merge of bill's code plus lint (plus, of course, cleanup of bill's bugs)
[unix-history] / usr / src / sys / ufs / lfs / lfs_vnops.c
index c1244b0..b5e8dce 100644 (file)
-/*     lfs_vnops.c     4.18    82/02/27        */
+/*     lfs_vnops.c     4.57    83/05/27        */
 
 #include "../h/param.h"
 #include "../h/systm.h"
 
 #include "../h/param.h"
 #include "../h/systm.h"
-#include "../h/mount.h"
-#include "../h/ino.h"
-#include "../h/reg.h"
-#include "../h/buf.h"
-#include "../h/filsys.h"
 #include "../h/dir.h"
 #include "../h/user.h"
 #include "../h/dir.h"
 #include "../h/user.h"
-#include "../h/inode.h"
+#include "../h/kernel.h"
 #include "../h/file.h"
 #include "../h/file.h"
-#include "../h/conf.h"
 #include "../h/stat.h"
 #include "../h/stat.h"
-#include "../h/inline.h"
+#include "../h/inode.h"
+#include "../h/fs.h"
+#include "../h/buf.h"
+#include "../h/proc.h"
+#include "../h/quota.h"
+#include "../h/uio.h"
 #include "../h/socket.h"
 #include "../h/socketvar.h"
 #include "../h/socket.h"
 #include "../h/socketvar.h"
+#include "../h/nami.h"
+#include "../h/mount.h"
+
+extern struct fileops inodeops;
+struct file *getinode();
 
 /*
 
 /*
- * the fstat system call.
+ * Change current working directory (``.'').
  */
  */
-fstat()
+chdir()
 {
 {
-       register struct file *fp;
-       register struct a {
-               int     fdes;
-               struct stat *sb;
-       } *uap;
 
 
-       uap = (struct a *)u.u_ap;
-       fp = getf(uap->fdes);
-       if (fp == NULL)
-               return;
-       if (fp->f_flag & FSOCKET)
-               u.u_error = sostat(fp->f_socket, uap->sb);
-       else
-               stat1(fp->f_inode, uap->sb);
+       chdirec(&u.u_cdir);
 }
 
 /*
 }
 
 /*
- * Stat system call.  This version does not follow links.
+ * Change notion of root (``/'') directory.
  */
  */
-stat()
+chroot()
+{
+
+       if (suser())
+               chdirec(&u.u_rdir);
+}
+
+/*
+ * Common routine for chroot and chdir.
+ */
+chdirec(ipp)
+       register struct inode **ipp;
 {
        register struct inode *ip;
 {
        register struct inode *ip;
-       register struct a {
+       struct a {
                char    *fname;
                char    *fname;
-               struct stat *sb;
-       } *uap;
+       };
 
 
-       uap = (struct a *)u.u_ap;
-       ip = namei(uchar, 0, 0);
+       ip = namei(uchar, LOOKUP, 1);
        if (ip == NULL)
                return;
        if (ip == NULL)
                return;
-       stat1(ip, uap->sb);
+       if ((ip->i_mode&IFMT) != IFDIR) {
+               u.u_error = ENOTDIR;
+               goto bad;
+       }
+       if (access(ip, IEXEC))
+               goto bad;
+       iunlock(ip);
+       if (*ipp)
+               irele(*ipp);
+       *ipp = ip;
+       return;
+
+bad:
        iput(ip);
 }
 
 /*
        iput(ip);
 }
 
 /*
- * Lstat system call.  This version does follow links.
+ * Open system call.
  */
  */
-lstat()
+open()
 {
 {
-       register struct inode *ip;
-       register struct a {
+       struct a {
                char    *fname;
                char    *fname;
-               struct stat *sb;
-       } *uap;
+               int     mode;
+               int     crtmode;
+       } *uap = (struct a *) u.u_ap;
 
 
-       uap = (struct a *)u.u_ap;
-       ip = namei(uchar, 0, 1);
-       if (ip == NULL)
+       copen(uap->mode-FOPEN, uap->crtmode);
+}
+
+/*
+ * Creat system call.
+ */
+creat()
+{
+       struct a {
+               char    *fname;
+               int     fmode;
+       } *uap = (struct a *)u.u_ap;
+
+       copen(FWRITE|FCREAT|FTRUNC, uap->fmode);
+}
+
+/*
+ * Common code for open and creat.
+ * Check permissions, allocate an open file structure,
+ * and call the device open routine if any.
+ */
+copen(mode, arg)
+       register int mode;
+       int arg;
+{
+       register struct inode *ip;
+       register struct file *fp;
+       int i;
+
+#ifdef notdef
+       if ((mode&(FREAD|FWRITE)) == 0) {
+               u.u_error = EINVAL;
+               return;
+       }
+#endif
+       if (mode&FCREAT) {
+               ip = namei(uchar, CREATE, 1);
+               if (ip == NULL) {
+                       if (u.u_error)
+                               return;
+                       ip = maknode(arg&07777&(~ISVTX));
+                       if (ip == NULL)
+                               return;
+                       mode &= ~FTRUNC;
+               } else {
+                       if (mode&FEXCL) {
+                               u.u_error = EEXIST;
+                               iput(ip);
+                               return;
+                       }
+                       mode &= ~FCREAT;
+               }
+       } else {
+               ip = namei(uchar, LOOKUP, 1);
+               if (ip == NULL)
+                       return;
+       }
+       if ((ip->i_mode & IFMT) == IFSOCK) {
+               u.u_error = EOPNOTSUPP;
+               goto bad;
+       }
+       if ((mode&FCREAT) == 0) {
+               if (mode&FREAD)
+                       if (access(ip, IREAD))
+                               goto bad;
+               if (mode&FWRITE) {
+                       if (access(ip, IWRITE))
+                               goto bad;
+                       if ((ip->i_mode&IFMT) == IFDIR) {
+                               u.u_error = EISDIR;
+                               goto bad;
+                       }
+               }
+       }
+       fp = falloc();
+       if (fp == NULL)
+               goto bad;
+       if (mode&FTRUNC)
+               itrunc(ip, (u_long)0);
+       iunlock(ip);
+       fp->f_flag = mode&FMASK;
+       fp->f_type = DTYPE_INODE;
+       fp->f_ops = &inodeops;
+       fp->f_data = (caddr_t)ip;
+       i = u.u_r.r_val1;
+#ifdef notdef
+       if (setjmp(&u.u_qsave)) {
+               if (u.u_error == 0)
+                       u.u_error = EINTR;
+               u.u_ofile[i] = NULL;
+               closef(fp);
+               return;
+       }
+#endif
+       u.u_error = openi(ip, mode);
+       if (u.u_error == 0)
                return;
                return;
-       stat1(ip, uap->sb, (off_t)0);
+       u.u_ofile[i] = NULL;
+       fp->f_count--;
+       irele(ip);
+       return;
+bad:
        iput(ip);
 }
 
