lint; checkoverlap code is conditional on DIAGNOSTIC
[unix-history] / usr / src / sys / ufs / ffs / ffs_vnops.c
index b7509eb..9c9015f 100644 (file)
@@ -14,7 +14,7 @@
  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
  *
  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
  *
- *     @(#)ffs_vnops.c 7.16 (Berkeley) %G%
+ *     @(#)ffs_vnops.c 7.39 (Berkeley) %G%
  */
 
 #include "param.h"
  */
 
 #include "param.h"
 #include "stat.h"
 #include "buf.h"
 #include "proc.h"
 #include "stat.h"
 #include "buf.h"
 #include "proc.h"
-#include "uio.h"
 #include "socket.h"
 #include "socketvar.h"
 #include "conf.h"
 #include "mount.h"
 #include "vnode.h"
 #include "socket.h"
 #include "socketvar.h"
 #include "conf.h"
 #include "mount.h"
 #include "vnode.h"
+#include "specdev.h"
+#include "../ufs/quota.h"
 #include "../ufs/inode.h"
 #include "../ufs/fs.h"
 #include "../ufs/inode.h"
 #include "../ufs/fs.h"
-#include "../ufs/quota.h"
 
 /*
  * Global vfs data structures for ufs
 
 /*
  * Global vfs data structures for ufs
@@ -64,48 +64,151 @@ int        ufs_lookup(),
        ufs_readlink(),
        ufs_abortop(),
        ufs_inactive(),
        ufs_readlink(),
        ufs_abortop(),
        ufs_inactive(),
+       ufs_reclaim(),
        ufs_lock(),
        ufs_unlock(),
        ufs_bmap(),
        ufs_lock(),
        ufs_unlock(),
        ufs_bmap(),
-       ufs_strategy();
+       ufs_strategy(),
+       ufs_print(),
+       ufs_islocked();
 
 struct vnodeops ufs_vnodeops = {
 
 struct vnodeops ufs_vnodeops = {
-       ufs_lookup,
-       ufs_create,
-       ufs_mknod,
-       ufs_open,
-       ufs_close,
-       ufs_access,
-       ufs_getattr,
-       ufs_setattr,
-       ufs_read,
-       ufs_write,
-       ufs_ioctl,
-       ufs_select,
-       ufs_mmap,
-       ufs_fsync,
-       ufs_seek,
-       ufs_remove,
-       ufs_link,
-       ufs_rename,
-       ufs_mkdir,
-       ufs_rmdir,
-       ufs_symlink,
-       ufs_readdir,
-       ufs_readlink,
-       ufs_abortop,
-       ufs_inactive,
-       ufs_lock,
-       ufs_unlock,
-       ufs_bmap,
-       ufs_strategy,
+       ufs_lookup,             /* lookup */
+       ufs_create,             /* create */
+       ufs_mknod,              /* mknod */
+       ufs_open,               /* open */
+       ufs_close,              /* close */
+       ufs_access,             /* access */
+       ufs_getattr,            /* getattr */
+       ufs_setattr,            /* setattr */
+       ufs_read,               /* read */
+       ufs_write,              /* write */
+       ufs_ioctl,              /* ioctl */
+       ufs_select,             /* select */
+       ufs_mmap,               /* mmap */
+       ufs_fsync,              /* fsync */
+       ufs_seek,               /* seek */
+       ufs_remove,             /* remove */
+       ufs_link,               /* link */
+       ufs_rename,             /* rename */
+       ufs_mkdir,              /* mkdir */
+       ufs_rmdir,              /* rmdir */
+       ufs_symlink,            /* symlink */
+       ufs_readdir,            /* readdir */
+       ufs_readlink,           /* readlink */
+       ufs_abortop,            /* abortop */
+       ufs_inactive,           /* inactive */
+       ufs_reclaim,            /* reclaim */
+       ufs_lock,               /* lock */
+       ufs_unlock,             /* unlock */
+       ufs_bmap,               /* bmap */
+       ufs_strategy,           /* strategy */
+       ufs_print,              /* print */
+       ufs_islocked,           /* islocked */
+};
+
+int    spec_lookup(),
+       spec_open(),
+       ufsspec_read(),
+       ufsspec_write(),
+       spec_strategy(),
+       spec_bmap(),
+       spec_ioctl(),
+       spec_select(),
+       ufsspec_close(),
+       spec_badop(),
+       spec_nullop();
+
+struct vnodeops spec_inodeops = {
+       spec_lookup,            /* lookup */
+       spec_badop,             /* create */
+       spec_badop,             /* mknod */
+       spec_open,              /* open */
+       ufsspec_close,          /* close */
+       ufs_access,             /* access */
+       ufs_getattr,            /* getattr */
+       ufs_setattr,            /* setattr */
+       ufsspec_read,           /* read */
+       ufsspec_write,          /* write */
+       spec_ioctl,             /* ioctl */
+       spec_select,            /* select */
+       spec_badop,             /* mmap */
+       spec_nullop,            /* fsync */
+       spec_badop,             /* seek */
+       spec_badop,             /* remove */
+       spec_badop,             /* link */
+       spec_badop,             /* rename */
+       spec_badop,             /* mkdir */
+       spec_badop,             /* rmdir */
+       spec_badop,             /* symlink */
+       spec_badop,             /* readdir */
+       spec_badop,             /* readlink */
+       spec_badop,             /* abortop */
+       ufs_inactive,           /* inactive */
+       ufs_reclaim,            /* reclaim */
+       ufs_lock,               /* lock */
+       ufs_unlock,             /* unlock */
+       spec_bmap,              /* bmap */
+       spec_strategy,          /* strategy */
+       ufs_print,              /* print */
+       ufs_islocked,           /* islocked */
 };
 
 };
 
