/* $NetBSD: nfs_subs.c,v 1.67 1999/03/16 23:22:57 fvdl Exp $ */ /* * 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. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 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_subs.c 8.8 (Berkeley) 5/22/95 */ #include "fs_nfs.h" #include "opt_nfsserver.h" #include "opt_iso.h" #include "opt_uvm.h" /* * These functions support the macros and help fiddle mbuf chains for * the nfs op functions. They do things like create the rpc header and * copy data between mbuf chains and uio lists. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if defined(UVM) #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef ISO #include #endif /* * Data items converted to xdr at startup, since they are constant * This is kinda hokey, but may save a little time doing byte swaps */ u_int32_t nfs_xdrneg1; u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_auth_kerb; u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false; /* And other global data */ static u_int32_t nfs_xid = 0; nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON }; nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, NFFIFO, NFNON }; enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; int nfs_ticks; extern struct nfs_public nfs_pub; /* NFS client/server stats. */ struct nfsstats nfsstats; /* * Mapping of old NFS Version 2 RPC numbers to generic numbers. */ int nfsv3_procid[NFS_NPROCS] = { NFSPROC_NULL, NFSPROC_GETATTR, NFSPROC_SETATTR, NFSPROC_NOOP, NFSPROC_LOOKUP, NFSPROC_READLINK, NFSPROC_READ, NFSPROC_NOOP, NFSPROC_WRITE, NFSPROC_CREATE, NFSPROC_REMOVE, NFSPROC_RENAME, NFSPROC_LINK, NFSPROC_SYMLINK, NFSPROC_MKDIR, NFSPROC_RMDIR, NFSPROC_READDIR, NFSPROC_FSSTAT, NFSPROC_NOOP, NFSPROC_NOOP, NFSPROC_NOOP, NFSPROC_NOOP, NFSPROC_NOOP, NFSPROC_NOOP, NFSPROC_NOOP, NFSPROC_NOOP }; /* * and the reverse mapping from generic to Version 2 procedure numbers */ int nfsv2_procid[NFS_NPROCS] = { NFSV2PROC_NULL, NFSV2PROC_GETATTR, NFSV2PROC_SETATTR, NFSV2PROC_LOOKUP, NFSV2PROC_NOOP, NFSV2PROC_READLINK, NFSV2PROC_READ, NFSV2PROC_WRITE, NFSV2PROC_CREATE, NFSV2PROC_MKDIR, NFSV2PROC_SYMLINK, NFSV2PROC_CREATE, NFSV2PROC_REMOVE, NFSV2PROC_RMDIR, NFSV2PROC_RENAME, NFSV2PROC_LINK, NFSV2PROC_READDIR, NFSV2PROC_NOOP, NFSV2PROC_STATFS, NFSV2PROC_NOOP, NFSV2PROC_NOOP, NFSV2PROC_NOOP, NFSV2PROC_NOOP, NFSV2PROC_NOOP, NFSV2PROC_NOOP, NFSV2PROC_NOOP, }; /* * Maps errno values to nfs error numbers. * Use NFSERR_IO as the catch all for ones not specifically defined in * RFC 1094. */ static u_char nfsrv_v2errmap[ELAST] = { NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, }; /* * Maps errno values to nfs error numbers. * Although it is not obvious whether or not NFS clients really care if * a returned error value is in the specified list for the procedure, the * safest thing to do is filter them appropriately. For Version 2, the * X/Open XNFS document is the only specification that defines error values * for each RPC (The RFC simply lists all possible error values for all RPCs), * so I have decided to not do this for Version 2. * The first entry is the default error return and the rest are the valid * errors for that RPC in increasing numeric order. */ static short nfsv3err_null[] = { 0, 0, }; static short nfsv3err_getattr[] = { NFSERR_IO, NFSERR_IO, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_setattr[] = { NFSERR_IO, NFSERR_PERM, NFSERR_IO, NFSERR_ACCES, NFSERR_INVAL, NFSERR_NOSPC, NFSERR_ROFS, NFSERR_DQUOT, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOT_SYNC, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_lookup[] = { NFSERR_IO, NFSERR_NOENT, NFSERR_IO, NFSERR_ACCES, NFSERR_NOTDIR, NFSERR_NAMETOL, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_access[] = { NFSERR_IO, NFSERR_IO, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_readlink[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_INVAL, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOTSUPP, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_read[] = { NFSERR_IO, NFSERR_IO, NFSERR_NXIO, NFSERR_ACCES, NFSERR_INVAL, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, NFSERR_JUKEBOX, 0, }; static short nfsv3err_write[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_INVAL, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_ROFS, NFSERR_DQUOT, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_create[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_EXIST, NFSERR_NOTDIR, NFSERR_NOSPC, NFSERR_ROFS, NFSERR_NAMETOL, NFSERR_DQUOT, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOTSUPP, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_mkdir[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_EXIST, NFSERR_NOTDIR, NFSERR_NOSPC, NFSERR_ROFS, NFSERR_NAMETOL, NFSERR_DQUOT, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOTSUPP, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_symlink[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_EXIST, NFSERR_NOTDIR, NFSERR_NOSPC, NFSERR_ROFS, NFSERR_NAMETOL, NFSERR_DQUOT, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOTSUPP, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_mknod[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_EXIST, NFSERR_NOTDIR, NFSERR_NOSPC, NFSERR_ROFS, NFSERR_NAMETOL, NFSERR_DQUOT, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOTSUPP, NFSERR_SERVERFAULT, NFSERR_BADTYPE, 0, }; static short