X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/2c3059b12521d35820e41bed7b6b9322bb45a87c..67361d543188004c23e8675506802b52a85fa12f:/usr/src/sys/kern/vfs_syscalls.c diff --git a/usr/src/sys/kern/vfs_syscalls.c b/usr/src/sys/kern/vfs_syscalls.c index e7531698d3..a5e4d5dbef 100644 --- a/usr/src/sys/kern/vfs_syscalls.c +++ b/usr/src/sys/kern/vfs_syscalls.c @@ -1,276 +1,1370 @@ -/* vfs_syscalls.c 4.4 %G% */ +/* + * 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. + * + * @(#)vfs_syscalls.c 7.3 (Berkeley) %G% + */ + +#include "param.h" +#include "systm.h" +#include "dir.h" +#include "user.h" +#include "kernel.h" +#include "file.h" +#include "stat.h" +#include "inode.h" +#include "fs.h" +#include "buf.h" +#include "proc.h" +#include "quota.h" +#include "uio.h" +#include "socket.h" +#include "socketvar.h" +#include "mount.h" + +extern struct fileops inodeops; +struct file *getinode(); + +/* + * Change current working directory (``.''). + */ +chdir() +{ + + chdirec(&u.u_cdir); +} + +/* + * Change notion of root (``/'') directory. + */ +chroot() +{ + + if (suser()) + chdirec(&u.u_rdir); +} + +/* + * Common routine for chroot and chdir. + */ +chdirec(ipp) + register struct inode **ipp; +{ + register struct inode *ip; + struct a { + char *fname; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; + + ndp->ni_nameiop = LOOKUP | FOLLOW; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); + if (ip == NULL) + return; + 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); +} -#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/inode.h" -#include "../h/file.h" -#include "../h/conf.h" -#include "../h/stat.h" -#include "../h/inline.h" +/* + * Open system call. + */ +open() +{ + struct a { + char *fname; + int mode; + int crtmode; + } *uap = (struct a *) u.u_ap; + + copen(uap->mode-FOPEN, uap->crtmode, uap->fname); +} /* - * the fstat system call. + * Creat system call. */ -fstat() +creat() { + struct a { + char *fname; + int fmode; + } *uap = (struct a *)u.u_ap; + + copen(FWRITE|FCREAT|FTRUNC, uap->fmode, uap->fname); +} + +/* + * Common code for open and creat. + * Check permissions, allocate an open file structure, + * and call the device open routine if any. + */ +copen(mode, arg, fname) + register int mode; + int arg; + caddr_t fname; +{ + register struct inode *ip; register struct file *fp; - register struct a { - int fdes; - struct stat *sb; - } *uap; + register struct nameidata *ndp = &u.u_nd; + int indx; - uap = (struct a *)u.u_ap; - fp = getf(uap->fdes); - if(fp == NULL) + fp = falloc(); + if (fp == NULL) + return; + indx = u.u_r.r_val1; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = fname; + if (mode&FCREAT) { + if (mode & FEXCL) + ndp->ni_nameiop = CREATE; + else + ndp->ni_nameiop = CREATE | FOLLOW; + ip = namei(ndp); + if (ip == NULL) { + if (u.u_error) + goto bad1; + ip = maknode(arg&07777&(~ISVTX), ndp); + if (ip == NULL) + goto bad1; + mode &= ~FTRUNC; + } else { + if (mode&FEXCL) { + u.u_error = EEXIST; + goto bad; + } + mode &= ~FCREAT; + } + } else { + ndp->ni_nameiop = LOOKUP | FOLLOW; + ip = namei(ndp); + if (ip == NULL) + goto bad1; + } + 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|FTRUNC)) { + if (access(ip, IWRITE)) + goto bad; + if ((ip->i_mode&IFMT) == IFDIR) { + u.u_error = EISDIR; + 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; + if (setjmp(&u.u_qsave)) { + if (u.u_error == 0) + u.u_error = EINTR; + u.u_ofile[indx] = NULL; + closef(fp); + return; + } + u.u_error = openi(ip, mode); + if (u.u_error == 0) return; - stat1(fp->f_inode, uap->sb, fp->f_flag&FPIPE? fp->f_un.f_offset: 