freebsd-nq/sys/nfs/nfs_common.c

2244 lines
51 KiB
C
Raw Normal View History

1994-05-24 10:09:53 +00:00
/*
* 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
1999-08-28 01:08:13 +00:00
* $FreeBSD$
1994-05-24 10:09:53 +00:00
*/
/*
* 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 <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/bio.h>
#include <sys/buf.h>
#include <sys/proc.h>
1994-05-24 10:09:53 +00:00
#include <sys/mount.h>
#include <sys/vnode.h>
#include <sys/namei.h>
#include <sys/mbuf.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/malloc.h>
1994-09-22 22:10:49 +00:00
#include <sys/sysent.h>
#include <sys/syscall.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
#include <vm/vm_zone.h>
1994-05-24 10:09:53 +00:00
#include <nfs/rpcv2.h>
#include <nfs/nfsproto.h>
1994-05-24 10:09:53 +00:00
#include <nfs/nfs.h>
#include <nfs/nfsnode.h>
1994-05-24 10:09:53 +00:00
#include <nfs/xdr_subs.h>
#include <nfs/nfsm_subs.h>
#include <nfs/nfsmount.h>
#include <nfs/nqnfs.h>
#include <nfs/nfsrtt.h>
#include <netinet/in.h>
/*
* 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,
1994-05-24 10:09:53 +00:00
rpc_auth_kerb;
u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
1994-05-24 10:09:53 +00:00
/* And other global data */
static u_int32_t nfs_xid = 0;
1995-12-17 21:14:36 +00:00
static 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;
int nfs_pbuf_freecnt = -1; /* start out unlimited */
struct nfs_reqq nfs_reqq;
struct nfssvc_sockhead nfssvc_sockhead;
int nfssvc_sockhead_flag;
struct nfsd_head nfsd_head;
int nfsd_head_flag;
struct nfs_bufq nfs_bufq;
struct nqtimerhead nqtimerhead;
struct nqfhhashhead *nqfhhashtbl;
u_long nqfhhash;
static void (*nfs_prev_lease_updatetime) __P((int));
static int nfs_prev_nfssvc_sy_narg;
static sy_call_t *nfs_prev_nfssvc_sy_call;
#ifndef NFS_NOSERVER
static vop_t *nfs_prev_vop_lease_check;
/*
* 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
};
#endif /* NFS_NOSERVER */
/*
* 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,
};
#ifndef NFS_NOSERVER
/*
* 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, NFSERR_IO, NFSERR_IO, NFSERR_IO,
NFSERR_IO /* << Last is 86 */
};
/*
* 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,
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,
};
#endif /* NFS_NOSERVER */
1994-05-24 10:09:53 +00:00
extern struct nfsrtt nfsrtt;
extern time_t nqnfsstarttime;
extern int nqsrv_clockskew;
extern int nqsrv_writeslack;
extern int nqsrv_maxlease;
extern struct nfsstats nfsstats;
extern int nqnfs_piggy[NFS_NPROCS];
extern nfstype nfsv2_type[9];
extern nfstype nfsv3_type[9];
extern struct nfsnodehashhead *nfsnodehashtbl;
extern u_long nfsnodehash;
1994-05-24 10:09:53 +00:00
1994-09-22 22:10:49 +00:00
struct nfssvc_args;
extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
LIST_HEAD(nfsnodehashhead, nfsnode);
int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
u_quad_t
nfs_curusec()
{
struct timeval tv;
getmicrotime(&tv);
return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
}
1994-05-24 10:09:53 +00:00
/*
* 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;
1994-05-24 10:09:53 +00:00
register caddr_t bpos;
struct mbuf *mb2;
struct nfsmount *nmp;
int nqflag;
MGET(mb, M_TRYWAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
if (hsiz >= MINCLSIZE)
MCLGET(mb, M_TRYWAIT);
1994-05-24 10:09:53 +00:00
mb->m_len = 0;
bpos = mtod(mb, caddr_t);
1995-05-30 08:16:23 +00:00
1994-05-24 10:09:53 +00:00
/*
* 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);
1994-05-24 10:09:53 +00:00
*tl++ = txdr_unsigned(nqflag);
*tl = txdr_unsigned(nmp->nm_leaseterm);
} else {
nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1994-05-24 10:09:53 +00:00
*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)
1994-05-24 10:09:53 +00:00
register struct ucred *cr;
int nmflag;
1994-05-24 10:09:53 +00:00
int procid;
int auth_type;
int auth_len;
char *auth_str;
int verf_len;
char *verf_str;
1994-05-24 10:09:53 +00:00
struct mbuf *mrest;
int mrest_len;
struct mbuf **mbp;
u_int32_t *xidp;
1994-05-24 10:09:53 +00:00
{
register struct mbuf *mb;
register u_int32_t *tl;
1994-05-24 10:09:53 +00:00
register caddr_t bpos;
register int i;
struct mbuf *mreq, *mb2;
int siz, grpsiz, authsiz;
authsiz = nfsm_rndup(auth_len);
MGETHDR(mb, M_TRYWAIT, MT_DATA);
if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
MCLGET(mb, M_TRYWAIT);
} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
1994-05-24 10:09:53 +00:00
} else {
MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
1994-05-24 10:09:53 +00:00
}
mb->m_len = 0;
mreq = mb;
bpos = mtod(mb, caddr_t);
/*
* First the RPC header.
