4.4BSD snapshot (revision 8.1); add 1993 to copyright
[unix-history] / usr / src / sys / ufs / ffs / ffs_vfsops.c
index 92af9c1..55530ef 100644 (file)
 /*
 /*
- * Copyright (c) 1982, 1986 Regents of the University of California.
- * All rights reserved.  The Berkeley software License Agreement
- * specifies the terms and conditions for redistribution.
+ * Copyright (c) 1989, 1991, 1993
+ *     The Regents of the University of California.  All rights reserved.
  *
  *
- *     @(#)ffs_vfsops.c        7.7 (Berkeley) %G%
+ * %sccs.include.redist.c%
+ *
+ *     @(#)ffs_vfsops.c        8.1 (Berkeley) %G%
  */
 
  */
 
-#include "param.h"
-#include "systm.h"
-#include "dir.h"
-#include "user.h"
-#include "inode.h"
-#include "proc.h"
-#include "fs.h"
-#include "buf.h"
-#include "mount.h"
-#include "file.h"
-#include "conf.h"
-#include "ioctl.h"
-#include "disklabel.h"
-#include "stat.h"
-#include "malloc.h"
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/namei.h>
+#include <sys/proc.h>
+#include <sys/kernel.h>
+#include <sys/vnode.h>
+#include <sys/socket.h>
+#include <sys/mount.h>
+#include <sys/buf.h>
+#include <sys/mbuf.h>
+#include <sys/file.h>
+#include <sys/disklabel.h>
+#include <sys/ioctl.h>
+#include <sys/errno.h>
+#include <sys/malloc.h>
+
+#include <miscfs/specfs/specdev.h>
 #include "ioctl.h"
 #include "disklabel.h"
 #include "stat.h"
 
 #include "ioctl.h"
 #include "disklabel.h"
 #include "stat.h"
 
-smount()
+#include <ufs/ufs/quota.h>
+#include <ufs/ufs/ufsmount.h>
+#include <ufs/ufs/inode.h>
+#include <ufs/ufs/ufs_extern.h>
+
+#include <ufs/ffs/fs.h>
+#include <ufs/ffs/ffs_extern.h>
+
+int ffs_sbupdate __P((struct ufsmount *, int));
+
+struct vfsops ufs_vfsops = {
+       ffs_mount,
+       ufs_start,
+       ffs_unmount,
+       ffs_root,
+       ufs_quotactl,
+       ffs_statfs,
+       ffs_sync,
+       ffs_vget,
+       ffs_fhtovp,
+       ffs_vptofh,
+       ffs_init,
+};
+
+extern u_long nextgennumber;
+
+/*
+ * Called by main() when ufs is going to be mounted as root.
+ *
+ * Name is updated by mount(8) after booting.
+ */
+#define ROOTNAME       "root_device"
+
+ffs_mountroot()
 {
 {
-       register struct a {
-               char    *fspec;
-               char    *freg;
-               int     ronly;
-       } *uap = (struct a *)u.u_ap;
-       dev_t dev;
-       register struct inode *ip;
+       extern struct vnode *rootvp;
        register struct fs *fs;
        register struct fs *fs;
-       register struct nameidata *ndp = &u.u_nd;
-       u_int len;
-
-       u.u_error = getmdev(&dev, uap->fspec);
-       if (u.u_error)
-               return;
-       ndp->ni_nameiop = LOOKUP | FOLLOW;
-       ndp->ni_segflg = UIO_USERSPACE;
-       ndp->ni_dirp = (caddr_t)uap->freg;
-       ip = namei(ndp);
-       if (ip == NULL)
-               return;
-       if (ip->i_count != 1) {
-               iput(ip);
-               u.u_error = EBUSY;
-               return;
-       }
-       if ((ip->i_mode&IFMT) != IFDIR) {
-               iput(ip);
-               u.u_error = ENOTDIR;
-               return;
+       register struct mount *mp;
+       struct proc *p = curproc;       /* XXX */
+       struct ufsmount *ump;
+       u_int size;
+       int error;
+       
+       /*
+        * Get vnodes for swapdev and rootdev.
+        */
+       if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp))
+               panic("ffs_mountroot: can't setup bdevvp's");
+
+       mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
+       bzero((char *)mp, (u_long)sizeof(struct mount));
+       mp->mnt_op = &ufs_vfsops;
+       mp->mnt_flag = MNT_RDONLY;
+       if (error = ffs_mountfs(rootvp, mp, p)) {
+               free(mp, M_MOUNT);
+               return (error);
        }
        }
-       fs = mountfs(dev, uap->ronly, ip);
-       if (fs == 0) {
-               iput(ip);
-               return;
+       if (error = vfs_lock(mp)) {
+               (void)ffs_unmount(mp, 0, p);
+               free(mp, M_MOUNT);
+               return (error);
        }
        }
-       (void) copyinstr(uap->freg, fs->fs_fsmnt, sizeof(fs->fs_fsmnt)-1, &len);
-       bzero(fs->fs_fsmnt + len, sizeof (fs->fs_fsmnt) - len);
+       rootfs = mp;
+       mp->mnt_next = mp;
+       mp->mnt_prev = mp;
+       mp->mnt_vnodecovered = NULLVP;
+       ump = VFSTOUFS(mp);
+       fs = ump->um_fs;
+       bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
+       fs->fs_fsmnt[0] = '/';
+       bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
+           MNAMELEN);
+       (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
+           &size);
+       bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
+       (void)ffs_statfs(mp, &mp->mnt_stat, p);
+       vfs_unlock(mp);
+       inittodr(fs->fs_time);
+       return (0);
 }
 
