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.
|
|
|
|
*
|
1997-02-10 02:22:35 +00:00
|
|
|
* @(#)nfs_vnops.c 8.16 (Berkeley) 5/27/95
|
1999-06-30 02:53:51 +00:00
|
|
|
* $Id: nfs_vnops.c,v 1.133 1999/06/28 12:34:40 peter Exp $
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1997-02-10 02:22:35 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
1995-06-27 11:07:30 +00:00
|
|
|
* vnode op calls for Sun NFS version 2 and 3
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1998-01-08 23:42:31 +00:00
|
|
|
#include "opt_inet.h"
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/kernel.h>
|
|
|
|
#include <sys/systm.h>
|
1995-06-27 11:07:30 +00:00
|
|
|
#include <sys/resourcevar.h>
|
|
|
|
#include <sys/proc.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/mount.h>
|
|
|
|
#include <sys/buf.h>
|
|
|
|
#include <sys/malloc.h>
|
|
|
|
#include <sys/mbuf.h>
|
|
|
|
#include <sys/namei.h>
|
1998-03-28 12:04:40 +00:00
|
|
|
#include <sys/socket.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/vnode.h>
|
|
|
|
#include <sys/dirent.h>
|
1995-06-27 11:07:30 +00:00
|
|
|
#include <sys/fcntl.h>
|
1994-08-08 17:31:01 +00:00
|
|
|
#include <sys/lockf.h>
|
1998-05-31 20:09:01 +00:00
|
|
|
#include <sys/stat.h>
|
1998-11-13 02:39:09 +00:00
|
|
|
#include <sys/sysctl.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
#include <vm/vm.h>
|
1995-12-07 12:48:31 +00:00
|
|
|
#include <vm/vm_extern.h>
|
1997-12-27 02:56:39 +00:00
|
|
|
#include <vm/vm_zone.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
#include <miscfs/fifofs/fifo.h>
|
1997-09-14 03:00:44 +00:00
|
|
|
#include <miscfs/specfs/specdev.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
#include <nfs/rpcv2.h>
|
1995-06-27 11:07:30 +00:00
|
|
|
#include <nfs/nfsproto.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <nfs/nfs.h>
|
|
|
|
#include <nfs/nfsnode.h>
|
|
|
|
#include <nfs/nfsmount.h>
|
|
|
|
#include <nfs/xdr_subs.h>
|
|
|
|
#include <nfs/nfsm_subs.h>
|
|
|
|
#include <nfs/nqnfs.h>
|
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
#include <net/if.h>
|
|
|
|
#include <netinet/in.h>
|
|
|
|
#include <netinet/in_var.h>
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/* Defs */
|
|
|
|
#define TRUE 1
|
|
|
|
#define FALSE 0
|
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
/*
|
|
|
|
* Ifdef for FreeBSD-current merged buffer cache. It is unfortunate that these
|
|
|
|
* calls are not in getblk() and brelse() so that they would not be necessary
|
|
|
|
* here.
|
|
|
|
*/
|
|
|
|
#ifndef B_VMIO
|
|
|
|
#define vfs_busy_pages(bp, f)
|
|
|
|
#endif
|
|
|
|
|
1995-10-29 15:33:36 +00:00
|
|
|
static int nfsspec_read __P((struct vop_read_args *));
|
|
|
|
static int nfsspec_write __P((struct vop_write_args *));
|
|
|
|
static int nfsfifo_read __P((struct vop_read_args *));
|
|
|
|
static int nfsfifo_write __P((struct vop_write_args *));
|
|
|
|
static int nfsspec_close __P((struct vop_close_args *));
|
|
|
|
static int nfsfifo_close __P((struct vop_close_args *));
|
1997-09-14 03:00:44 +00:00
|
|
|
#define nfs_poll vop_nopoll
|
1995-10-29 15:33:36 +00:00
|
|
|
static int nfs_flush __P((struct vnode *,struct ucred *,int,struct proc *,int));
|
|
|
|
static int nfs_setattrrpc __P((struct vnode *,struct vattr *,struct ucred *,struct proc *));
|
1997-10-05 12:28:59 +00:00
|
|
|
static int nfs_lookup __P((struct vop_lookup_args *));
|
1995-10-29 15:33:36 +00:00
|
|
|
static int nfs_create __P((struct vop_create_args *));
|
|
|
|
static int nfs_mknod __P((struct vop_mknod_args *));
|
|
|
|
static int nfs_open __P((struct vop_open_args *));
|
|
|
|
static int nfs_close __P((struct vop_close_args *));
|
|
|
|
static int nfs_access __P((struct vop_access_args *));
|
|
|
|
static int nfs_getattr __P((struct vop_getattr_args *));
|
|
|
|
static int nfs_setattr __P((struct vop_setattr_args *));
|
|
|
|
static int nfs_read __P((struct vop_read_args *));
|
|
|
|
static int nfs_mmap __P((struct vop_mmap_args *));
|
|
|
|
static int nfs_fsync __P((struct vop_fsync_args *));
|
|
|
|
static int nfs_remove __P((struct vop_remove_args *));
|
|
|
|
static int nfs_link __P((struct vop_link_args *));
|
|
|
|
static int nfs_rename __P((struct vop_rename_args *));
|
|
|
|
static int nfs_mkdir __P((struct vop_mkdir_args *));
|
|
|
|
static int nfs_rmdir __P((struct vop_rmdir_args *));
|
|
|
|
static int nfs_symlink __P((struct vop_symlink_args *));
|
|
|
|
static int nfs_readdir __P((struct vop_readdir_args *));
|
|
|
|
static int nfs_bmap __P((struct vop_bmap_args *));
|
|
|
|
static int nfs_strategy __P((struct vop_strategy_args *));
|
1998-05-31 17:48:07 +00:00
|
|
|
static int nfs_lookitup __P((struct vnode *, const char *, int,
|
|
|
|
struct ucred *, struct proc *, struct nfsnode **));
|
1995-10-29 15:33:36 +00:00
|
|
|
static int nfs_sillyrename __P((struct vnode *,struct vnode *,struct componentname *));
|
|
|
|
static int nfsspec_access __P((struct vop_access_args *));
|
|
|
|
static int nfs_readlink __P((struct vop_readlink_args *));
|
|
|
|
static int nfs_print __P((struct vop_print_args *));
|
|
|
|
static int nfs_advlock __P((struct vop_advlock_args *));
|
|
|
|
static int nfs_bwrite __P((struct vop_bwrite_args *));
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Global vfs data structures for nfs
|
|
|
|
*/
|
1995-11-09 08:17:23 +00:00
|
|
|
vop_t **nfsv2_vnodeop_p;
|
1995-10-29 15:33:36 +00:00
|
|
|
static struct vnodeopv_entry_desc nfsv2_vnodeop_entries[] = {
|
1997-10-26 20:55:39 +00:00
|
|
|
{ &vop_default_desc, (vop_t *) vop_defaultop },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_abortop_desc, (vop_t *) nfs_abortop },
|
|
|
|
{ &vop_access_desc, (vop_t *) nfs_access },
|
|
|
|
{ &vop_advlock_desc, (vop_t *) nfs_advlock },
|
|
|
|
{ &vop_bmap_desc, (vop_t *) nfs_bmap },
|
|
|
|
{ &vop_bwrite_desc, (vop_t *) nfs_bwrite },
|
|
|
|
{ &vop_close_desc, (vop_t *) nfs_close },
|
|
|
|
{ &vop_create_desc, (vop_t *) nfs_create },
|
|
|
|
{ &vop_fsync_desc, (vop_t *) nfs_fsync },
|
|
|
|
{ &vop_getattr_desc, (vop_t *) nfs_getattr },
|
1997-10-15 10:05:29 +00:00
|
|
|
{ &vop_getpages_desc, (vop_t *) nfs_getpages },
|
1998-03-06 09:46:52 +00:00
|
|
|
{ &vop_putpages_desc, (vop_t *) nfs_putpages },
|
1997-10-15 10:05:29 +00:00
|
|
|
{ &vop_inactive_desc, (vop_t *) nfs_inactive },
|
1997-10-26 20:55:39 +00:00
|
|
|
{ &vop_lease_desc, (vop_t *) vop_null },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_link_desc, (vop_t *) nfs_link },
|
VFS mega cleanup commit (x/N)
1. Add new file "sys/kern/vfs_default.c" where default actions for
VOPs go. Implement proper defaults for ABORTOP, BWRITE, LEASE,
POLL, REVOKE and STRATEGY. Various stuff spread over the entire
tree belongs here.
2. Change VOP_BLKATOFF to a normal function in cd9660.
3. Kill VOP_BLKATOFF, VOP_TRUNCATE, VOP_VFREE, VOP_VALLOC. These
are private interface functions between UFS and the underlying
storage manager layer (FFS/LFS/MFS/EXT2FS). The functions now
live in struct ufsmount instead.
4. Remove a kludge of VOP_ functions in all filesystems, that did
nothing but obscure the simplicity and break the expandability.
If a filesystem doesn't implement VOP_FOO, it shouldn't have an
entry for it in its vnops table. The system will try to DTRT
if it is not implemented. There are still some cruft left, but
the bulk of it is done.
5. Fix another VCALL in vfs_cache.c (thanks Bruce!)
1997-10-16 10:50:27 +00:00
|
|
|
{ &vop_lock_desc, (vop_t *) vop_sharedlock },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_lookup_desc, (vop_t *) nfs_lookup },
|
|
|
|
{ &vop_mkdir_desc, (vop_t *) nfs_mkdir },
|
|
|
|
{ &vop_mknod_desc, (vop_t *) nfs_mknod },
|
|
|
|
{ &vop_mmap_desc, (vop_t *) nfs_mmap },
|
|
|
|
{ &vop_open_desc, (vop_t *) nfs_open },
|
|
|
|
{ &vop_poll_desc, (vop_t *) nfs_poll },
|
|
|
|
{ &vop_print_desc, (vop_t *) nfs_print },
|
|
|
|
{ &vop_read_desc, (vop_t *) nfs_read },
|
|
|
|
{ &vop_readdir_desc, (vop_t *) nfs_readdir },
|
1997-10-15 10:05:29 +00:00
|
|
|
{ &vop_readlink_desc, (vop_t *) nfs_readlink },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_reclaim_desc, (vop_t *) nfs_reclaim },
|
|
|
|
{ &vop_remove_desc, (vop_t *) nfs_remove },
|
|
|
|
{ &vop_rename_desc, (vop_t *) nfs_rename },
|
|
|
|
{ &vop_rmdir_desc, (vop_t *) nfs_rmdir },
|
|
|
|
{ &vop_setattr_desc, (vop_t *) nfs_setattr },
|
1997-10-15 10:05:29 +00:00
|
|
|
{ &vop_strategy_desc, (vop_t *) nfs_strategy },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_symlink_desc, (vop_t *) nfs_symlink },
|
|
|
|
{ &vop_write_desc, (vop_t *) nfs_write },
|
1995-11-09 08:17:23 +00:00
|
|
|
{ NULL, NULL }
|
1994-05-24 10:09:53 +00:00
|
|
|
};
|
1995-10-29 15:33:36 +00:00
|
|
|
static struct vnodeopv_desc nfsv2_vnodeop_opv_desc =
|
1994-05-24 10:09:53 +00:00
|
|
|
{ &nfsv2_vnodeop_p, nfsv2_vnodeop_entries };
|
1994-09-21 03:47:43 +00:00
|
|
|
VNODEOP_SET(nfsv2_vnodeop_opv_desc);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Special device vnode ops
|
|
|
|
*/
|
1995-11-09 08:17:23 +00:00
|
|
|
vop_t **spec_nfsv2nodeop_p;
|
1997-10-16 22:01:05 +00:00
|
|
|
static struct vnodeopv_entry_desc nfsv2_specop_entries[] = {
|
1997-10-15 13:24:07 +00:00
|
|
|
{ &vop_default_desc, (vop_t *) spec_vnoperate },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_access_desc, (vop_t *) nfsspec_access },
|
|
|
|
{ &vop_close_desc, (vop_t *) nfsspec_close },
|
|
|
|
{ &vop_fsync_desc, (vop_t *) nfs_fsync },
|
|
|
|
{ &vop_getattr_desc, (vop_t *) nfs_getattr },
|
1997-10-15 10:05:29 +00:00
|
|
|
{ &vop_inactive_desc, (vop_t *) nfs_inactive },
|
VFS mega cleanup commit (x/N)
1. Add new file "sys/kern/vfs_default.c" where default actions for
VOPs go. Implement proper defaults for ABORTOP, BWRITE, LEASE,
POLL, REVOKE and STRATEGY. Various stuff spread over the entire
tree belongs here.
2. Change VOP_BLKATOFF to a normal function in cd9660.
3. Kill VOP_BLKATOFF, VOP_TRUNCATE, VOP_VFREE, VOP_VALLOC. These
are private interface functions between UFS and the underlying
storage manager layer (FFS/LFS/MFS/EXT2FS). The functions now
live in struct ufsmount instead.
4. Remove a kludge of VOP_ functions in all filesystems, that did
nothing but obscure the simplicity and break the expandability.
If a filesystem doesn't implement VOP_FOO, it shouldn't have an
entry for it in its vnops table. The system will try to DTRT
if it is not implemented. There are still some cruft left, but
the bulk of it is done.
5. Fix another VCALL in vfs_cache.c (thanks Bruce!)
1997-10-16 10:50:27 +00:00
|
|
|
{ &vop_lock_desc, (vop_t *) vop_sharedlock },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_print_desc, (vop_t *) nfs_print },
|
|
|
|
{ &vop_read_desc, (vop_t *) nfsspec_read },
|
|
|
|
{ &vop_reclaim_desc, (vop_t *) nfs_reclaim },
|
|
|
|
{ &vop_setattr_desc, (vop_t *) nfs_setattr },
|
|
|
|
{ &vop_write_desc, (vop_t *) nfsspec_write },
|
1995-11-09 08:17:23 +00:00
|
|
|
{ NULL, NULL }
|
1994-05-24 10:09:53 +00:00
|
|
|
};
|
1995-10-29 15:33:36 +00:00
|
|
|
static struct vnodeopv_desc spec_nfsv2nodeop_opv_desc =
|
1997-10-16 22:01:05 +00:00
|
|
|
{ &spec_nfsv2nodeop_p, nfsv2_specop_entries };
|
1994-09-21 03:47:43 +00:00
|
|
|
VNODEOP_SET(spec_nfsv2nodeop_opv_desc);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-11-09 08:17:23 +00:00
|
|
|
vop_t **fifo_nfsv2nodeop_p;
|
1997-10-16 22:01:05 +00:00
|
|
|
static struct vnodeopv_entry_desc nfsv2_fifoop_entries[] = {
|
1997-10-15 13:24:07 +00:00
|
|
|
{ &vop_default_desc, (vop_t *) fifo_vnoperate },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_access_desc, (vop_t *) nfsspec_access },
|
|
|
|
{ &vop_close_desc, (vop_t *) nfsfifo_close },
|
|
|
|
{ &vop_fsync_desc, (vop_t *) nfs_fsync },
|
|
|
|
{ &vop_getattr_desc, (vop_t *) nfs_getattr },
|
1997-10-15 10:05:29 +00:00
|
|
|
{ &vop_inactive_desc, (vop_t *) nfs_inactive },
|
VFS mega cleanup commit (x/N)
1. Add new file "sys/kern/vfs_default.c" where default actions for
VOPs go. Implement proper defaults for ABORTOP, BWRITE, LEASE,
POLL, REVOKE and STRATEGY. Various stuff spread over the entire
tree belongs here.
2. Change VOP_BLKATOFF to a normal function in cd9660.
3. Kill VOP_BLKATOFF, VOP_TRUNCATE, VOP_VFREE, VOP_VALLOC. These
are private interface functions between UFS and the underlying
storage manager layer (FFS/LFS/MFS/EXT2FS). The functions now
live in struct ufsmount instead.
4. Remove a kludge of VOP_ functions in all filesystems, that did
nothing but obscure the simplicity and break the expandability.
If a filesystem doesn't implement VOP_FOO, it shouldn't have an
entry for it in its vnops table. The system will try to DTRT
if it is not implemented. There are still some cruft left, but
the bulk of it is done.
5. Fix another VCALL in vfs_cache.c (thanks Bruce!)
1997-10-16 10:50:27 +00:00
|
|
|
{ &vop_lock_desc, (vop_t *) vop_sharedlock },
|
1997-10-15 09:22:02 +00:00
|
|
|
{ &vop_print_desc, (vop_t *) nfs_print },
|
|
|
|
{ &vop_read_desc, (vop_t *) nfsfifo_read },
|
|
|
|
{ &vop_reclaim_desc, (vop_t *) nfs_reclaim },
|
|
|
|
{ &vop_setattr_desc, (vop_t *) nfs_setattr },
|
|
|
|
{ &vop_write_desc, (vop_t *) nfsfifo_write },
|
1995-11-09 08:17:23 +00:00
|
|
|
{ NULL, NULL }
|
1994-05-24 10:09:53 +00:00
|
|
|
};
|
1995-10-29 15:33:36 +00:00
|
|
|
static struct vnodeopv_desc fifo_nfsv2nodeop_opv_desc =
|
1997-10-16 22:01:05 +00:00
|
|
|
{ &fifo_nfsv2nodeop_p, nfsv2_fifoop_entries };
|
1994-09-21 03:47:43 +00:00
|
|
|
VNODEOP_SET(fifo_nfsv2nodeop_opv_desc);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-11-21 15:51:39 +00:00
|
|
|
static int nfs_commit __P((struct vnode *vp, u_quad_t offset, int cnt,
|
|
|
|
struct ucred *cred, struct proc *procp));
|
|
|
|
static int nfs_mknodrpc __P((struct vnode *dvp, struct vnode **vpp,
|
|
|
|
struct componentname *cnp,
|
|
|
|
struct vattr *vap));
|
1998-05-31 17:48:07 +00:00
|
|
|
static int nfs_removerpc __P((struct vnode *dvp, const char *name,
|
|
|
|
int namelen,
|
1995-11-21 15:51:39 +00:00
|
|
|
struct ucred *cred, struct proc *proc));
|
1998-05-31 17:48:07 +00:00
|
|
|
static int nfs_renamerpc __P((struct vnode *fdvp, const char *fnameptr,
|
1995-11-21 15:51:39 +00:00
|
|
|
int fnamelen, struct vnode *tdvp,
|
1998-05-31 17:48:07 +00:00
|
|
|
const char *tnameptr, int tnamelen,
|
1995-11-21 15:51:39 +00:00
|
|
|
struct ucred *cred, struct proc *proc));
|
|
|
|
static int nfs_renameit __P((struct vnode *sdvp,
|
|
|
|
struct componentname *scnp,
|
|
|
|
struct sillyrename *sp));
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Global variables
|
|
|
|
*/
|
1998-05-31 20:09:01 +00:00
|
|
|
extern u_int32_t nfs_true, nfs_false;
|
|
|
|
extern u_int32_t nfs_xdrneg1;
|
1995-06-27 11:07:30 +00:00
|
|
|
extern struct nfsstats nfsstats;
|
|
|
|
extern nfstype nfsv3_type[9];
|
1994-05-24 10:09:53 +00:00
|
|
|
struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON];
|
1996-11-06 10:53:16 +00:00
|
|
|
struct nfsmount *nfs_iodmount[NFS_MAXASYNCDAEMON];
|
1994-05-24 10:09:53 +00:00
|
|
|
int nfs_numasync = 0;
|
|
|
|
#define DIRHDSIZ (sizeof (struct dirent) - (MAXNAMLEN + 1))
|
|
|
|
|
1999-02-16 10:49:55 +00:00
|
|
|
SYSCTL_DECL(_vfs_nfs);
|
|
|
|
|
Reimplement the NFS ACCESS RPC cache as an "accelerator" rather than a true
cache. If the cached result lets us say "yes", then go with that. If
we're not sure, or we think the answer might be "no", go to the wire to be
certain. This avoids all of the possible false veto cases, and allows us
to key the cached value with just the UID for which the cached value holds,
reducing the bloat of the nfsnode structure from 104 bytes to just 12 bytes.
Since the "yes" case is by far the most common, this should still provide
a substantial performance improvement. Also default the cache to on, with
a conservative timeout (2 seconds). This improves performance if NFS is
loaded as a KLD module, as there's not (yet) code to parse an option out
of the module arguments to set it, and sysctl doesn't work (yet) for OIDs
in modules.
The 'accelerator' mode was suggested by Bjoern Groenvall (bg@sics.se)
Feedback on this would be appreciated as testing has been necessarily
limited by Comdex, and it would be valuable to have this in 2.2.8.