-enum vtype iftovt_tab[8] = {
-       VNON, VCHR, VDIR, VBLK, VREG, VLNK, VSOCK, VBAD,
+#ifdef FIFO
+int    fifo_lookup(),
+       fifo_open(),
+       ufsfifo_read(),
+       ufsfifo_write(),
+       fifo_bmap(),
+       fifo_ioctl(),
+       fifo_select(),
+       ufsfifo_close(),
+       fifo_print(),
+       fifo_badop(),
+       fifo_nullop();
+
+struct vnodeops fifo_inodeops = {
+       fifo_lookup,            /* lookup */
+       fifo_badop,             /* create */
+       fifo_badop,             /* mknod */
+       fifo_open,              /* open */
+       ufsfifo_close,          /* close */
+       ufs_access,             /* access */
+       ufs_getattr,            /* getattr */
+       ufs_setattr,            /* setattr */
+       ufsfifo_read,           /* read */
+       ufsfifo_write,          /* write */
+       fifo_ioctl,             /* ioctl */
+       fifo_select,            /* select */
+       fifo_badop,             /* mmap */
+       fifo_nullop,            /* fsync */
+       fifo_badop,             /* seek */
+       fifo_badop,             /* remove */
+       fifo_badop,             /* link */
+       fifo_badop,             /* rename */
+       fifo_badop,             /* mkdir */
+       fifo_badop,             /* rmdir */
+       fifo_badop,             /* symlink */
+       fifo_badop,             /* readdir */
+       fifo_badop,             /* readlink */
+       fifo_badop,             /* abortop */
+       ufs_inactive,           /* inactive */
+       ufs_reclaim,            /* reclaim */
+       ufs_lock,               /* lock */
+       ufs_unlock,             /* unlock */
+       fifo_bmap,              /* bmap */
+       fifo_badop,             /* strategy */
+       ufs_print,              /* print */
+       ufs_islocked,           /* islocked */
 };
 };
-int    vttoif_tab[8] = {
-       0, IFREG, IFDIR, IFBLK, IFCHR, IFLNK, IFSOCK, IFMT,
+#endif /* FIFO */
+
+enum vtype iftovt_tab[16] = {
+       VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON,
+       VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD,
+};
+int    vttoif_tab[9] = {
+       0, IFREG, IFDIR, IFBLK, IFCHR, IFLNK, IFSOCK, IFIFO, IFMT,
 };
 
 /*
 };
 
 /*
@@ -133,28 +236,29 @@ ufs_mknod(ndp, vap, cred)
        struct ucred *cred;
        struct vattr *vap;
 {
        struct ucred *cred;
        struct vattr *vap;
 {
+       register struct vnode *vp;
        struct inode *ip;
        int error;
 
        if (error = maknode(MAKEIMODE(vap->va_type, vap->va_mode), ndp, &ip))
                return (error);
        struct inode *ip;
        int error;
 
        if (error = maknode(MAKEIMODE(vap->va_type, vap->va_mode), ndp, &ip))
                return (error);
-       if (vap->va_rdev) {
+       ip->i_flag |= IACC|IUPD|ICHG;
+       if (vap->va_rdev != VNOVAL) {
                /*
                 * Want to be able to use this to make badblock
                 * inodes, so don't truncate the dev number.
                 */
                /*
                 * Want to be able to use this to make badblock
                 * inodes, so don't truncate the dev number.
                 */
-               ITOV(ip)->v_rdev = ip->i_rdev = vap->va_rdev;
-               ip->i_flag |= IACC|IUPD|ICHG;
+               ip->i_rdev = vap->va_rdev;
        }
        }
-       iput(ip);
        /*
         * Remove inode so that it will be reloaded by iget and
         * checked to see if it is an alias of an existing entry
         * in the inode cache.
         */
        /*
         * Remove inode so that it will be reloaded by iget and
         * checked to see if it is an alias of an existing entry
         * in the inode cache.
         */
-       remque(ip);
-       ip->i_forw = ip;
-       ip->i_back = ip;
+       vp = ITOV(ip);
+       vput(vp);
+       vp->v_type = VNON;
+       vgone(vp);
        return (0);
 }
 
        return (0);
 }
 
@@ -186,18 +290,63 @@ ufs_close(vp, fflag, cred)
 {
        register struct inode *ip = VTOI(vp);
 
 {
        register struct inode *ip = VTOI(vp);
 
-       if (vp->v_count > 1 && !(ip->i_flag & ILOCKED))
+       if (vp->v_usecount > 1 && !(ip->i_flag & ILOCKED))
                ITIMES(ip, &time, &time);
        return (0);
 }
 
                ITIMES(ip, &time, &time);
        return (0);
 }
 
+/*
+ * Check mode permission on inode pointer. Mode is READ, WRITE or EXEC.
+ * The mode is shifted to select the owner/group/other fields. The
+ * super user is granted all permissions.
+ */
 ufs_access(vp, mode, cred)
        struct vnode *vp;
 ufs_access(vp, mode, cred)
        struct vnode *vp;
-       int mode;
+       register int mode;
        struct ucred *cred;
 {
        struct ucred *cred;
 {
+       register struct inode *ip = VTOI(vp);
+       register gid_t *gp;
+       int i, error;
 
 
-       return (iaccess(VTOI(vp), mode, cred));
+#ifdef DIAGNOSTIC
+       if (!VOP_ISLOCKED(vp)) {
+               vprint("ufs_access: not locked", vp);
+               panic("ufs_access: not locked");
+       }
+#endif
+#ifdef QUOTA
+       if (mode & VWRITE) {
+               switch (vp->v_type) {
+               case VREG: case VDIR: case VLNK:
+                       if (error = getinoquota(ip))
+                               return (error);
+               }
+       }
+#endif /* QUOTA */
+       /*
+        * If you're the super-user, you always get access.
+        */
+       if (cred->cr_uid == 0)
+               return (0);
+       /*
+        * 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 != ip->i_uid) {
+               mode >>= 3;
+               gp = cred->cr_groups;
+               for (i = 0; i < cred->cr_ngroups; i++, gp++)
+                       if (ip->i_gid == *gp)
+                               goto found;
+               mode >>= 3;
+found:
+               ;
+       }
+       if ((ip->i_mode & mode) != 0)
+               return (0);
+       return (EACCES);
 }
 