 /*
        iput(ip);
 }
 
 /*
- * The basic routine for fstat and stat:
- * get the inode and pass appropriate parts back.
+ * Mknod system call
  */
  */
-stat1(ip, ub)
-       register struct inode *ip;
-       struct stat *ub;
+mknod()
 {
 {
-       register struct dinode *dp;
-       register struct buf *bp;
-       struct stat ds;
+       register struct inode *ip;
+       register struct a {
+               char    *fname;
+               int     fmode;
+               int     dev;
+       } *uap;
 
 
-       IUPDAT(ip, &time, &time, 0);
-       /*
-        * First copy from inode table
-        */
-       ds.st_dev = ip->i_dev;
-       ds.st_ino = ip->i_number;
-       ds.st_mode = ip->i_mode;
-       ds.st_nlink = ip->i_nlink;
-       ds.st_uid = ip->i_uid;
-       ds.st_gid = ip->i_gid;
-       ds.st_rdev = (dev_t)ip->i_un.i_rdev;
-       ds.st_size = ip->i_size;
-       /*
-        * next the dates in the disk
-        */
-       bp = bread(ip->i_dev, itod(ip->i_number));
-       dp = bp->b_un.b_dino;
-       dp += itoo(ip->i_number);
-       ds.st_atime = dp->di_atime;
-       ds.st_mtime = dp->di_mtime;
-       ds.st_ctime = dp->di_ctime;
-       brelse(bp);
-       if (copyout((caddr_t)&ds, (caddr_t)ub, sizeof(ds)) < 0)
-               u.u_error = EFAULT;
+       uap = (struct a *)u.u_ap;
+       if (!suser())
+               return;
+       ip = namei(uchar, CREATE, 0);
+       if (ip != NULL) {
+               u.u_error = EEXIST;
+               goto out;
+       }
+       if (u.u_error)
+               return;
+       ip = maknode(uap->fmode);
+       if (ip == NULL)
+               return;
+       switch (ip->i_mode & IFMT) {
+
+       case IFCHR:
+       case IFBLK:
+               if (uap->dev) {
+                       /*
+                        * Want to be able to use this to make badblock
+                        * inodes, so don't truncate the dev number.
+                        */
+                       ip->i_rdev = uap->dev;
+                       ip->i_flag |= IACC|IUPD|ICHG;
+               }
+       }
+
+out:
+       iput(ip);
 }
 
 /*
 }
 
 /*
- * Return target name of a symbolic link
+ * link system call
  */
  */
-readlink()
+link()
 {
 {
-       register struct inode *ip;
+       register struct inode *ip, *xp;
        register struct a {
        register struct a {
-               char    *name;
-               char    *buf;
-               int     count;
+               char    *target;
+               char    *linkname;
        } *uap;
 
        } *uap;
 
-       ip = namei(uchar, 0, 0);
+       uap = (struct a *)u.u_ap;
+       ip = namei(uchar, LOOKUP, 1); /* well, this routine is doomed anyhow */
        if (ip == NULL)
                return;
        if (ip == NULL)
                return;
-       if ((ip->i_mode&IFMT) != IFLNK) {
-               u.u_error = ENXIO;
+       if ((ip->i_mode&IFMT) == IFDIR && !suser()) {
+               iput(ip);
+               return;
+       }
+       ip->i_nlink++;
+       ip->i_flag |= ICHG;
+       iupdat(ip, &time, &time, 1);
+       iunlock(ip);
+       u.u_dirp = (caddr_t)uap->linkname;
+       xp = namei(uchar, CREATE, 0);
+       if (xp != NULL) {
+               u.u_error = EEXIST;
+               iput(xp);
                goto out;
        }
                goto out;
        }
-       uap = (struct a *)u.u_ap;
-       u.u_offset = 0;
-       u.u_base = uap->buf;
-       u.u_count = uap->count;
-       u.u_segflg = 0;
-       readi(ip);
+       if (u.u_error)
+               goto out;
+       if (u.u_pdir->i_dev != ip->i_dev) {
+               iput(u.u_pdir);
+               u.u_error = EXDEV;
+               goto out;
+       }
+       u.u_error = direnter(ip);
 out:
 out:
-       iput(ip);
-       u.u_r.r_val1 = uap->count - u.u_count;
+       if (u.u_error) {
+               ip->i_nlink--;
+               ip->i_flag |= ICHG;
+       }
+       irele(ip);
 }
 