*/
nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
1998-04-06 11:41:07 +00:00
/* Get a pretty random xid to start with */
if (!nfs_xid)
nfs_xid = random();
/*
* Skip zero xid if it should ever happen.
*/
1994-05-24 10:09:53 +00:00
if (++nfs_xid == 0)
nfs_xid++;
1994-05-24 10:09:53 +00:00
*tl++ = *xidp = txdr_unsigned(nfs_xid);
*tl++ = rpc_call;
*tl++ = rpc_vers;
if (nmflag & NFSMNT_NQNFS) {
1994-05-24 10:09:53 +00:00
*tl++ = txdr_unsigned(NQNFS_PROG);
*tl++ = txdr_unsigned(NQNFS_VER3);
1994-05-24 10:09:53 +00:00
} else {
*tl++ = txdr_unsigned(NFS_PROG);
if (nmflag & NFSMNT_NFSV3)
*tl++ = txdr_unsigned(NFS_VER3);
else
*tl++ = txdr_unsigned(NFS_VER2);
1994-05-24 10:09:53 +00:00
}
if (nmflag & NFSMNT_NFSV3)
*tl++ = txdr_unsigned(procid);
else
*tl++ = txdr_unsigned(nfsv2_procid[procid]);
1994-05-24 10:09:53 +00:00
/*
* 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);
1994-05-24 10:09:53 +00:00
*tl++ = 0; /* stamp ?? */
*tl++ = 0; /* NULL hostname */
*tl++ = txdr_unsigned(cr->cr_uid);
*tl++ = txdr_unsigned(cr->cr_groups[0]);
grpsiz = (auth_len >> 2) - 5;
*tl++ = txdr_unsigned(grpsiz);
for (i = 1; i <= grpsiz; i++)
*tl++ = txdr_unsigned(cr->cr_groups[i]);
break;
case RPCAUTH_KERB4:
1994-05-24 10:09:53 +00:00
siz = auth_len;
while (siz > 0) {
if (M_TRAILINGSPACE(mb) == 0) {
MGET(mb2, M_TRYWAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
if (siz >= MINCLSIZE)
MCLGET(mb2, M_TRYWAIT);
1994-05-24 10:09:53 +00:00
mb->m_next = mb2;
mb = mb2;
mb->m_len = 0;
bpos = mtod(mb, caddr_t);
}
i = min(siz, M_TRAILINGSPACE(mb));
bcopy(auth_str, bpos, 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_TRYWAIT, MT_DATA);
if (siz >= MINCLSIZE)
MCLGET(mb2, M_TRYWAIT);
mb->m_next = mb2;
mb = mb2;
mb->m_len = 0;
bpos = mtod(mb, caddr_t);
}
i = min(siz, M_TRAILINGSPACE(mb));
bcopy(verf_str, bpos, 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;
}
1994-05-24 10:09:53 +00:00
mb->m_next = mrest;
mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
1994-05-24 10:09:53 +00:00
mreq->m_pkthdr.rcvif = (struct ifnet *)0;
*mbp = mb;
return (mreq);
}
/*
* copies mbuf chain to the uio scatter/gather list
*/
int
1994-05-24 10:09:53 +00:00
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)
bcopy(mbufcp, uiocp, 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 {
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
1994-05-24 10:09:53 +00:00
*/
int
1994-05-24 10:09:53 +00:00
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
1994-05-24 10:09:53 +00:00
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_TRYWAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
if (clflg)
MCLGET(mp, M_TRYWAIT);
1994-05-24 10:09:53 +00:00
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)
bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, 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;
}
uiop->uio_iov->iov_base += uiosiz;
uiop->uio_iov->iov_len -= uiosiz;
1994-05-24 10:09:53 +00:00
siz -= uiosiz;
}
if (rem > 0) {
if (rem > M_TRAILINGSPACE(mp)) {
MGET(mp, M_TRYWAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
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);
}
/*
* Help break down an mbuf chain by setting the first siz bytes contiguous
* pointed to by returned val.