0); + ILOCK(ip); +bad: + iput(ip); +bad1: + u.u_ofile[indx] = NULL; + fp->f_count--; } /* - * the stat system call. + * Mknod system call */ -stat() +mknod() { register struct inode *ip; register struct a { char *fname; - struct stat *sb; - } *uap; + int fmode; + int dev; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; - uap = (struct a *)u.u_ap; - ip = namei(uchar, 0); - if(ip == NULL) + if (!suser()) return; - stat1(ip, uap->sb, (off_t)0); + ndp->ni_nameiop = CREATE; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); + if (ip != NULL) { + u.u_error = EEXIST; + goto out; + } + if (u.u_error) + return; + ip = maknode(uap->fmode, ndp); + if (ip == NULL) + return; + switch (ip->i_mode & IFMT) { + + case IFMT: /* used by badsect to flag bad sectors */ + 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); } /* - * The basic routine for fstat and stat: - * get the inode and pass appropriate parts back. + * link system call */ -stat1(ip, ub, pipeadj) -register struct inode *ip; -struct stat *ub; -off_t pipeadj; +link() { - register struct dinode *dp; - register struct buf *bp; - struct stat ds; + register struct inode *ip, *xp; + register struct a { + char *target; + char *linkname; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; - 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 - pipeadj; + ndp->ni_nameiop = LOOKUP | FOLLOW; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->target; + ip = namei(ndp); /* well, this routine is doomed anyhow */ + if (ip == NULL) + return; + if ((ip->i_mode&IFMT) == IFDIR && !suser()) { + iput(ip); + return; + } + ip->i_nlink++; + ip->i_flag |= ICHG; + iupdat(ip, &time, &time, 1); + IUNLOCK(ip); + ndp->ni_nameiop = CREATE; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = (caddr_t)uap->linkname; + xp = namei(ndp); + if (xp != NULL) { + u.u_error = EEXIST; + iput(xp); + goto out; + } + if (u.u_error) + goto out; + if (ndp->ni_pdir->i_dev != ip->i_dev) { + iput(ndp->ni_pdir); + u.u_error = EXDEV; + goto out; + } + u.u_error = direnter(ip, ndp); +out: + if (u.u_error) { + ip->i_nlink--; + ip->i_flag |= ICHG; + } + irele(ip); +} + +/* + * symlink -- make a symbolic link + */ +symlink() +{ + register struct a { + char *target; + char *linkname; + } *uap = (struct a *)u.u_ap; + register struct inode *ip; + register char *tp; + register c, nc; + register struct nameidata *ndp = &u.u_nd; + + tp = uap->target; + nc = 0; + while (c = fubyte(tp)) { + if (c < 0) { + u.u_error = EFAULT; + return; + } + tp++; + nc++; + } + ndp->ni_nameiop = CREATE; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->linkname; + ip = namei(ndp); + if (ip) { + iput(ip); + u.u_error = EEXIST; + return; + } + if (u.u_error) + return; + ip = maknode(IFLNK | 0777, ndp); + if (ip == NULL) + return; + u.u_error = rdwri(UIO_WRITE, ip, uap->target, nc, (off_t)0, 0, + (int *)0); + /* handle u.u_error != 0 */ + iput(ip); +} + +/* + * Unlink system call. + * Hard to avoid races here, especially + * in unlinking directories. + */ +unlink() +{ + struct a { + char *fname; + } *uap = (struct a *)u.u_ap; + register struct inode *ip, *dp; + register struct nameidata *ndp = &u.u_nd; + + ndp->ni_nameiop = DELETE | LOCKPARENT; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); + if (ip == NULL) + return; + dp = ndp->ni_pdir; + if ((ip->i_mode&IFMT) == IFDIR && !suser()) + goto out; /* - * next the dates in the disk + * Don't unlink a mounted file. */ - 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; + 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(ndp)) { + ip->i_nlink--; + ip->i_flag |= ICHG; + } +out: + if (dp == ip) + irele(ip); + else + iput(ip); + iput(dp); } /* - * the dup system call. + * Seek system call */ -dup() +lseek() { register struct file *fp; register struct a { - int fdes; - int fdes2; - } *uap; - register i, m; - - uap = (struct a *)u.u_ap; - m = uap->fdes & ~077; - uap->fdes &= 077; - fp = getf(uap->fdes); - 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; - } - if (u.u_vrpages[i]) { - u.u_error = ETXTBSY; - return; - } - u.u_r.r_val1 = i; + int fd; + off_t off; + int sbase; + } *uap = (struct a *)u.u_ap; + + GETF(fp, uap->fd); + if (fp->f_type != DTYPE_INODE) { + u.u_error = ESPIPE; + return; } - if (i!=uap->fdes) { - if (u.u_ofile[i]!=NULL) - closef(u.u_ofile[i]); - u.u_ofile[i] = fp; - fp->f_count++; + switch (uap->sbase) { + + case L_INCR: + fp->f_offset += uap->off; + break; + + case L_XTND: + fp->f_offset = uap->off + ((struct inode *)fp->f_data)->i_size; + break; + + case L_SET: + fp->f_offset = uap->off; + break; + + default: + u.u_error = EINVAL; + return; } + u.u_r.r_off = fp->f_offset; } /* - * the mount system call. + * Access system call */ -smount() +saccess() { - dev_t dev; + register svuid, svgid; register struct inode *ip; - register struct mount *mp; - struct mount *smp; - register struct filsys *fp; - struct buf *bp; register struct a { - char *fspec; - char *freg; - int ronly; - } *uap; + char *fname; + int fmode; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; - uap = (struct a *)u.u_ap; - dev = getmdev(); - if(u.u_error) + svuid = u.u_uid; + svgid = u.u_gid; + u.u_uid = u.u_ruid; + u.u_gid = u.u_rgid; + ndp->ni_nameiop = LOOKUP | FOLLOW; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); + 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; +} + +/* + * Stat system call. This version follows links. + */ +stat() +{ + + stat1(FOLLOW); +} + +/* + * Lstat system call. This version does not follow links. + */ +lstat() +{ + + stat1(NOFOLLOW); +} + +stat1(follow) + int follow; +{ + register struct inode *ip; + register struct a { + char *fname; + struct stat *ub; + } *uap = (struct a *)u.u_ap; + struct stat sb; + register struct nameidata *ndp = &u.u_nd; + + ndp->ni_nameiop = LOOKUP | follow; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); + if (ip == NULL) return; - u.u_dirp = (caddr_t)uap->freg; - ip = namei(uchar, 0); - if(ip == NULL) + (void) ino_stat(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; + register struct nameidata *ndp = &u.u_nd; + int resid; + + ndp->ni_nameiop = LOOKUP; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->name; + ip = namei(ndp); + if (ip == NULL) return; - if(ip->i_count!=1 || (ip->i_mode&(IFBLK&IFCHR))!=0) + if ((ip->i_mode&IFMT) != IFLNK) { + u.u_error = EINVAL; 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) - 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; - brelse(bp); - ip->i_flag |= IMOUNT; - prele(ip); - return; - + u.u_error = rdwri(UIO_READ, ip, uap->buf, uap->count, (off_t)0, 0, + &resid); out: - u.u_error = EBUSY; -out1: + 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 = (struct a *)u.u_ap; + + if ((ip = owner(uap->fname, FOLLOW)) == NULL) + return; + u.u_error = chmod1(ip, uap->fmode); + iput(ip); +} + +/* + * Change mode of a file given a file descriptor. + */ +fchmod() +{ + struct a { + int fd; + int fmode; + } *uap = (struct a *)u.u_ap; + register struct inode *ip; + register struct file *fp; + + 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); + u.u_error = 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; +{ + + if (ip->i_fs->fs_ronly) + return (EROFS); + ip->i_mode &= ~07777; + if (u.u_uid) { + if ((ip->i_mode & IFMT) != IFDIR) + 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); + return (0); +} + +/* + * Set ownership given a path name. + */ +chown() +{ + struct inode *ip; + struct a { + char *fname; + int uid; + int gid; + } *uap = (struct a *)u.u_ap; + + if ((ip = owner(uap->fname, NOFOLLOW)) == NULL) + return; + u.u_error = chown1(ip, uap->uid, uap->gid); iput(ip); } /* - * the umount system call. + * Set ownership given a file descriptor. */ -sumount() +fchown() { - dev_t dev; + struct a { + int fd; + int uid; + int gid; + } *uap = (struct a *)u.u_ap; register struct inode *ip; - register struct mount *mp; - struct buf *bp; - int stillopen, flag; + register struct file *fp; + + fp = getinode(uap->fd); + if (fp == NULL) + return; + ip = (struct inode *)fp->f_data; + 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 (ip->i_fs->fs_ronly) + return (EROFS); + if (uid == -1) + uid = ip->i_uid; + if (gid == -1) + gid = ip->i_gid; + if (uid != ip->i_uid && !suser()) + return (u.u_error); + if (gid != ip->i_gid && !groupmember((gid_t)gid) && !suser()) + return (u.u_error); +#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_t)uid, 1); + return (u.u_error); /* should == 0 ALWAYS !! */ +#else + return (0); +#endif +} + +utimes() +{ 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(uap->fname, FOLLOW)) == NULL) + return; + if (ip->i_fs->fs_ronly) { + u.u_error = EROFS; + iput(ip); return; - xumount(dev); /* remove unused sticky files from text table */ + } + 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(); - for(mp = &mount[0]; mp < &mount[NMOUNT]; mp++) - if(mp->m_bufp != NULL && dev == mp->m_dev) - goto found; - u.u_error = EINVAL; - return; +} -found: - for(ip = &inode[0]; ip < &inode[NINODE]; ip++) - if (ip->i_number != 0 && dev == ip->i_dev) { - u.u_error = EBUSY; - 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; - plock(ip); +/* + * Truncate a file given its path name. + */ +truncate() +{ + struct a { + char *fname; + off_t length; + } *uap = (struct a *)u.u_ap; + struct inode *ip; + register struct nameidata *ndp = &u.u_nd; + + ndp->ni_nameiop = LOOKUP | FOLLOW; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); + if (ip == NULL) + return; + if (access(ip, IWRITE)) + goto bad; + if ((ip->i_mode&IFMT) == IFDIR) { + u.u_error = EISDIR; + goto bad; + } + itrunc(ip, (u_long)uap->length); +bad: 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); - if (!stillopen) { - mpurge(mp - &mount[0]); - (*bdevsw[major(dev)].d_close)(dev, flag); - binval(dev); +} + +/* + * Truncate a file given a file descriptor. + */ +ftruncate() +{ + struct a { + int fd; + off_t 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, (u_long)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); + if (fp->f_flag&FWRITE) + ip->i_flag |= ICHG; + 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 (it may be truncated by + * a concurrent `trunc' or `open' for creation). + * 2) Link source to destination. If destination already exists, + * delete it first. + * 3) Unlink source reference to inode if still around. 