 /*
 }
 
 /*
@@ -169,208 +300,1032 @@ symlink()
                nc++;
        }
        u.u_dirp = uap->linkname;
                nc++;
        }
        u.u_dirp = uap->linkname;
-       ip = namei(uchar, 1, 0);
+       ip = namei(uchar, CREATE, 0);
        if (ip) {
                iput(ip);
                u.u_error = EEXIST;
                return;
        }
        if (ip) {
                iput(ip);
                u.u_error = EEXIST;
                return;
        }
+       if (u.u_error)
+               return;
        ip = maknode(IFLNK | 0777);
        if (ip == NULL)
                return;
        ip = maknode(IFLNK | 0777);
        if (ip == NULL)
                return;
-       u.u_base = uap->target;
-       u.u_count = nc;
-       u.u_offset = 0;
-       u.u_segflg = 0;
-       writei(ip);
+       u.u_error = rdwri(UIO_WRITE, ip, uap->target, nc, 0, 0, (int *)0);
+       /* handle u.u_error != 0 */
        iput(ip);
 }
 
 /*
        iput(ip);
 }
 
 /*
- * the dup system call.
+ * Unlink system call.
+ * Hard to avoid races here, especially
+ * in unlinking directories.
  */
  */
-dup()
+unlink()
+{
+       struct a {
+               char    *fname;
+       };
+       register struct inode *ip, *dp;
+
+       ip = namei(uchar, DELETE | LOCKPARENT, 0);
+       if (ip == NULL)
+               return;
+       dp = u.u_pdir;
+       if ((ip->i_mode&IFMT) == IFDIR && !suser())
+               goto out;
+       /*
+        * Don't unlink a mounted file.
+        */
+       if (ip->i_dev != dp->i_dev) {
+               u.u_error = EBUSY;
+               goto out;
+       }
+       if (ip->i_flag&ITEXT)
+               xrele(ip);      /* try once to free text */
+       if (dirremove()) {
+               ip->i_nlink--;
+               ip->i_flag |= ICHG;
+       }
+out:
+       if (dp == ip)
+               irele(ip);
+       else
+               iput(ip);
+       iput(dp);
+}
+
+/*
+ * Seek system call
+ */
+lseek()
 {
        register struct file *fp;
        register struct a {
 {
        register struct file *fp;
        register struct a {
-               int     fdes;
-               int     fdes2;
+               int     fd;
+               off_t   off;
+               int     sbase;
        } *uap;
        } *uap;
-       register i, m;
 
        uap = (struct a *)u.u_ap;
 
        uap = (struct a *)u.u_ap;
-       m = uap->fdes & ~077;
-       uap->fdes &= 077;
-       fp = getf(uap->fdes);
+       fp = getinode(uap->fd);
        if (fp == NULL)
                return;
        if (fp == NULL)
                return;
-       if ((m&0100) == 0) {
-               if ((i = ufalloc()) < 0)
-                       return;
-       } else {
-               i = uap->fdes2;
-               if (i<0 || i>=NOFILE) {
-                       u.u_error = EBADF;
-                       return;
-               }
-               u.u_r.r_val1 = i;
-       }
-       if (i != uap->fdes) {
-               if (u.u_ofile[i]!=NULL)
-                       closef(u.u_ofile[i], 0);
-               if (u.u_error)
-                       return;
-               u.u_ofile[i] = fp;
-               fp->f_count++;
+       if (uap->sbase == L_INCR)
+               uap->off += fp->f_offset;
+       else if (uap->sbase == L_XTND)
+               uap->off += ((struct inode *)fp->f_data)->i_size;
+       fp->f_offset = uap->off;
+       u.u_r.r_off = uap->off;
+}
+
+/*
+ * Access system call
+ */
+saccess()
+{
+       register svuid, svgid;
+       register struct inode *ip;
+       register struct a {
+               char    *fname;
+               int     fmode;
+       } *uap;
+
+       uap = (struct a *)u.u_ap;
+       svuid = u.u_uid;
+       svgid = u.u_gid;
+       u.u_uid = u.u_ruid;
+       u.u_gid = u.u_rgid;
+       ip = namei(uchar, LOOKUP, 1);
+       if (ip != NULL) {
+               if ((uap->fmode&R_OK) && access(ip, IREAD))
+                       goto done;
+               if ((uap->fmode&W_OK) && access(ip, IWRITE))
+                       goto done;
+               if ((uap->fmode&X_OK) && access(ip, IEXEC))
+                       goto done;
+done:
+               iput(ip);
        }
        }
+       u.u_uid = svuid;
+       u.u_gid = svgid;
 }
 
 /*
 }
 
 /*
- * Mount system call.
+ * Stat system call.  This version follows links.
  */
  */
-smount()
+stat()
+{
+
+       stat1(1);
+}
+
+/*
+ * Lstat system call.  This version does not follow links.
+ */
+lstat()
+{
+
+       stat1(0);
+}
+
+stat1(follow)
+       int follow;
 {
 {
-       dev_t dev;
        register struct inode *ip;
        register struct inode *ip;
-       register struct mount *mp;
-       struct mount *smp;
-       register struct filsys *fp;
-       struct buf *bp;
        register struct a {
        register struct a {
-               char    *fspec;
-               char    *freg;
-               int     ronly;
+               char    *fname;
+               struct stat *ub;
        } *uap;
        } *uap;
-       register char *cp;
+       struct stat sb;
 
        uap = (struct a *)u.u_ap;
 