 /* ARGSUSED */
 }
 
 /* ARGSUSED */
@@ -219,8 +368,12 @@ ufs_getattr(vp, vap, cred)
        vap->va_uid = ip->i_uid;
        vap->va_gid = ip->i_gid;
        vap->va_rdev = (dev_t)ip->i_rdev;
        vap->va_uid = ip->i_uid;
        vap->va_gid = ip->i_gid;
        vap->va_rdev = (dev_t)ip->i_rdev;
-       vap->va_size = ip->i_ic.ic_size.val[0];
-       vap->va_size1 = ip->i_ic.ic_size.val[1];
+#ifdef tahoe
+       vap->va_size = ip->i_size;
+       vap->va_size_rsv = 0;
+#else
+       vap->va_qsize = ip->i_din.di_qsize;
+#endif
        vap->va_atime.tv_sec = ip->i_atime;
        vap->va_atime.tv_usec = 0;
        vap->va_mtime.tv_sec = ip->i_mtime;
        vap->va_atime.tv_sec = ip->i_atime;
        vap->va_atime.tv_usec = 0;
        vap->va_mtime.tv_sec = ip->i_mtime;
@@ -237,7 +390,7 @@ ufs_getattr(vp, vap, cred)
        else
                vap->va_blocksize = ip->i_fs->fs_bsize;
        vap->va_bytes = dbtob(ip->i_blocks);
        else
                vap->va_blocksize = ip->i_fs->fs_bsize;
        vap->va_bytes = dbtob(ip->i_blocks);
-       vap->va_bytes1 = -1;
+       vap->va_bytes_rsv = 0;
        vap->va_type = vp->v_type;
        return (0);
 }
        vap->va_type = vp->v_type;
        return (0);
 }
@@ -271,7 +424,7 @@ ufs_setattr(vp, vap, cred)
        if (vap->va_size != VNOVAL) {
                if (vp->v_type == VDIR)
                        return (EISDIR);
        if (vap->va_size != VNOVAL) {
                if (vp->v_type == VDIR)
                        return (EISDIR);
-               if (error = itrunc(ip, vap->va_size))
+               if (error = itrunc(ip, vap->va_size, 0)) /* XXX IO_SYNC? */
                        return (error);
        }
        if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
                        return (error);
        }
        if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) {
@@ -343,10 +496,13 @@ chown1(vp, uid, gid, cred)
        struct ucred *cred;
 {
        register struct inode *ip = VTOI(vp);
        struct ucred *cred;
 {
        register struct inode *ip = VTOI(vp);
+       uid_t ouid;
+       gid_t ogid;
+       int error = 0;
 #ifdef QUOTA
 #ifdef QUOTA
-       register long change;
+       register int i;
+       long change;
 #endif
 #endif
-       int error;
 
        if (uid == (u_short)VNOVAL)
                uid = ip->i_uid;
 