* This is used by the macros nfsm_dissect and nfsm_dissecton for tough
* cases. (The macros use the vars. dpos and dpos2)
*/
int
1994-05-24 10:09:53 +00:00
nfsm_disct(mdp, dposp, siz, left, cp2)
struct mbuf **mdp;
caddr_t *dposp;
int siz;
int left;
caddr_t *cp2;
{
register struct mbuf *mp, *mp2;
register int siz2, xfer;
register caddr_t p;
mp = *mdp;
while (left == 0) {
*mdp = mp = mp->m_next;
if (mp == NULL)
return (EBADRPC);
left = mp->m_len;
*dposp = mtod(mp, caddr_t);
}
if (left >= siz) {
*cp2 = *dposp;
*dposp += siz;
} else if (mp->m_next == NULL) {
return (EBADRPC);
} else if (siz > MHLEN) {
panic("nfs S too big");
} else {
MGET(mp2, M_TRYWAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
mp2->m_next = mp->m_next;
mp->m_next = mp2;
mp->m_len -= left;
mp = mp2;
*cp2 = p = mtod(mp, caddr_t);
bcopy(*dposp, p, left); /* Copy what was left */
siz2 = siz-left;
p += left;
mp2 = mp->m_next;
/* Loop around copying up the siz2 bytes */
while (siz2 > 0) {
if (mp2 == NULL)
return (EBADRPC);
xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
if (xfer > 0) {
bcopy(mtod(mp2, caddr_t), p, xfer);
NFSMADV(mp2, xfer);
mp2->m_len -= xfer;
p += xfer;
siz2 -= xfer;
}
if (siz2 > 0)
mp2 = mp2->m_next;
}
mp->m_len = siz;
*mdp = mp2;
*dposp = mtod(mp2, caddr_t);
}
return (0);
}
/*
* Advance the position in the mbuf chain.
*/
int
1994-05-24 10:09:53 +00:00
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
1994-05-24 10:09:53 +00:00
nfsm_strtmbuf(mb, bpos, cp, siz)
struct mbuf **mb;
char **bpos;
1998-05-31 17:48:07 +00:00
const char *cp;
1994-05-24 10:09:53 +00:00
long siz;
{
register struct mbuf *m1 = NULL, *m2;
1994-05-24 10:09:53 +00:00
long left, xfer, len, tlen;
u_int32_t *tl;
1994-05-24 10:09:53 +00:00
int putsize;
putsize = 1;
m2 = *mb;
left = M_TRAILINGSPACE(m2);
if (left > 0) {
tl = ((u_int32_t *)(*bpos));
1994-05-24 10:09:53 +00:00
*tl++ = txdr_unsigned(siz);
putsize = 0;
left -= NFSX_UNSIGNED;
m2->m_len += NFSX_UNSIGNED;
if (left > 0) {
bcopy(cp, (caddr_t) tl, left);
siz -= left;
cp += left;
m2->m_len += left;
left = 0;
}
}
/* Loop around adding mbufs */
while (siz > 0) {
MGET(m1, M_TRYWAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
if (siz > MLEN)
MCLGET(m1, M_TRYWAIT);
1994-05-24 10:09:53 +00:00
m1->m_len = NFSMSIZ(m1);
m2->m_next = m1;
m2 = m1;
tl = mtod(m1, u_int32_t *);
1994-05-24 10:09:53 +00:00
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;
}
bcopy(cp, (caddr_t) tl, xfer);
m1->m_len = len+tlen;
siz -= xfer;
cp += xfer;
}
*mb = m1;
*bpos = mtod(m1, caddr_t)+m1->m_len;
return (0);
}
/*
* Called once to initialize data structures...
*/
int
nfs_init(vfsp)
struct vfsconf *vfsp;
1994-05-24 10:09:53 +00:00
{
register int i;
nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
nfs_mount_type = vfsp->vfc_typenum;
1994-05-24 10:09:53 +00:00
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);
1994-05-24 10:09:53 +00:00
nfs_prog = txdr_unsigned(NFS_PROG);
nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1994-05-24 10:09:53 +00:00
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;
1994-05-24 10:09:53 +00:00
/* Ensure async daemons disabled */
for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1994-05-24 10:09:53 +00:00
nfs_iodwant[i] = (struct proc *)0;
nfs_iodmount[i] = (struct nfsmount *)0;
}
1994-05-24 10:09:53 +00:00
nfs_nhinit(); /* Init the nfsnode table */
#ifndef NFS_NOSERVER
1994-05-24 10:09:53 +00:00
nfsrv_init(0); /* Init server data structures */
nfsrv_initcache(); /* Init the server request cache */
#endif
1994-05-24 10:09:53 +00:00
/*
* Initialize the nqnfs server stuff.