 }
 
-struct fs *
-mountfs(dev, ronly, ip)
-       dev_t dev;
-       int ronly;
-       struct inode *ip;
-{
+/*
+ * VFS Operations.
+ *
+ * mount system call
+ */
+int
+ffs_mount(mp, path, data, ndp, p)
        register struct mount *mp;
        register struct mount *mp;
-       struct mount *fmp = NULL;
-       struct buf *tp = NULL;
-       register struct buf *bp = NULL;
+       char *path;
+       caddr_t data;
+       struct nameidata *ndp;
+       struct proc *p;
+{
+       struct vnode *devvp;
+       struct ufs_args args;
+       struct ufsmount *ump;
        register struct fs *fs;
        register struct fs *fs;
-       struct partinfo dpart;
-       int havepart = 0, blks;
-       caddr_t base, space;
-       int i, size;
-       register error;
-       int needclose = 0;
-
-       for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++) {
-               if (mp->m_fs == NULL) {
-                       if (fmp == NULL)
-                               fmp = mp;
-               } else if (dev == mp->m_dev) {
-                       u.u_error = EBUSY;              /* XXX */
-                       return ((struct fs *) NULL);
+       u_int size;
+       int error, flags;
+
+       if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)))
+               return (error);
+       /*
+        * If updating, check whether changing from read-only to
+        * read/write; if there is no device name, that's all we do.
+        */
+       if (mp->mnt_flag & MNT_UPDATE) {
+               ump = VFSTOUFS(mp);
+               fs = ump->um_fs;
+               error = 0;
+               if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
+                       flags = WRITECLOSE;
+                       if (mp->mnt_flag & MNT_FORCE)
+                               flags |= FORCECLOSE;
+                       if (vfs_busy(mp))
+                               return (EBUSY);
+                       error = ffs_flushfiles(mp, flags, p);
+                       vfs_unbusy(mp);
+               }
+               if (!error && (mp->mnt_flag & MNT_RELOAD))
+                       error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
+               if (error)
+                       return (error);
+               if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR))
+                       fs->fs_ronly = 0;
+               if (args.fspec == 0) {
+                       /*
+                        * Process export requests.
+                        */
+                       if (args.exflags & MNT_EXPORTED) {
+                               if (error = ufs_hang_addrlist(mp, &args))
+                                       return (error);
+                               mp->mnt_flag |= MNT_EXPORTED;
+                       }
+                       if (args.exflags & MNT_DELEXPORT) {
+                               ufs_free_addrlist(ump);
+                               mp->mnt_flag &=
+                                   ~(MNT_EXPORTED | MNT_DEFEXPORTED);
+                       }
+                       return (0);
                }
        }
                }
        }
-       if ((mp = fmp) == NULL) {
-               u.u_error = EMFILE;             /* needs translation      XXX */
-               return ((struct fs *) NULL);
+       /*
+        * Not an update, or updating the name: look up the name
+        * and verify that it refers to a sensible block device.
+        */
+       NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
+       if (error = namei(ndp))
+               return (error);
+       devvp = ndp->ni_vp;
+
+       if (devvp->v_type != VBLK) {
+               vrele(devvp);
+               return (ENOTBLK);
+       }
+       if (major(devvp->v_rdev) >= nblkdev) {
+               vrele(devvp);
+               return (ENXIO);
+       }
+       if ((mp->mnt_flag & MNT_UPDATE) == 0)
+               error = ffs_mountfs(devvp, mp, p);
+       else {
+               if (devvp != ump->um_devvp)
+                       error = EINVAL; /* needs translation */
+               else
+                       vrele(devvp);
        }
        }
-       mp->m_fs = (struct fs *)1;      /* just to reserve this slot */
-       mp->m_dev = dev;
-       error =
-           (*bdevsw[major(dev)].d_open)(dev, ronly ? FREAD : FREAD|FWRITE,
-               S_IFBLK);
        if (error) {
        if (error) {
-               u.u_error = error;
-               mp->m_fs = NULL;
-               return ((struct fs *) NULL);
+               vrele(devvp);
+               return (error);
        }
        }
-       needclose = 1;
-       if ((*bdevsw[major(dev)].d_ioctl)(dev, DIOCGPART,
-           (caddr_t)&dpart, FREAD) == 0) {
-               havepart = 1;
-               size = dpart.disklab->d_secsize;
-       } else
-               size = DEV_BSIZE;
-#ifdef SECSIZE
+       ump = VFSTOUFS(mp);
+       fs = ump->um_fs;
+       (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
+       bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
+       bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
+           MNAMELEN);
+       (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 
+           &size);
+       bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
+       (void)ffs_statfs(mp, &mp->mnt_stat, p);
+       return (0);
+}
+