        if (uid == (u_short)VNOVAL)
                uid = ip->i_uid;
@@ -361,27 +517,226 @@ chown1(vp, uid, gid, cred)
            !groupmember((gid_t)gid, cred)) &&
            (error = suser(cred, &u.u_acflag)))
                return (error);
            !groupmember((gid_t)gid, cred)) &&
            (error = suser(cred, &u.u_acflag)))
                return (error);
+       ouid = ip->i_uid;
+       ogid = ip->i_gid;
 #ifdef QUOTA
 #ifdef QUOTA
-       if (ip->i_uid == uid)           /* this just speeds things a little */
-               change = 0;
-       else
-               change = ip->i_blocks;
-       (void) chkdq(ip, -change, 1);
-       (void) chkiq(ip->i_dev, ip, ip->i_uid, 1);
-       dqrele(ip->i_dquot);
+       if (error = getinoquota(ip))
+               return (error);
+       if (ouid == uid) {
+               dqrele(vp, ip->i_dquot[USRQUOTA]);
+               ip->i_dquot[USRQUOTA] = NODQUOT;
+       }
+       if (ogid == gid) {
+               dqrele(vp, ip->i_dquot[GRPQUOTA]);
+               ip->i_dquot[GRPQUOTA] = NODQUOT;
+       }
+       change = ip->i_blocks;
+       (void) chkdq(ip, -change, cred, CHOWN);
+       (void) chkiq(ip, -1, cred, CHOWN);
+       for (i = 0; i < MAXQUOTAS; i++) {
+               dqrele(vp, ip->i_dquot[i]);
+               ip->i_dquot[i] = NODQUOT;
+       }
 #endif
        ip->i_uid = uid;
        ip->i_gid = gid;
 #endif
        ip->i_uid = uid;
        ip->i_gid = gid;
-       ip->i_flag |= ICHG;
-       ip->i_mode &= ~(ISUID|ISGID);
 #ifdef QUOTA
 #ifdef QUOTA
-       ip->i_dquot = inoquota(ip);
-       (void) chkdq(ip, change, 1);
-       (void) chkiq(ip->i_dev, (struct inode *)NULL, (uid_t)uid, 1);
-       return (u.u_error);             /* should == 0 ALWAYS !! */
-#else
-       return (0);
+       if ((error = getinoquota(ip)) == 0) {
+               if (ouid == uid) {
+                       dqrele(vp, ip->i_dquot[USRQUOTA]);
+                       ip->i_dquot[USRQUOTA] = NODQUOT;
+               }
+               if (ogid == gid) {
+                       dqrele(vp, ip->i_dquot[GRPQUOTA]);
+                       ip->i_dquot[GRPQUOTA] = NODQUOT;
+               }
+               if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
+                       if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
+                               return (0);
+                       else
+                               (void) chkdq(ip, -change, cred, CHOWN|FORCE);
+               }
+               for (i = 0; i < MAXQUOTAS; i++) {
+                       dqrele(vp, ip->i_dquot[i]);
+                       ip->i_dquot[i] = NODQUOT;
+               }
+       }
+       ip->i_uid = ouid;
+       ip->i_gid = ogid;
+       if (getinoquota(ip) == 0) {
+               if (ouid == uid) {
+                       dqrele(vp, ip->i_dquot[USRQUOTA]);
+                       ip->i_dquot[USRQUOTA] = NODQUOT;
+               }
+               if (ogid == gid) {
+                       dqrele(vp, ip->i_dquot[GRPQUOTA]);
+                       ip->i_dquot[GRPQUOTA] = NODQUOT;
+               }
+               (void) chkdq(ip, change, cred, FORCE);
+               (void) chkiq(ip, 1, cred, FORCE);
+       }
+       if (error)
+               return (error);
 #endif
 #endif
+       if (ouid != uid || ogid != gid)
+               ip->i_flag |= ICHG;
+       if (ouid != uid && cred->cr_uid != 0)
+               ip->i_mode &= ~ISUID;
+       if (ogid != gid && cred->cr_uid != 0)
+               ip->i_mode &= ~ISGID;
+       return (0);
+}
+
+/*
+ * Vnode op for reading.
+ */
+/* ARGSUSED */
+ufs_read(vp, uio, ioflag, cred)
+       struct vnode *vp;
+       register struct uio *uio;
+       int ioflag;
+       struct ucred *cred;
+{
+       register struct inode *ip = VTOI(vp);
+       register struct fs *fs;
+       struct buf *bp;
+       daddr_t lbn, bn, rablock;
+       int size, diff, error = 0;
+       long n, on, type;
+
+       if (uio->uio_rw != UIO_READ)
+               panic("ufs_read mode");
+       type = ip->i_mode & IFMT;
+       if (type != IFDIR && type != IFREG && type != IFLNK)
+               panic("ufs_read type");
+       if (uio->uio_resid == 0)
+               return (0);
+       if (uio->uio_offset < 0)
+               return (EINVAL);
+       ip->i_flag |= IACC;
+       fs = ip->i_fs;
+       do {
+               lbn = lblkno(fs, uio->uio_offset);
+               on = blkoff(fs, uio->uio_offset);
+               n = MIN((unsigned)(fs->fs_bsize - on), uio->uio_resid);
+               diff = ip->i_size - uio->uio_offset;
+               if (diff <= 0)
+                       return (0);
+               if (diff < n)
+                       n = diff;
+               size = blksize(fs, ip, lbn);
+               rablock = lbn + 1;
+               if (vp->v_lastr + 1 == lbn &&
+                   lblktosize(fs, rablock) < ip->i_size)
+                       error = breada(ITOV(ip), lbn, size, rablock,
+                               blksize(fs, ip, rablock), NOCRED, &bp);
+               else
+                       error = bread(ITOV(ip), lbn, size, NOCRED, &bp);
+               vp->v_lastr = lbn;
+               n = MIN(n, size - bp->b_resid);
+               if (error) {
+                       brelse(bp);
+                       return (error);
+               }
+               error = uiomove(bp->b_un.b_addr + on, (int)n, uio);
+               if (n + on == fs->fs_bsize || uio->uio_offset == ip->i_size)
+                       bp->b_flags |= B_AGE;
+               brelse(bp);
+       } while (error == 0 && uio->uio_resid > 0 && n != 0);
+       return (error);
+}
+
+/*
+ * Vnode op for writing.
+ */
+ufs_write(vp, uio, ioflag, cred)
+       register struct vnode *vp;
+       struct uio *uio;
+       int ioflag;
+       struct ucred *cred;
+{
+       register struct inode *ip = VTOI(vp);
+       register struct fs *fs;
+       struct buf *bp;
+       daddr_t lbn, bn;
+       u_long osize;
+       int i, n, on, flags;
+       int count, size, resid, error = 0;
+
+       if (uio->uio_rw != UIO_WRITE)
+               panic("ufs_write mode");
+       switch (vp->v_type) {
+       case VREG:
+               if (ioflag & IO_APPEND)
+                       uio->uio_offset = ip->i_size;
+               /* fall through */
+       case VLNK:
+               break;
+
+       case VDIR:
+               if ((ioflag & IO_SYNC) == 0)
+                       panic("ufs_write nonsync dir write");
+               break;
+
+       default:
+               panic("ufs_write type");
+       }
+       if (uio->uio_offset < 0)
+               return (EINVAL);
+       if (uio->uio_resid == 0)
+               return (0);
+       /*
+        * Maybe this should be above the vnode op call, but so long as
+        * file servers have no limits, i don't think it matters
+        */
+       if (vp->v_type == VREG &&
+           uio->uio_offset + uio->uio_resid >
+             u.u_rlimit[RLIMIT_FSIZE].rlim_cur) {
+               psignal(u.u_procp, SIGXFSZ);
+               return (EFBIG);
+       }
+       resid = uio->uio_resid;
+       osize = ip->i_size;
+       fs = ip->i_fs;
+       flags = 0;
+       if (ioflag & IO_SYNC)
+               flags = B_SYNC;
+       do {
+               lbn = lblkno(fs, uio->uio_offset);
+               on = blkoff(fs, uio->uio_offset);
+               n = MIN((unsigned)(fs->fs_bsize - on), uio->uio_resid);
+               if (n < fs->fs_bsize)
+                       flags |= B_CLRBUF;
+               else
+                       flags &= ~B_CLRBUF;
+               if (error = balloc(ip, lbn, (int)(on + n), &bp, flags))
+                       break;
+               bn = bp->b_blkno;
+               if (uio->uio_offset + n > ip->i_size)
+                       ip->i_size = uio->uio_offset + n;
+               size = blksize(fs, ip, lbn);
+               count = howmany(size, CLBYTES);
+               for (i = 0; i < count; i++)
+                       munhash(ip->i_devvp, bn + i * CLBYTES / DEV_BSIZE);
+               n = MIN(n, size - bp->b_resid);
+               error = uiomove(bp->b_un.b_addr + on, n, uio);
+               if (ioflag & IO_SYNC)
+                       (void) bwrite(bp);
+               else if (n + on == fs->fs_bsize) {
+                       bp->b_flags |= B_AGE;
+                       bawrite(bp);
+               } else
+                       bdwrite(bp);
+               ip->i_flag |= IUPD|ICHG;
+               if (cred->cr_uid != 0)
+                       ip->i_mode &= ~(ISUID|ISGID);
+       } while (error == 0 && uio->uio_resid > 0 && n != 0);
+       if (error && (ioflag & IO_UNIT)) {
+               (void) itrunc(ip, osize, ioflag & IO_SYNC);
+               uio->uio_offset -= resid - uio->uio_resid;
+               uio->uio_resid = resid;
+       }
+       return (error);
 }
 