        uap = (struct a *)u.u_ap;
-       dev = getmdev();
-       if (u.u_error)
+       ip = namei(uchar, LOOKUP, follow);
+       if (ip == NULL)
                return;
                return;
-       u.u_dirp = (caddr_t)uap->freg;
-       ip = namei(uchar, 0, 1);
+       (void) statinode(ip, &sb);
+       iput(ip);
+       u.u_error = copyout((caddr_t)&sb, (caddr_t)uap->ub, sizeof (sb));
+}
+
+/*
+ * Return target name of a symbolic link
+ */
+readlink()
+{
+       register struct inode *ip;
+       register struct a {
+               char    *name;
+               char    *buf;
+               int     count;
+       } *uap = (struct a *)u.u_ap;
+       int resid;
+
+       ip = namei(uchar, LOOKUP, 0);
        if (ip == NULL)
                return;
        if (ip == NULL)
                return;
-       if (ip->i_count!=1 || (ip->i_mode&IFMT) != IFDIR)
-               goto out;
-       smp = NULL;
-       for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++) {
-               if (mp->m_bufp != NULL) {
-                       if (dev == mp->m_dev)
-                               goto out;
-               } else
-               if (smp == NULL)
-                       smp = mp;
-       }
-       mp = smp;
-       if (mp == NULL)
-               goto out;
-       (*bdevsw[major(dev)].d_open)(dev, !uap->ronly);
-       if (u.u_error)
+       if ((ip->i_mode&IFMT) != IFLNK) {
+               u.u_error = ENXIO;
                goto out;
                goto out;
-       bp = bread(dev, SUPERB);
-       if (u.u_error) {
-               brelse(bp);
-               goto out1;
-       }
-       mp->m_inodp = ip;
-       mp->m_dev = dev;
-       bp->b_flags |= B_LOCKED;
-       mp->m_bufp = bp;
-       fp = bp->b_un.b_filsys;
-       fp->s_ilock = 0;
-       fp->s_flock = 0;
-       fp->s_ronly = uap->ronly & 1;
-       fp->s_nbehind = 0;
-       fp->s_lasti = 1;
-       u.u_dirp = uap->freg;
-       for (cp = fp->s_fsmnt; cp < &fp->s_fsmnt[sizeof (fp->s_fsmnt) - 1]; )
-               if ((*cp++ = uchar()) == 0)
-                       u.u_dirp--;             /* get 0 again */
-       *cp = 0;
-       brelse(bp);
-       ip->i_flag |= IMOUNT;
-       irele(ip);
-       return;
-
+       }
+       u.u_error = rdwri(UIO_READ, ip, uap->buf, uap->count, 0, 0, &resid);
 out:
 out:
-       u.u_error = EBUSY;
-out1:
        iput(ip);
        iput(ip);
+       u.u_r.r_val1 = uap->count - resid;
+}
+
+/*
+ * Change mode of a file given path name.
+ */
+chmod()
+{
+       struct inode *ip;
+       struct a {
+               char    *fname;
+               int     fmode;
+       } *uap;
+
+       uap = (struct a *)u.u_ap;
+       if ((ip = owner(1)) == NULL)
+               return;
+       chmod1(ip, uap->fmode);
+       iput(ip);
+}
+
+/*
+ * Change mode of a file given a file descriptor.
+ */
+fchmod()
+{
+       struct a {
+               int     fd;
+               int     fmode;
+       } *uap;
+       register struct inode *ip;
+       register struct file *fp;
+
+       uap = (struct a *)u.u_ap;
+       fp = getinode(uap->fd);
+       if (fp == NULL)
+               return;
+       ip = (struct inode *)fp->f_data;
+       if (u.u_uid != ip->i_uid && !suser())
+               return;
+       ilock(ip);
+       chmod1(ip, uap->fmode);
+       iunlock(ip);
+}
+
+/*
+ * Change the mode on a file.
+ * Inode must be locked before calling.
+ */
+chmod1(ip, mode)
+       register struct inode *ip;
+       register int mode;
+{
+
+       ip->i_mode &= ~07777;
+       if (u.u_uid) {
+               mode &= ~ISVTX;
+               if (!groupmember(ip->i_gid))
+                       mode &= ~ISGID;
+       }
+       ip->i_mode |= mode&07777;
+       ip->i_flag |= ICHG;
+       if (ip->i_flag&ITEXT && (ip->i_mode&ISVTX)==0)
+               xrele(ip);
+}
+
+/*
+ * Set ownership given a path name.
+ */
+chown()
+{
+       struct inode *ip;
+       struct a {
+               char    *fname;
+               int     uid;
+               int     gid;
+       } *uap;
+
+       uap = (struct a *)u.u_ap;
+       if (!suser() || (ip = owner(0)) == NULL)
+               return;
+       u.u_error = chown1(ip, uap->uid, uap->gid);
+       iput(ip);
+}
+
+/*
+ * Set ownership given a file descriptor.
+ */
+fchown()
+{
+       struct a {
+               int     fd;
+               int     uid;
+               int     gid;
+       } *uap;
+       register struct inode *ip;
+       register struct file *fp;
+
+       uap = (struct a *)u.u_ap;
+       fp = getinode(uap->fd);
+       if (fp == NULL)
+               return;
+       ip = (struct inode *)fp->f_data;
+       if (!suser())
+               return;
+       ilock(ip);
+       u.u_error = chown1(ip, uap->uid, uap->gid);
+       iunlock(ip);
+}
+
+/*
+ * Perform chown operation on inode ip;
+ * inode must be locked prior to call.
+ */
+chown1(ip, uid, gid)
+       register struct inode *ip;
+       int uid, gid;
+{
+#ifdef QUOTA
+       register long change;
+#endif
+
+       if (uid == -1)
+               uid = ip->i_uid;
+       if (gid == -1)
+               gid = ip->i_gid;
+#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);
+#endif
+       ip->i_uid = uid;
+       ip->i_gid = gid;
+       ip->i_flag |= ICHG;
+       if (u.u_ruid != 0)
+               ip->i_mode &= ~(ISUID|ISGID);
+#ifdef QUOTA
+       ip->i_dquot = inoquota(ip);
+       (void) chkdq(ip, change, 1);
+       (void) chkiq(ip->i_dev, (struct inode *)NULL, uid, 1);
+       return (u.u_error);             /* should == 0 ALWAYS !! */
+#else
+       return (0);
+#endif
 }
 
