887e73e..193d8b6cd5eb 100644 --- a/sys/fs/nfsclient/nfs_clvnops.c +++ b/sys/fs/nfsclient/nfs_clvnops.c @@ -1026,7 +1026,7 @@ nfs_getattr(struct vop_getattr_args *ap) * cached attributes should be ignored. */ if (nmp->nm_fhsize > 0 && ncl_getattrcache(vp, &vattr) == 0) { - ncl_copy_vattr(vap, &vattr); + ncl_copy_vattr(vp, vap, &vattr); /* * Get the local modify time for the case of a write @@ -1081,12 +1081,14 @@ nfs_setattr(struct vop_setattr_args *ap) #endif /* - * Only setting of UF_HIDDEN and UF_SYSTEM are supported and + * Only setting of UF_ARCHIVE, UF_HIDDEN and UF_SYSTEM are supported and * only for NFSv4 servers that support them. */ nmp = VFSTONFS(vp->v_mount); if (vap->va_flags != VNOVAL && (!NFSHASNFSV4(nmp) || - (vap->va_flags & ~(UF_HIDDEN | UF_SYSTEM)) != 0 || + (vap->va_flags & ~(UF_ARCHIVE | UF_HIDDEN | UF_SYSTEM)) != 0 || + ((vap->va_flags & UF_ARCHIVE) != 0 && + !NFSISSET_ATTRBIT(&np->n_vattr.na_suppattr, NFSATTRBIT_ARCHIVE)) || ((vap->va_flags & UF_HIDDEN) != 0 && !NFSISSET_ATTRBIT(&np->n_vattr.na_suppattr, NFSATTRBIT_HIDDEN)) || ((vap->va_flags & UF_SYSTEM) != 0 && @@ -1284,7 +1286,7 @@ nfs_lookup(struct vop_lookup_args *ap) bool is_nameddir, needs_nameddir, opennamed; dattrflag = 0; - *vpp = NULLVP; + *vpp = NULL; nmp = VFSTONFS(mp); opennamed = (flags & (OPENNAMED | ISLASTCN)) == (OPENNAMED | ISLASTCN); if (opennamed && (!NFSHASNFSV4(nmp) || !NFSHASNFSV4N(nmp))) @@ -1309,7 +1311,7 @@ nfs_lookup(struct vop_lookup_args *ap) /* * If the named attribute directory is needed, acquire it now. */ - newvp = NULLVP; + newvp = NULL; if (needs_nameddir) { KASSERT(np->n_v4 == NULL, ("nfs_lookup: O_NAMEDATTR when" " n_v4 not NULL")); @@ -1322,10 +1324,10 @@ nfs_lookup(struct vop_lookup_args *ap) } dvp = newvp; np = VTONFS(dvp); - newvp = NULLVP; + newvp = NULL; } else if (opennamed && cnp->cn_namelen == 1 && *cnp->cn_nameptr == '.') { - VREF(dvp); + vref(dvp); *vpp = dvp; return (0); } @@ -1399,7 +1401,7 @@ nfs_lookup(struct vop_lookup_args *ap) vput(newvp); else vrele(newvp); - *vpp = NULLVP; + *vpp = NULL; } else if (error == ENOENT) { if (VN_IS_DOOMED(dvp)) return (ENOENT); @@ -1450,7 +1452,7 @@ nfs_lookup(struct vop_lookup_args *ap) NFSUNLOCKMNT(nmp); #endif - newvp = NULLVP; + newvp = NULL; NFSINCRGLOBAL(nfsstatsv1.lookupcache_misses); nanouptime(&ts); error = nfsrpc_lookup(dvp, cnp->cn_nameptr, cnp->cn_namelen, @@ -1464,9 +1466,9 @@ nfs_lookup(struct vop_lookup_args *ap) } handle_error: if (error) { - if (newvp != NULLVP) { + if (newvp != NULL) { vput(newvp); - *vpp = NULLVP; + *vpp = NULL; } if (error != ENOENT) { @@ -1587,7 +1589,7 @@ handle_error: 0, 1); } else if (NFS_CMPFH(np, nfhp->nfh_fh, nfhp->nfh_len)) { free(nfhp, M_NFSFH); - VREF(dvp); + vref(dvp); newvp = dvp; if (attrflag) (void) nfscl_loadattrcache(&newvp, &nfsva, NULL, @@ -1782,7 +1784,7 @@ nfs_mknodrpc(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp, int error = 0, attrflag, dattrflag; u_int32_t rdev; - if (vap->va_type == VCHR || vap->va_type == VBLK) + if (VATTR_ISDEV(vap)) rdev = vap->va_rdev; else if (vap->va_type == VFIFO || vap->va_type == VSOCK) rdev = 0xffffffff; @@ -2863,7 +2865,7 @@ nfs_sillyrename(struct vnode *dvp, struct vnode *vp, struct componentname *cnp) M_NEWNFSREQ, M_WAITOK); sp->s_cred = crhold(cnp->cn_cred); sp->s_dvp = dvp; - VREF(dvp); + vref(dvp); /* * Fudge together a funny name. @@ -2961,7 +2963,7 @@ nfs_lookitup(struct vnode *dvp, char *name, int len, struct ucred *cred, newvp = NFSTOV(np); } else if (NFS_CMPFH(dnp, nfhp->nfh_fh, nfhp->nfh_len)) { free(nfhp, M_NFSFH); - VREF(dvp); + vref(dvp); newvp = dvp; } else { cn.cn_nameptr = name; @@ -3474,7 +3476,7 @@ nfs_advlock(struct vop_advlock_args *ap) u_quad_t size; struct nfsmount *nmp; - error = NFSVOPLOCK(vp, LK_SHARED); + error = NFSVOPLOCK(vp, LK_EXCLUSIVE); if (error != 0) return (EBADF); nmp = VFSTONFS(vp->v_mount); @@ -3511,11 +3513,6 @@ nfs_advlock(struct vop_advlock_args *ap) cred = p->p_ucred; else cred = td->td_ucred; - NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY); - if (VN_IS_DOOMED(vp)) { - error = EBADF; - goto out; - } /* * If this is unlocking a write locked region, flush and @@ -3899,11 +3896,15 @@ nfs_allocate(struct vop_allocate_args *ap) mtx_lock(&nmp->nm_mtx); nmp->nm_privflag |= NFSMNTP_NOALLOCATE; mtx_unlock(&nmp->nm_mtx); - error = EINVAL; + error = EOPNOTSUPP; } } else { + /* + * Pre-v4.2 NFS server that doesn't support it, or a newer + * NFS server that has indicated that it doesn't support it. + */ mtx_unlock(&nmp->nm_mtx); - error = EINVAL; + error = EOPNOTSUPP; } if (attrflag != 0) { ret = nfscl_loadattrcache(&vp, &nfsva, NULL, 0, 1); @@ -4027,31 +4028,51 @@ nfs_copy_file_range(struct vop_copy_file_range_args *ap) struct vattr va, *vap; struct uio io; struct nfsmount *nmp; + struct nfsnode *np; size_t len, len2; ssize_t r; int error, inattrflag, outattrflag, ret, ret2, invp_lock; off_t inoff, outoff; - bool consecutive, must_commit, tryoutcred; + bool consecutive, must_commit, onevp, toeof, tryclone, tryoutcred; + bool mustclone; /* * NFSv4.2 Copy is not permitted for infile == outfile. + * The NFSv4.2 Clone operation does work on non-overlapping + * byte ranges in the same file, but only if offsets + * (and len if not to EOF) are aligned properly. * TODO: copy_file_range() between multiple NFS mountpoints + * --> This is not possible now, since each mount appears to + * the NFSv4.n server as a separate client. */ - if (invp == outvp || invp->v_mount != outvp->v_mount) { + if ((invp == outvp && (ap->a_flags & COPY_FILE_RANGE_CLONE) == 0) || + (invp != outvp && invp->v_mount != outvp->v_mount)) { generic_copy: return (ENOSYS); } - - invp_lock = LK_SHARED; + if (invp == outvp) { + onevp = true; + invp_lock = LK_EXCLUSIVE; + } else { + onevp = false; + invp_lock = LK_SHARED; + } + mustclone = false; + if (onevp || (ap->a_flags & COPY_FILE_RANGE_CLONE) != 0) + mustclone = true; relock: + inoff = *ap->a_inoffp; + outoff = *ap->a_outoffp; - /* Lock both vnodes, avoiding risk of deadlock. */ + /* Lock vnode(s), avoiding risk of deadlock. */ do { mp = NULL; error = vn_start_write(outvp, &mp, V_WAIT); if (error == 0) { error = vn_lock(outvp, LK_EXCLUSIVE); if (error == 0) { + if (onevp) + break; error = vn_lock(invp, invp_lock | LK_NOWAIT); if (error == 0) break; @@ -4071,16 +4092,24 @@ relock: return (error); /* - * More reasons to avoid nfs copy: not NFSv4.2, or explicitly - * disabled. + * More reasons to avoid nfs copy/clone: not NFSv4.2, explicitly + * disabled or requires cloning and unable to clone. + * Only clone if the clone_blksize attribute is supported + * and the clone_blksize is greater than 0. + * Alignment of offsets and length will be checked later. */ nmp = VFSTONFS(invp->v_mount); + np = VTONFS(invp); mtx_lock(&nmp->nm_mtx); + if ((nmp->nm_privflag & NFSMNTP_NOCOPY) != 0) + mustclone = true; if (!NFSHASNFSV4(nmp) || nmp->nm_minorvers < NFSV42_MINORVERSION || - (nmp->nm_privflag & NFSMNTP_NOCOPY) != 0) { + (mustclone && (!NFSISSET_ATTRBIT(&np->n_vattr.na_suppattr, + NFSATTRBIT_CLONEBLKSIZE) || nmp->nm_cloneblksize == 0))) { mtx_unlock(&nmp->nm_mtx); VOP_UNLOCK(invp); - VOP_UNLOCK(outvp); + if (!onevp) + VOP_UNLOCK(outvp); /* For onevp, same as invp. */ if (mp != NULL) vn_finished_write(mp); goto generic_copy; @@ -4111,6 +4140,8 @@ relock: invp_obj = invp->v_object; if (invp_obj != NULL && vm_object_mightbedirty(invp_obj)) { if (invp_lock != LK_EXCLUSIVE) { + KASSERT(!onevp, ("nfs_copy_file_range: " + "invp_lock LK_SHARED for onevp")); invp_lock = LK_EXCLUSIVE; VOP_UNLOCK(invp); VOP_UNLOCK(outvp); @@ -4134,10 +4165,10 @@ relock: else consecutive = false; mtx_unlock(&nmp->nm_mtx); - inoff = *ap->a_inoffp; - outoff = *ap->a_outoffp; tryoutcred = true; must_commit = false; + toeof = false; + if (error == 0) { vap = &VTONFS(invp)->n_vattr.na_vattr; error = VOP_GETATTR(invp, vap, ap->a_incred); @@ -4169,29 +4200,63 @@ relock: if (error == 0 && ret != 0) error = ret; } - } else if (inoff + len > vap->va_size) + } else if (inoff + len >= vap->va_size) { + toeof = true; *ap->a_lenp = len = vap->va_size - inoff; + } } else error = 0; } /* + * For cloning, the offsets must be clone blksize aligned and + * the len must be blksize aligned unless it goes to EOF on + * the input file. + */ + tryclone = false; + if (len > 