*/
if (nqnfsstarttime == 0) {
nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
+ nqsrv_clockskew + nqsrv_writeslack;
NQLOADNOVRAM(nqnfsstarttime);
TAILQ_INIT(&nqtimerhead);
nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash);
1994-05-24 10:09:53 +00:00
}
/*
* Initialize reply list and start timer
*/
TAILQ_INIT(&nfs_reqq);
nfs_timer(0);
1994-09-22 22:10:49 +00:00
/*
* Set up lease_check and lease_updatetime so that other parts
* of the system can call us, if we are loadable.
*/
#ifndef NFS_NOSERVER
nfs_prev_vop_lease_check = default_vnodeop_p[VOFFSET(vop_lease)];
default_vnodeop_p[VOFFSET(vop_lease)] = (vop_t *)nqnfs_vop_lease_check;
#endif
nfs_prev_lease_updatetime = lease_updatetime;
1994-09-22 22:10:49 +00:00
lease_updatetime = nfs_lease_updatetime;
nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
1994-09-22 22:10:49 +00:00
sysent[SYS_nfssvc].sy_narg = 2;
nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
sysent[SYS_nfssvc].sy_call = (sy_call_t *)nfssvc;
1994-09-22 22:10:49 +00:00
nfs_pbuf_freecnt = nswbuf / 2 + 1;
return (0);
1994-05-24 10:09:53 +00:00
}
int
nfs_uninit(vfsp)
struct vfsconf *vfsp;
{
untimeout(nfs_timer, (void *)NULL, nfs_timer_handle);
nfs_mount_type = -1;
#ifndef NFS_NOSERVER
default_vnodeop_p[VOFFSET(vop_lease)] = nfs_prev_vop_lease_check;
#endif
lease_updatetime = nfs_prev_lease_updatetime;
sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
return (0);
}
1994-05-24 10:09:53 +00:00
/*
* 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
nfs_loadattrcache(vpp, mdp, dposp, vaper, dontshrink)
1994-05-24 10:09:53 +00:00
struct vnode **vpp;
struct mbuf **mdp;
caddr_t *dposp;
struct vattr *vaper;
int dontshrink;
1994-05-24 10:09:53 +00:00
{
register struct vnode *vp = *vpp;
register struct vattr *vap;
register struct nfs_fattr *fp;
register struct nfsnode *np;
register int32_t t1;
caddr_t cp2;
int error = 0, rdev;
1994-05-24 10:09:53 +00:00
struct mbuf *md;
enum vtype vtyp;
u_short vmode;
struct timespec mtime;
int v3 = NFS_ISV3(vp);
1994-05-24 10:09:53 +00:00
md = *mdp;
t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
fp = (struct nfs_fattr *)cp2;
if (v3) {
vtyp = nfsv3tov_type(fp->fa_type);
vmode = fxdr_unsigned(u_short, fp->fa_mode);
rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
fxdr_unsigned(int, fp->fa3_rdev.specdata2));
fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1994-05-24 10:09:53 +00:00
} else {
vtyp = nfsv2tov_type(fp->fa_type);
vmode = fxdr_unsigned(u_short, fp->fa_mode);
/*
* XXX
*
* The duplicate information returned in fa_type and fa_mode
* is an ambiguity in the NFS version 2 protocol.
*
* VREG should be taken literally as a regular file. If a
* server intents to return some type information differently
* in the upper bits of the mode field (e.g. for sockets, or
* FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
* leave the examination of the mode bits even in the VREG
* case to avoid breakage for bogus servers, but we make sure
* that there are actually type bits set in the upper part of
* fa_mode (and failing that, trust the va_type field).
*
* NFSv3 cleared the issue, and requires fa_mode to not
* contain any type information (while also introduing sockets
* and FIFOs for fa_type).