nfsv3err_remove[] = { NFSERR_IO, NFSERR_NOENT, NFSERR_IO, NFSERR_ACCES, NFSERR_NOTDIR, NFSERR_ROFS, NFSERR_NAMETOL, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_rmdir[] = { NFSERR_IO, NFSERR_NOENT, NFSERR_IO, NFSERR_ACCES, NFSERR_EXIST, NFSERR_NOTDIR, NFSERR_INVAL, NFSERR_ROFS, NFSERR_NAMETOL, NFSERR_NOTEMPTY, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOTSUPP, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_rename[] = { NFSERR_IO, NFSERR_NOENT, NFSERR_IO, NFSERR_ACCES, NFSERR_EXIST, NFSERR_XDEV, NFSERR_NOTDIR, NFSERR_ISDIR, NFSERR_INVAL, NFSERR_NOSPC, NFSERR_ROFS, NFSERR_MLINK, NFSERR_NAMETOL, NFSERR_NOTEMPTY, NFSERR_DQUOT, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOTSUPP, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_link[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_EXIST, NFSERR_XDEV, NFSERR_NOTDIR, NFSERR_INVAL, NFSERR_NOSPC, NFSERR_ROFS, NFSERR_MLINK, NFSERR_NAMETOL, NFSERR_DQUOT, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_NOTSUPP, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_readdir[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_NOTDIR, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_BAD_COOKIE, NFSERR_TOOSMALL, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_readdirplus[] = { NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_NOTDIR, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_BAD_COOKIE, NFSERR_NOTSUPP, NFSERR_TOOSMALL, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_fsstat[] = { NFSERR_IO, NFSERR_IO, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_fsinfo[] = { NFSERR_STALE, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_pathconf[] = { NFSERR_STALE, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short nfsv3err_commit[] = { NFSERR_IO, NFSERR_IO, NFSERR_STALE, NFSERR_BADHANDLE, NFSERR_SERVERFAULT, 0, }; static short *nfsrv_v3errmap[] = { nfsv3err_null, nfsv3err_getattr, nfsv3err_setattr, nfsv3err_lookup, nfsv3err_access, nfsv3err_readlink, nfsv3err_read, nfsv3err_write, nfsv3err_create, nfsv3err_mkdir, nfsv3err_symlink, nfsv3err_mknod, nfsv3err_remove, nfsv3err_rmdir, nfsv3err_rename, nfsv3err_link, nfsv3err_readdir, nfsv3err_readdirplus, nfsv3err_fsstat, nfsv3err_fsinfo, nfsv3err_pathconf, nfsv3err_commit, }; extern struct nfsrtt nfsrtt; extern time_t nqnfsstarttime; extern int nqsrv_clockskew; extern int nqsrv_writeslack; extern int nqsrv_maxlease; extern int nqnfs_piggy[NFS_NPROCS]; extern nfstype nfsv2_type[9]; extern nfstype nfsv3_type[9]; extern struct nfsnodehashhead *nfsnodehashtbl; extern u_long nfsnodehash; LIST_HEAD(nfsnodehashhead, nfsnode); u_long nfsdirhashmask; int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); /* * Create the header for an rpc request packet * The hsiz is the size of the rest of the nfs request header. * (just used to decide if a cluster is a good idea) */ struct mbuf * nfsm_reqh(vp, procid, hsiz, bposp) struct vnode *vp; u_long procid; int hsiz; caddr_t *bposp; { register struct mbuf *mb; register u_int32_t *tl; register caddr_t bpos; struct mbuf *mb2; struct nfsmount *nmp; int nqflag; MGET(mb, M_WAIT, MT_DATA); if (hsiz >= MINCLSIZE) MCLGET(mb, M_WAIT); mb->m_len = 0; bpos = mtod(mb, caddr_t); /* * For NQNFS, add lease request. */ if (vp) { nmp = VFSTONFS(vp->v_mount); if (nmp->nm_flag & NFSMNT_NQNFS) { nqflag = NQNFS_NEEDLEASE(vp, procid); if (nqflag) { nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED); *tl++ = txdr_unsigned(nqflag); *tl = txdr_unsigned(nmp->nm_leaseterm); } else { nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); *tl = 0; } } } /* Finally, return values */ *bposp = bpos; return (mb); } /* * Build the RPC header and fill in the authorization info. * The authorization string argument is only used when the credentials * come from outside of the kernel. * Returns the head of the mbuf list. */ struct mbuf * nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, verf_str, mrest, mrest_len, mbp, xidp) register struct ucred *cr; int nmflag; int procid; int auth_type; int auth_len; char *auth_str; int verf_len; char *verf_str; struct mbuf *mrest; int mrest_len; struct mbuf **mbp; u_int32_t *xidp; { register struct mbuf *mb; register u_int32_t *tl; register caddr_t bpos; register int i; struct mbuf *mreq, *mb2; int siz, grpsiz, authsiz; struct timeval tv; static u_int32_t base; authsiz = nfsm_rndup(auth_len); MGETHDR(mb, M_WAIT, MT_DATA); if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { MCLGET(mb, M_WAIT); } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); } else { MH_ALIGN(mb, 8 * NFSX_UNSIGNED); } mb->m_len = 0; mreq = mb; bpos = mtod(mb, caddr_t); /* * First the RPC header. */ nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); /* * derive initial xid from system time * XXX time is invalid if root not yet mounted */ if (!base && (rootvp)) { microtime(&tv); base = tv.tv_sec << 12; nfs_xid = base; } /* * Skip zero xid if it should ever happen. */ if (++nfs_xid == 0) nfs_xid++; *tl++ = *xidp = txdr_unsigned(nfs_xid); *tl++ = rpc_call; *tl++ = rpc_vers; if (nmflag & NFSMNT_NQNFS) { *tl++ = txdr_unsigned(NQNFS_PROG); *tl++ = txdr_unsigned(NQNFS_VER3); } else { *tl++ = txdr_unsigned(NFS_PROG); if (nmflag & NFSMNT_NFSV3) *tl++ = txdr_unsigned(NFS_VER3); else *tl++ = txdr_unsigned(NFS_VER2); } if (nmflag & NFSMNT_NFSV3) *tl++ = txdr_unsigned(procid); else *tl++ = txdr_unsigned(nfsv2_procid[procid]); /* * And then the authorization cred. */ *tl++ = txdr_unsigned(auth_type); *tl = txdr_unsigned(authsiz); switch (auth_type) { case RPCAUTH_UNIX: nfsm_build(tl, u_int32_t *, auth_len); *tl++ = 0; /* stamp ?? */ *tl++ = 0; /* NULL hostname */ *tl++ = txdr_unsigned(cr->cr_uid); *tl++ = txdr_unsigned(cr->cr_gid); grpsiz = (auth_len >> 2) - 5; *tl++ = txdr_unsigned(grpsiz); for (i = 0; i < grpsiz; i++) *tl++ = txdr_unsigned(cr->cr_groups[i]); break; case RPCAUTH_KERB4: siz = auth_len; while (siz > 0) { if (M_TRAILINGSPACE(mb) == 0) { MGET(mb2, M_WAIT, MT_DATA); if (siz >= MINCLSIZE) MCLGET(mb2, M_WAIT); mb->m_next = mb2; mb = mb2; mb->m_len = 0; bpos = mtod(mb, caddr_t); } i = min(siz, M_TRAILINGSPACE(mb)); memcpy(bpos, auth_str, i); mb->m_len += i; auth_str += i; bpos += i; siz -= i; } if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { for (i = 0; i < siz; i++) *bpos++ = '\0'; mb->m_len += siz; } break; }; /* * And the verifier... */ nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); if (verf_str) { *tl++ = txdr_unsigned(RPCAUTH_KERB4); *tl = txdr_unsigned(verf_len); siz = verf_len; while (siz > 0) { if (M_TRAILINGSPACE(mb) == 0) { MGET(mb2, M_WAIT, MT_DATA); if (siz >= MINCLSIZE) MCLGET(mb2, M_WAIT); mb->m_next = mb2; mb = mb2; mb->m_len = 0; bpos = mtod(mb, caddr_t); } i = min(siz, M_TRAILINGSPACE(mb)); memcpy(bpos, verf_str, i); mb->m_len += i; verf_str += i; bpos += i; siz -= i; } if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { for (i = 0; i < siz; i++) *bpos++ = '\0'; mb->m_len += siz; } } else { *tl++ = txdr_unsigned(RPCAUTH_NULL); *tl = 0; } mb->m_next = mrest; mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; mreq->m_pkthdr.rcvif = (struct ifnet *)0; *mbp = mb; return (mreq); } /* * copies mbuf chain to the uio scatter/gather list */ int nfsm_mbuftouio(mrep, uiop, siz, dpos) struct mbuf **mrep; register struct uio *uiop; int siz; caddr_t *dpos; { register char *mbufcp, *uiocp; register int xfer, left, len; register struct mbuf *mp; long uiosiz, rem; int error = 0; mp = *mrep; mbufcp = *dpos; len = mtod(mp, caddr_t)+mp->m_len-mbufcp; rem = nfsm_rndup(siz)-siz; while (siz > 0) { if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) return (EFBIG); left = uiop->uio_iov->iov_len; uiocp = uiop->uio_iov->iov_base; if (left > siz) left = siz; uiosiz = left; while (left > 0) { while (len == 0) { mp = mp->m_next; if (mp == NULL) return (EBADRPC); mbufcp = mtod(mp, caddr_t); len = mp->m_len; } xfer = (left > len) ? len : left; #ifdef notdef /* Not Yet.. */ if (uiop->uio_iov->iov_op != NULL) (*(uiop->uio_iov->iov_op)) (mbufcp, uiocp, xfer); else #endif if (uiop->uio_segflg == UIO_SYSSPACE) memcpy(uiocp, mbufcp, xfer); else copyout(mbufcp, uiocp, xfer); left -= xfer; len -= xfer; mbufcp += xfer; uiocp += xfer; uiop->uio_offset += xfer; uiop->uio_resid -= xfer; } if (uiop->uio_iov->iov_len <= siz) { uiop->uio_iovcnt--; uiop->uio_iov++; } else { (caddr_t)uiop->uio_iov->iov_base += uiosiz; uiop->uio_iov->iov_len -= uiosiz; } siz -= uiosiz; } *dpos = mbufcp; *mrep = mp; if (rem > 0) { if (len < rem) error = nfs_adv(mrep, dpos, rem, len); else *dpos += rem; } return (error); } /* * copies a uio scatter/gather list to an mbuf chain. * NOTE: can ony handle iovcnt == 1 */ int nfsm_uiotombuf(uiop, mq, siz, bpos) register struct uio *uiop; struct mbuf **mq; int siz; caddr_t *bpos; { register char *uiocp; register struct mbuf *mp, *mp2; register int xfer, left, mlen; int uiosiz, clflg, rem; char *cp; #ifdef DIAGNOSTIC if (uiop->uio_iovcnt != 1) panic("nfsm_uiotombuf: iovcnt != 1"); #endif if (siz > MLEN) /* or should it >= MCLBYTES ?? */ clflg = 1; else clflg = 0; rem = nfsm_rndup(siz)-siz; mp = mp2 = *mq; while (siz > 0) { left = uiop->uio_iov->iov_len; uiocp = uiop->uio_iov->iov_base; if (left > siz) left = siz; uiosiz = left; while (left > 0) { mlen = M_TRAILINGSPACE(mp); if (mlen == 0) { MGET(mp, M_WAIT, MT_DATA); if (clflg) MCLGET(mp, M_WAIT); mp->m_len = 0; mp2->m_next = mp; mp2 = mp; mlen = M_TRAILINGSPACE(mp); } xfer = (left > mlen) ? mlen : left; #ifdef notdef /* Not Yet.. */ if (uiop->uio_iov->iov_op != NULL) (*(uiop->uio_iov->iov_op)) (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); else #endif if (uiop->uio_segflg == UIO_SYSSPACE) memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer); else copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); mp->m_len += xfer; left -= xfer; uiocp += xfer; uiop->uio_offset += xfer; uiop->uio_resid -= xfer; } (caddr_t)uiop->uio_iov->iov_base += uiosiz; uiop->uio_iov->iov_len -= uiosiz; siz -= uiosiz; } if (rem > 0) { if (rem > M_TRAILINGSPACE(mp)) { MGET(mp, M_WAIT, MT_DATA); mp->m_len = 0; mp2->m_next = mp; } cp = mtod(mp, caddr_t)+mp->m_len; for (left = 0; left < rem; left++) *cp++ = '\0'; mp->m_len += rem; *bpos = cp; } else *bpos = mtod(mp, caddr_t)+mp->m_len; *mq = mp; return (0); } /* * Get at least "siz" bytes of correctly aligned data. * When called the mbuf pointers are not necessarily correct, * dsosp points to what ought to be in m_data and left contains * what ought to be in m_len. * This is used by the macros nfsm_dissect and nfsm_dissecton for tough * cases. (The macros use the vars. dpos and dpos2) */ int nfsm_disct(mdp, dposp, siz, left, cp2) struct mbuf **mdp; caddr_t *dposp; int siz; int left; caddr_t *cp2; { register struct mbuf *m1, *m2; struct mbuf *havebuf = NULL; caddr_t src = *dposp; caddr_t dst; int len; #ifdef DEBUG if (left < 0) panic("nfsm_disct: left < 0"); #endif m1 = *mdp; /* * Skip through the mbuf chain looking for an mbuf with * some data. If the first mbuf found has enough data * and it is correctly aligned return it. */ while (left == 0) { havebuf = m1; *mdp = m1 = m1->m_next; if (m1 == NULL) return (EBADRPC); src = mtod(m1, caddr_t); left = m1->m_len; /* * If we start a new mbuf and it is big enough * and correctly aligned just return it, don't * do any pull up. */ if (left >= siz && nfsm_aligned(src)) { *cp2 = src; *dposp = src + siz; return (0); } } if (m1->m_flags & M_EXT) { if (havebuf) { /* If the first mbuf with data has external data * and there is a previous empty mbuf use it * to move the data into. */ m2 = m1; *mdp = m1 = havebuf; if (m1->m_flags & M_EXT) { MEXTREMOVE(m1); } } else { /* * If the first mbuf has a external data * and there is no previous empty mbuf * allocate a new mbuf and move the external * data to the new mbuf. Also make the first * mbuf look empty. */ m2 = m_get(M_WAIT, MT_DATA); m2->m_ext = m1->m_ext; m2->m_data = src; m2->m_len = left; MCLADDREFERENCE(m1, m2); MEXTREMOVE(m1); m2->m_next = m1->m_next; m1->m_next = m2; } m1->m_len = 0; dst = m1->m_dat; } else { /* * If the first mbuf has no external data * move the data to the front of the mbuf. */ if ((dst = m1->m_dat) != src) memmove(dst, src, left); dst += left; m1->m_len = left; m2 = m1->m_next; } m1->m_flags &= ~M_PKTHDR; *cp2 = m1->m_data = m1->m_dat; /* data is at beginning of buffer */ *dposp = mtod(m1, caddr_t) + siz; /* * Loop through mbufs pulling data up into first mbuf until * the first mbuf is full or there is no more data to * pullup. */ while ((len = (MLEN - m1->m_len)) != 0 && m2) { if ((len = min(len, m2->m_len)) != 0) memcpy(dst, m2->m_data, len); m1->m_len += len; dst += len; m2->m_data += len; m2->m_len -= len; m2 = m2->m_next; } if (m1->m_len < siz) return (EBADRPC); return (0); } /* * Advance the position in the mbuf chain. */ int nfs_adv(mdp, dposp, offs, left) struct mbuf **mdp; caddr_t *dposp; int offs; int left; { register struct mbuf *m; register int s; m = *mdp; s = left; while (s < offs) { offs -= s; m = m->m_next; if (m == NULL) return (EBADRPC); s = m->m_len; } *mdp = m; *dposp = mtod(m, caddr_t)+offs; return (0); } /* * Copy a string into mbufs for the hard cases... */ int nfsm_strtmbuf(mb, bpos, cp, siz) struct mbuf **mb; char **bpos; const char *cp; long siz; { register struct mbuf *m1 = NULL, *m2; long left, xfer, len, tlen; u_int32_t *tl; int putsize; putsize = 1; m2 = *mb; left = M_TRAILINGSPACE(m2); if (left > 0) { tl = ((u_int32_t *)(*bpos)); *tl++ = txdr_unsigned(siz); putsize = 0; left -= NFSX_UNSIGNED; m2->m_len += NFSX_UNSIGNED; if (left > 0) { memcpy((caddr_t) tl, cp, left); siz -= left; cp += left; m2->m_len += left; left = 0; } } /* Loop around adding mbufs */ while (siz > 0) { MGET(m1, M_WAIT, MT_DATA); if (siz > MLEN) MCLGET(m1, M_WAIT); m1->m_len = NFSMSIZ(m1); m2->m_next = m1; m2 = m1; tl = mtod(m1, u_int32_t *); tlen = 0; if (putsize) { *tl++ = txdr_unsigned(siz); m1->m_len -= NFSX_UNSIGNED; tlen = NFSX_UNSIGNED; putsize = 0; } if (siz < m1->m_len) { len = nfsm_rndup(siz); xfer = siz; if (xfer < len) *(tl+(xfer>>2)) = 0; } else { xfer = len = m1->m_len; } memcpy((caddr_t) tl, cp, xfer); m1->m_len = len+tlen; siz -= xfer; cp += xfer; } *mb = m1; *bpos = mtod(m1, caddr_t)+m1->m_len; return (0); } /* * Directory caching routines. They work as follows: * - a cache is maintained per VDIR nfsnode. * - for each offset cookie that is exported to userspace, and can * thus be thrown back at us as an offset to VOP_READDIR, store * information in the cache. * - cached are: * - cookie itself * - blocknumber (essentially just a search key in the buffer cache) * - entry number in block. * - offset cookie of block in which this entry is stored * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. * - entries are looked up in a hash table * - also maintained is an LRU list of entries, used to determine * which ones to delete if the cache grows too large. * - if 32 <-> 64 translation mode is requested for a filesystem, * the cache also functions as a translation table * - in the translation case, invalidating the cache does not mean * flushing it, but just marking entries as invalid, except for * the <64bit cookie, 32bitcookie> pair which is still valid, to * still be able to use the cache as a translation table. * - 32 bit cookies are uniquely created by combining the hash table * entry value, and one generation count per hash table entry, * incremented each time an entry is appended to the chain. * - the cache is invalidated each time a direcory is modified * - sanity checks are also done; if an entry in a block turns * out not to have a matching cookie, the cache is invalidated * and a new block starting from the wanted offset is fetched from * the server. * - directory entries as read from the server are extended to contain * the 64bit and, optionally, the 32bit cookies, for sanity checking * the cache and exporting them to userspace through the cookie * argument to VOP_READDIR. */ u_long nfs_dirhash(off) off_t off; { int i; char *cp = (char *)&off; u_long sum = 0L; for (i = 0 ; i < sizeof (off); i++) sum += *cp++; return sum; } void nfs_initdircache(vp) struct vnode *vp; { struct nfsnode *np = VTONFS(vp); struct nfsmount *nmp = VFSTONFS(vp->v_mount); np->n_dircachesize = 0; np->n_dblkno = 1; np->n_dircache = hashinit(NFS_DIRHASHSIZ, M_NFSDIROFF, M_WAITOK, &nfsdirhashmask); TAILQ_INIT(&np->n_dirchain); if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { MALLOC(np->n_dirgens, unsigned *, NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, M_WAITOK); memset((caddr_t)np->n_dirgens, 0, NFS_DIRHASHSIZ * sizeof (unsigned)); } } static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0}; struct nfsdircache * nfs_searchdircache(vp, off, do32, hashent) struct vnode *vp; off_t off; int do32; int *hashent; { struct nfsdirhashhead *ndhp; struct nfsdircache *ndp = NULL; struct nfsnode *np = VTONFS(vp); unsigned ent; /* * Zero is always a valid cookie. */ if (off == 0) return &dzero; /* * We use a 32bit cookie as search key, directly reconstruct * the hashentry. Else use the hashfunction. */ if (do32) { ent = (u_int32_t)off >> 24; if (ent >= NFS_DIRHASHSIZ) return NULL; ndhp = &np->n_dircache[ent]; } else { ndhp = NFSDIRHASH(np, off); } if (hashent) *hashent = (int)(ndhp - np->n_dircache); if (do32) { for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) { if (ndp->dc_cookie32 == (u_int32_t)off) { /* * An invalidated entry will become the * start of a new block fetched from * the server. */ if (ndp->dc_blkno == -1) { ndp->dc_blkcookie = ndp->dc_cookie; ndp->dc_blkno = np->n_dblkno++; ndp->dc_entry = 0; } break; } } } else { for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) if (ndp->dc_cookie == off) break; } return ndp; } struct nfsdircache * nfs_enterdircache(vp, off, blkoff, en, blkno) struct vnode *vp; off_t off, blkoff; daddr_t blkno; int en; { struct nfsnode *np = VTONFS(vp); struct nfsdirhashhead *ndhp; struct nfsdircache *ndp = NULL, *first; struct nfsmount *nmp = VFSTONFS(vp->v_mount); int hashent, gen, overwrite; if (!np->n_dircache) /* * XXX would like to do this in nfs_nget but vtype * isn't known at that time. */ nfs_initdircache(vp); /* * XXX refuse entries for offset 0. amd(8) erroneously sets * cookie 0 for the '.' entry, making this necessary. This * isn't so bad, as 0 is a special case anyway. */ if (off == 0) return &dzero; ndp = nfs_searchdircache(vp, off, 0, &hashent); if (ndp && ndp->dc_blkno != -1) { /* * Overwriting an old entry. Check if it's the same. * If so, just return. If not, remove the old entry. */ if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) return ndp; TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); LIST_REMOVE(ndp, dc_hash); FREE(ndp, M_NFSDIROFF); ndp = 0; } ndhp = &np->n_dircache[hashent]; if (!ndp) { MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF, M_WAITOK); overwrite = 0; if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { /* * We're allocating a new entry, so bump the * generation number. */ gen = ++np->n_dirgens[hashent]; if (gen == 0) { np->n_dirgens[hashent]++; gen++; } ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); } } else