 /* ARGSUSED */
 }
 
 /* ARGSUSED */
@@ -393,18 +748,16 @@ ufs_ioctl(vp, com, data, fflag, cred)
        struct ucred *cred;
 {
 
        struct ucred *cred;
 {
 
-       printf("ufs_ioctl called with type %d\n", vp->v_type);
        return (ENOTTY);
 }
 
 /* ARGSUSED */
        return (ENOTTY);
 }
 
 /* ARGSUSED */
-ufs_select(vp, which, cred)
+ufs_select(vp, which, fflags, cred)
        struct vnode *vp;
        struct vnode *vp;
-       int which;
+       int which, fflags;
        struct ucred *cred;
 {
 
        struct ucred *cred;
 {
 
-       printf("ufs_select called with type %d\n", vp->v_type);
        return (1);             /* XXX */
 }
 
        return (1);             /* XXX */
 }
 
@@ -427,20 +780,18 @@ ufs_mmap(vp, fflags, cred)
  * Synch an open file.
  */
 /* ARGSUSED */
  * Synch an open file.
  */
 /* ARGSUSED */
-ufs_fsync(vp, fflags, cred)
+ufs_fsync(vp, fflags, cred, waitfor)
        struct vnode *vp;
        int fflags;
        struct ucred *cred;
        struct vnode *vp;
        int fflags;
        struct ucred *cred;
+       int waitfor;
 {
 {
-       register struct inode *ip = VTOI(vp);
-       int error;
+       struct inode *ip = VTOI(vp);
 
 
-       ILOCK(ip);
        if (fflags&FWRITE)
                ip->i_flag |= ICHG;
        if (fflags&FWRITE)
                ip->i_flag |= ICHG;
-       error = syncip(ip);
-       IUNLOCK(ip);
-       return (error);
+       vflushbuf(vp, waitfor == MNT_WAIT ? B_SYNC : 0);
+       return (iupdat(ip, &time, &time, waitfor == MNT_WAIT));
 }
 
 /*
 }
 
 /*
@@ -601,7 +952,10 @@ ufs_rename(fndp, tndp)
        if (oldparent != dp->i_number)
                newparent = dp->i_number;
        if (doingdirectory && newparent) {
        if (oldparent != dp->i_number)
                newparent = dp->i_number;
        if (doingdirectory && newparent) {
-               if (error = iaccess(ip, IWRITE, tndp->ni_cred))
+               VOP_LOCK(fndp->ni_vp);
+               error = ufs_access(fndp->ni_vp, VWRITE, tndp->ni_cred);
+               VOP_UNLOCK(fndp->ni_vp);
+               if (error)
                        goto bad;
                tndp->ni_nameiop = RENAME | LOCKPARENT | LOCKLEAF | NOCACHE;
                do {
                        goto bad;
                tndp->ni_nameiop = RENAME | LOCKPARENT | LOCKLEAF | NOCACHE;
                do {
@@ -698,7 +1052,7 @@ ufs_rename(fndp, tndp)
                if (doingdirectory) {
                        if (--xp->i_nlink != 0)
                                panic("rename: linked directory");
                if (doingdirectory) {
                        if (--xp->i_nlink != 0)
                                panic("rename: linked directory");
-                       error = itrunc(xp, (u_long)0);
+                       error = itrunc(xp, (u_long)0, IO_SYNC);
                }
                xp->i_flag |= ICHG;
                iput(xp);
                }
                xp->i_flag |= ICHG;
                iput(xp);
@@ -742,20 +1096,22 @@ ufs_rename(fndp, tndp)
                if (doingdirectory && newparent) {
                        dp->i_nlink--;
                        dp->i_flag |= ICHG;
                if (doingdirectory && newparent) {
                        dp->i_nlink--;
                        dp->i_flag |= ICHG;
-                       error = rdwri(UIO_READ, xp, (caddr_t)&dirbuf,
+                       error = vn_rdwr(UIO_READ, ITOV(xp), (caddr_t)&dirbuf,
                                sizeof (struct dirtemplate), (off_t)0,
                                sizeof (struct dirtemplate), (off_t)0,
-                               UIO_SYSSPACE, tndp->ni_cred, (int *)0);
+                               UIO_SYSSPACE, IO_NODELOCKED, 
+                               tndp->ni_cred, (int *)0);
                        if (error == 0) {
                                if (dirbuf.dotdot_namlen != 2 ||
                                    dirbuf.dotdot_name[0] != '.' ||
                                    dirbuf.dotdot_name[1] != '.') {
                        if (error == 0) {
                                if (dirbuf.dotdot_namlen != 2 ||
                                    dirbuf.dotdot_name[0] != '.' ||
                                    dirbuf.dotdot_name[1] != '.') {
-                                       printf("rename: mangled dir\n");
+                                       dirbad(xp, 12, "rename: mangled dir");
                                } else {
                                        dirbuf.dotdot_ino = newparent;
                                } else {
                                        dirbuf.dotdot_ino = newparent;
-                                       (void) rdwri(UIO_WRITE, xp,
+                                       (void) vn_rdwr(UIO_WRITE, ITOV(xp),
                                            (caddr_t)&dirbuf,
                                            sizeof (struct dirtemplate),
                                            (off_t)0, UIO_SYSSPACE,
                                            (caddr_t)&dirbuf,
                                            sizeof (struct dirtemplate),
                                            (off_t)0, UIO_SYSSPACE,
+                                           IO_NODELOCKED|IO_SYNC,
                                            tndp->ni_cred, (int *)0);
                                        cache_purge(ITOV(dp));
                                }
                                            tndp->ni_cred, (int *)0);
                                        cache_purge(ITOV(dp));
                                }
@@ -819,25 +1175,26 @@ ufs_mkdir(ndp, vap)
         * directory.  The entry is made later
         * after writing "." and ".." entries out.
         */
         * directory.  The entry is made later
         * after writing "." and ".." entries out.
         */
-       error = ialloc(dp, dirpref(dp->i_fs), dmode, &tip);
-       if (error) {
+       if (error = ialloc(dp, dirpref(dp->i_fs), dmode, ndp->ni_cred, &tip)) {
                iput(dp);
                return (error);
        }
        ip = tip;
                iput(dp);
                return (error);
        }
        ip = tip;
+       ip->i_uid = ndp->ni_cred->cr_uid;
+       ip->i_gid = dp->i_gid;
 #ifdef QUOTA
 #ifdef QUOTA
-       if (ip->i_dquot != NODQUOT)
-               panic("mkdir: dquot");
+       if ((error = getinoquota(ip)) ||
+           (error = chkiq(ip, 1, ndp->ni_cred, 0))) {
+               ifree(ip, ip->i_number, dmode);
+               iput(ip);
+               iput(dp);
+               return (error);
+       }
 #endif
        ip->i_flag |= IACC|IUPD|ICHG;
        ip->i_mode = dmode;
        ITOV(ip)->v_type = VDIR;        /* Rest init'd in iget() */
        ip->i_nlink = 2;
 #endif
        ip->i_flag |= IACC|IUPD|ICHG;
        ip->i_mode = dmode;
        ITOV(ip)->v_type = VDIR;        /* Rest init'd in iget() */
        ip->i_nlink = 2;
-       ip->i_uid = ndp->ni_cred->cr_uid;
-       ip->i_gid = dp->i_gid;
-#ifdef QUOTA
-       ip->i_dquot = inoquota(ip);
-#endif
        error = iupdat(ip, &time, &time, 1);
 