1998-11-15 20:36:18 +00:00
|
|
|
static int nfsaccess_cache_timeout = 2;
|
1998-11-13 02:39:09 +00:00
|
|
|
SYSCTL_INT(_vfs_nfs, OID_AUTO, access_cache_timeout, CTLFLAG_RW,
|
|
|
|
&nfsaccess_cache_timeout, 0, "NFS ACCESS cache timeout");
|
|
|
|
|
|
|
|
static int nfsaccess_cache_hits;
|
|
|
|
SYSCTL_INT(_vfs_nfs, OID_AUTO, access_cache_hits, CTLFLAG_RD,
|
|
|
|
&nfsaccess_cache_hits, 0, "NFS ACCESS cache hit count");
|
|
|
|
|
|
|
|
static int nfsaccess_cache_fills;
|
|
|
|
SYSCTL_INT(_vfs_nfs, OID_AUTO, access_cache_fills, CTLFLAG_RD,
|
|
|
|
&nfsaccess_cache_fills, 0, "NFS ACCESS cache fill count");
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* nfs access vnode op.
|
1995-06-27 11:07:30 +00:00
|
|
|
* For nfs version 2, just return ok. File accesses may fail later.
|
|
|
|
* For nfs version 3, use the access rpc to check accessibility. If file modes
|
|
|
|
* are changed on the server, accesses might still fail later.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_access(ap)
|
|
|
|
struct vop_access_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
int a_mode;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
int error = 0, attrflag;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1998-11-13 02:39:09 +00:00
|
|
|
u_int32_t mode, rmode, wmode;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(vp);
|
1998-11-13 02:39:09 +00:00
|
|
|
struct nfsnode *np = VTONFS(vp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-10-22 09:32:48 +00:00
|
|
|
/*
|
|
|
|
* Disallow write attempts on filesystems mounted read-only;
|
|
|
|
* unless the file is a socket, fifo, or a block or character
|
|
|
|
* device resident on the filesystem.
|
|
|
|
*/
|
|
|
|
if ((ap->a_mode & VWRITE) && (vp->v_mount->mnt_flag & MNT_RDONLY)) {
|
|
|
|
switch (vp->v_type) {
|
1998-05-31 17:27:58 +00:00
|
|
|
case VREG:
|
|
|
|
case VDIR:
|
|
|
|
case VLNK:
|
1995-10-22 09:32:48 +00:00
|
|
|
return (EROFS);
|
1998-05-31 17:27:58 +00:00
|
|
|
default:
|
|
|
|
break;
|
1995-10-22 09:32:48 +00:00
|
|
|
}
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
1998-11-13 02:39:09 +00:00
|
|
|
* For nfs v3, check to see if we have done this recently, and if
|
|
|
|
* so return our cached result instead of making an ACCESS call.
|
|
|
|
* If not, do an access rpc, otherwise you are stuck emulating
|
1994-05-24 10:09:53 +00:00
|
|
|
* ufs_access() locally using the vattr. This may not be correct,
|
|
|
|
* since the server may apply other access criteria such as
|
1998-11-13 02:39:09 +00:00
|
|
|
* client uid-->server uid mapping that we do not know about.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ap->a_mode & VREAD)
|
1995-06-27 11:07:30 +00:00
|
|
|
mode = NFSV3ACCESS_READ;
|
1994-05-24 10:09:53 +00:00
|
|
|
else
|
1995-06-27 11:07:30 +00:00
|
|
|
mode = 0;
|
1998-05-31 17:27:58 +00:00
|
|
|
if (vp->v_type != VDIR) {
|
1995-06-27 11:07:30 +00:00
|
|
|
if (ap->a_mode & VWRITE)
|
1998-05-31 17:27:58 +00:00
|
|
|
mode |= (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (ap->a_mode & VEXEC)
|
1998-05-31 17:27:58 +00:00
|
|
|
mode |= NFSV3ACCESS_EXECUTE;
|
1995-06-27 11:07:30 +00:00
|
|
|
} else {
|
|
|
|
if (ap->a_mode & VWRITE)
|
1998-05-31 17:27:58 +00:00
|
|
|
mode |= (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
|
|
|
|
NFSV3ACCESS_DELETE);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (ap->a_mode & VEXEC)
|
1998-05-31 17:27:58 +00:00
|
|
|
mode |= NFSV3ACCESS_LOOKUP;
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
1998-11-13 02:39:09 +00:00
|
|
|
/* XXX safety belt, only make blanket request if caching */
|
|
|
|
if (nfsaccess_cache_timeout > 0) {
|
|
|
|
wmode = NFSV3ACCESS_READ | NFSV3ACCESS_MODIFY |
|
|
|
|
NFSV3ACCESS_EXTEND | NFSV3ACCESS_EXECUTE |
|
|
|
|
NFSV3ACCESS_DELETE | NFSV3ACCESS_LOOKUP;
|
|
|
|
} else {
|
|
|
|
wmode = mode;
|
|
|
|
}
|
|
|
|
|
Reimplement the NFS ACCESS RPC cache as an "accelerator" rather than a true
cache. If the cached result lets us say "yes", then go with that. If
we're not sure, or we think the answer might be "no", go to the wire to be
certain. This avoids all of the possible false veto cases, and allows us
to key the cached value with just the UID for which the cached value holds,
reducing the bloat of the nfsnode structure from 104 bytes to just 12 bytes.
Since the "yes" case is by far the most common, this should still provide
a substantial performance improvement. Also default the cache to on, with
a conservative timeout (2 seconds). This improves performance if NFS is
loaded as a KLD module, as there's not (yet) code to parse an option out
of the module arguments to set it, and sysctl doesn't work (yet) for OIDs
in modules.
The 'accelerator' mode was suggested by Bjoern Groenvall (bg@sics.se)
Feedback on this would be appreciated as testing has been necessarily
limited by Comdex, and it would be valuable to have this in 2.2.8.
1998-11-15 20:36:18 +00:00
|
|
|
/*
|
|
|
|
* Does our cached result allow us to give a definite yes to
|
|
|
|
* this request?
|
|
|
|
*/
|
1998-11-13 02:39:09 +00:00
|
|
|
if ((time_second < (np->n_modestamp + nfsaccess_cache_timeout)) &&
|
Reimplement the NFS ACCESS RPC cache as an "accelerator" rather than a true
cache. If the cached result lets us say "yes", then go with that. If
we're not sure, or we think the answer might be "no", go to the wire to be
certain. This avoids all of the possible false veto cases, and allows us
to key the cached value with just the UID for which the cached value holds,
reducing the bloat of the nfsnode structure from 104 bytes to just 12 bytes.
Since the "yes" case is by far the most common, this should still provide
a substantial performance improvement. Also default the cache to on, with
a conservative timeout (2 seconds). This improves performance if NFS is
loaded as a KLD module, as there's not (yet) code to parse an option out
of the module arguments to set it, and sysctl doesn't work (yet) for OIDs
in modules.
The 'accelerator' mode was suggested by Bjoern Groenvall (bg@sics.se)
Feedback on this would be appreciated as testing has been necessarily
limited by Comdex, and it would be valuable to have this in 2.2.8.
1998-11-15 20:36:18 +00:00
|
|
|
(ap->a_cred->cr_uid == np->n_modeuid) &&
|
|
|
|
((np->n_mode & mode) == mode)) {
|
1998-11-13 02:39:09 +00:00
|
|
|
nfsaccess_cache_hits++;
|
|
|
|
} else {
|
Reimplement the NFS ACCESS RPC cache as an "accelerator" rather than a true
cache. If the cached result lets us say "yes", then go with that. If
we're not sure, or we think the answer might be "no", go to the wire to be
certain. This avoids all of the possible false veto cases, and allows us
to key the cached value with just the UID for which the cached value holds,
reducing the bloat of the nfsnode structure from 104 bytes to just 12 bytes.
Since the "yes" case is by far the most common, this should still provide
a substantial performance improvement. Also default the cache to on, with
a conservative timeout (2 seconds). This improves performance if NFS is
loaded as a KLD module, as there's not (yet) code to parse an option out
of the module arguments to set it, and sysctl doesn't work (yet) for OIDs
in modules.
The 'accelerator' mode was suggested by Bjoern Groenvall (bg@sics.se)
Feedback on this would be appreciated as testing has been necessarily
limited by Comdex, and it would be valuable to have this in 2.2.8.
1998-11-15 20:36:18 +00:00
|
|
|
/*
|
|
|
|
* Either a no, or a don't know. Go to the wire.
|
|
|
|
*/
|
1998-11-13 02:39:09 +00:00
|
|
|
nfsstats.rpccnt[NFSPROC_ACCESS]++;
|
|
|
|
nfsm_reqhead(vp, NFSPROC_ACCESS, NFSX_FH(v3) + NFSX_UNSIGNED);
|
|
|
|
nfsm_fhtom(vp, v3);
|
|
|
|
nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
|
|
|
|
*tl = txdr_unsigned(wmode);
|
|
|
|
nfsm_request(vp, NFSPROC_ACCESS, ap->a_p, ap->a_cred);
|
|
|
|
nfsm_postop_attr(vp, attrflag);
|
|
|
|
if (!error) {
|
|
|
|
nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
|
|
|
|
rmode = fxdr_unsigned(u_int32_t, *tl);
|
|
|
|
/*
|
|
|
|
* The NFS V3 spec does not clarify whether or not
|
|
|
|
* the returned access bits can be a superset of
|
|
|
|
* the ones requested, so...
|
|
|
|
*/
|
|
|
|
if ((rmode & mode) != mode) {
|
|
|
|
error = EACCES;
|
|
|
|
} else if (nfsaccess_cache_timeout > 0) {
|
|
|
|
/* cache the result */
|
|
|
|
nfsaccess_cache_fills++;
|
|
|
|
np->n_mode = rmode;
|
Reimplement the NFS ACCESS RPC cache as an "accelerator" rather than a true
cache. If the cached result lets us say "yes", then go with that. If
we're not sure, or we think the answer might be "no", go to the wire to be
certain. This avoids all of the possible false veto cases, and allows us
to key the cached value with just the UID for which the cached value holds,
reducing the bloat of the nfsnode structure from 104 bytes to just 12 bytes.
Since the "yes" case is by far the most common, this should still provide
a substantial performance improvement. Also default the cache to on, with
a conservative timeout (2 seconds). This improves performance if NFS is
loaded as a KLD module, as there's not (yet) code to parse an option out
of the module arguments to set it, and sysctl doesn't work (yet) for OIDs
in modules.
The 'accelerator' mode was suggested by Bjoern Groenvall (bg@sics.se)
Feedback on this would be appreciated as testing has been necessarily
limited by Comdex, and it would be valuable to have this in 2.2.8.
1998-11-15 20:36:18 +00:00
|
|
|
np->n_modeuid = ap->a_cred->cr_uid;
|
1998-11-13 02:39:09 +00:00
|
|
|
np->n_modestamp = time_second;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
nfsm_reqdone;
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
1997-05-09 13:18:42 +00:00
|
|
|
} else {
|
1999-01-27 22:45:13 +00:00
|
|
|
if ((error = nfsspec_access(ap)) != 0)
|
1997-05-09 13:18:42 +00:00
|
|
|
return (error);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Attempt to prevent a mapped root from accessing a file
|
|
|
|
* which it shouldn't. We try to read a byte from the file
|
|
|
|
* if the user is root and the file is not zero length.
|
|
|
|
* After calling nfsspec_access, we should have the correct
|
|
|
|
* file size cached.
|
|
|
|
*/
|
|
|
|
if (ap->a_cred->cr_uid == 0 && (ap->a_mode & VREAD)
|
|
|
|
&& VTONFS(vp)->n_size > 0) {
|
|
|
|
struct iovec aiov;
|
|
|
|
struct uio auio;
|
|
|
|
char buf[1];
|
|
|
|
|
|
|
|
aiov.iov_base = buf;
|
|
|
|
aiov.iov_len = 1;
|
|
|
|
auio.uio_iov = &aiov;
|
|
|
|
auio.uio_iovcnt = 1;
|
|
|
|
auio.uio_offset = 0;
|
|
|
|
auio.uio_resid = 1;
|
|
|
|
auio.uio_segflg = UIO_SYSSPACE;
|
|
|
|
auio.uio_rw = UIO_READ;
|
|
|
|
auio.uio_procp = ap->a_p;
|
|
|
|
|
|
|
|
if (vp->v_type == VREG)
|
|
|
|
error = nfs_readrpc(vp, &auio, ap->a_cred);
|
|
|
|
else if (vp->v_type == VDIR) {
|
1997-11-07 09:21:01 +00:00
|
|
|
char* bp;
|
1999-05-03 20:59:14 +00:00
|
|
|
bp = malloc(NFS_DIRBLKSIZ, M_TEMP, M_WAITOK);
|
1997-11-07 09:21:01 +00:00
|
|
|
aiov.iov_base = bp;
|
1999-05-03 20:59:14 +00:00
|
|
|
aiov.iov_len = auio.uio_resid = NFS_DIRBLKSIZ;
|
1997-05-09 13:18:42 +00:00
|
|
|
error = nfs_readdirrpc(vp, &auio, ap->a_cred);
|
1997-11-07 09:21:01 +00:00
|
|
|
free(bp, M_TEMP);
|
1999-01-27 22:45:49 +00:00
|
|
|
} else if (vp->v_type == VLNK)
|
1997-05-09 13:18:42 +00:00
|
|
|
error = nfs_readlinkrpc(vp, &auio, ap->a_cred);
|
|
|
|
else
|
|
|
|
error = EACCES;
|
|
|
|
}
|
|
|
|
return (error);
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs open vnode op
|
|
|
|
* Check to see if the type is ok
|
|
|
|
* and that deletion is not in progress.
|
|
|
|
* For paged in text files, you will need to flush the page cache
|
|
|
|
* if consistency is lost.
|
|
|
|
*/
|
|
|
|
/* ARGSUSED */
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_open(ap)
|
|
|
|
struct vop_open_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
int a_mode;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
struct nfsnode *np = VTONFS(vp);
|
|
|
|
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
|
|
|
|
struct vattr vattr;
|
|
|
|
int error;
|
|
|
|
|
1998-05-31 17:27:58 +00:00
|
|
|
if (vp->v_type != VREG && vp->v_type != VDIR && vp->v_type != VLNK) {
|
1998-05-31 18:23:24 +00:00
|
|
|
#ifdef DIAGNOSTIC
|
1998-05-31 17:27:58 +00:00
|
|
|
printf("open eacces vtyp=%d\n",vp->v_type);
|
1998-05-31 18:23:24 +00:00
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EACCES);
|
1998-05-31 17:27:58 +00:00
|
|
|
}
|
1994-08-29 06:09:15 +00:00
|
|
|
/*
|
|
|
|
* Get a valid lease. If cached data is stale, flush it.
|
|
|
|
*/
|
|
|
|
if (nmp->nm_flag & NFSMNT_NQNFS) {
|
1995-06-27 11:07:30 +00:00
|
|
|
if (NQNFS_CKINVALID(vp, np, ND_READ)) {
|
1994-05-24 10:09:53 +00:00
|
|
|
do {
|
1995-06-27 11:07:30 +00:00
|
|
|
error = nqnfs_getlease(vp, ND_READ, ap->a_cred,
|
|
|
|
ap->a_p);
|
1994-05-24 10:09:53 +00:00
|
|
|
} while (error == NQNFS_EXPIRED);
|
1998-05-16 15:21:29 +00:00
|
|
|
if (error)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
if (np->n_lrev != np->n_brev ||
|
|
|
|
(np->n_flag & NQNFSNONCACHE)) {
|
|
|
|
if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred,
|
1998-05-16 15:21:29 +00:00
|
|
|
ap->a_p, 1)) == EINTR)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
np->n_brev = np->n_lrev;
|
|
|
|
}
|
|
|
|
}
|
1994-08-29 06:09:15 +00:00
|
|
|
} else {
|
1994-05-24 10:09:53 +00:00
|
|
|
if (np->n_flag & NMODIFIED) {
|
|
|
|
if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred,
|
1998-05-16 15:21:29 +00:00
|
|
|
ap->a_p, 1)) == EINTR)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
np->n_attrstamp = 0;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (vp->v_type == VDIR)
|
|
|
|
np->n_direofoffset = 0;
|
1994-10-02 17:27:07 +00:00
|
|
|
error = VOP_GETATTR(vp, &vattr, ap->a_cred, ap->a_p);
|
1998-05-16 15:21:29 +00:00
|
|
|
if (error)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
1996-09-19 18:21:32 +00:00
|
|
|
np->n_mtime = vattr.va_mtime.tv_sec;
|
1994-05-24 10:09:53 +00:00
|
|
|
} else {
|
1994-10-02 17:27:07 +00:00
|
|
|
error = VOP_GETATTR(vp, &vattr, ap->a_cred, ap->a_p);
|
1998-05-16 15:21:29 +00:00
|
|
|
if (error)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
1996-09-19 18:21:32 +00:00
|
|
|
if (np->n_mtime != vattr.va_mtime.tv_sec) {
|
1995-06-27 11:07:30 +00:00
|
|
|
if (vp->v_type == VDIR)
|
|
|
|
np->n_direofoffset = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
if ((error = nfs_vinvalbuf(vp, V_SAVE,
|
1998-05-16 15:21:29 +00:00
|
|
|
ap->a_cred, ap->a_p, 1)) == EINTR)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
1996-09-19 18:21:32 +00:00
|
|
|
np->n_mtime = vattr.va_mtime.tv_sec;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
}
|
1994-08-29 06:09:15 +00:00
|
|
|
}
|
|
|
|
if ((nmp->nm_flag & NFSMNT_NQNFS) == 0)
|
1994-05-24 10:09:53 +00:00
|
|
|
np->n_attrstamp = 0; /* For Open/Close consistency */
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs close vnode op
|
1995-06-27 11:07:30 +00:00
|
|
|
* What an NFS client should do upon close after writing is a debatable issue.
|
|
|
|
* Most NFS clients push delayed writes to the server upon close, basically for
|
|
|
|
* two reasons:
|
|
|
|
* 1 - So that any write errors may be reported back to the client process
|
|
|
|
* doing the close system call. By far the two most likely errors are
|
|
|
|
* NFSERR_NOSPC and NFSERR_DQUOT to indicate space allocation failure.
|
|
|
|
* 2 - To put a worst case upper bound on cache inconsistency between
|
|
|
|
* multiple clients for the file.
|
|
|
|
* There is also a consistency problem for Version 2 of the protocol w.r.t.
|
|
|
|
* not being able to tell if other clients are writing a file concurrently,
|
|
|
|
* since there is no way of knowing if the changed modify time in the reply
|
|
|
|
* is only due to the write for this client.
|
|
|
|
* (NFS Version 3 provides weak cache consistency data in the reply that
|
|
|
|
* should be sufficient to detect and handle this case.)
|
|
|
|
*
|
|
|
|
* The current code does the following:
|
|
|
|
* for NFS Version 2 - play it safe and flush/invalidate all dirty buffers
|
|
|
|
* for NFS Version 3 - flush dirty buffers to the server but don't invalidate
|
|
|
|
* or commit them (this satisfies 1 and 2 except for the
|
|
|
|
* case where the server crashes after this close but
|
|
|
|
* before the commit RPC, which is felt to be "good
|
|
|
|
* enough". Changing the last argument to nfs_flush() to
|
|
|
|
* a 1 would force a commit operation, if it is felt a
|
|
|
|
* commit is necessary now.
|
|
|
|
* for NQNFS - do nothing now, since 2 is dealt with via leases and
|
|
|
|
* 1 should be dealt with via an fsync() system call for
|
|
|
|
* cases where write errors are important.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
/* ARGSUSED */
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_close(ap)
|
|
|
|
struct vop_close_args /* {
|
|
|
|
struct vnodeop_desc *a_desc;
|
|
|
|
struct vnode *a_vp;
|
|
|
|
int a_fflag;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
if (vp->v_type == VREG) {
|
|
|
|
if ((VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) == 0 &&
|
|
|
|
(np->n_flag & NMODIFIED)) {
|
1995-07-13 17:55:12 +00:00
|
|
|
if (NFS_ISV3(vp)) {
|
1995-06-27 11:07:30 +00:00
|
|
|
error = nfs_flush(vp, ap->a_cred, MNT_WAIT, ap->a_p, 0);
|
1995-07-13 17:55:12 +00:00
|
|
|
np->n_flag &= ~NMODIFIED;
|
|
|
|
} else
|
1995-06-27 11:07:30 +00:00
|
|
|
error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1);
|
1994-05-24 10:09:53 +00:00
|
|
|
np->n_attrstamp = 0;
|
|
|
|
}
|
|
|
|
if (np->n_flag & NWRITEERR) {
|
|
|
|
np->n_flag &= ~NWRITEERR;
|
|
|
|
error = np->n_error;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs getattr call from vfs.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_getattr(ap)
|
|
|
|
struct vop_getattr_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct vattr *a_vap;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
|
|
|
register int32_t t1, t2;
|
1994-05-24 10:09:53 +00:00
|
|
|
caddr_t bpos, dpos;
|
|
|
|
int error = 0;
|
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(vp);
|
1997-02-10 02:22:35 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Update local times for special files.
|
|
|
|
*/
|
|
|
|
if (np->n_flag & (NACC | NUPD))
|
|
|
|
np->n_flag |= NCHG;
|
|
|
|
/*
|
|
|
|
* First look in the cache.
|
|
|
|
*/
|
|
|
|
if (nfs_getattrcache(vp, ap->a_vap) == 0)
|
|
|
|
return (0);
|
|
|
|
nfsstats.rpccnt[NFSPROC_GETATTR]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(vp, NFSPROC_GETATTR, NFSX_FH(v3));
|
|
|
|
nfsm_fhtom(vp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_request(vp, NFSPROC_GETATTR, ap->a_p, ap->a_cred);
|
1998-05-31 17:27:58 +00:00
|
|
|
if (!error) {
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_loadattr(vp, ap->a_vap);
|
1998-05-31 17:27:58 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs setattr call.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_setattr(ap)
|
|
|
|
struct vop_setattr_args /* {
|
|
|
|
struct vnodeop_desc *a_desc;
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct vattr *a_vap;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
register struct vattr *vap = ap->a_vap;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error = 0;
|
|
|
|
u_quad_t tsize;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
#ifndef nolint
|
|
|
|
tsize = (u_quad_t)0;
|
|
|
|
#endif
|
1998-05-31 19:24:19 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Setting of flags is not supported.