0) { + if (error == 0 && NFSISSET_ATTRBIT(&np->n_vattr.na_suppattr, + NFSATTRBIT_CLONEBLKSIZE) && nmp->nm_cloneblksize != 0 && + (inoff % nmp->nm_cloneblksize) == 0 && + (outoff % nmp->nm_cloneblksize) == 0 && + (toeof || (len % nmp->nm_cloneblksize) == 0)) + tryclone = true; + else if (mustclone) + error = ENOSYS; + } + + /* * len will be set to 0 upon a successful Copy RPC. - * As such, this only loops when the Copy RPC needs to be retried. + * As such, this only loops when the Copy/Clone RPC needs to be retried. */ while (len > 0 && error == 0) { inattrflag = outattrflag = 0; len2 = len; - if (tryoutcred) - error = nfsrpc_copy_file_range(invp, ap->a_inoffp, - outvp, ap->a_outoffp, &len2, ap->a_flags, - &inattrflag, &innfsva, &outattrflag, &outnfsva, - ap->a_outcred, consecutive, &must_commit); - else - error = nfsrpc_copy_file_range(invp, ap->a_inoffp, - outvp, ap->a_outoffp, &len2, ap->a_flags, - &inattrflag, &innfsva, &outattrflag, &outnfsva, - ap->a_incred, consecutive, &must_commit); + if (tryclone) { + if (tryoutcred) + error = nfsrpc_clone(invp, ap->a_inoffp, outvp, + ap->a_outoffp, &len2, toeof, &inattrflag, + &innfsva, &outattrflag, &outnfsva, + ap->a_outcred); + else + error = nfsrpc_clone(invp, ap->a_inoffp, outvp, + ap->a_outoffp, &len2, toeof, &inattrflag, + &innfsva, &outattrflag, &outnfsva, + ap->a_incred); + } else { + if (tryoutcred) + error = nfsrpc_copy_file_range(invp, + ap->a_inoffp, outvp, ap->a_outoffp, &len2, + ap->a_flags, &inattrflag, &innfsva, + &outattrflag, &outnfsva, + ap->a_outcred, consecutive, &must_commit); + else + error = nfsrpc_copy_file_range(invp, + ap->a_inoffp, outvp, ap->a_outoffp, &len2, + ap->a_flags, &inattrflag, &innfsva, + &outattrflag, &outnfsva, + ap->a_incred, consecutive, &must_commit); + } if (inattrflag != 0) ret = nfscl_loadattrcache(&invp, &innfsva, NULL, 0, 1); if (outattrflag != 0) @@ -4230,6 +4295,13 @@ relock: /* Try again with incred. */ tryoutcred = false; error = 0; + } else if (tryclone && error != 0) { + if (mustclone) { + error = ENOSYS; + } else { + tryclone = false; + error = 0; + } } if (error == NFSERR_STALEWRITEVERF) { /* @@ -4243,11 +4315,12 @@ relock: } } VOP_UNLOCK(invp); - VOP_UNLOCK(outvp); + if (!onevp) + VOP_UNLOCK(outvp); /* For onevp, same as invp. */ if (mp != NULL) vn_finished_write(mp); if (error == NFSERR_NOTSUPP || error == NFSERR_OFFLOADNOREQS || - error == NFSERR_ACCES) { + error == NFSERR_ACCES || error == ENOSYS) { /* * Unlike the NFSv4.2 Copy, vn_generic_copy_file_range() can * use a_incred for the read and a_outcred for the write, so @@ -4255,7 +4328,7 @@ relock: * For NFSERR_NOTSUPP and NFSERR_OFFLOADNOREQS, the Copy can * never succeed, so disable it. */ - if (error != NFSERR_ACCES) { + if (error != NFSERR_ACCES && error != ENOSYS) { /* Can never do Copy on this mount. */ mtx_lock(&nmp->nm_mtx); nmp->nm_privflag |= NFSMNTP_NOCOPY; @@ -4596,6 +4669,7 @@ nfs_pathconf(struct vop_pathconf_args *ap) struct nfsmount *nmp; struct thread *td = curthread; off_t off; + uint32_t clone_blksize; bool eof, has_namedattr, named_enabled; int attrflag, error; struct nfsnode *np; @@ -4604,19 +4678,23 @@ nfs_pathconf(struct vop_pathconf_args *ap) np = VTONFS(vp); named_enabled = false; has_namedattr = false; + clone_blksize = 0; if ((NFS_ISV34(vp) && (ap->a_name == _PC_LINK_MAX || ap->a_name == _PC_NAME_MAX || ap->a_name == _PC_CHOWN_RESTRICTED || - ap->a_name == _PC_NO_TRUNC)) || + ap->a_name == _PC_NO_TRUNC || + ap->a_name == _PC_CASE_INSENSITIVE)) || (NFS_ISV4(vp) && (ap->a_name == _PC_ACL_NFS4 || - ap->a_name == _PC_HAS_NAMEDATTR))) { + ap->a_name == _PC_HAS_NAMEDATTR || + ap->a_name == _PC_CLONE_BLKSIZE))) { /* - * Since only the above 4 a_names are returned by the NFSv3 + * Since only the above 5 a_names are returned by the NFSv3 * Pathconf RPC, there is no point in doing it for others. * For NFSv4, the Pathconf RPC (actually a Getattr Op.) can - * be used for _PC_ACL_NFS4 and _PC_HAS_NAMEDATTR as well. + * be used for _PC_ACL_NFS4, _PC_HAS_NAMEDATTR and + * _PC_CLONE_BLKSIZE as well. */ - error = nfsrpc_pathconf(vp, &pc, &has_namedattr, td->td_ucred, - td, &nfsva, &attrflag); + error = nfsrpc_pathconf(vp, &pc, &has_namedattr, &clone_blksize, + td->td_ucred, td, &nfsva, &attrflag); if (attrflag != 0) (void) nfscl_loadattrcache(&vp, &nfsva, NULL, 0, 1); if (error != 0) @@ -4764,6 +4842,8 @@ nfs_pathconf(struct vop_pathconf_args *ap) break; case _PC_HAS_HIDDENSYSTEM: if (NFS_ISV4(vp) && NFSISSET_ATTRBIT(&np->n_vattr.na_suppattr, + NFSATTRBIT_ARCHIVE) && + NFSISSET_ATTRBIT(&np->n_vattr.na_suppattr, NFSATTRBIT_HIDDEN) && NFSISSET_ATTRBIT(&np->n_vattr.na_suppattr, NFSATTRBIT_SYSTEM)) @@ -4771,6 +4851,12 @@ nfs_pathconf(struct vop_pathconf_args *ap) else *ap->a_retval = 0; break; + case _PC_CLONE_BLKSIZE: + *ap->a_retval = clone_blksize; + break; + case _PC_CASE_INSENSITIVE: + *ap->a_retval = pc.pc_caseinsensitive; + break; default: error = vop_stdpathconf(ap); diff --git a/sys/fs/nfsclient/nfsmount.h b/sys/fs/nfsclient/nfsmount.h index 37b84a015dab..ef876dd30e59 100644 --- a/sys/fs/nfsclient/nfsmount.h +++ b/sys/fs/nfsclient/nfsmount.h @@ -87,6 +87,7 @@ struct nfsmount { /* unclipped, wraps to 0 */ struct __rpc_client *nm_aconn[NFS_MAXNCONN - 1]; /* Additional nconn */ /* Locked via nm_sockreq.nr_mtx */ + uint32_t nm_cloneblksize; /* Block cloning alignment */ u_int16_t nm_krbnamelen; /* Krb5 host principal, if any */ u_int16_t nm_dirpathlen; /* and mount dirpath, for V4 */ u_int16_t nm_srvkrbnamelen; /* and the server's target name */ |