+/*
+ * Reload all incore data for a filesystem (used after running fsck on
+ * the root filesystem and finding things to fix). The filesystem must
+ * be mounted read-only.
+ *
+ * Things to do to update the mount:
+ *     1) invalidate all cached meta-data.
+ *     2) re-read superblock from disk.
+ *     3) re-read summary information from disk.
+ *     4) invalidate all inactive vnodes.
+ *     5) invalidate all cached file data.
+ *     6) re-read inode data for all active vnodes.
+ */
+ffs_reload(mountp, cred, p)
+       register struct mount *mountp;
+       struct ucred *cred;
+       struct proc *p;
+{
+       register struct vnode *vp, *nvp, *devvp;
+       struct inode *ip;
+       struct dinode *dp;
+       struct csum *space;
+       struct buf *bp;
+       struct fs *fs;
+       int i, blks, size, error;
+
+       if ((mountp->mnt_flag & MNT_RDONLY) == 0)
+               return (EINVAL);
        /*
        /*
-        * If possible, determine hardware sector size
-        * and adjust fsbtodb to correspond.
+        * Step 1: invalidate all cached meta-data.
         */
         */
-#endif SECSIZE
-       if ((*bdevsw[major(dev)].d_ioctl)(dev, DIOCGPART,
-           (caddr_t)&dpart, FREAD) == 0) {
-               havepart = 1;
-               size = dpart.disklab->d_secsize;
-#ifdef SECSIZE
-               if (size < MINSECSIZE) {
-                       error = EINVAL;
-                       goto out;
+       devvp = VFSTOUFS(mountp)->um_devvp;
+       if (vinvalbuf(devvp, 0, cred, p, 0, 0))
+               panic("ffs_reload: dirty1");
+       /*
+        * Step 2: re-read superblock from disk.
+        */
+       if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp))
+               return (error);
+       fs = bp->b_un.b_fs;
+       if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
+           fs->fs_bsize < sizeof(struct fs)) {
+               brelse(bp);
+               return (EIO);           /* XXX needs translation */
+       }
+       fs = VFSTOUFS(mountp)->um_fs;
+       bcopy((caddr_t)&fs->fs_csp[0], (caddr_t)&bp->b_un.b_fs->fs_csp[0],
+           sizeof(fs->fs_csp));
+       bcopy((caddr_t)bp->b_un.b_addr, (caddr_t)fs, (u_int)fs->fs_sbsize);
+       if (fs->fs_sbsize < SBSIZE)
+               bp->b_flags |= B_INVAL;
+       brelse(bp);
+       ffs_oldfscompat(fs);
+       /*
+        * Step 3: re-read summary information from disk.
+        */
+       blks = howmany(fs->fs_cssize, fs->fs_fsize);
+       space = fs->fs_csp[0];
+       for (i = 0; i < blks; i += fs->fs_frag) {
+               size = fs->fs_bsize;
+               if (i + fs->fs_frag > blks)
+                       size = (blks - i) * fs->fs_fsize;
+               if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
+                   NOCRED, &bp))
+                       return (error);
+               bcopy((caddr_t)bp->b_un.b_addr, fs->fs_csp[fragstoblks(fs, i)],
+                   (u_int)size);
+               brelse(bp);
+       }
+loop:
+       for (vp = mountp->mnt_mounth; vp; vp = nvp) {
+               nvp = vp->v_mountf;
+               /*
+                * Step 4: invalidate all inactive vnodes.
+                */
+               if (vp->v_usecount == 0) {
+                       vgone(vp);
+                       continue;
                }
                }
-#endif SECSIZE
-       } else
+               /*
+                * Step 5: invalidate all cached file data.
+                */
+               if (vget(vp))
+                       goto loop;
+               if (vinvalbuf(vp, 0, cred, p, 0, 0))
+                       panic("ffs_reload: dirty2");
+               /*
+                * Step 6: re-read inode data for all active vnodes.
+                */
+               ip = VTOI(vp);
+               if (error = bread(devvp, fsbtodb(fs, itod(fs, ip->i_number)),
+                   (int)fs->fs_bsize, NOCRED, &bp)) {
+                       vput(vp);
+                       return (error);
+               }
+               dp = bp->b_un.b_dino;
+               dp += itoo(fs, ip->i_number);
+               ip->i_din = *dp;
+               brelse(bp);
+               vput(vp);
+               if (vp->v_mount != mountp)
+                       goto loop;
+       }
+       return (0);
+}
+
+/*
+ * Common code for mount and mountroot
+ */
+int
+ffs_mountfs(devvp, mp, p)
+       register struct vnode *devvp;
+       struct mount *mp;
+       struct proc *p;
+{
+       register struct ufsmount *ump;
+       struct buf *bp;
+       register struct fs *fs;
+       dev_t dev = devvp->v_rdev;
+       struct partinfo dpart;
+       int havepart = 0, blks;
+       caddr_t base, space;
+       int havepart = 0, blks;
+       int error, i, size;
+       int ronly;
+       extern struct vnode *rootvp;
+
+       if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p))
+               return (error);
+       if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
                size = DEV_BSIZE;
                size = DEV_BSIZE;
-#ifdef SECSIZE
-       tp = bread(dev, (daddr_t)(SBOFF / size), SBSIZE, size);
-#else SECSIZE
-       tp = bread(dev, SBLOCK, SBSIZE);
-#endif SECSIZE
-       if (tp->b_flags & B_ERROR) {
-               mp->m_fs = NULL;
-               goto out;
+       else {
+               havepart = 1;