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 = (struct a *)u.u_ap; + register struct inode *ip, *xp, *dp; + struct dirtemplate dirbuf; + int doingdirectory = 0, oldparent = 0, newparent = 0; + register struct nameidata *ndp = &u.u_nd; + int error = 0; + + ndp->ni_nameiop = DELETE | LOCKPARENT; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->from; + ip = namei(ndp); + if (ip == NULL) + return; + dp = ndp->ni_pdir; + if ((ip->i_mode&IFMT) == IFDIR) { + register struct direct *d; + + d = &ndp->ni_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) || (ip->i_flag & IRENAME)) { + iput(dp); + if (dp == ip) + irele(ip); + else + iput(ip); + u.u_error = EINVAL; + return; + } + ip->i_flag |= IRENAME; + 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. + */ + ndp->ni_nameiop = CREATE | LOCKPARENT | NOCACHE; + ndp->ni_dirp = (caddr_t)uap->to; + xp = namei(ndp); + if (u.u_error) { + error = u.u_error; + goto out; + } + dp = ndp->ni_pdir; + /* + * If ".." must be changed (ie the directory gets a new + * parent) then the source directory must not be in the + * directory heirarchy above the target, as this would + * orphan everything below the source directory. Also + * the user must have write permission in the source so + * as to be able to change "..". We must repeat the call + * to namei, as the parent directory is unlocked by the + * call to checkpath(). + */ + if (oldparent != dp->i_number) + newparent = dp->i_number; + if (doingdirectory && newparent) { + if (access(ip, IWRITE)) + goto bad; + do { + dp = ndp->ni_pdir; + if (xp != NULL) + iput(xp); + u.u_error = checkpath(ip, dp); + if (u.u_error) + goto out; + xp = namei(ndp); + if (u.u_error) { + error = u.u_error; + goto out; + } + } while (dp != ndp->ni_pdir); + } + /* + * 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; + } + /* + * Account for ".." in new directory. + * When source and destination have the same + * parent we don't fool with the link count. + */ + if (doingdirectory && newparent) { + dp->i_nlink++; + dp->i_flag |= ICHG; + iupdat(dp, &time, &time, 1); + } + error = direnter(ip, ndp); + 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; + /* + * If the parent directory is "sticky", then the user must + * own the parent directory, or the destination of the rename, + * otherwise the destination may not be changed (except by + * root). This implements append-only directories. + */ + if ((dp->i_mode & ISVTX) && u.u_uid != 0 && + u.u_uid != dp->i_uid && xp->i_uid != u.u_uid) { + error = EPERM; + 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, dp->i_number) || xp->i_nlink > 2) { + error = ENOTEMPTY; + goto bad; + } + if (!doingdirectory) { + error = ENOTDIR; + goto bad; + } + cacheinval(dp); + } else if (doingdirectory) { + error = EISDIR; + goto bad; + } + dirrewrite(dp, ip, ndp); + 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 the remaining link would point to + * a directory without "." or ".." entries. + */ + 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. + */ + ndp->ni_nameiop = DELETE | LOCKPARENT; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->from; + xp = namei(ndp); + if (xp != NULL) + dp = ndp->ni_pdir; + else + dp = NULL; + /* + * Insure that the directory entry still exists and has not + * changed while the new name has been entered. If the source is + * a file then the entry may have been unlinked or renamed. In + * either case there is no further work to be done. If the source + * is a directory then it cannot have been rmdir'ed; its link + * count of three would cause a rmdir to fail with ENOTEMPTY. + * The IRENAME flag insures that it cannot be moved by another + * rename. + */ + if (xp != ip) { + if (doingdirectory) + panic("rename: lost dir entry"); + } else { + /* + * If the source is a directory with a + * new parent, the link count of the old + * parent directory must be decremented + * and ".." set to point to the new parent. + */ + if (doingdirectory && newparent) { + dp->i_nlink--; + dp->i_flag |= ICHG; + error = rdwri(UIO_READ, xp, (caddr_t)&dirbuf, + sizeof (struct dirtemplate), (off_t)0, 1, + (int *)0); + if (error == 0) { + if (dirbuf.dotdot_namlen != 2 || + dirbuf.dotdot_name[0] != '.' || + dirbuf.dotdot_name[1] != '.') { + printf("rename: mangled dir\n"); + } else { + dirbuf.dotdot_ino = newparent; + (void) rdwri(UIO_WRITE, xp, + (caddr_t)&dirbuf, + sizeof (struct dirtemplate), + (off_t)0, 1, (int *)0); + cacheinval(dp); + } + } + } + if (dirremove(ndp)) { + xp->i_nlink--; + xp->i_flag |= ICHG; + } + xp->i_flag &= ~IRENAME; + if (error == 0) /* XXX conservative */ + error = u.u_error; + } + if (dp) + iput(dp); + if (xp) + iput(xp); + irele(ip); + if (error) + u.u_error = error; + return; + +bad: + iput(dp); +bad1: + if (xp) + iput(xp); +out: + ip->i_nlink--; + ip->i_flag |= ICHG; + irele(ip); + if (error) + u.u_error = error; +} + +/* + * Make a new file. + */ +struct inode * +maknode(mode, ndp) + int mode; + register struct nameidata *ndp; { - dev_t dev; register struct inode *ip; + register struct inode *pdir = ndp->ni_pdir; + ino_t ipref; - if (!suser()) - return(NODEV); - ip = namei(uchar, 0); - 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; + if ((mode & IFMT) == IFDIR) + ipref = dirpref(pdir->i_fs); + else + ipref = pdir->i_number; + ip = ialloc(pdir, ipref, mode); + if (ip == NULL) { + iput(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 = pdir->i_gid; + if (ip->i_mode & ISGID && !groupmember(ip->i_gid) && !suser()) + 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, ndp); + 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 = (struct a *)u.u_ap; + register struct inode *ip, *dp; + struct dirtemplate dirtemplate; + register struct nameidata *ndp = &u.u_nd; + + ndp->ni_nameiop = CREATE; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->name; + ip = namei(ndp); + if (u.u_error) + return; + if (ip != NULL) { + iput(ip); + u.u_error = EEXIST; + return; + } + dp = ndp->ni_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; + } + if (DIRBLKSIZ > ip->i_fs->fs_fsize) + panic("mkdir: blksize"); /* XXX - should grow with bmap() */ + else + ip->i_size = DIRBLKSIZ; + /* + * Directory all set up, now + * install the entry for it in + * the parent directory. + */ + u.u_error = direnter(ip, ndp); + dp = NULL; + if (u.u_error) { + ndp->ni_nameiop = LOOKUP | NOCACHE; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->name; + dp = namei(ndp); + 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); - return(dev); +} + +/* + * Rmdir system call. + */ +rmdir() +{ + struct a { + char *name; + } *uap = (struct a *)u.u_ap; + register struct inode *ip, *dp; + register struct nameidata *ndp = &u.u_nd; + + ndp->ni_nameiop = DELETE | LOCKPARENT; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->name; + ip = namei(ndp); + if (ip == NULL) + return; + dp = ndp->ni_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, dp->i_number)) { + 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(ndp) == 0) + goto out; + dp->i_nlink--; + dp->i_flag |= ICHG; + cacheinval(dp); + 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); + cacheinval(ip); +out: + if (dp) + iput(dp); + iput(ip); +} + +struct file * +getinode(fdes) + int fdes; +{ + struct file *fp; + + if ((unsigned)fdes >= NOFILE || (fp = u.u_ofile[fdes]) == NULL) { + u.u_error = EBADF; + return ((struct file *)0); + } + if (fp->f_type != DTYPE_INODE) { + u.u_error = EINVAL; + return ((struct file *)0); + } + return (fp); +} + +/* + * mode mask for creation of files + */ +umask() +{ + register struct a { + int mask; + } *uap = (struct a *)u.u_ap; + + u.u_r.r_val1 = u.u_cmask; + u.u_cmask = uap->mask & 07777; }