|
|
|
|
*/
|
|
|
|
if (vap->va_flags != VNOVAL)
|
|
|
|
return (EOPNOTSUPP);
|
|
|
|
|
1995-10-22 09:32:48 +00:00
|
|
|
/*
|
|
|
|
* Disallow write attempts if the filesystem is mounted read-only.
|
|
|
|
*/
|
1997-02-10 02:22:35 +00:00
|
|
|
if ((vap->va_flags != VNOVAL || vap->va_uid != (uid_t)VNOVAL ||
|
1996-09-19 18:21:32 +00:00
|
|
|
vap->va_gid != (gid_t)VNOVAL || vap->va_atime.tv_sec != VNOVAL ||
|
|
|
|
vap->va_mtime.tv_sec != VNOVAL || vap->va_mode != (mode_t)VNOVAL) &&
|
1995-10-22 09:32:48 +00:00
|
|
|
(vp->v_mount->mnt_flag & MNT_RDONLY))
|
|
|
|
return (EROFS);
|
1994-10-17 17:47:45 +00:00
|
|
|
if (vap->va_size != VNOVAL) {
|
1995-06-27 11:07:30 +00:00
|
|
|
switch (vp->v_type) {
|
|
|
|
case VDIR:
|
|
|
|
return (EISDIR);
|
|
|
|
case VCHR:
|
|
|
|
case VBLK:
|
1997-02-10 02:22:35 +00:00
|
|
|
case VSOCK:
|
|
|
|
case VFIFO:
|
1996-09-19 18:21:32 +00:00
|
|
|
if (vap->va_mtime.tv_sec == VNOVAL &&
|
|
|
|
vap->va_atime.tv_sec == VNOVAL &&
|
1998-05-31 20:09:01 +00:00
|
|
|
vap->va_mode == (mode_t)VNOVAL &&
|
1995-06-27 11:07:30 +00:00
|
|
|
vap->va_uid == (uid_t)VNOVAL &&
|
|
|
|
vap->va_gid == (gid_t)VNOVAL)
|
1994-10-17 17:47:45 +00:00
|
|
|
return (0);
|
1995-06-27 11:07:30 +00:00
|
|
|
vap->va_size = VNOVAL;
|
|
|
|
break;
|
|
|
|
default:
|
1995-10-22 09:32:48 +00:00
|
|
|
/*
|
|
|
|
* Disallow write attempts if the filesystem is
|
|
|
|
* mounted read-only.
|
|
|
|
*/
|
|
|
|
if (vp->v_mount->mnt_flag & MNT_RDONLY)
|
|
|
|
return (EROFS);
|
1998-11-09 07:00:14 +00:00
|
|
|
vnode_pager_setsize(vp, vap->va_size);
|
1997-02-10 02:22:35 +00:00
|
|
|
if (np->n_flag & NMODIFIED) {
|
|
|
|
if (vap->va_size == 0)
|
1995-06-27 11:07:30 +00:00
|
|
|
error = nfs_vinvalbuf(vp, 0,
|
|
|
|
ap->a_cred, ap->a_p, 1);
|
1997-02-10 02:22:35 +00:00
|
|
|
else
|
1995-06-27 11:07:30 +00:00
|
|
|
error = nfs_vinvalbuf(vp, V_SAVE,
|
|
|
|
ap->a_cred, ap->a_p, 1);
|
1998-09-29 23:28:32 +00:00
|
|
|
if (error) {
|
1998-11-09 07:00:14 +00:00
|
|
|
vnode_pager_setsize(vp, np->n_size);
|
1995-06-27 11:07:30 +00:00
|
|
|
return (error);
|
1998-09-29 23:28:32 +00:00
|
|
|
}
|
1997-02-10 02:22:35 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
tsize = np->n_size;
|
|
|
|
np->n_size = np->n_vattr.va_size = vap->va_size;
|
|
|
|
};
|
1996-09-19 18:21:32 +00:00
|
|
|
} else if ((vap->va_mtime.tv_sec != VNOVAL ||
|
1997-02-10 02:22:35 +00:00
|
|
|
vap->va_atime.tv_sec != VNOVAL) && (np->n_flag & NMODIFIED) &&
|
1995-06-27 11:07:30 +00:00
|
|
|
vp->v_type == VREG &&
|
|
|
|
(error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred,
|
|
|
|
ap->a_p, 1)) == EINTR)
|
|
|
|
return (error);
|
|
|
|
error = nfs_setattrrpc(vp, vap, ap->a_cred, ap->a_p);
|
1995-07-24 16:38:05 +00:00
|
|
|
if (error && vap->va_size != VNOVAL) {
|
|
|
|
np->n_size = np->n_vattr.va_size = tsize;
|
1998-11-09 07:00:14 +00:00
|
|
|
vnode_pager_setsize(vp, np->n_size);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do an nfs setattr rpc.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_setattrrpc(vp, vap, cred, procp)
|
|
|
|
register struct vnode *vp;
|
|
|
|
register struct vattr *vap;
|
|
|
|
struct ucred *cred;
|
|
|
|
struct proc *procp;
|
|
|
|
{
|
|
|
|
register struct nfsv2_sattr *sp;
|
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
1998-05-31 20:09:01 +00:00
|
|
|
u_int32_t *tl;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error = 0, wccflag = NFSV3_WCCRATTR;
|
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
|
|
|
int v3 = NFS_ISV3(vp);
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsstats.rpccnt[NFSPROC_SETATTR]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(vp, NFSPROC_SETATTR, NFSX_FH(v3) + NFSX_SATTR(v3));
|
|
|
|
nfsm_fhtom(vp, v3);
|
|
|
|
if (v3) {
|
1998-12-25 10:34:27 +00:00
|
|
|
nfsm_v3attrbuild(vap, TRUE);
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
*tl = nfs_false;
|
1994-05-24 10:09:53 +00:00
|
|
|
} else {
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
|
1998-05-31 20:09:01 +00:00
|
|
|
if (vap->va_mode == (mode_t)VNOVAL)
|
1998-05-31 19:00:19 +00:00
|
|
|
sp->sa_mode = nfs_xdrneg1;
|
1995-06-27 11:07:30 +00:00
|
|
|
else
|
|
|
|
sp->sa_mode = vtonfsv2_mode(vp->v_type, vap->va_mode);
|
|
|
|
if (vap->va_uid == (uid_t)VNOVAL)
|
1998-05-31 19:00:19 +00:00
|
|
|
sp->sa_uid = nfs_xdrneg1;
|
1995-06-27 11:07:30 +00:00
|
|
|
else
|
|
|
|
sp->sa_uid = txdr_unsigned(vap->va_uid);
|
|
|
|
if (vap->va_gid == (gid_t)VNOVAL)
|
1998-05-31 19:00:19 +00:00
|
|
|
sp->sa_gid = nfs_xdrneg1;
|
1995-06-27 11:07:30 +00:00
|
|
|
else
|
|
|
|
sp->sa_gid = txdr_unsigned(vap->va_gid);
|
|
|
|
sp->sa_size = txdr_unsigned(vap->va_size);
|
|
|
|
txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
|
|
|
|
txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_request(vp, NFSPROC_SETATTR, procp, cred);
|
|
|
|
if (v3) {
|
|
|
|
nfsm_wcc_data(vp, wccflag);
|
|
|
|
} else
|
|
|
|
nfsm_loadattr(vp, (struct vattr *)0);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs lookup call, one step at a time...
|
1997-10-05 12:28:59 +00:00
|
|
|
* First look in cache
|
|
|
|
* If not found, unlock the directory nfsnode and do the rpc
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_lookup(ap)
|
1997-10-05 12:28:59 +00:00
|
|
|
struct vop_lookup_args /* {
|
1994-05-24 10:09:53 +00:00
|
|
|
struct vnodeop_desc *a_desc;
|
|
|
|
struct vnode *a_dvp;
|
|
|
|
struct vnode **a_vpp;
|
|
|
|
struct componentname *a_cnp;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
1998-05-31 17:27:58 +00:00
|
|
|
struct componentname *cnp = ap->a_cnp;
|
|
|
|
struct vnode *dvp = ap->a_dvp;
|
|
|
|
struct vnode **vpp = ap->a_vpp;
|
|
|
|
int flags = cnp->cn_flags;
|
|
|
|
struct vnode *newvp;
|
1998-05-31 20:09:01 +00:00
|
|
|
u_int32_t *tl;
|
1998-05-31 17:27:58 +00:00
|
|
|
caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
int32_t t1, t2;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct nfsmount *nmp;
|
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
|
|
|
long len;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsfh_t *fhp;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct nfsnode *np;
|
1995-06-27 11:07:30 +00:00
|
|
|
int lockparent, wantparent, error = 0, attrflag, fhsize;
|
|
|
|
int v3 = NFS_ISV3(dvp);
|
1997-02-10 02:22:35 +00:00
|
|
|
struct proc *p = cnp->cn_proc;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-05-13 05:47:09 +00:00
|
|
|
*vpp = NULLVP;
|
1997-10-05 12:28:59 +00:00
|
|
|
if ((flags & ISLASTCN) && (dvp->v_mount->mnt_flag & MNT_RDONLY) &&
|
|
|
|
(cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
|
|
|
|
return (EROFS);
|
|
|
|
if (dvp->v_type != VDIR)
|
|
|
|
return (ENOTDIR);
|
1994-05-24 10:09:53 +00:00
|
|
|
lockparent = flags & LOCKPARENT;
|
|
|
|
wantparent = flags & (LOCKPARENT|WANTPARENT);
|
|
|
|
nmp = VFSTONFS(dvp->v_mount);
|
|
|
|
np = VTONFS(dvp);
|
1997-10-05 12:28:59 +00:00
|
|
|
if ((error = cache_lookup(dvp, vpp, cnp)) && error != ENOENT) {
|
|
|
|
struct vattr vattr;
|
|
|
|
int vpid;
|
|
|
|
|
1999-01-27 22:45:13 +00:00
|
|
|
if ((error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, p)) != 0) {
|
1997-10-05 12:28:59 +00:00
|
|
|
*vpp = NULLVP;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
newvp = *vpp;
|
|
|
|
vpid = newvp->v_id;
|
|
|
|
/*
|
|
|
|
* See the comment starting `Step through' in ufs/ufs_lookup.c
|
|
|
|
* for an explanation of the locking protocol
|
|
|
|
*/
|
|
|
|
if (dvp == newvp) {
|
|
|
|
VREF(newvp);
|
|
|
|
error = 0;
|
|
|
|
} else if (flags & ISDOTDOT) {
|
|
|
|
VOP_UNLOCK(dvp, 0, p);
|
|
|
|
error = vget(newvp, LK_EXCLUSIVE, p);
|
|
|
|
if (!error && lockparent && (flags & ISLASTCN))
|
|
|
|
error = vn_lock(dvp, LK_EXCLUSIVE, p);
|
|
|
|
} else {
|
|
|
|
error = vget(newvp, LK_EXCLUSIVE, p);
|
|
|
|
if (!lockparent || error || !(flags & ISLASTCN))
|
|
|
|
VOP_UNLOCK(dvp, 0, p);
|
|
|
|
}
|
|
|
|
if (!error) {
|
|
|
|
if (vpid == newvp->v_id) {
|
|
|
|
if (!VOP_GETATTR(newvp, &vattr, cnp->cn_cred, p)
|
|
|
|
&& vattr.va_ctime.tv_sec == VTONFS(newvp)->n_ctime) {
|
|
|
|
nfsstats.lookupcache_hits++;
|
|
|
|
if (cnp->cn_nameiop != LOOKUP &&
|
|
|
|
(flags & ISLASTCN))
|
|
|
|
cnp->cn_flags |= SAVENAME;
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
cache_purge(newvp);
|
|
|
|
}
|
|
|
|
vput(newvp);
|
|
|
|
if (lockparent && dvp != newvp && (flags & ISLASTCN))
|
|
|
|
VOP_UNLOCK(dvp, 0, p);
|
|
|
|
}
|
|
|
|
error = vn_lock(dvp, LK_EXCLUSIVE, p);
|
|
|
|
*vpp = NULLVP;
|
|
|
|
if (error)
|
|
|
|
return (error);
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
error = 0;
|
1995-06-27 11:07:30 +00:00
|
|
|
newvp = NULLVP;
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsstats.lookupcache_misses++;
|
|
|
|
nfsstats.rpccnt[NFSPROC_LOOKUP]++;
|
|
|
|
len = cnp->cn_namelen;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(dvp, NFSPROC_LOOKUP,
|
|
|
|
NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len));
|
|
|
|
nfsm_fhtom(dvp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN);
|
|
|
|
nfsm_request(dvp, NFSPROC_LOOKUP, cnp->cn_proc, cnp->cn_cred);
|
|
|
|
if (error) {
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_postop_attr(dvp, attrflag);
|
|
|
|
m_freem(mrep);
|
|
|
|
goto nfsmout;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_getfh(fhp, fhsize, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Handle RENAME case...
|
|
|
|
*/
|
|
|
|
if (cnp->cn_nameiop == RENAME && wantparent && (flags & ISLASTCN)) {
|
1995-06-27 11:07:30 +00:00
|
|
|
if (NFS_CMPFH(np, fhp, fhsize)) {
|
1994-05-24 10:09:53 +00:00
|
|
|
m_freem(mrep);
|
|
|
|
return (EISDIR);
|
|
|
|
}
|
1998-05-31 17:27:58 +00:00
|
|
|
error = nfs_nget(dvp->v_mount, fhp, fhsize, &np);
|
|
|
|
if (error) {
|
1994-05-24 10:09:53 +00:00
|
|
|
m_freem(mrep);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
newvp = NFSTOV(np);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
|
|
|
nfsm_postop_attr(newvp, attrflag);
|
|
|
|
nfsm_postop_attr(dvp, attrflag);
|
|
|
|
} else
|
|
|
|
nfsm_loadattr(newvp, (struct vattr *)0);
|
1994-05-24 10:09:53 +00:00
|
|
|
*vpp = newvp;
|
|
|
|
m_freem(mrep);
|
|
|
|
cnp->cn_flags |= SAVENAME;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!lockparent)
|
1997-02-10 02:22:35 +00:00
|
|
|
VOP_UNLOCK(dvp, 0, p);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
if (flags & ISDOTDOT) {
|
1997-02-10 02:22:35 +00:00
|
|
|
VOP_UNLOCK(dvp, 0, p);
|
1995-06-27 11:07:30 +00:00
|
|
|
error = nfs_nget(dvp->v_mount, fhp, fhsize, &np);
|
|
|
|
if (error) {
|
1998-05-16 16:03:10 +00:00
|
|
|
vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, p);
|
1995-06-27 11:07:30 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
newvp = NFSTOV(np);
|
|
|
|
if (lockparent && (flags & ISLASTCN) &&
|
1997-02-10 02:22:35 +00:00
|
|
|
(error = vn_lock(dvp, LK_EXCLUSIVE, p))) {
|
1995-06-27 11:07:30 +00:00
|
|
|
vput(newvp);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
} else if (NFS_CMPFH(np, fhp, fhsize)) {
|
1994-05-24 10:09:53 +00:00
|
|
|
VREF(dvp);
|
|
|
|
newvp = dvp;
|
|
|
|
} else {
|
1998-05-31 17:27:58 +00:00
|
|
|
error = nfs_nget(dvp->v_mount, fhp, fhsize, &np);
|
|
|
|
if (error) {
|
1994-05-24 10:09:53 +00:00
|
|
|
m_freem(mrep);
|
|
|
|
return (error);
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!lockparent || !(flags & ISLASTCN))
|
1997-02-10 02:22:35 +00:00
|
|
|
VOP_UNLOCK(dvp, 0, p);
|
1994-05-24 10:09:53 +00:00
|
|
|
newvp = NFSTOV(np);
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
|
|
|
nfsm_postop_attr(newvp, attrflag);
|
|
|
|
nfsm_postop_attr(dvp, attrflag);
|
|
|
|
} else
|
|
|
|
nfsm_loadattr(newvp, (struct vattr *)0);
|
1994-05-24 10:09:53 +00:00
|
|
|
if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN))
|
|
|
|
cnp->cn_flags |= SAVENAME;
|
|
|
|
if ((cnp->cn_flags & MAKEENTRY) &&
|
|
|
|
(cnp->cn_nameiop != DELETE || !(flags & ISLASTCN))) {
|
1996-09-19 18:21:32 +00:00
|
|
|
np->n_ctime = np->n_vattr.va_ctime.tv_sec;
|
1995-06-27 11:07:30 +00:00
|
|
|
cache_enter(dvp, newvp, cnp);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
*vpp = newvp;
|
|
|
|
nfsm_reqdone;
|
|
|
|
if (error) {
|
1997-06-25 08:32:33 +00:00
|
|
|
if (newvp != NULLVP) {
|
1995-06-27 11:07:30 +00:00
|
|
|
vrele(newvp);
|
1997-06-25 08:32:33 +00:00
|
|
|
*vpp = NULLVP;
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
if ((cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME) &&
|
|
|
|
(flags & ISLASTCN) && error == ENOENT) {
|
|
|
|
if (!lockparent)
|
1997-02-10 02:22:35 +00:00
|
|
|
VOP_UNLOCK(dvp, 0, p);
|
1995-10-22 09:32:48 +00:00
|
|
|
if (dvp->v_mount->mnt_flag & MNT_RDONLY)
|
|
|
|
error = EROFS;
|
|
|
|
else
|
|
|
|
error = EJUSTRETURN;
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
|
|
|
if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN))
|
|
|
|
cnp->cn_flags |= SAVENAME;
|
|
|
|
}
|
|
|
|
return (error);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs read call.
|
|
|
|
* Just call nfs_bioread() to do the work.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_read(ap)
|
|
|
|
struct vop_read_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct uio *a_uio;
|
|
|
|
int a_ioflag;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
|
|
|
|
if (vp->v_type != VREG)
|
|
|
|
return (EPERM);
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
return (nfs_bioread(vp, ap->a_uio, ap->a_ioflag, ap->a_cred));
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs readlink call
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_readlink(ap)
|
|
|
|
struct vop_readlink_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct uio *a_uio;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
|
|
|
|
if (vp->v_type != VLNK)
|
1998-06-01 10:59:23 +00:00
|
|
|
return (EINVAL);
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
return (nfs_bioread(vp, ap->a_uio, 0, ap->a_cred));
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do a readlink rpc.