*/
if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
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;
1994-05-24 10:09:53 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* If v_type == VNON it is a new node, so fill in the v_type,
1995-05-30 08:16:23 +00:00
* n_mtime fields. Check to see if it represents a special
1994-05-24 10:09:53 +00:00
* 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;
1994-05-24 10:09:53 +00:00
if (vp->v_type == VFIFO) {
vp->v_op = fifo_nfsv2nodeop_p;
}
if (vp->v_type == VCHR || vp->v_type == VBLK) {
vp->v_op = spec_nfsv2nodeop_p;
This patch corrects the first round of panics and hangs reported with the new snapshot code. Update addaliasu to correctly implement the semantics of the old checkalias function. When a device vnode first comes into existence, check to see if an anonymous vnode for the same device was created at boot time by bdevvp(). If so, adopt the bdevvp vnode rather than creating a new vnode for the device. This corrects a problem which caused the kernel to panic when taking a snapshot of the root filesystem. Change the calling convention of vn_write_suspend_wait() to be the same as vn_start_write(). Split out softdep_flushworklist() from softdep_flushfiles() so that it can be used to clear the work queue when suspending filesystem operations. Access to buffers becomes recursive so that snapshots can recursively traverse their indirect blocks using ffs_copyonwrite() when checking for the need for copy on write when flushing one of their own indirect blocks. This eliminates a deadlock between the syncer daemon and a process taking a snapshot. Ensure that softdep_process_worklist() can never block because of a snapshot being taken. This eliminates a problem with buffer starvation. Cleanup change in ffs_sync() which did not synchronously wait when MNT_WAIT was specified. The result was an unclean filesystem panic when doing forcible unmount with heavy filesystem I/O in progress. Return a zero'ed block when reading a block that was not in use at the time that a snapshot was taken. Normally, these blocks should never be read. However, the readahead code will occationally read them which can cause unexpected behavior. Clean up the debugging code that ensures that no blocks be written on a filesystem while it is suspended. Snapshots must explicitly label the blocks that they are writing during the suspension so that they do not cause a `write on suspended filesystem' panic. Reorganize ffs_copyonwrite() to eliminate a deadlock and also to prevent a race condition that would permit the same block to be copied twice. This change eliminates an unexpected soft updates inconsistency in fsck caused by the double allocation. Use bqrelse rather than brelse for buffers that will be needed soon again by the snapshot code. This improves snapshot performance.
2000-07-24 05:28:33 +00:00
vp = addaliasu(vp, rdev);
np->n_vnode = vp;
1994-05-24 10:09:53 +00:00
}
np->n_mtime = mtime.tv_sec;
1994-05-24 10:09:53 +00:00
}
vap = &np->n_vattr;
vap->va_type = vtyp;
vap->va_mode = (vmode & 07777);
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
vap->va_rdev = rdev;
1994-05-24 10:09:53 +00:00
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);
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;
1994-05-24 10:09:53 +00:00
} 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);
vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
vap->va_bytes = (u_quad_t)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);
1994-05-24 10:09:53 +00:00
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);
1994-05-24 10:09:53 +00:00
vap->va_filerev = 0;
}
np->n_attrstamp = time_second;
1994-05-24 10:09:53 +00:00
if (vap->va_size != np->n_size) {
if (vap->va_type == VREG) {
if (dontshrink && vap->va_size < np->n_size) {
/*
* We've been told not to shrink the file;
* zero np->n_attrstamp to indicate that
* the attributes are stale.
*/
vap->va_size = np->n_size;
np->n_attrstamp = 0;
} else if (np->n_flag & NMODIFIED) {
1994-05-24 10:09:53 +00:00
if (vap->va_size < np->n_size)
vap->va_size = np->n_size;
else
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
}
vnode_pager_setsize(vp, np->n_size);
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
}
1994-05-24 10:09:53 +00:00
}
if (vaper != NULL) {
bcopy((caddr_t)vap, (caddr_t)vaper, 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;
1994-05-24 10:09:53 +00:00
}
}
return (0);
}
#ifdef NFS_ACDEBUG
#include <sys/sysctl.h>
1999-02-17 13:59:29 +00:00
SYSCTL_DECL(_vfs_nfs);
static int nfs_acdebug;
SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
#endif
1994-05-24 10:09:53 +00:00
/*
* Check the time stamp
* If the cache is valid, copy contents to *vap and return 0
* otherwise return an error
*/
int
1994-05-24 10:09:53 +00:00
nfs_getattrcache(vp, vaper)
register struct vnode *vp;
struct vattr *vaper;
{
register struct nfsnode *np;
1994-05-24 10:09:53 +00:00
register struct vattr *vap;
struct nfsmount *nmp;
int timeo;
np = VTONFS(vp);
vap = &np->n_vattr;
nmp = VFSTONFS(vp->v_mount);
/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
timeo = (time_second - np->n_mtime) / 10;
#ifdef NFS_ACDEBUG
if (nfs_acdebug>1)
printf("nfs_getattrcache: initial timeo = %d\n", timeo);
#endif
if (vap->va_type == VDIR) {
if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
timeo = nmp->nm_acdirmin;
else if (timeo > nmp->nm_acdirmax)
timeo = nmp->nm_acdirmax;
} else {
if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
timeo = nmp->nm_acregmin;
else if (timeo > nmp->nm_acregmax)
timeo = nmp->nm_acregmax;
}
1994-05-24 10:09:53 +00:00
#ifdef NFS_ACDEBUG
if (nfs_acdebug > 2)
printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
nmp->nm_acregmin, nmp->nm_acregmax,
nmp->nm_acdirmin, nmp->nm_acdirmax);
if (nfs_acdebug)
printf("nfs_getattrcache: age = %d; final timeo = %d\n",
(time_second - np->n_attrstamp), timeo);
#endif
if ((time_second - np->n_attrstamp) >= timeo) {
1994-05-24 10:09:53 +00:00
nfsstats.attrcache_misses++;
return (ENOENT);
}
nfsstats.attrcache_hits++;
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;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
}
vnode_pager_setsize(vp, np->n_size);
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
}
1994-05-24 10:09:53 +00:00
}
bcopy((caddr_t)vap, (caddr_t)vaper, 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;
1994-05-24 10:09:53 +00:00
}
return (0);
}
#ifndef NFS_NOSERVER
1994-05-24 10:09:53 +00:00
/*
* 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.