 }
 
+#ifndef NOCOMPAT
 /*
 /*
- * the umount system call.
+ * Set IUPD and IACC times on file.
+ * Can't set ICHG.
  */
  */
-sumount()
+outime()
 {
 {
-       dev_t dev;
+       register struct a {
+               char    *fname;
+               time_t  *tptr;
+       } *uap = (struct a *)u.u_ap;
        register struct inode *ip;
        register struct inode *ip;
-       register struct mount *mp;
-       struct buf *bp;
-       int stillopen, flag;
+       time_t tv[2];
+       struct timeval tv0, tv1;
+
+       if ((ip = owner(1)) == NULL)
+               return;
+       u.u_error = copyin((caddr_t)uap->tptr, (caddr_t)tv, sizeof (tv));
+       if (u.u_error == 0) {
+               ip->i_flag |= IACC|IUPD|ICHG;
+               tv0.tv_sec = tv[0]; tv0.tv_usec = 0;
+               tv1.tv_sec = tv[1]; tv1.tv_usec = 0;
+               iupdat(ip, &tv0, &tv1, 0);
+       }
+       iput(ip);
+}
+#endif
+
+utimes()
+{
        register struct a {
        register struct a {
-               char    *fspec;
-       };
+               char    *fname;
+               struct  timeval *tptr;
+       } *uap = (struct a *)u.u_ap;
+       register struct inode *ip;
+       struct timeval tv[2];
 
 
-       dev = getmdev();
-       if (u.u_error)
+       if ((ip = owner(1)) == NULL)
                return;
                return;
-       xumount(dev);   /* remove unused sticky files from text table */
-       update(0);
-       for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++)
-               if (mp->m_bufp != NULL && dev == mp->m_dev)
-                       goto found;
-       u.u_error = EINVAL;
-       return;
+       u.u_error = copyin((caddr_t)uap->tptr, (caddr_t)tv, sizeof (tv));
+       if (u.u_error == 0) {
+               ip->i_flag |= IACC|IUPD|ICHG;
+               iupdat(ip, &tv[0], &tv[1], 0);
+       }
+       iput(ip);
+}
+
+/*
+ * Flush any pending I/O.
+ */
+sync()
+{
+
+       update();
+}
+
+/*
+ * Apply an advisory lock on a file descriptor.
+ */
+flock()
+{
+       register struct a {
+               int     fd;
+               int     how;
+       } *uap = (struct a *)u.u_ap;
+       register struct file *fp;
+       register int cmd, flags;
 
 
-found:
-       stillopen = 0;
-       for (ip = inode; ip < inodeNINODE; ip++)
-               if (ip->i_number != 0 && dev == ip->i_dev) {
-                       u.u_error = EBUSY;
+       fp = getinode(uap->fd);
+       if (fp == NULL)
+               return;
+       cmd = uap->how;
+       flags = u.u_pofile[uap->fd] & (UF_SHLOCK|UF_EXLOCK);
+       if (cmd&LOCK_UN) {
+               if (flags == 0) {
+                       u.u_error = EINVAL;
                        return;
                        return;
-               } else if (ip->i_number != 0 && (ip->i_mode&IFMT) == IFBLK &&
-                   ip->i_un.i_rdev == dev)
-                       stillopen++;
-       ip = mp->m_inodp;
-       ip->i_flag &= ~IMOUNT;
-       ilock(ip);
+               }
+               funlocki((struct inode *)fp->f_data, flags);
+               u.u_pofile[uap->fd] &= ~(UF_SHLOCK|UF_EXLOCK);
+               return;
+       }
+       /*
+        * No reason to write lock a file we've already
+        * write locked, similarly with a read lock.
+        */
+       if ((flags&UF_EXLOCK) && (cmd&LOCK_EX) ||
+           (flags&UF_SHLOCK) && (cmd&LOCK_SH))
+               return;
+       u.u_pofile[uap->fd] =
+           flocki((struct inode *)fp->f_data, u.u_pofile[uap->fd], cmd);
+}
+
+/*
+ * Truncate a file given its path name.
+ */
+truncate()
+{
+       struct a {
+               char    *fname;
+               u_long  length;
+       } *uap = (struct a *)u.u_ap;
+       struct inode *ip;
+
+       ip = namei(uchar, LOOKUP, 1);
+       if (ip == NULL)
+               return;
+       if (access(ip, IWRITE))
+               goto bad;
+       if ((ip->i_mode&IFMT) == IFDIR) {
+               u.u_error = EISDIR;
+               goto bad;
+       }
+       itrunc(ip, uap->length);
+bad:
        iput(ip);
        iput(ip);
-       if ((bp = getblk(dev, SUPERB)) != mp->m_bufp)
-               panic("umount");
-       bp->b_flags &= ~B_LOCKED;
-       flag = !bp->b_un.b_filsys->s_ronly;
-       mp->m_bufp = NULL;
-       brelse(bp);
-       mpurge(mp - &mount[0]);
-       if (!stillopen) {
-               (*bdevsw[major(dev)].d_close)(dev, flag);
-               binval(dev);
+}
+
+/*
+ * Truncate a file given a file descriptor.
+ */
+ftruncate()
+{
+       struct a {
+               int     fd;
+               u_long  length;
+       } *uap = (struct a *)u.u_ap;
+       struct inode *ip;
+       struct file *fp;
+
+       fp = getinode(uap->fd);
+       if (fp == NULL)
+               return;
+       if ((fp->f_flag&FWRITE) == 0) {
+               u.u_error = EINVAL;
+               return;
        }
        }
+       ip = (struct inode *)fp->f_data;
+       ilock(ip);
+       itrunc(ip, uap->length);
+       iunlock(ip);
 }
 