|
|
|
|
* Called by nfs_doio() from below the buffer cache.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
nfs_readlinkrpc(vp, uiop, cred)
|
|
|
|
register struct vnode *vp;
|
|
|
|
struct uio *uiop;
|
|
|
|
struct ucred *cred;
|
|
|
|
{
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1994-05-24 10:09:53 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error = 0, len, attrflag;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(vp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
nfsstats.rpccnt[NFSPROC_READLINK]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(vp, NFSPROC_READLINK, NFSX_FH(v3));
|
|
|
|
nfsm_fhtom(vp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_request(vp, NFSPROC_READLINK, uiop->uio_procp, cred);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3)
|
|
|
|
nfsm_postop_attr(vp, attrflag);
|
|
|
|
if (!error) {
|
|
|
|
nfsm_strsiz(len, NFS_MAXPATHLEN);
|
1999-06-30 02:53:51 +00:00
|
|
|
if (len == NFS_MAXPATHLEN) {
|
|
|
|
struct nfsnode *np = VTONFS(vp);
|
|
|
|
if (np->n_size && np->n_size < NFS_MAXPATHLEN)
|
|
|
|
len = np->n_size;
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_mtouio(uiop, len);
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs read rpc call
|
|
|
|
* Ditto above
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
nfs_readrpc(vp, uiop, cred)
|
|
|
|
register struct vnode *vp;
|
|
|
|
struct uio *uiop;
|
|
|
|
struct ucred *cred;
|
|
|
|
{
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1994-05-24 10:09:53 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
|
|
|
struct nfsmount *nmp;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error = 0, len, retlen, tsiz, eof, attrflag;
|
|
|
|
int v3 = NFS_ISV3(vp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
#ifndef nolint
|
|
|
|
eof = 0;
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
nmp = VFSTONFS(vp->v_mount);
|
|
|
|
tsiz = uiop->uio_resid;
|
1998-05-30 16:33:58 +00:00
|
|
|
if (uiop->uio_offset + tsiz > nmp->nm_maxfilesize)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EFBIG);
|
|
|
|
while (tsiz > 0) {
|
|
|
|
nfsstats.rpccnt[NFSPROC_READ]++;
|
|
|
|
len = (tsiz > nmp->nm_rsize) ? nmp->nm_rsize : tsiz;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(vp, NFSPROC_READ, NFSX_FH(v3) + NFSX_UNSIGNED * 3);
|
|
|
|
nfsm_fhtom(vp, v3);
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED * 3);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
1999-06-05 05:35:03 +00:00
|
|
|
txdr_hyper(uiop->uio_offset, tl);
|
1994-05-24 10:09:53 +00:00
|
|
|
*(tl + 2) = txdr_unsigned(len);
|
|
|
|
} else {
|
|
|
|
*tl++ = txdr_unsigned(uiop->uio_offset);
|
|
|
|
*tl++ = txdr_unsigned(len);
|
|
|
|
*tl = 0;
|
|
|
|
}
|
|
|
|
nfsm_request(vp, NFSPROC_READ, uiop->uio_procp, cred);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
|
|
|
nfsm_postop_attr(vp, attrflag);
|
|
|
|
if (error) {
|
|
|
|
m_freem(mrep);
|
|
|
|
goto nfsmout;
|
|
|
|
}
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
eof = fxdr_unsigned(int, *(tl + 1));
|
|
|
|
} else
|
|
|
|
nfsm_loadattr(vp, (struct vattr *)0);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strsiz(retlen, nmp->nm_rsize);
|
|
|
|
nfsm_mtouio(uiop, retlen);
|
|
|
|
m_freem(mrep);
|
1995-06-27 11:07:30 +00:00
|
|
|
tsiz -= retlen;
|
|
|
|
if (v3) {
|
|
|
|
if (eof || retlen == 0)
|
|
|
|
tsiz = 0;
|
|
|
|
} else if (retlen < len)
|
1994-05-24 10:09:53 +00:00
|
|
|
tsiz = 0;
|
|
|
|
}
|
|
|
|
nfsmout:
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs write call
|
|
|
|
*/
|
|
|
|
int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_writerpc(vp, uiop, cred, iomode, must_commit)
|
1994-05-24 10:09:53 +00:00
|
|
|
register struct vnode *vp;
|
1995-06-27 11:07:30 +00:00
|
|
|
register struct uio *uiop;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct ucred *cred;
|
1995-06-27 11:07:30 +00:00
|
|
|
int *iomode, *must_commit;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2, backup;
|
1994-05-24 10:09:53 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
|
|
|
|
int error = 0, len, tsiz, wccflag = NFSV3_WCCRATTR, rlen, commit;
|
|
|
|
int v3 = NFS_ISV3(vp), committed = NFSV3WRITE_FILESYNC;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
#ifndef DIAGNOSTIC
|
|
|
|
if (uiop->uio_iovcnt != 1)
|
|
|
|
panic("nfs: writerpc iovcnt > 1");
|
|
|
|
#endif
|
|
|
|
*must_commit = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
tsiz = uiop->uio_resid;
|
1998-05-30 16:33:58 +00:00
|
|
|
if (uiop->uio_offset + tsiz > nmp->nm_maxfilesize)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EFBIG);
|
|
|
|
while (tsiz > 0) {
|
|
|
|
nfsstats.rpccnt[NFSPROC_WRITE]++;
|
|
|
|
len = (tsiz > nmp->nm_wsize) ? nmp->nm_wsize : tsiz;
|
|
|
|
nfsm_reqhead(vp, NFSPROC_WRITE,
|
1995-06-27 11:07:30 +00:00
|
|
|
NFSX_FH(v3) + 5 * NFSX_UNSIGNED + nfsm_rndup(len));
|
|
|
|
nfsm_fhtom(vp, v3);
|
|
|
|
if (v3) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
|
1999-06-05 05:35:03 +00:00
|
|
|
txdr_hyper(uiop->uio_offset, tl);
|
1994-05-24 10:09:53 +00:00
|
|
|
tl += 2;
|
1995-06-27 11:07:30 +00:00
|
|
|
*tl++ = txdr_unsigned(len);
|
|
|
|
*tl++ = txdr_unsigned(*iomode);
|
1998-05-31 19:28:15 +00:00
|
|
|
*tl = txdr_unsigned(len);
|
1994-05-24 10:09:53 +00:00
|
|
|
} else {
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t x;
|
1998-05-31 19:28:15 +00:00
|
|
|
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
|
1998-05-31 19:28:15 +00:00
|
|
|
/* Set both "begin" and "current" to non-garbage. */
|
1998-05-31 20:09:01 +00:00
|
|
|
x = txdr_unsigned((u_int32_t)uiop->uio_offset);
|
1998-05-31 19:28:15 +00:00
|
|
|
*tl++ = x; /* "begin offset" */
|
|
|
|
*tl++ = x; /* "current offset" */
|
|
|
|
x = txdr_unsigned(len);
|
|
|
|
*tl++ = x; /* total to this offset */
|
|
|
|
*tl = x; /* size of this write */
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
nfsm_uiotom(uiop, len);
|
|
|
|
nfsm_request(vp, NFSPROC_WRITE, uiop->uio_procp, cred);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
|
|
|
wccflag = NFSV3_WCCCHK;
|
|
|
|
nfsm_wcc_data(vp, wccflag);
|
|
|
|
if (!error) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED
|
|
|
|
+ NFSX_V3WRITEVERF);
|
1995-06-27 11:07:30 +00:00
|
|
|
rlen = fxdr_unsigned(int, *tl++);
|
|
|
|
if (rlen == 0) {
|
|
|
|
error = NFSERR_IO;
|
1998-05-31 18:25:32 +00:00
|
|
|
m_freem(mrep);
|
1995-06-27 11:07:30 +00:00
|
|
|
break;
|
|
|
|
} else if (rlen < len) {
|
|
|
|
backup = len - rlen;
|
|
|
|
uiop->uio_iov->iov_base -= backup;
|
|
|
|
uiop->uio_iov->iov_len += backup;
|
|
|
|
uiop->uio_offset -= backup;
|
|
|
|
uiop->uio_resid += backup;
|
|
|
|
len = rlen;
|
|
|
|
}
|
|
|
|
commit = fxdr_unsigned(int, *tl++);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return the lowest committment level
|
|
|
|
* obtained by any of the RPCs.
|
|
|
|
*/
|
|
|
|
if (committed == NFSV3WRITE_FILESYNC)
|
|
|
|
committed = commit;
|
|
|
|
else if (committed == NFSV3WRITE_DATASYNC &&
|
|
|
|
commit == NFSV3WRITE_UNSTABLE)
|
|
|
|
committed = commit;
|
1998-05-31 17:27:58 +00:00
|
|
|
if ((nmp->nm_state & NFSSTA_HASWRITEVERF) == 0){
|
1995-06-27 11:07:30 +00:00
|
|
|
bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf,
|
|
|
|
NFSX_V3WRITEVERF);
|
1998-05-19 07:11:27 +00:00
|
|
|
nmp->nm_state |= NFSSTA_HASWRITEVERF;
|
1995-06-27 11:07:30 +00:00
|
|
|
} else if (bcmp((caddr_t)tl,
|
|
|
|
(caddr_t)nmp->nm_verf, NFSX_V3WRITEVERF)) {
|
|
|
|
*must_commit = 1;
|
|
|
|
bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf,
|
|
|
|
NFSX_V3WRITEVERF);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else
|
|
|
|
nfsm_loadattr(vp, (struct vattr *)0);
|
|
|
|
if (wccflag)
|
1996-09-19 18:21:32 +00:00
|
|
|
VTONFS(vp)->n_mtime = VTONFS(vp)->n_vattr.va_mtime.tv_sec;
|
1994-05-24 10:09:53 +00:00
|
|
|
m_freem(mrep);
|
1998-05-31 18:27:07 +00:00
|
|
|
if (error)
|
|
|
|
break;
|
1994-05-24 10:09:53 +00:00
|
|
|
tsiz -= len;
|
|
|
|
}
|
|
|
|
nfsmout:
|
1997-06-03 13:56:55 +00:00
|
|
|
if (vp->v_mount->mnt_flag & MNT_ASYNC)
|
|
|
|
committed = NFSV3WRITE_FILESYNC;
|
1995-06-27 11:07:30 +00:00
|
|
|
*iomode = committed;
|
1994-05-24 10:09:53 +00:00
|
|
|
if (error)
|
|
|
|
uiop->uio_resid = tsiz;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
1995-06-27 11:07:30 +00:00
|
|
|
* nfs mknod rpc
|
|
|
|
* For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
|
|
|
|
* mode set to specify the file type and the size field for rdev.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_mknodrpc(dvp, vpp, cnp, vap)
|
|
|
|
register struct vnode *dvp;
|
|
|
|
register struct vnode **vpp;
|
|
|
|
register struct componentname *cnp;
|
|
|
|
register struct vattr *vap;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
|
|
|
register struct nfsv2_sattr *sp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct vnode *newvp = (struct vnode *)0;
|
|
|
|
struct nfsnode *np = (struct nfsnode *)0;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct vattr vattr;
|
|
|
|
char *cp2;
|
|
|
|
caddr_t bpos, dpos;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error = 0, wccflag = NFSV3_WCCRATTR, gotvp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1998-05-31 20:09:01 +00:00
|
|
|
u_int32_t rdev;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(dvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
if (vap->va_type == VCHR || vap->va_type == VBLK)
|
|
|
|
rdev = txdr_unsigned(vap->va_rdev);
|
1995-06-27 11:07:30 +00:00
|
|
|
else if (vap->va_type == VFIFO || vap->va_type == VSOCK)
|
1998-05-31 19:29:28 +00:00
|
|
|
rdev = nfs_xdrneg1;
|
1994-05-24 10:09:53 +00:00
|
|
|
else {
|
|
|
|
VOP_ABORTOP(dvp, cnp);
|
|
|
|
return (EOPNOTSUPP);
|
|
|
|
}
|
1999-01-27 22:45:13 +00:00
|
|
|
if ((error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) != 0) {
|
1994-05-24 10:09:53 +00:00
|
|
|
VOP_ABORTOP(dvp, cnp);
|
|
|
|
return (error);
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsstats.rpccnt[NFSPROC_MKNOD]++;
|
|
|
|
nfsm_reqhead(dvp, NFSPROC_MKNOD, NFSX_FH(v3) + 4 * NFSX_UNSIGNED +
|
|
|
|
+ nfsm_rndup(cnp->cn_namelen) + NFSX_SATTR(v3));
|
|
|
|
nfsm_fhtom(dvp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
1998-12-25 10:34:27 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
*tl++ = vtonfsv3_type(vap->va_type);
|
1998-12-25 10:34:27 +00:00
|
|
|
nfsm_v3attrbuild(vap, FALSE);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (vap->va_type == VCHR || vap->va_type == VBLK) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
|
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
|
|
|
*tl++ = txdr_unsigned(umajor(vap->va_rdev));
|
|
|
|
*tl = txdr_unsigned(uminor(vap->va_rdev));
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
} else {
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
|
|
|
|
sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
|
1998-05-31 19:00:19 +00:00
|
|
|
sp->sa_uid = nfs_xdrneg1;
|
|
|
|
sp->sa_gid = nfs_xdrneg1;
|
1995-06-27 11:07:30 +00:00
|
|
|
sp->sa_size = rdev;
|
|
|
|
txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
|
|
|
|
txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_request(dvp, NFSPROC_MKNOD, cnp->cn_proc, cnp->cn_cred);
|
|
|
|
if (!error) {
|
|
|
|
nfsm_mtofh(dvp, newvp, v3, gotvp);
|
|
|
|
if (!gotvp) {
|
|
|
|
if (newvp) {
|
|
|
|
vput(newvp);
|
|
|
|
newvp = (struct vnode *)0;
|
|
|
|
}
|
|
|
|
error = nfs_lookitup(dvp, cnp->cn_nameptr,
|
|
|
|
cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc, &np);
|
|
|
|
if (!error)
|
|
|
|
newvp = NFSTOV(np);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (v3)
|
|
|
|
nfsm_wcc_data(dvp, wccflag);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (error) {
|
|
|
|
if (newvp)
|
|
|
|
vput(newvp);
|
|
|
|
} else {
|
|
|
|
if (cnp->cn_flags & MAKEENTRY)
|
|
|
|
cache_enter(dvp, newvp, cnp);
|
|
|
|
*vpp = newvp;
|
|
|
|
}
|
1997-09-21 04:24:27 +00:00
|
|
|
zfree(namei_zone, cnp->cn_pnbuf);
|
1994-05-24 10:09:53 +00:00
|
|
|
VTONFS(dvp)->n_flag |= NMODIFIED;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!wccflag)
|
|
|
|
VTONFS(dvp)->n_attrstamp = 0;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs mknod vop
|
|
|
|
* just call nfs_mknodrpc() to do the work.
|
|
|
|
*/
|
|
|
|
/* ARGSUSED */
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_mknod(ap)
|
|
|
|
struct vop_mknod_args /* {
|
|
|
|
struct vnode *a_dvp;
|
|
|
|
struct vnode **a_vpp;
|
|
|
|
struct componentname *a_cnp;
|
|
|
|
struct vattr *a_vap;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
struct vnode *newvp;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
error = nfs_mknodrpc(ap->a_dvp, &newvp, ap->a_cnp, ap->a_vap);
|
|
|
|
if (!error)
|
|
|
|
vput(newvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
static u_long create_verf;
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* nfs file create call
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_create(ap)
|
|
|
|
struct vop_create_args /* {
|
|
|
|
struct vnode *a_dvp;
|
|
|
|
struct vnode **a_vpp;
|
|
|
|
struct componentname *a_cnp;
|
|
|
|
struct vattr *a_vap;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *dvp = ap->a_dvp;
|
|
|
|
register struct vattr *vap = ap->a_vap;
|
|
|
|
register struct componentname *cnp = ap->a_cnp;
|
|
|
|
register struct nfsv2_sattr *sp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct nfsnode *np = (struct nfsnode *)0;
|
|
|
|
struct vnode *newvp = (struct vnode *)0;
|
1994-05-24 10:09:53 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error = 0, wccflag = NFSV3_WCCRATTR, gotvp = 0, fmode = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
|
|
|
struct vattr vattr;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(dvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
/*
|
|
|
|
* Oops, not for me..
|
|
|
|
*/
|
|
|
|
if (vap->va_type == VSOCK)
|
|
|
|
return (nfs_mknodrpc(dvp, ap->a_vpp, cnp, vap));
|
|
|
|
|
1999-01-27 22:45:13 +00:00
|
|
|
if ((error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) != 0) {
|
1994-05-24 10:09:53 +00:00
|
|
|
VOP_ABORTOP(dvp, cnp);
|
|
|
|
return (error);
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
if (vap->va_vaflags & VA_EXCLUSIVE)
|
|
|
|
fmode |= O_EXCL;
|
|
|
|
again:
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsstats.rpccnt[NFSPROC_CREATE]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(dvp, NFSPROC_CREATE, NFSX_FH(v3) + 2 * NFSX_UNSIGNED +
|
|
|
|
nfsm_rndup(cnp->cn_namelen) + NFSX_SATTR(v3));
|
|
|
|
nfsm_fhtom(dvp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (fmode & O_EXCL) {
|
1998-12-25 10:34:27 +00:00
|
|
|
*tl = txdr_unsigned(NFSV3CREATE_EXCLUSIVE);
|
|
|
|
nfsm_build(tl, u_int32_t *, NFSX_V3CREATEVERF);
|
1998-01-08 23:42:31 +00:00
|
|
|
#ifdef INET
|
1998-12-25 10:34:27 +00:00
|
|
|
if (!TAILQ_EMPTY(&in_ifaddrhead))
|
|
|
|
*tl++ = IA_SIN(in_ifaddrhead.tqh_first)->sin_addr.s_addr;
|
|
|
|
else
|
1998-01-08 23:42:31 +00:00
|
|
|
#endif
|
1998-12-25 10:34:27 +00:00
|
|
|
*tl++ = create_verf;
|
|
|
|
*tl = ++create_verf;
|
1995-06-27 11:07:30 +00:00
|
|
|
} else {
|
1998-12-25 10:34:27 +00:00
|
|
|
*tl = txdr_unsigned(NFSV3CREATE_UNCHECKED);
|
|
|
|
nfsm_v3attrbuild(vap, FALSE);
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
} else {
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
|
|
|
|
sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
|
1998-05-31 19:00:19 +00:00
|
|
|
sp->sa_uid = nfs_xdrneg1;
|
|
|
|
sp->sa_gid = nfs_xdrneg1;
|
1995-06-27 11:07:30 +00:00
|
|
|
sp->sa_size = 0;
|
|
|
|
txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
|
|
|
|
txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
nfsm_request(dvp, NFSPROC_CREATE, cnp->cn_proc, cnp->cn_cred);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!error) {
|
|
|
|
nfsm_mtofh(dvp, newvp, v3, gotvp);
|
|
|
|
if (!gotvp) {
|
|
|
|
if (newvp) {
|
|
|
|
vput(newvp);
|
|
|
|
newvp = (struct vnode *)0;
|
|
|
|
}
|
|
|
|
error = nfs_lookitup(dvp, cnp->cn_nameptr,
|
|
|
|
cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc, &np);
|
|
|
|
if (!error)
|
|
|
|
newvp = NFSTOV(np);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (v3)
|
|
|
|
nfsm_wcc_data(dvp, wccflag);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (error) {
|
|
|
|
if (v3 && (fmode & O_EXCL) && error == NFSERR_NOTSUPP) {
|
|
|
|
fmode &= ~O_EXCL;
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
if (newvp)
|
|
|
|
vput(newvp);
|
|
|
|
} else if (v3 && (fmode & O_EXCL))
|
|
|
|
error = nfs_setattrrpc(newvp, vap, cnp->cn_cred, cnp->cn_proc);
|
|
|
|
if (!error) {
|
|
|
|
if (cnp->cn_flags & MAKEENTRY)
|
|
|
|
cache_enter(dvp, newvp, cnp);
|
|
|
|
*ap->a_vpp = newvp;
|
|
|
|
}
|
1999-02-13 09:47:30 +00:00
|
|
|
if (error || (cnp->cn_flags & SAVESTART) == 0)
|
|
|
|
zfree(namei_zone, cnp->cn_pnbuf);
|
1994-05-24 10:09:53 +00:00
|
|
|
VTONFS(dvp)->n_flag |= NMODIFIED;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!wccflag)
|
|
|
|
VTONFS(dvp)->n_attrstamp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs file remove call
|
|
|
|
* To try and make nfs semantics closer to ufs semantics, a file that has
|
|
|
|
* other processes using the vnode is renamed instead of removed and then
|
|
|
|
* removed later on the last close.
|
|
|
|
* - If v_usecount > 1
|
|
|
|
* If a rename is not already in the works
|
|
|
|
* call nfs_sillyrename() to set it up
|
|
|
|
* else
|
|
|
|
* do the remove rpc
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_remove(ap)
|
|
|
|
struct vop_remove_args /* {
|
|
|
|
struct vnodeop_desc *a_desc;
|
|
|
|
struct vnode * a_dvp;
|
|
|
|
struct vnode * a_vp;
|
|
|
|
struct componentname * a_cnp;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct vnode *dvp = ap->a_dvp;
|
|
|
|
register struct componentname *cnp = ap->a_cnp;
|
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
int error = 0;
|
1994-10-17 17:47:45 +00:00
|
|
|
struct vattr vattr;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
#ifndef DIAGNOSTIC
|
|
|
|
if ((cnp->cn_flags & HASBUF) == 0)
|
|
|
|
panic("nfs_remove: no name");
|
|
|
|
if (vp->v_usecount < 1)
|
|
|
|
panic("nfs_remove: bad v_usecount");
|
|
|
|
#endif
|
1998-05-31 18:28:45 +00:00
|
|
|
if (vp->v_type == VDIR)
|
|
|
|
error = EPERM;
|
|
|
|
else if (vp->v_usecount == 1 || (np->n_sillyrename &&
|
1995-06-27 11:07:30 +00:00
|
|
|
VOP_GETATTR(vp, &vattr, cnp->cn_cred, cnp->cn_proc) == 0 &&
|
|
|
|
vattr.va_nlink > 1)) {
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Purge the name cache so that the chance of a lookup for
|
|
|
|
* the name succeeding while the remove is in progress is
|
|
|
|
* minimized. Without node locking it can still happen, such
|
|
|
|
* that an I/O op returns ESTALE, but since you get this if
|
|
|
|
* another host removes the file..
|
|
|
|
*/
|
|
|
|
cache_purge(vp);
|
|
|
|
/*
|
1995-06-27 11:07:30 +00:00
|
|
|
* throw away biocache buffers, mainly to avoid
|
|
|
|
* unnecessary delayed writes later.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
error = nfs_vinvalbuf(vp, 0, cnp->cn_cred, cnp->cn_proc, 1);
|
|
|
|
/* Do the rpc */
|
1995-06-27 11:07:30 +00:00
|
|
|
if (error != EINTR)
|
|
|
|
error = nfs_removerpc(dvp, cnp->cn_nameptr,
|
|
|
|
cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc);
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Kludge City: If the first reply to the remove rpc is lost..
|
|
|
|
* the reply to the retransmitted request will be ENOENT
|
|
|
|
* since the file was in fact removed
|
|
|
|
* Therefore, we cheat and return success.
|
|
|
|
*/
|
|
|
|
if (error == ENOENT)
|
|
|
|
error = 0;
|
1995-06-27 11:07:30 +00:00
|
|
|
} else if (!np->n_sillyrename)
|
|
|
|
error = nfs_sillyrename(dvp, vp, cnp);
|
1997-09-21 04:24:27 +00:00
|
|
|
zfree(namei_zone, cnp->cn_pnbuf);
|
1994-05-24 10:09:53 +00:00
|
|
|
np->n_attrstamp = 0;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs file remove rpc called from nfs_inactive
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
nfs_removeit(sp)
|
|
|
|
register struct sillyrename *sp;
|
|
|
|
{
|
1995-06-27 11:07:30 +00:00
|
|
|
|
|
|
|
return (nfs_removerpc(sp->s_dvp, sp->s_name, sp->s_namlen, sp->s_cred,
|
|
|
|
(struct proc *)0));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Nfs remove rpc, called from nfs_remove() and nfs_removeit().