overwrite = 1; /* * If the entry number is 0, we are at the start of a new block, so * allocate a new blocknumber. */ if (en == 0) ndp->dc_blkno = np->n_dblkno++; else ndp->dc_blkno = blkno; ndp->dc_cookie = off; ndp->dc_blkcookie = blkoff; ndp->dc_entry = en; if (overwrite) return ndp; /* * If the maximum directory cookie cache size has been reached * for this node, take one off the front. The idea is that * directories are typically read front-to-back once, so that * the oldest entries can be thrown away without much performance * loss. */ if (np->n_dircachesize == NFS_MAXDIRCACHE) { first = np->n_dirchain.tqh_first; TAILQ_REMOVE(&np->n_dirchain, first, dc_chain); LIST_REMOVE(first, dc_hash); FREE(first, M_NFSDIROFF); } else np->n_dircachesize++; LIST_INSERT_HEAD(ndhp, ndp, dc_hash); TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); return ndp; } void nfs_invaldircache(vp, forcefree) struct vnode *vp; int forcefree; { struct nfsnode *np = VTONFS(vp); struct nfsdircache *ndp = NULL; struct nfsmount *nmp = VFSTONFS(vp->v_mount); #ifdef DIAGNOSTIC if (vp->v_type != VDIR) panic("nfs: invaldircache: not dir"); #endif if (!np->n_dircache) return; if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { while ((ndp = np->n_dirchain.tqh_first)) { TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); LIST_REMOVE(ndp, dc_hash); FREE(ndp, M_NFSDIROFF); } np->n_dircachesize = 0; if (forcefree && np->n_dirgens) { FREE(np->n_dirgens, M_NFSDIROFF); } } else { for (ndp = np->n_dirchain.tqh_first; ndp; ndp = ndp->dc_chain.tqe_next) ndp->dc_blkno = -1; } np->n_dblkno = 1; } /* * Called once before VFS init to initialize shared and * server-specific data structures. */ void nfs_init() { #if !defined(alpha) && defined(DIAGNOSTIC) /* * Check to see if major data structures haven't bloated. */ if (sizeof (struct nfsnode) > NFS_NODEALLOC) { printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); printf("Try reducing NFS_SMALLFH\n"); } if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); printf("Try reducing NFS_UIDHASHSIZ\n"); } if (sizeof (struct nfsuid) > NFS_UIDALLOC) { printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); printf("Try unionizing the nu_nickname and nu_flag fields\n"); } #endif nfsrtt.pos = 0; rpc_vers = txdr_unsigned(RPC_VER2); rpc_call = txdr_unsigned(RPC_CALL); rpc_reply = txdr_unsigned(RPC_REPLY); rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); rpc_mismatch = txdr_unsigned(RPC_MISMATCH); rpc_autherr = txdr_unsigned(RPC_AUTHERR); rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); nfs_prog = txdr_unsigned(NFS_PROG); nqnfs_prog = txdr_unsigned(NQNFS_PROG); nfs_true = txdr_unsigned(TRUE); nfs_false = txdr_unsigned(FALSE); nfs_xdrneg1 = txdr_unsigned(-1); nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; if (nfs_ticks < 1) nfs_ticks = 1; #ifdef NFSSERVER nfsrv_init(0); /* Init server data structures */ nfsrv_initcache(); /* Init the server request cache */ #endif /* NFSSERVER */ /* * Initialize the nqnfs data structures. */ if (nqnfsstarttime == 0) { nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease + nqsrv_clockskew + nqsrv_writeslack; NQLOADNOVRAM(nqnfsstarttime); CIRCLEQ_INIT(&nqtimerhead); nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, M_WAITOK, &nqfhhash); } /* * Initialize reply list and start timer */ TAILQ_INIT(&nfs_reqq); nfs_timer(NULL); } #ifdef NFS /* * Called once at VFS init to initialize client-specific data structures. */ void nfs_vfs_init() { register int i; /* Ensure async daemons disabled */ for (i = 0; i < NFS_MAXASYNCDAEMON; i++) { nfs_iodwant[i] = (struct proc *)0; nfs_iodmount[i] = (struct nfsmount *)0; } nfs_nhinit(); /* Init the nfsnode table */ } /* * Attribute cache routines. * nfs_loadattrcache() - loads or updates the cache contents from attributes * that are on the mbuf list * nfs_getattrcache() - returns valid attributes if found in cache, returns * error otherwise */ /* * Load the attribute cache (that lives in the nfsnode entry) with * the values on the mbuf list and * Iff vap not NULL * copy the attributes to *vaper */ int nfsm_loadattrcache(vpp, mdp, dposp, vaper) struct vnode **vpp; struct mbuf **mdp; caddr_t *dposp; struct vattr *vaper; { register int32_t t1; caddr_t cp2; int error = 0; struct mbuf *md; int v3 = NFS_ISV3(*vpp); md = *mdp; t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); if (error) return (error); return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper); } int nfs_loadattrcache(vpp, fp, vaper) struct vnode **vpp; struct nfs_fattr *fp; struct vattr *vaper; { register struct vnode *vp = *vpp; register struct vattr *vap; int v3 = NFS_ISV3(vp); enum vtype vtyp; u_short vmode; struct timespec mtime; struct vnode *nvp; int32_t rdev; register struct nfsnode *np; extern int (**spec_nfsv2nodeop_p) __P((void *)); if (v3) { vtyp = nfsv3tov_type(fp->fa_type); vmode = fxdr_unsigned(u_short, fp->fa_mode); rdev = makedev(fxdr_unsigned(u_char, fp->fa3_rdev.specdata1), fxdr_unsigned(u_char, fp->fa3_rdev.specdata2)); fxdr_nfsv3time(&fp->fa3_mtime, &mtime); } else { vtyp = nfsv2tov_type(fp->fa_type); vmode = fxdr_unsigned(u_short, fp->fa_mode); if (vtyp == VNON || vtyp == VREG) vtyp = IFTOVT(vmode); rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); fxdr_nfsv2time(&fp->fa2_mtime, &mtime); /* * Really ugly NFSv2 kludge. */ if (vtyp == VCHR && rdev == 0xffffffff) vtyp = VFIFO; } /* * If v_type == VNON it is a new node, so fill in the v_type, * n_mtime fields. Check to see if it represents a special * device, and if so, check for a possible alias. Once the * correct vnode has been obtained, fill in the rest of the * information. */ np = VTONFS(vp); if (vp->v_type != vtyp) { vp->v_type = vtyp; if (vp->v_type == VFIFO) { extern int (**fifo_nfsv2nodeop_p) __P((void *)); vp->v_op = fifo_nfsv2nodeop_p; } if (vp->v_type == VCHR || vp->v_type == VBLK) { vp->v_op = spec_nfsv2nodeop_p; nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); if (nvp) { /* * Discard unneeded vnode, but save its nfsnode. * Since the nfsnode does not have a lock, its * vnode lock has to be carried