X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/4f083fd7a06675f7d351f02e7d69eb5c7b9fcc8f..18b0bce61728e5c0943caa5386425644f81d9aec:/usr/src/sys/ufs/ffs/ufs_vnops.c diff --git a/usr/src/sys/ufs/ffs/ufs_vnops.c b/usr/src/sys/ufs/ffs/ufs_vnops.c index a81a284f0d..016a175485 100644 --- a/usr/src/sys/ufs/ffs/ufs_vnops.c +++ b/usr/src/sys/ufs/ffs/ufs_vnops.c @@ -1,22 +1,30 @@ -/* ufs_vnops.c 4.42 82/11/13 */ - -#include "../h/param.h" -#include "../h/systm.h" -#include "../h/dir.h" -#include "../h/user.h" -#include "../h/kernel.h" -#include "../h/file.h" -#include "../h/stat.h" -#include "../h/inode.h" -#include "../h/fs.h" -#include "../h/buf.h" -#include "../h/proc.h" -#include "../h/quota.h" -#include "../h/descrip.h" -#include "../h/uio.h" -#include "../h/socket.h" -#include "../h/socketvar.h" -#include "../h/nami.h" +/* + * 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. + * + * @(#)ufs_vnops.c 7.4 (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 (``.''). @@ -46,9 +54,13 @@ chdirec(ipp) register struct inode *ip; struct a { char *fname; - }; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; - ip = namei(uchar, LOOKUP, 1); + 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) { @@ -57,7 +69,7 @@ chdirec(ipp) } if (access(ip, IEXEC)) goto bad; - iunlock(ip); + IUNLOCK(ip); if (*ipp) irele(*ipp); *ipp = ip; @@ -72,80 +84,84 @@ bad: */ open() { - register struct inode *ip; - register struct a { + struct a { char *fname; - int flags; int mode; - } *uap; - int checkpermissions = 1, flags; + int crtmode; + } *uap = (struct a *) u.u_ap; - uap = (struct a *)u.u_ap; - flags = uap->flags + 1; - if ((flags&FTRUNCATE) && (flags&FWRITE) == 0) { - u.u_error = EINVAL; - return; - } - if (flags&FCREATE) { - ip = namei(uchar, CREATE, 1); - if (ip == NULL) { - if (u.u_error) - return; - ip = maknode(uap->mode&07777&(~ISVTX)); - checkpermissions = 0; - flags &= ~FTRUNCATE; - } - } else - ip = namei(uchar, LOOKUP, 1); - if (ip == NULL) - return; - open1(ip, flags, checkpermissions); + copen(uap->mode-FOPEN, uap->crtmode, uap->fname); } -#ifndef NOCOMPAT /* * Creat system call. */ -ocreat() +creat() { - register struct inode *ip; - register struct a { + struct a { char *fname; int fmode; - } *uap; + } *uap = (struct a *)u.u_ap; - uap = (struct a *)u.u_ap; - ip = namei(uchar, CREATE, 1); - if (ip == NULL) { - if (u.u_error) - return; - ip = maknode(uap->fmode&07777&(~ISVTX)); - if (ip == NULL) - return; - open1(ip, FWRITE, 0); - } else - open1(ip, FWRITE|FTRUNCATE, 1); + copen(FWRITE|FCREAT|FTRUNC, uap->fmode, uap->fname); } -#endif /* * Common code for open and creat. - * Check permissions (if we haven't done so already), - * allocate an open file structure, and call - * the device open routine, if any. + * Check permissions, allocate an open file structure, + * and call the device open routine if any. */ -open1(ip, mode, checkpermissions) - register struct inode *ip; - register mode; +copen(mode, arg, fname) + register int mode; + int arg; + caddr_t fname; { + register struct inode *ip; register struct file *fp; - int i, flags; + register struct nameidata *ndp = &u.u_nd; + int indx; - if (checkpermissions) { + 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) { + if (mode&(FWRITE|FTRUNC)) { if (access(ip, IWRITE)) goto bad; if ((ip->i_mode&IFMT) == IFDIR) { @@ -154,40 +170,29 @@ open1(ip, mode, checkpermissions) } } } - - /* - * Check locking on inode. Release "inode lock" - * while doing so in case we block inside flocki. - */ - flags = 0; - if (mode&(FSHLOCK|FEXLOCK)) { - iunlock(ip); - flags = flocki(ip, 0, mode); - ilock(ip); - if (u.u_error) - goto bad; - } - if (mode&FTRUNCATE) + if (mode&FTRUNC) itrunc(ip, (u_long)0); - iunlock(ip); - if ((fp = falloc()) == NULL) - goto out; - fp->f_flag = mode & FMODES; - fp->f_type = DTYPE_FILE; - i = u.u_r.r_val1; - fp->f_inode = ip; - u.u_error = openi(ip, mode); - if (u.u_error == 0) { - u.u_pofile[i] = flags; + 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_ofile[i] = NULL; - fp->f_count--; -out: - irele(ip); - return; + u.u_error = openi(ip, mode); + if (u.u_error == 0) + return; + ILOCK(ip); bad: iput(ip); +bad1: + u.u_ofile[indx] = NULL; + fp->f_count--; } /* @@ -200,28 +205,37 @@ mknod() char *fname; int fmode; int dev; - } *uap; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; - uap = (struct a *)u.u_ap; - if (suser()) { - ip = namei(uchar, CREATE, 0); - if (ip != NULL) { - u.u_error = EEXIST; - goto out; - } + if (!suser()) + return; + 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); + ip = maknode(uap->fmode, ndp); if (ip == NULL) return; - 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; + 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: @@ -237,10 +251,13 @@ link() register struct a { char *target; char *linkname; - } *uap; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; - uap = (struct a *)u.u_ap; - ip = namei(uchar, LOOKUP, 1); /* well, this routine is doomed anyhow */ + 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()) { @@ -250,9 +267,11 @@ link() 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); + 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); @@ -260,12 +279,12 @@ link() } if (u.u_error) goto out; - if (u.u_pdir->i_dev != ip->i_dev) { - iput(u.u_pdir); + if (ndp->ni_pdir->i_dev != ip->i_dev) { + iput(ndp->ni_pdir); u.u_error = EXDEV; goto out; } - direnter(ip); + u.u_error = direnter(ip, ndp); out: if (u.u_error) { ip->i_nlink--; @@ -282,12 +301,12 @@ symlink() register struct a { char *target; char *linkname; - } *uap; + } *uap = (struct a *)u.u_ap; register struct inode *ip; register char *tp; register c, nc; + register struct nameidata *ndp = &u.u_nd; - uap = (struct a *)u.u_ap; tp = uap->target; nc = 0; while (c = fubyte(tp)) { @@ -298,8 +317,10 @@ symlink() tp++; nc++; } - u.u_dirp = uap->linkname; - ip = namei(uchar, CREATE, 0); + 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; @@ -307,10 +328,11 @@ symlink() } if (u.u_error) return; - ip = maknode(IFLNK | 0777); + ip = maknode(IFLNK | 0777, ndp); if (ip == NULL) return; - u.u_error = rdwri(UIO_WRITE, ip, uap->target, nc, 0, 0, (int *)0); + u.u_error = rdwri(UIO_WRITE, ip, uap->target, nc, (off_t)0, 0, + (int *)0); /* handle u.u_error != 0 */ iput(ip); } @@ -324,13 +346,17 @@ unlink() { struct a { char *fname; - }; + } *uap = (struct a *)u.u_ap; register struct inode *ip, *dp; + register struct nameidata *ndp = &u.u_nd; - ip = namei(uchar, DELETE | LOCKPARENT, 0); + ndp->ni_nameiop = DELETE | LOCKPARENT; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); if (ip == NULL) return; - dp = u.u_pdir; + dp = ndp->ni_pdir; if ((ip->i_mode&IFMT) == IFDIR && !suser()) goto out; /* @@ -342,7 +368,7 @@ unlink() } if (ip->i_flag&ITEXT) xrele(ip); /* try once to free text */ - if (dirremove()) { + if (dirremove(ndp)) { ip->i_nlink--; ip->i_flag |= ICHG; } @@ -364,22 +390,32 @@ lseek() int fd; off_t off; int sbase; - } *uap; + } *uap = (struct a *)u.u_ap; - uap = (struct a *)u.u_ap; - fp = getf(uap->fd); - if (fp == NULL) - return; - if (fp->f_type == DTYPE_SOCKET) { + GETF(fp, uap->fd); + if (fp->f_type != DTYPE_INODE) { u.u_error = ESPIPE; return; } - if (uap->sbase == FSEEK_RELATIVE) - uap->off += fp->f_offset; - else if (uap->sbase == FSEEK_EOF) - uap->off += fp->f_inode->i_size; - fp->f_offset = uap->off; - u.u_r.r_off = uap->off; + 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; } /* @@ -392,20 +428,23 @@ saccess() register struct a { char *fname; int fmode; - } *uap; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; - 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); + ndp->ni_nameiop = LOOKUP | FOLLOW; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); if (ip != NULL) { - if ((uap->fmode&FACCESS_READ) && access(ip, IREAD)) + if ((uap->fmode&R_OK) && access(ip, IREAD)) goto done; - if ((uap->fmode&FACCESS_WRITE) && access(ip, IWRITE)) + if ((uap->fmode&W_OK) && access(ip, IWRITE)) goto done; - if ((uap->fmode&FACCESS_EXECUTE) && access(ip, IEXEC)) + if ((uap->fmode&X_OK) && access(ip, IEXEC)) goto done; done: iput(ip); @@ -414,99 +453,44 @@ done: u.u_gid = svgid; } -/* - * the fstat system call. - */ -fstat() -{ - register struct file *fp; - register struct a { - int fd; - struct stat *sb; - } *uap; - - uap = (struct a *)u.u_ap; - fp = getf(uap->fd); - if (fp == NULL) - return; - if (fp->f_type == DTYPE_SOCKET) - u.u_error = sostat(fp->f_socket, uap->sb); - else - stat1(fp->f_inode, uap->sb); -} - /* * Stat system call. This version follows links. */ stat() { - register struct inode *ip; - register struct a { - char *fname; - struct stat *sb; - } *uap; - uap = (struct a *)u.u_ap; - ip = namei(uchar, LOOKUP, 1); - if (ip == NULL) - return; - stat1(ip, uap->sb); - iput(ip); + 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 *sb; - } *uap; + struct stat *ub; + } *uap = (struct a *)u.u_ap; + struct stat sb; + register struct nameidata *ndp = &u.u_nd; - uap = (struct a *)u.u_ap; - ip = namei(uchar, LOOKUP, 0); + ndp->ni_nameiop = LOOKUP | follow; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); if (ip == NULL) return; - stat1(ip, uap->sb); + (void) ino_stat(ip, &sb); iput(ip); -} - -/* - * The basic routine for fstat and stat: - * get the inode and pass appropriate parts back. - */ -stat1(ip, ub) - register struct inode *ip; - struct stat *ub; -{ - struct stat ds; - - IUPDAT(ip, &time, &time, 0); - /* - * 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_rdev; - ds.st_size = ip->i_size; - ds.st_atime = ip->i_atime; - ds.st_mtime = ip->i_mtime; - ds.st_ctime = ip->i_ctime; - /* this doesn't belong here */ - if ((ip->i_mode&IFMT) == IFBLK) - ds.st_blksize = BLKDEV_IOSIZE; - else if ((ip->i_mode&IFMT) == IFCHR) - ds.st_blksize = MAXBSIZE; - else - ds.st_blksize = ip->i_fs->fs_bsize; - if (copyout((caddr_t)&ds, (caddr_t)ub, sizeof(ds)) < 0) - u.u_error = EFAULT; + u.u_error = copyout((caddr_t)&sb, (caddr_t)uap->ub, sizeof (sb)); } /* @@ -520,16 +504,21 @@ readlink() char *buf; int count; } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; int resid; - ip = namei(uchar, LOOKUP, 0); + ndp->ni_nameiop = LOOKUP; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->name; + ip = namei(ndp); if (ip == NULL) return; if ((ip->i_mode&IFMT) != IFLNK) { - u.u_error = ENXIO; + u.u_error = EINVAL; goto out; } - u.u_error = rdwri(UIO_READ, ip, uap->buf, uap->count, 0, 0, &resid); + u.u_error = rdwri(UIO_READ, ip, uap->buf, uap->count, (off_t)0, 0, + &resid); out: iput(ip); u.u_r.r_val1 = uap->count - resid; @@ -544,12 +533,11 @@ chmod() struct a { char *fname; int fmode; - } *uap; + } *uap = (struct a *)u.u_ap; - uap = (struct a *)u.u_ap; - if ((ip = owner(1)) == NULL) + if ((ip = owner(uap->fname, FOLLOW)) == NULL) return; - chmod1(ip, uap->fmode); + u.u_error = chmod1(ip, uap->fmode); iput(ip); } @@ -561,24 +549,19 @@ fchmod() struct a { int fd; int fmode; - } *uap; + } *uap = (struct a *)u.u_ap; register struct inode *ip; register struct file *fp; - uap = (struct a *)u.u_ap; - fp = getf(uap->fd); + fp = getinode(uap->fd); if (fp == NULL) return; - if (fp->f_type == DTYPE_SOCKET) { - u.u_error = EINVAL; - return; - } - ip = fp->f_inode; + ip = (struct inode *)fp->f_data; if (u.u_uid != ip->i_uid && !suser()) return; - ilock(ip); - chmod1(ip, uap->fmode); - iunlock(ip); + ILOCK(ip); + u.u_error = chmod1(ip, uap->fmode); + IUNLOCK(ip); } /* @@ -589,27 +572,21 @@ chmod1(ip, mode) register struct inode *ip; register int mode; { - register int *gp; + if (ip->i_fs->fs_ronly) + return (EROFS); ip->i_mode &= ~07777; if (u.u_uid) { - mode &= ~ISVTX; - for (gp = u.u_groups; gp < &u.u_groups[NGROUPS]; gp++) - if (*gp == ip->i_gid) - goto ok; - mode &= ~ISGID; -ok: - ; -#ifdef MUSH - if (u.u_quota->q_syflags & QF_UMASK && u.u_uid != 0 && - (ip->i_mode & IFMT) != IFCHR) - mode &= ~u.u_cmask; -#endif + 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); } /* @@ -622,12 +599,16 @@ chown() char *fname; int uid; int gid; - } *uap; + } *uap = (struct a *)u.u_ap; + register struct nameidata *ndp = &u.u_nd; - uap = (struct a *)u.u_ap; - if (!suser() || (ip = owner(0)) == NULL) + ndp->ni_nameiop = LOOKUP | NOFOLLOW; + ndp->ni_segflg = UIO_USERSPACE; + ndp->ni_dirp = uap->fname; + ip = namei(ndp); + if (ip == NULL) return; - chown1(ip, uap->uid, uap->gid); + u.u_error = chown1(ip, uap->uid, uap->gid); iput(ip); } @@ -640,24 +621,17 @@ fchown() int fd; int uid; int gid; - } *uap; + } *uap = (struct a *)u.u_ap; register struct inode *ip; register struct file *fp; - uap = (struct a *)u.u_ap; - fp = getf(uap->fd); + fp = getinode(uap->fd); if (fp == NULL) return; - if (fp->f_type == DTYPE_SOCKET) { - u.u_error = EINVAL; - return; - } - ip = fp->f_inode; - if (!suser()) - return; - ilock(ip); - chown1(ip, uap->uid, uap->gid); - iunlock(ip); + ip = (struct inode *)fp->f_data; + ILOCK(ip); + u.u_error = chown1(ip, uap->uid, uap->gid); + IUNLOCK(ip); } /* @@ -670,77 +644,66 @@ chown1(ip, 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; /* - * This doesn't allow for holes in files (which hopefully don't - * happen often in files that we chown), and is not accurate anyway - * (eg: it totally ignores 3 level indir blk files - but hopefully - * noone who can make a file that big will have a quota) + * If we don't own the file, are trying to change the owner + * of the file, or are not a member of the target group, + * the caller must be superuser or the call fails. */ - if (ip->i_uid == uid) + if ((u.u_uid != ip->i_uid || uid != ip->i_uid || + !