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_removerpc(dvp, name, namelen, cred, proc)
|
|
|
|
register struct vnode *dvp;
|
1998-05-31 17:48:07 +00:00
|
|
|
const char *name;
|
1995-06-27 11:07:30 +00:00
|
|
|
int namelen;
|
|
|
|
struct ucred *cred;
|
|
|
|
struct proc *proc;
|
|
|
|
{
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
int error = 0, wccflag = NFSV3_WCCRATTR;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(dvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
nfsstats.rpccnt[NFSPROC_REMOVE]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(dvp, NFSPROC_REMOVE,
|
|
|
|
NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(namelen));
|
|
|
|
nfsm_fhtom(dvp, v3);
|
|
|
|
nfsm_strtom(name, namelen, NFS_MAXNAMLEN);
|
|
|
|
nfsm_request(dvp, NFSPROC_REMOVE, proc, cred);
|
|
|
|
if (v3)
|
|
|
|
nfsm_wcc_data(dvp, wccflag);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
1995-06-27 11:07:30 +00:00
|
|
|
VTONFS(dvp)->n_flag |= NMODIFIED;
|
|
|
|
if (!wccflag)
|
|
|
|
VTONFS(dvp)->n_attrstamp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs file rename call
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_rename(ap)
|
|
|
|
struct vop_rename_args /* {
|
|
|
|
struct vnode *a_fdvp;
|
|
|
|
struct vnode *a_fvp;
|
|
|
|
struct componentname *a_fcnp;
|
|
|
|
struct vnode *a_tdvp;
|
|
|
|
struct vnode *a_tvp;
|
|
|
|
struct componentname *a_tcnp;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *fvp = ap->a_fvp;
|
|
|
|
register struct vnode *tvp = ap->a_tvp;
|
|
|
|
register struct vnode *fdvp = ap->a_fdvp;
|
|
|
|
register struct vnode *tdvp = ap->a_tdvp;
|
|
|
|
register struct componentname *tcnp = ap->a_tcnp;
|
|
|
|
register struct componentname *fcnp = ap->a_fcnp;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
#ifndef DIAGNOSTIC
|
|
|
|
if ((tcnp->cn_flags & HASBUF) == 0 ||
|
|
|
|
(fcnp->cn_flags & HASBUF) == 0)
|
|
|
|
panic("nfs_rename: no name");
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
/* Check for cross-device rename */
|
|
|
|
if ((fvp->v_mount != tdvp->v_mount) ||
|
|
|
|
(tvp && (fvp->v_mount != tvp->v_mount))) {
|
|
|
|
error = EXDEV;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
1999-02-06 07:48:56 +00:00
|
|
|
/*
|
|
|
|
* We have to flush B_DELWRI data prior to renaming
|
|
|
|
* the file. If we don't, the delayed-write buffers
|
|
|
|
* can be flushed out later after the file has gone stale
|
|
|
|
* under NFSV3. NFSV2 does not have this problem because
|
|
|
|
* ( as far as I can tell ) it flushes dirty buffers more
|
|
|
|
* often.
|
|
|
|
*/
|
|
|
|
|
|
|
|
VOP_FSYNC(fvp, fcnp->cn_cred, MNT_WAIT, fcnp->cn_proc);
|
|
|
|
if (tvp)
|
|
|
|
VOP_FSYNC(tvp, tcnp->cn_cred, MNT_WAIT, tcnp->cn_proc);
|
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
/*
|
|
|
|
* If the tvp exists and is in use, sillyrename it before doing the
|
|
|
|
* rename of the new file over it.
|
1996-12-31 07:10:19 +00:00
|
|
|
* XXX Can't sillyrename a directory.
|
1995-06-27 11:07:30 +00:00
|
|
|
*/
|
|
|
|
if (tvp && tvp->v_usecount > 1 && !VTONFS(tvp)->n_sillyrename &&
|
1996-12-31 07:10:19 +00:00
|
|
|
tvp->v_type != VDIR && !nfs_sillyrename(tdvp, tvp, tcnp)) {
|
1997-04-04 17:49:35 +00:00
|
|
|
vput(tvp);
|
1995-06-27 11:07:30 +00:00
|
|
|
tvp = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = nfs_renamerpc(fdvp, fcnp->cn_nameptr, fcnp->cn_namelen,
|
|
|
|
tdvp, tcnp->cn_nameptr, tcnp->cn_namelen, tcnp->cn_cred,
|
|
|
|
tcnp->cn_proc);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
if (fvp->v_type == VDIR) {
|
|
|
|
if (tvp != NULL && tvp->v_type == VDIR)
|
|
|
|
cache_purge(tdvp);
|
|
|
|
cache_purge(fdvp);
|
|
|
|
}
|
1998-11-13 02:39:09 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
out:
|
1999-02-13 09:47:30 +00:00
|
|
|
VOP_ABORTOP(tdvp, tcnp);
|
1994-05-24 10:09:53 +00:00
|
|
|
if (tdvp == tvp)
|
|
|
|
vrele(tdvp);
|
|
|
|
else
|
|
|
|
vput(tdvp);
|
|
|
|
if (tvp)
|
|
|
|
vput(tvp);
|
1999-02-13 09:47:30 +00:00
|
|
|
VOP_ABORTOP(fdvp, fcnp);
|
1994-05-24 10:09:53 +00:00
|
|
|
vrele(fdvp);
|
|
|
|
vrele(fvp);
|
|
|
|
/*
|
|
|
|
* Kludge: Map ENOENT => 0 assuming that it is a reply to a retry.
|
|
|
|
*/
|
|
|
|
if (error == ENOENT)
|
|
|
|
error = 0;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs file rename rpc called from nfs_remove() above
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_renameit(sdvp, scnp, sp)
|
|
|
|
struct vnode *sdvp;
|
|
|
|
struct componentname *scnp;
|
|
|
|
register struct sillyrename *sp;
|
1995-06-27 11:07:30 +00:00
|
|
|
{
|
|
|
|
return (nfs_renamerpc(sdvp, scnp->cn_nameptr, scnp->cn_namelen,
|
|
|
|
sdvp, sp->s_name, sp->s_namlen, scnp->cn_cred, scnp->cn_proc));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do an nfs rename rpc. Called from nfs_rename() and nfs_renameit().
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_renamerpc(fdvp, fnameptr, fnamelen, tdvp, tnameptr, tnamelen, cred, proc)
|
|
|
|
register struct vnode *fdvp;
|
1998-05-31 17:48:07 +00:00
|
|
|
const char *fnameptr;
|
1995-06-27 11:07:30 +00:00
|
|
|
int fnamelen;
|
|
|
|
register struct vnode *tdvp;
|
1998-05-31 17:48:07 +00:00
|
|
|
const char *tnameptr;
|
1995-06-27 11:07:30 +00:00
|
|
|
int tnamelen;
|
|
|
|
struct ucred *cred;
|
|
|
|
struct proc *proc;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
int error = 0, fwccflag = NFSV3_WCCRATTR, twccflag = NFSV3_WCCRATTR;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(fdvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
nfsstats.rpccnt[NFSPROC_RENAME]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(fdvp, NFSPROC_RENAME,
|
|
|
|
(NFSX_FH(v3) + NFSX_UNSIGNED)*2 + nfsm_rndup(fnamelen) +
|
|
|
|
nfsm_rndup(tnamelen));
|
|
|
|
nfsm_fhtom(fdvp, v3);
|
|
|
|
nfsm_strtom(fnameptr, fnamelen, NFS_MAXNAMLEN);
|
|
|
|
nfsm_fhtom(tdvp, v3);
|
|
|
|
nfsm_strtom(tnameptr, tnamelen, NFS_MAXNAMLEN);
|
|
|
|
nfsm_request(fdvp, NFSPROC_RENAME, proc, cred);
|
|
|
|
if (v3) {
|
|
|
|
nfsm_wcc_data(fdvp, fwccflag);
|
|
|
|
nfsm_wcc_data(tdvp, twccflag);
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
1995-06-27 11:07:30 +00:00
|
|
|
VTONFS(fdvp)->n_flag |= NMODIFIED;
|
|
|
|
VTONFS(tdvp)->n_flag |= NMODIFIED;
|
|
|
|
if (!fwccflag)
|
|
|
|
VTONFS(fdvp)->n_attrstamp = 0;
|
|
|
|
if (!twccflag)
|
|
|
|
VTONFS(tdvp)->n_attrstamp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs hard link create call
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_link(ap)
|
|
|
|
struct vop_link_args /* {
|
|
|
|
struct vnode *a_tdvp;
|
1995-08-01 18:51:02 +00:00
|
|
|
struct vnode *a_vp;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct componentname *a_cnp;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct vnode *tdvp = ap->a_tdvp;
|
|
|
|
register struct componentname *cnp = ap->a_cnp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
int error = 0, wccflag = NFSV3_WCCRATTR, attrflag = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1998-09-29 23:39:37 +00:00
|
|
|
int v3;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
if (vp->v_mount != tdvp->v_mount) {
|
1998-05-31 01:03:07 +00:00
|
|
|
VOP_ABORTOP(tdvp, cnp);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EXDEV);
|
|
|
|
}
|
|
|
|
|
1994-10-17 17:47:45 +00:00
|
|
|
/*
|
|
|
|
* Push all writes to the server, so that the attribute cache
|
|
|
|
* doesn't get "out of sync" with the server.
|
|
|
|
* XXX There should be a better way!
|
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
VOP_FSYNC(vp, cnp->cn_cred, MNT_WAIT, cnp->cn_proc);
|
1994-10-17 17:47:45 +00:00
|
|
|
|
1998-09-29 23:39:37 +00:00
|
|
|
v3 = NFS_ISV3(vp);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsstats.rpccnt[NFSPROC_LINK]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(vp, NFSPROC_LINK,
|
|
|
|
NFSX_FH(v3)*2 + NFSX_UNSIGNED + nfsm_rndup(cnp->cn_namelen));
|
|
|
|
nfsm_fhtom(vp, v3);
|
|
|
|
nfsm_fhtom(tdvp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_request(vp, NFSPROC_LINK, cnp->cn_proc, cnp->cn_cred);
|
|
|
|
if (v3) {
|
|
|
|
nfsm_postop_attr(vp, attrflag);
|
|
|
|
nfsm_wcc_data(tdvp, wccflag);
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
1997-09-21 04:24:27 +00:00
|
|
|
zfree(namei_zone, cnp->cn_pnbuf);
|
1995-06-27 11:07:30 +00:00
|
|
|
VTONFS(tdvp)->n_flag |= NMODIFIED;
|
|
|
|
if (!attrflag)
|
|
|
|
VTONFS(vp)->n_attrstamp = 0;
|
|
|
|
if (!wccflag)
|
|
|
|
VTONFS(tdvp)->n_attrstamp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
|
|
|
|
*/
|
|
|
|
if (error == EEXIST)
|
|
|
|
error = 0;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs symbolic link create call
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_symlink(ap)
|
|
|
|
struct vop_symlink_args /* {
|
|
|
|
struct vnode *a_dvp;
|
|
|
|
struct vnode **a_vpp;
|
|
|
|
struct componentname *a_cnp;
|
|
|
|
struct vattr *a_vap;
|
|
|
|
char *a_target;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *dvp = ap->a_dvp;
|
|
|
|
register struct vattr *vap = ap->a_vap;
|
|
|
|
register struct componentname *cnp = ap->a_cnp;
|
|
|
|
register struct nfsv2_sattr *sp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
int slen, error = 0, wccflag = NFSV3_WCCRATTR, gotvp;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct vnode *newvp = (struct vnode *)0;
|
|
|
|
int v3 = NFS_ISV3(dvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
nfsstats.rpccnt[NFSPROC_SYMLINK]++;
|
|
|
|
slen = strlen(ap->a_target);
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(dvp, NFSPROC_SYMLINK, NFSX_FH(v3) + 2*NFSX_UNSIGNED +
|
|
|
|
nfsm_rndup(cnp->cn_namelen) + nfsm_rndup(slen) + NFSX_SATTR(v3));
|
|
|
|
nfsm_fhtom(dvp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
1998-12-25 10:34:27 +00:00
|
|
|
nfsm_v3attrbuild(vap, FALSE);
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strtom(ap->a_target, slen, NFS_MAXPATHLEN);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!v3) {
|
|
|
|
nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
|
|
|
|
sp->sa_mode = vtonfsv2_mode(VLNK, vap->va_mode);
|
1998-05-31 19:00:19 +00:00
|
|
|
sp->sa_uid = nfs_xdrneg1;
|
|
|
|
sp->sa_gid = nfs_xdrneg1;
|
|
|
|
sp->sa_size = nfs_xdrneg1;
|
1995-06-27 11:07:30 +00:00
|
|
|
txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
|
|
|
|
txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
nfsm_request(dvp, NFSPROC_SYMLINK, cnp->cn_proc, cnp->cn_cred);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
|
|
|
if (!error)
|
|
|
|
nfsm_mtofh(dvp, newvp, v3, gotvp);
|
|
|
|
nfsm_wcc_data(dvp, wccflag);
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (newvp)
|
|
|
|
vput(newvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
VTONFS(dvp)->n_flag |= NMODIFIED;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!wccflag)
|
|
|
|
VTONFS(dvp)->n_attrstamp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
|
|
|
|
*/
|
|
|
|
if (error == EEXIST)
|
|
|
|
error = 0;
|
1999-02-13 09:47:30 +00:00
|
|
|
/*
|
|
|
|
* cnp's buffer expected to be freed if SAVESTART not set or
|
|
|
|
* if an error was returned.
|
|
|
|
*/
|
|
|
|
if (error || (cnp->cn_flags & SAVESTART) == 0)
|
|
|
|
zfree(namei_zone, cnp->cn_pnbuf);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs make dir call
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_mkdir(ap)
|
|
|
|
struct vop_mkdir_args /* {
|
|
|
|
struct vnode *a_dvp;
|
|
|
|
struct vnode **a_vpp;
|
|
|
|
struct componentname *a_cnp;
|
|
|
|
struct vattr *a_vap;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *dvp = ap->a_dvp;
|
|
|
|
register struct vattr *vap = ap->a_vap;
|
|
|
|
register struct componentname *cnp = ap->a_cnp;
|
|
|
|
register struct nfsv2_sattr *sp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1994-05-24 10:09:53 +00:00
|
|
|
register int len;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct nfsnode *np = (struct nfsnode *)0;
|
|
|
|
struct vnode *newvp = (struct vnode *)0;
|
1994-05-24 10:09:53 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
1995-10-29 15:33:36 +00:00
|
|
|
int error = 0, wccflag = NFSV3_WCCRATTR;
|
|
|
|
int gotvp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
|
|
|
struct vattr vattr;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(dvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1999-01-27 22:45:13 +00:00
|
|
|
if ((error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) != 0) {
|
1994-05-24 10:09:53 +00:00
|
|
|
VOP_ABORTOP(dvp, cnp);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
len = cnp->cn_namelen;
|
|
|
|
nfsstats.rpccnt[NFSPROC_MKDIR]++;
|
|
|
|
nfsm_reqhead(dvp, NFSPROC_MKDIR,
|
1995-06-27 11:07:30 +00:00
|
|
|
NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len) + NFSX_SATTR(v3));
|
|
|
|
nfsm_fhtom(dvp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
1998-12-25 10:34:27 +00:00
|
|
|
nfsm_v3attrbuild(vap, FALSE);
|
1994-05-24 10:09:53 +00:00
|
|
|
} else {
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
|
|
|
|
sp->sa_mode = vtonfsv2_mode(VDIR, vap->va_mode);
|
1998-05-31 19:00:19 +00:00
|
|
|
sp->sa_uid = nfs_xdrneg1;
|
|
|
|
sp->sa_gid = nfs_xdrneg1;
|
|
|
|
sp->sa_size = nfs_xdrneg1;
|
1995-06-27 11:07:30 +00:00
|
|
|
txdr_nfsv2time(&vap->va_atime, &sp->sa_atime);
|
|
|
|
txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
nfsm_request(dvp, NFSPROC_MKDIR, cnp->cn_proc, cnp->cn_cred);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!error)
|
|
|
|
nfsm_mtofh(dvp, newvp, v3, gotvp);
|
|
|
|
if (v3)
|
|
|
|
nfsm_wcc_data(dvp, wccflag);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
|
|
|
VTONFS(dvp)->n_flag |= NMODIFIED;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!wccflag)
|
|
|
|
VTONFS(dvp)->n_attrstamp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Kludge: Map EEXIST => 0 assuming that you have a reply to a retry
|
|
|
|
* if we can succeed in looking up the directory.
|
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
if (error == EEXIST || (!error && !gotvp)) {
|
|
|
|
if (newvp) {
|
|
|
|
vrele(newvp);
|
|
|
|
newvp = (struct vnode *)0;
|
|
|
|
}
|
|
|
|
error = nfs_lookitup(dvp, cnp->cn_nameptr, len, cnp->cn_cred,
|
|
|
|
cnp->cn_proc, &np);
|
|
|
|
if (!error) {
|
|
|
|
newvp = NFSTOV(np);
|
|
|
|
if (newvp->v_type != VDIR)
|
|
|
|
error = EEXIST;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
if (error) {
|
|
|
|
if (newvp)
|
|
|
|
vrele(newvp);
|
|
|
|
} else
|
|
|
|
*ap->a_vpp = newvp;
|
1999-02-13 09:47:30 +00:00
|
|
|
if (error || (cnp->cn_flags & SAVESTART) == 0)
|
|
|
|
zfree(namei_zone, cnp->cn_pnbuf);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs remove directory call
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_rmdir(ap)
|
|
|
|
struct vop_rmdir_args /* {
|
|
|
|
struct vnode *a_dvp;
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct componentname *a_cnp;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct vnode *dvp = ap->a_dvp;
|
|
|
|
register struct componentname *cnp = ap->a_cnp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
int error = 0, wccflag = NFSV3_WCCRATTR;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(dvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-05-31 18:30:42 +00:00
|
|
|
if (dvp == vp)
|
|
|
|
return (EINVAL);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsstats.rpccnt[NFSPROC_RMDIR]++;
|
|
|
|
nfsm_reqhead(dvp, NFSPROC_RMDIR,
|
1995-06-27 11:07:30 +00:00
|
|
|
NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(cnp->cn_namelen));
|
|
|
|
nfsm_fhtom(dvp, v3);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN);
|
|
|
|
nfsm_request(dvp, NFSPROC_RMDIR, cnp->cn_proc, cnp->cn_cred);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3)
|
|
|
|
nfsm_wcc_data(dvp, wccflag);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_reqdone;
|
1997-09-21 04:24:27 +00:00
|
|
|
zfree(namei_zone, cnp->cn_pnbuf);
|
1994-05-24 10:09:53 +00:00
|
|
|
VTONFS(dvp)->n_flag |= NMODIFIED;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!wccflag)
|
|
|
|
VTONFS(dvp)->n_attrstamp = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
cache_purge(dvp);
|
|
|
|
cache_purge(vp);
|
|
|
|
/*
|
|
|
|
* Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
|
|
|
|
*/
|
|
|
|
if (error == ENOENT)
|
|
|
|
error = 0;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nfs readdir call
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_readdir(ap)
|
|
|
|
struct vop_readdir_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct uio *a_uio;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
register struct uio *uio = ap->a_uio;
|
|
|
|
int tresid, error;
|
|
|
|
struct vattr vattr;
|
|
|
|
|
|
|
|
if (vp->v_type != VDIR)
|
|
|
|
return (EPERM);
|
|
|
|
/*
|
|
|
|
* First, check for hit on the EOF offset cache
|
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
if (np->n_direofoffset > 0 && uio->uio_offset >= np->n_direofoffset &&
|
1994-05-24 10:09:53 +00:00
|
|
|
(np->n_flag & NMODIFIED) == 0) {
|
|
|
|
if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) {
|
1995-06-27 11:07:30 +00:00
|
|
|
if (NQNFS_CKCACHABLE(vp, ND_READ)) {
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsstats.direofcache_hits++;
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
} else if (VOP_GETATTR(vp, &vattr, ap->a_cred, uio->uio_procp) == 0 &&
|
1996-09-19 18:21:32 +00:00
|
|
|
np->n_mtime == vattr.va_mtime.tv_sec) {
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsstats.direofcache_hits++;
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Call nfs_bioread() to do the real work.
|
|
|
|
*/
|
|
|
|
tresid = uio->uio_resid;
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
error = nfs_bioread(vp, uio, 0, ap->a_cred);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
if (!error && uio->uio_resid == tresid)
|
|
|
|
nfsstats.direofcache_misses++;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Readdir rpc call.
|
|
|
|
* Called from below the buffer cache by nfs_doio().