 /*
 }
 
 /*
- * 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.
+ * Synch an open file.
  */
  */
-dev_t
-getmdev()
+fsync()
+{
+       struct a {
+               int     fd;
+       } *uap = (struct a *)u.u_ap;
+       struct inode *ip;
+       struct file *fp;
+
+       fp = getinode(uap->fd);
+       if (fp == NULL)
+               return;
+       ip = (struct inode *)fp->f_data;
+       ilock(ip);
+       syncip(ip);
+       iunlock(ip);
+}
+
+/*
+ * Rename system call.
+ *     rename("foo", "bar");
+ * is essentially
+ *     unlink("bar");
+ *     link("foo", "bar");
+ *     unlink("foo");
+ * but ``atomically''.  Can't do full commit without saving state in the
+ * inode on disk which isn't feasible at this time.  Best we can do is
+ * always guarantee the target exists.
+ *
+ * Basic algorithm is:
+ *
+ * 1) Bump link count on source while we're linking it to the
+ *    target.  This also insure the inode won't be deleted out
+ *    from underneath us while we work.
+ * 2) Link source to destination.  If destination already exists,
+ *    delete it first.
+ * 3) Unlink source reference to inode if still around.
+ * 4) If a directory was moved and the parent of the destination
+ *    is different from the source, patch the ".." entry in the
+ *    directory.
+ *
+ * Source and destination must either both be directories, or both
+ * not be directories.  If target is a directory, it must be empty.
+ */
+rename()
+{
+       struct a {
+               char    *from;
+               char    *to;
+       } *uap;
+       register struct inode *ip, *xp, *dp;
+       int oldparent, parentdifferent, doingdirectory;
+       int error = 0;
+
+       uap = (struct a *)u.u_ap;
+       ip = namei(uchar, DELETE | LOCKPARENT, 0);
+       if (ip == NULL)
+               return;
+       dp = u.u_pdir;
+       oldparent = 0, doingdirectory = 0;
+       if ((ip->i_mode&IFMT) == IFDIR) {
+               register struct direct *d;
+
+               d = &u.u_dent;
+               /*
+                * Avoid ".", "..", and aliases of "." for obvious reasons.
+                */
+               if ((d->d_namlen == 1 && d->d_name[0] == '.') ||
+                   (d->d_namlen == 2 && bcmp(d->d_name, "..", 2) == 0) ||
+                   (dp == ip)) {
+                       iput(dp);
+                       if (dp == ip)
+                               irele(ip);
+                       else
+                               iput(ip);
+                       u.u_error = EINVAL;
+                       return;
+               }
+               oldparent = dp->i_number;
+               doingdirectory++;
+       }
+       iput(dp);
+
+       /*
+        * 1) Bump link count while we're moving stuff
+        *    around.  If we crash somewhere before
+        *    completing our work, the link count
+        *    may be wrong, but correctable.
+        */
+       ip->i_nlink++;
+       ip->i_flag |= ICHG;
+       iupdat(ip, &time, &time, 1);
+       iunlock(ip);
+
+       /*
+        * When the target exists, both the directory
+        * and target inodes are returned locked.
+        */
+       u.u_dirp = (caddr_t)uap->to;
+       xp = namei(uchar, CREATE | LOCKPARENT, 0);
+       if (u.u_error) {
+               error = u.u_error;
+               goto out;
+       }
+       dp = u.u_pdir;
+       /*
+        * If ".." must be changed (ie the directory gets a new
+        * parent) then the user must have write permission.
+        */
+       parentdifferent = oldparent != dp->i_number;
+       if (doingdirectory && parentdifferent && access(ip, IWRITE))
+               goto bad;
+       /*
+        * 2) If target doesn't exist, link the target
+        *    to the source and unlink the source. 
+        *    Otherwise, rewrite the target directory
+        *    entry to reference the source inode and
+        *    expunge the original entry's existence.
+        */
+       if (xp == NULL) {
+               if (dp->i_dev != ip->i_dev) {
+                       error = EXDEV;
+                       goto bad;
+               }
+               /*
+                * Disallow rename(foo, foo/bar).
+                */
+               if (dp->i_number == ip->i_number) {
+                       error = EEXIST;
+                       goto bad;
+               }
+               /*
+                * Account for ".." in directory.
+                * When source and destination have the
+                * same parent we don't fool with the
+                * link count -- this isn't required
+                * because we do a similar check below.
+                */
+               if (doingdirectory && parentdifferent) {
+                       dp->i_nlink++;
+                       dp->i_flag |= ICHG;
+                       iupdat(dp, &time, &time, 1);
+               }
+               error = direnter(ip);
+               if (error)
+                       goto out;
+       } else {
+               if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev) {
+                       error = EXDEV;
+                       goto bad;
+               }
+               /*
+                * Short circuit rename(foo, foo).
+                */
+               if (xp->i_number == ip->i_number)