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 { - register struct fs *fs = ip->i_fs; - - if (ip->i_size > (change = NDADDR * fs->fs_bsize)) { - register off_t size; - - size = blkroundup(fs, ip->i_size) - change; - change += size; - change += fs->fs_bsize; - /* this assumes NIADDR <= 2 */ - if (size > NINDIR(fs) * fs->fs_bsize) - change += fs->fs_bsize; - } else - change = fragroundup(fs, ip->i_size); - change /= DEV_BSIZE; - } - (void)chkdq(ip, -change, 1); - (void)chkiq(ip->i_dev, ip, ip->i_uid, 1); + 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 - /* - * keep uid/gid's in sane range -- no err, - * so chown(file, uid, -1) will do something useful - */ - if (uid >= 0 && uid <= 32767) /* should have a constant */ - ip->i_uid = uid; - if (gid >= 0 && gid <= 32767) /* same here */ - ip->i_gid = gid; + 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); + (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 } -/* - * Set IUPD and IACC times on file. - * Can't set ICHG. - */ -outime() +utimes() { register struct a { char *fname; - time_t *tptr; - } *uap; + struct timeval *tptr; + } *uap = (struct a *)u.u_ap; register struct inode *ip; - time_t tv[2]; - struct timeval tv0, tv1; + struct timeval tv[2]; - uap = (struct a *)u.u_ap; - if ((ip = owner(1)) == NULL) + if ((ip = owner(uap->fname, FOLLOW)) == NULL) return; - if (copyin((caddr_t)uap->tptr, (caddr_t)tv, sizeof(tv))) { - u.u_error = EFAULT; - } else { + if (ip->i_fs->fs_ronly) { + u.u_error = EROFS; + iput(ip); + 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); + iupdat(ip, &tv[0], &tv[1], 0); } iput(ip); } @@ -754,47 +717,6 @@ sync() update(); } -/* - * Apply an advisory lock on a file descriptor. - */ -flock() -{ - struct a { - int fd; - int how; - } *uap; - register struct file *fp; - register int cmd, flags; - - uap = (struct a *)u.u_ap; - fp = getf(uap->fd); - if (fp == NULL) - return; - if (fp->f_type == DTYPE_SOCKET) { /* XXX */ - u.u_error = EINVAL; - return; - } - cmd = uap->how; - flags = u.u_pofile[uap->fd] & (SHLOCK|EXLOCK); - if (cmd&FUNLOCK) { - if (flags == 0) { - u.u_error = EINVAL; - return; - } - funlocki(fp->f_inode, flags); - u.u_pofile[uap->fd] &= ~(SHLOCK|EXLOCK); - return; - } - /* - * No reason to write lock a file we've already - * write locked, similarly with a read lock. - */ - if ((flags&EXLOCK) && (cmd&FEXLOCK) || - (flags&SHLOCK) && (cmd&FSHLOCK)) - return; - u.u_pofile[uap->fd] = flocki(fp->f_inode, u.u_pofile[uap->fd], cmd); -} - /* * Truncate a file given its path name. */ @@ -802,11 +724,15 @@ truncate() { struct a { char *fname; - u_long length; + off_t length; } *uap = (struct a *)u.u_ap; struct inode *ip; + register struct nameidata *ndp = &u.u_nd; - ip = namei(uchar, LOOKUP, 1); + 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)) @@ -815,7 +741,7 @@ truncate() u.u_error = EISDIR; goto bad; } - itrunc(ip, uap->length); + itrunc(ip, (u_long)uap->length); bad: iput(ip); } @@ -827,26 +753,22 @@ ftruncate() { struct a { int fd; - u_long length; + off_t