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
nfs_readdirrpc(vp, uiop, cred)
|
1995-06-27 11:07:30 +00:00
|
|
|
struct vnode *vp;
|
|
|
|
register struct uio *uiop;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct ucred *cred;
|
1995-06-27 11:07:30 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1995-06-27 11:07:30 +00:00
|
|
|
register int len, left;
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
register struct dirent *dp = NULL;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
register nfsuint64 *cookiep;
|
1994-05-24 10:09:53 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsuint64 cookie;
|
|
|
|
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
|
|
|
|
struct nfsnode *dnp = VTONFS(vp);
|
1995-10-29 15:33:36 +00:00
|
|
|
u_quad_t fileno;
|
|
|
|
int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
|
|
|
|
int attrflag;
|
1995-06-27 11:07:30 +00:00
|
|
|
int v3 = NFS_ISV3(vp);
|
|
|
|
|
|
|
|
#ifndef DIAGNOSTIC
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
if (uiop->uio_iovcnt != 1 || (uiop->uio_offset & (DIRBLKSIZ - 1)) ||
|
|
|
|
(uiop->uio_resid & (DIRBLKSIZ - 1)))
|
1995-06-27 11:07:30 +00:00
|
|
|
panic("nfs readdirrpc bad uio");
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/*
|
1996-10-21 10:07:52 +00:00
|
|
|
* If there is no cookie, assume directory was stale.
|
1995-06-27 11:07:30 +00:00
|
|
|
*/
|
|
|
|
cookiep = nfs_getcookie(dnp, uiop->uio_offset, 0);
|
|
|
|
if (cookiep)
|
|
|
|
cookie = *cookiep;
|
|
|
|
else
|
1996-10-21 10:07:52 +00:00
|
|
|
return (NFSERR_BAD_COOKIE);
|
1995-06-27 11:07:30 +00:00
|
|
|
/*
|
|
|
|
* Loop around doing readdir rpc's of size nm_readdirsize
|
|
|
|
* truncated to a multiple of DIRBLKSIZ.
|
1994-05-24 10:09:53 +00:00
|
|
|
* The stopping criteria is EOF or buffer full.
|
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
while (more_dirs && bigenough) {
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsstats.rpccnt[NFSPROC_READDIR]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(vp, NFSPROC_READDIR, NFSX_FH(v3) +
|
|
|
|
NFSX_READDIR(v3));
|
|
|
|
nfsm_fhtom(vp, v3);
|
|
|
|
if (v3) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
*tl++ = cookie.nfsuquad[0];
|
|
|
|
*tl++ = cookie.nfsuquad[1];
|
|
|
|
*tl++ = dnp->n_cookieverf.nfsuquad[0];
|
|
|
|
*tl++ = dnp->n_cookieverf.nfsuquad[1];
|
|
|
|
} else {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
*tl++ = cookie.nfsuquad[0];
|
|
|
|
}
|
|
|
|
*tl = txdr_unsigned(nmp->nm_readdirsize);
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsm_request(vp, NFSPROC_READDIR, uiop->uio_procp, cred);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (v3) {
|
|
|
|
nfsm_postop_attr(vp, attrflag);
|
|
|
|
if (!error) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *,
|
|
|
|
2 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
dnp->n_cookieverf.nfsuquad[0] = *tl++;
|
|
|
|
dnp->n_cookieverf.nfsuquad[1] = *tl;
|
|
|
|
} else {
|
|
|
|
m_freem(mrep);
|
|
|
|
goto nfsmout;
|
|
|
|
}
|
|
|
|
}
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
|
1994-05-24 10:09:53 +00:00
|
|
|
more_dirs = fxdr_unsigned(int, *tl);
|
1997-02-10 02:22:35 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/* loop thru the dir entries, doctoring them to 4bsd form */
|
1995-06-27 11:07:30 +00:00
|
|
|
while (more_dirs && bigenough) {
|
|
|
|
if (v3) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *,
|
|
|
|
3 * NFSX_UNSIGNED);
|
1999-06-05 05:35:03 +00:00
|
|
|
fileno = fxdr_hyper(tl);
|
1995-06-27 11:07:30 +00:00
|
|
|
len = fxdr_unsigned(int, *(tl + 2));
|
|
|
|
} else {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *,
|
|
|
|
2 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
fileno = fxdr_unsigned(u_quad_t, *tl++);
|
|
|
|
len = fxdr_unsigned(int, *tl);
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
if (len <= 0 || len > NFS_MAXNAMLEN) {
|
|
|
|
error = EBADRPC;
|
|
|
|
m_freem(mrep);
|
|
|
|
goto nfsmout;
|
|
|
|
}
|
|
|
|
tlen = nfsm_rndup(len);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (tlen == len)
|
|
|
|
tlen += 4; /* To ensure null termination */
|
|
|
|
left = DIRBLKSIZ - blksiz;
|
|
|
|
if ((tlen + DIRHDSIZ) > left) {
|
|
|
|
dp->d_reclen += left;
|
|
|
|
uiop->uio_iov->iov_base += left;
|
|
|
|
uiop->uio_iov->iov_len -= left;
|
|
|
|
uiop->uio_offset += left;
|
|
|
|
uiop->uio_resid -= left;
|
|
|
|
blksiz = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
if ((tlen + DIRHDSIZ) > uiop->uio_resid)
|
|
|
|
bigenough = 0;
|
|
|
|
if (bigenough) {
|
|
|
|
dp = (struct dirent *)uiop->uio_iov->iov_base;
|
|
|
|
dp->d_fileno = (int)fileno;
|
|
|
|
dp->d_namlen = len;
|
|
|
|
dp->d_reclen = tlen + DIRHDSIZ;
|
|
|
|
dp->d_type = DT_UNKNOWN;
|
|
|
|
blksiz += dp->d_reclen;
|
|
|
|
if (blksiz == DIRBLKSIZ)
|
|
|
|
blksiz = 0;
|
|
|
|
uiop->uio_offset += DIRHDSIZ;
|
|
|
|
uiop->uio_resid -= DIRHDSIZ;
|
|
|
|
uiop->uio_iov->iov_base += DIRHDSIZ;
|
|
|
|
uiop->uio_iov->iov_len -= DIRHDSIZ;
|
|
|
|
nfsm_mtouio(uiop, len);
|
|
|
|
cp = uiop->uio_iov->iov_base;
|
|
|
|
tlen -= len;
|
|
|
|
*cp = '\0'; /* null terminate */
|
|
|
|
uiop->uio_iov->iov_base += tlen;
|
|
|
|
uiop->uio_iov->iov_len -= tlen;
|
|
|
|
uiop->uio_offset += tlen;
|
|
|
|
uiop->uio_resid -= tlen;
|
|
|
|
} else
|
|
|
|
nfsm_adv(nfsm_rndup(len));
|
|
|
|
if (v3) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *,
|
|
|
|
3 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
} else {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *,
|
|
|
|
2 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
|
|
|
if (bigenough) {
|
|
|
|
cookie.nfsuquad[0] = *tl++;
|
|
|
|
if (v3)
|
|
|
|
cookie.nfsuquad[1] = *tl++;
|
|
|
|
} else if (v3)
|
|
|
|
tl += 2;
|
|
|
|
else
|
|
|
|
tl++;
|
1994-05-24 10:09:53 +00:00
|
|
|
more_dirs = fxdr_unsigned(int, *tl);
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* If at end of rpc data, get the eof boolean
|
|
|
|
*/
|
|
|
|
if (!more_dirs) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
|
1994-05-24 10:09:53 +00:00
|
|
|
more_dirs = (fxdr_unsigned(int, *tl) == 0);
|
|
|
|
}
|
|
|
|
m_freem(mrep);
|
|
|
|
}
|
|
|
|
/*
|
1995-06-27 11:07:30 +00:00
|
|
|
* Fill last record, iff any, out to a multiple of DIRBLKSIZ
|
1994-05-24 10:09:53 +00:00
|
|
|
* by increasing d_reclen for the last record.
|
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
if (blksiz > 0) {
|
|
|
|
left = DIRBLKSIZ - blksiz;
|
|
|
|
dp->d_reclen += left;
|
|
|
|
uiop->uio_iov->iov_base += left;
|
|
|
|
uiop->uio_iov->iov_len -= left;
|
|
|
|
uiop->uio_offset += left;
|
|
|
|
uiop->uio_resid -= left;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We are now either at the end of the directory or have filled the
|
|
|
|
* block.
|
|
|
|
*/
|
|
|
|
if (bigenough)
|
|
|
|
dnp->n_direofoffset = uiop->uio_offset;
|
|
|
|
else {
|
|
|
|
if (uiop->uio_resid > 0)
|
|
|
|
printf("EEK! readdirrpc resid > 0\n");
|
|
|
|
cookiep = nfs_getcookie(dnp, uiop->uio_offset, 1);
|
|
|
|
*cookiep = cookie;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
nfsmout:
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
1995-06-27 11:07:30 +00:00
|
|
|
* NFS V3 readdir plus RPC. Used in place of nfs_readdirrpc().
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_readdirplusrpc(vp, uiop, cred)
|
1994-05-24 10:09:53 +00:00
|
|
|
struct vnode *vp;
|
|
|
|
register struct uio *uiop;
|
|
|
|
struct ucred *cred;
|
|
|
|
{
|
1995-06-27 11:07:30 +00:00
|
|
|
register int len, left;
|
|
|
|
register struct dirent *dp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
register struct vnode *newvp;
|
|
|
|
register nfsuint64 *cookiep;
|
|
|
|
caddr_t bpos, dpos, cp2, dpossav1, dpossav2;
|
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2, *mdsav1, *mdsav2;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct nameidata nami, *ndp = &nami;
|
|
|
|
struct componentname *cnp = &ndp->ni_cnd;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsuint64 cookie;
|
|
|
|
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
|
|
|
|
struct nfsnode *dnp = VTONFS(vp), *np;
|
|
|
|
nfsfh_t *fhp;
|
1995-10-29 15:33:36 +00:00
|
|
|
u_quad_t fileno;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error = 0, tlen, more_dirs = 1, blksiz = 0, doit, bigenough = 1, i;
|
1995-10-29 15:33:36 +00:00
|
|
|
int attrflag, fhsize;
|
1995-06-27 11:07:30 +00:00
|
|
|
|
|
|
|
#ifndef nolint
|
|
|
|
dp = (struct dirent *)0;
|
|
|
|
#endif
|
|
|
|
#ifndef DIAGNOSTIC
|
|
|
|
if (uiop->uio_iovcnt != 1 || (uiop->uio_offset & (DIRBLKSIZ - 1)) ||
|
|
|
|
(uiop->uio_resid & (DIRBLKSIZ - 1)))
|
|
|
|
panic("nfs readdirplusrpc bad uio");
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
ndp->ni_dvp = vp;
|
|
|
|
newvp = NULLVP;
|
1995-06-27 11:07:30 +00:00
|
|
|
|
|
|
|
/*
|
1996-10-21 10:07:52 +00:00
|
|
|
* If there is no cookie, assume directory was stale.
|
1995-06-27 11:07:30 +00:00
|
|
|
*/
|
|
|
|
cookiep = nfs_getcookie(dnp, uiop->uio_offset, 0);
|
|
|
|
if (cookiep)
|
|
|
|
cookie = *cookiep;
|
|
|
|
else
|
1996-10-21 10:07:52 +00:00
|
|
|
return (NFSERR_BAD_COOKIE);
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
1995-06-27 11:07:30 +00:00
|
|
|
* Loop around doing readdir rpc's of size nm_readdirsize
|
|
|
|
* truncated to a multiple of DIRBLKSIZ.
|
1994-05-24 10:09:53 +00:00
|
|
|
* The stopping criteria is EOF or buffer full.
|
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
while (more_dirs && bigenough) {
|
|
|
|
nfsstats.rpccnt[NFSPROC_READDIRPLUS]++;
|
|
|
|
nfsm_reqhead(vp, NFSPROC_READDIRPLUS,
|
|
|
|
NFSX_FH(1) + 6 * NFSX_UNSIGNED);
|
|
|
|
nfsm_fhtom(vp, 1);
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
*tl++ = cookie.nfsuquad[0];
|
|
|
|
*tl++ = cookie.nfsuquad[1];
|
|
|
|
*tl++ = dnp->n_cookieverf.nfsuquad[0];
|
|
|
|
*tl++ = dnp->n_cookieverf.nfsuquad[1];
|
|
|
|
*tl++ = txdr_unsigned(nmp->nm_readdirsize);
|
|
|
|
*tl = txdr_unsigned(nmp->nm_rsize);
|
|
|
|
nfsm_request(vp, NFSPROC_READDIRPLUS, uiop->uio_procp, cred);
|
|
|
|
nfsm_postop_attr(vp, attrflag);
|
|
|
|
if (error) {
|
|
|
|
m_freem(mrep);
|
|
|
|
goto nfsmout;
|
|
|
|
}
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
dnp->n_cookieverf.nfsuquad[0] = *tl++;
|
|
|
|
dnp->n_cookieverf.nfsuquad[1] = *tl++;
|
1994-05-24 10:09:53 +00:00
|
|
|
more_dirs = fxdr_unsigned(int, *tl);
|
1995-05-30 08:16:23 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/* loop thru the dir entries, doctoring them to 4bsd form */
|
|
|
|
while (more_dirs && bigenough) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
|
1999-06-05 05:35:03 +00:00
|
|
|
fileno = fxdr_hyper(tl);
|
1995-06-27 11:07:30 +00:00
|
|
|
len = fxdr_unsigned(int, *(tl + 2));
|
1994-05-24 10:09:53 +00:00
|
|
|
if (len <= 0 || len > NFS_MAXNAMLEN) {
|
|
|
|
error = EBADRPC;
|
|
|
|
m_freem(mrep);
|
|
|
|
goto nfsmout;
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
tlen = nfsm_rndup(len);
|
|
|
|
if (tlen == len)
|
|
|
|
tlen += 4; /* To ensure null termination*/
|
|
|
|
left = DIRBLKSIZ - blksiz;
|
|
|
|
if ((tlen + DIRHDSIZ) > left) {
|
|
|
|
dp->d_reclen += left;
|
|
|
|
uiop->uio_iov->iov_base += left;
|
|
|
|
uiop->uio_iov->iov_len -= left;
|
|
|
|
uiop->uio_offset += left;
|
|
|
|
uiop->uio_resid -= left;
|
|
|
|
blksiz = 0;
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
if ((tlen + DIRHDSIZ) > uiop->uio_resid)
|
|
|
|
bigenough = 0;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (bigenough) {
|
1994-05-24 10:09:53 +00:00
|
|
|
dp = (struct dirent *)uiop->uio_iov->iov_base;
|
1995-06-27 11:07:30 +00:00
|
|
|
dp->d_fileno = (int)fileno;
|
1994-05-24 10:09:53 +00:00
|
|
|
dp->d_namlen = len;
|
|
|
|
dp->d_reclen = tlen + DIRHDSIZ;
|
1995-06-27 11:07:30 +00:00
|
|
|
dp->d_type = DT_UNKNOWN;
|
|
|
|
blksiz += dp->d_reclen;
|
|
|
|
if (blksiz == DIRBLKSIZ)
|
|
|
|
blksiz = 0;
|
|
|
|
uiop->uio_offset += DIRHDSIZ;
|
1994-05-24 10:09:53 +00:00
|
|
|
uiop->uio_resid -= DIRHDSIZ;
|
|
|
|
uiop->uio_iov->iov_base += DIRHDSIZ;
|
|
|
|
uiop->uio_iov->iov_len -= DIRHDSIZ;
|
|
|
|
cnp->cn_nameptr = uiop->uio_iov->iov_base;
|
|
|
|
cnp->cn_namelen = len;
|
|
|
|
nfsm_mtouio(uiop, len);
|
|
|
|
cp = uiop->uio_iov->iov_base;
|
|
|
|
tlen -= len;
|
1995-06-27 11:07:30 +00:00
|
|
|
*cp = '\0';
|
1994-05-24 10:09:53 +00:00
|
|
|
uiop->uio_iov->iov_base += tlen;
|
|
|
|
uiop->uio_iov->iov_len -= tlen;
|
1995-06-27 11:07:30 +00:00
|
|
|
uiop->uio_offset += tlen;
|
1994-05-24 10:09:53 +00:00
|
|
|
uiop->uio_resid -= tlen;
|
1995-06-27 11:07:30 +00:00
|
|
|
} else
|
|
|
|
nfsm_adv(nfsm_rndup(len));
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (bigenough) {
|
|
|
|
cookie.nfsuquad[0] = *tl++;
|
|
|
|
cookie.nfsuquad[1] = *tl++;
|
|
|
|
} else
|
|
|
|
tl += 2;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Since the attributes are before the file handle
|
|
|
|
* (sigh), we must skip over the attributes and then
|
|
|
|
* come back and get them.
|
|
|
|
*/
|
|
|
|
attrflag = fxdr_unsigned(int, *tl);
|
|
|
|
if (attrflag) {
|
|
|
|
dpossav1 = dpos;
|
|
|
|
mdsav1 = md;
|
|
|
|
nfsm_adv(NFSX_V3FATTR);
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
doit = fxdr_unsigned(int, *tl);
|
|
|
|
if (doit) {
|
|
|
|
nfsm_getfh(fhp, fhsize, 1);
|
|
|
|
if (NFS_CMPFH(dnp, fhp, fhsize)) {
|
|
|
|
VREF(vp);
|
|
|
|
newvp = vp;
|
|
|
|
np = dnp;
|
|
|
|
} else {
|
1998-05-31 17:27:58 +00:00
|
|
|
error = nfs_nget(vp->v_mount, fhp,
|
|
|
|
fhsize, &np);
|
|
|
|
if (error)
|
1995-06-27 11:07:30 +00:00
|
|
|
doit = 0;
|
|
|
|
else
|
|
|
|
newvp = NFSTOV(np);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (doit) {
|
|
|
|
dpossav2 = dpos;
|
|
|
|
dpos = dpossav1;
|
|
|
|
mdsav2 = md;
|
|
|
|
md = mdsav1;
|
|
|
|
nfsm_loadattr(newvp, (struct vattr *)0);
|
|
|
|
dpos = dpossav2;
|
|
|
|
md = mdsav2;
|
|
|
|
dp->d_type =
|
|
|
|
IFTODT(VTTOIF(np->n_vattr.va_type));
|
|
|
|
ndp->ni_vp = newvp;
|
1994-05-24 10:09:53 +00:00
|
|
|
cnp->cn_hash = 0;
|
1995-06-27 11:07:30 +00:00
|
|
|
for (cp = cnp->cn_nameptr, i = 1; i <= len;
|
|
|
|
i++, cp++)
|
|
|
|
cnp->cn_hash += (unsigned char)*cp * i;
|
1. Add a {pointer, v_id} pair to the vnode to store the reference to the
".." vnode. This is cheaper storagewise than keeping it in the
namecache, and it makes more sense since it's a 1:1 mapping.
2. Also handle the case of "." more intelligently rather than stuff
the namecache with pointless entries.
3. Add two lists to the vnode and hang namecache entries which go from
or to this vnode. When cleaning a vnode, delete all namecache
entries it invalidates.
4. Never reuse namecache enties, malloc new ones when we need it, free
old ones when they die. No longer a hard limit on how many we can
have.
5. Remove the upper limit on namelength of namecache entries.
6. Make a global list for negative namecache entries, limit their number
to a sysctl'able (debug.ncnegfactor) fraction of the total namecache.
Currently the default fraction is 1/16th. (Suggestions for better
default wanted!)
7. Assign v_id correctly in the face of 32bit rollover.
8. Remove the LRU list for namecache entries, not needed. Remove the
#ifdef NCH_STATISTICS stuff, it's not needed either.
9. Use the vnode freelist as a true LRU list, also for namecache accesses.
10. Reuse vnodes more aggresively but also more selectively, if we can't
reuse, malloc a new one. There is no longer a hard limit on their
number, they grow to the point where we don't reuse potentially
usable vnodes. A vnode will not get recycled if still has pages in
core or if it is the source of namecache entries (Yes, this does
indeed work :-) "." and ".." are not namecache entries any longer...)
11. Do not overload the v_id field in namecache entries with whiteout
information, use a char sized flags field instead, so we can get
rid of the vpid and v_id fields from the namecache struct. Since
we're linked to the vnodes and purged when they're cleaned, we don't
have to check the v_id any more.
12. NFS knew about the limitation on name length in the namecache, it
shouldn't and doesn't now.
Bugs:
The namecache statistics no longer includes the hits for ".."
and "." hits.
Performance impact:
Generally in the +/- 0.5% for "normal" workstations, but
I hope this will allow the system to be selftuning over a
bigger range of "special" applications. The case where
RAM is available but unused for cache because we don't have
any vnodes should be gone.
Future work:
Straighten out the namecache statistics.
"desiredvnodes" is still used to (bogusly ?) size hash
tables in the filesystems.
I have still to find a way to safely free unused vnodes
back so their number can shrink when not needed.
There is a few uses of the v_id field left in the filesystems,
scheduled for demolition at a later time.
Maybe a one slot cache for unused namecache entries should
be implemented to decrease the malloc/free frequency.
1997-05-04 09:17:38 +00:00
|
|
|
cache_enter(ndp->ni_dvp, ndp->ni_vp, cnp);
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
} else {
|
1995-06-27 11:07:30 +00:00
|
|
|
/* Just skip over the file handle */
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
|
1995-06-27 11:07:30 +00:00
|
|
|
i = fxdr_unsigned(int, *tl);
|
|
|
|
nfsm_adv(nfsm_rndup(i));
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
if (newvp != NULLVP) {
|
1995-06-27 11:07:30 +00:00
|
|
|
vrele(newvp);
|
|
|
|
newvp = NULLVP;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
|
1994-05-24 10:09:53 +00:00
|
|
|
more_dirs = fxdr_unsigned(int, *tl);
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* If at end of rpc data, get the eof boolean
|
|
|
|
*/
|
|
|
|
if (!more_dirs) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
|
1994-05-24 10:09:53 +00:00
|
|
|
more_dirs = (fxdr_unsigned(int, *tl) == 0);
|
|
|
|
}
|
|
|
|
m_freem(mrep);
|
|
|
|
}
|
|
|
|
/*
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
* Fill last record, iff any, out to a multiple of DIRBLKSIZ
|
1994-05-24 10:09:53 +00:00
|
|
|
* by increasing d_reclen for the last record.
|
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
if (blksiz > 0) {
|
|
|
|
left = DIRBLKSIZ - blksiz;
|
|
|
|
dp->d_reclen += left;
|
|
|
|
uiop->uio_iov->iov_base += left;
|
|
|
|
uiop->uio_iov->iov_len -= left;
|
|
|
|
uiop->uio_offset += left;
|
|
|
|
uiop->uio_resid -= left;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We are now either at the end of the directory or have filled the
|
|
|
|
* block.
|
|
|
|
*/
|
|
|
|
if (bigenough)
|
|
|
|
dnp->n_direofoffset = uiop->uio_offset;
|
|
|
|
else {
|
|
|
|
if (uiop->uio_resid > 0)
|
|
|
|
printf("EEK! readdirplusrpc resid > 0\n");
|
|
|
|
cookiep = nfs_getcookie(dnp, uiop->uio_offset, 1);
|
|
|
|
*cookiep = cookie;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
nfsmout:
|
1995-06-27 11:07:30 +00:00
|
|
|
if (newvp != NULLVP) {
|
|
|
|
if (newvp == vp)
|
|
|
|
vrele(newvp);
|
|
|
|
else
|
|
|
|
vput(newvp);
|
|
|
|
newvp = NULLVP;
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Silly rename. To make the NFS filesystem that is stateless look a little
|
|
|
|
* more like the "ufs" a remove of an active vnode is translated to a rename
|
|
|
|
* to a funny looking filename that is removed by nfs_inactive on the
|
|
|
|
* nfsnode. There is the potential for another process on a different client
|
|
|
|
* to create the same funny name between the nfs_lookitup() fails and the
|
|
|
|
* nfs_rename() completes, but...