+                       goto bad;
+               /*
+                * Target must be empty if a directory
+                * and have no links to it.
+                * Also, insure source and target are
+                * compatible (both directories, or both
+                * not directories).
+                */
+               if ((xp->i_mode&IFMT) == IFDIR) {
+                       if (!dirempty(xp) || xp->i_nlink > 2) {
+                               error = ENOTEMPTY;
+                               goto bad;
+                       }
+                       if (!doingdirectory) {
+                               error = ENOTDIR;
+                               goto bad;
+                       }
+               } else if (doingdirectory) {
+                       error = EISDIR;
+                       goto bad;
+               }
+               dirrewrite(dp, ip);
+               if (u.u_error) {
+                       error = u.u_error;
+                       goto bad1;
+               }
+               /*
+                * Adjust the link count of the target to
+                * reflect the dirrewrite above.  If this is
+                * a directory it is empty and there are
+                * no links to it, so we can squash the inode and
+                * any space associated with it.  We disallowed
+                * renaming over top of a directory with links to
+                * it above, as we've no way to determine if
+                * we've got a link or the directory itself, and
+                * if we get a link, then ".." will be screwed up.
+                */
+               xp->i_nlink--;
+               if (doingdirectory) {
+                       if (--xp->i_nlink != 0)
+                               panic("rename: linked directory");
+                       itrunc(xp, (u_long)0);
+               }
+               xp->i_flag |= ICHG;
+               iput(xp);
+               xp = NULL;
+       }
+
+       /*
+        * 3) Unlink the source.
+        */
+       u.u_dirp = uap->from;
+       dp = namei(uchar, DELETE, 0);
+       /*
+        * Insure directory entry still exists and
+        * has not changed since the start of all
+        * this.  If either has occured, forget about
+        * about deleting the original entry and just
+        * adjust the link count in the inode.
+        */
+       if (dp == NULL || u.u_dent.d_ino != ip->i_number) {
+               ip->i_nlink--;
+               ip->i_flag |= ICHG;
+       } else {
+               /*
+                * If source is a directory, must adjust
+                * link count of parent directory also.
+                * If target didn't exist and source and
+                * target have the same parent, then we
+                * needn't touch the link count, it all
+                * balances out in the end.  Otherwise, we
+                * must do so to reflect deletion of ".."
+                * done above.
+                */
+               if (doingdirectory && (xp != NULL || parentdifferent)) {
+                       dp->i_nlink--;
+                       dp->i_flag |= ICHG;
+               }
+               if (dirremove()) {
+                       ip->i_nlink--;
+                       ip->i_flag |= ICHG;
+               }
+               if (error == 0)         /* conservative */
+                       error = u.u_error;
+       }
+       irele(ip);
+       if (dp)
+               iput(dp);
+
+       /*
+        * 4) Renaming a directory with the parent
+        *    different requires ".." to be rewritten.
+        *    The window is still there for ".." to
+        *    be inconsistent, but this is unavoidable,
+        *    and a lot shorter than when it was done
+        *    in a user process.
+        */
+       if (doingdirectory && parentdifferent && error == 0) {
+               struct dirtemplate dirbuf;
+
+               u.u_dirp = uap->to;
+               ip = namei(uchar, LOOKUP | LOCKPARENT, 0);
+               if (ip == NULL) {
+                       printf("rename: .. went away\n");
+                       return;
+               }
+               dp = u.u_pdir;
+               if ((ip->i_mode&IFMT) != IFDIR) {
+                       printf("rename: .. not a directory\n");
+                       goto stuck;
+               }
+               error = rdwri(UIO_READ, ip, (caddr_t)&dirbuf,
+                       sizeof (struct dirtemplate), (off_t)0, 1, (int *)0);
+               if (error == 0) {
+                       dirbuf.dotdot_ino = dp->i_number;
+                       (void) rdwri(UIO_WRITE, ip, (caddr_t)&dirbuf,
+                         sizeof (struct dirtemplate), (off_t)0, 1, (int *)0);
+               }
+stuck:
+               irele(dp);
+               iput(ip);
+       }
+       goto done;
+
+bad:
+       iput(dp);
+bad1:
+       if (xp)
+               iput(xp);
+out:
+       ip->i_nlink--;
+       ip->i_flag |= ICHG;
+       irele(ip);
+done:
+       if (error)
+               u.u_error = error;
+}
+
+/*
+ * Make a new file.
+ */
+struct inode *
+maknode(mode)
+       int mode;
 {
 {
-       dev_t dev;
        register struct inode *ip;
        register struct inode *ip;
+       ino_t ipref;
 