        /*
        error = iupdat(ip, &time, &time, 1);
 
        /*
@@ -857,9 +1214,9 @@ ufs_mkdir(ndp, vap)
        dirtemplate = mastertemplate;
        dirtemplate.dot_ino = ip->i_number;
        dirtemplate.dotdot_ino = dp->i_number;
        dirtemplate = mastertemplate;
        dirtemplate.dot_ino = ip->i_number;
        dirtemplate.dotdot_ino = dp->i_number;
-       error = rdwri(UIO_WRITE, ip, (caddr_t)&dirtemplate,
+       error = vn_rdwr(UIO_WRITE, ITOV(ip), (caddr_t)&dirtemplate,
                sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
                sizeof (dirtemplate), (off_t)0, UIO_SYSSPACE,
-               ndp->ni_cred, (int *)0);
+               IO_NODELOCKED|IO_SYNC, ndp->ni_cred, (int *)0);
        if (error) {
                dp->i_nlink--;
                dp->i_flag |= ICHG;
        if (error) {
                dp->i_nlink--;
                dp->i_flag |= ICHG;
@@ -956,7 +1313,7 @@ ufs_rmdir(ndp)
         * worry about them later.
         */
        ip->i_nlink -= 2;
         * worry about them later.
         */
        ip->i_nlink -= 2;
-       error = itrunc(ip, (u_long)0);
+       error = itrunc(ip, (u_long)0, IO_SYNC);
        cache_purge(ITOV(ip));
 out:
        if (ndp->ni_dvp)
        cache_purge(ITOV(ip));
 out:
        if (ndp->ni_dvp)
@@ -979,8 +1336,8 @@ ufs_symlink(ndp, vap, target)
        error = maknode(IFLNK | vap->va_mode, ndp, &ip);
        if (error)
                return (error);
        error = maknode(IFLNK | vap->va_mode, ndp, &ip);
        if (error)
                return (error);
-       error = rdwri(UIO_WRITE, ip, target, strlen(target), (off_t)0,
-               UIO_SYSSPACE, ndp->ni_cred, (int *)0);
+       error = vn_rdwr(UIO_WRITE, ITOV(ip), target, strlen(target), (off_t)0,
+               UIO_SYSSPACE, IO_NODELOCKED, ndp->ni_cred, (int *)0);
        iput(ip);
        return (error);
 }
        iput(ip);
        return (error);
 }
@@ -988,29 +1345,27 @@ ufs_symlink(ndp, vap, target)
 /*
  * Vnode op for read and write
  */
 /*
  * Vnode op for read and write
  */
-ufs_readdir(vp, uio, offp, cred)
+ufs_readdir(vp, uio, cred, eofflagp)
        struct vnode *vp;
        register struct uio *uio;
        struct vnode *vp;
        register struct uio *uio;
-       off_t *offp;
        struct ucred *cred;
        struct ucred *cred;
+       int *eofflagp;
 {
 {
-       register struct inode *ip = VTOI(vp);
-       int count, error;
+       int count, lost, error;
 
 
-       ILOCK(ip);
-       uio->uio_offset = *offp;
        count = uio->uio_resid;
        count &= ~(DIRBLKSIZ - 1);
        count = uio->uio_resid;
        count &= ~(DIRBLKSIZ - 1);
-       if (vp->v_type != VDIR || uio->uio_iovcnt != 1 ||
-           (count < DIRBLKSIZ) || (uio->uio_offset & (DIRBLKSIZ -1))) {
-               IUNLOCK(ip);
+       lost = uio->uio_resid - count;
+       if (count < DIRBLKSIZ || (uio->uio_offset & (DIRBLKSIZ -1)))
                return (EINVAL);
                return (EINVAL);
-       }
        uio->uio_resid = count;
        uio->uio_iov->iov_len = count;
        uio->uio_resid = count;
        uio->uio_iov->iov_len = count;
-       error = readip(ip, uio, cred);
-       *offp += count - uio->uio_resid;
-       IUNLOCK(ip);
+       error = ufs_read(vp, uio, 0, cred);
+       uio->uio_resid += lost;
+       if ((VTOI(vp)->i_size - uio->uio_offset) <= 0)
+               *eofflagp = 1;
+       else
+               *eofflagp = 0;
        return (error);
 }
 
        return (error);
 }
 
@@ -1023,7 +1378,7 @@ ufs_readlink(vp, uiop, cred)
        struct ucred *cred;
 {
 
        struct ucred *cred;
 {
 
-       return (readip(VTOI(vp), uiop, cred));
+       return (ufs_read(vp, uiop, 0, cred));
 }
 