+               size = dpart.disklab->d_secsize;
        }
        }
-       fs = tp->b_un.b_fs;
-       if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
-           fs->fs_bsize < sizeof(struct fs)) {
-               error = EINVAL;         /* also needs translation */
+
+       bp = NULL;
+       ump = NULL;
+       if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp))
+               goto out;
+       fs = bp->b_un.b_fs;
+               error = EINVAL;         /* XXX needs translation */
                goto out;
        }
                goto out;
        }
-       mp->m_fs = (struct fs *)malloc(fs->fs_sbsize, M_SUPERBLK, M_WAITOK);
-       bcopy((caddr_t)tp->b_un.b_addr, (caddr_t)mp->m_fs,
+       ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
+       bzero((caddr_t)ump, sizeof *ump);
+       ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
+           M_WAITOK);
+       bcopy((caddr_t)bp->b_un.b_addr, (caddr_t)ump->um_fs,
           (u_int)fs->fs_sbsize);
           (u_int)fs->fs_sbsize);
-       brelse(tp);
-       tp = NULL;
-       fs = mp->m_fs;
-       fs->fs_ronly = (ronly != 0);
+       if (fs->fs_sbsize < SBSIZE)
+               bp->b_flags |= B_INVAL;
+       brelse(bp);
+       bp = NULL;
+       fs = ump->um_fs;
+       fs->fs_ronly = ronly;
        if (ronly == 0)
                fs->fs_fmod = 1;
        if (havepart) {
        if (ronly == 0)
                fs->fs_fmod = 1;
        if (havepart) {
@@ -161,7 +376,6 @@ mountfs(dev, ronly, ip)
                dpart.part->p_fsize = fs->fs_fsize;
                dpart.part->p_frag = fs->fs_frag;
                dpart.part->p_cpg = fs->fs_cpg;
                dpart.part->p_fsize = fs->fs_fsize;
                dpart.part->p_frag = fs->fs_frag;
                dpart.part->p_cpg = fs->fs_cpg;
-               fs->fs_dbsize = size;
        }
 #ifdef SECSIZE
        /*
        }
 #ifdef SECSIZE
        /*
@@ -189,11 +403,8 @@ mountfs(dev, ronly, ip)
                fs->fs_dbsize = size;
        }
        blks = howmany(fs->fs_cssize, fs->fs_fsize);
                fs->fs_dbsize = size;
        }
        blks = howmany(fs->fs_cssize, fs->fs_fsize);
-       base = space = (caddr_t)malloc(fs->fs_cssize, M_SUPERBLK, M_WAITOK);
-       if (space == NULL) {
-               error = ENOMEM;
-               goto out;
-       }
+       base = space = malloc((u_long)fs->fs_cssize, M_UFSMNT,
+           M_WAITOK);
        for (i = 0; i < blks; i += fs->fs_frag) {
                size = fs->fs_bsize;
                if (i + fs->fs_frag > blks)
        for (i = 0; i < blks; i += fs->fs_frag) {
                size = fs->fs_bsize;
                if (i + fs->fs_frag > blks)
@@ -202,118 +413,428 @@ mountfs(dev, ronly, ip)
                tp = bread(dev, fsbtodb(fs, fs->fs_csaddr + i), size,
                    fs->fs_dbsize);
 #else SECSIZE
                tp = bread(dev, fsbtodb(fs, fs->fs_csaddr + i), size,
                    fs->fs_dbsize);
 #else SECSIZE
-               tp = bread(dev, fsbtodb(fs, fs->fs_csaddr + i), size);
-#endif SECSIZE
-               if (tp->b_flags&B_ERROR) {
-                       free(base, M_SUPERBLK);
+               error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
+                       NOCRED, &bp);
+               if (error) {
+                       free(base, M_UFSMNT);
                        goto out;
                }
                        goto out;
                }
-               bcopy((caddr_t)tp->b_un.b_addr, space, (u_int)size);
+               bcopy((caddr_t)bp->b_un.b_addr, space, (u_int)size);
                fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
                space += size;
                fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
                space += size;
-               brelse(tp);
-               tp = NULL;
+               brelse(bp);
+               bp = NULL;
        }
        }
-       mp->m_inodp = ip;
-       if (ip) {
-               ip->i_flag |= IMOUNT;
-               cacheinval(ip);
-               iunlock(ip);
+       mp->mnt_data = (qaddr_t)ump;
+       mp->mnt_stat.f_fsid.val[0] = (long)dev;
+       mp->mnt_stat.f_fsid.val[1] = MOUNT_UFS;
+       mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
+       mp->mnt_flag |= MNT_LOCAL;
+       ump->um_mountp = mp;
+       ump->um_dev = dev;
+       ump->um_devvp = devvp;
+       ump->um_nindir = fs->fs_nindir;
+       ump->um_bptrtodb = fs->fs_fsbtodb;
+       ump->um_seqinc = fs->fs_frag;
+       for (i = 0; i < MAXQUOTAS; i++)
+               ump->um_quotas[i] = NULLVP;
+       devvp->v_specflags |= SI_MOUNTEDON;
+       ffs_oldfscompat(fs);
+       return (0);
+out:
+       if (bp)
+               brelse(bp);
+       (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
+       if (ump) {
+               free(ump->um_fs, M_UFSMNT);
+               free(ump, M_UFSMNT);
+               mp->mnt_data = (qaddr_t)0;
        }
        }
-       /* Sanity checks for old file systems.                     XXX */
-       fs->fs_npsect = MAX(fs->fs_npsect, fs->fs_nsect);       /* XXX */
-       fs->fs_interleave = MAX(fs->fs_interleave, 1);          /* XXX */
+       return (error);
+}
+
+/*
+ * Sanity checks for old file systems.
+ *
+ * XXX - goes away some day.
+ */
+ffs_oldfscompat(fs)
+       struct fs *fs;
+{
+       int i;
 