over. */ nvp->v_vnlock = vp->v_vnlock; vp->v_vnlock = NULL; nvp->v_data = vp->v_data; vp->v_data = NULL; vp->v_op = spec_vnodeop_p; vrele(vp); vgone(vp); /* * Reinitialize aliased node. */ np->n_vnode = nvp; *vpp = vp = nvp; } } np->n_mtime = mtime.tv_sec; } vap = np->n_vattr; vap->va_type = vtyp; vap->va_mode = vmode & ALLPERMS; vap->va_rdev = (dev_t)rdev; vap->va_mtime = mtime; vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; if (v3) { vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); vap->va_size = fxdr_hyper(&fp->fa3_size); if (vtyp == VDIR) vap->va_blocksize = NFS_DIRFRAGSIZ; else vap->va_blocksize = NFS_FABLKSIZE; vap->va_bytes = fxdr_hyper(&fp->fa3_used); vap->va_fileid = fxdr_unsigned(int32_t, fp->fa3_fileid.nfsuquad[1]); fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); vap->va_flags = 0; vap->va_filerev = 0; } else { vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); if (vtyp == VDIR) vap->va_blocksize = NFS_DIRFRAGSIZ; else vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize); vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) * NFS_FABLKSIZE; vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); vap->va_flags = 0; vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t, fp->fa2_ctime.nfsv2_sec); vap->va_ctime.tv_nsec = 0; vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); vap->va_filerev = 0; } if (vap->va_size != np->n_size) { if (vap->va_type == VREG) { if (np->n_flag & NMODIFIED) { if (vap->va_size < np->n_size) vap->va_size = np->n_size; else np->n_size = vap->va_size; } else np->n_size = vap->va_size; #if defined(UVM) uvm_vnp_setsize(vp, np->n_size); #else vnode_pager_setsize(vp, np->n_size); #endif } else np->n_size = vap->va_size; } np->n_attrstamp = time.tv_sec; if (vaper != NULL) { memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); if (np->n_flag & NCHG) { if (np->n_flag & NACC) vaper->va_atime = np->n_atim; if (np->n_flag & NUPD) vaper->va_mtime = np->n_mtim; } } return (0); } /* * Check the time stamp * If the cache is valid, copy contents to *vap and return 0 * otherwise return an error */ int nfs_getattrcache(vp, vaper) register struct vnode *vp; struct vattr *vaper; { register struct nfsnode *np = VTONFS(vp); register struct vattr *vap; if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { nfsstats.attrcache_misses++; return (ENOENT); } nfsstats.attrcache_hits++; vap = np->n_vattr; if (vap->va_size != np->n_size) { if (vap->va_type == VREG) { if (np->n_flag & NMODIFIED) { if (vap->va_size < np->n_size) vap->va_size = np->n_size; else np->n_size = vap->va_size; } else np->n_size = vap->va_size; #if defined(UVM) uvm_vnp_setsize(vp, np->n_size); #else vnode_pager_setsize(vp, np->n_size); #endif } else np->n_size = vap->va_size; } memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); if (np->n_flag & NCHG) { if (np->n_flag & NACC) vaper->va_atime = np->n_atim; if (np->n_flag & NUPD) vaper->va_mtime = np->n_mtim; } return (0); } /* * Heuristic to see if the server XDR encodes directory cookies or not. * it is not supposed to, but a lot of servers may do this. Also, since * most/all servers will implement V2 as well, it is expected that they * may return just 32 bits worth of cookie information, so we need to * find out in which 32 bits this information is available. We do this * to avoid trouble with emulated binaries that can't handle 64 bit * directory offsets. */ void nfs_cookieheuristic(vp, flagp, p, cred) struct vnode *vp; int *flagp; struct proc *p; struct ucred *cred; { struct uio auio; struct iovec aiov; caddr_t buf, cp; struct dirent *dp; off_t *cookies = NULL, *cop; int error, eof, nc, len; MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); aiov.iov_base = buf; aiov.iov_len = NFS_DIRFRAGSIZ; auio.uio_iov = &aiov; auio.uio_iovcnt = 1; auio.uio_rw = UIO_READ; auio.uio_segflg = UIO_SYSSPACE; auio.uio_procp = p; auio.uio_resid = NFS_DIRFRAGSIZ; auio.uio_offset = 0; error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); len = NFS_DIRFRAGSIZ - auio.uio_resid; if (error || len == 0) { FREE(buf, M_TEMP); if (cookies) FREE(cookies, M_TEMP); return; } /* * Find the first valid entry and look at its offset cookie. */ cp = buf; for (cop = cookies; len > 0; len -= dp->d_reclen) { dp = (struct dirent *)cp; if (dp->d_fileno != 0 && len >= dp->d_reclen) { if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { *flagp |= NFSMNT_SWAPCOOKIE; nfs_invaldircache(vp, 0); nfs_vinvalbuf(vp, 0, cred, p, 1); } break; } cop++; cp += dp->d_reclen; } FREE(buf, M_TEMP); FREE(cookies, M_TEMP); } #endif /* NFS */ /* * Set up nameidata for a lookup() call and do it. * * If pubflag is set, this call is done for a lookup operation on the * public filehandle. In that case we allow crossing mountpoints and * absolute pathnames. However, the caller is expected to check that * the lookup result is within the public fs, and deny access if * it is not. */ int nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag) register struct nameidata *ndp; fhandle_t *fhp; int len; struct nfssvc_sock *slp; struct mbuf *nam; struct mbuf **mdp; caddr_t *dposp; struct vnode **retdirp; struct proc *p; int kerbflag, pubflag; { register int i, rem; register struct mbuf *md; register char *fromcp, *tocp, *cp; struct iovec aiov; struct uio auio; struct vnode *dp; int error, rdonly, linklen; struct componentname *cnp = &ndp->ni_cnd; *retdirp = (struct vnode *)0; MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK); /* * Copy the name from the mbuf list to ndp->ni_pnbuf * and set the various ndp fields appropriately. */ fromcp = *dposp; tocp = cnp->cn_pnbuf; md = *mdp; rem = mtod(md, caddr_t) + md->m_len - fromcp; for (i = 0; i < len; i++) { while (rem == 0) { md = md->m_next; if (md == NULL) { error = EBADRPC; goto out; } fromcp = mtod(md, caddr_t); rem = md->m_len; } if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { error = EACCES; goto out; } *tocp++ = *fromcp++; rem--; } *tocp = '\0'; *mdp = md; *dposp = fromcp; len = nfsm_rndup(len)-len; if (len > 0) { if (rem >= len) *dposp += len; else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) goto