*
* nfs_namei() clears out garbage fields that namei() might leave garbage.
* This is mainly ni_vp and ni_dvp when an error occurs, and ni_dvp when no
* error occurs but the parent was not requested.
*
* dirp may be set whether an error is returned or not, and must be
* released by the caller.
1994-05-24 10:09:53 +00:00
*/
int
nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
1994-05-24 10:09:53 +00:00
register struct nameidata *ndp;
fhandle_t *fhp;
int len;
struct nfssvc_sock *slp;
struct sockaddr *nam;
1994-05-24 10:09:53 +00:00
struct mbuf **mdp;
caddr_t *dposp;
struct vnode **retdirp;
1994-05-24 10:09:53 +00:00
struct proc *p;
int kerbflag, pubflag;
1994-05-24 10:09:53 +00:00
{
register int i, rem;
register struct mbuf *md;
register char *fromcp, *tocp, *cp;
struct iovec aiov;
struct uio auio;
1994-05-24 10:09:53 +00:00
struct vnode *dp;
int error, rdonly, linklen;
1994-05-24 10:09:53 +00:00
struct componentname *cnp = &ndp->ni_cnd;
*retdirp = (struct vnode *)0;
cnp->cn_pnbuf = zalloc(namei_zone);
1994-05-24 10:09:53 +00:00
/*
* 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;
1994-05-24 10:09:53 +00:00
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;
1994-05-24 10:09:53 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* Extract and set starting directory.
*/
error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
nam, &rdonly, kerbflag, pubflag);
if (error)
1994-05-24 10:09:53 +00:00
goto out;
if (dp->v_type != VDIR) {
vrele(dp);
1994-05-24 10:09:53 +00:00
error = ENOTDIR;
goto out;
}
if (rdonly)
cnp->cn_flags |= RDONLY;
/*
* Set return directory. Reference to dp is implicitly transfered
* to the returned pointer
*/
*retdirp = dp;
if (pubflag) {
/*
* Oh joy. For WebNFS, handle those pesky '%' escapes,
* and the 'native path' indicator.
*/
cp = zalloc(namei_zone);
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;
zfree(namei_zone, cp);
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;
zfree(namei_zone, cp);
goto out;
}
} else
*tocp++ = *fromcp++;
}
*tocp = '\0';
zfree(namei_zone, cnp->cn_pnbuf);
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;
}
/*
* Initialize for scan, set ni_startdir and bump ref on dp again
* becuase lookup() will dereference ni_startdir.
*/
cnp->cn_proc = p;
VREF(dp);
1994-05-24 10:09:53 +00:00
ndp->ni_startdir = dp;
for (;;) {
cnp->cn_nameptr = cnp->cn_pnbuf;
/*
* Call lookup() to do the real work. If an error occurs,
* ndp->ni_vp and ni_dvp are left uninitialized or NULL and
* we do not have to dereference anything before returning.
* In either case ni_startdir will be dereferenced and NULLed
* out.
*/
error = lookup(ndp);
if (error)
break;
/*
* Check for encountering a symbolic link. Trivial
* termination occurs if no symlink encountered.
* Note: zfree is safe because error is 0, so we will
* not zfree it again when we break.