 
+       fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);       /* XXX */
+       fs->fs_interleave = max(fs->fs_interleave, 1);          /* XXX */
+       if (fs->fs_postblformat == FS_42POSTBLFMT)              /* XXX */
+               fs->fs_nrpos = 8;                               /* XXX */
+       if (fs->fs_inodefmt < FS_44INODEFMT) {                  /* XXX */
+               quad_t sizepb = fs->fs_bsize;                   /* XXX */
+                                                               /* XXX */
+               fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1; /* XXX */
+               for (i = 0; i < NIADDR; i++) {                  /* XXX */
+                       sizepb *= NINDIR(fs);                   /* XXX */
+                       fs->fs_maxfilesize += sizepb;           /* XXX */
+               }                                               /* XXX */
+               fs->fs_qbmask = ~fs->fs_bmask;                  /* XXX */
+               fs->fs_qfmask = ~fs->fs_fmask;                  /* XXX */
+       }                                                       /* XXX */
 
 
-       return (fs);
-out:
-       if (needclose)
-               (void) closei(dev, IFBLK, ronly? FREAD : FREAD|FWRITE);
-       if (mp->m_fs) {
-               free((caddr_t)mp->m_fs, M_SUPERBLK);
-               mp->m_fs = NULL;
+       return (0);
+}
+
+/*
+ * unmount system call
+ */
+int
+ffs_unmount(mp, mntflags, p)
+       struct mount *mp;
+       int mntflags;
+       struct proc *p;
+{
+       register struct ufsmount *ump;
+       register struct fs *fs;
+       int error, flags, ronly;
+
+       flags = 0;
+       if (mntflags & MNT_FORCE) {
+               if (mp == rootfs)
+                       return (EINVAL);
+               flags |= FORCECLOSE;
        }
        }
-       if (tp)
-               brelse(tp);
-       u.u_error = error ? error : EIO;                        /* XXX */
-       return ((struct fs *) NULL);
+       if (error = ffs_flushfiles(mp, flags, p))
+               return (error);
+       ump = VFSTOUFS(mp);
+       fs = ump->um_fs;
+       ronly = !fs->fs_ronly;
+       ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
+       error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE,
+               NOCRED, p);
+       vrele(ump->um_devvp);
+       free(fs->fs_csp[0], M_UFSMNT);
+       free(fs, M_UFSMNT);
+       free(ump, M_UFSMNT);
+       mp->mnt_data = (qaddr_t)0;
+       mp->mnt_flag &= ~MNT_LOCAL;
+       return (error);
 }
 
 }
 
-umount()
+/*
+ * Flush out all the files in a filesystem.
+ */
+ffs_flushfiles(mp, flags, p)
+       register struct mount *mp;
+       int flags;
+       struct proc *p;
 {
 {
-       struct a {
-               char    *fspec;
-       } *uap = (struct a *)u.u_ap;
+       extern int doforce;
+       register struct ufsmount *ump;
+       int i, error;
 