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_sillyrename(dvp, vp, cnp)
|
|
|
|
struct vnode *dvp, *vp;
|
|
|
|
struct componentname *cnp;
|
|
|
|
{
|
|
|
|
register struct sillyrename *sp;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct nfsnode *np;
|
1994-05-24 10:09:53 +00:00
|
|
|
int error;
|
|
|
|
short pid;
|
|
|
|
|
|
|
|
cache_purge(dvp);
|
|
|
|
np = VTONFS(vp);
|
1995-06-27 11:07:30 +00:00
|
|
|
#ifndef DIAGNOSTIC
|
|
|
|
if (vp->v_type == VDIR)
|
|
|
|
panic("nfs: sillyrename dir");
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
MALLOC(sp, struct sillyrename *, sizeof (struct sillyrename),
|
|
|
|
M_NFSREQ, M_WAITOK);
|
|
|
|
sp->s_cred = crdup(cnp->cn_cred);
|
|
|
|
sp->s_dvp = dvp;
|
|
|
|
VREF(dvp);
|
|
|
|
|
|
|
|
/* Fudge together a funny name */
|
|
|
|
pid = cnp->cn_proc->p_pid;
|
1996-01-25 00:45:37 +00:00
|
|
|
sp->s_namlen = sprintf(sp->s_name, ".nfsA%04x4.4", pid);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
/* Try lookitups until we get one that isn't there */
|
1995-06-27 11:07:30 +00:00
|
|
|
while (nfs_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred,
|
|
|
|
cnp->cn_proc, (struct nfsnode **)0) == 0) {
|
1994-05-24 10:09:53 +00:00
|
|
|
sp->s_name[4]++;
|
|
|
|
if (sp->s_name[4] > 'z') {
|
|
|
|
error = EINVAL;
|
|
|
|
goto bad;
|
|
|
|
}
|
|
|
|
}
|
1998-05-31 17:27:58 +00:00
|
|
|
error = nfs_renameit(dvp, cnp, sp);
|
|
|
|
if (error)
|
1994-05-24 10:09:53 +00:00
|
|
|
goto bad;
|
1995-06-27 11:07:30 +00:00
|
|
|
error = nfs_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred,
|
|
|
|
cnp->cn_proc, &np);
|
1994-05-24 10:09:53 +00:00
|
|
|
np->n_sillyrename = sp;
|
|
|
|
return (0);
|
|
|
|
bad:
|
|
|
|
vrele(sp->s_dvp);
|
|
|
|
crfree(sp->s_cred);
|
|
|
|
free((caddr_t)sp, M_NFSREQ);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
1995-06-27 11:07:30 +00:00
|
|
|
* Look up a file name and optionally either update the file handle or
|
|
|
|
* allocate an nfsnode, depending on the value of npp.
|
|
|
|
* npp == NULL --> just do the lookup
|
|
|
|
* *npp == NULL --> allocate a new nfsnode and make sure attributes are
|
|
|
|
* handled too
|
|
|
|
* *npp != NULL --> update the file handle in the vnode
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_lookitup(dvp, name, len, cred, procp, npp)
|
|
|
|
register struct vnode *dvp;
|
1998-05-31 17:48:07 +00:00
|
|
|
const char *name;
|
1995-06-27 11:07:30 +00:00
|
|
|
int len;
|
|
|
|
struct ucred *cred;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct proc *procp;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct nfsnode **npp;
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
1994-05-24 10:09:53 +00:00
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct vnode *newvp = (struct vnode *)0;
|
|
|
|
struct nfsnode *np, *dnp = VTONFS(dvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
caddr_t bpos, dpos, cp2;
|
1995-06-27 11:07:30 +00:00
|
|
|
int error = 0, fhlen, attrflag;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsfh_t *nfhp;
|
|
|
|
int v3 = NFS_ISV3(dvp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
nfsstats.rpccnt[NFSPROC_LOOKUP]++;
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqhead(dvp, NFSPROC_LOOKUP,
|
|
|
|
NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len));
|
|
|
|
nfsm_fhtom(dvp, v3);
|
|
|
|
nfsm_strtom(name, len, NFS_MAXNAMLEN);
|
|
|
|
nfsm_request(dvp, NFSPROC_LOOKUP, procp, cred);
|
|
|
|
if (npp && !error) {
|
|
|
|
nfsm_getfh(nfhp, fhlen, v3);
|
|
|
|
if (*npp) {
|
|
|
|
np = *npp;
|
|
|
|
if (np->n_fhsize > NFS_SMALLFH && fhlen <= NFS_SMALLFH) {
|
|
|
|
free((caddr_t)np->n_fhp, M_NFSBIGFH);
|
|
|
|
np->n_fhp = &np->n_fh;
|
|
|
|
} else if (np->n_fhsize <= NFS_SMALLFH && fhlen>NFS_SMALLFH)
|
|
|
|
np->n_fhp =(nfsfh_t *)malloc(fhlen,M_NFSBIGFH,M_WAITOK);
|
|
|
|
bcopy((caddr_t)nfhp, (caddr_t)np->n_fhp, fhlen);
|
|
|
|
np->n_fhsize = fhlen;
|
|
|
|
newvp = NFSTOV(np);
|
|
|
|
} else if (NFS_CMPFH(dnp, nfhp, fhlen)) {
|
|
|
|
VREF(dvp);
|
|
|
|
newvp = dvp;
|
|
|
|
} else {
|
|
|
|
error = nfs_nget(dvp->v_mount, nfhp, fhlen, &np);
|
|
|
|
if (error) {
|
|
|
|
m_freem(mrep);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
newvp = NFSTOV(np);
|
|
|
|
}
|
|
|
|
if (v3) {
|
|
|
|
nfsm_postop_attr(newvp, attrflag);
|
|
|
|
if (!attrflag && *npp == NULL) {
|
|
|
|
m_freem(mrep);
|
|
|
|
if (newvp == dvp)
|
|
|
|
vrele(newvp);
|
|
|
|
else
|
|
|
|
vput(newvp);
|
|
|
|
return (ENOENT);
|
|
|
|
}
|
|
|
|
} else
|
|
|
|
nfsm_loadattr(newvp, (struct vattr *)0);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
nfsm_reqdone;
|
|
|
|
if (npp && *npp == NULL) {
|
|
|
|
if (error) {
|
1999-05-06 18:13:11 +00:00
|
|
|
if (newvp) {
|
1995-06-27 11:07:30 +00:00
|
|
|
if (newvp == dvp)
|
|
|
|
vrele(newvp);
|
|
|
|
else
|
|
|
|
vput(newvp);
|
1999-05-06 18:13:11 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
} else
|
|
|
|
*npp = np;
|
|
|
|
}
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Nfs Version 3 commit rpc
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_commit(vp, offset, cnt, cred, procp)
|
|
|
|
register struct vnode *vp;
|
|
|
|
u_quad_t offset;
|
|
|
|
int cnt;
|
|
|
|
struct ucred *cred;
|
|
|
|
struct proc *procp;
|
|
|
|
{
|
|
|
|
register caddr_t cp;
|
1998-05-31 20:09:01 +00:00
|
|
|
register u_int32_t *tl;
|
|
|
|
register int32_t t1, t2;
|
1995-06-27 11:07:30 +00:00
|
|
|
register struct nfsmount *nmp = VFSTONFS(vp->v_mount);
|
|
|
|
caddr_t bpos, dpos, cp2;
|
|
|
|
int error = 0, wccflag = NFSV3_WCCRATTR;
|
|
|
|
struct mbuf *mreq, *mrep, *md, *mb, *mb2;
|
|
|
|
|
1998-05-19 07:11:27 +00:00
|
|
|
if ((nmp->nm_state & NFSSTA_HASWRITEVERF) == 0)
|
1995-06-27 11:07:30 +00:00
|
|
|
return (0);
|
|
|
|
nfsstats.rpccnt[NFSPROC_COMMIT]++;
|
|
|
|
nfsm_reqhead(vp, NFSPROC_COMMIT, NFSX_FH(1));
|
|
|
|
nfsm_fhtom(vp, 1);
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_build(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
|
1999-06-05 05:35:03 +00:00
|
|
|
txdr_hyper(offset, tl);
|
1995-06-27 11:07:30 +00:00
|
|
|
tl += 2;
|
|
|
|
*tl = txdr_unsigned(cnt);
|
|
|
|
nfsm_request(vp, NFSPROC_COMMIT, procp, cred);
|
|
|
|
nfsm_wcc_data(vp, wccflag);
|
|
|
|
if (!error) {
|
1998-05-31 20:09:01 +00:00
|
|
|
nfsm_dissect(tl, u_int32_t *, NFSX_V3WRITEVERF);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (bcmp((caddr_t)nmp->nm_verf, (caddr_t)tl,
|
|
|
|
NFSX_V3WRITEVERF)) {
|
|
|
|
bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf,
|
|
|
|
NFSX_V3WRITEVERF);
|
|
|
|
error = NFSERR_STALEWRITEVERF;
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
nfsm_reqdone;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Kludge City..
|
|
|
|
* - make nfs_bmap() essentially a no-op that does no translation
|
1995-06-27 11:07:30 +00:00
|
|
|
* - do nfs_strategy() by doing I/O with nfs_readrpc/nfs_writerpc
|
1994-05-24 10:09:53 +00:00
|
|
|
* (Maybe I could use the process's page mapping, but I was concerned that
|
|
|
|
* Kernel Write might not be enabled and also figured copyout() would do
|
|
|
|
* a lot more work than bcopy() and also it currently happens in the
|
|
|
|
* context of the swapper process (2).
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_bmap(ap)
|
|
|
|
struct vop_bmap_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
daddr_t a_bn;
|
|
|
|
struct vnode **a_vpp;
|
|
|
|
daddr_t *a_bnp;
|
|
|
|
int *a_runp;
|
1995-09-04 00:21:16 +00:00
|
|
|
int *a_runb;
|
1994-05-24 10:09:53 +00:00
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
|
|
|
|
if (ap->a_vpp != NULL)
|
|
|
|
*ap->a_vpp = vp;
|
|
|
|
if (ap->a_bnp != NULL)
|
|
|
|
*ap->a_bnp = ap->a_bn * btodb(vp->v_mount->mnt_stat.f_iosize);
|
1995-02-03 06:46:28 +00:00
|
|
|
if (ap->a_runp != NULL)
|
|
|
|
*ap->a_runp = 0;
|
1995-09-04 00:21:16 +00:00
|
|
|
if (ap->a_runb != NULL)
|
|
|
|
*ap->a_runb = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Strategy routine.
|
|
|
|
* For async requests when nfsiod(s) are running, queue the request by
|
|
|
|
* calling nfs_asyncio(), otherwise just all nfs_doio() to do the
|
|
|
|
* request.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_strategy(ap)
|
|
|
|
struct vop_strategy_args *ap;
|
|
|
|
{
|
|
|
|
register struct buf *bp = ap->a_bp;
|
|
|
|
struct ucred *cr;
|
|
|
|
struct proc *p;
|
|
|
|
int error = 0;
|
|
|
|
|
1999-03-12 02:24:58 +00:00
|
|
|
KASSERT(!(bp->b_flags & B_DONE), ("nfs_strategy: buffer %p unexpectedly marked B_DONE", bp));
|
1999-06-28 12:34:40 +00:00
|
|
|
KASSERT(BUF_REFCNT(bp) > 0, ("nfs_strategy: buffer %p not locked", bp));
|
1999-03-12 02:24:58 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
if (bp->b_flags & B_PHYS)
|
|
|
|
panic("nfs physio");
|
1999-01-21 08:29:12 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
if (bp->b_flags & B_ASYNC)
|
|
|
|
p = (struct proc *)0;
|
|
|
|
else
|
|
|
|
p = curproc; /* XXX */
|
1999-01-21 08:29:12 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
if (bp->b_flags & B_READ)
|
|
|
|
cr = bp->b_rcred;
|
|
|
|
else
|
|
|
|
cr = bp->b_wcred;
|
1999-01-21 08:29:12 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* If the op is asynchronous and an i/o daemon is waiting
|
|
|
|
* queue the request, wake it up and wait for completion
|
|
|
|
* otherwise just do it ourselves.
|
|
|
|
*/
|
|
|
|
if ((bp->b_flags & B_ASYNC) == 0 ||
|
1999-05-06 20:00:34 +00:00
|
|
|
nfs_asyncio(bp, NOCRED, p))
|
1994-05-24 10:09:53 +00:00
|
|
|
error = nfs_doio(bp, cr, p);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Mmap a file
|
|
|
|
*
|
|
|
|
* NB Currently unsupported.
|
|
|
|
*/
|
|
|
|
/* ARGSUSED */
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_mmap(ap)
|
|
|
|
struct vop_mmap_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
int a_fflags;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
1995-06-27 11:07:30 +00:00
|
|
|
* fsync vnode op. Just call nfs_flush() with commit == 1.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
/* ARGSUSED */
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_fsync(ap)
|
|
|
|
struct vop_fsync_args /* {
|
|
|
|
struct vnodeop_desc *a_desc;
|
|
|
|
struct vnode * a_vp;
|
|
|
|
struct ucred * a_cred;
|
|
|
|
int a_waitfor;
|
|
|
|
struct proc * a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
1995-06-27 11:07:30 +00:00
|
|
|
|
|
|
|
return (nfs_flush(ap->a_vp, ap->a_cred, ap->a_waitfor, ap->a_p, 1));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Flush all the blocks associated with a vnode.
|
|
|
|
* Walk through the buffer pool and push any dirty pages
|
|
|
|
* associated with the vnode.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_flush(vp, cred, waitfor, p, commit)
|
|
|
|
register struct vnode *vp;
|
|
|
|
struct ucred *cred;
|
|
|
|
int waitfor;
|
|
|
|
struct proc *p;
|
|
|
|
int commit;
|
|
|
|
{
|
1994-05-24 10:09:53 +00:00
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
register struct buf *bp;
|
1995-06-27 11:07:30 +00:00
|
|
|
register int i;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct buf *nbp;
|
1995-06-27 11:07:30 +00:00
|
|
|
struct nfsmount *nmp = VFSTONFS(vp->v_mount);
|
|
|
|
int s, error = 0, slptimeo = 0, slpflag = 0, retv, bvecpos;
|
|
|
|
int passone = 1;
|
1997-06-03 10:03:40 +00:00
|
|
|
u_quad_t off, endoff, toff;
|
1995-06-28 17:33:39 +00:00
|
|
|
struct ucred* wcred = NULL;
|
1997-06-03 10:03:40 +00:00
|
|
|
struct buf **bvec = NULL;
|
1995-06-27 11:07:30 +00:00
|
|
|
#ifndef NFS_COMMITBVECSIZ
|
|
|
|
#define NFS_COMMITBVECSIZ 20
|
|
|
|
#endif
|
1997-06-03 10:03:40 +00:00
|
|
|
struct buf *bvec_on_stack[NFS_COMMITBVECSIZ];
|
|
|
|
int bvecsize = 0, bveccount;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
if (nmp->nm_flag & NFSMNT_INT)
|
|
|
|
slpflag = PCATCH;
|
1995-06-27 11:07:30 +00:00
|
|
|
if (!commit)
|
|
|
|
passone = 0;
|
|
|
|
/*
|
|
|
|
* A b_flags == (B_DELWRI | B_NEEDCOMMIT) block has been written to the
|
|
|
|
* server, but nas not been committed to stable storage on the server
|
|
|
|
* yet. On the first pass, the byte range is worked out and the commit
|
|
|
|
* rpc is done. On the second pass, nfs_writebp() is called to do the
|
|
|
|
* job.
|
|
|
|
*/
|
|
|
|
again:
|
1997-06-03 10:03:40 +00:00
|
|
|
off = (u_quad_t)-1;
|
|
|
|
endoff = 0;
|
1995-06-27 11:07:30 +00:00
|
|
|
bvecpos = 0;
|
|
|
|
if (NFS_ISV3(vp) && commit) {
|
|
|
|
s = splbio();
|
1997-06-03 10:03:40 +00:00
|
|
|
/*
|
|
|
|
* Count up how many buffers waiting for a commit.
|
|
|
|
*/
|
|
|
|
bveccount = 0;
|
1998-10-31 15:31:29 +00:00
|
|
|
for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
|
|
|
|
nbp = TAILQ_NEXT(bp, b_vnbufs);
|
1999-06-26 02:47:16 +00:00
|
|
|
if (BUF_REFCNT(bp) == 0 &&
|
|
|
|
(bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
|
|
|
|
== (B_DELWRI | B_NEEDCOMMIT))
|
1997-06-03 10:03:40 +00:00
|
|
|
bveccount++;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Allocate space to remember the list of bufs to commit. It is
|
|
|
|
* important to use M_NOWAIT here to avoid a race with nfs_write.
|
|
|
|
* If we can't get memory (for whatever reason), we will end up
|
|
|
|
* committing the buffers one-by-one in the loop below.
|
|
|
|
*/
|
|
|
|
if (bveccount > NFS_COMMITBVECSIZ) {
|
|
|
|
if (bvec != NULL && bvec != bvec_on_stack)
|
|
|
|
free(bvec, M_TEMP);
|
|
|
|
bvec = (struct buf **)
|
|
|
|
malloc(bveccount * sizeof(struct buf *),
|
|
|
|
M_TEMP, M_NOWAIT);
|
|
|
|
if (bvec == NULL) {
|
|
|
|
bvec = bvec_on_stack;
|
|
|
|
bvecsize = NFS_COMMITBVECSIZ;
|
|
|
|
} else
|
|
|
|
bvecsize = bveccount;
|
|
|
|
} else {
|
|
|
|
bvec = bvec_on_stack;
|
|
|
|
bvecsize = NFS_COMMITBVECSIZ;
|
|
|
|
}
|
1998-10-31 15:31:29 +00:00
|
|
|
for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
|
|
|
|
nbp = TAILQ_NEXT(bp, b_vnbufs);
|
1997-06-03 10:03:40 +00:00
|
|
|
if (bvecpos >= bvecsize)
|
1995-06-27 11:07:30 +00:00
|
|
|
break;
|
1999-06-26 02:47:16 +00:00
|
|
|
if ((bp->b_flags & (B_DELWRI | B_NEEDCOMMIT)) !=
|
|
|
|
(B_DELWRI | B_NEEDCOMMIT) ||
|
|
|
|
BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT))
|
1995-06-27 11:07:30 +00:00
|
|
|
continue;
|
|
|
|
bremfree(bp);
|
1995-06-28 17:33:39 +00:00
|
|
|
/*
|
|
|
|
* Work out if all buffers are using the same cred
|
|
|
|
* so we can deal with them all with one commit.
|
1999-03-12 02:24:58 +00:00
|
|
|
*
|
|
|
|
* NOTE: we are not clearing B_DONE here, so we have
|
|
|
|
* to do it later on in this routine if we intend to
|
|
|
|
* initiate I/O on the bp.
|
1995-06-28 17:33:39 +00:00
|
|
|
*/
|
|
|
|
if (wcred == NULL)
|
|
|
|
wcred = bp->b_wcred;
|
|
|
|
else if (wcred != bp->b_wcred)
|
|
|
|
wcred = NOCRED;
|
1999-06-26 02:47:16 +00:00
|
|
|
bp->b_flags |= B_WRITEINPROG;
|
1995-06-27 11:07:30 +00:00
|
|
|
vfs_busy_pages(bp, 1);
|
1999-03-12 02:24:58 +00:00
|
|
|
|
|
|
|
/*
|
1999-06-26 02:47:16 +00:00
|
|
|
* bp is protected by being locked, but nbp is not
|
1999-03-12 02:24:58 +00:00
|
|
|
* and vfs_busy_pages() may sleep. We have to
|
|
|
|
* recalculate nbp.
|
|
|
|
*/
|
|
|
|
nbp = TAILQ_NEXT(bp, b_vnbufs);
|
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
/*
|
|
|
|
* A list of these buffers is kept so that the
|
|
|
|
* second loop knows which buffers have actually
|
|
|
|
* been committed. This is necessary, since there
|
|
|
|
* may be a race between the commit rpc and new
|
|
|
|
* uncommitted writes on the file.
|
|
|
|
*/
|
|
|
|
bvec[bvecpos++] = bp;
|
|
|
|
toff = ((u_quad_t)bp->b_blkno) * DEV_BSIZE +
|
|
|
|
bp->b_dirtyoff;
|
|
|
|
if (toff < off)
|
|
|
|
off = toff;
|
|
|
|
toff += (u_quad_t)(bp->b_dirtyend - bp->b_dirtyoff);
|
|
|
|
if (toff > endoff)
|
|
|
|
endoff = toff;
|
|
|
|
}
|
|
|
|
splx(s);
|
|
|
|
}
|
|
|
|
if (bvecpos > 0) {
|
|
|
|
/*
|
|
|
|
* Commit data on the server, as required.