*/
if ((cnp->cn_flags & ISSYMLINK) == 0) {
nfsrv_object_create(ndp->ni_vp);
if (cnp->cn_flags & (SAVENAME | SAVESTART))
cnp->cn_flags |= HASBUF;
else
zfree(namei_zone, cnp->cn_pnbuf);
break;
}
/*
* Validate symlink
*/
1994-05-24 10:09:53 +00:00
if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
VOP_UNLOCK(ndp->ni_dvp, 0, p);
if (!pubflag) {
error = EINVAL;
goto badlink2;
}
if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
error = ELOOP;
goto badlink2;
}
if (ndp->ni_pathlen > 1)
cp = zalloc(namei_zone);
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) {
badlink1:
if (ndp->ni_pathlen > 1)
zfree(namei_zone, cp);
badlink2:
vrele(ndp->ni_dvp);
vput(ndp->ni_vp);
break;
}
linklen = MAXPATHLEN - auio.uio_resid;
if (linklen == 0) {
error = ENOENT;
goto badlink1;
}
if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
error = ENAMETOOLONG;
goto badlink1;
}
/*
* Adjust or replace path
*/
if (ndp->ni_pathlen > 1) {
bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
zfree(namei_zone, cnp->cn_pnbuf);
cnp->cn_pnbuf = cp;
} else
cnp->cn_pnbuf[linklen] = '\0';
ndp->ni_pathlen += linklen;
/*
* Cleanup refs for next loop and check if root directory
* should replace current directory. Normally ni_dvp
* becomes the new base directory and is cleaned up when
* we loop. Explicitly null pointers after invalidation
* to clarify operation.
*/
vput(ndp->ni_vp);
ndp->ni_vp = NULL;
if (cnp->cn_pnbuf[0] == '/') {
vrele(ndp->ni_dvp);
ndp->ni_dvp = ndp->ni_rootdir;
VREF(ndp->ni_dvp);
}
ndp->ni_startdir = ndp->ni_dvp;
ndp->ni_dvp = NULL;
1994-05-24 10:09:53 +00:00
}
/*
* nfs_namei() guarentees that fields will not contain garbage
* whether an error occurs or not. This allows the caller to track
* cleanup state trivially.
*/
1994-05-24 10:09:53 +00:00
out:
if (error) {
zfree(namei_zone, cnp->cn_pnbuf);
ndp->ni_vp = NULL;
ndp->ni_dvp = NULL;
ndp->ni_startdir = NULL;
cnp->cn_flags &= ~HASBUF;
} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
ndp->ni_dvp = NULL;
}
1994-05-24 10:09:53 +00:00
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)
1994-05-24 10:09:53 +00:00
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);
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
fp->fa3_rdev.specdata1 = txdr_unsigned(umajor(vap->va_rdev));
fp->fa3_rdev.specdata2 = txdr_unsigned(uminor(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);
}
}
1994-05-24 10:09:53 +00:00
/*
* 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)
1994-05-24 10:09:53 +00:00
fhandle_t *fhp;
int lockflag;
struct vnode **vpp;
struct ucred *cred;
struct nfssvc_sock *slp;
struct sockaddr *nam;
1994-05-24 10:09:53 +00:00
int *rdonlyp;
int kerbflag;
int pubflag;
1994-05-24 10:09:53 +00:00
{
struct proc *p = curproc; /* XXX */
1994-05-24 10:09:53 +00:00
register struct mount *mp;
register int i;
struct ucred *credanon;
int error, exflags;
#ifdef MNT_EXNORESPORT /* XXX needs mountd and /etc/exports help yet */
struct sockaddr_int *saddr;
#endif
1994-05-24 10:09:53 +00:00
*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)
1994-05-24 10:09:53 +00:00
return (ESTALE);
error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
if (error)
return (error);
error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
#ifdef MNT_EXNORESPORT
if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
saddr = (struct sockaddr_in *)nam;
if (saddr->sin_family == AF_INET &&
ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
vput(*vpp);
*vpp = NULL;
return (NFSERR_AUTHERR | AUTH_TOOWEAK);
}
}
#endif
1994-05-24 10:09:53 +00:00
/*
* Check/setup credentials.