 
-       u.u_error = unmount1(uap->fspec, 0);
+       if (!doforce)
+               flags &= ~FORCECLOSE;
+       ump = VFSTOUFS(mp);
+               return (error);
+#ifdef QUOTA
+       if (mp->mnt_flag & MNT_QUOTA) {
+               if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags))
+                       return (error);
+               for (i = 0; i < MAXQUOTAS; i++) {
+                       if (ump->um_quotas[i] == NULLVP)
+                               continue;
+                       quotaoff(p, mp, i);
+               }
+               /*
+                * Here we fall through to vflush again to ensure
+                * that we have gotten rid of all the system vnodes.
+                */
+       }
+#endif
+ * Get file system statistics.
+ */
+int
+ffs_statfs(mp, sbp, p)
+       struct mount *mp;
+       register struct statfs *sbp;
+       struct proc *p;
+{
+       register struct ufsmount *ump;
+       register struct fs *fs;
+
+       ump = VFSTOUFS(mp);
+       fs = ump->um_fs;
+       if (fs->fs_magic != FS_MAGIC)
+               panic("ffs_statfs");
+       sbp->f_type = MOUNT_UFS;
+       sbp->f_bsize = fs->fs_fsize;
+       sbp->f_iosize = fs->fs_bsize;
+       sbp->f_blocks = fs->fs_dsize;
+       sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
+               fs->fs_cstotal.cs_nffree;
+       sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
+               (fs->fs_dsize - sbp->f_bfree);
+       sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
+       sbp->f_ffree = fs->fs_cstotal.cs_nifree;
+       if (sbp != &mp->mnt_stat) {
+               bcopy((caddr_t)mp->mnt_stat.f_mntonname,
+                       (caddr_t)&sbp->f_mntonname[0], MNAMELEN);
+               bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
+                       (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
+       }
+       return (0);
 }
 
 }
 
-unmount1(fname, forcibly)
-       caddr_t fname;
-       int forcibly;
+/*
+ * Go through the disk queues to initiate sandbagged IO;
+ * go through the inodes to write those that have been modified;
+ * initiate the writing of the super block if it has been modified.
+ *
+ * Note: we are always called with the filesystem marked `MPBUSY'.
+ */
+int
+ffs_sync(mp, waitfor, cred, p)
+       struct mount *mp;
+       int waitfor;
+       struct ucred *cred;
+       struct proc *p;
 {
 {
-       dev_t dev;
-       register struct mount *mp;
-       int error;
+       register struct vnode *vp;
        register struct inode *ip;
        register struct inode *ip;
+       register struct ufsmount *ump = VFSTOUFS(mp);
        register struct fs *fs;
        register struct fs *fs;
+       int error, allerror = 0;
 