length; } *uap = (struct a *)u.u_ap; struct inode *ip; struct file *fp; - fp = getf(uap->fd); + fp = getinode(uap->fd); if (fp == NULL) return; - if (fp->f_type == DTYPE_SOCKET) { - u.u_error = EINVAL; - return; - } if ((fp->f_flag&FWRITE) == 0) { u.u_error = EINVAL; return; } - ip = fp->f_inode; - ilock(ip); - itrunc(ip, uap->length); - iunlock(ip); + ip = (struct inode *)fp->f_data; + ILOCK(ip); + itrunc(ip, (u_long)uap->length); + IUNLOCK(ip); } /* @@ -860,17 +782,15 @@ fsync() struct inode *ip; struct file *fp; - fp = getf(uap->fd); + fp = getinode(uap->fd); if (fp == NULL) return; - if (fp->f_type == DTYPE_SOCKET) { - u.u_error = EINVAL; - return; - } - ip = fp->f_inode; - ilock(ip); + ip = (struct inode *)fp->f_data; + ILOCK(ip); + if (fp->f_flag&FWRITE) + ip->i_flag |= ICHG; syncip(ip); - iunlock(ip); + IUNLOCK(ip); } /* @@ -888,11 +808,12 @@ fsync() * * 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. + * 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. - * 4) If a directory was moved and the parent of the destination + * 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. * @@ -904,35 +825,43 @@ rename() struct a { char *from; char *to; - } *uap; + } *uap = (struct a *)u.u_ap; register struct inode *ip, *xp, *dp; - int oldparent, parentdifferent, doingdirectory; - - uap = (struct a *)u.u_ap; - ip = namei(uchar, LOOKUP | LOCKPARENT, 0); + 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 = u.u_pdir; - oldparent = 0, doingdirectory = 0; + dp = ndp->ni_pdir; if ((ip->i_mode&IFMT) == IFDIR) { register struct direct *d; - d = &u.u_dent; + d = &ndp->ni_dent; /* - * Avoid "." and ".." for obvious reasons. + * Avoid ".", "..", and aliases of "." for obvious reasons. */ - if (d->d_name[0] == '.') { - if (d->d_namlen == 1 || - (d->d_namlen == 2 && d->d_name[1] == '.')) { - u.u_error = EINVAL; + 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); - return; - } + u.u_error = EINVAL; + return; } + ip->i_flag |= IRENAME; oldparent = dp->i_number; doingdirectory++; } - irele(dp); + iput(dp); /* * 1) Bump link count while we're moving stuff @@ -943,17 +872,49 @@ rename() ip->i_nlink++; ip->i_flag |= ICHG; iupdat(ip, &time, &time, 1); - iunlock(ip); + 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) + 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 = u.u_pdir; + } + 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. @@ -961,171 +922,193 @@ rename() * entry to reference the source inode and * expunge the original entry's existence. */ - parentdifferent = oldparent != dp->i_number; if (xp == NULL) { if (dp->i_dev != ip->i_dev) { - u.u_error = EXDEV; + error = EXDEV; 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. + * Account for ".." in new directory. + * When source and destination have the same + * parent we don't fool with the link count. */ - if (doingdirectory && parentdifferent) { + if (doingdirectory && newparent) { dp->i_nlink++; dp->i_flag |= ICHG; iupdat(dp, &time, &time, 1); } - direnter(ip); - if (u.u_error) + error = direnter(ip, ndp); + if (error) goto out; } else { if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev) { - u.u_error = EXDEV; + error = EXDEV; goto bad; } /* - * Target must be empty if a directory. + * 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)) { - u.u_error = EEXIST; /* XXX */ + if (!dirempty(xp, dp->i_number) || xp->i_nlink > 2) { + error = ENOTEMPTY; goto bad; } if (!doingdirectory) { - u.u_error = ENOTDIR; + error = ENOTDIR; goto bad; } + cacheinval(dp); } else if (doingdirectory) { - u.u_error = EISDIR; + error = EISDIR; goto bad; } - dirrewrite(dp, ip); - if (u.u_error) + dirrewrite(dp, ip, ndp); + if (u.u_error) { + error = u.u_error; goto bad1; + } /* - * If this is a directory we know it is - * empty and we can squash the inode and - * any space associated with it. Otherwise, - * we've got a plain file and the link count - * simply needs to be adjusted. + * 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) { - xp->i_nlink = 0; + if (--xp->i_nlink != 0) + panic("rename: linked directory"); itrunc(xp, (u_long)0); - } else - xp->i_nlink--; + } xp->i_flag |= ICHG; iput(xp); + xp = NULL; } /* * 3) Unlink the source. */ - u.u_dirp = uap->from; - dp = namei(uchar, DELETE, 0); + 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 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. + * 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 (dp == NULL || u.u_dent.d_ino != ip->i_number) { - ip->i_nlink--; - ip->i_flag |= ICHG; + if (xp != ip) { + if (doingdirectory) + panic("rename: lost dir entry"); } 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 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 && (xp != NULL || parentdifferent)) { + 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()) { - ip->i_nlink--; - ip->i_flag |= ICHG; + if (dirremove(ndp)) { + xp->i_nlink--; + xp->i_flag |= ICHG; } + xp->i_flag &= ~IRENAME; + if (error == 0) /* XXX 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 && u.u_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; - } - u.u_error = rdwri(UIO_READ, ip, (caddr_t)&dirbuf, - sizeof (struct dirtemplate), (off_t)0, 1, (int *)0); - if (u.u_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); - } + if (xp) + iput(xp); + irele(ip); + if (error) + u.u_error = error; return; + bad: - iput(u.u_pdir); + iput(dp); bad1: if (xp) - irele(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) +maknode(mode, ndp) int mode; + register struct nameidata *ndp; { register struct inode *ip; + register struct inode *pdir = ndp->ni_pdir; ino_t ipref; if ((mode & IFMT) == IFDIR) - ipref = dirpref(u.u_pdir->i_fs); + ipref = dirpref(pdir->i_fs); else - ipref = u.u_pdir->i_number; - ip = ialloc(u.u_pdir, ipref, mode); + ipref = pdir->i_number; + ip = ialloc(pdir, ipref, mode); if (ip == NULL) { - iput(u.u_pdir); + iput(pdir); return (NULL); } #ifdef QUOTA @@ -1138,7 +1121,9 @@ maknode(mode) 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; + 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 @@ -1147,11 +1132,11 @@ maknode(mode) * Make sure inode goes to disk before directory entry. */ iupdat(ip, &time, &time, 1); - direnter(ip); + u.u_error = direnter(ip, ndp); if (u.u_error) { /* - * write error occurred trying to update directory - * so must deallocate the inode + * Write error occurred trying to update directory + * so must deallocate the inode. */ ip->i_nlink = 0; ip->i_flag |= ICHG; @@ -1160,3 +1145,235 @@ maknode(mode) } 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); +} + +/* + * 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; +}