|
1995-06-28 17:33:39 +00:00
|
|
|
* If all bufs are using the same wcred, then use that with
|
|
|
|
* one call for all of them, otherwise commit each one
|
|
|
|
* separately.
|
1995-06-27 11:07:30 +00:00
|
|
|
*/
|
1995-06-28 17:33:39 +00:00
|
|
|
if (wcred != NOCRED)
|
|
|
|
retv = nfs_commit(vp, off, (int)(endoff - off),
|
|
|
|
wcred, p);
|
|
|
|
else {
|
|
|
|
retv = 0;
|
|
|
|
for (i = 0; i < bvecpos; i++) {
|
|
|
|
off_t off, size;
|
|
|
|
bp = bvec[i];
|
|
|
|
off = ((u_quad_t)bp->b_blkno) * DEV_BSIZE +
|
|
|
|
bp->b_dirtyoff;
|
|
|
|
size = (u_quad_t)(bp->b_dirtyend
|
|
|
|
- bp->b_dirtyoff);
|
|
|
|
retv = nfs_commit(vp, off, (int)size,
|
|
|
|
bp->b_wcred, p);
|
|
|
|
if (retv) break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
if (retv == NFSERR_STALEWRITEVERF)
|
|
|
|
nfs_clearcommit(vp->v_mount);
|
1999-03-12 02:24:58 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
/*
|
|
|
|
* Now, either mark the blocks I/O done or mark the
|
|
|
|
* blocks dirty, depending on whether the commit
|
|
|
|
* succeeded.
|
|
|
|
*/
|
|
|
|
for (i = 0; i < bvecpos; i++) {
|
|
|
|
bp = bvec[i];
|
|
|
|
bp->b_flags &= ~(B_NEEDCOMMIT | B_WRITEINPROG);
|
|
|
|
if (retv) {
|
1999-03-12 02:24:58 +00:00
|
|
|
/*
|
|
|
|
* Error, leave B_DELWRI intact
|
|
|
|
*/
|
|
|
|
vfs_unbusy_pages(bp);
|
|
|
|
brelse(bp);
|
1995-06-27 11:07:30 +00:00
|
|
|
} else {
|
1999-03-12 02:24:58 +00:00
|
|
|
/*
|
|
|
|
* Success, remove B_DELWRI ( bundirty() ).
|
|
|
|
*
|
|
|
|
* b_dirtyoff/b_dirtyend seem to be NFS
|
|
|
|
* specific. We should probably move that
|
|
|
|
* into bundirty(). XXX
|
|
|
|
*/
|
|
|
|
s = splbio();
|
|
|
|
vp->v_numoutput++;
|
|
|
|
bp->b_flags |= B_ASYNC;
|
|
|
|
bundirty(bp);
|
|
|
|
bp->b_flags &= ~(B_READ|B_DONE|B_ERROR);
|
|
|
|
bp->b_dirtyoff = bp->b_dirtyend = 0;
|
|
|
|
splx(s);
|
|
|
|
biodone(bp);
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Start/do any write(s) that are required.
|
|
|
|
*/
|
1994-05-24 10:09:53 +00:00
|
|
|
loop:
|
|
|
|
s = splbio();
|
1998-10-31 15:31:29 +00:00
|
|
|
for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
|
|
|
|
nbp = TAILQ_NEXT(bp, b_vnbufs);
|
1999-06-26 02:47:16 +00:00
|
|
|
if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
|
1995-06-27 11:07:30 +00:00
|
|
|
if (waitfor != MNT_WAIT || passone)
|
1994-05-24 10:09:53 +00:00
|
|
|
continue;
|
1999-06-26 02:47:16 +00:00
|
|
|
error = BUF_TIMELOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL,
|
|
|
|
"nfsfsync", slpflag, slptimeo);
|
1994-05-24 10:09:53 +00:00
|
|
|
splx(s);
|
1999-06-26 02:47:16 +00:00
|
|
|
if (error == 0)
|
|
|
|
panic("nfs_fsync: inconsistent lock");
|
|
|
|
if (error == ENOLCK)
|
|
|
|
goto loop;
|
|
|
|
if (nfs_sigintr(nmp, (struct nfsreq *)0, p)) {
|
1997-06-03 10:03:40 +00:00
|
|
|
error = EINTR;
|
|
|
|
goto done;
|
1999-06-26 02:47:16 +00:00
|
|
|
}
|
|
|
|
if (slpflag == PCATCH) {
|
1994-05-24 10:09:53 +00:00
|
|
|
slpflag = 0;
|
|
|
|
slptimeo = 2 * hz;
|
|
|
|
}
|
|
|
|
goto loop;
|
|
|
|
}
|
|
|
|
if ((bp->b_flags & B_DELWRI) == 0)
|
|
|
|
panic("nfs_fsync: not dirty");
|
1999-06-26 02:47:16 +00:00
|
|
|
if ((passone || !commit) && (bp->b_flags & B_NEEDCOMMIT)) {
|
|
|
|
BUF_UNLOCK(bp);
|
1995-06-27 11:07:30 +00:00
|
|
|
continue;
|
1999-06-26 02:47:16 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
bremfree(bp);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (passone || !commit)
|
1999-06-26 02:47:16 +00:00
|
|
|
bp->b_flags |= B_ASYNC;
|
1995-06-27 11:07:30 +00:00
|
|
|
else
|
1999-06-26 02:47:16 +00:00
|
|
|
bp->b_flags |= (B_ASYNC | B_WRITEINPROG | B_NEEDCOMMIT);
|
1994-05-24 10:09:53 +00:00
|
|
|
splx(s);
|
1999-06-16 23:27:55 +00:00
|
|
|
VOP_BWRITE(bp->b_vp, bp);
|
1994-05-24 10:09:53 +00:00
|
|
|
goto loop;
|
|
|
|
}
|
|
|
|
splx(s);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (passone) {
|
|
|
|
passone = 0;
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
if (waitfor == MNT_WAIT) {
|
1994-05-24 10:09:53 +00:00
|
|
|
while (vp->v_numoutput) {
|
|
|
|
vp->v_flag |= VBWAIT;
|
|
|
|
error = tsleep((caddr_t)&vp->v_numoutput,
|
|
|
|
slpflag | (PRIBIO + 1), "nfsfsync", slptimeo);
|
|
|
|
if (error) {
|
1997-06-03 10:03:40 +00:00
|
|
|
if (nfs_sigintr(nmp, (struct nfsreq *)0, p)) {
|
|
|
|
error = EINTR;
|
|
|
|
goto done;
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
if (slpflag == PCATCH) {
|
|
|
|
slpflag = 0;
|
|
|
|
slptimeo = 2 * hz;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
1998-10-31 15:31:29 +00:00
|
|
|
if (!TAILQ_EMPTY(&vp->v_dirtyblkhd) && commit) {
|
1994-05-24 10:09:53 +00:00
|
|
|
goto loop;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (np->n_flag & NWRITEERR) {
|
|
|
|
error = np->n_error;
|
|
|
|
np->n_flag &= ~NWRITEERR;
|
|
|
|
}
|
1997-06-03 10:03:40 +00:00
|
|
|
done:
|
|
|
|
if (bvec != NULL && bvec != bvec_on_stack)
|
|
|
|
free(bvec, M_TEMP);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* NFS advisory byte-level locks.
|
|
|
|
* Currently unsupported.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_advlock(ap)
|
|
|
|
struct vop_advlock_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
caddr_t a_id;
|
|
|
|
int a_op;
|
|
|
|
struct flock *a_fl;
|
|
|
|
int a_flags;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
1994-08-08 17:31:01 +00:00
|
|
|
register struct nfsnode *np = VTONFS(ap->a_vp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1994-08-08 17:31:01 +00:00
|
|
|
/*
|
|
|
|
* The following kludge is to allow diskless support to work
|
|
|
|
* until a real NFS lockd is implemented. Basically, just pretend
|
|
|
|
* that this is a local lock.
|
|
|
|
*/
|
|
|
|
return (lf_advlock(ap, &(np->n_lockf), np->n_size));
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Print out the contents of an nfsnode.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfs_print(ap)
|
|
|
|
struct vop_print_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
|
1994-10-02 17:27:07 +00:00
|
|
|
printf("tag VT_NFS, fileid %ld fsid 0x%lx",
|
1994-05-24 10:09:53 +00:00
|
|
|
np->n_vattr.va_fileid, np->n_vattr.va_fsid);
|
|
|
|
if (vp->v_type == VFIFO)
|
|
|
|
fifo_printinfo(vp);
|
|
|
|
printf("\n");
|
1994-05-25 09:21:21 +00:00
|
|
|
return (0);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
/*
|
|
|
|
* Just call nfs_writebp() with the force argument set to 1.
|
1999-03-12 02:24:58 +00:00
|
|
|
*
|
|
|
|
* NOTE: B_DONE may or may not be set in a_bp on call.
|
1995-06-27 11:07:30 +00:00
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_bwrite(ap)
|
|
|
|
struct vop_bwrite_args /* {
|
|
|
|
struct vnode *a_bp;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
1999-05-06 20:00:34 +00:00
|
|
|
return (nfs_writebp(ap->a_bp, 1, curproc));
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This is a clone of vn_bwrite(), except that B_WRITEINPROG isn't set unless
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
* the force flag is one and it also handles the B_NEEDCOMMIT flag. We set
|
|
|
|
* B_CACHE if this is a VMIO buffer.
|
1995-06-27 11:07:30 +00:00
|
|
|
*/
|
|
|
|
int
|
1999-05-06 20:00:34 +00:00
|
|
|
nfs_writebp(bp, force, procp)
|
1995-06-27 11:07:30 +00:00
|
|
|
register struct buf *bp;
|
|
|
|
int force;
|
1999-05-06 20:00:34 +00:00
|
|
|
struct proc *procp;
|
1995-06-27 11:07:30 +00:00
|
|
|
{
|
1998-03-08 09:59:44 +00:00
|
|
|
int s;
|
1999-03-12 02:24:58 +00:00
|
|
|
int oldflags = bp->b_flags;
|
|
|
|
int retv = 1;
|
1995-06-27 11:07:30 +00:00
|
|
|
off_t off;
|
|
|
|
|
1999-06-26 02:47:16 +00:00
|
|
|
if (BUF_REFCNT(bp) == 0)
|
|
|
|
panic("bwrite: buffer is not locked???");
|
1995-06-27 11:07:30 +00:00
|
|
|
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
if (bp->b_flags & B_INVAL) {
|
|
|
|
brelse(bp);
|
|
|
|
return(0);
|
|
|
|
}
|
|
|
|
|
|
|
|
bp->b_flags |= B_CACHE;
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
|
1999-03-12 02:24:58 +00:00
|
|
|
/*
|
The VFS/BIO subsystem contained a number of hacks in order to optimize
piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
1999-05-02 23:57:16 +00:00
|
|
|
* Undirty the bp. We will redirty it later if the I/O fails.
|
1999-03-12 02:24:58 +00:00
|
|
|
*/
|
1995-06-27 11:07:30 +00:00
|
|
|
|
1999-03-12 02:24:58 +00:00
|
|
|
s = splbio();
|
|
|
|
bundirty(bp);
|
|
|
|
bp->b_flags &= ~(B_READ|B_DONE|B_ERROR);
|
1995-06-27 11:07:30 +00:00
|
|
|
|
|
|
|
bp->b_vp->v_numoutput++;
|
|
|
|
curproc->p_stats->p_ru.ru_oublock++;
|
1998-03-08 09:59:44 +00:00
|
|
|
splx(s);
|
1995-06-27 11:07:30 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If B_NEEDCOMMIT is set, a commit rpc may do the trick. If not
|
|
|
|
* an actual write will have to be scheduled via. VOP_STRATEGY().
|
|
|
|
* If B_WRITEINPROG is already set, then push it with a write anyhow.
|
|
|
|
*/
|
1997-02-28 17:56:27 +00:00
|
|
|
vfs_busy_pages(bp, 1);
|
1997-02-10 02:22:35 +00:00
|
|
|
if ((oldflags & (B_NEEDCOMMIT | B_WRITEINPROG)) == B_NEEDCOMMIT) {
|
1995-06-27 11:07:30 +00:00
|
|
|
off = ((u_quad_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
|
|
|
|
bp->b_flags |= B_WRITEINPROG;
|
|
|
|
retv = nfs_commit(bp->b_vp, off, bp->b_dirtyend-bp->b_dirtyoff,
|
1999-05-06 20:00:34 +00:00
|
|
|
bp->b_wcred, procp);
|
1995-06-27 11:07:30 +00:00
|
|
|
bp->b_flags &= ~B_WRITEINPROG;
|
|
|
|
if (!retv) {
|
|
|
|
bp->b_dirtyoff = bp->b_dirtyend = 0;
|
|
|
|
bp->b_flags &= ~B_NEEDCOMMIT;
|
|
|
|
biodone(bp);
|
1999-03-12 02:24:58 +00:00
|
|
|
} else if (retv == NFSERR_STALEWRITEVERF) {
|
1995-06-27 11:07:30 +00:00
|
|
|
nfs_clearcommit(bp->b_vp->v_mount);
|
1999-03-12 02:24:58 +00:00
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
|
|
|
if (retv) {
|
|
|
|
if (force)
|
|
|
|
bp->b_flags |= B_WRITEINPROG;
|
1999-06-26 02:47:16 +00:00
|
|
|
BUF_KERNPROC(bp);
|
1998-07-04 20:45:42 +00:00
|
|
|
VOP_STRATEGY(bp->b_vp, bp);
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if( (oldflags & B_ASYNC) == 0) {
|
|
|
|
int rtval = biowait(bp);
|
|
|
|
|
|
|
|
if (oldflags & B_DELWRI) {
|
1998-03-08 09:59:44 +00:00
|
|
|
s = splbio();
|
1995-06-27 11:07:30 +00:00
|
|
|
reassignbuf(bp, bp->b_vp);
|
1998-03-08 09:59:44 +00:00
|
|
|
splx(s);
|
1995-06-27 11:07:30 +00:00
|
|
|
}
|
This mega-commit is meant to fix numerous interrelated problems. There
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
1998-03-07 21:37:31 +00:00
|
|
|
|
1995-06-27 11:07:30 +00:00
|
|
|
brelse(bp);
|
|
|
|
return (rtval);
|
|
|
|
}
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* nfs special file access vnode op.
|
|
|
|
* Essentially just get vattr and then imitate iaccess() since the device is
|
|
|
|
* local to the client.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsspec_access(ap)
|
|
|
|
struct vop_access_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
int a_mode;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vattr *vap;
|
|
|
|
register gid_t *gp;
|
|
|
|
register struct ucred *cred = ap->a_cred;
|
1995-10-22 09:32:48 +00:00
|
|
|
struct vnode *vp = ap->a_vp;
|
1994-05-24 10:09:53 +00:00
|
|
|
mode_t mode = ap->a_mode;
|
|
|
|
struct vattr vattr;
|
|
|
|
register int i;
|
|
|
|
int error;
|
|
|
|
|
1995-10-22 09:32:48 +00:00
|
|
|
/*
|
|
|
|
* Disallow write attempts on filesystems mounted read-only;
|
|
|
|
* unless the file is a socket, fifo, or a block or character
|
|
|
|
* device resident on the filesystem.
|
|
|
|
*/
|
|
|
|
if ((mode & VWRITE) && (vp->v_mount->mnt_flag & MNT_RDONLY)) {
|
|
|
|
switch (vp->v_type) {
|
1998-05-31 17:27:58 +00:00
|
|
|
case VREG:
|
|
|
|
case VDIR:
|
|
|
|
case VLNK:
|
1995-10-22 09:32:48 +00:00
|
|
|
return (EROFS);
|
1998-05-31 17:27:58 +00:00
|
|
|
default:
|
|
|
|
break;
|
1995-10-22 09:32:48 +00:00
|
|
|
}
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* If you're the super-user,
|
|
|
|
* you always get access.
|
|
|
|
*/
|
|
|
|
if (cred->cr_uid == 0)
|
|
|
|
return (0);
|
|
|
|
vap = &vattr;
|
1997-02-10 02:22:35 +00:00
|
|
|
error = VOP_GETATTR(vp, vap, cred, ap->a_p);
|
1994-10-02 17:27:07 +00:00
|
|
|
if (error)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
/*
|
|
|
|
* Access check is based on only one of owner, group, public.
|
|
|
|
* If not owner, then check group. If not a member of the
|
|
|
|
* group, then check public access.
|
|
|
|
*/
|
|
|
|
if (cred->cr_uid != vap->va_uid) {
|
|
|
|
mode >>= 3;
|
|
|
|
gp = cred->cr_groups;
|
|
|
|
for (i = 0; i < cred->cr_ngroups; i++, gp++)
|
|
|
|
if (vap->va_gid == *gp)
|
|
|
|
goto found;
|
|
|
|
mode >>= 3;
|
|
|
|
found:
|
|
|
|
;
|
|
|
|
}
|
1995-06-27 11:07:30 +00:00
|
|
|
error = (vap->va_mode & mode) == mode ? 0 : EACCES;
|
|
|
|
return (error);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read wrapper for special devices.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsspec_read(ap)
|
|
|
|
struct vop_read_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct uio *a_uio;
|
|
|
|
int a_ioflag;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct nfsnode *np = VTONFS(ap->a_vp);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set access flag.
|
|
|
|
*/
|
|
|
|
np->n_flag |= NACC;
|
1998-05-16 16:20:50 +00:00
|
|
|
getnanotime(&np->n_atim);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Write wrapper for special devices.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsspec_write(ap)
|
|
|
|
struct vop_write_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct uio *a_uio;
|
|
|
|
int a_ioflag;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct nfsnode *np = VTONFS(ap->a_vp);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set update flag.
|
|
|
|
*/
|
|
|
|
np->n_flag |= NUPD;
|
1998-05-16 16:20:50 +00:00
|
|
|
getnanotime(&np->n_mtim);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Close wrapper for special devices.
|
|
|
|
*
|
|
|
|
* Update the times on the nfsnode then do device close.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsspec_close(ap)
|
|
|
|
struct vop_close_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
int a_fflag;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
struct vattr vattr;
|
|
|
|
|
|
|
|
if (np->n_flag & (NACC | NUPD)) {
|
|
|
|
np->n_flag |= NCHG;
|
|
|
|
if (vp->v_usecount == 1 &&
|
|
|
|
(vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
|
|
|
|
VATTR_NULL(&vattr);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (np->n_flag & NACC)
|
|
|
|
vattr.va_atime = np->n_atim;
|
|
|
|
if (np->n_flag & NUPD)
|
|
|
|
vattr.va_mtime = np->n_mtim;
|
1994-05-24 10:09:53 +00:00
|
|
|
(void)VOP_SETATTR(vp, &vattr, ap->a_cred, ap->a_p);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read wrapper for fifos.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsfifo_read(ap)
|
|
|
|
struct vop_read_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct uio *a_uio;
|
|
|
|
int a_ioflag;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct nfsnode *np = VTONFS(ap->a_vp);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set access flag.
|
|
|
|
*/
|
|
|
|
np->n_flag |= NACC;
|
1998-05-16 16:20:50 +00:00
|
|
|
getnanotime(&np->n_atim);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Write wrapper for fifos.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsfifo_write(ap)
|
|
|
|
struct vop_write_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
struct uio *a_uio;
|
|
|
|
int a_ioflag;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct nfsnode *np = VTONFS(ap->a_vp);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set update flag.
|
|
|
|
*/
|
|
|
|
np->n_flag |= NUPD;
|
1998-05-16 16:20:50 +00:00
|
|
|
getnanotime(&np->n_mtim);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Close wrapper for fifos.
|
|
|
|
*
|
|
|
|
* Update the times on the nfsnode then do fifo close.
|
|
|
|
*/
|
1995-10-29 15:33:36 +00:00
|
|
|
static int
|
1994-05-24 10:09:53 +00:00
|
|
|
nfsfifo_close(ap)
|
|
|
|
struct vop_close_args /* {
|
|
|
|
struct vnode *a_vp;
|
|
|
|
int a_fflag;
|
|
|
|
struct ucred *a_cred;
|
|
|
|
struct proc *a_p;
|
|
|
|
} */ *ap;
|
|
|
|
{
|
|
|
|
register struct vnode *vp = ap->a_vp;
|
|
|
|
register struct nfsnode *np = VTONFS(vp);
|
|
|
|
struct vattr vattr;
|
1998-05-16 16:20:50 +00:00
|
|
|
struct timespec ts;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
if (np->n_flag & (NACC | NUPD)) {
|
1998-05-16 16:20:50 +00:00
|
|
|
getnanotime(&ts);
|
|
|
|
if (np->n_flag & NACC)
|
|
|
|
np->n_atim = ts;
|
|
|
|
if (np->n_flag & NUPD)
|
|
|
|
np->n_mtim = ts;
|
1994-05-24 10:09:53 +00:00
|
|
|
np->n_flag |= NCHG;
|
|
|
|
if (vp->v_usecount == 1 &&
|
|
|
|
(vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
|
|
|
|
VATTR_NULL(&vattr);
|
1995-06-27 11:07:30 +00:00
|
|
|
if (np->n_flag & NACC)
|
|
|
|
vattr.va_atime = np->n_atim;
|
|
|
|
if (np->n_flag & NUPD)
|
|
|
|
vattr.va_mtime = np->n_mtim;
|
1994-05-24 10:09:53 +00:00
|
|
|
(void)VOP_SETATTR(vp, &vattr, ap->a_cred, ap->a_p);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap));
|
|
|
|
}
|