*/
if (exflags & MNT_EXKERB) {
if (!kerbflag) {
1994-05-24 10:09:53 +00:00
vput(*vpp);
*vpp = NULL;
return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1994-05-24 10:09:53 +00:00
}
} else if (kerbflag) {
vput(*vpp);
*vpp = NULL;
return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1994-05-24 10:09:53 +00:00
} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
cred->cr_uid = credanon->cr_uid;
for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
cred->cr_groups[i] = credanon->cr_groups[i];
cred->cr_ngroups = i;
1994-05-24 10:09:53 +00:00
}
if (exflags & MNT_EXRDONLY)
*rdonlyp = 1;
else
*rdonlyp = 0;
nfsrv_object_create(*vpp);
1994-05-24 10:09:53 +00:00
if (!lockflag)
VOP_UNLOCK(*vpp, 0, p);
1994-05-24 10:09:53 +00:00
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);
}
#endif /* NFS_NOSERVER */
1994-05-24 10:09:53 +00:00
/*
* 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
1994-05-24 10:09:53 +00:00
netaddr_match(family, haddr, nam)
int family;
union nethostaddr *haddr;
struct sockaddr *nam;
1994-05-24 10:09:53 +00:00
{
register struct sockaddr_in *inetaddr;
switch (family) {
case AF_INET:
inetaddr = (struct sockaddr_in *)nam;
1994-05-24 10:09:53 +00:00
if (inetaddr->sin_family == AF_INET &&
inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
return (1);
break;
default:
break;
};
return (0);
}
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
static nfsuint64 nfs_nullcookie = { { 0, 0 } };
/*
* This function finds the directory cookie that corresponds to the
* logical byte offset given.
*/
nfsuint64 *
nfs_getcookie(np, off, add)
register struct nfsnode *np;
off_t off;
int add;
{
register struct nfsdmap *dp, *dp2;
register int pos;
pos = (uoff_t)off / NFS_DIRBLKSIZ;
if (pos == 0 || off < 0) {
#ifdef DIAGNOSTIC
if (add)
panic("nfs getcookie add at <= 0");
#endif
return (&nfs_nullcookie);
}
pos--;
dp = np->n_cookies.lh_first;
if (!dp) {
if (add) {
MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
M_NFSDIROFF, M_WAITOK);
dp->ndm_eocookie = 0;
LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
} else
return ((nfsuint64 *)0);
}
while (pos >= NFSNUMCOOKIES) {
pos -= NFSNUMCOOKIES;
if (dp->ndm_list.le_next) {
if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
pos >= dp->ndm_eocookie)
return ((nfsuint64 *)0);
dp = dp->ndm_list.le_next;
} else if (add) {
MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
M_NFSDIROFF, M_WAITOK);
dp2->ndm_eocookie = 0;
LIST_INSERT_AFTER(dp, dp2, ndm_list);
dp = dp2;
} else
return ((nfsuint64 *)0);
}
if (pos >= dp->ndm_eocookie) {
if (add)
dp->ndm_eocookie = pos + 1;
else
return ((nfsuint64 *)0);
}
return (&dp->ndm_cookies[pos]);
}
/*
* Invalidate cached directory information, except for the actual directory
* blocks (which are invalidated separately).
* Done mainly to avoid the use of stale offset cookies.
*/
void
nfs_invaldir(vp)
register struct vnode *vp;
{
register struct nfsnode *np = VTONFS(vp);
#ifdef DIAGNOSTIC
if (vp->v_type != VDIR)
panic("nfs: invaldir not dir");
#endif
np->n_direofoffset = 0;
np->n_cookieverf.nfsuquad[0] = 0;
np->n_cookieverf.nfsuquad[1] = 0;
if (np->n_cookies.lh_first)
np->n_cookies.lh_first->ndm_eocookie = 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
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
* and B_CLUSTEROK flags. Once done the new write verifier can be set for the
* mount point.
*
* B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
* writes are not clusterable.
*/
void
nfs_clearcommit(mp)
struct mount *mp;
{
register struct vnode *vp, *nvp;
register struct buf *bp, *nbp;
int s;
GIANT_REQUIRED;
s = splbio();
mtx_lock(&mntvnode_mtx);
loop:
for (vp = LIST_FIRST(&mp->mnt_vnodelist); vp; vp = nvp) {
if (vp->v_mount != mp) /* Paranoia */
goto loop;
nvp = LIST_NEXT(vp, v_mntvnodes);
for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
nbp = TAILQ_NEXT(bp, b_vnbufs);
if (BUF_REFCNT(bp) == 0 &&
(bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
== (B_DELWRI | B_NEEDCOMMIT))
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
}
}
mtx_unlock(&mntvnode_mtx);
splx(s);
}
#ifndef NFS_NOSERVER
/*
* 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);
}
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
int
nfsrv_object_create(vp)
struct vnode *vp;
{
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
if (vp == NULL || vp->v_type != VREG)
return (1);
return (vfs_object_create(vp, curproc,
curproc ? curproc->p_ucred : NULL));
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
}
/*
* 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 bcmp().
*/
void
nfsrv_setcred(incred, outcred)
register struct ucred *incred, *outcred;
{
register int i;
bzero((caddr_t)outcred, sizeof (struct ucred));
outcred->cr_ref = 1;
outcred->cr_uid = incred->cr_uid;
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);
}
#endif /* NFS_NOSERVER */