 
-       forcibly = 0;                                   /* XXX */
-       forcibly = 0;                                   /* XXX */
-       error = getmdev(&dev, fname);
-       if (error)
-               return (error);
-       for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++)
-               if (mp->m_fs != NULL && dev == mp->m_dev)
-                       goto found;
-       return (EINVAL);
-found:
-       xumount(dev);   /* remove unused sticky files from text table */
-       nchinval(dev);  /* flush the name cache */
-       update();
+       fs = ump->um_fs;
+       /*
+        * Write back modified superblock.
+        * Consistency check that the superblock
+        * is still in the buffer cache.
+        */
+       if (fs->fs_fmod != 0) {
+               if (fs->fs_ronly != 0) {                /* XXX */
+                       printf("fs = %s\n", fs->fs_fsmnt);
+                       panic("update: rofs mod");
+               }
+               fs->fs_fmod = 0;
+               fs->fs_time = time.tv_sec;
+               allerror = ffs_sbupdate(ump, waitfor);
+       }
+       /*
+        * Write back each (modified) inode.
+        */
+loop:
+       for (vp = mp->mnt_mounth; vp; vp = vp->v_mountf) {
+               /*
+                * If the vnode that we are about to sync is no longer
+                * associated with this mount point, start over.
+                */
+               if (vp->v_mount != mp)
+                       goto loop;
+               if (VOP_ISLOCKED(vp))
+                       continue;
+               ip = VTOI(vp);
+               if ((ip->i_flag & (IMOD|IACC|IUPD|ICHG)) == 0 &&
+                   vp->v_dirtyblkhd.le_next == NULL)
+                       continue;
+               if (vget(vp))
+                       goto loop;
+               if (error = VOP_FSYNC(vp, cred, waitfor, p))
+                       allerror = error;
+               vput(vp);
+       }
+       /*
+        * Force stale file system control information to be flushed.
+        */
+       if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p))
+               allerror = error;
 #ifdef QUOTA
 #ifdef QUOTA
-       if ((error = iflush(dev, mp->m_qinod)) && !forcibly)
-#else
-       if ((error = iflush(dev)) && !forcibly)
+       qsync(mp);
 #endif
 #endif
+       return (allerror);
+}
+
+/*
+ * Look up a FFS dinode number to find its incore vnode.
+ * If it is not in core, read it in from the specified device.
+ * If it is in core, wait for the lock bit to clear, then
+ * return the inode locked. Detection and handling of mount
+ * points must be done by the calling routine.
+ */
+int
+ffs_vget(mp, ino, vpp)
+       struct mount *mp;
+       ino_t ino;
+       struct vnode **vpp;
+{
+       register struct fs *fs;
+       register struct inode *ip;
+       struct ufsmount *ump;
+       struct buf *bp;
+       struct dinode *dp;
+       struct vnode *vp;
+       union ihead *ih;
+       dev_t dev;
+       int i, type, error;
+
+       ump = VFSTOUFS(mp);
+       dev = ump->um_dev;
+       if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
+               return (0);
+
+       /* Allocate a new vnode/inode. */
+       if (error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) {
+               *vpp = NULL;
                return (error);
                return (error);
+       }
+       type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
+       MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
+       bzero((caddr_t)ip, sizeof(struct inode));
+       vp->v_data = ip;
+       ip->i_vnode = vp;
+       ip->i_fs = fs = ump->um_fs;
+       ip->i_dev = dev;
+       ip->i_number = ino;
 #ifdef QUOTA
 #ifdef QUOTA
-       closedq(mp);
+       for (i = 0; i < MAXQUOTAS; i++)
+               ip->i_dquot[i] = NODQUOT;
+#endif
        /*
        /*
-        * Here we have to iflush again to get rid of the quota inode.
-        * A drag, but it would be ugly to cheat, & this doesn't happen often.
+        * Put it onto its hash chain and lock it so that other requests for
+        * this inode will block if they arrive while we are sleeping waiting
+        * for old data structures to be purged or for the contents of the
+        * disk portion of this inode to be read.
         */
         */
-       (void)iflush(dev, (struct inode *)NULL);
-#endif
-       ip = mp->m_inodp;
-       ip->i_flag &= ~IMOUNT;
-       fs = mp->m_fs;
-       free((caddr_t)fs->fs_csp[0], M_SUPERBLK);
-       free((caddr_t)mp->m_fs, M_SUPERBLK);
-       mp->m_fs = NULL;
-       mp->m_dev = NODEV;
-       mpurge(mp - &mount[0]);
-       error = closei(dev, IFBLK, fs->fs_ronly? FREAD : FREAD|FWRITE);
-       irele(ip);
-       return (error);
+       ufs_ihashins(ip);
+
+       /* Read in the disk contents for the inode, copy into the inode. */
+       if (error = bread(ump->um_devvp, fsbtodb(fs, itod(fs, ino)),
+           (int)fs->fs_bsize, NOCRED, &bp)) {
+               /*
+                * The inode does not contain anything useful, so it would
+                * be misleading to leave it on its hash chain. With mode
+                * still zero, it will be unlinked and returned to the free
+                * list by vput().
+                */
+               vput(vp);
+               brelse(bp);
+               *vpp = NULL;
+               return (error);
+       }
+       dp = bp->b_un.b_dino;
+       dp += itoo(fs, ino);
+       ip->i_din = *dp;
+       brelse(bp);
+
+       /*
+        * Initialize the vnode from the inode, check for aliases.
+        * Note that the underlying vnode may have changed.
+        */
+       if (error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp)) {
+               vput(vp);
+               *vpp = NULL;
+               return (error);
+       }
+       /*
+        * Finish inode initialization now that aliasing has been resolved.
+        */
+       ip->i_devvp = ump->um_devvp;
+       VREF(ip->i_devvp);
+       /*
+        * Set up a generation number for this inode if it does not
+        * already have one. This should only happen on old filesystems.
+        */
+       if (ip->i_gen == 0) {
+               if (++nextgennumber < (u_long)time.tv_sec)
+                       nextgennumber = time.tv_sec;
+               ip->i_gen = nextgennumber;
+               if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
+                       ip->i_flag |= IMOD;
+       }
+       /*
+        * Ensure that uid and gid are correct. This is a temporary
+        * fix until fsck has been changed to do the update.
+        */
+       if (fs->fs_inodefmt < FS_44INODEFMT) {          /* XXX */
+               ip->i_uid = ip->i_din.di_ouid;          /* XXX */
+               ip->i_gid = ip->i_din.di_ogid;          /* XXX */
+       }                                               /* XXX */
+
+       *vpp = vp;
+       return (0);
 }
 
 }
 
-sbupdate(mp)
-       struct mount *mp;
+/*
+ * File handle to vnode
+ *
+ * Have to be really careful about stale file handles:
+ * - check that the inode number is valid
+ * - call ffs_vget() to get the locked inode
+ * - check for an unallocated inode (i_mode == 0)
+ * - check that the given client host has export rights and return
+ *   those rights via. exflagsp and credanonp
+ */
+int
+ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
+       register struct mount *mp;
+       struct fid *fhp;
+       struct mbuf *nam;
+       struct vnode **vpp;
+       int *exflagsp;
+       struct ucred **credanonp;
+{
+       register struct ufid *ufhp;
+       struct fs *fs;
+
+       ufhp = (struct ufid *)fhp;
+       fs = VFSTOUFS(mp)->um_fs;
+       if (ufhp->ufid_ino < ROOTINO ||
+           ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
+               return (ESTALE);
+       return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
+}
+
+/*
+ * Vnode pointer to File handle
+ */
+/* ARGSUSED */
+ffs_vptofh(vp, fhp)
+       struct vnode *vp;
+       struct fid *fhp;
+{
+       register struct inode *ip;
+       register struct ufid *ufhp;
+
+       ip = VTOI(vp);
+       ufhp = (struct ufid *)fhp;
+       ufhp->ufid_len = sizeof(struct ufid);
+       ufhp->ufid_ino = ip->i_number;
+       ufhp->ufid_gen = ip->i_gen;
+       return (0);
+}
+
+/*
+ * Write a superblock and associated information back to disk.
+ */
+int
+ffs_sbupdate(mp, waitfor)
+       struct ufsmount *mp;
+       int waitfor;
 {
 {
-       register struct fs *fs = mp->m_fs;
+       register struct fs *fs = mp->um_fs;
        register struct buf *bp;
        int blks;
        caddr_t space;
        register struct buf *bp;
        int blks;
        caddr_t space;
-       int i, size;
+       int i, size, error = 0;
 