 /*
 }
 
 /*
@@ -1033,23 +1388,23 @@ ufs_readlink(vp, uiop, cred)
 ufs_abortop(ndp)
        register struct nameidata *ndp;
 {
 ufs_abortop(ndp)
        register struct nameidata *ndp;
 {
-       register struct inode *ip;
 
 
-       if (ndp->ni_vp) {
-               ip = VTOI(ndp->ni_vp);
-               if (ip->i_flag & ILOCKED)
-                       IUNLOCK(ip);
-               vrele(ndp->ni_vp);
-       }
        if (ndp->ni_dvp) {
        if (ndp->ni_dvp) {
-               ip = VTOI(ndp->ni_dvp);
-               if (ip->i_flag & ILOCKED)
-                       IUNLOCK(ip);
+               if (VOP_ISLOCKED(ndp->ni_dvp))
+                       VOP_UNLOCK(ndp->ni_dvp);
                vrele(ndp->ni_dvp);
        }
                vrele(ndp->ni_dvp);
        }
+       if (ndp->ni_vp) {
+               if (VOP_ISLOCKED(ndp->ni_vp))
+                       VOP_UNLOCK(ndp->ni_vp);
+               vrele(ndp->ni_vp);
+       }
        return;
 }
 
        return;
 }
 
+/*
+ * Lock an inode.
+ */
 ufs_lock(vp)
        struct vnode *vp;
 {
 ufs_lock(vp)
        struct vnode *vp;
 {
@@ -1059,6 +1414,9 @@ ufs_lock(vp)
        return (0);
 }
 
        return (0);
 }
 
+/*
+ * Unlock an inode.
+ */
 ufs_unlock(vp)
        struct vnode *vp;
 {
 ufs_unlock(vp)
        struct vnode *vp;
 {
@@ -1070,6 +1428,18 @@ ufs_unlock(vp)
        return (0);
 }
 
        return (0);
 }
 
+/*
+ * Check for a locked inode.
+ */
+ufs_islocked(vp)
+       struct vnode *vp;
+{
+
+       if (VTOI(vp)->i_flag & ILOCKED)
+               return (1);
+       return (0);
+}
+
 /*
  * Get access to bmap
  */
 /*
  * Get access to bmap
  */
@@ -1085,19 +1455,196 @@ ufs_bmap(vp, bn, vpp, bnp)
                *vpp = ip->i_devvp;
        if (bnp == NULL)
                return (0);
                *vpp = ip->i_devvp;
        if (bnp == NULL)
                return (0);
-       return (bmap(ip, bn, bnp, (daddr_t *)0, (int *)0));
+       return (bmap(ip, bn, bnp));
 }
 
 /*
 }
 
 /*
- * Just call the device strategy routine
+ * Calculate the logical to physical mapping if not done already,
+ * then call the device strategy routine.
  */
  */
+int checkoverlap = 0;
+
 ufs_strategy(bp)
        register struct buf *bp;
 {
 ufs_strategy(bp)
        register struct buf *bp;
 {
-       (*bdevsw[major(bp->b_dev)].d_strategy)(bp);
+       register struct inode *ip = VTOI(bp->b_vp);
+       struct vnode *vp;
+       int error;
+
+       if (bp->b_vp->v_type == VBLK || bp->b_vp->v_type == VCHR)
+               panic("ufs_strategy: spec");
+       if (bp->b_blkno == bp->b_lblkno) {
+               if (error = bmap(ip, bp->b_lblkno, &bp->b_blkno))
+                       return (error);
+               if ((long)bp->b_blkno == -1)
+                       clrbuf(bp);
+       }
+       if ((long)bp->b_blkno == -1) {
+               biodone(bp);
+               return (0);
+       }
+#ifdef DIAGNOSTIC
+       if (checkoverlap) {
+               register struct buf *ep;
+               struct buf *ebp;
+               daddr_t start, last;
+
+               ebp = &buf[nbuf];
+               start = bp->b_blkno;
+               last = start + btodb(bp->b_bcount) - 1;
+               for (ep = buf; ep < ebp; ep++) {
+                       if (ep == bp || (ep->b_flags & B_INVAL) ||
+                           ep->b_vp == NULLVP)
+                               continue;
+                       if (VOP_BMAP(ep->b_vp, (daddr_t)0, &vp, (daddr_t)0))
+                               continue;
+                       if (vp != ip->i_devvp)
+                               continue;
+                       /* look for overlap */
+                       if (ep->b_bcount == 0 || ep->b_blkno > last ||
+                           ep->b_blkno + btodb(ep->b_bcount) <= start)
+                               continue;
+                       vprint("Disk overlap", vp);
+                       printf("\tstart %d, end %d overlap start %d, end %d\n",
+                               start, last, ep->b_blkno,
+                               ep->b_blkno + btodb(ep->b_bcount) - 1);
+                       panic("Disk buffer overlap");
+               }
+       }
+#endif /* DIAGNOSTIC */
+       vp = ip->i_devvp;
+       bp->b_dev = vp->v_rdev;
+       (*(vp->v_op->vn_strategy))(bp);
        return (0);
 }
 