 
-       if (!suser())
-               return(NODEV);
-       ip = namei(uchar, 0, 1);
+       if ((mode & IFMT) == IFDIR)
+               ipref = dirpref(u.u_pdir->i_fs);
+       else
+               ipref = u.u_pdir->i_number;
+       ip = ialloc(u.u_pdir, ipref, mode);
+       if (ip == NULL) {
+               iput(u.u_pdir);
+               return (NULL);
+       }
+#ifdef QUOTA
+       if (ip->i_dquot != NODQUOT)
+               panic("maknode: dquot");
+#endif
+       ip->i_flag |= IACC|IUPD|ICHG;
+       if ((mode & IFMT) == 0)
+               mode |= IFREG;
+       ip->i_mode = mode & ~u.u_cmask;
+       ip->i_nlink = 1;
+       ip->i_uid = u.u_uid;
+       ip->i_gid = u.u_pdir->i_gid;
+       if (ip->i_mode & ISGID && !groupmember(ip->i_gid))
+               ip->i_mode &= ~ISGID;
+#ifdef QUOTA
+       ip->i_dquot = inoquota(ip);
+#endif
+
+       /*
+        * Make sure inode goes to disk before directory entry.
+        */
+       iupdat(ip, &time, &time, 1);
+       u.u_error = direnter(ip);
+       if (u.u_error) {
+               /*
+                * Write error occurred trying to update directory
+                * so must deallocate the inode.
+                */
+               ip->i_nlink = 0;
+               ip->i_flag |= ICHG;
+               iput(ip);
+               return (NULL);
+       }
+       return (ip);
+}
+
+/*
+ * A virgin directory (no blushing please).
+ */
+struct dirtemplate mastertemplate = {
+       0, 12, 1, ".",
+       0, DIRBLKSIZ - 12, 2, ".."
+};
+
+/*
+ * Mkdir system call
+ */
+mkdir()
+{
+       struct a {
+               char    *name;
+               int     dmode;
+       } *uap;
+       register struct inode *ip, *dp;
+       struct dirtemplate dirtemplate;
+
+       uap = (struct a *)u.u_ap;
+       ip = namei(uchar, CREATE, 0);
+       if (u.u_error)
+               return;
+       if (ip != NULL) {
+               iput(ip);
+               u.u_error = EEXIST;
+               return;
+       }
+       dp = u.u_pdir;
+       uap->dmode &= 0777;
+       uap->dmode |= IFDIR;
+       /*
+        * Must simulate part of maknode here
+        * in order to acquire the inode, but
+        * not have it entered in the parent
+        * directory.  The entry is made later
+        * after writing "." and ".." entries out.
+        */
+       ip = ialloc(dp, dirpref(dp->i_fs), uap->dmode);
+       if (ip == NULL) {
+               iput(dp);
+               return;
+       }
+#ifdef QUOTA
+       if (ip->i_dquot != NODQUOT)
+               panic("mkdir: dquot");
+#endif
+       ip->i_flag |= IACC|IUPD|ICHG;
+       ip->i_mode = uap->dmode & ~u.u_cmask;
+       ip->i_nlink = 2;
+       ip->i_uid = u.u_uid;
+       ip->i_gid = dp->i_gid;
+#ifdef QUOTA
+       ip->i_dquot = inoquota(ip);
+#endif
+       iupdat(ip, &time, &time, 1);
+
+       /*
+        * Bump link count in parent directory
+        * to reflect work done below.  Should
+        * be done before reference is created
+        * so reparation is possible if we crash.
+        */
+       dp->i_nlink++;
+       dp->i_flag |= ICHG;
+       iupdat(dp, &time, &time, 1);
+
+       /*
+        * Initialize directory with "."
+        * and ".." from static template.
+        */
+       dirtemplate = mastertemplate;
+       dirtemplate.dot_ino = ip->i_number;
+       dirtemplate.dotdot_ino = dp->i_number;
+       u.u_error = rdwri(UIO_WRITE, ip, (caddr_t)&dirtemplate,
+               sizeof (dirtemplate), (off_t)0, 1, (int *)0);
+       if (u.u_error) {
+               dp->i_nlink--;
+               dp->i_flag |= ICHG;
+               goto bad;
+       }
+       /*
+        * Directory all set up, now
+        * install the entry for it in
+        * the parent directory.
+        */
+       u.u_error = direnter(ip);
+       dp = NULL;
+       if (u.u_error) {
+               u.u_dirp = uap->name;
+               dp = namei(uchar, LOOKUP, 0);
+               if (dp) {
+                       dp->i_nlink--;
+                       dp->i_flag |= ICHG;
+               }
+       }
+bad:
+       /*
+        * No need to do an explicit itrunc here,
+        * irele will do this for us because we set
+        * the link count to 0.
+        */
+       if (u.u_error) {
+               ip->i_nlink = 0;
+               ip->i_flag |= ICHG;
+       }
+       if (dp)
+               iput(dp);
+       iput(ip);
+}
+
+/*
+ * Rmdir system call.
+ */
+rmdir()
+{
+       struct a {
+               char    *name;
+       };
+       register struct inode *ip, *dp;
+
+       ip = namei(uchar, DELETE | LOCKPARENT, 0);
        if (ip == NULL)
        if (ip == NULL)
-               return(NODEV);
-       if ((ip->i_mode&IFMT) != IFBLK)
-               u.u_error = ENOTBLK;
-       dev = (dev_t)ip->i_un.i_rdev;
-       if (major(dev) >= nblkdev)
-               u.u_error = ENXIO;
+               return;
+       dp = u.u_pdir;
+       /*
+        * No rmdir "." please.
+        */
+       if (dp == ip) {
+               irele(dp);
+               iput(ip);
+               u.u_error = EINVAL;
+               return;
+       }
+       if ((ip->i_mode&IFMT) != IFDIR) {
+               u.u_error = ENOTDIR;
+               goto out;
+       }
+       /*
+        * Don't remove a mounted on directory.
+        */
+       if (ip->i_dev != dp->i_dev) {
+               u.u_error = EBUSY;
+               goto out;
+       }
+       /*
+        * Verify the directory is empty (and valid).
+        * (Rmdir ".." won't be valid since
+        *  ".." will contain a reference to
+        *  the current directory and thus be
+        *  non-empty.)
+        */
+       if (ip->i_nlink != 2 || !dirempty(ip)) {
+               u.u_error = ENOTEMPTY;
+               goto out;
+       }
+       /*
+        * Delete reference to directory before purging
+        * inode.  If we crash in between, the directory
+        * will be reattached to lost+found,
+        */
+       if (dirremove() == 0)
+               goto out;
+       dp->i_nlink--;
+       dp->i_flag |= ICHG;
+       iput(dp);
+       dp = NULL;
+       /*
+        * Truncate inode.  The only stuff left
+        * in the directory is "." and "..".  The
+        * "." reference is inconsequential since
+        * we're quashing it.  The ".." reference
+        * has already been adjusted above.  We've
+        * removed the "." reference and the reference
+        * in the parent directory, but there may be
+        * other hard links so decrement by 2 and
+        * worry about them later.
+        */
+       ip->i_nlink -= 2;
+       itrunc(ip, (u_long)0);
+out:
+       if (dp)
+               iput(dp);
        iput(ip);
        iput(ip);
-       return(dev);
+}
+
+struct file *
+getinode(fdes)
+       int fdes;
+{
+       register struct file *fp;
+
+       fp = getf(fdes);
+       if (fp == 0)
+               return (0);
+       if (fp->f_type != DTYPE_INODE) {
+               u.u_error = EINVAL;
+               return (0);
+       }
+       return (fp);
+}
+
+/*
+ * mode mask for creation of files
+ */
+umask()
+{
+       register struct a {
+               int     mask;
+       } *uap;
+
+       uap = (struct a *)u.u_ap;
+       u.u_r.r_val1 = u.u_cmask;
+       u.u_cmask = uap->mask & 07777;
 }
 }