out; } /* * Extract and set starting directory. */ error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, nam, &rdonly, kerbflag, pubflag); if (error) goto out; if (dp->v_type != VDIR) { vrele(dp); error = ENOTDIR; goto out; } if (rdonly) cnp->cn_flags |= RDONLY; *retdirp = dp; if (pubflag) { /* * Oh joy. For WebNFS, handle those pesky '%' escapes, * and the 'native path' indicator. */ MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); fromcp = cnp->cn_pnbuf; tocp = cp; if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { switch ((unsigned char)*fromcp) { case WEBNFS_NATIVE_CHAR: /* * 'Native' path for us is the same * as a path according to the NFS spec, * just skip the escape char. */ fromcp++; break; /* * More may be added in the future, range 0x80-0xff */ default: error = EIO; FREE(cp, M_NAMEI); goto out; } } /* * Translate the '%' escapes, URL-style. */ while (*fromcp != '\0') { if (*fromcp == WEBNFS_ESC_CHAR) { if (fromcp[1] != '\0' && fromcp[2] != '\0') { fromcp++; *tocp++ = HEXSTRTOI(fromcp); fromcp += 2; continue; } else { error = ENOENT; FREE(cp, M_NAMEI); goto out; } } else *tocp++ = *fromcp++; } *tocp = '\0'; FREE(cnp->cn_pnbuf, M_NAMEI); cnp->cn_pnbuf = cp; } ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; ndp->ni_segflg = UIO_SYSSPACE; if (pubflag) { ndp->ni_rootdir = rootvnode; ndp->ni_loopcnt = 0; if (cnp->cn_pnbuf[0] == '/') dp = rootvnode; } else { cnp->cn_flags |= NOCROSSMOUNT; } cnp->cn_proc = p; VREF(dp); for (;;) { cnp->cn_nameptr = cnp->cn_pnbuf; ndp->ni_startdir = dp; /* * And call lookup() to do the real work */ error = lookup(ndp); if (error) break; /* * Check for encountering a symbolic link */ if ((cnp->cn_flags & ISSYMLINK) == 0) { if (cnp->cn_flags & (SAVENAME | SAVESTART)) { cnp->cn_flags |= HASBUF; return (0); } break; } else { if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1) VOP_UNLOCK(ndp->ni_dvp, 0); if (!pubflag) { vrele(ndp->ni_dvp); vput(ndp->ni_vp); ndp->ni_vp = NULL; error = EINVAL; break; } if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { error = ELOOP; break; } if (ndp->ni_pathlen > 1) MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, 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_procp = (struct proc *)0; auio.uio_resid = MAXPATHLEN; error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); if (error) { badlink: if (ndp->ni_pathlen > 1) FREE(cp, M_NAMEI); break; } linklen = MAXPATHLEN - auio.uio_resid; if (linklen == 0) { error = ENOENT; goto badlink; } if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { error = ENAMETOOLONG; goto badlink; } if (ndp->ni_pathlen > 1) { memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); FREE(cnp->cn_pnbuf, M_NAMEI); cnp->cn_pnbuf = cp; } else cnp->cn_pnbuf[linklen] = '\0'; ndp->ni_pathlen += linklen; vput(ndp->ni_vp); dp = ndp->ni_dvp; /* * Check if root directory should replace current directory. */ if (cnp->cn_pnbuf[0] == '/') { vrele(dp); dp = ndp->ni_rootdir; VREF(dp); } } } out: FREE(cnp->cn_pnbuf, M_NAMEI); return (error); } /* * A fiddled version of m_adj() that ensures null fill to a long * boundary and only trims off the back end */ void nfsm_adj(mp, len, nul) struct mbuf *mp; register int len; int nul; { register struct mbuf *m; register int count, i; register char *cp; /* * Trim from tail. Scan the mbuf chain, * calculating its length and finding the last mbuf. * If the adjustment only affects this mbuf, then just * adjust and return. Otherwise, rescan and truncate * after the remaining size. */ count = 0; m = mp; for (;;) { count += m->m_len; if (m->m_next == (struct mbuf *)0) break; m = m->m_next; } if (m->m_len > len) { m->m_len -= len; if (nul > 0) { cp = mtod(m, caddr_t)+m->m_len-nul; for (i = 0; i < nul; i++) *cp++ = '\0'; } return; } count -= len; if (count < 0) count = 0; /* * Correct length for chain is "count". * Find the mbuf with last data, adjust its length, * and toss data from remaining mbufs on chain. */ for (m = mp; m; m = m->m_next) { if (m->m_len >= count) { m->m_len = count; if (nul > 0) { cp = mtod(m, caddr_t)+m->m_len-nul; for (i = 0; i < nul; i++) *cp++ = '\0'; } break; } count -= m->m_len; } for (m = m->m_next;m;m = m->m_next) m->m_len = 0; } /* * Make these functions instead of macros, so that the kernel text size * doesn't get too big... */ void nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) struct nfsrv_descript *nfsd; int before_ret; register struct vattr *before_vap; int after_ret; struct vattr *after_vap; struct mbuf **mbp; char **bposp; { register struct mbuf *mb = *mbp, *mb2; register char *bpos = *bposp; register u_int32_t *tl; if (before_ret) { nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); *tl = nfs_false; } else { nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); *tl++ = nfs_true; txdr_hyper(before_vap->va_size, tl); tl += 2; txdr_nfsv3time(&(before_vap->va_mtime), tl); tl += 2; txdr_nfsv3time(&(before_vap->va_ctime), tl); } *bposp = bpos; *mbp = mb; nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); } void nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) struct nfsrv_descript *nfsd; int after_ret; struct vattr *after_vap; struct mbuf **mbp; char **bposp; { register struct mbuf *mb = *mbp, *mb2; register char *bpos = *bposp; register u_int32_t *tl; register struct nfs_fattr *fp; if (after_ret) { nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); *tl = nfs_false; } else { nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); *tl++ = nfs_true; fp = (struct nfs_fattr *)tl; nfsm_srvfattr(nfsd, after_vap, fp); } *mbp = mb; *bposp = bpos; } void nfsm_srvfattr(nfsd, vap, fp) register struct nfsrv_descript *nfsd; register struct vattr *vap; register struct nfs_fattr *fp; { fp->fa_nlink = txdr_unsigned(vap->va_nlink); fp->fa_uid = txdr_unsigned(vap->va_uid); fp->fa_gid = txdr_unsigned(vap->va_gid); if (nfsd->nd_flag & ND_NFSV3) { fp->fa_type = vtonfsv3_type(vap->va_type); fp->fa_mode = vtonfsv3_mode(vap->va_mode); txdr_hyper(vap->va_size, &fp->fa3_size); txdr_hyper(vap->va_bytes, &fp->fa3_used); fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); fp->fa3_fsid.nfsuquad[0] = 0; fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); fp->fa3_fileid.nfsuquad[0] = 0; fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); } else { fp->fa_type = vtonfsv2_type(vap->va_type); fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); fp->fa2_size = txdr_unsigned(vap->va_size); fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); if (vap->va_type == VFIFO) fp->fa2_rdev = 0xffffffff; else fp->fa2_rdev = txdr_unsigned(vap->va_rdev); fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); fp->fa2_fsid = txdr_unsigned(vap->va_fsid); fp->fa2_fileid = txdr_unsigned(vap->va_fileid); txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); } } /* * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) * - look up fsid in mount list (if not found ret error) * - get vp and export rights by calling VFS_FHTOVP() * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon * - if not lockflag unlock it with VOP_UNLOCK() */ int nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag) fhandle_t *fhp; int lockflag; struct vnode **vpp; struct ucred *cred; struct nfssvc_sock *slp; struct mbuf *nam; int *rdonlyp; int kerbflag; { register struct mount *mp; register int i; struct ucred *credanon; int error, exflags; struct sockaddr_in *saddr; *vpp = (struct vnode *)0; if (nfs_ispublicfh(fhp)) { if (!pubflag || !nfs_pub.np_valid) return (ESTALE); fhp = &nfs_pub.np_handle; } mp = vfs_getvfs(&fhp->fh_fsid); if (!mp) return (ESTALE); error = VFS_CHECKEXP(mp, nam, &exflags, &credanon); if (error) return (error); error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); if (error) return (error); if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { saddr = mtod(nam, struct sockaddr_in *); if (saddr->sin_family == AF_INET && ntohs(saddr->sin_port) >= IPPORT_RESERVED) { vput(*vpp); return (NFSERR_AUTHERR | AUTH_TOOWEAK); } } /* * Check/setup credentials. */ if (exflags & MNT_EXKERB) { if (!kerbflag) { vput(*vpp); return (NFSERR_AUTHERR | AUTH_TOOWEAK); } } else if (kerbflag) { vput(*vpp); return (NFSERR_AUTHERR | AUTH_TOOWEAK); } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { cred->cr_uid = credanon->cr_uid; cred->cr_gid = credanon->cr_gid; for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) cred->cr_groups[i] = credanon->cr_groups[i]; cred->cr_ngroups = i; } if (exflags & MNT_EXRDONLY) *rdonlyp = 1; else *rdonlyp = 0; if (!lockflag) VOP_UNLOCK(*vpp, 0); return (0); } /* * WebNFS: check if a filehandle is a public filehandle. For v3, this * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has * transformed this to all zeroes in both cases, so check for it. */ int nfs_ispublicfh(fhp) fhandle_t *fhp; { char *cp = (char *)fhp; int i; for (i = 0; i < NFSX_V3FH; i++) if (*cp++ != 0) return (FALSE); return (TRUE); } /* * This function compares two net addresses by family and returns TRUE * if they are the same host. * If there is any doubt, return FALSE. * The AF_INET family is handled as a special case so that address mbufs * don't need to be saved to store "struct in_addr", which is only 4 bytes. */ int netaddr_match(family, haddr, nam) int family; union nethostaddr *haddr; struct mbuf *nam; { register struct sockaddr_in *inetaddr; switch (family) { case AF_INET: inetaddr = mtod(nam, struct sockaddr_in *); if (inetaddr->sin_family == AF_INET && inetaddr->sin_addr.s_addr == haddr->had_inetaddr) return (1); break; #ifdef ISO case AF_ISO: { register struct sockaddr_iso *isoaddr1, *isoaddr2; isoaddr1 = mtod(nam, struct sockaddr_iso *); isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); if (isoaddr1->siso_family == AF_ISO && isoaddr1->siso_nlen > 0 && isoaddr1->siso_nlen == isoaddr2->siso_nlen && SAME_ISOADDR(isoaddr1, isoaddr2)) return (1); break; } #endif /* ISO */ default: break; }; return (0); } /* * The write verifier has changed (probably due to a server reboot), so all * B_NEEDCOMMIT blocks will have to be written again. Since they are on the * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT * flag. Once done the new write verifier can be set for the mount point. */ void nfs_clearcommit(mp) struct mount *mp; { register struct vnode *vp, *nvp; register struct buf *bp, *nbp; int s; s = splbio(); loop: for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { if (vp->v_mount != mp) /* Paranoia */ goto loop; nvp = vp->v_mntvnodes.le_next; for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { nbp = bp->b_vnbufs.le_next; if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) == (B_DELWRI | B_NEEDCOMMIT)) bp->b_flags &= ~B_NEEDCOMMIT; } } splx(s); } /* * Map errnos to NFS error numbers. For Version 3 also filter out error * numbers not specified for the associated procedure. */ int nfsrv_errmap(nd, err) struct nfsrv_descript *nd; register int err; { register short *defaulterrp, *errp; if (nd->nd_flag & ND_NFSV3) { if (nd->nd_procnum <= NFSPROC_COMMIT) { errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; while (*++errp) { if (*errp == err) return (err); else if (*errp > err) break; } return ((int)*defaulterrp); } else return (err & 0xffff); } if (err <= ELAST) return ((int)nfsrv_v2errmap[err - 1]); return (NFSERR_IO); } /* * Sort the group list in increasing numerical order. * (Insertion sort by Chris Torek, who was grossed out by the bubble sort * that used to be here.) */ void nfsrvw_sort(list, num) register gid_t *list; register int num; { register int i, j; gid_t v; /* Insertion sort. */ for (i = 1; i < num; i++) { v = list[i]; /* find correct slot for value v, moving others up */ for (j = i; --j >= 0 && v < list[j];) list[j + 1] = list[j]; list[j + 1] = v; } } /* * copy credentials making sure that the result can be compared with memcmp(). */ void nfsrv_setcred(incred, outcred) register struct ucred *incred, *outcred; { register int i; memset((caddr_t)outcred, 0, sizeof (struct ucred)); outcred->cr_ref = 1; outcred->cr_uid = incred->cr_uid; outcred->cr_gid = incred->cr_gid; outcred->cr_ngroups = incred->cr_ngroups; for (i = 0; i < incred->cr_ngroups; i++) outcred->cr_groups[i] = incred->cr_groups[i]; nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); }