 #ifdef SECSIZE
        bp = getblk(mp->m_dev, (daddr_t)fsbtodb(fs, SBOFF / fs->fs_fsize),
            (int)fs->fs_sbsize, fs->fs_dbsize);
 #else SECSIZE
 
 #ifdef SECSIZE
        bp = getblk(mp->m_dev, (daddr_t)fsbtodb(fs, SBOFF / fs->fs_fsize),
            (int)fs->fs_sbsize, fs->fs_dbsize);
 #else SECSIZE
-       bp = getblk(mp->m_dev, SBLOCK, (int)fs->fs_sbsize);
+       bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
 #endif SECSIZE
        bcopy((caddr_t)fs, bp->b_un.b_addr, (u_int)fs->fs_sbsize);
 #endif SECSIZE
        bcopy((caddr_t)fs, bp->b_un.b_addr, (u_int)fs->fs_sbsize);
+       /* Restore compatibility to old file systems.              XXX */
+       if (fs->fs_postblformat == FS_42POSTBLFMT)              /* XXX */
+               bp->b_un.b_fs->fs_nrpos = -1;                   /* XXX */
 #ifdef SECSIZE
 #ifdef tahoe
        /* restore standard fsbtodb shift */
 #ifdef SECSIZE
 #ifdef tahoe
        /* restore standard fsbtodb shift */
@@ -321,7 +842,10 @@ sbupdate(mp)
        bp->b_un.b_fs->fs_sparecon[0] = 0;
 #endif
 #endif SECSIZE
        bp->b_un.b_fs->fs_sparecon[0] = 0;
 #endif
 #endif SECSIZE
-       bwrite(bp);
+       if (waitfor == MNT_WAIT)
+               error = bwrite(bp);
+       else
+               bawrite(bp);
        blks = howmany(fs->fs_cssize, fs->fs_fsize);
        space = (caddr_t)fs->fs_csp[0];
        for (i = 0; i < blks; i += fs->fs_frag) {
        blks = howmany(fs->fs_cssize, fs->fs_fsize);
        space = (caddr_t)fs->fs_csp[0];
        for (i = 0; i < blks; i += fs->fs_frag) {
@@ -332,46 +856,15 @@ sbupdate(mp)
                bp = getblk(mp->m_dev, fsbtodb(fs, fs->fs_csaddr + i), size,
                    fs->fs_dbsize);
 #else SECSIZE
                bp = getblk(mp->m_dev, fsbtodb(fs, fs->fs_csaddr + i), size,
                    fs->fs_dbsize);
 #else SECSIZE
-               bp = getblk(mp->m_dev, fsbtodb(fs, fs->fs_csaddr + i), size);
+               bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
+                   size, 0, 0);
 #endif SECSIZE
                bcopy(space, bp->b_un.b_addr, (u_int)size);
                space += size;
 #endif SECSIZE
                bcopy(space, bp->b_un.b_addr, (u_int)size);
                space += size;
-               bwrite(bp);
+               if (waitfor == MNT_WAIT)
+                       error = bwrite(bp);
+               else
+                       bawrite(bp);
        }
        }
-}
-
-/*
- * Common code for mount and umount.
- * Check that the user's argument is a reasonable
- * thing on which to mount, and return the device number if so.
- */
-getmdev(pdev, fname)
-       caddr_t fname;
-       dev_t *pdev;
-{
-       dev_t dev;
-       register struct inode *ip;
-       register struct nameidata *ndp = &u.u_nd;
-
-       if (!suser())
-               return (u.u_error);
-       ndp->ni_nameiop = LOOKUP | FOLLOW;
-       ndp->ni_segflg = UIO_USERSPACE;
-       ndp->ni_dirp = fname;
-       ip = namei(ndp);
-       if (ip == NULL) {
-               if (u.u_error == ENOENT)
-                       return (ENODEV); /* needs translation */
-               return (u.u_error);
-       }
-       if ((ip->i_mode&IFMT) != IFBLK) {
-               iput(ip);
-               return (ENOTBLK);
-       }
-       dev = (dev_t)ip->i_rdev;
-       iput(ip);
-       if (major(dev) >= nblkdev)
-               return (ENXIO);
-       *pdev = dev;
-       return (0);
+       return (error);
 }
 }