        return (0);
 }
 
+/*
+ * Print out the contents of an inode.
+ */
+ufs_print(vp)
+       struct vnode *vp;
+{
+       register struct inode *ip = VTOI(vp);
+
+       printf("tag VT_UFS, ino %d, on dev %d, %d", ip->i_number,
+               major(ip->i_dev), minor(ip->i_dev));
+#ifdef FIFO
+       if (vp->v_type == VFIFO)
+               fifo_printinfo(vp);
+#endif /* FIFO */
+       printf("%s\n", (ip->i_flag & ILOCKED) ? " (LOCKED)" : "");
+       if (ip->i_spare0 == 0)
+               return;
+       printf("\towner pid %d", ip->i_spare0);
+       if (ip->i_spare1)
+               printf(" waiting pid %d", ip->i_spare1);
+       printf("\n");
+}
+
+/*
+ * Read wrapper for special devices.
+ */
+ufsspec_read(vp, uio, ioflag, cred)
+       struct vnode *vp;
+       struct uio *uio;
+       int ioflag;
+       struct ucred *cred;
+{
+
+       /*
+        * Set access flag.
+        */
+       VTOI(vp)->i_flag |= IACC;
+       return (spec_read(vp, uio, ioflag, cred));
+}
+
+/*
+ * Write wrapper for special devices.
+ */
+ufsspec_write(vp, uio, ioflag, cred)
+       struct vnode *vp;
+       struct uio *uio;
+       int ioflag;
+       struct ucred *cred;
+{
+
+       /*
+        * Set update and change flags.
+        */
+       VTOI(vp)->i_flag |= IUPD|ICHG;
+       return (spec_write(vp, uio, ioflag, cred));
+}
+
+/*
+ * Close wrapper for special devices.
+ *
+ * Update the times on the inode then do device close.
+ */
+ufsspec_close(vp, fflag, cred)
+       struct vnode *vp;
+       int fflag;
+       struct ucred *cred;
+{
+       register struct inode *ip = VTOI(vp);
+
+       if (vp->v_usecount > 1 && !(ip->i_flag & ILOCKED))
+               ITIMES(ip, &time, &time);
+       return (spec_close(vp, fflag, cred));
+}
+
+#ifdef FIFO
+/*
+ * Read wrapper for fifo's
+ */
+ufsfifo_read(vp, uio, ioflag, cred)
+       struct vnode *vp;
+       struct uio *uio;
+       int ioflag;
+       struct ucred *cred;
+{
+
+       /*
+        * Set access flag.
+        */
+       VTOI(vp)->i_flag |= IACC;
+       return (fifo_read(vp, uio, ioflag, cred));
+}
+
+/*
+ * Write wrapper for fifo's.
+ */
+ufsfifo_write(vp, uio, ioflag, cred)
+       struct vnode *vp;
+       struct uio *uio;
+       int ioflag;
+       struct ucred *cred;
+{
+
+       /*
+        * Set update and change flags.
+        */
+       VTOI(vp)->i_flag |= IUPD|ICHG;
+       return (fifo_write(vp, uio, ioflag, cred));
+}
+
+/*
+ * Close wrapper for fifo's.
+ *
+ * Update the times on the inode then do device close.
+ */
+ufsfifo_close(vp, fflag, cred)
+       struct vnode *vp;
+       int fflag;
+       struct ucred *cred;
+{
+       register struct inode *ip = VTOI(vp);
+
+       if (vp->v_usecount > 1 && !(ip->i_flag & ILOCKED))
+               ITIMES(ip, &time, &time);
+       return (fifo_close(vp, fflag, cred));
+}
+#endif /* FIFO */
+
 /*
  * Make a new file.
  */
 /*
  * Make a new file.
  */
@@ -1113,49 +1660,57 @@ maknode(mode, ndp, ipp)
        int error;
 
        *ipp = 0;
        int error;
 
        *ipp = 0;
+       if ((mode & IFMT) == 0)
+               mode |= IFREG;
        if ((mode & IFMT) == IFDIR)
                ipref = dirpref(pdir->i_fs);
        else
                ipref = pdir->i_number;
        if ((mode & IFMT) == IFDIR)
                ipref = dirpref(pdir->i_fs);
        else
                ipref = pdir->i_number;
-       error = ialloc(pdir, ipref, mode, &tip);
-       if (error) {
+       if (error = ialloc(pdir, ipref, mode, ndp->ni_cred, &tip)) {
                iput(pdir);
                return (error);
        }
        ip = tip;
                iput(pdir);
                return (error);
        }
        ip = tip;
+       ip->i_uid = ndp->ni_cred->cr_uid;
+       ip->i_gid = pdir->i_gid;
 #ifdef QUOTA
 #ifdef QUOTA
-       if (ip->i_dquot != NODQUOT)
-               panic("maknode: dquot");
+       if ((error = getinoquota(ip)) ||
+           (error = chkiq(ip, 1, ndp->ni_cred, 0))) {
+               ifree(ip, ip->i_number, mode);
+               iput(ip);
+               iput(pdir);
+               return (error);
+       }
 #endif
        ip->i_flag |= IACC|IUPD|ICHG;
 #endif
        ip->i_flag |= IACC|IUPD|ICHG;
-       if ((mode & IFMT) == 0)
-               mode |= IFREG;
        ip->i_mode = mode;
        ITOV(ip)->v_type = IFTOVT(mode);        /* Rest init'd in iget() */
        ip->i_nlink = 1;
        ip->i_mode = mode;
        ITOV(ip)->v_type = IFTOVT(mode);        /* Rest init'd in iget() */
        ip->i_nlink = 1;
-       ip->i_uid = ndp->ni_cred->cr_uid;
-       ip->i_gid = pdir->i_gid;
        if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, ndp->ni_cred) &&
            suser(ndp->ni_cred, NULL))
                ip->i_mode &= ~ISGID;
        if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, ndp->ni_cred) &&
            suser(ndp->ni_cred, NULL))
                ip->i_mode &= ~ISGID;
-#ifdef QUOTA
-       ip->i_dquot = inoquota(ip);
-#endif
 
        /*
         * Make sure inode goes to disk before directory entry.
         */
 
        /*
         * Make sure inode goes to disk before directory entry.
         */
-       if ((error = iupdat(ip, &time, &time, 1)) ||
-           (error = direnter(ip, ndp))) {
-               /*
-                * Write error occurred trying to update the inode
-                * or the directory so must deallocate the inode.
-                */
-               ip->i_nlink = 0;
-               ip->i_flag |= ICHG;
-               iput(ip);
-               return (error);
+       if (error = iupdat(ip, &time, &time, 1))
+               goto bad;
+       if (error = direnter(ip, ndp)) {
+               pdir = NULL;
+               goto bad;
        }
        *ipp = ip;
        return (0);
        }
        *ipp = ip;
        return (0);
+
+bad:
+       /*
+        * Write error occurred trying to update the inode
+        * or the directory so must deallocate the inode.
+        */
+       if (pdir)
+               iput(pdir);
+       ip->i_nlink = 0;
+       ip->i_flag |= ICHG;
+       iput(ip);
+       return (error);
 }
 }